The Polar Organizing Protein PopZ Is Fundamental for Proper Cell Division and Segregation of Cellular Content in Magnetospirillum gryphiswaldense

Magnetotactic bacteria (MTB) share the unique capability of magnetic navigation, one of the most complex behavioral responses found in prokaryotes, by means of magnetosomes, which act as an internal compass. Due to formation of these unique nanoparticles, MTB have emerged as a model to study prokaryotic organelle formation and cytoskeletal organization in conjunction with complex motility systems. Despite the high degree of subcellular organization required in MTB, less is known about cell-cycle-related factors or proteins responsible for spatiotemporal polarity control. Here, we investigate the function of the polar organizer PopZ in the magnetotactic alphaproteobacterium Magnetospirillum gryphiswaldense. Although PopZ is widely distributed among the alphaproteobacteria, its function in MTB belonging to this class has remained unexplored. Our results suggest that in M. gryphiswaldense, PopZ has a key role during cell division and subcellular organization. Furthermore, we show that PopZ localization and function differ from other nonmagnetotactic alphaproteobacterial model organisms.

implementation within the cell-cycle-dependent interaction network of the heterologous host. In summary, our results reveal a key role of PopZ Mgr in subcellular organization and provide the first fundamental insights into its function in cell cycle control and polarity determination in MTB. Thus, this work also demonstrates the usefulness of M. gryphiswaldense as a potential and emerging model to scrutinize the bacterial cell cycle and its coordination to spatiotemporal organelle organization.

RESULTS
PopZ Mgr localizes to both cell poles. In M. gryphiswaldense, PopZ Mgr is encoded in a conserved genomic region, similar to C. crescentus, next to putative genes coding for a valyl-tRNA synthetase and an outer membrane efflux protein (see Fig. S1A in the supplemental material). To study its localization pattern throughout the cell cycle, PopZ Mgr was translationally fused to green fluorescent protein (GFP) by integration of an M. gryphiswaldense codon-optimized gfp gene (magegfp [29]) within this genomic region. Expression of popZ Mgr -gfp from its native promoter was verified via immunoblotting using an antibody against GFP (not shown). In vivo time-lapse fluorescence microscopy revealed that PopZ Mgr -GFP localized to both cell poles and exhibited a bipolar localization pattern throughout the cell cycle ( Fig. 1A; see Movie S1 in the supplemental material). PopZ foci at the future new poles appeared at the end of the cell cycle (Fig. 1B). Since conventional wide-field microscopy did not allow us to judge with high confidence if cells with PopZ foci present at the division plane had already completed cytokinesis, we imaged dividing cells with superresolution three-dimensional structured illumination microscopy (3D-SIM). Using FM4-64 membrane staining, 3D-SIM revealed two adjacent PopZ foci (ϳ250 nm apart) at the cell division site ( Fig. 1Ci; see Fig. S2 in the supplemental material). In general, all cells with two PopZ foci present at the division site had already completed separation of their membranes ( Fig. 1Ci and Fig. S2). In contrast, no PopZ foci were observed in cells with membranes and cytoplasm still connected, but which already had undergone partial membrane constriction ( Fig. 1Cii and Fig. S2). These results indicated that formation of PopZ-rich zones at the new poles occurs very late during or shortly after completion of cytokinesis.
Deletion of popZ Mgr causes severe cell division defects. To study the effects of popZ absence in M. gryphiswaldense, a markerless in-frame deletion mutant was constructed. The ΔpopZ Mgr strain was viable, but showed severely impaired growth and increased cell length (Fig. 2). Some cells were elongated up to 60 m (Fig. 2Ai), equivalent to ϳ20-fold the length of a newborn wild-type cell (ϳ3 m). Elongated cells contained between 1 and 3 abnormally long magnetosome chains running in parallel, which were sometimes interspaced by segments of unknown origin and compositioni.e., parts of the cell body that appeared brighter in the electron microscope (Fig. 2Aii, black arrowheads). Elongated magnetosome chains were up to ϳ20 m in length (allowing gaps not larger than 150 nm). Elongated cells with more than 700 particles were observed; however, cell length and number of magnetosomes were well correlated, resulting in ϳ12 particles/m (Fig. 2B). For comparison, we determined that the wild type typically exhibited a median chain length of ϳ1 m and median magnetosome numbers of 35 particles per cell. In contrast to the ΔpopZ Mgr strain, particle number and cell length were only poorly correlated in the wild type (Fig. 2B, inset), similar to previous correlative estimations of magnetosome particle numbers versus cell area (30). Moreover, even in the most highly elongated cells, magnetosome chains were mostly absent from the regions near the cell poles (Fig. 2Aiii). As is also commonly observed in the wild type (31), cells of the ΔpopZ Mgr strain contained large amounts of polyphosphate and polyhydroxybutyrate (PHB) granules ( Fig. 2Ai to Aiii).
During time-lapse microscopy ( Fig. 2C), the wild type divided at a median cell length of 3.3 m, whereas in the ΔpopZ Mgr strain, the cell elongation and unequal division produced a much broader length distribution with a median cell length of 3.9 m during division (Fig. 2D). Furthermore, the time required to complete a division cycle was less regular for the ΔpopZ Mgr strain ( Fig. 2C and Movie S1). While wild-type cells divide approximately every 4 to 5 h, cell division in the ΔpopZ Mgr strain occurred with variable timing and at ectopic positions, suggesting that the generation time as well as division septum positioning were affected. To measure potential asymmetry during cell division, the difference in length of both newborn daughter cells was calculated from time-lapse series (Fig. 2E). Thus, the median cell length differences for wild-type daughter cells were 0.29 and 0.94 m for the popZ Mgr deletion strain, confirming that the ΔpopZ Mgr population contained cells with high variation in length caused by unequal division. Also, a slight asymmetry (ϳ11% off-center) was detected for newborn daughter cells of the wild-type strain, in agreement with previous observations (on average, 15% off-center) by Katzmann et al. (6).
CET ultrastructural analysis of the ⌬popZ Mgr mutant reveals missegregation of cellular content and chemotactic receptor arrays, septum mislocalization, and minicell formation. Prompted by the observed cell elongation and impairment of division, we performed cryo-electron tomography (CET) to further investigate the cell . First row, bright-field; second row, fluorescence channel; third row, overlay of bright-field and GFP channel. PopZ Mgr localizes to both cell poles. In dividing cells, an additional spot appears at the cell division site (fourth and ninth frames, white arrowheads). Generation time during time-lapse series was approximately 4 to 5 h. Numbers indicate hours and minutes. (B) Demograph of cells expressing PopZ Mgr -GFP (n ϭ 642 cells). The appearance of the signal at midcell is marked with an arrowhead. (C) Structured illumination microscopy (3D-SIM) allows us to resolve two PopZ foci in close proximity at the division plane with a distance near the resolution limit of conventional epifluorescence microscopy (ϳ250 nm). Micrographs are maximum-intensity projections of z-stack images from representative FM4-64-stained dividing cells. First row, bright-field (left image) and GFP channel (right image); second row, FM4-64 channel (left image) and overlay of FM4-64 and GFP channel (right image). Insets are magnified xy, xz, and yz projections of PopZ foci (GFP channel) and division plane (FM4-64 channel). Note cells with PopZ foci present at the site of division (Ci) had already completed compartmentalization and separation of their membranes, whereas no foci were observed in cells that were still connected (Cii). A representative number of dividing cells was imaged (n ϭ 29 [additional cells are shown in Fig. S2]). All scale bars not indicated in the figure correspond to 3 m.  confirms the formation of distinct minicells at the poles that were also observed in time-lapse imaging experiments ( Fig. 2C and Movie S1). CET and further segmentation of the cell pole showed the typical arrangement of magnetosomes by the actin-like MamK filament ( Fig. 3Bi and Bii). In addition, a chemoreceptor array was observed at the pole close to the lateral cytoplasmic membrane (inset from Fig. 3Bi and purple objects in Bii), as observed previously in wild-type cells (11). Additionally, a tomographic slice of the cell pole revealed the presence of chemoreceptor arrays also within the minicell (Fig. 3Bi and vi), highlighted in the tomogram segmentation (purple objects in Fig. 3Bii and Biii). Moreover, additional structures inside the minicell, such as magnetosome membrane vesicles, MamK filaments, and ribosomes, were visible ( Fig. 3Biv to vi), indicating putative defects in segregation of cellular content. The presence of chemoreceptor arrays at both the pole and within the adjacent minicell (Fig. 3B), and the observation of double chemoreceptor arrays at the cell poles (see Fig. S3Ai in the supplemental material) argue for an improper polar chemoreceptor array localization (CET micrographs of chemoreceptor arrays in the wild-type strain can be found in reference 11 for comparison). Localization of magnetosome chains did not seem to be affected in the ΔpopZ Mgr strain, as tomographic slices of the cell body showed a properly localized magnetosome chain across the long cell axis (Fig. 3Ci). Remarkably, a deep unidirectional constriction of the membrane located distant from midcell indicated a putative septum mispositioning (black-white arrowheads in Fig. 3A and Cii and white arrowheads in Fig. 3Ciii and Civ). Furthermore, both the misplaced septum invaginations far-off midcell ( Fig. S3Aii and Aiv and Bi, Bii, and Cii) and minicell formation ( Fig. S3Bi to Biii) were commonly observed by CET. Therefore, minicell formation is likely caused by ectopic septum localization, confirming the cell division impairment in the ΔpopZ Mgr mutant. Thus, ΔpopZ Mgr cells are likely unable to properly control the FtsZ ring localization.
Since one of the major functions of PopZ in both C. crescentus and A. tumefaciens is also in regulation of chromosome segregation (17,18,24), we quantified DNA content of ΔpopZ Mgr minicells by staining with DAPI (4=,6-diamidino-2-phenylindole), a dye specific for DNA (Fig. 4A). Remarkably, minicells had an approximately 2.8-fold or 1.5-fold reduced mean cell fluorescence (Fig. 4B) compared to either other cells or their polar regions (to account for possible volume differences of minicells, since cells get thinner toward the poles), further suggesting that PopZ in M. gryphiswaldense also contributes to proper chromosome segregation.
Deletion of popZ Mgr impairs motility and magneto-aerotaxis. In M. gryphiswaldense, magnetotaxis is tightly coupled to aerotaxis in order to govern directed swimming toward optimal low-oxygen levels (13,32). Remarkably, it was also found that cells that perform polar magneto-aerotaxis and display a distinct swimming polarity bias within the magnetic field (preferentially north or south seeking) can be enriched within only few generations (13). Thus, during cytokinesis the proper segregation of magnetosome chains with an inherent physically imprinted magnetic polarity must be coordinated with the determination of the magnetotactic swimming direction. In order to analyze if the ΔpopZ Mgr strain is affected with respect to motility and magnetoaerotaxis, we performed tracking microscopy ( cells of M. gryphiswaldense wild-type, ΔpopZ Mgr , and popZ Mgr ::popZ Mgr -gfp strains measured from time-lapse series. Note that the cell length distribution of the popZ Mgr ::popZ Mgr -gfp strain, as well as growth (not shown), did not differ noticeably from those of the wild type (specific growth rates determined under oxic conditions at 28°C: popZ Mgr ::popZ Mgr -gfp strain, 0.169 Ϯ 0.002 h Ϫ1 ; wild type, 0.171 Ϯ 0.006 h Ϫ1 ). In contrast, for the ΔpopZ Mgr strain, which exhibited cells with high variation in length, a prolonged lag phase and decreased specific growth rate of 0.140 Ϯ 0.021 h Ϫ1 were seen. In box plots, the bar indicates the median, the box the interquartile range, and the whiskers the 5th and 95th percentiles. The mean is shown as ϩ. The number of measured cells (n) and mean and standard deviation (SD) are indicated in the graph. P values were determined by Kruskal-Wallis test with Dunn's multiple-comparison posttest. *, P Ͻ 0.05, **, P Ͻ 0.01, ***, P Ͻ 0.001, ****, P Ͻ 0.0001; ns, not significant (P Ն 0.05).
Pfeiffer et al. PopZ in Magnetospirillum gryphiswaldense ® the wild type ( Fig. 2Aiii and 2iv), but occasionally additional flagella were observed along the cell body ( Fig. 2Aii). Elongated cells of the popZ Mgr deletion strain were motile and despite their highly increased length still aligned to an external magnetic field when observed by dark-field microscopy ( Fig. 5Aii; see Movie S3 in the supplemental material) or optical measurement of their magnetic response (C mag ). In general, the ΔpopZ Mgr population was heterogeneous, consisting of smaller fast-swimming cells and more elongated cells that moved at a lower speed than the wild type (Movie S3 [movies of the wild-type strain can be found in reference 13]). Shorter cells, which had higher swimming speeds and traveled longer distances within the time frame of observation, were observed to swim in circular motions and were not well aligned within the magnetic field ( Fig. 5Aii and Movie S3), which might be due to their altered cell length or the existence of no or only short magnetosome chains. In contrast, longer cells, which displayed low swimming speeds and traveled only short distances, were very well aligned within the magnetic field, presumably due to their overly elongated magnetosome chains. In summary, due to the higher number of highly motile short cells, the mean overall alignment of the ΔpopZ Mgr population was reduced compared to that of the wild type ( Fig. 5Aiii), whereas the swimming speed distribution did not differ significantly from the wild type ( Fig. 5Aiv). Accordingly, ΔpopZ Mgr cultures also reproducibly displayed a slightly lower C mag than the wild type (as determined from at least triplicate cultures [values are given in the legend to Fig. 5Aiii and Aiv]). In contrast to the wild type, the formation of aerotactic swim halos in semisolid medium was almost completely abolished ( Fig. 5B). Moreover, in comparison to the wild type, which forms sharp aerotactic bands in soft agar tubes, the ΔpopZ Mgr mutant grew only in a diffuse zone close to the surface (Fig. 5C). When soft agar assays were performed in the presence of a magnetic field ( Fig. 5D; see Fig. S4C in the supplemental material), spreading of the popZ Mgr deletion mutant parallel to the magnetic field was only observed after prolonged incubation (Ͼ4 days), confirming that cells are still able to align and distribute along the magnetic field lines, but in a slow and possibly only growth-dependent manner. We also failed to restore a swimming polarity bias in elongated ΔpopZ Mgr cells by magnetic selection. Whereas the wild-type and popZ Mgr :: popZ Mgr -gfp strains displayed a south-seeking polarity bias upon repeated passaging in O 2 gradients within a superimposed vertical Southern Hemisphere-like magnetic field, ΔpopZ Mgr cells were rather equally distributed toward both magnetic poles ( Fig. 5D and Fig. S4C). These findings were confirmed by the hanging drop assay (not shown). In general, only few cells of the ΔpopZ Mgr strain accumulated in equal proportions at the northern and southern magnetic pole (facing the air-adjacent borders of the drop), speaking for a general impairment of aerotaxis. In summary, deletion of popZ Mgr severely impaired motility and aerotaxis. Since we observed an increased tendency of highly elongated ΔpopZ Mgr cells to intertwine and aggregate (Fig. 5E), reduced motility in soft agar is also partially explained by the formation of cell clumps. MamK filament dynamics is independent of PopZ Mgr . In M. gryphiswaldense, magnetosome chains are recruited to midcell to ensure equal partitioning of magnetosomes to both daughter cells. It has been recently found that the MamK filament has a particular dynamic behavior, growing from both cell poles, elongating toward midcell, and undergoing treadmilling (7).
Characterization of ΔpopZ Mgr cell motility and magneto-aerotactic behavior. (Ai and ii) Representative tracking results from magnetic nonpolar wild type (Ai; C mag ϭ 1.21, n ϭ 400 tracks) and ΔpopZ Mgr (Aii; C mag ϭ 0.91, n ϭ 49 tracks) cells in a 500-T magnetic field determined at 24°C. "Swimming trajectories" (trajectory length corresponds to the appropriate track length) and "heading" (depicted as colored arrow, where arrow length corresponds to the appropriate object speed) are plotted in a polar coordinate system. The ranges for "line length" and "line speed" (indicated by the inner ring [dark green] and outer ring [light green] of the polar coordinate system) are given below the graphs. Magnetic north corresponds to 0°and magnetic south to 180°, respectively. Note the ΔpopZ Mgr population consisted of small cells that traveled long distances and displayed high swimming speeds, but were not very well aligned within the magnetic field, and slow-swimming elongated cells, which were well aligned. (Aiii and iv) Overall alignment and swimming speed distributions within a 500-T magnetic field for the wild-type and ΔpopZ Mgr strains estimated by motility tracking. Results were pooled from data sets of at least 3 independent experiments (performed on different days), resulting in total numbers of 1,115 tracks for the wild type (mean C mag ϭ 1.2 Ϯ 0.07) and 243 tracks for the ΔpopZ Mgr strain (mean C mag ϭ 0.94 Ϯ 0.03). Box plots are as described in legend to Fig. 2. P values were determined by Mann-Whitney U test (****, P Ͻ 0.0001; ns, not significant [P Ն 0.05]). (B) Aerotactic swim halo formation of M. gryphiswaldense wild-type, ΔpopZ Mgr , and popZ Mgr ::popZ Mgr -gfp strains in 0.2% motility agar 3 days after inoculation and incubation at 28°C and 2% atmospheric oxygen (scale bar ϭ 1 cm). Deletion of popZ Mgr almost completely abolished aerotactic swim ring formation, whereas integration of gfp at the chromosomal popZ Mgr locus did not negatively affect motility. Swim ring diameters were estimated from at least three independent experiments. Values are given as mean ϩ SD. Statistical analysis in panel B was performed as described in the legend to Fig. 2. (C) Aerotactic band formation (marked with an asterisk) in 0.3% soft agar tubes is impaired in the ΔpopZ Mgr strain. (D) Swim ring assay of polar wild-type, ΔpopZ Mgr , and popZ Mgr ::popZ Mgr -gfp strains in the presence of a magnetic field. Cells were grown for several passages under conditions permissive to enrich cells with a south-seeking (SS) swimming polarity bias before the experiment. Six-well plates with 0.2% motility agar were incubated between two strong permanent magnets (ϳ100 mT) for 2 days at 28°C under atmospheric conditions (20% oxygen). Growth-dependent migration along the magnetic field for the ΔpopZ Mgr strain was only observed after additional 3 days of incubation (indicated in the figure with a total incubation time of 5 days). Polar cultures (SS) of the wild-type and popZ Mgr ::popZ Mgr -gfp strains exhibited a biased movement toward the northern magnetic pole (south seeking), which was confirmed by the hanging drop assay (not shown), whereas polar behavior of the ΔpopZ Mgr strain was apparently lost. The direction of the magnetic field is indicated in the figure. Similar results were obtained with a coil setup ( Fig. S4B and C), which provides a weaker and more uniform magnetic field. (E) Cell filamentation of the ΔpopZ Mgr strain raises the tendency of flocculation and cell clumping. Pictures show magnifications of cell aggregates. Scale bars correspond to 100 m (left picture) and 10 m (right picture), respectively. PopZ in Magnetospirillum gryphiswaldense ® Therefore, based on the PopZ Mgr localization pattern, we asked whether PopZ Mgr is involved in or influences the MamK dynamics. To examine this hypothesis, we performed photokinetic analysis of the MamK filament in ΔpopZ Mgr cells. Fluorescence recovery after photobleaching (FRAP) of MamK filaments using an mCherry-MamK fusion showed a half-time fluorescence recovery (t 1/2 ) of 87.6 Ϯ 17.9 s (Fig. 6A). Recently, it was reported that the mCherry-MamK translational fusion expressed in M. gryphiswaldense wild-type cells from a plasmid and chromosomally showed t 1/2 values of 71.8 Ϯ 6.6 and 68.3 Ϯ 4.8 s, respectively (7). A one-way analysis of variance (ANOVA) followed by a Tukey's multiple-comparison test determined that the mCherry-MamK filament t 1/2 in the absence of popZ Mgr is statistically not significant compared to the previously reported values for the wild-type strain (P Ͻ 0.05). Furthermore, the MamK pole-to-midcell growth and its treadmilling behavior are not affected upon absence of popZ Mgr (Fig. 6B). Thus, it can be concluded that the MamK filament dynamics, especially the directed pole-to-midcell growth, is independent of PopZ Mgr . PopZ Mgr forms a polar exclusion zone devoid of macromolecules and chromosomal DNA. To study the effect of popZ Mgr overexpression, PopZ Mgr and PopZ Mgr -GFP were overproduced in trans under the control of P tet (anhydrotetracycline-inducible promoter) in the M. gryphiswaldense wild type and ΔpopZ Mgr strain. Upon reintroduction of popZ Mgr (or popZ Mgr -gfp), cell morphology of the wild type, formation of swim halos, and growth were restored in the ΔpopZ Mgr strain (see Fig. S5 in the supplemental material). Furthermore, prolonged overexpression of popZ Mgr or popZ Mgr -gfp in the wild-type strain caused severe cell filamentation and delayed growth, whereas expression in the ΔpopZ Mgr strain had a lesser effect on growth and cell length, likely due to the absence of endogenous PopZ Mgr (Fig. 6D and Fig. S5). Moreover, PopZ Mgr -GFP overproduction in the M. gryphiswaldense wild type caused either (i) cells with two large polar foci and multiple smaller foci distributed across the cell and in between the PHB granules (stained with the lipophilic dye Nile red, specific for membranes and polyhydroxybutyrate [PHB]) (Fig. 6C, inset) or (ii) cells with a large PopZ accumulation cluster expanding from one pole (Fig. 6C). In the latter, some cells had an additional smaller PopZ Mgr -GFP cluster at the opposite pole. The PopZ expansion zone encompassed several micrometers in length and presented a reduced cell diameter, resulting in a tail-like appearance. Additional staining with DAPI and Nile red revealed that chromosomal DNA and PHB granules were excluded from the expanded PopZ area. Transmission electron microscopy (TEM) analysis confirmed that this zone was depleted of larger cytoplasmic structures such as PHB or polyphosphate granules (Fig. 6D). Furthermore, the brighter appearance indicated that the putative polar PopZ-rich region is mostly devoid of electron-dense cytoplasmic structures and macromolecules (e.g., ribosomes). Even upon PopZ Mgr overexpression, magnetosome chains were still located at midcell, resembling the wild-type phenotype, but in a few cases were also embedded into the outermost part of the PopZ expansion zone. Of note, the formation of flagella at the PopZ-rich poles was not impaired (Fig. 6D). In summary, PopZ Mgr forms a polar expansion zone that is depleted in larger macromolecules and organelles, similar to previously reported observations regarding PopZ in C. crescentus (15,17). Bipolar PopZ Mgr localization requires host-specific factors. In C. crescentus, PopZ (abbreviated PopZ Cc ) first localizes to the old pole and undergoes a transition from monopolar to bipolar after completion of cell division (17). The C. crescentus life cycle is highly asymmetric, generating a smaller and motile swarmer cell and a stalked cell that possesses a tubular extension at the old pole, required for surface attachment. The distinct bipolar localization pattern in M. gryphiswaldense (Fig. 1) prompted us to investigate the localization pattern of PopZ Mgr in C. crescentus ( Fig. 7; see Fig. S6A in the supplemental material). When PopZ Mgr -GFP was heterologously produced in C. crescentus NA1000 (in trans expressed from P tet in the presence of endogenous PopZ Cc ), a unipolar-to-bipolar transition pattern was revealed (Fig. 7A, 1-h time-point and Fig. S6A), similar to the localization pattern of PopZ Cc . PopZ Mgr and PopZ Cc are conserved in their N-and C-terminal regions (37.2% identity and 51.3% similarity, in a global alignment, including some of the most related orthologs [ Fig. S1B to D]), which are known to be important in C. crescentus for interaction with the ParA/ParB chromosome segregation machinery and PopZ cluster formation, respectively. Thus, the observed localization pattern of PopZ Mgr in C. crescentus might be explained by a direct interaction between PopZ Mgr and PopZ Cc and/or with other known PopZ interactors present in C. crescentus, such as ParA/ParB. Upon prolonged expression of popZ Mgr -gfp in C. crescentus, cells became heavily elongated and aberrantly shaped ( Fig. 7A and B), indicating that overexpression of PopZ Mgr also interferes with cell division in C. crescentus. In addition, heterologous popZ Mgr -gfp overproduction in C. crescentus resulted in the appearance of multiple PopZ foci and large polar PopZ exclusion zones (Fig. 7A),   (Continued on next page) Pfeiffer et al. similar to the previously described observations regarding the overproduction of native PopZ Cc in C. crescentus (15,17). As observed for M. gryphiswaldense (Fig. 6C and D), PopZ Mgr exclusion zones in C. crescentus were devoid of DNA and PHB storage granules (Fig. 7C). Heterologous expression of popZ Mgr -gfp in the ΔpopZ Cc background partially restored the cellular morphology of the C. crescentus wild-type strain (Fig. 7A), resulting in a reduced cell length close to wild-type-like levels (Fig. 7B). However, in some cells, we observed PopZ Mgr -GFP foci located at ectopic positions opposite to the stalked pole (Fig. 7A, 12-h time point, yellow arrowheads), indicating that the absence of PopZ Cc was not fully transcomplemented by PopZ Mgr -GFP. In order to avoid artifacts caused by altered expression levels (due to expression from a random ectopic locus under the control of P tet or the presence of endogenous PopZ Cc ), we constructed a C. crescentus strain harboring a site-specific chromosomal replacement of popZ Cc against popZ Mgr -gfp. We also performed the reciprocal experiment and constructed an M. gryphiswaldense strain that carries a site-specific chromosomal replacement of popZ Mgr for an mCherry-popZ Cc fusion. Notably, PopZ Mgr -GFP localized in a monopolar-to-bipolar fashion in C. crescentus (Fig. 7Di and Ei), when expressed from the endogenous popZ Cc -promoter as the sole source of PopZ present. However, transition of PopZ Mgr -GFP to the new poles occurred with ectopic timings, and not all PopZ Mgr was retained in polar regions, as indicated by the diffuse cytoplasmic fluorescence signal ( Fig. 7Di and Ei). In M. gryphiswaldense, mCherry-PopZ Cc in general did localize to both cell poles ( Fig. 7Dii and Eii). However, one cell pole displayed much stronger fluorescence, and filamentous localization patterns were observed (Fig. 7Dii, elongated cell at the bottom left). Furthermore, the appearance of PopZ foci at the new cell poles was delayed, and disappearance of polar foci was observed in some cases (Fig. 7Eii). Together, these results indicated that bipolar PopZ localization is regulated by host-specific proteins in M. gryphiswaldense and that both PopZ orthologs differ to an extent that does not allow full functionality within the cell cycle of the heterologous host. The findings that bipolar PopZ subcellular localization is not inherent to the protein itself but rather host specific were further corroborated by expression of PopZ Mgr -GFP in various other bacterial hosts (as shown in Fig. S6B to E), including only distantly related Escherichia coli and the two other alphaproteobacteria Rhodobacter sphaeroides and Rhodospirillum rubrum, which contain no PopZ (R. sphaeroides) or an endogenous PopZ ortholog (R. rubrum [49% identical to PopZ Mgr ]) (28). Expression of PopZ Mgr -GFP in E. coli WM3064 resulted in the formation of large fluorescent clusters in polar nucleoid-free regions, exhibiting unipolar localization at either the new or old pole (Fig. S6B), similar to the observations made upon expression of PopZ Cc in E. coli (17,18). When we studied the localization of PopZ Mgr -GFP in spheroid 2,6-diaminopimelic acid (DAP)-auxotrophic WM3064 cells formed after depletion of DAP, random foci were formed close to the cell periphery (Fig. S6C), indicating that localization of PopZ Mgr does not depend on geometrical constraints. In ovoid rod-shaped R. sphaeroides cells, PopZ Mgr generally localized only at the old pole (Fig. S6D). A similar bipolar pattern to that in M. gryphiswaldense was observed in Numbers indicate hours and minutes. Note both PopZ orthologs display delayed or erratic polar localization when expressed in the opposing parent strain. (F) Analysis of cell length of C. crescentus (Fi) and M. gryphiswaldense (Fii) strains harboring a chromosomal site-specific replacement of their PopZ ortholog compared to the respective wild-type and ΔpopZ strains. C. crescentus strains were grown in PYE to an OD 600 of ϳ0.17. M. gryphiswaldense strains were grown under microaerobic conditions to an OD 565 of ϳ0.2 in FSM medium. Note both PopZ orthologs are capable to partially replace their reciprocal functions with respect to the cell-length phenotype compared to strains harboring popZ deletions. (G) The C. crescentus wild-type NA1000, ΔpopZ Cc , and popZ Cc ::popZ Mgr -gfp strains were grown in HIGG minimal medium containing 0.1 mM phosphate to an OD 600 of ϳ1.5 for the analysis of stalks. (Gi) Epifluorescence micrographs of all three strains stained with FM4-64 and DAPI. Note since FM4-64 did not reliably stain stalks of all cells, DIC microscopy and DAPI staining (which unspecifically stains stalks) were included to identify and measure stalks. (Gii and Giii) Analysis of (Gii) stalk length and (Giii) stalk number (given as percentage of cells with no stalks or mono-or bipolar stalks). Note PopZ Mgr is able to partially complement the function of PopZ Cc with respect to proper stalk formation. Measurements shown in panel B were taken from one representative induction experiment. Experiments shown in panels F and G were performed in biological triplicates (independent cultures). The total number of analyzed cells (n) is indicated in the graphs. Box plots and statistical analysis are similar as described in the legend to Fig. 2. Epifluorescence micrographs shown in panels C to E and G are maximum-intensity projections of deconvolved z-stacks. Fluorescence channels are indicated in the graph. Representative stalks and cell division events are exemplarily marked with yellow and white arrowheads, respectively. All scale bars ϭ 3 m.

PopZ in Magnetospirillum gryphiswaldense
® spirillum-shaped R. rubrum cells, with two new PopZ foci emerging at the site of cell division (Fig. S6E). In summary, these results suggest that monopolar accumulation in DNA-free polar regions occurs by a mechanism that is inherent to PopZ Mgr , whereas bipolar localization apparently depends on distinct alphaproteobacterium-specific host factors.
To further investigate whether PopZ orthologs can replace their functionalities, we compared median cell lengths of M. gryphiswaldense and C. crescentus strains harboring reciprocal PopZ orthologs with the respective wild-type and ΔpopZ strains. The median cell length of the C. crescentus popZ Cc ::popZ Mgr -gfp strain was 1.5-fold reduced compared to the ΔpopZ Cc strain (and 1.2-fold higher than that of the NA1000 wild-type strain [ Fig. 7Fi]), whereas the median cell length of the M. gryphiswaldense popZ Mgr :: mCherry-popZ Cc strain was 1.4-fold lower than that of the ΔpopZ Mgr strain (but only 1.05-fold larger than that of the M. gryphiswaldense wild-type strain [ Fig. 7Fii]). These results indicated that in both strains, the loss of the respective PopZ ortholog can be partially rescued by expression of the reciprocal ortholog.
To analyze whether PopZ Mgr is capable to accomplish functions with respect to stalk formation, which is impaired in the C. crescentus popZ deletion strain (17,18), we also analyzed stalk length and frequency in cells grown under phosphate-limiting conditions, known to cause severe stalk elongation (33), to facilitate detection and analysis of stalks (Fig. 7G). In contrast to previous reports that the ΔpopZ Cc strain does not form stalks (3,21), we found that the ΔpopZ Cc strain grown under phosphate starvation is still able to form stalks, but of 2.5-fold-reduced median length (Fig. 7Gii) and at a lower frequency (Fig. 7Giii) than the NA1000 wild-type strain. In comparison, the fraction of cells with monopolar stalks was 1.6-fold lower in the ΔpopZ Cc mutant than in the NA1000 wild-type strain. In contrast, the fraction of cells without stalks was 2.3-fold reduced in the popZ Cc ::popZ Mgr -gfp strain compared to the ΔpopZ Cc strain, whereas the number of cells with monopolar stalks was 1.9-fold higher, bringing both values closer to wild-type levels. A minor fraction of cells with bipolar stalks was detected within the ΔpopZ Cc and popZ Cc ::popZ Mgr -gfp populations, whereas no cells containing bipolar stalks were found in the NA1000 wild type. Median stalk lengths of the popZ Cc ::popZ Mgrgfp strain were still 1.4-fold lower relative to the wild type but 1.8-fold larger in comparison to the ΔpopZ Cc strain. Notably, stalk formation was also restored upon expression of popZ Mgr -gfp in the ΔpopZ Cc strain from P tet (Fig. 7A, 8 and 12 h, yellow arrowheads), and overexpression of PopZ Mgr -GFP in NA1000 caused aberrant bipolar stalk formation (Fig. 7C, yellow arrowheads). In summary, these results argue that PopZ Mgr is able to partially accomplish functions inherent to PopZ Cc with respect to stalk formation in C. crescentus.

DISCUSSION
In C. crescentus, PopZ has been described as an important landmark protein, generating a polar hub domain for multiple proteins involved in cell cycle control and polar morphogenesis (14-18, 34, 35). In addition to C. crescentus, PopZ has been studied in A. tumefaciens (23)(24)(25)(26), which exhibits unipolar growth by addition of peptidoglycan at the new "growth pole" (22). Here, we report that PopZ in the magnetotactic model organism M. gryphiswaldense plays a similar, but somewhat distinct role. In contrast to C. crescentus and A. tumefaciens, where cell division results in morphologically distinct cells and/or daughter cells that differ in cell cycle progression, division in M. gryphiswaldense gives rise to morphologically nearly equal daughter cells. Deletion and overexpression of popZ in M. gryphiswaldense resulted in severe cell division defects (Fig. 2, Fig. 3, Fig. S3, Movie S1, Movie S2, and Fig. 6C and D and Fig. S5A and B, respectively) and DNA missegregation (Fig. 4), consistent with previous observations in C. crescentus (17,18) and A. tumefaciens (24,25). However, we did not observe formation of ectopic poles and cell branching as in A. tumefaciens (25,26). In accordance with reported results in C. crescentus (15,17), we have observed formation of large exclusion zones upon overproduction of PopZ Mgr in M. gryphiswaldense (Fig. 6C  and D). These results imply that PopZ Mgr may have an important role as a putative landmark protein and in the control of cell-cycle-related factors in M. gryphiswaldense. As for now, it can only be speculated that the severe cell elongation and minicell formation of the ΔpopZ Mgr strain are due to an indirect impairment in FtsZ ring positioning. In C. crescentus, MipZ inhibits FtsZ polymerization by generating a gradient with the highest concentration in polar regions via ParB-PopZ-dependent retention of MipZ (19,20), thus creating a region with the lowest MipZ concentration at midcell with suitable conditions for FtsZ ring positioning and formation. Since orthologs of the ParA/ParB chromosome segregation system and MipZ spatial regulator are present in M. gryphiswaldense, it is likely that PopZ Mgr contributes to stabilization of the MipZ gradient and, thereby, proper placement of the division site. However, the specific functions of ParA, ParB, and MipZ in M. gryphiswaldense remain to be elucidated.
Deletion of popZ in M. gryphiswaldense severely affected motility and apparently polar magneto-aerotactic behavior (Fig. 5, Fig. S4, and Movie S3). Inheritance of a specific magnetotactic pole-seeking polarity was hypothesized to rely on a yet elusive superimposed mechanism of cellular polarity control, by defining a cellular polarity axis in addition to the magnetosome chain's magnetic dipole (8,13). However, the affected motility and loss of swimming polarity are supposedly not directly caused by the absence of popZ Mgr . In contrast, the aforementioned phenotypes are likely explained by a general impairment of aerotaxis in the ΔpopZ Mgr strain (i.e., due to improper localization of motility-related structures, as discussed below) and as an indirect effect due to formation of short cells that are highly motile, but only weakly aligned within the magnetic field, as well as severe cell elongation, which affects hydrodynamic properties of cells' propulsion during swimming (as also previously observed for artificially elongated cells caused by cephalexin treatment [6]). Furthermore, we observed an increased tendency of elongated ΔpopZ Mgr cells to form aggregates (Fig. 5E), which might contribute to the strong motility phenotype observed in soft-agar-based assays (Fig. 5B, C, and D). An increased tendency of ΔpopZ cells to aggregate has also been observed in A. tumefaciens and might be caused by an altered formation of extracellular polysaccharides (25), but in the case of M. gryphiswaldense also due to the helical nature of intertwined elongated cells. It can be further hypothesized that the disturbed aerotactic behavior in the ΔpopZ Mgr strain may be due to a delayed or impaired signal transduction from the chemotactic machinery to the flagellar motors, since some cells contained improperly placed (Fig. 3B) or additional chemosensory clusters (as confirmed by fluorescence microscopy of various methyl-accepting chemotaxis proteins [MCPs] fused to GFP in the ΔpopZ Mgr parent strain [results not shown]) as well as occasional flagella located in nonpolar regions (Fig. 2Aii). Altered localization of MCPs, chemoreceptor-associated histidine kinase CheA, and flagellar basal body proteins FliG and FliM upon popZ deletion has been also reported for C. crescentus (17) and A. tumefaciens (25) or in artificially elongated cephalexin-treated E. coli cells (36,37). However, only a mild effect on motility in swim plate assays has been observed upon popZ deletion in A. tumefaciens (25), and artificially elongated E. coli cells were only affected in their swimming speed, but were still able to perform chemotaxis (36). Hence, due to their different flagellation patterns, cell shapes, and chemotactic behaviors, the experimental results among different strains are not directly comparable.
In addition, severe cell elongation upon popZ deletion in M. gryphiswaldense resulted in drastically elongated magnetosome chains and a highly increased number of particles per cell ( Fig. 2A and B). Our results imply that magnetosome number and chain length are likely directly related to cell length, resembling previously published observations on artificially elongated cephalexin-treated cells (6). Besides, it has recently been shown that increased gene dosage by genomic multiplication of the magnetosome island results in increased particle numbers as well (38), but with several chains running in parallel or cells closely packed with magnetosomes that lack an ordered chain-like arrangement. Presumably, elongated ΔpopZ Mgr cells also possess an increased number of gene copies due to the presence of multiple chromosomes (albeit we were not able to identify distinct individual chromosomes by DAPI staining, without any specific treatment to condense DNA [ Fig. 4A]). However, in contrast to the overproducer strain (38), the amount of magnetosomes and gene copies per cell volume in elongated ΔpopZ Mgr cells can be assumed to be roughly in the same range as for the wild type.
Magnetosome synthesis, midcell positioning and proper segregation of magnetosome chains are controlled by the treadmilling behavior of the actin-like MamK, which forms dynamic filaments (6,7). MamK-dependent repositioning of magnetosome chains was not affected in the ΔpopZ Mgr strain ( Fig. 2A and Fig. 6A and B), suggesting that PopZ Mgr does not play a role in magnetosome organelle segregation or positioning by exerting direct control of the MamK dynamics. Since magnetosome chain segregation is tightly coupled to cell division, it can be hypothesized that PopZ Mgr may influence magnetosome segregation indirectly-likely by regulating the FtsZ ring localization. Thus, lack of PopZ Mgr causes unequal cell division and misdistribution of chains during cell division as a side effect.
Most strikingly, a consistent bipolar localization pattern of PopZ in M. gryphiswaldense was observed (Fig. 1, Fig. S2, and Movie S1), contrasting with the reported monopolar-to-bipolar transition in C. crescentus (17,18) and unipolar localization in A. tumefaciens (23)(24)(25). Ortholog substitution experiments between C. crescentus and M. gryphiswaldense ( Fig. 7 and Fig. S6) indicated that bipolar PopZ localization is not inherent to the protein itself but rather is host specific. PopZ Mgr and PopZ Cc have conserved N and C termini (Fig. S1) and were capable of partially substituting their reciprocal functionalities (Fig. 7). Thus, the observed localization pattern of both orthologs may be explained by a direct interaction with the respective PopZ interactors present in each host, albeit our results also indicate that both orthologs have diverged to an extent that does not allow full conservation of all PopZ-dependent interactions. For C. crescentus, several factors for control of PopZ localization have been discussed (14,15,34,39). Polar localization of PopZ relies on its self-assembly into higher-order structures in DNA-free polar regions, and the unipolar-to-bipolar transition is coupled to the asymmetric distribution of ParA during the cell cycle (14). The chromosome segregation system adaptor protein ParB and the ParA ATPase, which act together to spatially separate replicated chromosomes in C. crescentus (1), might be suitable candidates for control of bipolar PopZ localization in M. gryphiswaldense. Recently, the zinc finger protein ZitP (28,40) and muramidase homolog SpmX (41) have been described as additional important factors to nucleate new PopZ microdomains in C. crescentus. An ortholog of ZitP (locus tag MGR_3358) is also encoded in the M. gryphiswaldense genome (23% identity and 39% similarity compared to ZitP Cc , respectively), whereas no protein orthologous to SpmX is present. Further investigation is needed to identify PopZ interactors in M. gryphiswaldense and elucidate how they differ in function from those of other alphaproteobacteria.
In conclusion, protein functions depend on the genetic context, and can be implemented in different ways, even in closely related species. Thus, M. gryphiswaldense also serves as an appropriate and interesting model organism to study the function of cell cycle factors and its coordination with organelle synthesis and segregation. In the near future, these cell-cycle-related studies will also help to understand how polar magnetotaxis is functionally controlled and inherited in MTB.

MATERIALS AND METHODS
Bacterial strains, plasmids, and culture conditions. Bacterial strains and plasmids used in this study are listed in Text S1 in the supplemental material. Cells were grown using previously described standard procedures described in detail in Text S1.
Molecular and genetic techniques. Oligonucleotides (sequences are listed in Text S1) were purchased from Sigma-Aldrich (Steinheim, Germany). Plasmids were constructed by standard recombinant techniques (as described in Text S1), employing a homologous recombination-based counterselectable system for the construction of in-site deletion and insertion mutants (31) and a Tn5-based anhydrotetracycline-inducible expression vector (29,42) for the construction of transcomplementation and overexpression constructs. All constructs were sequenced by Macrogen Europe (Amsterdam, Netherlands).
Motility assay. Motility soft agar assays were performed as described by Popp et al. (13) and in Text S1. Single-cell tracking was performed at 24°C on a Nikon FN1 Eclipse microscope (Fig. S4A) equipped with an S Plan Fluor 20ϫ differential inference contrast (DIC) N1 objective (NA0.5), a dark-field condenser