Genome Editing Method for the Anaerobic Magnetotactic Bacterium Desulfovibrio magneticus RS-1

Magnetotactic bacteria (MTB) are a group of organisms that form intracellular nanometer-scale magnetic crystals though a complex process involving lipid and protein scaffolds. These magnetic crystals and their lipid membranes, termed magnetosomes, are model systems for studying bacterial cell biology and biomineralization and are potential platforms for biotechnological applications. Due to a lack of genetic tools and unculturable representatives, the mechanisms of magnetosome formation in phylogenetically deeply branching MTB remain unknown. These MTB contain elongated bullet-/tooth-shaped magnetite and greigite crystals that likely form in a manner distinct from that of the cubooctahedral-shaped magnetite crystals of the genetically tractable MTB within the Alphaproteobacteria. Here, we present a method for genome editing in Desulfovibrio magneticus RS-1, a cultured representative of the deeply branching MTB of the class Deltaproteobacteria. This marks a crucial step in developing D. magneticus as a model for studying diverse mechanisms of magnetic particle formation by MTB.

M agnetotactic bacteria (MTB) are a group of diverse microorganisms that align along magnetic fields via their intracellular chains of magnetic crystals (1,2). Each magnetic crystal consists of either magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ) and is synthesized within a complex organelle called a magnetosome (3). The first cultured MTB were microaerophilic members of the Alphaproteobacteria, which form cubooctahedralshaped magnetite crystals and have served as model organisms for understanding magnetosome formation (4)(5)(6)(7). Early studies on Magnetospirillum spp. revealed a lipid-bilayer membrane, with a unique suite of proteins, surrounding each magnetite crystal (8)(9)(10). The development of genetic tools in Magnetospirillum magneticum AMB-1 and Magnetospirillum gryphiswaldense MSR-1 revealed a conserved magnetosome gene island (MAI) that contains the factors necessary and sufficient for the formation of the magnetosome membrane, magnetite biomineralization within the lumen of the magnetosome, and alignment of the magnetosomes in a chain along the length of the cell (3,11). These molecular advances, along with the magnetic properties of magnetosomes, have made MTB ideal models for the study of compartmentalization and biomineralization in bacteria as well as a target for the development of biomedical and industrial applications. Improvements in isolation techniques and sequencing have revealed that MTB are ubiquitous in many aquatic environments. On the basis of phylogeny and magnetosome morphology, MTB can be categorized into two subgroups. The first subgroup includes members of the Alphaproteobacteria and Gammaproteobacteria, such as Magnetospirillum spp., that synthesize cubooctahedral, elongated octahedral, or elongated prisms of magnetite (12). The second subgroup comprises MTB from more deepbranching lineages, including members of the Deltaproteobacteria class and the Nitrospirae and Omnitrophica phyla, which synthesize elongated bullet-/tooth-shaped magnetite and/or greigite crystals (13,14). While all MTB sequenced to date have their putative magnetosome genes arranged in distinct regions of their genomes (3,(15)(16)(17), many of the genes essential for magnetosome biogenesis in Magnetospirillum spp. are missing from the genomes of deep-branching MTB (14). Likewise, a conserved set of mad (magnetosome-associated Deltaproteobacteria) genes are only found in deep-branching MTB (14,(18)(19)(20). This suggests a genetic diversity underpinning the control of magnetosome morphology and physiology in nonmodel MTB that is distinct from that of the well-characterized Magnetospirillum spp.
Desulfovibrio magneticus RS-1, one of the few cultured MTB outside the Alphaproteobacteria, is an anaerobic sulfate-reducing member of the Deltaproteobacteria that forms irregular bullet-shaped crystals of magnetite (21,22). As with the Magnetospirillum spp., the magnetosome genes of D. magneticus are located within a MAI and include homologs to some mam genes as well as mad genes (14,18,23). Recently, we used a forward genetic screen combining random chemical and UV mutagenesis with whole-genome resequencing to identify mutations that resulted in nonmagnetic phenotypes. These included many mutants that had the entire MAI deleted (ΔMAI) as well as mutants with point mutations, frameshift mutations, and transposon insertions in 10 mam and mad genes of the D. magneticus MAI that resulted in nonmagnetic phenotypes (20). However, this screen relied on a strict selection scheme for nonmagnetic mutants. As such, we likely missed magnetosome genes that are important for regulating the shape, size, and arrangement of magnetosomes. To elucidate the degree of conservation between mam genes and determine the function of the proteins encoded by mad genes in D. magneticus, a reverse genetic method for targeted mutagenesis is necessary.
D. magneticus and other Desulfovibrio spp. have gained much attention for their importance in the global cycling of numerous elements, in biocorrosion, and in the bioremediation of toxic metal ions (24,25). The development of genetic tools, such as expression vectors, transposons, and targeted genome-editing systems, has enabled a more detailed examination of the important activities of a few Desulfovibrio spp. (26,27). Targeted mutagenesis using a one-step double recombination method was first achieved in Desulfovibrio fructosivorans and, more recently, in Desulfovibrio gigas and Desulfovibrio desulfuricans ND132 (28)(29)(30). With this method, plasmids that are electro-porated into the cell are thought to be rapidly linearized by endogenous restriction modification systems (30)(31)(32). The linearized plasmid DNA, carrying a selectable marker flanked by upstream and downstream regions of homology to a target gene, can then undergo double recombination into the chromosome in one step (Fig. 1A). This efficient one-step method, which is dependent on electroporation of the plasmid (28)(29)(30), is unlikely to be applicable for D. magneticus, because plasmid uptake has only been demonstrated using conjugal transfer (20). The second targeted mutagenesis method, used in Desulfovibrio vulgaris Hildenborough, is a two-step double recombination that makes use of a nonreplicative, or suicide, vector (31,32). In the first step of this method, a suicide vector, with sequences upstream and downstream of the target gene, integrates into the genome upon the first homologous recombination event (Fig.  1B). Next, a second recombination event occurs whereby the vector is excised from the genome, and cells with the desired genotype are selected with an antibiotic marker and/or a counterselection marker (31,32) (Fig. 1B). For many bacteria, including D. magneticus, plasmid uptake and integration occur at frequencies that are too low for genetic manipulation via suicide vectors (20). Plasmids (black lines) were designed to replace a target gene (X, aqua arrows) in the chromosome (blue lines) with a streptomycin resistance cassette (strAB, purple arrows). Regions upstream (*) and downstream (**) of the target gene (blue boxes) on the chromosome undergo recombination (red lines) with homologous regions that are cloned into the deletion plasmid. Key steps, such as recombination events (red crosses), are indicated in the boxes, and the selection steps are labeled in red. (A) Double recombination can occur in one step after plasmids are linearized (dashed lines) by endogenous restriction enzymes. Mutants are selected using the marker (e.g., strAB) that was exchanged with the target gene. (B) Two-step double recombination is possible when suicide vectors integrate into the chromosome in the first homologous recombination event and then recombine out after the second homologous recombination event. The first step and second step are selected for with antibiotic resistance markers (e.g., npt) and counterselectable markers (e.g., sacB), respectively. (C) A replicative deletion plasmid designed to target genes for deletion may undergo double recombination in one or two steps as shown in panels A and B, respectively. After passaging the cells without antibiotic, the mutants are selected with an antibiotic resistance cassette (e.g., strAB) and a counterselectable marker (e.g., sacB). mob, mobilization genes (mobA=, mobB, mobC); npt, kanamycin-resistance gene; ori Dm , origin of replication for D. magneticus; ori Ec , origin of replication for E. coli.
Here, we describe the method we developed for targeted gene deletion using a replicative plasmid, thereby bypassing the need for suicide vector integration (Fig. 1C). We generated a mutant resistant to 5-fluorouracil by making a markerless deletion of the upp gene, which encodes an enzyme in the pyrimidine salvage pathway that is nonessential under standard laboratory conditions. Additionally, we deleted kupM, a gene encoding a potassium transporter that acts as a magnetosome formation factor (20), via marker exchange with a streptomycin resistance cassette. The deletion of both upp and kupM conferred the expected phenotypes, which were subsequently complemented in trans. Overall, our results show that targeted mutagenesis using a replicative plasmid is possible in D. magneticus. It may also be suitable for other bacteria for which replicative plasmid uptake is possible but at a rate too low for suicide vector integration.

RESULTS
Design of a replicative deletion plasmid using sacB counterselection. Targeted genetic manipulation in most bacteria requires a method to efficiently deliver foreign DNA destined for integration into the chromosome. One commonly used method involves suicide vector uptake and integration prior to the first selection step (Fig. 1B). In D. magneticus, plasmid transfer has only been achieved via conjugation at low efficiencies, making the uptake and subsequent integration of suicide vectors into its chromosome an unlikely event (20). As such, we attempted to bypass the use of suicide vectors and use a stable replicative plasmid designed to delete specific genes via homologous recombination (Fig. 1C). Two features of this method enable the isolation of desired mutants: (i) a selectable marker is used to identify double recombination events at the targeted site and (ii) a counterselectable marker distinguishes the desired mutant cells, which have lost all remaining copies of the plasmid.
sacB is a common counterselection marker that is effective in many bacteria. The sacB gene from Bacillus subtilis encodes levansucrase, which converts sucrose to levans that are lethal to many Gram-negative bacteria, including D. vulgaris Hildenborough (31,33,34). To test its functionality in D. magneticus, we inserted sacB under the expression of the mamA promoter of D. magneticus (described in reference 20) in a plasmid that replicates in both Escherichia coli and D. magneticus ( Fig. 2A). This plasmid (pAK914) and a control plasmid were then conjugated into D. magneticus. We found no growth inhibition for D. magneticus cells with the control plasmid in the presence of sucrose and kanamycin. In contrast, cells expressing sacB were unable to grow with kanamycin and sucrose concentrations of 1% (wt/vol) or higher (data not shown). To test if the plasmids could be cured, D. magneticus with pAK914 was passaged two times in liquid medium containing no antibiotic and plated on 1% sucrose. Individual sucrose-resistant (Suc r ) colonies were inoculated and screened for kanamycin sensitivity (Kan s ). All isolated colonies (n ϭ 16) were Kan s , suggesting that the cells had lost the plasmid. These experiments demonstrate that sacB is a suitable counterselection marker in D. magneticus.
Construction of a ⌬upp strain by markerless deletion. To test our replicative deletion method, we chose to target the upp gene, the mutation of which has a selectable phenotype. The upp gene encodes uracil phosphoribosyltransferase (UPR-Tase), a key enzyme in the pyrimidine salvage pathway that catalyzes the reaction of uracil with 5-phosphoribosyl-␣-1-pyrophosphate (PRPP) to UMP and PP i (35) (Fig. 3A). When given the pyrimidine analog 5-fluorouracil (5-FU), UPRTase catalyzes the production of 5-fluoroxyuridine monophosphate (5-FUMP). 5-FUMP is further metabolized and incorporated into DNA, RNA, and sugar nucleotides resulting in eventual cell death (Fig.  3A) (36,37). Previous studies have shown that Δupp mutants of D. vulgaris Hildenborough are resistant to 5-FU, while wild-type (WT) cells are effectively killed by the pyrimidine analog (32,38). The D. magneticus genome has a homolog (DMR_08390) to the D. vulgaris Hildenborough upp gene that is likely functional, as detected by the sensitivity of D. magneticus to 5-FU ( Fig. 3B and 4A). To show that the upp gene product confers 5-FU sensitivity and to validate our replicative deletion system, we chose to target the D. magneticus upp gene for markerless deletion.
To construct a upp deletion vector, a markerless cassette containing the regions upstream and downstream of the upp gene were inserted into plasmid pAK914 (Fig. 2B). The resulting plasmid (pAK1126) was transferred to WT D. magneticus by conjugation and single kanamycin-resistant (Kan r ) colonies were isolated and passaged in growth medium containing no antibiotic. Since D. magneticus has interesting features independent of its magnetosomes, the same deletion procedure was also carried out in a nonmagnetic strain (ΔMAI) isolated in our previous genetic studies (20). After the third passage, upp mutants that had lost the vector backbone were selected for with 5-FU and sucrose. Compared with those obtained using a control plasmid (pAK914), Ͼ20fold more 5-FU-resistant (5-FU r ) mutants were generated using pAK1126 at a frequency of approximately 10 Ϫ6 . PCR of the region flanking the upp gene confirmed that the 5-FU r colonies harboring pAK1126 resulted from a markerless deletion of upp (Δupp), while 5-FU r colonies harboring pAK914 were likely the result of point mutations (Fig. 3B  and D). Similar to the results obtained for D. vulgaris Hildenborough (32), the Δupp mutant of D. magneticus grew in the presence of 5-FU ( Fig. 4B and Table 1). Complementation of the upp gene in trans restored UPRTase function, and the cells no longer grew with 5-FU ( Fig. 2C and 4C and Table 1). These experiments demonstrate that a replicative plasmid can be used to directly edit the D. magneticus genome.
Construction of a ⌬kup strain by marker exchange mutagenesis. Because many genetic mutations do not confer a selectable phenotype, we sought to develop our replicative deletion plasmid for marker exchange mutagenesis. To test this system, we chose to replace a gene with a known phenotype, kupM (DMR_40800), with a streptomycin-resistance gene cassette (strAB). kupM is located in the D. magneticus MAI and encodes a functional potassium transporter (20). Mutant alleles in kupM, including missense, nonsense, and frameshift mutations, were previously identified in our screen for nonmagnetic mutants (20). These kupM mutations resulted in cells that rarely contained electron-dense particles and were unable to turn in a magnetic field, as measured by the coefficient of magnetism (C mag ) (20).
To mutate kupM, we inserted a marker exchange cassette, with regions upstream and downstream of kupM flanking strAB, into pAK914 (Fig. 2D) to create the deletion plasmid pAK941. Following conjugation, single colonies of D. magneticus harboring pAK941 were isolated by kanamycin selection. After three passages in growth medium without selection, potential mutants were isolated at a frequency of approximately 10 Ϫ6 on plates containing streptomycin and sucrose. Single colonies that were streptomycin resistant (Str r ) and Suc r were inoculated in liquid medium and screened for Kan s . Of the isolates screened (n ϭ 48), 20% were Kan s and 4% had the correct genotype (ΔkupM::strAB) as confirmed by PCR and sequencing (Fig. 3C and E). Similar to the phenotypes previously observed in kupM mutants (20), ΔkupM::strAB cells were severely defective in magnetosome synthesis and turning in a magnetic field (Fig. 5). Although a slight C mag was measured, few cells contained electron-dense particles or magnetosomes. Importantly, the WT phenotype was rescued by expressing kupM from a plasmid in the ΔkupM::strAB mutant (Fig. 5). These results confirm that the replicative deletion plasmid method described here can be used successfully for marker exchange mutagenesis.

DISCUSSION
In this study, we expand the genetic toolbox for D. magneticus to include a replicative plasmid method for targeted mutagenesis (Fig. 1C). We show the utility of this method for markerless deletion of genes with a selectable phenotype and for marker exchange mutagenesis. Some of the earliest examples of targeted mutagenesis in Gram-negative bacteria used replicative plasmids, similar to the method described here (34,39). These studies, which predated the application of suicide vectors, relied on plasmid instability by introducing a second plasmid of the same incompatibility group or by limiting nutrients in the growth medium (34,39).
Because the D. magneticus genetic toolbox has a limited number of plasmids, antibiotic markers, and narrow growth constraints, we used a replicative plasmid and established sacB as a counterselection marker to generate and isolate mutants. While sacB counterselection was ultimately successful, a large number of false positives were also isolated at the sucrose selection step. Mutations in sacB have been found to occur at a high frequency in many bacteria (31,(40)(41)(42)(43). Indeed, we found that deletions and mutations in P mamA -sacB are abundant in the false-positive Suc r Str r isolates (data not shown). Alternative counterselection markers, including upp, have been shown to select for fewer false positives (32,(43)(44)(45). Since D. magneticus is sensitive to 5-FU only when the upp gene is present (Fig. 4), the upp mutants generated in this study may be used as the parent strains for future targeted mutagenesis with upp, rather than sacB, serving as a counterselectable marker. Additionally, the combined use of upp and sacB for counterselection might reduce the false-positive background that results from the accumulation of mutations in these markers.
The replicative deletion plasmid described here was designed to replace a target gene with an antibiotic resistance marker. As such, the construction of strains with multiple directed mutations will be complicated by the need for additional antibioticresistance markers, which are limited in D. magneticus. These limitations may be overcome by removing the chromosomal antibiotic marker in subsequent steps (34,46,47). Ultimately, improvements in conjugation efficiency or methods for electroporation with high transformation efficiency are desired. Similar to the ongoing development of genetics in D. vulgaris Hildenborough, the establishment of a suicide vector delivery system in D. magneticus will enable more high-throughput targeted mutagenesis and even the construction of markerless deletion mutants (26,32).
Overall, we demonstrated the utility of a replicative deletion plasmid to generate targeted mutants of D. magneticus. This method marks a crucial step in developing D. magneticus as a model for the study of anaerobic sulfate reduction and diverse mechanisms of magnetic particle formation by MTB. Both MTB and sulfate-reducing bacteria have been singled out for their role in the global cycling of numerous elements and for potential applications, such as bioremediation (24,25,48,49). D. magneticus, in particular, may be useful in the bioremediation of heavy metals and in the global cycling of iron, since it can form both magnetosomes and other iron-containing organelles (50,51). Through genetic manipulation of D. magneticus, pathways of elemental cycling and heavy-metal turnover may now be explored. Additionally, genetic manipulation of D. magneticus will further our understanding of magnetosome formation and provide answers to many longstanding questions for the deeply branching MTB. Which proteins regulate and control magnetosome formation? To what extent are lipid membranes involved in forming these crystals? How is the elongated and irregular crystal shape achieved? Finally, in addition to D. magneticus, the method described here may extend to other bacteria that are not amenable to targeted mutagenesis with suicide vectors but are able to accommodate replicative plasmids.

MATERIALS AND METHODS
Strains, media, and growth conditions. The bacterial strains used in this study are listed in Table  2. All E. coli strains were cultured aerobically with continuous shaking at 250 rpm at 37°C in lysogeny broth (LB). D. magneticus strains were grown anaerobically at 30°C in sealed Balch tubes with a N 2 headspace containing RS-1 growth medium (RGM) that was degassed with N 2 , unless otherwise stated (51). Sodium pyruvate (10 mM) was used as an electron donor with fumaric acid disodium (10 mM) as the terminal electron acceptor. RGM was buffered with HEPES, and the pH was adjusted to 6.7 with NaOH (20). Before inoculating with cells, RGM was supplemented with 0.8% (vol/vol) Wolfe's vitamins, 100 M ferric malate, and 285 M cysteine-HCl (51). Solid agar plates were prepared by adding 1.5% agar (wt/vol) to LB and 1% agar (wt/vol) to RGM. Vitamins (0.8% [vol/vol]), ferric malate (20 M), and cysteine (285 M), as well as antibiotics and selective agents, were added to the molten RGM agar as needed. For D. magneticus, all plating steps were carried out aerobically, and the bacteria were transferred to an anaerobic jar and incubated at 30°C for 10 to 14 days, as described previously (20). The antibiotics and selective agents used are as follows: kanamycin (50 g/ml for E. coli strains, 125 g/ml for D. magneticus strains), streptomycin (50 g/ml for E. coli and D. magneticus strains), diaminopimelic acid (300 M for E. coli WM3064), 5-FU (2.5 g/ml for D. magneticus strains), and sucrose (1% for D. magneticus strains).
Plasmids and cloning. All plasmids used in this work are listed in Table 2. All cloning was performed in E. coli DH5␣ pir using the Gibson method (52) or restriction enzyme ligation. For PCR amplification, KOD (EMD Millipore, Germany) and GoTaq (Promega, USA) DNA polymerases were used with the primers listed in Table 3. All upstream and downstream homology regions were amplified from D. magneticus genomic DNA. strAB and P npt were amplified from pBMS6 and pLR6, respectively, and subcloned into pBMC7 to make pAK920, which served as the template for amplifying P npt -strAB for the deletion vectors. sacB was amplified from pAK0 and inserted into pLR6 digested with SalI and XbaI to create pAK914. To construct a plasmid for the targeted deletion of upp (DMR_08390), 991 bp upstream and 1,012 bp downstream of upp were amplified and inserted into pAK914 digested with XbaI and SacI using a 3-piece Gibson assembly. To create the upp complementation plasmid, pAK914 was digested with BamHI and SacI, and the upp gene, with its promoter, was PCR amplified from D. magneticus genomic DNA. To construct pAK941 for marker exchange mutagenesis of kupM, a cassette of a 1,064-bp upstream region and 1,057-bp downstream region flanking P npt -strAB was assembled using Gibson cloning. The cassette was amplified and inserted into pAK914 digested with XbaI using a two-piece Gibson assembly. upp and kup mutant generation and complementation. Replicative deletion plasmids were transformed into E. coli WM3064 by heat shock and transferred to D. magneticus by conjugation, as described previously (20). Single colonies of Kan r D. magneticus were isolated and inoculated in RGM containing no antibiotic. Cultures were passaged and, after the third passage, approximately 2 ϫ 10 8 cells were spread on 1% agar RGM plates containing either 50 g/ml streptomycin and 1% sucrose or 2.5 g/ml 5-FU and 1% sucrose. 5-FU r Suc r and Str r Suc r colonies harboring plasmids pAK1126 and pAK941, respectively, were recovered at a frequency of approximately 10 Ϫ6 . Single colonies were screened for Kan s and by PCR using the primers listed in Table 3. Successful upp and kup mutants were confirmed by Sanger sequencing. The expression plasmids for the complementation of Δkup::strAB and Δupp, as well as empty vectors for controls, were transferred to D. magneticus strains as described above. Transconjugants were inoculated in RGM containing kanamycin to maintain the plasmids.
Mutant phenotype and complementation analyses. The growth and coefficient of magnetism (C mag ) of D. magneticus strains were measured in a Spec20 spectrophotometer at an optical density of 650 nm (OD 650 ), as described previously (10,51). For upp mutant and complementation analyses, RGM was supplemented with 5-FU (1.25 g/ml in 0.01% dimethyl sulfoxide [DMSO]) or DMSO (0.01%), and the growth was measured for WT and Δupp strains with an empty vector (pAK914) and for the Δupp strain with the complementation plasmid pAK1127. For kup mutant and complementation analyses, the C mag was measured by placing a large bar magnet parallel or perpendicular to the sample to measure the maximum or minimum absorbance, respectively, as the D. magneticus strains rotate 90°with the magnetic field. The ratio of maximum to minimum absorbances was calculated as the C mag (10). Whole-cell transmission electron microscopy (TEM) was performed as previously described (51). The C mag calculations and TEM were performed for WT D. magneticus with an empty vector (pBMK7) and the Δkup::strAB strain with an empty vector (pBMK7) or complementation plasmid (pLR41). For all growth measurements, C mag measurements, and TEM, the cells harboring the plasmids were maintained with 125 g/ml kanamycin.