Potential and whole-genome sequence-based mechanism of elongated-prismatic magnetite magnetosome formation in Acidithiobacillus ferrooxidans BYM

A magnetosome-producing bacterium Acidithiobacillus ferrooxidans BYM (At. ferrooxidans BYM) was isolated and magnetically screened. The magnetosome yield from 0.5896 to 13.1291 mg/g was achieved under different aeration rates, ferrous sulfate, ammonium sulfate, and gluconic acid concentrations at 30 ℃. TEM observed 6–9 magnetosomes in size of 20–80 nm irregularly dispersed in a cell. STEM-EDXS and HRTEM-FFT implied that the elongated-prismatic magnetite magnetosomes with {110} crystal faces grown along the [111] direction. Whole-genome sequencing and annotation of BYM showed that 3.2 Mb chromosome and 47.11 kb plasmid coexisted, and 322 genes associated with iron metabolism were discovered. Ten genes shared high similarity with magnetosome genes were predicted, providing sufficient evidence for the magnetosome-producing potential of BYM. Accordingly, we first proposed a hypothetic model of magnetosome formation including vesicle formation, iron uptake and mineralization, and magnetite crystal maturation in At. ferrooxidans. These indicated that At. ferrooxidans BYM would be used as a commercial magnetosome-producing microorganism.


Introduction
Magnetotactic bacteria (MTB) are a multi-phyletic group of gram-negative prokaryotes and widely distribute in the soil, lakes, oceans, and underwater sludge, even exist in extreme environments (Lin et al. 2017). They exhibit various morphotypes (rods, cocci, spirilla, vibrios, and ovoid) and are classified into 13 bacterial phyla according to the genome taxonomy database (GTDB) taxonomy and 7 phyla based on the NCBI taxonomy, including Proteobacteria, Omnitrophica, Nitrospirae, Planctomycetota, Latescibacteria, Fibrobacterota and Desulfobacterota etc. (Lin et al. 2020). They are characterized by their ability to produce magnetosomes and swim along the geomagnetic field lines. Only about 70 strains of MTB have been isolated in pure culture since one of them was independently discovered by Salvatore Bellini andRichard Blakemore in 1963 and1975, respectively, such as Magnetospirillum magneticum MGT-1, M. magneticum AMB-1, M. gryphiswaldense MSR-1, Magnetospirillum sp. MSM3, Magnetococcus sp. MC-1, Desulfamplus magnetovallimortis BW-1 and so on (Yan et al. 2012a, b). Bacterial magnetosome is composed of phospholipid bilayer membrane-enclosed crystals of magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ). The size of mature magnetite crystals ranging from 35 to 120 nm is in a stable single magnetic domain (Yan et al. 2017a, b). The formation of magnetosomes were strictly controlled by the genes and consisted of several processes, including magnetosome vesicle formation, iron transportation and accumulation, redox control Dan Zhao and Jiani Yang have contributed equally to the manuscript. of magnetic crystal biomineralization, magnetosome chain assembly and morphological regulation (Yan et al. 2012a, b). It has been reported that more than 40 genes involved in the biosynthesis of magnetosomes (Dieudonnéet al. 2019). The complete genome sequencing of MTB strains was first achieved in a model microorganism M. magneticum AMB-1 and then the genomes of about 17 strains of MTB have also become available. Comparative genome analysis of both cultured and uncultured MTB revealed the presence of a well-conserved region called the magnetosome island (MAI) in the chromosome and confirmed its functional significance in magnetosome formation. It has been suggested that MAI harbored the magnetosome gene clusters, structured as operons including mamAB, mamGFDC, mms6, mamXY and feoAB, can be acquired and transmitted via horizontal gene transfer (Dieudonné et al. 2019). Due to the unique properties such as ferrimagnetism, nanoscale size, narrow size distribution, dispersal ability, membrane-bound structure, and good biocompatibility, the magnetosomes could be applied in biomedicine, biotechnology, material science and environmental protection (Yan et al. 2017a, b). Additionally, their application in geology, mineralogy, oceanography and astrobiology would be worth exploring further.
Therefore, the high-yield production of magnetosomes from microorganisms is critical for commercial application. Most studies have been conducted to enhance magnetosome yields by optimizing conditions for bacterial growth or/and magnetosome synthesis, genetic construction of magnetosome high-yielding mutants, and protein-mediated biomimetic synthesis of magnetosome-like particles in vitro (Kashyap et al. 2014). Unfortunately, it is challenging to find a suitable artificial environmental condition to cultivate wild and mutant MTB due to their fastidious lifestyle, creating a major hurdle to relocate magnetosome production from laboratory level to industrial or commercial scale (Basit et al. 2020). In addition, the magnetosome-like nanoparticles synthesized by the physical, chemical and biomimetic method in vitro are often non-uniform, incompletely crystalline, compositionally nonhomogeneous and easy to aggregate because of the magnetic attraction (Yan et al. 2012a, b). Furthermore, these artificial synthetic techniques are costly and energy-intensive due to the involvement of expensive toxic or hazardous chemicals which can be harmful to humans and the environment (Basit et al. 2020). Thus, it is necessary to introduce unfastidious magnetosome-producing microbes regretless of their magnetotaxis.
It has been discovered that several easy-to-cultivate non-magnetotactic iron-oxidizing bacteria including At. ferrooxidans, Ferroplasma thermophilum and Leptospirillum ferriphilum can produce magnetosomes (Yan et al. 2017a, b). Among them, At. ferrooxidans is a gram-negative, extremely acidophilic chemolithoautotrophic bacterium and has been recognized as an important member involved in biogeochemical iron cycling (Zhang et al. 2018). It wildly distributes in nature and obtains energy through the oxidation of ferrous iron and/or reduced inorganic sulfur. It has been demonstrated that the magnetosomes synthesized by At. ferrooxidans exhibit the above-mentioned excellent characteristics possessed by MTB magnetosomes (Yan et al. 2012a, b). It was speculated that the formation mechanism of magnetosomes in At. ferrooxidans might be different from that in MTB, since a chromosomal region similar to MAI was not found in the whole genome of At. ferrooxidans. Although more than 500 isolates and several genomes of At. ferrooxidans have been reported, the insight into the molecular mechanism of magnetosome formation by utilizing the complete genome sequence is not provided. Additionally, the synthetic ability of magnetosomes of At. ferrooxidans were preliminarily evaluated in the previous studies at the shake-flask scale by monitoring the indirect indexes including intracellular iron content and the cellular coefficient of magnetism (C mag ) (Yan et al. 2016). However, the magnetosome-producing potential should be assayed at the bioreactor scale by measuring the direct parameters such as the weight and number of magnetosomes, which would help to have an in-depth understanding of the magnetosome production ability of At. ferrooxidans and its potential as a magnetosome-producing bacterial resource.
In the present study, At. ferrooxidans BYM was isolated and screened by a combination of liquid and semisolid-plate magnetophoresis, and was phylogenetically identified by the 16S rRNA gene. The magnetosome-producing potential of At. ferrooxidans BYM cultivated in a 5 L aerated bioreactor under various incubation conditions were appraised by a direct weighing method. The shape, crystal lattice parameter and element composition of the magnetosomes were observed by STEM-EDXS and HRTEM-FFT. Additionally, the genome of At. ferrooxidans BYM was wholly sequenced and annotated to generate a comprehensive description of the mechanism that how At. ferrooxidans synthesized magnetosomes.

Isolation and magnetic screening
Soil samples were obtained from Baiyin copper mine (N36° 55′ 41.02″, E104° 08′ 5.25″) of Gansu Province, China. Two gram of soil sample was inoculated in 40 mL modified 9 K medium consisting of 2.4 g/L (NH 4 ) 2 SO 4 , 0.1 g/L KCl, 0.5 g/L K 2 HPO 4 , 0.5 g/L MgSO 4 ·7H 2 O and 0.01 g/L Ca (NO 3 ) 2 , 40 g/L FeSO 4 ·7H 2 O at pH 2.0 (Yan et al. 2016). The cultures were incubated aerobically in a shaking incubator at 120 rpm and 30 °C until the color of the medium turned reddish-brown. The isolation was carried out on an agar 9 K medium plate (1.0% agar) using a serial dilution method. Discrete colonies were obtained after 7-10 days and incubated in 9 K liquid medium. This procedure was repeated three times to obtain a purified strain and magnetic screening was performed. Briefly, a small hard ferrite magnet (2 mT) was immediately placed close to a wall of microcosms. After about 1-2 h of magnetic enrichment, the bacteria adjacent to the magnet were transferred to the wide mouth of Pasteur pipette to further magnetic separation by the "race track" method (Morillo et al. 2014). The tip-end pipette containing the accumulated cells was broken off after collection for 30-60 min and then incubated in 9 K liquid medium. The stationary phase cells were collected by centrifugation and suspended in medium. Filter paper strips with a length equal to diameter of the petri dish were soaked briefly in bacterial suspension and then inoculated on a semisolid agar plate (0.7% agar) with aerobically cultured at 30 °C. The test group plate was placed on an artificial magnetic field created by a permanent magnet (8 cm × 6 cm × 1 cm, 18 mT), whereas the control group plates was placed on the geomagnetic field (Liu et al. 2008a, b). The bacteria adjacent to the magnet were picked into the 9 K liquid medium, and the brownish-red culture was diluted and plated on the solid plate after incubation. The cycling of magnetic screening was performed 10 times (Fig. S1). The well-isolated colony was picked and inoculated into 9 K liquid medium for cell collection and analysis. All reagents used in this work were analytical grade.

16S rDNA-based identification
Two-gram bacterial cells from the logarithmic phase were collected according to the previous report (Yan et al. 2017a, b). The genomic DNA was extracted from the collected cells according to the manufacturer's instructions by a commercially available DNA extraction kit (Biomed Biotechnology, Beijing, China), and was detected by agarose gel electrophoresis (1.0%). The identification was done by 16S rDNA amplification using the forward primer 27F (5′-AGA GTT TGA TCC TGG CTA G-3′) and the reverse primer 1492R (5′-TAC GYT ACC TTG TTA CGA CTT-3′) purchased from Shanghai sangon Biotech. The amplification reaction was carried out in a total volume of 25 μL consisting of 1 μL DNA, 1 μL each of forward and reverse primer (10 μM), 12.5 μL 2 × Taq PCR Master Mix (Biomed Biotechnology, Beijing, China) and 9.5 μL ddH 2 O. PCR was performed on a 2720 Thermal Cycler Instrument (Applied Biosystems, Foster City, USA) using the following program: initial denaturation at 94 ℃ for 5 min, followed by 30 cycles of denaturation of 94 °C for 30 s, annealing at 55 °C for 45 s, extension at 72 °C for 1 min and a final extension of 5 min at 72 °C. The PCR products were checked by 1.0% agarose gel electrophoresis and purified by AxyGEN Nucleic Acid Purification Kit (Axygen scientific, Inc, USA). The gel-purified products were sequenced by Sangon Biotech (Shanghai) Co., Ltd. The 16S rDNA gene contiguous sequences were analyzed using the ChromasPro software version 2.6.5 (Technelysium Pty Ltd, Tewantin, Australia) and assembled by DNAMAN software version 6.0 (Lynnon Biosoft, Quebec, Canada). The nucleotide sequence of 16S rDNA was used to find maximum similarity to the bacterial strains in the NCBI database. The phylogenetic tree was drawn using the neighbor-joining method in MEGA 5.0 (Temple University, Philadelphia, PA, USA) with the reference sequences obtained from NCBI GenBank.

Batch magnetosome-producing potential assay
At. ferrooxidans BYM was cultivated in a 5 L aerated bioreactor ( Fig. S2) to improve the production of bacterial biomass. The air filtered by 0.22 µm Millipore membrane was pumped into bioreactor through a central glass draught tube installed 5 cm above the bottom by a CT-202 air pump (SenSen Group Co., Ltd., Zhoushan, China). The ventilatory capacity was controlled by LZB-4WB gas flowmeter (Xiangjin Flowmeter Factory, Xinghua, China). The air inlet and outlet of bioreactor were installed with filters (0.22 μm × 25 mm, Lab instrument Co., Ltd., Shanghai, China). The temperature was controlled using a DK-S24 thermostatic water bath (Senxin Experimental Instrument Co., Ltd., Shanghai, China). Batch fermentation were carried out in the bioreactor with 10% (v/v) inoculation under different ferrous sulfate concentration (15, 25, 30, and 40 g/L), ammonium sulfate concentration (0.24, 0.72, 1.2, and 1.8 g/L), gluconic acid concentration (5, 15, 30, and 50 mM), and ventilatory capacity (0.5, 0.8, 1.2, and 1.5 L/ min). After 90% ferrous ions oxidation, the bacteria were thought to have reached the beginning of the stable growth phase (Yan et al. 2013). The cells were then collected and the magnetosomes were extracted to evaluate the magnetosome-producing potential of At. ferrooxidans BYM.

Extraction of magnetosomes
The cultures of At. ferrooxidans BYM were filtered using a porous metal filter (Chuangwei Filtration Equipment Factory, Haining, China) with the filter membranes (0.22 μm × 50 mm, Xinya Purification Device Factory, Shanghai, China) and then washed with pH 2.0 H 2 SO 4 . The bacterial cells were collected until the filtrate did not contain iron ions monitored by the potassium thiocyanate (KSCN) method. The magnetosomes were extracted according to the protocol described by Yan et al. (Yan et al. 2012a, b), with some modifications. Briefly, cell suspension cultures containing approximately 0.5 g (wet weight) of At. ferrooxidans BYM cells, DNase (50 μg/ml), RNase (100 μg/ml), MgCl 2 (5 mM) and 1% (w/v) lysozyme in 20 ml of 10 mM Tris-HCl buffer (pH 7.4), was incubated at 37 °C for 1 h. Then the intermittent ultrasonic treatment was performed for 45 min in an ice bath with 3 s ultrasound and intermittent 5 s by using an ultrasonic generator (Ningbo Xinzhi Biotechnology Co., Ltd. Ningbo, China) operating at 40 kHz and 382.5 W. The treated cell suspension was diluted 2 times by Tris-HCl and transferred into another flat-bottomed test tube which was placed on the S pole of the magnet (2 mT). After magnetic adsorption for 1 h, the supernatant was removed and the magnetosomes gathered at the bottom were repeatedly washed with 10 mM HEPES (pH 7.4) buffer at least ten times and stored in 4 °C. The extracted magnetosomes at each time were accurately weighted used an electronic analytical balance (Shanghai Tianmei balance instrument Co., Ltd., Shanghai, China).

TEM, STEM and HRTEM characterization
The harvested At. ferrooxidans BYM cells were fixed by glutaraldehyde (2.5%, v/v) at 4 °C overnight, post-fixed in 1% (w/v) osmium tetroxide for 1 h, dehydrated through a graded ethanol series (70%, 80%, 90%, 100%), and embedded with a mixture of EPON 812 and acetone. Ultrathin sections were obtained using a Leica UC7 ultramicrotome (Co. Leica Microsystems GmbH, Germany), stained with both uranyl acetate and lead citrate. Conventional TEM observations were performed with a JEM-2100 TEM (JEOL, Tokyo, Japan) at an accelerating voltage of 120 kV. Crystal lengths (along the long axis) and widths (perpendicular to the long axis) were measured from TEM images. The concentrated magnetosome samples of At. ferrooxidans BYM were dispersed by ultrasound in ethanol. The analyte solutions were dropped onto duplex film and used for investigation. Highresolution TEM (HRTEM) and scanning TEM (STEM) observations in high-angle annular dark field (HAADF) mode were carried out on an FEI Tecnai G2 F20 microscope (FEI Ltd., Hillsboro, OR, USA) operating at 200 kV. Element analysis was carried out by energy-dispersive X-ray spectroscopy (EDXS). The fast Fourier transform (FFT) analysis of HRTEM images was performed to investigate the crystals' structural parameters using RADIUS software (EMSIS GmbH, Mendelstrasse, Muenster, Germany).

Whole-genome sequencing and proposal of magnetosome formation mechanism
The bacterial genomic DNA was extracted and detected as mentioned above. The qualified genomic DNA samples were subjected to whole-genome sequencing using Illumina MiSeq and PacBio in Personal Biotechnology Co. Ltd (Shanghai, China). The filtered reads were assembled into contigs by the A5-miseq version 20160825. and the Canu v 1.6 (Maryland Bioinformatics Labs, Park, MD, USA), and then corrected by Pilon v 1.22 (Free Software Foundation, Inc., Boston, MA, USA) to gain the final genome sequence. The coding sequences (CDS) in the genome were predicted by GeneMarkS v4.32 (http:// exon. gatech. edu/ GeneM ark/). The tRNAs and rRNAs of the genome were predicted by the software programs tRNAscan-SE v1.3.1 (http:// lowel ab. ucsc. edu/ tRNAs can-SE) and Barrnap v1.2 (https:// github. com/ tseem ann/ barrn ap), respectively.
The assembled genome was annotated by the kyoto encyclopedia of genes and genomes (KEGG) (https:// www. kegg. jp/ kegg/), the gene ontology (GO) (http:// geneo ntolo gy. org/) , and the evolutionary genealogy of genes: non-supervised orthologous groups (eggNOG) (http:// eggno gdb. embl. de/#/ app/ emapp er). The circular map of the genome was generated using Cgview v1.7.11 (http:// circos. ca). The direct repeats (DRs) and Spacers (interval region) were detected by CRISPRs finder tool (http:// crispr. i2bc. paris-saclay. fr/ Server/). Gene island was predicted by Island Viewer 4 (http:// www. patho genom ics. sfu. ca/ islan dview er/). The FASTA formatted sequences of the magnetosome genes obtained from NCBI database were separately blasted against the genome sequence of At. ferrooxidans BYM. The sequences shared high similarity with the magnetosome genes in the genome of At. ferrooxidans BYM were selected applying a cut-off value of E < 0.01 and functionally predicted based on the published information.

Analysis
The oxidation ratio of ferrous iron was the ratio of oxidized and initial ferrous ion content. The concentration of ferrous iron was tested by the potassium dichromate titration method according to the previous study (Yan et al. 2016). The magnetosome-producing potential was evaluated based on the yields of magnetosomes (mg/g), which was expressed as the dry weight ratio of magnetosomes (mg) to bacterial cells (g). All experiments were conducted in triplicate and the results were expressed as mean ± standard deviation. The statistical analysis was performed with OriginPro 9.0 software (OriginLab, Northampton, USA) by one-way analysis of variance followed by Student's t-test for pairwise comparison. The difference with p < 0.05 was considered statistically significant.

Nucleotide accession number
The accession numbers for the 16S rRNA gene sequence and the whole genome sequence of At. ferrooxidans BYM (plasmid) were deposited in the NCBI database with the accession number KY923140 and CP082238 (CP082239), respectively.

Isolation, magnetic screening and identification
The distinct and vivid color change from colorless to redbrown can be easily observed by naked eyes after inoculating the soil samples into 9 K medium in a shaking incubator with 120 rpm at 30 °C for 2-3 weeks. The serially diluted cultures were separately spread on the surface of 9 K solid medium and incubated at 30 °C for 10 days. It can be observed that the bacterial colonies with an average diameter of 1.0 ± 0.2 mm appeared smooth, round or oval with the rust-colored raised center surrounded by a yellow zone ( Fig. 1a), which was similar to the results reported by Yan et al. (Yan et al. 2017a, b). The single colony was scraped and inoculated into 9 K liquid medium for incubation and magnetic enrichment. The Pasteur tube appeared a brown liquid after few minutes under the "race-track" liquid magnetic separation (Fig. 1b). It can be found that the bacterial cells from the tip end of a cut-off Pasteur tube showed weak magnetotaxis on the semisolid 9 K plates under an artificial magnetic field, while the magnetostatic phenomenon did not occur in the case of no artificial magnetic field (Fig. 1c). After repeating these procedures, the bacterial cells were collected. SEM analysis showed that the morphology of the bacterium is rod-shaped with 0.2-0.4 µm in width and 0.8-1.4 µm in length (Fig. 1d), which was consistent with the previous report .
Sequencing results showed that the length of the 16S rDNA fragment of the isolate was 1429 bp. The NCBI BLAST sequence homology analysis showed the 16S rDNA of the isolate shared 99.93% similarity with At. ferrooxidans ATCC 19859, thus the isolate was named At. ferrooxidans BYM (Fig. 1e). It also can be found that the 16S rDNA of At. ferrooxidans BYM exhibited high similarity (> 97.0%) with six strains of Acidithiobacillus genus, whereas only showed 80.80% homology with M. gryphiswaldense MSR-1 (Fig. 1e).

Magnetosome-producing potential assay
MTB usually mineralize membrane-bound intracellular iron crystals (magnetosomes) which enable the cells to align and swim along magnetic field lines in a preferred direction. The inorganic component of magnetosomes is magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ). In order to biomineralize magnetite, MTB require several orders of magnitude more iron than non-MTB, reaching up to 4.0% of their cell dry weight (Yan et al. 2012a, b). There are many similar characteristics between MTB and At. ferrooxidans such as morphological types, trophic type, motility, ecological distribution, and growth temperature. It has been reported that At. ferrooxidans can synthesize magnetosomes and contain more than 2.0% iron based on dry weight (Yan et al. 2016). Thus, sufficient iron is the crucial substrate for magnetosome-producing bacteria to form magnetosomes. Results showed that At. ferrooxidans BYM can produce magnetosomes within a wide range of ferrous sulfate concentrations (15-40 g/L) (Fig. 2). It appeared that the cell dry weight of At. ferrooxidans BYM increased with the increase of ferrous sulfate concentration, but the yields of magnetosomes were found to increase first and then decrease. The maximum yields of 7.1137 mg/g magnetosomes occurred at 30 g/L ferrous sulfate. However, the magnetosome yields declined to 4.5294 mg/g when the ferrous sulfate concentration increased to 40 g/L. This finding was supported by the previous study in which the high concentration of ferrous iron promoted the growth of bacteria but inhibited the synthesis of magnetosomes (Yang et al. 2001). Additionally, the previous works suggested that the intracellular iron content of M. magneticum AMB-1 cells presented a logarithmic increase with increasing external iron concentration from 10 to 100 µM. However, a mutant of M. magneticum AMB-1 lacking MAI which was unable to form magnetosomes can still incorporate a large amount of iron (Amor et al. 2020). Therefore, the iron demand for the growth of MTB cell was not completely consistent with that for magnetosomes synthesis, which is similar to our result.
The genes associated with the synthesis of magnetosomes such as mpsA-like, magA-like and mamB-like existed in the chromosome of At. ferrooxidans BYM. It has been reported that the highest expression level of these genes appeared at 150-200 mM of ferrous concentration, while the iron-deficient or iron-excessive conditions (Below or above 150-200 mM ferrous concentration) could adversely affect their expression (Liu et al. 2008a, b). Additionally, the jarosite precipitation formed by hydroxylation would increase with the increase of the ferrous concentration, which could gradually cover the bacterial surface and hinder the electron transfer from iron as well as proton diffusion from bacterial cells, resulting in the reduction of magnetosome formation (Yan et al. 2016).
One of the most important factors affecting microbial growth and magnetosome synthesis is oxygen. It has been demonstrated that magnetosome formation in MTB occurred only in a narrow range of low oxygen concentration. Dissolved oxygen (DO) was the major factor affecting the magnetosome production of M. gryphiswaldense MSR-1 and the yields in a microaerobic condition with 5-10 ppm oxygen was found to be 1.43 times higher than that in anaerobic condition (Jajan et al. 2019). The formation of magnetosomes in M. magneticum AMB-1 was favorably carried out at the low level of O 2 (≤ 18.7 µM), but was strongly inhibited under high O 2 concentrations (50-100 µM) (Popa et al. 2009). The effect of oxygen on the magnetosome-producing potential of At. ferrooxidans BYM were investigated under different aeration rates (0.5-1.5 L/min). Results showed that the formation of magnetosomes produced at any given aeration rate and the yields of magnetosomes ranged from 0.5896 to 4.5336 mg/g (Fig. 2). It was observed that the bacterial biomass significantly increased with the increase of ventilatory capacity due to the typical aerobic characteristic of At. ferrooxidans BYM, while the magnetosome yields tended to saturation and decreased gradually with further increase of air supply. The low aeration rate might cause the low level of oxygen dissolved in 9 K medium, which could inhibit the growth of At. ferrooxidans BYM, resulting in a lower final cell density and further a lower magnetosome production. This finding indicated that the oxygen demand for cell growth of At. ferrooxidans BYM was higher than that for magnetosome formation.
It appeared that the magnetosome yields began to decrease once the aeration rate exceeded 1.2 L/min (Fig. 2). The enhancement of oxygen mass transfer in the fermentation system possibly caused by the increase of aeration rate could cause high autooxidation of ferrous ion (Zhang et al. 2012), which would compete with biological oxidation mediated by bacteria, leading to an advice effect on the growth of At. ferrooxidans BYM. The reactive oxygen species (ROS) produced by the metabolism of At. ferrooxidans BYM with O 2 as the terminal electron acceptor can damage DNA ultimately leading to cell death. It has been reported that the formation of magnetosomes can efficiently scavenge the intracellular ROS in MTB (Guo et al. 2012). Therefore, the magnetosome formation in At. ferrooxidans might play critical role in removing ROS to avoid damage to cells. These indicated that the suitable aeration rate can not only satisfy the oxygen requirement for bacterial oxidation of ferrous ions, but also maintain the balance between ROS scavenging and magnetosome formation. Nitrogen source is also an important impact factor for the growth of MTB and magnetite magnetosome biomineralization. It can be seen from Fig. 2 that the ammonium sulfate in the concentration range of 0.24-1.80 g/L was well documented to favor magnetosome formation. The magnetosome yields increased from 2.0099 to 13.1291 mg/g with ammonium sulfate concentration (Fig. 2). It has been reported that there is a substantial demand for nitrogen to form proteins for the assembly of magnetosome vesicles (Yan et al. 2016). Additionally, the magnetosome-associated membrane proteins play a significant role in magnetosome crystals growth inside MTB (Cornejo et al. 2016). Therefore, in the case of a sufficient supply of nitrogen sources, At. ferrooxidans BYM can efficiently form magnetosomes. Generally, the formation of magnetosomes would be adversely affected once the concentration of ammonium sulfate exceeds a certain value. It can be observed that the magnetosome yields decreased to 7.6584 mg/g when the ammonium sulfate concentration exceeded 1.2 g/L. In such cases, the production of massive amounts of jarosite layer on the cell surface at high ammonia ion concentration could prevent the metabolism of bacteria (Zhang et al. 2012), resulting in the reduction of magnetosome yield. Therefore, the nitrogen source was not only necessary for the cell growth of magnetosome-producing bacteria but was also essential for magnetosome formation. Our finding was consistent with the previous studies on MTB (Jajan et al. 2019;Wang et al. 2020). It has been confirmed that NaNO 3 was more likely to favor the magnetosome formation of M. gryphiswaldense MSR-1 than NH 4 Cl and (NH 4 ) 2 SO 4 , since NaNO 3 was used as terminal electron acceptor to alternate oxygen for the survival of microaerobic MTB. Although the cell growth rate of M. gryphiswaldense MSR-1 increased by 27% with the increase of NaNO 3 concentrations from 20 to 60 mM, the magnetosome yields decreased (Jajan et al. 2019). Low nitrate concentration (0.004 M) was found to promote the magnetosome formation Although organic compounds usually appear adverse effects on the growth of At. ferrooxidans, our results suggested that the gluconic acid in concentrations ranging from 5 to 15 mM can promote the yields of magnetosomes from 2.6009 to 6.8854 mg/g (Fig. 2). The gluconic acid might act as an iron-chelating agent to chelate ferric ions and enhance the bioavailability of iron by forming ferric-ligand complexes. This phenomenon was in agreement with the results of Edouard Alphandéry et al., who noted the growth of M. magneticum AMB-1 and the production of magnetosomes were improved by the chelating agents including ethylenediaminetetraacetic acid (EDTA), ascorbic acid, citric acid, alendronic acid, and neridronic acid (Alphandéry et al. 2012). The magnetosome formation depended on siderophores with a high and specific affinity for the ferric ion to incorporate iron, which has been described in diverse groups of MTB (Yan et al. 2012a, b). It has been reported that At. ferrooxidans can synthesize several potential siderophore outer membrane receptors (OMRs) involved in transporting ferric ions to the cytoplasm (Quatrini et al. 2005). Therefore, the ferric-chelating agents can save energy consumption for synthesizing siderophores and OMRs, thus providing more energy for cell growth. Additionally, gluconic acid could alleviate the passivation formation and remove the precipitate layer on the surface of bacteria, since it can chelate with ferric iron of jarosite which is the main component for passivation and precipitate layer (Alphandéry et al. 2012). Thus, the adverse effect of ferrous iron metabolic by-products on nutrient mass transfer and energy metabolism of At. ferrooxidans BYM could be relieved to a certain extent, resulting in a favorable conidiation for bacterial growth and magnetosome formation. Moreover, the redox potential of the cultivation system of At. ferrooxidans is dominated by the ratio of dissolved ferric to ferrous ions. Adding gluconic acid might affect the redox potential and pH because jarosite formation can be eliminated or partially alleviated in high redox potential and low pH (Bevilaqua et al. 2014). It was observed that the magnetosome yields began to decrease with the increase of the gluconic acid concentration ranged from 15 to 50 mM (Fig. 2). Two possible explanations for this phenomenon can be considered. One is that high concentration of organic chelating agent might cause toxicity to autotrophic At. ferrooxidans BYM and affect its growth. The other reason is that the reactive oxygen species generated as the result of the alcohol functional unbinding from iron could damage the out membrane of bacterial cells.

TEM, STEM and HRTEM observation
It has been documented that the magnetosomes exist as one, two, or multiple chains inside living MTB (Yan et al. 2012a, b). Due to the presence of these magnetosomes, MTB have the ability to move along the applied external magnetic field. It can be observed that At. ferrooxidans BYM was capable of synthesizing intracellular magnetosomes which were not arranged in chain but distributed irregularly in cells (Fig. 3a). The magnetotaxis and magnetism of At. ferrooxidans BYM are very weak (Fig. 1c), which might be attributed to the random distribution of magnetosomes, which is a way of storing iron sources in the cell (Zhang et al. 2012).
TEM observation indicated that the number of magnetosomes per cell was 6-9 in At. ferrooxidans BYM (Fig. 3a, white arrows). A comparison of the magnetosome number between At. ferrooxidans BYM to other magnetosome-producing bacteria was shown in Table 1  Although the magnetosome number per cell of At. ferrooxidans BYM is similar to or lower than most MTB (Table 1), At. ferrooxidans BYM is still promising for magnetosome production due to its mild culture condition, simple nutritional requirement, and short incubation time. The particle size of the magnetosomes in At. ferrooxidans BYM varied from 20 to 80 nm ( Fig. 3a and b), which was similar to the finding in At. ferrooxidans BY-3 (Yan et al. 2013). It has been addressed that the size of the mature magnetosomes in most MTB fall within a narrow range of 35-120 nm, which were in the magnetic-single-domain range (Yan et al. 2012a, b). Therefore, the magnetosomes formed in At. ferrooxidans BYM appeared a single-domain (SD) size, possibly resulting in the permanent magnetism of the magnetosomes and thus the weak magnetotaxis behavior of cells.
We performed STEM-EDXS elemental mapping and HRTEM investigations on the magnetosomes to localize and characterize the intracellular iron mineral phases. It appeared that At. ferrooxidans BYM accumulated iron and intracellularly formed amorphous granules with an elemental composition dominated by Fe, C, N and O (Fig. 3c-h). This indicated that the inorganic crystal was magnetite and composed of organic membrane. The finding was supported by the previous studies, in which the magnetosomes of MTB and ferrous oxidizers from freshwater were consisted of magnetite crystal and phospholipid bilayer membrane (Yan et al. 2017a, b). It has been proved that the magnetosomes from At. ferrooxidans BYM characterized by TEM and FTIR in our previous study were wrapped by magnetosome membrane consisted of carboxylate group, hydroxyl group, and primary amino group (Zhang et al. 2021). It can be seen from Fig. 3b (white arrows) and Fig. 3i that the crystals mainly exhibited elongated prismatic shape which usually existed in the MTB affiliated with the alpha, eta and gamma subclasses of Proteobacteria (Dieudonné et al. 2019). The presence of the irregular shaped crystals in Fig. 3b (black arrows) might be due to immaturity in magnetosome biomineralization. Additionally, no crystal defects such as twinning, stacking fault and cation vacancies were detected in the isolated magnetosomes (Fig. 3b, i), indicating that no obvious oxidization occurred within magnetite crystals during magnetosome extraction. The fast fourier transform (FFT) analyses indicated that the magnetosome crystals expressed {110} crystal faces (Fig. 3j), and the lattice spacing of nanoparticles was about 2.8 Å, which was consistent with the distance of the {111} crystal face. These results implied that the magnetite crystals elongated along the [111] direction. The similar finding was also reported in the previous studies. The lattice spacings of nanoparticles in A. ferrooxidans ATCC 23270 were found to vary from 2.43 to 2.95 in the previous study (Wu et al. 2019). It has been reported that the magnetosomes in M. magnetotacticum MS-1 were elongated-prismatic and grew parallel to [111] direction (Pósfai et al. 2013).

Whole-genome sequencing and magnetosome-producing candidate gene prediction
The rapid development of genome sequencing has obtained the clear insight on the magnetosome-producing mechanism of MTB. Deciphering the whole genome sequence is necessary for understanding the genes involved in the formation process of magnetosomes. It has been reported that ten strains among more than 500 At. ferrooxidans sensu stricto isolates have been sequenced so far, but only three strains (ATCC 23270, ATCC 53993, YNTRS-40) were sequenced to obtain the complete genome sequences (Table 2). Illumina sequencing showed that the genome of At. ferrooxidans BYM consisted of a circular chromosome of about 3208.389 kb with 58.54% GC content, ORF number of 3260, 3 CRISPRs, and a plasmid of 47.11 kb with a GC content of 56.44%, ORF number of 54. The genes categorized in "biological process", "cellular component" and "molecular function" were determined in the genome of At. ferrooxidans BYM. The presence of numerous signal transduction components suggested that the At. ferrooxidans could greatly regulate the cellular function to adapt to varying environmental conditions. GO functional annotation indicated that a total of 314 genes and 8 genes closely were related to iron metabolism on chromosome and plasmid, respectively (Fig. 4). Among them, 94 genes were associated with iron-sulfur cluster binding, followed by 49 genes related to iron ion binding. The genes concerned with ferric iron transport, ferrous iron transport, iron assimilation and iron chelate transport participated in the multiple metabolic pathways of At. ferrooxidans BYM.
Blast analysis indicated that several genes possibly related to magnetosome formation existed in the genome of At. ferrooxidans BYM, including the major category mam genes, followed by feoAB, magA, mpsA and mms6 (Table 3) et al. 2003;Rong et al. 2008;Lefèvre et al. 2012;Lefevre et al. 2012;Dziuba et al. 2016). The formation mechanism of magnetosomes in At. ferrooxidans might be different from that in MTB due to the apparent lack of MAI. Therefore, the similarity of magnetosome formation gene in At. ferrooxidans was slightly lower than that in MTB. Thus, we set > 60% as a proper cut-off value to select magnetosome genes to provide a basis for transcriptome analysis. The result indicated that the mamA-like gene might exist in At. ferrooxidans BYM because the gene fragment in the genome of BYM shared 87.88% sequence identity with mamA from Alphaproteobacterium LM-1. It has been proved that the magnetosome-associated protein MamA encoded by mamA gene (651 bp) covers the outside of magnetosomes and plays a critical role in protein sorting through protein-protein interactions (Zeytuni et al. 2011;Nguyen et al. 2016). The function was found to be based on its special structure because MamA folds as a sequential tetratricopeptide repeat (TPR) protein with three protein-protein interaction sites including a concave site, a convex site, and a putative TPR repeat in M. magneticum AMB-1 and M. gryphiswaldense MSR-1. MamA self-assembles through its putative TPR motif and its concave site create a large homooligomeric scaffold that can interact with other magnetosome-associated proteins via the convex site (Zeytuni et al. 2011). It has been demonstrated that the TPR motifs existed in a wide variety of proteins from prokaryotes are important for cells to execute the desired functions such as protein transport, protein folding, transcription and splicing, and cell cycle control (Nguyen et al. 2016). Additionally, another function based on studies of a mamA deletion mutant has been proposed that MamA appears to activate or prime the preformed magnetosomes to biomineralization. Deletion of mamA has no effect on magnetosome membrane invagination, but fewer crystals were formed in M. magneticum AMB-1 (Komeili et al. 2004).
The genes of mamB in M. moscoviense BB-1and mamM in Gammaproteobacterium SS-5 separately shared 83.78-87.88% similarity with the sequences in At. ferrooxidans BYM (Table 3). It has been reported that MamB and MamM are conserved and abundant integral membrane magnetosome-associated proteins with a molecular weight of 31.9-34.4 kDa, respectively (Nudelman and Zarivach 2014). They were thought to be involved in iron transport during magnetosome formation because of their high similarity to the members of the cation diffusion facilitator family (CDF), which generally transport metal cations such as ferrous ions from the cytoplasm into intracellular compartments or into the extracellular space using the proton motive force (Uebe et al. 2011). These might be the reason that could account for the increase of magnetosome yields of At. ferrooxidans BYM with increasing ferrous sulfate concentration. It has been confirmed that MamB not only serves as iron transporter in magnetite biomineralization, but also acts as a landmark protein to initiating the formation of magnetosome vesicle in a transport independent process (Uebe et al. 2018). The previous studies suggested that the stable expression of MamB depended on the presence of MamM which could protect MamB from proteolytic degradation due to complex formation (Uebe et al. 2011). Additionally, the stabilization of MamB and MamM was found to directly involve in the regulation of magnetite crystal growth in magnetosome formation (Siponen et al. 2012). Therefore, the potential proteins coded by mamB-like and mamM-like genes in At. ferrooxidans BYM might execute iron transport and crystal size control.
It has been reported that mamE and mamO genes might play critical roles in the early stage of magnetite magnetosome formation in M. magneticum AMB-1. MamE, a HtrA protease consisted of 655 amino acids in M. gryphiswaldense MSR-1, which is responsible for the regulation of the Fe(II)/Fe(III) ratio to promote magnetite nucleation, and also appears to play an important role in protein sorting to the magnetosome membrane (Quinlan et al. 2011;Siponen et al. 2012;Nudelman and Zarivach 2014). Deletion of mamE can cause the empty magnetosome vesicles and the mislocalization of several proteins within the cell (Quinlan et al. 2011). It was found that mamE gene in Gammaproteobacteria SS-5 shared 70.19% identity with the gene fragment (mamE-like) in At. ferrooxidans BYM (Table 3). The previous study showed that Gammaproteobacteria SS-5 isolated from the southeastern shore of the Salton Sea could mineralize octahedral magnetite crystal chains (Lefevre et al. 2012). Similar to MamE, MamO is a second HtrA/DegP family protease containing 632 amino acids (Murat et al. 2010). Additionally, MamO contains a degenerate active site, rendering it incapable of exerting protease activity. It has been reported that MamO can promote magnetosome formation through two genetically distinct, noncatalytic activities including the activation of MamE-dependent proteolysis of biomineralization factors and direct binding to transition metal ions (Hershey et al. 2016). The GO results indicated that 49 genes related to iron ion binding existed in At. ferrooxidans BYM. The mamO in Magnetospirillum marisnigri SP-1 displayed a strong homology (92.59%) with the sequence in At. ferrooxidans BYM, suggesting that the potential mamO-like gene in At. ferrooxidans BYM might specifically promote magnetite nucleation. It has been proved that MamK and MamJ play key roles in the formation of magnetosome chain in MTB (Nudelman and Zarivach 2014). However, we did not find the genes with high similarity to mamK and mamJ in genome of At. ferrooxidans BYM, which was supported by the TEM observation that the magnetosomes appeared to arrange dispersively. Table 3 showed that the mamQ gene in Alphaproteobacterium SS-4 shared 80.39% sequence identity with the possible gene in At. ferrooxidans BYM. It has been suggested that magnetosome membrane formation was genetically induced via the mamQ gene by acting as a hub for the early organization of magnetosome proteins prior to membrane invagination. The deletion of mamQ in M. magneticum AMB-1 cells resulted in the complete loss of magnetosome formation (Murat et al. 2010). It has been reported that an integral membrane protein MamQ with 273 amino acids was found to be homologous to the LemA protein family which bears a potential resemblance to BAR domain proteins and involves bending membranes in eukaryotic cells (Nudelman and Zarivach 2014).
It can be found that the magA-like gene might exist in At. ferrooxidans BYM due to it shared high similarity (79.55%) with magA sequence in M. magnetotacticum MS-1. MagA consisted of 434 amino acids was found to have high homology with the cation efflux proteins including the potassium ion-translocating protein KefC in Escherichia coli, and the putative Na + /H + -antiporter NapA from Enterococcus hirae. It has been reported that MagA was an iron transporter in E. coli which was confirmed by direct iron uptake measurements in membrane vesicles (Nakamura et al. 1995). More importantly, the gene fragment with high similarity with magA gene was also found to exist in At. ferrooxidans 23270 and its maximum expression level occurred at 150-200 mmol/L ferrous (Liu et al. 2008a, b). In the present study, the highest magnetosome yield was observed when the concentration of ferrous sulfate was 30 g/L (Fig. 2). These results indicated that the potential protein coded by magA-like gene in At. ferrooxidans BYM could participate in the iron transport during the process of magnetosome formation.
It has been reported that mpsA with a length of 954 bp plays a key role in the formation of magnetosome membrane invaginated from cytoplasmic membrane. The MpsA protein consisted of 317 amino acids has been identified from M. magneticum AMB-1 . It is homologous to acyl-CoA carboxylase of E. coli, which was classified under the category of lipid metabolism, providing  Uebe et al. (2011) evidence that the invagination of the cytoplasmic membrane to form magnetosome membrane were mediated by acylation (Tanaka et al. 2006). The mpsA gene in M. magneticum AMB-1 was found to share 64.82% identity with the sequence in At. ferrooxidans BYM. These indicated that mpsA-like gene might play a critical role in the process of magnetosome formation in At. ferrooxidans BYM, which served a similar function in MTB. The previous report suggested that the mpsA gene in At. ferrooxidans reached the highest expression level at 150-200 mmol/L of ferrous and its expression increased with the decrease of oxygen concentration ). These findings indicated that mpsAlike gene was significantly affected by ferrous concentration and oxygen concentration and tightly associated with the magnetosome formation in At. ferrooxidans.
Mms6, a small acidic protein consisted of 136 amino acids in M. gryphiswaldense MSR-1 strain, was predicted to control the size of the magnetite crystal (Tanaka et al. 2011). It was also isolated from M. magneticum AMB-1 along with Mms5, Mms7 and Mms13. The Mms6 protein has a low complexity hydrophobic N-terminal region with a hydrophobic glycine-leucine sequence, and a hydrophilic C-terminal region rich in acidic amino acids (aspartic-glutamic) with iron-binding activity (Rawlings et al. 2020). It has been reported that the mms6 gene deletion mutant of M. magneticum AMB-1 was found to synthesize the smaller magnetite crystals with uncommon crystal faces, while the wild-type and complementation strains synthesized highly ordered cubo-octahedral crystals (Tanaka et al. 2011

Hypothetical mechanism of magnetosome formation in At. ferrooxidans
So far, the understanding of magnetosome formation and iron biomineralization process in the magnetosome-producing bacteria mainly comes from two model MTB strains M. gryphiswaldense MSR-1 and M. magneticum AMB-1 (Dieudonné et al. 2019). It has been demonstrated that the synthesis of bacterial magnetite magnetosomes is not a simple transformation from ionic iron to nano-magnetite particles, but includes a series of processes such as invagination of the cytoplasmic membrane, formation of magnetosome vesicles, iron uptake and transport, formation of iron mineral precursors and crystal nucleus growth and maturation (Yan et al. 2012a, b). The magnetosome gene island (MAI), a set of about 40 genes located in five operons including mamG-FDC, mms6, mamAB, mamXYZ, and feoAB, was proved to involve in the step-wise magnetosome genesis mentioned above.
Therefore, based on the prediction of the magnetosomeproducing candidate genes in At. ferrooxidans BYM and the information of magnetosome formation mechanism in MTB, we proposed a possible mechanism for the formation of magnetosomes in At. ferrooxidans. The hypothetical model of magnetosome formation in At. ferrooxidans is present in Fig. 5. The formation of magnetosomes in At. ferrooxidans is hypothesized to comprise three independent and constitutionally coupled processes. (1) The protein encoded by mamA-like gene could initiate invagination of the cytoplasmic membrane to form magnetosome membrane once the ferric uptake regulator (Fur) is able to sense the external ferrous iron ions. (2) The Fe(II) then transport from the outer membrane into the periplasm via the iron transporter FeoAB and/or Mnth. Meanwhile, a portion of Fe(II) is oxidized by Cyc1 and/or CycA1 to Fe(III) (Zhang et al. 2018). Then MamB-like, MamM-like and MagA-like proteins transport Fe(II) and Fe(III) into the magnetosome vesicle to generate hydrated ferric oxide (Fe 2 O 3 ·nH 2 O), which triggers the growth of magnetosome vesicle involving the proteins of MamB-like, MamQ-like and MpsA-like. (3) The dehydration and redox of hydrated ferric oxide simultaneously occurred to form the magnetite precursor with the involvement of MamE-like, MamM-like, MamO-like and Mms6like (Yan et al. 2012a, b). A stoichiometric ratio of Fe(III)/ Fe(II) = 2:1 and the low redox potential inside magnetosome vesicle were regarded as necessary for crystal formation. Thus, the occurrence of the crystal nucleation mediated by MamM-like and MamO-like might result in the maturation of magnetosome from magnetite precursor.

Conclusion
A magnetosome-producing bacterium was successfully isolated and identified as At. ferrooxidans BYM capable of forming magnetosomes at 30 °C under a wide range of operating parameters such as ferrous sulfate (15-40 g/L), ammonium sulfate (0.24-1.8 g/L), gluconic acid (5-50 mM), and aeration rate (0.5-1.5 L/min). TEM showed the magnetosomes in the size of 20-80 mM distributed irregularly in cells. STEM-EDXS and HRTEM-FFT indicated that the elongated prismatic magnetite magnetosome crystal with {110} crystal faces elongated along the [111] direction. The genome of At. ferrooxidans BYM is consisted of the chromosome and the plasmid, harboring a total of 322 genes closely related to iron metabolism. Ten genes involved in magnetosome formation were predicted but were not acquired by multiple gene-transfer events during the bacterial evolution. A hypothetic mechanism of magnetosome formation in At. ferrooxidans comprising magnetosome vesicle formation, iron uptake and mineralization, and magnetite crystal maturation was proposed. These findings would lay the foundation for more profound theoretical research and commercial development of the magnetosomes of At. ferrooxidans BYM in the future.
Author contributions LY contributed to conceptualization and validation, investigation and funding acquisition; DZ performed writingoriginal draft preparation; DZ and JNY performed writing-reviewing and editing; WDW provided methodology; SZ and GJZ performed data curation and visualization; DL was involved in supervision; LY performed project administration. All authors read and agreed to publish the final manuscript.