Development of a Genetically Encoded Magnetic Platform in Magnetospirillum gryphiswaldense MSR-1 for Downstream Processing of Protein Expression System

Background: Protein downstream processing remains a challenge in protein production, especially in low yields of products, in spite of ensuring effective disruption of cell and separation of target proteins. It is complicated, expensive and time-consuming. Here, we report a novel nano-bio-purification system for producing recombinant proteins of interest with automatic purification from engineered bacteria. Results: This system employed a complete genetic engineering downstream processing platform for proteins at low expression levels, referred to as a genetically encoded magnetic platform (GEMP). GEMP consists of four elements as follows. (1) A truncated phage lambda lysis cassette (RRz/Rz1) is controllable for lysis of Magnetospirillum gryphiswaldense MSR-1 (host cell). (2) A surface-expressed nuclease (NucA) is to reduce viscosity of homogenate by hydrolyzing long chain nucleic acids. (3) A bacteriogenic magnetic nanoparticle, known as magnetosome, allows an easy separation system in a magnetic field. (4) An intein realizes abscission of products (nanobodies against tetrabromobisphenol A) from magnetosome. Conclusions: In this work, removal of most impurities greatly simplified the subsequent purification procedure. The system also facilitated the bioproduction of nanomaterials. The developed platform can substantially simplify industrial protein production and reduce its cost.


Background
Native and recombinant peptides and proteins are becoming increasingly important as enzymes and noncatalytic functional products (e.g., antibodies, hormones, factors and vaccines) for industrial, nutritional, medical, and agricultural applications [1]. A prokaryotic system is most widely used for protein production at both laboratory and industrial scales. Such a system allows to rapidly produce large quantities of recombinant proteins [2]. Downstream processing refers to the recovery and puri cation of biosynthetic products and remains a challenge in biotechnology. Protein puri cation is the most time-consuming and costliest step in protein production [3,4]. In prokaryotes, puri cation is more complex because recombinant proteins are mainly produced inside the cells [5]. To purify recombinant proteins in large quantities, it is imperative to ensure effective lysis of cells and separation of target proteins. The most commonly used methods for cell lysis are the enzymatic digestion and physical disruption by sonication or high pressure homogenization. Separation processes are designed to isolate target molecules in the presence of a variety of impurities from a fermentation broth that contains not only many different biomolecules, but also cell debris and salts. The processes typically require numerous steps: ltration, centrifugation, occulation, sedimentation or crystallization, and chromatographic separation. Those steps require special equipment and often suffer from loss of sample during processing, heat generation, contamination, and high cost. Therefore, downstream processing often reaches to 50-90% of the total production costs in most biotechnological products [6], especially for low yield products.
Magnetic separation utilizes a magnet to attract magnetic substances out of non-magnetic components in a mixture. It enables the fast and direct capture of target molecules in fermentation broths [7]. However, magnetic separation requires high-gradient magnetic elds for target complex separation and hazardous elution buffers such as imidazole to weaken interactions between desired products and magnetic particles [8]. An alternative separation carrier is probably bacterial magnetosome or bacterial magnetic nanoparticle, which has been recently studied for drug delivery [9]. Magnetosomes are special organelles of magnetotactic bacteria such as Magnetospirillum gryphiswaldense MSR-1. Magnetosomes are composed of membrane-enveloped magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ) crystals in diameters ranging from 30 to 120 nm [10]. Magnetosomes have a perfect crystalline core, a quasi-spherical shape and a fair uniform particle size. Their magnetism is generally much stronger than the arti cial magnetic nanoparticles of iron oxides [11]. When used as a separation carrier, magnetosomes do not need highgradient magnetic elds. Magnetosomes are inside bacteria, which can be well used for tailored strategy in downstream processing [12].
The feasibility to fusion express antibodies, uorophores, enzymes and receptors on magnetosome surfaces (also known as magnetosome surface display) was recently demonstrated and highly attractive for many biomedical and biotechnological applications [13]. Nanobody (Nb) was functionally expressed on the surface of magnetosome via a surface display technique [14]. Nbs are de ned as single-domain variable fragments of camelid-derived heavy-chain antibodies. Nbs have arisen as an alternative to conventional antibodies and show great potential in diagnostics and therapy [15]. In 2020, the global Nbs market was worth of U.S. $132 million, and by 2028, it is projected to be worth of U.S. $9,192 million. Nb expression is extensively studied in both prokaryotic and eukaryotic systems, though most of these works fail to reach high production levels so far [16]. However, few reports have been focused on downstream Nb separation and puri cation up to now.
Tetrabromobisphenol A (TBBPA) is a brominated ame retardant used in 90% of epoxy coated circuit boards [17]. Its potential toxicity to human and bioaccumulation property have brought close attention [18]. An anti-TBBPA Nb immunoassay is a promising method for monitoring TBBPA in the environment [19]. Compared with the classical polyclonal antibody-based immunoassay, the Nb-based immunoassays show excellent sensitivity of TBBPA detection [20].
This study aimed to purify recombinant protein in a simple and economical way, and a novel genetically encoded magnetic platform (GEMP) (Fig. 1), which was capable of simultaneous biosynthesis of proteins of interest (e.g. anti-TBBPA Nb) and automatic separation upon their expression by engineered bacteria, was constructed and evaluated. We chose gram-negative bacteria MSR-1 with magnetosomes as a prokaryotic expression system for subsequent construction to ensure production and automatic separation of recombinant proteins. The system contained a truncated phage lambda lysis cassette (SRRz/Rz1), a non-speci c nuclease NucA, a protein-splicing element of Pyrococcus abyssi DNA polymerase II (Pab PolII intein), and a Nb against TBBPA. The separation procedure of the GEMP included three automatic steps ( Fig. 1): cell lysis of bacterial host, magnetic separation of magnetosome-Nb complexes, and release of Nbs from the complexes. Anti-TBBPA Nb was used as a model protein. The results showed a promise for a smart platform for extraction and puri cation of target proteins.

Functional components of GEMP
As a component of GEMP, the phage lambda lysis cassette consisted of four genes: S, R, Rz and Rz1. The S gene encodes holin and its inhibitor. The R gene encodes the endolysin. Rz and Rz1 are nested genes encoding Rz and Rz1 proteins, respectively, involved in resolving the oligopeptide linkages. Holin is a small membrane protein, which forms µm-scale holes in host cytoplasmic membrane (inner membrane). These holes result in the release of R endolysin, Rz and Rz1 proteins, thus accessing to their substrate (host cell wall), and at last, the disruption of host cell [21]. The surface attached extracellular nuclease (Nuc) of Staphylococcus aureus is a secreted enzyme that possesses a long 60-residue Sec signal sequence. The secreted form of Nuc, known as NucB, is processed by most S. aureus strains to a shorter form called NucA [22]. The fusion of NucB and the signal sequence of E. coli's major outer membrane protein (OmpA) resulted in accumulation of NucA in the periplasm of E. coli [23]. NucA can hydrolyze DNA and RNA, whether double or single strand, into oligo-and mononucleotides. NucA was here responsible for reducing extract viscosity by completely digesting genomic DNA [24].
To avoid cell lysis during cultivation, the elements mentioned above need to be expressed at different subcellular locations. The truncated phage lambda lysis cassette and the NucA were expressed in cytoplasm and periplasm, respectively, while the intein and Nb were expressed on the surface of magnetosome (Fig. S1). These elements were cloned into different plasmid vectors, introduced into E. coli S17-1 via transformation, followed by function veri cation of the rst two elements. All of vectors was then transferred into MSR-1 via bacterial biparental conjugation.
Transfer of nucB and RRz/Rz1 into E. coli S17-1 In the present work, only R, Rz and Rz1 in the phage lysis cassette were employed and cloned into a plasmid vector pBBR1MCS-2 to construct recombinant strains. The S gene was omitted to avoid host lysis during cultivation. The native promoter P R' from λ phage was cloned and employed for the expression of RRz/Rz1 genes. Besides, the DNA fragment encoding OmpA signal peptide was fused with nucB, and then cloned into the plasmid vector pBBR1MCS-2, along with the RRz/Rz1. The lac promoter from pBBR1MCS-2 was employed for the expression of NucA. The resulting plasmid was referred to as pBBRONL ( Fig. 2A). Another recombinant plasmid pBBRPNL was also constructed, which was the same as pBBRONL, except that the signal peptide sequence was from a phaZ1 gene instead of ompA ( Fig. 2A). The PhaZ1 here was an "intracellular" poly(3-hydroxybutyrate) (PHB) depolymerase of Rhodospirillum rubrum, which is a periplasm-located protein with speci city for native PHB and with structural similarity to extracellular PHB depolymerases [25]. R. rubrum is closely related to Magnetospirillum spp., of which 90% of the 16S rRNA sequence is identical to that of wild type (WT) MSR-1 [26]. The plasmids pBBRONL and pBBRPNL were then transferred into S17-1 and the resulting recombinant strains were referred to as ONL and PNL, respectively ( Fig. 2A).
The target genes nucB and RRz/Rz1 in recombinants were identi ed by colony polymerase chain reaction (PCR) and sequencing. On isopropyl β-D-1-thiogalactopyranoside (IPTG)/DNase agar plates, transparent zones surrounding the colonies of ONL and PNL were observed, while those of controls (S17-1) did not occur (Fig. 2B). NucA was presumably transported across cytoplasmic membrane into periplasmic space under the direct of signal peptide from OmpA or PhaZ1. To con rm this, two recombinant strains and their host strain (S17-1) were cultivated in shaking asks. Their cell pellets were harvested by centrifugation and resuspended into 3 mL of Tris-HCl (20 mM, pH 8.0), followed by the addition of CHCl 3 (20 µL). The majority of ONL and PNL cells was disrupted within 0.5 h, while no apparent changes were observed from S17-1 cells (Fig. 2C), indicating the successful expression of RRz/Rz1 in the cytoplasm of two recombinant strains. Compared to the viscosity of the homogenate from modi ed S17-1 strains which expressed RRz/Rz1 only, the viscosity from disrupted S17-1 expressing both NucA and RRz/Rz1 was lower (Fig. S2).
Construction of an engineered MSR-1 harboring nucB, RRz/Rz1 and Nb RRz/Rz1 and nucA were successfully expressed in E. coli S17-1, and their protein products exhibited speci c activities in the process of cell self-lysis and nucleic acid hydrolysis. We assumed that RRz/Rz1 and nucA could be expressed in other Gram-negative bacteria including magnetotactic bacteria. Herein, a recombinant MSR-1 was developed for automatic downstream processing. The plasmid pBBRPNL with the RRz/Rz1 and recombinant nucB was employed and transferred into MSR-1. The signal peptide sequence of the recombinant nucB gene in this plasmid was from phaZ1 gene of R. rubrum, which was closely related to MSR-1. A fusion gene of the anti-TBBPA Nb, an intein, and MamC was also introduced into MSR-1 via another plasmid vector. The intein gene was cloned from DNA polymerase II of Pyrococcus abyssi, abbreviated as the Pab PolII intein [27]. MamC was the most abundant protein on the magnetosome surface, which helped to express the anti-TBBPA Nb and the Pab PolII intein and anchor them on the surface of magnetosomes [28]. Different from the plasmid pBBRPNL, a suicide plasmid vector pK18mobSacB was employed for cloning the fusion gene of Nb and intein, resulting a recombinant plasmid pKTBInC. After being transferred into MSR-1, the anti-TBBPA Nb and Pab PolII intein genes were integrated into the host chromosome. The recombinant MSR-1 strain was referred to as TBInCPNL (Fig. 3A). An additional control strain was also constructed and referred to as TBInC, which contained the fusion gene of anti-TBBPA Nb and Pab PolII intein in its chromosome but did not harbor the plasmid pBBRPNL (i.e., without the RRz/Rz1 and nucB genes in this strain).
The western-blot analysis showed that the proteins RRz/Rz1, NucA and Nbs-Intein-MamC extracted from TBInCPNL migrated as expected (

Culture Of Tbincpnl
The growth and magnetic response of TBInCPNL in shaking culture To evaluate the impacts of exogenous genes on the growth of recombinant strains, the OD 565 value and magnetic response (Cmag) of TBInC and TBInCPNL growing in 100 mL of sodium lactate medium (SLM) were detected and compared with those of WT MSR-1 (Fig. 4A). The growth curves of three strains almost overlapped. The maximum OD 565 values of MSR-1, TBInC and TBInCPNL were 1.53, 1.46 and 1.49, respectively. After 53 h of cultivation, cells were harvested via centrifugation to yield 0.92, 0.96, and 0.90 g (wet weights, ww) of MSR-1, TBInC, and TBInCPNL, respectively. Although the cell yields varied slightly, the strains showed different magnetic responses. Remarkably, the average Cmag values of TBInC and TBInCPNL were approximately 1.5-and 2-fold higher than that of WT MSR-1, respectively.

Magnetosome's characteristics
The morphology of magnetosomes was characterized by transmission electron microscope (TEM).
Magnetosomes from different strains all appeared in a chain ( Fig. 4B (i-iii)). The size and yield of magnetosomes were determined with the ImageJ (Fig. 4B (iv-v)). The diameters of magnetosomes were mostly distributed in a range of 20-50 nm ( Fig. 4B(iv)). No signi cant differences were observed in the size of magnetosomes biosynthesized by MSR-1, TBInC, and TBInCPNL, with an average diameter of 32.7 ± 7.9, 33.1 ± 7.7, and 33.0 ± 7.4 nm, respectively ( Fig. 4B(iv)). However, dramatic differences were observed in the numbers of magnetosomes biosynthesized by single cell (Fig. 4B(v)). The numbers of magnetosomes in a single cell of WT were distributed in a range of 5-15 and those in a single cell of TBInC and TBInCPNL were distributed in a range of 5-25 ( Fig. 4B(v)). The average number of magnetosomes biomineralized in a single cell of WT, TBInC, and TBInCPNL was 11 ± 5, 14 ± 6, and 16 ± 7, respectively ( Fig. 4B(v)). These results demonstrated that the transfer of these exogenous genes into MSR-1 exhibited little inhibition on the proliferation of cells but promoted the biomineralization of magnetosomes. It is implied a positive correlation between the magnetic responses and the number of magnetosomes from various strains.
Suitability of high cell density cultivation: Subsequently, TBInCPNL was incubated in fed-batch cultivation in a 7.5-L fermenter. Figure 4C showed a typical growth curve including the lag, exponential, stationary, and decline phases. The peak value of OD 565 was 19.6, appearing at 114 h. The curve of magnetic response from growing TBInCPNL could be divided into two parts: a decreasing curve and a parabolic curve. Typically, MSR-1 biosynthesizes magnetosomes under a low concentration of dissolved oxygen (dO 2 < 1%). When TBInCPNL cells were transferred into the fermenter, dO 2 of the culture medium was enhanced and the biosynthesis of magnetosomes in cells was temporarily inhibited, leading to the initial decrease of Cmag. After around 25 h culture, dO 2 was gradually driven down to a level suitable for the biomineralization of magnetosomes and thus, Cmag values started to increase. The Cmag value reached to the peak 1.11 and thereafter declined again with the increase of dO 2 , due to the high stirring rate in partial. After 120 h, TBInCPNL was harvested and the yields of cells and magnetosomes were 106.3 g and 6.8 g (ww), respectively. Hence, in spite of exogenous genes, with nucB and RRz/Rz1 in particular, it was not a problem to carry out a high-density culture of TBInCPNL at a large scale.

Cascade Cell Lysis Of Tbincpnl And Hydrolysis Of Nucleic Acids
The function of RRz/Rz1 and nucB had been demonstrated in the recombinant E. coli strains (PNL and ONL) above in the section of 3.1. We investigated whether these genes worked in the recombinant MSR-1 strain (TBInCPNL). TBInC, the recombinant strain harboring Nb but not RRz/Rz1 and nucB, was employed as a control. Cells of TBInC and TBInCPNL were harvested from a shake-ask culture (100 mL) and resuspended in 3 mL of phosphate buffered saline (PBS: 10 mM, pH 7.4), followed by the addition of 20 µL of CHCl 3 . After 2 h incubation, the majority of TBInCPNL cells were broken, whereas TBInC cells changed slightly (Fig. 5A). The lysis rate of TBInCPNL cells was over 90% within 0.5 h and approximately 99% within 1 h, as determined with a blood counting chamber (Fig. 5B). Another method also showed that RRz/Rz1 worked well in TBInCPNL. When the frozen TBInCPNL cells were transferred from liquid nitrogen to room temperature, they had almost been completely lysed within 10 min (Fig. 5C). These results indicated that RRz/Rz1 was functional in TBInCPNL, and cascade cell lysis would occur at the suitable condition (e.g., CHCl 3 or liquid nitrogen).
Extracts from the periplasmic space of TBInCPNL by osmotic shock were able to hydrolyze the plasmid pBBRPNL and genomic DNA of MSR-1 at 37°C (Fig. 5D and E), indicating the expression of the functional NucA. After the addition of lysozyme into the cell suspension of TBInC and TBInCPNL (details in Materials and Methods), the mixtures were incubated at room temperature for 2 h. In the absence of Ca 2+ , nucleic acids in homogenates of both TBInC and TBInCPNL could be detected within 2 h. While in the presence of Ca 2+ , nucleic acids in the homogenate of TBInCPNL were hardly detectable after incubation for 1 h, but detectable in the homogenate of TBInC within 2 h (Fig. 5F). These results illustrated that NucA was expressed in the periplasmic space of TBInCPNL and showed a strong non-speci c hydrolysis capability for nucleic acids in the presence of Ca 2+ at room temperature.
The activities of NucA and RRz/Rz1 were further evaluated in TBInCPNL cells cultivated in a 7.5-L fermenter. Here, WT MSR-1 was used as a control. Cells were harvested at the later stage of exponential phase of the growth curve and then treated with CHCl 3 . One hour after the addition of CHCl 3 , over 75% WT cells were intact, while approximately 90% TBInCPNL cells were disrupted ( Fig. 5G and H). In a gravity ow experiment [29], the homogenate of disrupted TBInCPNL cells had a lower viscosity than that of WT cells (Fig. 5I), indicating that the cascade cell lysis and hydrolysis of nucleic acids occurred in TBInCPNL from the large-scale culture.

Extraction And Isolation Of Nbs
In general, the cells should be harvested before a drastic drop of Cmag values to ensure a high yield of magnetosomes. Herein, when the Cmag values of TBInCPNL dropped to approximately 1.0 from the peak, cells were harvested even though they were still in the exponential phase of growth. According to the curves of magnetic response and cell growth (Fig. 4C), a 5-L cultivation of TBInCPNL was harvested at 72 h (OD 565 = 8.28, Cmag = 1.07). Cells were separated from the culture medium by centrifugation. The medium supernatant was then concentrated to approximately 25 mL, containing proteins at a concentration of 9.05 mg mL − 1 . The proteins in the supernatant showed no binding activity to TBBPA or its hapten T5 conjugated with horseradish peroxidase (HRP) (T5-HRP) by enzyme-linked immunosorbent assays (ELISAs), suggesting that Nbs were hardly secreted into medium. Afterwards, the cascadeampli ed lysis of cells suspended in PBS (with Ca 2+ ) was carried out using CHCl 3 . Magnetosomes were separated from cell broth under a magnetic eld and cleaned up by washing with PBS (pH 7.4). The resultant magnetosomes exhibited a strong binding activity to T5-HRP by a non-competitive ELISA (Fig.  6A), and to TBBPA by a competitive ELISA (Fig. 6B). The results indicated the attachment of Nbs to magnetosomes. The cell broth debris was removed via centrifugation. The binding activities of proteins in the supernatant and precipitant to antigens (TBBPA or T5-HRP) were evaluated. No obvious binding activities were observed. Then the ratios of Nbs in supernatant (soluble proteins) and precipitant (insoluble proteins) with whole cell were further analyzed at two different stages: i) In exponential phase (OD 565 = 8.28, Cmag = 1.07), only ~ 2% Nbs were detected in the supernatant and ~ 8% Nbs were detected in the precipitant (Fig. S3A); ii) In decline phase (OD 565 = 18.72, Cmag = 0.54), only ~ 1% Nbs were detected in the supernatant and ~ 14% Nbs were detected in the precipitant (Fig. S3B). These results supported that the majority of Nbs was attached to magnetosomes.
In the present study, the Nb was immobilized on the surface of magnetosomes, via a Pab Pol intein as a bridge of Nb and MamC, to form a fusion protein on magnetosome, MamC-Intein-Nb (Fig. 6C). Pab PolII intein could promote protein splicing in vitro at high temperature [27]. Therefore, the temperature controlled self-splicing of intein would be accompanied with the separation of Nbs from magnetosomes.
Under the optimized cleaving conditions: 0.1 g magnetosome complexes (MamC-Intein-Nb) were suspended in 300 µL of PBS (pH 6.0) containing 200 mM dithiothreitol (DTT) and incubated at 50°C for 30 min. Nbs were detected in PBS by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Fig. 6D) and showed binding activities to T5-HRP and TBBPA ( Fig. 6E and F). Magnetosome complexes showed a slight binding capability to TBBPA after the splicing treatment (data not shown), indicating that Nbs were split from the complexes. The yield of Nbs released from magnetosome complexes was approximately 0.60 µg mg − 1 .

Discussion
Advances of molecular biology have improved cloning and culture methods, which has diversi ed applications of recombinant proteins, ranging from enzymes used in laundry detergents to antibodies employed in cancer therapy [1]. Extensive studies have been carried out to nd suitable production systems for high-level expression of recombinant proteins [30,31]. However, downstream processing often contributes more to production cost than upstream and fermentation process [8]. Most efforts were previously focused on extracellular expression or production to reduce the downstream processing cost, but in general, the yields of extracellular protein were much lower than those produced intracellularly [32].
The downstream processing of intracellular products needs multiple steps, including host cell harvest, cell disruption, and target protein isolation and puri cation. Though there are several reports on controllable cell lysis [33], no complete genetic engineering platform for downstream processing was available.
The methods for the productions of recombinant proteins are quite diverse. The recombinant proteins are often produced in high yields with maturity technology, while the production capacities of newly constructed recombinant proteins are usually low [32]. Developing a system for the isolation of low-level product, in a sense, is a more di cult task, because it must remove much larger proportion of impurities. In this work, we developed a complete set of genetic engineering downstream processing platform for low expression level proteins, from cell harvest to target protein isolation. Since the magnetotactic bacteria strain MSR-1 was employed as the host strain, the cells with target proteins could be harvested by magnetic separation, instead of traditional centrifugations. Magnetic separation makes the collection of host cells much easier for continuous pipeline transportation, which is a progressive and economically advantageous mode of industrial transportation [34]. Cell disruption is also a key downstream processing step. This is a costly step for industrial production, in which high pressure homogenization was often employed. Autolysis or controllable lysis of host cell is the most convenient way to release intracellular products [35]. However, after cell lysis, the release of long chain nucleic acids led to a very high viscosity, which makes di cult for the next downstream processing steps [36]. We here developed a surface expressed nuclease to degrade nucleic acids and decreased the viscosity, which would simplify the rest of downstream processing steps (Fig. 1). The chromatographic procedure has been used in the isolation and puri cation of most protein products, as it has the advantages of having mild elution conditions and strong and speci c binding. However, in the industrial scale, these methods generally need large equipment, and complicated and laborious procedures for chromatographic matrix cleaning and regeneration [37]. The anti-TBBPA Nb was expressed on the surface of magnetosomes in this work. Nbs could be retrieved by magnetic separation from cell debris and undesired molecules in the host crude extract. The Nb was then abscised via intein on the engineered magnetosomes, and thus separated the protein product and magnetosome (Fig. 1). Although the product Nbs need further puri cation in this work, most impurities were removed (Fig. 6D), which greatly simpli es subsequent puri cation procedures. This system could also be employed to express insoluble proteins or membrane proteins. In traditional systems, these proteins usually accumulated in cytoplasm as inclusion bodies. While, in this system, the recombinant protein was expressed on the magnetosome surface instead of forming inclusion bodies. Besides, the system also facilitates bio-nano engineering and the bioproduction of functional nanomaterials, such as metal, alloy, or metallic oxide. These rigid materials, in traditional procedures, often cause serious wear of high-pressure homogenizer during host cell disruption [38]. In contrast, the damage from the homogenization could be completely avoided in the system described here.
The host strain, MSR-1, also has the potential for high cell density cultivation. Traditionally, cultures with high microbial cell density have a high metabolic oxygen demand. In these cultures, the oxygen transfer rate of the bioreactor determines the maximum biomass concentration. Unfortunately, the solubility of oxygen decreases with increasing cell densities due to higher viscosity of the cultivation. Therefore, high speed agitation and a large amount of pure oxygen are required at the latter part of cultivation, which is an energy-intensive and costly process [39]. MSR-1 is a microaerophilic strain. Compared with prevalent host strains, such as the strains of E. coli, Bacillus subtilis, or Saccharomyces cerevisiae, MSR-1 demands much less oxygen during cultivation when reaching the same cell density [40,41], which made the procedure energy-e cient and economical. It has been reported that the stirring speed was over 750 rpm, when the culture density of E. coli reached to an OD 600 of 20 [39]. While, at similar cell density, only around 400 rpm was required for MSR-1 cultivation. Fortunately, MSR-1 is also a typical gram-negative bacterium like E. coli. Most plasmid vectors, genetic elements, and gene expression systems of E. coli also work in MSR-1. We thus believe a good possibility to construct a microaerobic gene expression system of MSR-1 in the near future, which will make a great progress of recombinant protein production [9]. Nonetheless, much effort is required to develop a gene expressing system in a microaerophilic bacterium, especially to set up easier cultivation methods, well-developed genetic manipulation system. Besides, magnetosome numbers in each cell and the percentage of MamC in the membrane proteins need to be substantially increased to achieve high protein productions.

Conclusion
We constructed a new complete genetic engineering downstream processing platform for low expression level of proteins in the magnetotactic bacterium MSR-1. This smart magnetosome-based platform was e ciently controlled by three functional components including RRz/Rz1, NucA, and magnetosome complex (MamC-Intein-Nbs), which were amenable to the lysis of host cell, the reducing viscosity of homogenate, and the separation and puri cation of Nbs, respectively. Using this platform, most impurities were removed from the medium and the subsequent puri cation procedure of Nbs was greatly simpli ed. Such a platform is an innovative protein puri cation platform to advance and overcome current bioprocessing challenges.

Bacterial strains and culture conditions
The bacterial strains and plasmids used in this study are listed in Table S1. Escherichia coli strains were cultured in Luria broth (LB) or NDase Agar plates for detecting NucA at 37°C. MSR-1 cells were cultured in SLM with 20 µM ferric citrate or sodium glutamate medium (SGM) at 30°C (substitute 4 g SGM for NH 4 Cl and yeast extract of SLM) [42]. MSR-1 cells were cultured in 100-mL serum bottles lled with 50 mL of medium or 250-mL serum bottles lled with 100 mL of medium. In addition, MSR-1 cells were cultured in a 7.5-L fermenter with Fed-batch culture described as previous [43]. Antibiotics used were as follows: for E. coli, ampicillin at 100 µg mL − 1 and kanamycin (Km) at 50 µg mL − 1 ; for MSR-1, Km at 5 µg mL − 1 and nalidixic acid at 5 µg mL − 1 . The growth (OD 565 ) and Cmag of MSR-1 were measured as described previously [44].

Plasmid And Strain Construction
All plasmids used in this study are listed in Table S1. All cloning was performed in E. coli using restriction enzyme ligation (TaKaRa, Japan). For PCR ampli cation, 2 × Phanta Max Master Mix (Vazyme, China) were used with the primers listed in Table S2. nucB was ampli ed from S. aureus NCTC 8325 genomic DNA that the signal peptide from E. coli OmpA or R. rubrum PhaZ1 was expressed at N-terminal. RRz/Rz1 expressed using the native promoter P R' was ampli ed from λ phage genomic DNA. HA and Flag tags were fusion expressed at C-terminal of NucB and RRz/Rz1, respectively. Then nucB and RRz/Rz1 were assembled into a cassette by fusing PCR ampli cation, named as ONL (signal peptide from E. coli OmpA) or PNL (signal peptide from R. rubrum PhaZ1). ONL and PNL were inserted into the plasmid pBBR1MCS-2 digested with BamH I and Xba I using T4 DNA ligase to generate recombination plasmid pBBRONL and pBBRPNL, respectively. Finally, pBBRONL and pBBRPNL were transferred into E. coli S17-1, named as ONL and PNL, respectively. mamC and its upstream and downstream homology regions were ampli ed from MSR-1 genomic DNA. Anti-TBBPA Nb with 6×his tag at its C-terminal was ampli ed from a plasmid (pecan 45) containing the Nb-alkaline phosphatase (AP) fragment [45]. The fusion gene consisted of Nb and mamC was cloned by fusing PCR ampli cation with primers TB-F (BamH I) and C-R (Xba I), named as TBC. uTBCd, a cassette of mamC's upstream region, TBC, and mamC's downstream region, was assembled with fusing PCR ampli cation. The gene sequence coding Pab Pol intein, synthesized with Myc tag expressed at Cterminal by Tsingke Biotechnology Co., Ltd., was inserted into uTBCd to generate uTBInCd by PCR ampli cation. Finally, uTBInCd was subcloned into pK18mobSacB with EcoR I and Xba I to create recombination suicide plasmid pKTBInC, which was transferred into E. coli S17-1 and then WT MSR-1 by biparental conjugation to obtain TBInC. The smart engineering strain was referred to as TBInCPNL generated from TBInC and PNL by biparental conjugation. The lysate was transferred into a new Eppendorf tube at different time (0, 0.5, 1, and 2 h) and the collection volume was at 200 µL each time. Phenol-chloroform-isopentanol was added with an equal volume (200 µL), vibrated drastically for 10 s, and centrifuged at 12,000g for 5 min. The supernatant was transferred into another new Eppendorf tube again. Sodium acetate (3 M, pH 5.2) that was 1/10 volume supernatant and ice-cold anhydrous alcohol that was 2-2.5 volume supernatant were successively added, mixed, maintained at least 5 min on ice, and centrifugated at 12,000g for about 10 min. The precipitates were washed twice with 70% alcohol (1 mL) after supernatant moved. Nucleic acids were dissolved in ddH 2 O and detected with agarose gel electrophoresis.

Target Protein Expression In Different Compartments (Periplasmic
Viscosity: The viscosity of homogenate from cells incubated with CHCl 3 or lysis buffer (200 mM NaOH, 1% SDS (w/v)) were tested by gravity ow experiments [29].

Transmission Electron Microscopy
Cells were collected at ~ 1.0 of OD 565 , placed on copper grid, washed thrice with ddH 2 O, and observed under a JEM-1230 TEM (JEOL, Tokyo, Japan). Numbers and diameters of magnetosomes were analyzed statistically with ImageJ (National Institutes of Health; Bethesda, MD, USA), a Java-based imageprocessing program.

Conjugation Of Tbbpa Derivative To Protein
The hapten (T5) of TBBPA (Fig. S4) was available from our previous study [46] and coupled to HRP or bovine serum albumin (BSA) according to the method described previously [47]. The concentrations of T5-HRP or T5-BSA were determined with BCA protein assay.

Puri cation And Quantitation Of Anti-tbbpa Nbs
Strains were cultured in a 7.5-L fermenter according to the fed-batch culture described previously [43].
Magnetosome-Nb complexes were harvested from the cell homogenates (whole cell) under a magnetic eld. The rest of homogenates was collected (1 mL) and separated by centrifugation at 4°C with 12,000g to generated supernatant (cytoplasm) and precipitate (cell debris). The precipitate was resuspended in 1 mL PBS (pH 7.4). Proteins from whole cell, supernatant and precipitate were separated on SDS-PAGE, detected with western blotting based on His tag at C-terminal of anti-TBBPA Nbs, and analyzed with ImageJ. The ratios of anti-TBBPA Nbs in different components with whole cell proteins were calculated.

Availability of data and materials
Data sharing not applicable to this article as no data-sets were generated. Data analysis in the current study was performed using publicly available datasets.
Ethics approval and consent to participate Not Applicable.

Consent for publication
Not Applicable.

Competing interests
The authors declare no competing interests.
47. Schneider Peter, Hammock BD. In uence of the ELISA format and the hapten-enzyme conjugate on the sensitivity of an immunoassay for s-triazine herbicides using monoclonal antibodies. J Agric Food Chem. 1992; 40:525-530. Figure 1 A GEMP for downstream extraction and puri cation of Nbs. "OM": out membrane, "PG": peptidoglycan, "IM": in membrane.