Functional Analysis of the Magnetosome Island in Magnetospirillum gryphiswaldense: The mamAB Operon Is Sufficient for Magnetite Biomineralization

Bacterial magnetosomes are membrane-enveloped, nanometer-sized crystals of magnetite, which serve for magnetotactic navigation. All genes implicated in the synthesis of these organelles are located in a conserved genomic magnetosome island (MAI). We performed a comprehensive bioinformatic, proteomic and genetic analysis of the MAI in Magnetospirillum gryphiswaldense. By the construction of large deletion mutants we demonstrate that the entire region is dispensable for growth, and the majority of MAI genes have no detectable function in magnetosome formation and could be eliminated without any effect. Only <25% of the region comprising four major operons could be associated with magnetite biomineralization, which correlated with high expression of these genes and their conservation among magnetotactic bacteria. Whereas only deletion of the mamAB operon resulted in the complete loss of magnetic particles, deletion of the conserved mms6, mamGFDC, and mamXY operons led to severe defects in morphology, size and organization of magnetite crystals. However, strains in which these operons were eliminated together retained the ability to synthesize small irregular crystallites, and weakly aligned in magnetic fields. This demonstrates that whereas the mamGFDC, mms6 and mamXY operons have crucial and partially overlapping functions for the formation of functional magnetosomes, the mamAB operon is the only region of the MAI, which is necessary and sufficient for magnetite biomineralization. Our data further reduce the known minimal gene set required for magnetosome formation and will be useful for future genome engineering approaches.


Introduction
The ability of magnetotactic bacteria (MTB) to orient in the earth's magnetic field is based on specific organelles, the magnetosomes. In the a-proteobacterium Magnetospirillum gryphiswaldense and related MTB, magnetosomes consist of magnetite (Fe 3 O 4 ) crystals enclosed by a phospholipid membrane. This magnetosome membrane (MM) contains a specific set of .20 proteins, which direct the biomineralization of highly ordered crystals [1,2,3]. Synthesis of magnetosomes has recently emerged as a model for prokaryotic organelle formation and biomineralization [4,5] In addition, magnetosomes represent biogenic magnetic nanoparticles with unique characteristics, which make them attractive for use in a wide range of biomedical and biotechnological applications [4,6,7]. Although the mechanism of magnetosome synthesis is not understood in detail, several recent studies revealed that the formation of functional magnetosomes depends on several steps, which include the invagination of MM vesicles from the inner membrane [8,9], the transport of iron and crystallization of magnetite within these vesicles [10], and the assembly of mature crystals into a linear chain along a filamentous cytoskeletal structure [9,11,12,13]. It has been also become clear that each of these steps is under strict genetic control. By proteomic analysis of M. gryphiswaldense (in the following referred to as MSR), genes encoding the MM-specific proteins were identified within a single genomic magnetosome island (MAI) [14,15]. The functional significance of this region was confirmed by a comparative genomics approach, which revealed that magnetotaxis signature genes are predominantly located within the MAI [16]. Because of their general conservation in other cultivated and uncultivated a-proteobacterial MTB [3,17,18,19] it has been suggested that the MAI was transferred horizontally [15,16,18,20,21]. This was further corroborated by the recent discovery of homologous gene clusters in the d-proteobacteria Desulfovibrio magneticus RS-1 [22] and the multicellular magnetotactic prokaryote (MMP) [23], as well as in the deep-branching Nitrospirae-phylum [21]. In addition to all genes, so far implicated in magnetosome biomineralization, the MAI of MSR contains a number of genes with unknown functions and numerous transposase genes that account for .20% of the coding region [14]. Owing to frequent homologous recombinations between the numerous direct or inverted repeats associated with transposase genes, the MAI is genetically unstable, resulting in frequent spontaneous loss of the magnetic phenotype [15,24]. In MSR all known magnetosome genes are comprised within four gene clusters known as mms6, mamGFDC, mamAB, and mamXY operons. First experimental indications for their functional significance in magnetosome formation came from the isolation of a nonmagnetic mutant strain, which had lost 40 kb of the MAI by a spontaneous deletion that included the mamAB, mms6 and mamGFDC operons [25]. Targeted deletion of the entire mamGFDC operon revealed that the small MamGFDC proteins, which account for .35% of all magnetosome-associated proteins, are not essential, but involved in size control, since mutant cells formed smaller and less regular magnetite crystals [26]. In a recent study by Murat et al. deletion analysis of the MAI in M. magneticum AMB-1 (referred to as AMB) revealed three regions, which are crucial for magnetite crystal formation [27]. Whereas the deletion of the R2 and R3 regions including parts of the mamGFDC and mms6 operons led to severe defects in the size and morphology of the crystals, loss of the mamAB operon resulted in cells entirely devoid of magnetite crystals [27]. Only the deletion of mamE, M, N, O, L, I, and also of mamQ and mamB, if co-deleted with their respective duplicates outside the mamAB operon, entirely abolished magnetite synthesis. Non-magnetic cells were also observed upon deletion of this operon in MSR [25]. This suggested that only the mamAB operon may contains genes that are absolutely essential [27]. However, it has remained unknown whether this region is also sufficient for magnetosome biomineralization in the absence of other magnetosome genes, since possible genetic redundancy was suggested by the identification of genes, which are identical or similar to genes from mamAB operon and partially encoded within a ''magnetosome islet'' located elsewhere in the genome of AMB [28].
Despite morphological similarities between the strains AMB and MSR, previous studies suggested that function of orthologous genes might be somewhat distinct in these organisms depending on their different genetic context [8], since only about 50% of all genes are shared by the genomes of these two strains [16]. In particular, the MAI regions flanking the magnetosome operons show a divergent organization, gene content and were speculated to possibly harbor additional determinants for magnetosome formation [16,18]. Here, we show that highly expressed and conserved genes within the MAI of MSR are mostly confined to the mms6, mamGFDC, mamXY, and mamAB operons. By deletion of these operons, either independently or in combination, we demonstrate that all four of them have crucial and partially overlapping functions in the synthesis of functional magnetosomes, whereas only the mamAB operon is absolutely essential for magnetite biomineralization. Intriguingly, even in the absence of all other three operons as well of further parts of the MAI, the mamAB operon proved sufficient to maintain synthesis of small magnetite crystals. A further motivation for this study was to explore the potential for reduction of dispensable or instable gene content from the residual MAI. By using an improved Cre-loxbased technique, we demonstrate that 115 kb of the MAI can be deleted without any consequences for growth, while 59 kb have no obvious function in magnetosome synthesis.

Expression of MAI genes coincides with their conservation and operon localization
Besides numerous (.50) transposase and phage related genes, the mam and mms operons within the MAI are flanked by a number of ORFs, mostly annotated as hypothetical genes, which may represent either unrecognized determinants for magnetosome formation, genes with unknown different functions, or simply pseudogenes or misannotations. To tentatively distinguish between regions of predicted relevance and those not likely involved in magnetotaxis, we reasoned that putative magnetosome genes are expected (I) to lack strong prediction of other cellular functions, (II) to be highly conserved among MTB, and (III) to be expressed during magnetosome synthesis. We therefore reassessed functional annotation of the MAI against current databases. Only 12 of the MAI genes have functionally predicted homologs outside MTB (Fig. 1), which encode three hemerythrin-like proteins, putative regulatory proteins, secretion components, a sensory transduction histidine kinase, a partition-related protein, and an IdiA fragment (Table S1). To identify conserved genes, we tested by blastp analysis the presence of all genes from the MAI of MSR against all genomic information available from cultivated MTB (Fig. 1, Table  S1). Genes that are highly conserved between several MTB were found mostly confined to the mam and mms operons, where ten ORFs (mamE, K, M, O, A, Q, B, T, and with lower similarity mamI and mamP) are conserved in all analyzed strains including MSR, AMB, Desulfovibrio magneticus RS-1, M. magnetotacticum MS-1, Magnetococcus marinus MC-1, and Magnetovibrio blakemorei MV-1. MamE, I, K, M, O, P, A, Q, B genes were also detected in the metagenomic MAI fragment Fos001, whereas a second metagenomic clone Fos002 lacks mamI but contains mamT [20]. MamE, I, M, P, A, B, and two mamQ homologs were also found in the incomplete MAI sequence of ''Candidatus Magnetobacterium bavaricum'' [21]. Nine ORFs have homologs in only one other MTB (Fig. 1), and 41 genes are shared by at least all magnetospirilla (Fig. 1). However, only 7 of these genes show positional conservation within the MAI of AMB, whereas the rest is located elsewhere in the genome in the latter strain. 22 genes, which are mostly confined to larger regions close to the putative boundaries of the MAI, are specific for MSR (i. e., have no homolog in any other organism), and appear less likely to represent determinants required for magnetosome formation. Thus, hypothetical genes outside the mam and mms operons are poorly conserved, with none of them found shared by all sequenced MTB.
To identify expressed products of ORFs encoded within the MAI, we performed proteomic analyses of magnetosomes, as well as intracellular soluble and membrane-enriched protein fractions of cells grown under magnetite forming conditions. In total, 923 proteins were identified by 1D LC-MS/MS analysis, or from spots detected on 2D gels. In summary, only 33 proteins encoded within the MAI were found expressed in the membrane or magnetosome fraction of MSR. These for instance include, with the exception of Mgr4074, MamI, MamL, and MamX, all proteins encoded by the mamAB, mamGFDC, mms6, and mamXY operons, whereas only seven genes outside the mam and mms operons were found expressed (mgr4041, mamW, mgr4067, mgr4106, mgr4109, mgr4115; mgr4152, Fig. 1; Table S1) as well as one gene barely inside the boundaries of the 130 kb region (mgr4022) [29]. With the exception of MamK, none of the MAI proteins was detected within the soluble protein fraction among the analyzed spots.

Mutagenesis of MAI genes
By excluding putatively essential genes such as tRNA and rRNA genes, we predicted a core region of 115 kb from mgr4026 to mgr4074, comprising 149 ORFs that are probably not important for central metabolic functions and including all so far known magnetosome genes. According to bioinformatic prediction and expression data, this region was divided into partially overlapping target regions for mutagenesis ( Fig. 1). We constructed 13 mutant strains in which single or several of these targets were excised, resulting in deletions between 400 bp and 61 kb. Shorter deletions (up to 7 kb) were generated by allelic replacement (double crossover mediated by homologous recombination, Fig. S1A) [30], whereas Cre-lox excision ( Fig. S1B; Fig. S2) [25,31], was used for the construction of larger deletions between 5 and 53 kb. We noticed that success of deletion mutagenesis was not fully predictable. For instance, whereas we previously generated the DA17 deletion in the MSR-1B background [25], we failed to enforce deletion of parts of that region (DA2) in the WT background despite of repeated attempts. With few exceptions described below, all mutants including the longest deletion (DA14) extending over 58.9 kb exhibited WT-like growth, indicating that no central metabolic functions are encoded by deleted MAI genes. However, Cmag measurements and TEM of mutant strains revealed three different classes of phenotypes with respect to magnetosome formation: (I) Mutants that were unaffected in magnetosome formation, i. e. cells were virtually WT-like with respect to crystal appearance (shape, size, number per cell and alignment) including the long deletions DA3 (9.8 kb), DA4 (27.8 kb), and DA5 (19.7 kb), as well as DmamW (411 bp), eliminating a protein that was previously identified as associated with magnetosomes in MSR [15,16]. (II) Mutants in which magnetosome formation was entirely abolished, as indicated by a pale pink to orange cell pellet (in contrast to the black appearance of the WT), lack of a magnetic response (Cmag = 0) and the absence of any electron dense particles. The non-magnetic mutants DA19, in which an additional 19.7 kb fragment was excised in the background of deletion mutant MSR-1B, and DA15 comprising the mamJKL genes, had in common a deletion of either the entire mamAB operon or parts of it, similar to strains MSR-1B, DA16, DA17 and DA18, which had been generated in previous studies [15,25]. (III) A third class of mutant strains still exhibited a magnetic response, but cells were gradually affected in magnetosome biomineralization or assembly, resulting in fewer, smaller and irregular crystals or distorted chains (Fig. 2). Mutants of this class could be recognized by variable intensities of brownish color of colonies and cell pellets (Fig. 1). Single-operon deletions of mms6 (DA10) and mamXY (DA8) showed a significantly reduced magnetic response, but still contained electron-dense particles with different sizes and shapes (Table 1). Strain DA10 had smaller crystals ( Table 1) that were scattered throughout the cell or aligned in irregular, widely spaced ''pseudo-chains'' (i. e., with ,10 crystals per chain; Fig. 2). Crystals between 25 and 30 nm were predominant, whereas particles larger than 50 nm were not observed, unlike WT crystals that were most frequently between 40 and 50 nm with a maximum size up to 70 nm (data not shown). Besides cubo-octahedral crystals also heterogeneous crystal shapes were observed (Fig. 2). Complementation with fragments comprising genes mgr4072, mgr4073, and mgr4074 restored size, shape and alignment of crystals to WT range within about one third of the cells (data not shown). Strain DA8 had an inconsistent phenotype. TEM revealed a variety of magnetosome appearances between different cells, including those lacking any electron-dense particles ( Fig. 3 A), and those with non-uniform, small crystals lacking any chain configuration (Fig. 3 B-F). Remarkably, many cells contained two distinct types of crystals: short chains of almost regular (i.e., cubicle-shaped) crystals, which were flanked by irregular particles with poorly defined morphologies ( Fig. 3 G-K). Analysis of about 350 crystals from cells of the latter phenotype revealed that approximately 66% of the crystals were irregular and less electron dense, slightly elongate and poorly crystalline particles (Fig. 2). The different particles had distinct size distributions: Among irregular particles, sizes between 15 and 25 nm were most abundant, whereas the regular-shaped crystals had a maximum size of 60 nm, and diameters between 35 to 45 nm were most frequent among them (Fig. 4). The WT-like phenotype could be restored by transcomplementation with plasmid pmamXY containing the entire mamXY cluster (mgr4147 to mgr4150; data not shown). A similar phenotype was observed for the mutant DA7 (Fig. 2) in which the deletion included the regions A4 and A5 in addition to the mamXY operon ( Fig. 1; Table 1), resulting in an average crystal size of 23.5 nm. Crystal number per cell was not significantly affected in comparison to WT (Table 1). Operons whose single deletions had magnetosome phenotypes were also deleted in combination with each other. This was also achieved by modification of the previously described Cre-lox method [25] by using altered lox sequences [32] that enabled the construction of strains bearing multiple unmarked deletions by sequential rounds of insertions and excisions (Fig. S1). In strain DA12 the entire mms6 operon was deleted in addition to the adjacent mamGFDC operon. This resulted in a stronger phenotype compared to its parent strain DGFDC [26], i. e. it formed even fewer and smaller magnetosomes that were aberrantly shaped and less regularly aligned (Fig. 2). The deletion of both operons also resulted in a particle size reduction of 52% compared to the WT, although crystals were only slightly smaller than in a deletion of mms6 operon alone (Table 1). While crystal numbers per cell were only slightly reduced in comparison to the mms6 operon mutant, the magnetic response of DA12 culture was markedly weaker (Cmag [DA12] = 0.6; Table 1). The DA11 double deletion mutant of mamXY and mamGFDC operons showed a reduced Cmag (Cmag [DA11] = 1.2; Table 1) and a phenotype as inconsistent as strain DA8 (Fig. 3). Compared to DA8, particles were smaller (Fig. 4), fewer per cell and less frequently aligned in chain-like structures (Fig. 2). Also, the number of crystals with regular morphology was reduced to 21.8%.
We also eliminated mms6, mamGFDC, and mamXY operons altogether using two approaches: While sequential triple deletion by allelic replacement of the three regions resulted in strain DA13, deletion of the mamGFDC and mms6 operons in a parental background (DA7) that already lacked the entire right arm of the MAI (about 53 kb) containing the mamXY operon resulted in strain DA14 (Fig. 1). Remarkably, both strains still displayed a detectable, although weak magnetic response (Cmag [DA13] = 0.3; Cmag [DA14] = 0.5) and contained tiny misshapen electron dense crystallites ( Fig. 2; Table 1). Crystal sizes were decreased to 54.8% of WT size and 84.8% of DA8 size, but were identical between DA13 and DA14 strains (Table 1). From all mutants, both strains DA13 and DA14 contained the fewest magnetosome number per cell (12-13 in average) and crystal shapes resembled the irregular morphologies found in strains DA7, DA8, DA10, DA11, and DA12. Thus, the phenotype of DA13 and DA14 is characterized by the coexistence of distinct particle morphologies found in the respective single operon deletion mutants (Fig. 5).

Discussion
We performed a comprehensive investigation of the MAI in MSR by combined bioinformatic, proteomic and genetic analysis. With the exception of mgr4041 and mgr4106, which are MSRspecific, all other genes from the 115 kb core region that were found expressed are also highly conserved in magnetospirilla or even all MTB. The majority of expressed genes (26 of 33) were localized within the mms6, mamGFDC, mamAB, and mamXY operons [25,27]. These were also the only regions, which displayed a magnetosome phenotype upon their deletion. Thus, in contrast to previous observations in AMB [27], conservation and expression of MAI genes showed a strong correlation with a function in magnetosome formation.
We used a Cre-lox based method [25,31], which allows the efficient excision of large fragments. The largest single deletion obtained by this method comprised 53 kb in strain DA7. Using modified lox-sites enabled multiple sequential rounds of markerless deletions. This resulted in strains in which up to 59 kb were deleted, comprising about 50% of the MAI and encoding 78 ORFs. Despite of repeated attempts, no deletion of the A2 region ( Fig. 1) was obtained. Whereas the DA17 (MSR_SU12) deletion was straightforwardly generated in the MSR-1B background in a previous approach [25], we failed to partially delete this region (DA2) in the WT background. It remains to be shown whether this was due to low efficiency, or if deletion of this region would be lethal only in the presence of the residual MAI genes.
The absence of detectable phenotypes apart from magnetosome formation in the deletion strains indicates that the MAI encodes no important functions for growth under laboratory conditions. Whereas less than 25% of the MAI region could be associated with magnetosome formation, more than 50% of the MAI seems to have no obvious functions. Remarkably, among the genes with no phenotype are several of the magnetospirilla-specific genes, such as mgr4067, mgr4109, mgr4115, mgr4152, and mgr4057 (mamW), which had been previously implicated in magnetite synthesis because of its magnetosome expression [16]. We also failed to detect a phenotype for the two hemerythrin-like genes harbored within the deleted A3 region. Because of their MAI localization and the known functions of hemerythrins from other organisms in the sensing or transport of oxygen and iron, it was speculated that these proteins may play a role in magnetoaerotaxis and magnetosome formation [33,34]. However, it cannot be excluded that their loss can be compensated by the Our deletion analysis confirmed several results of previous studies, in which the functional significance of several regions, such as mamAB, mms6, and mamGFDC were shown for AMB [27], and partially for MSR [25,26]. However, despite of the high similarity of targeted genes, we also found several striking differences between the two organisms. One example is the conserved mamXY operon, which contains several magnetotaxis signature genes, and for which a key role was predicted mostly based on comparative genome analysis [16]. While MamY was recently implicated in MM biogenesis in AMB [35], mamX has similarity to the serine like proteases MamE and MamS, whereas MamZ is an ortholog of MamH and resembles permeases of the major facilitator superfamily. The FtsZ-like gene has homology to the tubulin-like protein, which is involved in cell division in many bacteria [36]. In contrast to the mamXY operon deletion in AMB, which did not show a strong effect [27], we found that mamXY genes have a crucial function in magnetite biomineralization of MSR. This is consistent with the results obtained by Ding et al., who reported that the deletion of the ftsZ-like gene alone already resulted in the synthesis of smaller, predominantly superparamagnetic particles [37]. The deletion of all mamXY genes had an even stronger effect, which is different from all previously reported magnetosome phenotypes. Strikingly, all deletions including this operon had an inconsistent phenotype, which varied between different cells. In addition to size reduction, this was characterized by the coexistence of various distinct magnetosome morphologies within many single cells.
The deletion of genes from the mms6 operon had slightly different effects in AMB and MSR too. Single deletion of the mms6   gene in AMB already caused smaller and elongated crystals [38], thus resembling the R3 mutant constructed by Murat et al. [27], which comprised deletion of genes from both the mms6 and mamGFDC operons. In contrast, 58% of crystals within cells of the single operon deletion mutant in MSR (strain DA10) still had cubicle-shaped appearance, whereas elongate crystals were absent from the mutants DA10 and DA12. Although the phenotypes cannot be directly compared, since the extents of deletions are not fully congruent, this might point towards slightly distinct functions of the homologous regions in AMB and MSR. In MSR co-deletion of the mms6 operon together with mamGFDC in strain DA12 resulted in a further reduction of shape regularity and alignment of crystals, but only in a slight decrease of size, whereas the number of particles per cell was similar to strain DA10 (Dmms6). This argues for a certain functional overlap between the two operons, which is consistent with the high similarity between some of the encoded proteins, such as MmsF and MamF, which share 61% identity, and Mms6, which shares a conspicuous LG-rich motif with MamG and MamD [2,39]. However, single operon mutant phenotypes suggest that genes of the mms6 operon have a more pronounced effect on crystal size, number and alignment than mamGFDC, perhaps by direct binding to the surface of nascent crystallites through hydrophilic domains [40], or by enlarging the surface and curvature of MM vesicles, which spatially constrain the growth of magnetite crystals [26].
Intriguingly, even in the DA14 and DA13 strains, in which the mms6, mamGFDC, and mamXY operons were deleted in triple, magnetite formation was not entirely abolished and cells still weakly aligned in magnetic fields, although crystal sizes were further decreased and elongate crystals were present. Despite of a functional overlap in size control of magnetite crystals, the roles of the mms6, mamGFDC, and mamXY genes are not fully redundant, as indicated by the distinct morphologies found in their respective single operon deletions. While simultaneous excision of the mamGFDC and mms6 operon lead to heterogeneous cubicle-shaped crystals, loss of mamXY operon lead to poorly crystallin and elongate crystals, which were also detected within the double deletion mutant of mamXY and mamGFDC. Interestingly, these effects are superimposed in the mamGFDC, mms6, mamXY triple deletion strains (DA13 and DA14), in which crystallites of both morphologies are present. Altogether, these observations indicate that the mamGFDC, mms6 and mamXY operons have important and additive functions for the formation of regularly shaped crystals that are sufficiently large to be functional for interaction with the weak geomagnetic field [39,41].
Consistent with observations for AMB [27], only the mamAB operon contains genes, which are essential for magnetosome formation. However, our data for the first time demonstrate that the mamAB operon is the only region of the MAI, which is necessary and sufficient to maintain magnetite biomineralization even in the absence of the mamGFDC, mms6, and mamXY clusters. Although it cannot be precluded that additional, so far unrecognized determinants might be encoded outside the MAI, this further reduces the minimal gene set, which is likely required for biomineralization. As the MamJ and MamK proteins were already shown to have roles in magnetosome chain assembly rather than in biomineralization [8,42], the core set of MAI genes essential for magnetite biomineralization in MSR can be expected to be less than 15, and according to the identification of further non-essential genes in the mamAB operon of AMB (mamA, H, U, V, P, T, R, S) [27] this number is likely to shrink further.
Our results will be also useful for future genome reduction approaches. Comparable experiments in other bacteria have shown that large-scale deletions of target sequences are extremely powerful in engineering of strains optimized for biotechnological processes [43,44,45]. By stepwise removal of unnecessary or problematic genomic regions, in future approaches also strains of MSR can be engineered for the production of magnetosome particles, which may exhibit increased genetic stability due to the elimination of repeats and transposases, or might show improved growth or increased magnetosome yields because of reduced gene content. In summary, deletion analysis of MAI indicates that whereas only the mamAB operon is essential, different regions have important functions in control of size and morphology of magnetosomes. Thus, modular deletion or expression of various magnetosome genes and operons could be used for the production of engineered magnetic nanoparticles with tailored properties.

Molecular and genetic techniques
The working draft of M. gryphiswaldense genome sequence (GenBank accession number No. CU459003) was used for primer design. Oligonucleotids (Table S3) were purchased from Sigma-Aldrich (Steinheim, Germany). Chromosomal DNA of M. gryphiswaldense was isolated as described previously [3]. Plasmids were constructed by standard recombinant techniques as described in detail in Materials and Methods S1. All constructs were sequenced on an ABI 3700 capillary sequencer (Applied Biosystems, Darmstadt, Germany), utilizing BigDye Terminator v3.1. Sequence data were analyzed with Software Vector NTI AdvanceH 11.5 (Invitrogen, Darmstadt, Germany).

Analytical methods
Magnetic reaction of cells was checked by light microscopy applying a bar magnet.
Optical density and magnetic response (Cmag) of exponentially growing cells were measured photometrical at 565 nm as previously reported [48]. For Cmag messurement a magnetic field of approximately 70 millitesla was used [48]. As this field can possibly magnetize small magnetosomes in the superparamagnetic size range and cause artificially high Cmag readings, all putative magnetosome phenotypes were verified by transmission electron microscopy (TEM). For TEM analysis, exponential cells were 10fold concentrated and adsorbed onto carbon-coated copper grids. Samples were viewed and recorded with a TECNAI FEI20 microscope (FEI, Eindhoven, Netherlands). Magnetosome crystals were analyzed with respect to size, shape and numbers per cell. Magnetosome crystals were scored for chain formation as described by [8]. For pictures of cell pellets, cells were cultivated anaerobic in FSM and 10 9 cells were concentrated by centrifugation.
Cell fractionation, protein digestion, mass spectrometry, and data analysis For proteomic analysis M. gryphiswaldense WT was grown in microaerobic 1-liter batch cultures and cell fractions (membraneenriched, soluble, and magnetosomes) were prepared as previously described [2,29]. Soluble proteins were separated in 2D PAGE (pH 4-7 and 3-10). Analysis of 2D gels including relative quantification was done with the Delta2D software (Decodon, Greifswald, Germany). Protein spots were cut from 2D gels, transferred into microtiter plates, and tryptically digested using the Ettan Spot Handling Workstation (GE Healthcare, Munich, Germany). Mass spectra of protein fragments were measured by MALDI-TOF-MS/MS using a Proteome Analyzer 4800 (Applied Biosystems, Munich, Germany). The parameters for measurements were set as described in [49]. The spectra were searched against the published genome sequence from M. gryphiswaldense by using the JCoast 1.6 software [50], and proteins were identified using the Mascot search engine. For analysis of magnetosomes and membrane proteins, gel lanes obtained from 1D-SDS-PAGE were cut into 10 equal slices. Gel slices were digested manually with trypsin and analysed by LC coupled mass spectrometry performed as described by [51]. Relative quantification of membrane proteins was based on spectral counting using Scaffold [52]. Figure S1 Schematic illustration of methods for generation of deletions within the MAI. (A) Allelic replacement of target genes using double cross-over followed by removal of selection marker with Cre-lox mediated excision. (B) Cre-lox recombination using the modified sequences lox71 and lox66 for specific excision of large chromosomal regions and construction of marker-less mutant strains. After excision the modified lox* sequence remains in the genome, but is poorly recognized by Cre recombinase making multiple recombination events possible. (TIF) Figure S2 Constructed suicide plasmids (pAL01 to pAL11_term) for integration of modified lox sequences. Regions (AL01 to AL11) within the MAI of M. gryphiswaldense used for site-specific plasmid insertion via homologous recombination to enable subsequent excision between lox sites of double insertions. (TIF)