A Magnetosome-specific GTPase from the Magnetic BacteriumMagnetospirillum magneticum AMB-1*

Magnetic bacteria produce intracellular vesicles that envelope single domain magnetite crystals. Although many proteins are present in this intracellular vesicle membrane, five are specific to this membrane. A 16-kDa protein, designated Mms16, is the most abundant of the magnetosome-specific proteins, and to establish its function we cloned and sequenced its gene fromMagnetospirillum magneticum AMB-1. This was achieved by determination of the N-terminal amino acid sequence of the protein following two dimensional polyacrylamide gel electrophoresis, and sequencing of the gene was performed by gene walking using anchored polymerase chain reaction. Mms16 contains a putative ATP/GTP binding motif (P-loop). Recombinant Mms16 with a hemagglutinin tag, was expressed in Escherichia coli and purified. Recombinant Mms16 protein could bind GTP and showed GTPase activity. GTP was the preferred substrate for Mms16-catalyzed nucleotide triphosphate hydrolysis. These results suggest that a novel protein specifically localized on the magnetic particle membrane, Mms16, is a GTPase. Mms16 protein showed similar characteristics to small GTPases involved in the formation of intracellular vesicles. Furthermore, addition of the GTPase inhibitor AlF4 − also inhibited magnetic particle synthesis, suggesting that GTPase is required for magnetic particles synthesis.

Magnetic bacteria synthesize intracellular magnetite particles that are aligned in chains of around 10 particles per cell. They are covered with a thin lipid membrane and have an average diameter of 50 -100 nm. The morphology of these bacterial magnetic particles (BMPs) 1 is species-dependent, and it is therefore reasonable to hypothesize that species-specific bi-ological factors located in the BMP membrane mediate magnetite crystallization (1). However, the mechanism of BMP synthesis, including vesicle formation, still remains unclear. A complete understanding of biomineralization and of membrane vesicle formation at the molecular level will also have important implications for studying biomineralization in general and vesicle formation in prokaryotes in particular.
Several processes are involved in BMP synthesis, and one of the most important is vesicle formation. Numerous studies have been devoted to the investigation of eukaryotic intracellular vesicle formation, and hence the molecular machinery is well understood (for reviews, see Refs. 2 and 3). In contrast, there are few molecular studies of vesicle formation or of the events leading to invagination of the cytoplasmic membrane in prokaryotes. We have hypothesized that the BMP membrane is derived from the cytoplasmic membrane and formed through the invagination of the cytoplasmic membrane by a process similar that which occurs in eukaryotes (4).
A second process in BMP synthesis is magnetite crystallization, and this has been studied in more detail. It appears that ferric iron is reduced on the cell surface, taken into the cytoplasm, transferred into vesicles (magnetosomes), and finally oxidized to produce magnetite (5). Several proteins appear to be required for magnetite crystallization, and the first of these reported, magA, was cloned from a non-magnetic mutant obtained by transposon mutagenesis of Magnetospirillum magneticum AMB-1 (6). Internal localization analysis of the MagA protein using a MagA-luciferase fusion protein indicated that MagA is localized on the BMP membrane where it transports iron into the BMP vesicles (6,7). Thus magnetosomal membrane proteins play an important role in magnetite crystal formation.
Other proteins on the BMP membrane have been partially characterized. Gorby et al. observed two unique proteins in the BMP membrane fraction from M. magnetotacticum MS-1 (8). Okuda et al. identified three additional specific proteins in the BMP membrane of MS-1, and they determined the DNA/amino acid sequence of a 22-kDa protein (9). However, since gene transfer and its use in studying gene expression has not been successful in strain MS-1, the function of these proteins is still unknown. In strain AMB-1, three BMP membrane-specific proteins were identified and partially characterized. One of these, MpsA, which shows homology with acetyl-CoA carboxylase (transferase), containing acyl-CoA binding motif (4).
Recently, we have found two additional BMP-specific proteins, a 12-kDa and a 16-kDa protein (10) bringing the total number in strain AMB-1 to five (Table I). The 16-kDa protein was found to be the most abundantly expressed of the five BMP-specific proteins. In this work, we therefore cloned, sequenced, and partially characterized this 16-kDa protein. Although the MagA protein is important for BMP formation, it is not specific to the BMP membrane but is also present on the * This work was funded in part by Grant in Aid for Scientific Research on Priority Area (A) 10145102 and Specially Promoted Research 13002005 from the Ministry of Education, Culture, Science, Sports and Technology of Japan. It was also supported by the New Energy and Industrial Technology Development Organization's Proposal-based Advanced Industrial Technology Research and Development Program 1413. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM /EBI Data Bank with accession number(s) AB051013.
Strains and Growth Conditions-Escherichia coli DH5␣ was used for gene cloning and protein expression. E. coli was cultured in Luria broth at 37°C after adding the appropriate antibiotics. M. magneticum AMB-1 (ATCC 700264) (12) was grown anaerobically at 26°C in modified magnetic spirillum growth medium at pH 6.75 (13) as described previously (12).
Cloning and Sequencing of a Gene Encoding a 16-kDa Protein-The BMP membrane protein sample was prepared from ϳ10 12 cells according to the previously described (10). The 16-kDa protein from 200 g of protein sample was isolated by two dimensional polyacrylamide gel electrophoresis (PAGE) using automatic electrophoresis, TEP-2 unit (Shimadzu, Kyoto, Japan) according to the procedure described by Nokihara et al. (14). After staining the two dimensional gel with Coomassie Brilliant Blue R-250, the 16-kDa protein spot was excised from the gel and blotted onto a polyvinylidene difluoride (PVDF) membrane. The PVDF-blotted protein was subjected to the N-terminal amino acid sequence by Edman degradation using a gas-phase sequencer (PPSQ-10, Shimadzu, Kyoto, Japan).
Oligonucleotide primers were designed using the N-terminal amino acid sequence of the 16-kDa protein. The codon usage pattern of the magA gene isolated from strain AMB-1 was used (6). Polymerase chain reaction (PCR) was performed according to standard protocols (15) with 1-10 ng of genomic DNA as template. 132 bp of an amplified DNA fragment was cloned into the vector pGEM-T-easy (pGEM-T-easy Vector System, Promega, WI) and sequenced using automatic DNA sequencers, DSQ-2000L (Shimadzu) and ABI PRISM 377 (PerkinElmer Life Sciences). Gene walking for obtaining the whole gene encoding the 16-kDa protein was performed as follows. XbaI-HindIII-digested genomic fragments were ligated to oligonucleotide cassettes containing a restriction site (XbaI or HindIII) and a 35-bp consensus sequence as a primer (Takara, Shiga, Japan) (16). PCR amplification using LA PCR in vitro cloning kit (Takara) with a cassette primer and specific primers designed from the partial DNA sequence of the 16-kDa protein were performed to obtain upstream and downstream DNA sequences.
A computer software package, LASERGENE (DNASTAR Inc. Madison, WI), was used for DNA and protein sequence analysis. The sequence was further analyzed by performing homology searches using programs of FASTA (17,18) and BLAST (19) against the GenBank and EMBL DNA data bases.
For purification of the 16-kDa-HA fusion protein, anti-HA antibody (BabCo) was immobilized on protein G-Sepharose (Amersham Pharmacia Biotech) by affinity (22) and resuspended at 50% (v/v) of concentration in phosphate-buffered saline. The cell lysate (total protein, 4 mg) was incubated with the antibody-Sepharose complex suspension (500 l) for 1 h at 4°C. Then, after centrifugation at 500 ϫ g the supernatant was removed, and the Sepharose complex was washed with lysis buffer five times. The Sepharose complex was resuspended with lysis buffer at 50% (v/v) and used for the GTPase assay.
SDS-PAGE and Western Blotting-Preparation of samples for electrophoresis were as follows (23). Cell lysate or 16-kDa protein-Sepharose complex was mixed in half-volume of 3ϫ sample buffer (0.1875 M Tris-HCl, pH6.8, 15% 2-mercaptoethanol, 6% sodium dodecyl sulfate, 15% sucrose, and 0.006% bromphenol blue) and denatured. SDS-PAGE was performed at 15% (w/v) acrylamide gel. The proteins were stained with silver. The molecular weight was calculated from a standard linear regression curve by using low-molecular-weight calibration kit (Amersham Pharmacia Biotech). Western blotting was carried out as follows. Polyacrylamide gel was blotted onto PVDF membrane by electroblotting (24). For immunostaining of Western blots, monoclonal mouse anti-HA antibody (described above) was used at 1:5000 dilution and developed with goat anti-mouse IgG secondary antibody coupled to alkaline phosphatase (Zymed Laboratories Inc.).
GTP Binding Assay-The cell lysate and 1 l of [ 35 S]GTP␥S (100 nM, 1,250 Ci/mmol) were incubated in 20 l of binding buffer (50 mM Tris-HCl, pH7.5, 5 mM MgCl 2, 1 mM EDTA, and 0.3% Tween 20) for 1 h at 30°C (25). The mixtures were irradiated by UV-light (modeUVM-57, UVP Inc., Upland, CA) for 20 min on ice for ultraviolet-cross-linking of GTP to target protein (21). Half volume of 3ϫ sample buffer was added in the reaction mixture and denatured by heating. Supernatants were applied on 15% (w/v) acrylamide gel. After SDS-PAGE, the gel was dried, wrapped with tight-cling plastic wrap, and placed on an imaging plate (BAS-MS, Fuji Photo Film, Kanagawa, Japan) to visualize labeled bands by bioimaging analyzer (BAS-1500, Fujifilm, Kanagawa, Japan).
GTPase Assay-The 16-kDa protein-Sepharose complexes were incubated at 30°C for 1 h in 30 l of binding buffer with [␣-32 P]GTP or [␥-32 P]GTP (3 Ci/nmol). The nucleotides in supernatant of reaction mixture were resolved by thin layer chromatography according to the procedure described by RayChaudhuri and Park (21). The chromatography was contacted on the imaging plate, and the radioactive spots were analyzed by bioimaging analyzer. Competitive inhibition experiments were performed as for the GTPase assay in the presence of each competitor. Competitors were added at a final concentration of 1 mM each of unlabeled nucleotides such as GTP, GDP, ATP, CTP, UTP, and ITP on separate experiments.
Electron Microscopy-To observe the effect of AlF 4 Ϫ on BMP synthesis, M. magneticum AMB-1 was grown as described under "Strains and Growth Conditions", except that sodium fluoride was added to 5 mM and AlCl 3 was added from 5 to 100 M (26). After growth until stationary phase, cells were fixed for 24 h in 2% glutaraldehyde at 4°C. Observation was carried out using transmission electric microscopy (H-700H, Hitachi, Tokyo, Japan).

Isolation of the Gene Encoding Mms16 and Its Sequence
Analysis-58 amino acid residues from the N-terminal sequence of the 16-kDa protein, designated Mms16 (magnetic particle membrane-specific protein), were determined (Fig.  1A). Primers (S1 primer: 5Ј-CATAAGCAGACCGAGCAGTTCT- TCGA-3Ј; S2 primer; 5Ј-TTGGCCTGGGTCAGGGCCTCGAT-GTT-Ј) for amplification of the DNA fragment encoding the N-terminal peptide (44 amino acids) were designed using the amino acid sequences. The amplified DNA fragment (132 bp) was sequenced, and it was confirmed that the deduced amino acid sequence was identical to that of Mms16 (Fig. 1C). To obtain a complete nucleotide sequence of the mms16 gene, PCR amplification for gene walking was conducted. A 0.7-kb fragment of upstream region of mms16 was amplified using the cassette primer and S4 primer (5Ј-CGCTGGTTGGCGACGAT-GGTCTCGACATCC-3Ј) (Fig. 1B). A 0.6-kb fragment of downstream region was amplified using the cassette primer and S3 primer (5Ј-AAGTATCTGGGCGATTTCAAGGTTCC-3Ј) (Fig.  1B) Fig. 1C.
FASTA search analysis showed that the whole deduced amino acid sequence of mms16 has insignificant homology with known proteins in the data base. On the other hand, the sequence of the Mms16 protein in a specific region showed homology with several GTP/ATP-binding proteins (e.g. signal recognition particle protein, two component regulator, initiation factor IF-2, or vacuolar H ϩ -ATPase subunit) following BLAST search analysis. A P-loop motif (ATP/GTP-binding motif) GXXXXGK (X being any amino acid residue) that is present in a large number of nucleotide-binding protein (27) has also been observed as a similar sequence ( 82 GSPQGK 87 ) in Mms16.
Purification of Mms16-HA Tag Fusion Protein-Overexpression of Mms16-HA tag fusion protein in E. coli was performed. Although remarkable induction of Mms16-HA tag fusion pro-tein in transformant was not observed at the existence of IPTG, a specific band was observed as shown in Fig. 2A, lane 4 (indicated by arrow) after beads purification, and it was confirmed to be the Mms16-HA tag fusion protein by Western blots (Fig. 2B). The bands of H and L chain were derived from Sepharose beads on which anti-HA antibody was immobilized. The purified Mms16-HA tag fusion protein was used for GTPase assay. Fig. 3A shows GTP cross-linking to the Mms16-HA tag fusion protein. The Mms16-HA tag fusion protein from transformant, which was confirmed by Western blot (Fig. 3A, a), cross-linked with [ 35 S]GTP␥S (Fig. 3A, b). The ability of the purified protein to convert [␣-32 P]GTP to [␣-32 P]GDP is shown in Fig. 3B where Mms16-catalyzed GTP hydrolysis occurred. Changes in amount of GTP hydrolysis increased depending on time (Fig.  3C). For biochemical characterization of the isolated Mms16, the activity of GTP hydrolysis was measured under varied GTP concentrations from 1 M to 20 mM. Results showed that activity increased with increasing GTP concentration dependently (Fig. 4). These results indicate that Mms16 is a GTPase.

Activity of GTP Binding and GTP Hydrolysis of Mms16 -
The specificity of Mms16 GTPase activity was tested by adding nucleotide competitors (Fig. 5). Excess unlabeled GTP could reduced Mms16 GTPase activity, whereas remarkable reduction of activity was not observed in the presence of other NTPs (CTP, ATP, UTP, and ITP). On the other hand, GDP enhanced the ability of Mms16 GTPase. These results show that GTP is the preferred substrate for Mms16-catalyzed ribonucleotide triphosphate hydrolysis.

Relation of GTPase Activity to Bacterial Magnetic Particle Synthesis-AlF 4
Ϫ has been shown to inhibit ATPase and GTPase activity (26,28). AlF 4 Ϫ also inhibit transport on an in vitro system that reconstitutes vesicular transport (29,30). To investigate the relationship between GTPase and BMP synthesis, increasing concentrations of AlCl 3 with 5 mM NaF were added to AMB-1 culture. Although the AlF 4 Ϫ complex did not inhibit growth, cells lost magnetism as the concentration of AlF 4 Ϫ increased (Table II). Observation by electron microscopy reveals that cells grown with AlF 4 Ϫ contained interrupted chains of BMPs (Fig. 6B). The number of BMP decreased com- pared with that of normal condition cell. Thus, magnetism of cells becomes weaker following the increased concentration of AlF 4 Ϫ . These results suggest, at least, that GTPase activity is required for BMP synthesis.

DISCUSSION
This work reports the functional analysis of a magnetosomespecific protein from a magnetic bacterium for the first time. Mms16 protein, identified as the most abundant protein in magnetosomes, contains a putative P-loop sequence; however, this putative P-loop sequence has only three residues of any combination of four between Gly and Gly-Lys, and it contains only one motif of the four conserved sequence motifs in the GTPase superfamily (31). Despite these characteristics, Mms16 protein shows GTPase activity. The Ras, Rho, Rab, and Ran of small GTPase family were also observed to have spontaneous GTPase activity (32). These results suggest that the isolated Mms16 is similar to these GTPases. For example, the intrinsic GTP dissociation rates of Ras are about 10 5 times lower than in Ras in the presence of an exchange factor (33). Thus, interactions of other proteins would be required for binding to the membrane, GTP/GDP exchange, or stimulation of the activities in situ similar to that in eukaryotic cells. Probably, this Mms16 is also controlled by some factors in vivo, hence GTPase is expected to act in much lower levels of GTP concentration.  The Mms16 was localized in the cytoplasm of E. coli but was strongly associated with the BMP membrane since it was not removed when subjected to over 20 times stringent washing with sonication. Therefore, it appears to be directly associated with the membrane. Motif analysis of Mms16 indicated that it contained a myristoylation site; however, the site is not at the N terminus. ARF, one of the family of mammalian small GTPases, is anchored in the membrane by myristoylation of Gly at the N terminus (34). From observation of such a membrane-bound form and GTP hydrolysis activity, the Mms16 has similar properties to eukaryotic small GTPases. Therefore, it is anticipated that its function may also be similar.
It does not seem to be similar to other prokaryotic GTPbinding proteins that have also been widely studied and fall into five groups; (i) initiation factor (IF-2), elongation factor (EF-Tu) (35), (ii) cytokinesis protein FtsZ, which forms a ring at the leading edge of the cell division site (36) and is a GTPase that has significant structural similarity to tubulin (37), (iii) signal recognition particle (SRP) and SRP receptor ␣ subunit (38), (iv) Era protein, which encodes an essential protein (39 -41) that may play a role in DNA replication or chromosome partitioning in E. coli (42, 43), and (v) Obg protein, which plays a crucial role in sporulation and DNA replication in Bacillus subtilis (44,45). Although its precise cellular function is unknown, Era and Obg have been implicated in a wide array of cellular functions.
Eukaryotic cells have an elaborate network of organelles, many of which are in constant and bi-directional communication through a flow of small transport vesicles. The small transport vesicles that mediate membrane trafficking between intracellular organelles are encased in a protein coat. The small GTPase family is involved in the priming and budding of trafficking vesicles. Well studied examples of vesicle formation include the Golgi-derived COP (coat protein)-coated vesicles (46,47). In events leading to the formation of COP-coated vesicles, N-myristoylated ARF (48) anchored in the membrane is recruited concomitantly with GTP/GDP exchange at the Golgi surface (49) where it triggers assembly of the coatmer (50,51). Coat assembly drives vesicle budding (52). GTP hydrolysis is required to release the coatmer complex and ARF from the vesicles (46,47).
In several bacteria, intracellular membranal structures clearly originate from the cytoplasmic membrane. Although the origin of the thylakoid membrane of cyanobacteria are not known, in most phototrophic bacteria, intracytoplasmic membranes that are invaginations of the cytoplasmic membrane, are the site of the photosynthetic apparatus (53). Also chemolithotrophic, nitrifying bacteria possess complex arrangements of internal membranes (54). Spheroplasts of Nitrosomonas do not contain internal vesicles (55). This suggests that internal membranes of those organisms are invaginations. Methanotrophs generally contain intracytoplasmic membranes (56), which appear to be involved in the methane oxidation pathway. Azoarcus sp. BH72 is a chemoheterotroph, and their intracytoplasmic membranes, related to nitrogen fixation, often appear as stacks of flattened vesicles similar to those of many phototrophic bacteria (57). Surprisingly, the process of intracytoplasmic membrane formation is poorly understood in prokaryotes and no genes involved in this process have yet been isolated.
In magnetic bacteria, the origin of membrane vesicles that envelope the magnetite particles is unknown. We can hypothesize that magnetosomes arise through invagination of the cytoplasmic membrane. The BMP specific GTPase may also be involved in vesicle formation through a process similar to that which occurs in eukaryotic vesicle formation. Inhibition of GTPase activity by AlF 4 Ϫ causes inhibition of BMP synthesis, suggesting that GTPase activity is required for BMP formation.
In previous studies, we have reported two proteins, MagA and MpsA, presented on the BMP membrane. MagA is expressed on both BMP and cytoplasmic membrane and showed iron transport activity. Interestingly, MagA topology is inversely oriented between the cytoplasmic membrane and the BMP membrane (6,7), something that would occur if magnetosomes were formed by membrane invagination. MpsA, acyl-CoA transferase homologue, is present specifically on the BMP membrane in higher amounts than MagA, but its function is unknown. In eukaryotes, ARF requires acylation for membranal invagination (58). Taken together, our previous studies and this study, we have proposed a mechanism for the primary formation of BMP. The Mms16 binds with the membrane. And this serves to prime the invagination from the cytoplasmic membrane. Acyl-CoA and GTP hydrolysis might be required during vesicle budding. MagA takes up ferric iron into the vesicle released from cytoplasmic membrane.