Spatiotemporal Organization of Chemotaxis Pathways in Magnetospirillum gryphiswaldense

Magnetotactic bacteria (MTB) use the geomagnetic field for navigation in aquatic redox gradients. However, the highly complex signal transduction networks in these environmental microbes are poorly understood. Here, we analyzed the localization of selected chemotaxis proteins to spatially and temporally resolve chemotaxis array localization in Magnetospirillum gryphiswaldense. Our findings suggest that bipolar localization of chemotaxis arrays related to the key signaling pathway CheOp1 is important for aerotaxis and that CheOp1 signaling units assemble independent of the three other chemotaxis pathways present in M. gryphiswaldense. Overall, our results provide deeper insights into the complex organization of signaling pathways in MTB and add to the general understanding of environmental bacteria possessing multiple chemotaxis pathways.

ABSTRACT

Magnetospirillum gryphiswaldense employs iron-rich nanoparticles for magnetic navigation within environmental redox gradients. This behavior termed magneto-aerotaxis was previously shown to rely on the sensory pathway CheOp1, but the precise localization of CheOp1-related chemoreceptor arrays during the cell cycle and its possible interconnection with three other chemotaxis pathways have remained unstudied. Here, we analyzed the localization of chemoreceptor-associated adaptor protein CheW₁ and histidine kinase CheA₁ by superresolution microscopy in a spatiotemporal manner. CheW₁ localized in dynamic clusters that undergo occasional segregation and fusion events at lateral sites of both cell poles. Newly formed smaller clusters originating at midcell before completion of cytokinesis were found to grow in size during the cell cycle. Bipolar CheA₁ localization and formation of aerotactic swim halos were affected depending on the fluorescent protein tag, indicating that CheA₁ localization is important for aerotaxis. Furthermore, polar CheW₁ localization was independent of cheOp2 to cheOp4 but lost in the absence of cheOp1 or cheA₁. Results were corroborated by the detection of a direct protein interaction between CheA₁ and CheW₁ and by the observation that cheOp2- and cheOp3-encoded CheW paralogs localized in spatially distinct smaller clusters at the cell boundary. Although the findings of a minor aerotaxis-related CheOp4 phenotype and weak protein interactions between CheOp1 and CheOp4 by two-hybrid analysis implied that CheW₁ and CheW₄ might be part of the same chemoreceptor array, CheW₄ was localized in spatially distinct polar-lateral arrays independent of CheOp1, suggesting that CheOp1 and CheOp4 are also not connected at the molecular level.

IMPORTANCE Magnetotactic bacteria (MTB) use the geomagnetic field for navigation in aquatic redox gradients. However, the highly complex signal transduction networks in these environmental microbes are poorly understood. Here, we analyzed the localization of selected chemotaxis proteins to spatially and temporally resolve chemotaxis array localization in Magnetospirillum gryphiswaldense. Our findings suggest that bipolar localization of chemotaxis arrays related to the key signaling pathway CheOp1 is important for aerotaxis and that CheOp1 signaling units assemble independent of the three other chemotaxis pathways present in M. gryphiswaldense. Overall, our results provide deeper insights into the complex organization of signaling pathways in MTB and add to the general understanding of environmental bacteria possessing multiple chemotaxis pathways.

INTRODUCTION

The alphaproteobacterium Magnetospirillum gryphiswaldense employs membrane-enveloped magnetite (Fe₃O₄) nanoparticles (magnetosomes) organized in intracellular chains for magnetic alignment and navigation within the Earth’s magnetic field (magnetotaxis). In contrast to Escherichia coli, one of the most well-studied model bacteria for motility and chemotaxis, which possesses flagella distributed over the entire cell body (peritrichous flagellation) and uses tumbles to reorient (1), M. gryphiswaldense swims by means of polar flagella (commonly one flagellum at each cell pole) (2). Swimming trajectories of M. gryphiswaldense are characterized by long runs interrupted by reversal events resulting in an immediate directional change of almost 180° (3, 4). This motility pattern suits the distinct lifestyle of magnetotactic bacteria (MTB), which use the geomagnetic field to restrict their flagellum-mediated movement into quasi one-dimensional swimming up- and downward along oxygen (or other redox) gradients (5). Magnetotaxis in M. gryphiswaldense is closely associated with aerotaxis (i.e., magneto-aerotaxis [MTX]), characterized by the formation of sharp aerotactic bands within oxygen gradients (3, 4) and is believed to facilitate the orientation in aquatic freshwater gradients (5).

At the molecular level, aerotaxis in M. gryphiswaldense (which might be part of a general metabolism-dependent energy taxis) is mediated by the signal transduction pathway carried in chemotaxis operon 1 (cheOp1), which contains the canonical set of chemotaxis genes cheAWYBR (3). Deletion of cheOp1 resulted in a null phenotype for aerotaxis, that is, loss of a detectable response to oxygen and the inability to form aerotactic bands in glass microcapillaries (3). Remarkably, CheOp1-mediated aerotactic signaling was essential for polar MTX, i.e., north- or south-seeking swimming polarity, characterized by the preferential migration toward one of the magnetic poles (3).

The individual roles of all chemotaxis proteins encoded in cheOp1 have not been analyzed in detail, but based on the knowledge of well-studied chemotaxis models, such as E. coli, a chemotactic response is mediated via a stimulus perceived by methyl-accepting chemotaxis proteins (MCPs), resulting in autophosphorylation of the chemoreceptor-associated histidine kinase CheA and subsequent phosphotransfer to the diffusible response regulator CheY (1, 6). CheY in its phosphorylated state binds to the flagellar motor to trigger a switch in its rotational direction. MCPs are arranged in organized array structures, formed by MCP receptor trimers of dimers, and stabilized by baseplate interactions of CheA and the adaptor protein CheW (7). Various factors controlling array formation and (often polar) intracellular localization have been discussed, including stochastic self-assembly and membrane curvature, but also landmark and cell cycle-related proteins, depending on the genetic background and type of chemoreceptor arrays (i.e., membrane associated or cytoplasmic) (8, 9).

In M. gryphiswaldense, three additional che pathways (cheOp2, cheOp3, and cheOp4) are present apart from cheOp1, as well as putative chemotaxis-related proteins encoded outside cheOp1 to cheOp4 (e.g., multiple additional CheY paralogs and ∼56 genes encoding MCPs) (3). In comparison to the canonical gene set cheAWYBR in cheOp1, additional chemotaxis-related proteins are encoded within cheOp2 to cheOp4, such as MCPs, sensory transduction histidine kinases, a ParA-like ATPase (protein involved in the intracellular positioning of cytoplasmic chemotaxis-arrays) (6, 8, 9), and EAL/GGDEF-domain containing proteins related to the second messenger cyclic diguanylate (c-di-GMP) (see Fig. S1 in the supplemental material). Individual deletions of cheOp2 to cheOp4 had no obvious (cheOp2/4) or only a slight (cheOp3) impact on aerotactic swim halo formation, suggesting that CheOp1 is the major pathway to control swimming reversals and aerotaxis in M. gryphiswaldense (3).

In M. gryphiswaldense, chemotaxis arrays (of unknown molecular composition) were observed at polar-lateral sites by cryo‐electron tomography (CET) (10). Furthermore, a CheW₁-green fluorescent protein (GFP) fluorescent fusion localized to both cell poles, with two new foci emerging at midcell before cell division (11). cheW₁ deletion resulted in cells incapable of reacting to changes in oxygen level, whereas the sequestration of CheW₁ away from polar chemoreceptor arrays by an intracellular nanotrap caused a gradual knockdown of aerotaxis (11), suggesting that CheW₁ is required to maintain the functional integrity of CheOp1-related chemoreceptor arrays. However, the detailed cell cycle-dependent localization, formation, and inheritance of CheOp1-related chemotaxis clusters, as well as the role of CheA₁, have not been analyzed. In addition, it remained unknown whether CheOp2 to CheOp4-related chemotaxis arrays localize to distinct subcellular sites or if they are spatially and functionally linked with CheOp1.

Here, we employed three-dimensional structured illumination microscopy (3D-SIM) to analyze the cell cycle-dependent localization of CheOp1-related chemotaxis clusters (by means of CheW₁ and CheA₁ localization) with high spatiotemporal resolution. Our results suggest that CheOp1-related signaling units are predominantly located at polar-lateral sites of both cell poles and that their bipolar localization is important for aerotaxis. Furthermore, we provide evidence that CheOp1-related chemoreceptor arrays assemble in the absence of cheOp2 to cheOp4 and at spatially distinct subcellular sites, highlighting its role as an independent major pathway for aerotaxis in M. gryphiswaldense.

RESULTS

CheOp1-related chemotaxis arrays localize at both cell poles.

To perform an in-depth analysis of the cell cycle-dependent localization of CheOp1-related chemotaxis arrays, we used a previously generated strain expressing a chromosomally encoded functional CheW₁-GFP fusion (11). 3D-SIM revealed clusters located at both cell poles, often at polar-lateral sites, as well as new clusters formed at midcell before division (in agreement with reference 11) (Fig. 1A and B). Large clusters were mostly observed at the old poles, whereas clusters located at midcell were smaller, suggesting that the size of newborn clusters increases during the cell cycle (Fig. 1C and D). We occasionally observed clusters located at subpolar positions, which on average were smaller than clusters at the old cell poles and midcell (Fig. 1C and D), suggesting that those were not formed in the same temporal order as midcell clusters. The observed distances of CheOp1-related chemotaxis clusters were >1 μm (median, 3.1 μm) and up to 6 μm with increasing cell lengths (Fig. 1E). Time-lapse microscopy confirmed that clusters formed at midcell in predivisional cells grew in size and further revealed that polar clusters are dynamic and apparently undergo occasional segregation and fusion events (Fig. 1F; see Movie S1 in the supplemental material), which also have been observed in other Alphaproteobacteria (12).

FIG 1.

Cell cycle-dependent localization of CheW₁. (A) Representative cell (outlined by purple dashed lines) expressing CheW₁-GFP from the native chromosomal locus (cheW₁::cheW₁-gfp) with (sub)polar-lateral and midcell clusters. The main micrograph is a 3D-SIM z-stack maximum intensity projection. The calibration bar denotes the intensity of fluorescence. Insets depict an overlay of brightfield and fluorescence channels and z-cross sections (100-nm scale bars) at positions indicated by white dashed lines. (B) 3D-SIM demographic analysis of CheW₁-GFP localization (n = 471 cells). To generate the graph, single-cell fluorescence profiles were normalized and stacked on top of each other sorted by cell length. (C) Distribution of detected fluorescent maxima (n = 1,384 maxima in 471 cells) relative to cell length. Based on their area, fluorescent foci were sorted into 4 groups, from small to large, colored in dark green, light green, light magenta, and light blue, respectively. The dot size reflects the size group. (Di) Areas of maxima plotted against their relative position. Maxima were sorted into three groups based on their relative positions. Extreme values are clipped to better indicate distribution of the data (n = 1,378 maxima are shown in the graph). (Dii) Maximum and minimum Feret diameters of the fluorescent maxima groups shown in panel Di (n = 218, 85, and 1,081 maxima in groups 1, 2, and 3, respectively), and overall maximum and minimum Feret diameter of all fluorescent maxima (n = 1,384 maxima). Violin plots depict the frequency distribution of the data. The dashed line depicts the median, and dotted lines depict the quartiles. (E) Mean spacing between fluorescent maxima relative to the cell length. Dots are colored according to the total number of maxima present per cell (as indicated in the key). The inset figure (violin plot) depicts the overall distribution of mean maxima spacing estimates. Cells with less than 2 maxima were excluded since no spacing distance can be calculated. Estimates of n = 457 cells are shown. The data shown in panels B to E were derived from three different experiments. (F) 3D-SIM time-lapse microscopy of cells expressing CheW₁-GFP. White arrowheads indicate cell division events. Two foci that appeared at midcell in slice 2 are marked throughout the time lapse by magenta and yellow symbols, respectively. Double arrowheads indicate the appearance of fluorescent foci by separation from polar clusters. Time values are given in h and min.

Bipolar CheA₁ localization is beneficial for aerotaxis.

Various attempts to label CheA₁ with GFP resulted in nonfunctional fusion proteins (i.e., nonaerotactic cells) which localized to only one cell pole (results not shown), which we hypothesized might be related to the tendency of enhanced GFP (eGFP) to oligomerize (13), possibly interfering with the formation of ordered chemotaxis arrays or functionality of CheA₁. In contrast, chromosomally encoded fusions of the monomeric fluorescent proteins mCherry (14) and mTurquoise2 (mTurq2) (15) to the N terminus of CheA₁ provided functionality when present as the sole source of CheA₁, as indicated by the formation of wild-type-like swim halos if the respective strains were compared to nonaerotactic ΔcheOp1 (3) and ΔcheA₁ strains, which did not form a clearly apparent aerotactic swim halo in motility soft agar after 3 days of incubation (Fig. 2Ai). However, aerotactic efficiency (as measured by swim halo diameter) of mCherry- and mTurq2-CheA₁-expressing cells was decreased to 67% and 53% of the wild type, respectively (Fig. 2Aii), suggesting that both fluorescent proteins partially interfere with CheA₁ localization and/or signaling. Remarkably, both fluorescent fusions localized in divergent patterns (Fig. 2B and C); similar to the CheW₁-GFP fusion (Fig. 1), mCherry-CheA₁ was localized at both cell poles (61% of cells) or in a slightly asymmetric bipolar pattern (22% of cells), and additional foci at midcell were observed in 13% of cells. In contrast, mTurq2-CheA₁ predominantly localized in a monopolar (53% of cells) or highly asymmetric bipolar fashion (i.e., one cell pole exhibiting a much stronger fluorescence) (in 32% of cells), and fluorescent foci at midcell were rarely observed, indicating that the formation of new chemoreceptor arrays is delayed in mTurq2-CheA₁-expressing cells and occurs mostly postdivision at the new cell poles. The monopolar or highly asymmetric bipolar pattern and stronger reduction in swim halo diameters of mTurq2-CheA₁-expressing cells suggested that mTurq2 had a stronger negative impact on CheA₁ functionality and localization than mCherry and that bipolar CheA₁ localization is important for efficient aerotaxis in M. gryphiswaldense.

FIG 2.

Fluorescent labeling of CheA₁ affects aerotaxis and CheA₁ localization. (Ai) Representative swim halos of the wild type (WT) and of strains expressing chromosomally encoded fusions of mCherry-CheA₁ (cheA₁::mCherry-cheA₁) and mTurq2-CheA₁ (cheA₁::mTurq2-cheA₁) after 3 days. For comparison, nonaerotactic ΔcheA₁ and ΔcheOp1 strains are shown. (Aii) Corresponding swim halo diameters (mean + SD; n = 3 analyzed cultures per strain). A statistical comparison was performed by analysis of variance (ANOVA) with Tukey’s multiple-comparison test. ****, P < 0.0001; ns, not significant [P ≥ 0.05]. (Bi) Localization of mCherry-CheA₁ (3D-SIM) and mTurq2-CheA₁ (conventional epifluorescence). Shown is a merge of fluorescence and defocused brightfield channels. (Bii) Representative dividing cell outlined by green dashed lines with mCherry-CheA₁ foci at midcell (3D-SIM). The calibration bar denotes the intensity of fluorescence. The inset at the top left corner is an overlay of the brightfield and fluorescence channels. Additional insets indicated by white dashed lines depict z-cross sections (100-nm scale bars). All other fluorescence micrographs in panel B are z-stack maximum intensity projections. (C) Distribution of fluorescent foci in mCherry-CheA₁- and mTurq2-CheA₁-expressing cells. Colored dots indicate the outcome of n = 3 individual experiments. Bars denote the mean, and error bars indicate standard deviation. For the individual experiments, n = 95, 104, and 111 (mCherry-CheA₁) and 154, 133, and 198 (mTurq2-CheA₁) cells were analyzed.

Polar localization of CheW₁ is independent of cheOp2 to cheOp4.

To analyze if chemoreceptor arrays related to the major chemotaxis pathway CheOp1 assemble independently of CheOp2 to CheOp4, we investigated whether CheW₁ localization depends on Che proteins encoded by cheOp1 to cheOp4. A CheW₁-GFP fusion expressed in trans from a constitutive promoter did localize to the cell poles in both the wild type and a ΔcheW₁ strain (11) (Fig. 3A), indicating that targeting of CheW₁-GFP to chemoreceptor arrays does not require the presence of native untagged CheW₁ (in agreement with reference 11). Polar localization of CheW₁-GFP was lost in ΔcheOp1 and ΔcheA₁ strains, instead resulting in an almost homogenous distribution of fluorescence, except of intracellular storage granules, indicating that CheA₁ is required to form stable polar chemoreceptor arrays together with CheW₁. In contrast, polar localization of CheW₁-GFP was retained in a strain lacking cheOp2, cheOp3, and cheOp4 (i.e., the ΔcheOp2-4 strain) (3), revealing that none of the respective histidine kinases (CheA₂, CheA₃, and CheA₄), adaptor proteins (CheW_2.1, CheW_2.2, CheW_3.1, CheW_3.2, and CheW₄), or MCPs encoded in cheOp2 to cheOp4 are required for polar localization of CheW₁.

FIG 3.

Localization of CheW paralogs in the wild-type (WT) and different mutant backgrounds. (A) Polar localization of CheW₁-GFP is lost in the absence of cheOp1 or cheA₁ but not in a ΔcheOp2-4 strain. (B) Localization of cheOp2-, cheOp3-, and cheOp4-encoded CheW proteins in the WT, ΔcheOp1, and ΔcheOp2-4 background. All fluorescent fusions were expressed from the constitutive P_mamDC45 promoter after a single-copy chromosomal insertion of a Tn5-based expression cassette in the respective strains. Note, the CheW_3.1-GFP fusion was only poorly expressed, resulting in low quality of reconstructed 3D-SIM images. Fluorescent foci are exemplary marked by white arrowheads. Scale bars for the 3D-SIM and brightfield channel inset micrographs correspond to 1 μm.

Since the deletion of cheOp3 previously caused a slight reduction in aerotactic swim halo diameter (3), we further analyzed whether selected proteins encoded within cheOp1 and cheOp3 interact with each other. Bacterial two-hybrid analysis (BACTH) (16, 17) revealed possible interactions between CheA₁ and CheW₁, and between CheA₃ and CheW_3.1, as well as homo-oligomerization of CheA₁, CheA₃, and the GGDEF-domain-containing CheY response regulator encoded in cheOp3 (Fig. 4A). We failed to clearly detect hypothetical CheA-CheY interactions within individual CheOps, presumably since those are likely rather transient. No interactions between cheOp1- and cheOp3-encoded proteins were observed, providing evidence that both pathways are not connected at the molecular level. Furthermore, fluorescent fusions of CheW_3.1 and CheW_3.2, as well as CheW_2.1 and CheW_2.2, expressed in the wild-type background were found to localize not only at the poles but also mostly in small clusters close to the cell boundary (Fig. 3B), suggesting that cheOp2- and cheOp3-encoded CheW paralogs are present in spatially distinct arrays, contrasting with the predominant polar-lateral localization of CheW₁-GFP (Fig. 1A). The signal of a fluorescent fusion of the ParA-like protein encoded in cheOp2 was found to condense into distinct, equidistant foci when cells were costained with 4′,6-diamidino-2-phenylindole (DAPI) (see Fig. S3A in the supplemental material), indicating that it might localize in a DNA-dependent manner. In the case of CheW_3.1-GFP, distinct foci were observed only in a few cells, suggesting that this fusion is only poorly expressed and/or available binding sites are limited, likely due to very low expression levels of other cheOp3-encoded proteins (3) (see also Fig. S4 in the supplemental material). Similarly, distinct foci of a MCP_Op3-GFP fusion expressed from the native chromosomal cheOp3 locus (Fig. S3B) were only observed in 8.3% of cells (n = 1,113 cells), providing further evidence that cheOp3-encoded proteins are expressed only at a very low level under the chosen experimental conditions or require specific environmental stimuli to localize at the cell boundary (18). cheOp2- and cheOp3-encoded CheW paralogs localized at the cell boundary independent of cheOp1 (Fig. 3B), further suggesting that they do not form mixed clusters with cheOp1-encoded base plate-forming proteins. In contrast, localization in distinct foci was lost in the ΔcheOp2-4 background (resulting in an almost homogenous fluorescence). Few occasional foci of CheW_2.2- and CheW_3.2-GFP were still observed in some ΔcheOp2-4 cells (Fig. 3B), indicating that both fusion proteins have a weak tendency to aggregate.

FIG 4.

Two-hybrid analysis of selected chemotaxis proteins. (A) CheOp1 and CheOp3. (B) CheOp1 and CheOp4. E. coli BTH101 cells expressing protein fusions to the Bordetella pertussis adenylate cyclase catalytic domain complementary fragments T18 and T25 were spotted onto M63 mineral salt agar. Negative controls (T18 or T25 fusions with the corresponding empty vector) are shown in the first row(s) and column(s) at the very top and left of the panels, respectively. The positive control (leucine zipper) and corresponding negative controls (pUT18C-zip, pKT25; pUT18C, and pKT25-zip) are marked by black boxes.

Remarkably, a CheW₄-GFP fusion did localize in a similar pattern as CheW₁-GFP toward midcell and polar-lateral sites (Fig. 3B). BACTH revealed possible interactions of weak-to-moderate strength (considering the background level of the negative controls and strength of the CheA₁-CheW₁ interaction) between CheA₄ and CheW₄, as well as CheA₄ and CheW₄ self-interactions, and interpathway interactions between CheA₁ and CheW₄ and between CheA₄ and CheW₁ (Fig. 4B). However, polar localization of CheW₄-GFP was not obviously affected in the ΔcheOp1 strain but was entirely lost in the absence of cheOp2 to cheOp4 (Fig. 3B) and in a ΔcheOp4 strain (3) (not shown in the figure), suggesting that in M. gryphiswaldense, CheW₁ and CheW₄ predominantly interact with their respective CheA counterparts encoded within the same che operon. To finally prove whether CheW₁ and CheW₄ localize to spatially distinct or similar polar-lateral arrays, we coexpressed CheW₁-GFP and CheW₄-mCherry fluorescent fusions and imaged cells by 3D-SIM (Fig. 5). In agreement with the localization of CheW₁ and CheW₄ in different cheOp-deletion backgrounds (Fig. 3), colocalization analysis revealed that CheW₁ and CheW₄ localize to spatially distinct arrays (Pearson’s correlation coefficient mean ± SD, 0.0089 ± 0.0812; n = 28 cells) that are often located in close proximity (Fig. 5), indicating that the detected interactions between CheOp1 and CheOp4 (Fig. 4B) are false positives. Due to the absence of an obvious CheOp4-related aerotaxis phenotype in previous analyses (3), we also reevaluated aerotactic behavior of the ΔcheOp4 strain in microcapillaries (see Fig. S5 and Movie S2 in the supplemental material). These experiments confirmed that CheOp4 is not mandatory for aerotaxis but bears a so far unrecognized aerotaxis-related function, contributing to a sharper separation of the aerotactic band toward higher oxygen concentrations.

FIG 5.

CheW₁ and CheW₄ localize to distinct arrays. 3D-SIM (z-series maximum intensity projection) of a representative cell expressing CheW₁-GFP (green) from the native chromosomal locus (cheW₁::cheW₁-gfp) and CheW₄-mCherry (magenta) from the constitutive P_mamDC45 promoter (brightfield image shown as inset). Yellow and white dashed lines indicate, respectively, cell contours and orthogonal cross sections shown at the right and bottom. Calibration bars indicate the intensity of fluorescence. Scale bars = 1 μm.

In summary, these observations suggested that CheOp1-related chemoreceptor arrays form independent of cheOp2 to cheOp4 and also that CheOp4 appears to be involved in the regulation of flagellar motility in addition to CheOp1.

DISCUSSION

Compared to E. coli, which possesses one canonical chemotaxis pathway and 5 MCP-encoding genes, signal transduction systems in environmental bacteria are often more complex, comprising multiple chemoreceptors and several chemotaxis pathways that are also involved in the regulation of nonmotility-related traits (6, 19). M. gryphiswaldense also exhibits a highly complex signal transduction network that includes 4 chemotaxis pathways (3). Our results provide strong evidence that signaling units related to the major pathway CheOp1 are consistently located at both cell poles throughout the cell cycle by initiation of the formation of new chemoreceptor clusters at midcell before completion of cytokinesis (Fig. 1 and 2). This bipolar localization pattern agrees with previous observations of CheW₁ localization in M. gryphiswaldense (11) or localization of selected MCPs in Magnetospirillum magneticum (20) and might have implications on signaling given the fact that each polar cluster is located in close proximity to one flagellar motor. Most bacteria are too small to spatially sense gradients along their cell body (8), i.e., chemoreceptor arrays at both cell poles are likely exposed to the same stimulus. However, the close proximity of an individual array adjacent to each flagellar motor in bipolar flagellated bacteria might be advantageous to maintain similar intracellular levels of phosphorylated CheY (CheY-P) at both cell poles and thus maximize sensitivity of the chemotaxis system. Occasionally observed CheOp1-related clusters at nonpolar sites (Fig. 1) might contribute to minimizing heterogeneity in intracellular CheY-P levels. Calculations based on simulated diffusion of CheY-P (21) suggest that positioning of the chemosensory array close to the flagellum is not essential for efficient chemosensory signaling in peritrichously flagellated E. coli cells. However, CheY-P diffusion might limit signaling for distances of >2 μm (9, 21, 22), i.e., in M. gryphiswaldense cells where opposing polar chemoreceptor arrays are spaced 3 to 8 μm apart (Fig. 1), each array might exhibit a stronger influence on its adjacent flagellar motor. How the rotational bias of bipolar flagellar motors in magnetospirilla is exactly controlled (and maybe coordinated) is unknown. Different models of flagellar motor control in bipolarly flagellated spirilla have been postulated (23, 24). Asymmetric rotation of both flagellar motors in opposite senses (as suggested in M. magneticum [23]) must invoke different CheY-P levels at both cell poles, or differences in the molecular architecture and switching mechanisms of opposing flagellar motors, whereas the counterclockwise rotation of both motors during straight swimming (facilitated by a rolling motion of the leading flagellar hook, as suggested in the nonmagnetotactic spirillum Campylobacter jejuni [24]), would be in-line with a molecular model that assumes that both flagellar motors are mechanistically controlled in the same manner. Given the complexity of the chemotaxis system in magnetospirilla (e.g., multiple CheY paralogs in magnetospirilla, in contrast to only one in C. jejuni [24]) direct comparison of both systems must be considered with caution.

Asymmetric distribution or activity of chemotaxis-related proteins was also implicated to function as a cellular landmark or polarity marker during the control of polar MTX (3). Such a mechanism might apply with respect to the asymmetric inheritance and maintenance of a monopolar flagellation pattern in MTB (25), but with respect to the transmission of the chemotaxis signal, our results rather suggest that bipolar CheA₁ localization is beneficial for efficient aerotaxis (Fig. 2). Furthermore, we observed similar mCherry- and mTurq2-CheA₁ localization patterns in polar M. gryphiswaldense populations (not shown), disfavoring that polar MTX is controlled by repositioning of large chemoreceptor arrays. The observed differences in aerotactic efficiency between mTurq2- or mCherry-CheA₁-expressing cells might be related not only to CheA₁ localization but also to an impaired CheA₁ functionality. Since mTurq2 and mCherry are both considered monomeric (13), distinct localization patterns (Fig. 2) might be caused by properties of both fluorescent proteins other than oligomeric state (e.g., maturation time). In the case of mTurq2-CheA₁-expressing cells, it remains unknown if predominant monopolar localization of mTurq2-CheA₁ confers mislocalization of entire array structures, which would require further analysis based on the colabeling of CheA₁ and CheW₁ or associated MCPs or an analysis of chemoreceptor array size by CET.

We further show that CheA₁ is essential for polar localization of CheW₁ (Fig. 3), providing evidence that the formation of stable CheOp1-related chemotaxis arrays is not possible in the absence of CheA₁. The requirement of CheA to form stable arrays is not universal, e.g., in Vibrio cholerae polar chemoreceptor arrays of cluster II form in the absence of CheA since its structural role is substituted by a specific ParP protein that is also involved in the polar positioning of chemoreceptor arrays (26). In contrast, cheOp2 to cheOp4 were not required for polar localization of CheW₁ (Fig. 3A). These observations agree with the previously observed cheOp-deletion phenotypes in M. gryphiswaldense, which revealed that CheOp1 is the major pathway to control aerotaxis and swimming reversals (3), i.e., CheOp1-related signaling units assemble and function in the absence of cheOp2 to cheOp4. The distinct localization mostly at the cell periphery (separated by larger distances from polar flagellar motors) of cheOp2- and cheOp3-encoded CheW paralogs (and absence of interactions between cheOp1- and cheOp3-encoded proteins) (Fig. 3B and Fig. 4A) is in-line with the prediction of CheOp2 and CheOp3 as alternative cellular functions (ACF) class chemotaxis systems (MiST3 database [27]; Fig. S1) (28). Hence, both pathways likely act independently of CheOp1 and mediate traits other than control of flagellar motor output.

CheOp3 bears similarities to well-studied ACF-type pathways (e.g., a multidomain CheY-protein with GGDEF-domain), such as the Wsp pathway, which controls biofilm formation and flagellation in Pseudomonas aeruginosa in a c-di-GMP manner in response to cell surface stimuli (6). Thus, it might be hypothesized that CheOp3 is also involved in the regulation of flagellation in M. gryphiswaldense, supported by its close association with flagellar genes on the genomic level.

Similar to CheOp3, the exact function of CheOp2 is unknown (3). cheOp2 encodes a ParA-like protein (Fig. S1 and S3), a type of protein which is involved in the segregation of cytoplasmic chemoreceptor arrays in Rhodobacter sphaeorides (29, 30). However, in contrast to R. sphaeorides, both MCPs assigned to cheOp2 are predicted to be membrane bound, raising the question of whether MCPs and other chemotaxis proteins encoded in cheOp2 comprise functional membrane-associated signaling units which are positioned at the cell boundary by the help of this ParA-like protein.

In contrast to the distinct localization at the cell boundary of cheOp2- and cheOp3-encoded CheW paralogs, we found that CheW₄ localizes in a similar pattern as CheW₁ (Fig. 3B). The chemotaxis system of the alphaproteobacterium Azospirillum brasilense includes two operons named Che1 and Che4 (18) with similar genetic organization and classification as CheOp1 and CheOp4 in M. gryphiswaldense (3). However, in contrast to M. gryphiswaldense, Che4 is the major pathway in A. brasilense to regulate swimming reversals (31) (compared to CheOp1 in M. gryphiswaldense [3]), whereas Che1 affects swimming speeds (32). Chemoreceptors of different length classes (as determined by their number of heptad [H] repeats) segregate into distinct arrays (7, 33). Both the strict separation (i.e., maintaining signal cooperativity within individual chemoreceptor arrays) and also the integration of distinct chemotaxis signaling pathways into the same chemoreceptor array have been observed (6, 34). In A. brasilense, membrane-bound chemoreceptors with different cytoplasmic domain lengths form two distinct arrays, which are composed of baseplates that consist of mixed CheW/A paralogs from both Che1 and Che4 (18). CheW and CheA paralogs of both pathways physically interact in A. brasilense, and CheW₁ and CheW₄ mislocalize only in the absence of both che1 and che4 (18), suggesting a stronger association between both pathways than that in M. gryphiswaldense. Considering that CheW₁ and CheW₄ in M. gryphiswaldense localize in spatially distinct arrays (Fig. 5), and independently of cheOp4 and cheOp1 (Fig. 3), respectively, a model where cheOp1- and cheOp4-encoded CheW and CheA paralogs mix within a single polar-lateral cluster, or within two separate clusters composed of 36H and 38H length-class MCPs as in A. brasilense, appears unlikely. Assuming that the detected interactions between CheOp1 and CheOp4 using BACTH (Fig. 4B) were artificially or indirectly caused by elevated expression levels of protein fusions in E. coli (35) or the presence of the intrinsic E. coli F7-type (28) chemotaxis system, our data support a model favoring the localization of cheOp1- and cheOp4-encoded CheW (and CheA) paralogs into spatially distinct polar-lateral arrays. Also, the strength of the interaction (Fig. 4B) and abundance of individual Che proteins (Fig. S4) are likely important to determine in vivo connectivity during chemoreceptor array formation (36, 37).

Bioinformatic analysis forecasted the existence of five different MCP length classes in M. gryphiswaldense (24H [1×], 36H [2×], 38H [39×], 40H [4×], and 44H [8×]). Similar to A. brasilense, CheOp1 and CheOp4 are classified as F5 and F7 chemotaxis systems involved in flagellar motility (MiST3 database [27]; Fig. S1), which are likely to interact with 38H and 36H receptors, forming arrays of ∼28 nm and ∼30 nm in height, respectively (18). The previous observation of a polar-lateral chemoreceptor array with a height of ∼28 nm in M. gryphiswaldense (Fig. 3H in reference 10) suggests that this array might be related to CheOp1. Localization analysis of selected candidate aerotaxis MCPs (encoded outside cheOp1 to cheOp4) belonging to the 38H length class indeed confirmed that their polar localization requires the presence of cheOp1 (which lacks MCP genes) but not cheOp2 to cheOp4 (D. Pfeiffer, unpublished results), in support for a model where predominating 38H MCPs do form clusters and feed into CheOp1. Predictions that most chemoreceptors are utilized by only one pathway have been also made, e.g., in Pseudomonas aeruginosa, which has four che pathways and its main pathway controlling chemotaxis utilizes 23 out of a total complement of 26 chemoreceptors (28). A model that assumes the segregation of CheOp1 and CheOp4 into distinct arrays is in-line with major (3) and minor (Fig. S5) aerotaxis-related functions of both pathways considered that each pathway might signal through its own cognate response regulator. CheOp4 in particular might fine-tune aerotaxis and help to set threshold oxygen levels to minimize the exposure of cells to suboptimal higher oxygen levels. Apart from aerotaxis, CheOp1 (and CheOp4) might further mediate chemotactic responses that so far have not been identified under laboratory conditions but are relevant in environmental habitats, given the vast number of ∼56 putative MCPs encoded within the M. gryphiswaldense genome.

In conclusion, we provided insights into the spatiotemporal organization of highly complex chemotaxis systems in M. gryphiswaldense, focusing onto the major pathway for aerotactic signaling CheOp1. Our findings suggest that bipolar positioning of CheOp1-related chemoreceptor arrays is important for efficient aerotaxis in M. gryphiswaldense and that CheOp1 signaling units assemble independently of all other chemotaxis pathways present. Theoretical predictions of receptor-chemotaxis pathway interlinkage allow us to forecast that CheOp1 utilizes and integrates signaling input of most chemoreceptors within M. gryphiswaldense but need to be further experimentally confirmed to also reveal the functions of cheOp2 and cheOp3.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

Strains are listed in Table 1. M. gryphiswaldense was grown in flask standard medium (FSM) (38) at 28°C, either in 6-well plates without agitation or in polypropylene tubes with moderate shaking (120 rpm). E. coli strains were cultivated in lysogeny broth (LB) medium at 37°C and shaking (180 rpm). For cultivation of E. coli WM3064 (W. Metcalf, unpublished), the medium was supplemented with 1 mM diaminopimelic acid (DAP). Media were solidified by the addition of 1.5% (wt/vol) agar. Selection was achieved by addition of kanamycin at a concentration of 5 μg/ml (M. gryphiswaldense) or 25 μg/ml (E. coli).

TABLE 1.

Characteristics of strains^a and plasmids

Strain or vector	Relevant characteristic(s)	Reference and/or source
Strains
E. coli
DH5α	Host for cloning; F+ endA1 glnV44 thi-1 recA1 relA1 gyrA96 deoR nupG purB20 φ80dlacZΔM15 Δ(lacZYA-argF)U169, hsdR17(rK–mK+), λ–	48
WM3064	Conjugation strain; thrB1004 pro thi rpsL hsdS lacZΔM15 RP4-1360 Δ(araBAD)567 ΔdapA1341::[erm pir]	William Metcalf at UIUC
BTH101	Two-hybrid reporter strain; F- cya-99 araD139 galE15 galK16 rpsL1 (Strr) hsdR2 mcrA1 mcrB1	Euromedex
M. gryphiswaldense
Wild type	MSR-1 R3/S1; Rifr, Smr	49
ΔcheOp1	Chemotaxis operon 1 deletion strain	3
ΔcheOp4	Chemotaxis operon 4 deletion strain	3
ΔcheOp2-4	Strain lacking the chemotaxis operons 2, 3, and 4	3
ΔcheW1	cheW1 deletion strain	11
ΔcheA1	cheA1 deletion strain	This study
cheW1::cheW1-gfp	Strain with CheW1-GFP fusion expressed from the native chromosomal cheW1 locus	11
cheA1::mCherry-cheA1	Strain with mCherry-CheA1 fusion expressed from the native chromosomal cheA1 locus	This study
cheA1::mTurq2-cheA1	Strain with mTurquoise2-CheA1 fusion expressed from the native chromosomal cheA1 locus	This study
parAOp2::parAOp2-gfp	Strain with ParAOp2-GFP fusion expressed from the native chromosomal parA locus in cheOp2	This study
mcpAOp3::mcpOp3-gfp	Strain with MCPOp3-GFP fusion expressed from the native chromosomal locus in cheOp3	This study
Vectors
pORFM	Universal in-frame deletion/in-frame fusion vector for GalK-based counterselection; npt galK tetR mobRK2	39
pORFM-ΔcheA1	Vector for chromosomal deletion of cheA1	This study
pORFM-mCherry-cheA1	Vector for insertion of mCherry at the chromosomal cheA1 locus	This study
pORFM-mTurq2-cheA1	Vector for insertion of mTurq2 at the chromosomal cheA1 locus	This study
pORFM-parAOp2-gfp	Vector for insertion of codon-optimized egfp gene (magegfp) at the chromosomal parAOp2 locus	This study
pORFM-mcpOp3-gfp	Vector for insertion of codon-optimized egfp gene (magegfp) at the chromosomal mcpOp3 locus	This study
pJH39	Tn5 transposon vector for random single-copy chromosomal insertion; pBAM1 oriR6K, PmamDC45-mamC-maggbp-gbp, Kmr, Ampr, tnpA	11
pBAM-PmamDC45-cheW1-gfp	Tn5-based integrative plasmid for expression of CheW1-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW2.1-gfp	Tn5-based integrative plasmid for expression of CheW2.1-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW2.2-gfp	Tn5-based integrative plasmid for expression of CheW2.2-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW3.1-gfp	Tn5-based integrative plasmid for expression of CheW3.1-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW3.2-gfp	Tn5-based integrative plasmid for expression of CheW3.2-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW4-gfp	Tn5-based integrative plasmid for expression of CheW4-GFP under control of the constitutive PmamDC45 promoter	This study
pBAM-PmamDC45-cheW4-mCherry	Tn5-based integrative plasmid for expression of CheW4-mCherry under control of the constitutive PmamDC45 promoter	This study
pUT18C	BACTH vector designed to express a given polypeptide fused in frame at its N‐terminal end with T18 fragment; ColE1 ori; Ampr	50
pUT18	BACTH vector designed to express a given polypeptide fused in frame at its C‐terminal end with T18 fragment; ColE1 ori; Ampr	50
pKT25	BACTH vector designed to express a given polypeptide fused in frame at its N‐terminal end with T25 fragment; p15 ori; Kmr	50
pKNT25	BACTH vector designed to express a given polypeptide fused in frame at its C‐terminal end with T25 fragment; p15 ori; Kmr	50
pUT18C-zip; pKT25-zip	Derivatives of pUT18C and pKT25 with a 114-bp DNA fragment encoding a leucine zipper (positive control for two‐hybrid assays)	50
pUT18C-cheA1	cheA1Δ183-211 amplified with primers 305 and 306 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheA1	cheA1Δ183-211 amplified with primers 305 and 306 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheY1	cheY1 amplified with primers 307 and 308 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheY1	cheY1 amplified with primers 307 and 308 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheW1	cheW1 amplified with primers 309 and 310 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheW1	cheW1 amplified with primers 309 and 310 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheA3	cheA3 amplified with primers 311 and 312 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheA3	cheA3 amplified with primers 311 and 312 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheY3	cheY3 amplified with primers 313 and 314 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheY3	cheY3 amplified with primers 313 and 314 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheY3-GGDEF	cheY3-GGDEF amplified with primers 315 and 316 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheY3-GGDEF	cheY3-GGDEF amplified with primers 315 and 316 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheW3.1	cheW3.1 amplified with primers 317 and 318 cloned into pUT18C via XbaI and SmaI	This study
pKT25-cheW3.1	cheW3.1 amplified with primers 317 and 318 cloned into pKT25 via XbaI and SmaI	This study
pUT18C-cheW4	cheW4 amplified with primers 601 and 602 cloned into pUT18C via XbaI and KpnI	This study
pUT18-cheW4	cheW4 amplified with primers 601 and 602 cloned into pUT18 via XbaI and KpnI	This study
pKT25-cheW4	cheW4 amplified with primers 601 and 602 cloned into pKT25 via XbaI and KpnI	This study
pKNT25-cheW4	cheW4 amplified with primers 601 and 602 cloned into pKNT25 via XbaI and KpnI	This study
pUT18C-cheA4	cheA4 amplified with primers 603 and 604 cloned into pUT18C via XbaI and KpnI	This study
pUT18-cheA4	cheA4 amplified with primers 603 and 604 cloned into pUT18 via XbaI and KpnI	This study
pKT25-cheA4	cheA4 amplified with primers 603 and 604 cloned into pKT25 via XbaI and KpnI	This study
pKNT25-cheA4	cheA4 amplified with primers 603 and 604 cloned into pKNT25 via XbaI and KpnI	This study

^aIndividual strains harboring plasmids or chromosomal insertions of Tn5-based expression cassettes are not listed but were generated based on the according parental strains and vectors given in the table.

Molecular and genetic techniques.

Oligonucleotides (listed in Table 2) were purchased from Sigma-Aldrich (Steinheim, Germany). Genes of interest were amplified using Phusion (Thermo Scientific) and Q5 (New England BioLabs [NEB]) proof‐reading DNA polymerases. Flanking primers used for “blunt-end” cloning were phosphorylated with the T4 polynucleotide kinase (Thermo Scientific). The corresponding vector backbone was dephosphorylated using FastAP thermosensitive alkaline phosphatase (Thermo Scientific). Plasmids were constructed by standard molecular biology techniques as described in detail below, employing FastDigest restriction enzymes and T4 DNA ligase (Thermo Scientific). All constructs were sequenced by Macrogen Europe (Amsterdam, Netherlands).

TABLE 2.

Primers used in this study

No.	Primer name	Sequence (5′–3′)a
5	HLINK-oeGFP_fwd	GCTAGCCTGGCCGAAGCCGCGG
6	HLINK-oeGFP_rev	TCACTTATACAGCTCGTCCATGCCCAGG
19	ParA-oeGFP_up_fwd	TCCGGTGACCGCCACGCCTTATGC
20	ParA-oeGFP_up_rev	CTTCGGCCAGGCTAGCCAACCCGATGGCCGAAGACAGC
21	ParA-oeGFP_dw_fwd	CGAGCTGTATAAGTGAAACCCGTCATCGTATTTTCTTGAAAAACC
22	ParA-oeGFP_dw_rev	CAGACGCTGAGCGGTGCGATCC
83	oeGFP_BamHI_rev	CGGGATCCTCACTTATACAGCTCGTCCATGCCCAGG
179	MGR_0209-oeGFP_up_fwd	GCACGCCGCCGCCCAGGAAATCG
180	MGR_0209-oeGFP_up_rev	CTTCGGCCAGGCTAGCGCCGGGGTCCAGGCGGAAGCGG
181	MGR_0209-oeGFP_dw_fwd	CGAGCTGTATAAGTGAAAACGGCGGCGCCCCGGTGATGACC
182	MGR_0209-oeGFP_dw_rev	AGGCGGCGGCTGTGGGTGACGG
305	CheA-Op1_fwd_XbaI	GCTCTAGACATGGATGATCTCCTTAGCGAATTTCTGACG
306	CheA-Op1_rev_SmaI	GGGCTGCAGCACCTTTAAGGGTGGCG
307	CheY-Op1_fwd_XbaI	GCTCTAGACATGAAGTCCTGTCTGATCGTCGATGATTCC
308	CheY-Op1_rev_SmaI	GGGCCAGCAGACCGACCTGCGAGAACTTCG
309	CheW-Op1_fwd_XbaI	GCTCTAGACATGAATCAGATCGTCCCCGCCTCC
310	CheW-Op1_rev_SmaI	GGGCTTCGCTTTTGTTGAAATCCAGAAGCTTGG
311	CheA-Op3_fwd_XbaI	GCTCTAGACATGGCCTTGGACGTCAAACGCTTTGTCG
312	CheA-Op3_rev_SmaI	GGGCCCCCAGCAGACTTTTCAGCGTGTCG
313	CheY-Op3_fwd_XbaI	GCTCTAGACATGAAGATTCTGGTGGTCGACGACGACG
314	CheY-Op3_rev_SmaI	GGGCGGCGCGGCGCGCCATCACCGCG
315	Dual-CheY-Op3_fwd_XbaI	GCTCTAGACATGGAAAATCGATCCCCCTCGTTGC
316	Dual-CheY-Op3_rev_SmaI	GGGCTGGCGCCACCTGGACCTGATTGCG
317	CheW-Op3_fwd_XbaI	GCTCTAGACATGACCGACGACATGTCCCTCGACG
318	CheW-Op3_rev_SmaI	GGGCAAGGCCCCCGGCCCGGTAATCC
429	cheA1_N_up_fwd	CGGTGGTGATCCCGAGACCCTGG
430	cheA1_N_up_rev	GGCCGACTCCAAAACTCCACGCCATC
431	cheA1_N_dw_fwd	TGGTGGCGGAGGTAGCATGGATGATCTCCTTAGCGAATTTCTGACGG
432	cheA1_N_dw_rev	AAAGATCGGTGTCGCTGGCGGCACG
433	GGGGS_3x_linker	GGTGGAGGAGGTTCTGGAGGCGGTGGAAGTGGTGGCGGAGGTAGC
434	mch_fwd	AGTTTTGGAGTCGGCCATGGTGAGCAAGGGCGAGGAGGATAACATGG
435	mch_rev	CAGAACCTCCTCCACCCTTGTACAGCTCGTCCATGCCGCC
436	mTurq2_fwd	AGTTTTGGAGTCGGCCATGAGCAAGGGCGAGGAGCTGTTTACC
437	mTurq2_rev	CAGAACCTCCTCCACCTTTATATAGCTCGTCCATACCCAAGGTGATCC
448	cheW1_N_NdeI_fwd	GGGAATTCCATATGAATCAGATCGTCCCCGCCTCC
449	cheW1_N_GL_rev	CAGAACCTCCTCCACCTTCGCTTTTGTTGAAATCCAGAAGCTTGG
450	oeGFP_GL_fwd	TGGTGGCGGAGGTAGCATGGTGTCGAAGGGCGAGGAACTG
577	cheW2.1_fwd	CCTGCGAAGCTTAGGAGATCAGCATATGGATGCCGCCGCCAAGGGT
578	cheW2.1_rev	CGCCTCCAGAACCTCCTCCACCTGACGGGGCGTCGTCCCACAGG
579	cheW2.2_fwd	CCTGCGAAGCTTAGGAGATCAGCATATGACCCAGCCCGGCACCACC
580	cheW2.2_rev	CGCCTCCAGAACCTCCTCCACCCAGGCTACGGCGGTAATCGGC
581	cheW3.1_fwd	CCTGCGAAGCTTAGGAGATCAGCATATGACCGACGACATGTCCCTC
582	cheW3.1_rev	CGCCTCCAGAACCTCCTCCACCAAGGCCCCCGGCCCGGTA
583	cheW3.2_fwd	CCTGCGAAGCTTAGGAGATCAGCATATGGATCTGGCCGGACATGGG
584	cheW3.2_rev	CGCCTCCAGAACCTCCTCCACCTCGGACCAACAATTCCGGCGG
585	cheW4_fwd	CCTGCGAAGCTTAGGAGATCAGCATATGCAGAACGCTACCCCCGCT
586	cheW4_rev	CGCCTCCAGAACCTCCTCCACCCGCGGCCGCGGTGATGGC
601	cheW4_fwd_XbaI	GCTCTAGACCAGAACGCTACCCCCGCTTCCG
602	cheW4_rev_KpnI	GGGGTACCCCCGCGGCCGCGGTGATGGCGTCG
603	cheA4_fwd_XbaI	GCTCTAGACTCCGACGACCAAGGCTATGACC
604	cheA4_rev_KpnI	GGGGTACCCCTTGCGCCGCCAGGGCCAATGCC
628	mcherry-GL_fwd	TGGAAGTGGTGGCGGAGGTAGCGTGAGCAAGGGCGAGGAGG
629	mcherry-BamHI_rev	CGAACGGTAGGGACCCGGATCCTTACTTGTACAGCTCGTCCATGCCG
	Op1_Ins_for	GTCGCTGGAAGCACGCCTG
	cheA1del_5'_rev	GGCCGACTCCAAAACTCCAC
	cheA1del_3'_for	GTTTTGGAGTCGGCCGCATGAATCAGATCGTCCCC
	cheA1del_3'_rev	CGACCTTAAATCCTTTCGTGG

^aRestriction sites are underlined. Reverse complementary sequences used to fuse fragments via PCR are indicated in bold.

Strain construction.

Markerless site-specific chromosomal insertions and deletions were conducted using the homologous recombination-based GalK-counterselection system (39). To construct pORFM-mCherry-cheA₁ and pORFM-mTurq2-cheA₁, a 667-bp DNA fragment located upstream of the cheA₁ start codon and a 521-bp fragment comprising the first 505 bp of cheA₁ (located before a GC-rich stretch within cheA₁, see below) were amplified with primers 429/430 and 431/432 from M. gryphiswaldense genomic DNA, respectively. A sequence coding for a flexible 3× glycine-linker (GGGGS)₃ was added to the 521-bp fragment via a second PCR employing primers 433/432. Coding sequences for mCherry (14) and mTurquoise2 (15) were amplified with primers 434/435 and 436/437, respectively, and fused in between the two fragments amplified with primers 429/430 and 433/432 via overlap extension PCR. Subsequently, both resulting PCR fragments were ligated into an EcoRV-digested and dephosphorylated pORFM (39) vector. Construction of the cheA₁ deletion plasmid was conducted in a similar manner. In brief, homologous regions of 1,216 bp and 967 bp located up- and downstream of cheA₁ were amplified from M. gryphiswaldense genomic DNA with primer pairs Op1_Ins_for/cheA1del_5′_rev and cheA1del_3′_for/cheA1del_3′_rev, respectively. Subsequently, both PCR fragments were fused via overlap extension PCR, and the resulting fragment was ligated into pORFM. For the construction of pORFM-parA_Op2-gfp and pORFM-mcp_Op3-gfp, fragments up- and downstream of the respective gene’s stop codon were amplified with primer pair 19/20 and 21/22 and pair 179/180 and 181/182 and fused via overlap extension PCR with a fragment (amplified with primers 5/6) encompassing a codon-optimized egfp gene (magegfp) (40) and a sequence coding for a 4-helix linker (LAEAAAKEAAAKEAAAKEAAAKAAA), resulting in a fusion with gfp between both fragments. The two resulting fusion fragments were cloned into pORFM via EcoRV. After conjugative transfer of all resulting plasmids to M. gryphiswaldense and GalK-based counterselection, colonies were screened for deletion of cheA₁ or insertion of fluorophore-encoding genes by PCR and confirmed by DNA sequencing.

For the construction of pBAM-P_mamDC45-cheW₁-gfp, cheW₁ and an M. gryphiswaldense codon-optimized egfp gene (magegfp) (40) were amplified with primers 448/449 and 450/83, respectively. Subsequently, both fragments were fused via overlap extension PCR employing primers 448/83 after addition of a coding sequence for a 3× glycine-linker (GGGGS)₃ to the magegfp gene via PCR with primers 433/83. The resulting fragment was cut with NdeI and BamHI (primers 448 and 83 contained restriction site overhangs) and ligated into the backbone of pJH39 (11), a Tn5-based insertion vector harboring the constitutive P_mamDC45 promoter. For the construction of related pBAM-plasmids coding for paralogous CheW-GFP fusions, cheW₁ was replaced via mega-primer PCR (41), employing pBAM-P_mamDC45-cheW₁-gfp as the template. Mega-primer PCR products were generated with primers 577/578 (cheW_2.1), 579/580 (cheW_2.2), 581/582 (cheW_3.1), 583/584 (cheW_3.2), and 585/586 (cheW₄). Similarly, for the construction of pBAM-P_mamDC45-cheW₄-mcherry, gfp was replaced using a mCherry mega-primer (generated with primers 628/629) and pBAM-P_mamDC45-cheW₄-gfp as the template. All resulting plasmids were transferred into different M. gryphiswaldense strains via conjugation. Due to the random Tn5-based insertion of the expression cassette into the chromosome, at least 3 different colonies per strain and construct combination were analyzed via fluorescence microscopy to account for differences in the expression level.

Motility assay.

Motility soft agar swimming assays were performed as described previously (42). In brief, cells were grown for several passages in 6-well plates under defined microoxic conditions (2% headspace oxygen; microoxic incubator; Scholzen Microbiology Systems AG). Cultures were adjusted to an optical density at 565 nm of 0.1, and 5 μl of the diluted cell suspension was pipetted into 0.2% (wt/vol) motility soft agar, consisting of FSM (38) with a lowered sodium lactate concentration of 1.5 mM (14-cm-diameter petri dishes filled with 120 ml of agar). Soft agar plates were incubated at 28°C under atmospheric conditions (21% headspace oxygen) for 3 days.

Two-hybrid assay.

Protein interaction studies were conducted using the adenylate cyclase two‐hybrid assay, which is based on the reconstitution of adenylate cyclase (CyaA) activity in the cyaA-deficient E. coli BTH101 reporter strain (16, 17). Genes of interest were amplified from isolated chromosomal DNA of M. gryphiswaldense and cloned into two-hybrid variant vectors (primers and restriction endonuclease sites are specified in Table 2). Since numerous attempts to amplify the full-length cheA₁ gene from genomic DNA resulted in internal truncations due to a GC-rich stretch within the gene (see Fig. S2 in the supplemental material), we amplified cheA₁ from pFP11 (3) harboring a cheA₁ variant without the GC-rich region that codes for a functional CheA₁ protein lacking an alanine- and proline-rich disordered linker sequence (amino acid [aa] 183 to 211) between its histidine phosphotransferase (HPT) and dimerization domains. The functionally of this CheA₁ variant was confirmed previously by complementation analysis (3), but the absence of a linker sequence was noticed more recently upon completion of the M. gryphiswaldense genome (43). The function of the GC stretch is unknown, but it might exhibit a regulatory role at the transcriptional or translational level. All resulting T18- and T25-based plasmids were cotransformed into E. coli BTH101. Cells were plated onto LB agar supplemented with 40 μg ml⁻¹ 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal), 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG), ampicillin (100 μg ml⁻¹), and kanamycin sulfate (50 μg ml⁻¹). Plates were incubated at 28°C, and colonies were subsequently analyzed for their degree of blue color formation. To visualize the experimental outcome, several colonies per plasmid combination were grown overnight at 28°C in LB liquid medium containing IPTG (0.5 mM), ampicillin (100 μg ml⁻¹), and kanamycin sulfate (50 μg ml⁻¹). Subsequently, 3 μl of culture was spotted onto M63 mineral salt agar supplemented with 0.2% (wt/vol) maltose, X-Gal (40 μg ml⁻¹), 0.5 mM IPTG, ampicillin (50 μg ml⁻¹), and kanamycin sulfate (25 μg ml⁻¹). Plates were incubated at 28°C for about 1.5 to 2 days and documented with a Lumix FZ38 camera (Panasonic). The outcome was considered a possible positive interaction if combinations gave signals that were obviously more intense (blue color formation) than the controls (included on the same plate). Constructs carrying a leucine zipper fused to the T18 and T25 fragment were used as a positive control. Cotransformants harboring constructs coding for the respective T18 and T25 protein fusion in combination with the corresponding T25 and T18 subunit alone served as negative controls.

Fluorescence microscopy.

For fluorescence microscopy, M. gryphiswaldense strains were grown in 15-ml polypropylene tubes with sealed screw caps (culture volume, 10 ml), and 3 μl of the cell suspension was immobilized on agarose pads (44). Conventional epifluorescence imaging and 3D structured illumination microscopy (3D-SIM; striped illumination at 3 angles and 5 phases) were performed using the Eclipse Ti2-E N-SIM E fluorescence microscope (Nikon), equipped with a CFI SR Apo TIRF AC 100× H NA1.49 oil lens objective, a hardware-based “perfect focus system” (Nikon), an Orca Flash4.0 LT Plus sCMOS camera (Hamamatsu), a Spectra X epifluorescence illuminator (Lumencor), and CFP/YFP/mCherry-Triple filter for imaging of mTurquoise2, as well as a LU-N3-SIM laser unit (Nikon) with 488 nm and 561 nm laser lines and EM525/50 and EM700/75 filters for 3D-SIM imaging of GFP and mCherry, respectively. 3D-SIM sample preparation employing high-precision coverslips (0.17-mm thickness; no. 15H; Marienfeld), calibration of the objective correction collar and SIM grating focus using TetraSpeck fluorescent beads (T-7279 TetraSpeck microspheres), acquisition of 3D-SIM z-series (z-step spacing 120 nm or 150 nm at a total thickness of 1.8 to 2.2 μm), and image reconstruction using NIS-Elements 5.01 (Nikon) were performed as described previously (45). For 3D-SIM colocalization microscopy, the system was calibrated using fluorescent beads (T-7279 TetraSpeck microspheres) and the N-SIM color registration dialog to correct for color shift between different channels during image reconstruction. Epifluorescence imaging of cells shown in the supplemental material was performed using an Olympus BX81 microscope equipped with a 100× UPLSAPO100XO objective (NA1.4), an Orca-ER camera (Hamamatsu), and differential inference contrast (DIC).

Time-lapse microscopy.

3D-SIM time-lapse imaging was performed at 28°C using an incubation chamber (Tokai Hit). To capture fluorescent foci residing in different focal planes of helical cells (while minimizing phototoxicity), piezo stage 3D-SIM z-series were acquired every 20 min at 1.0-μm total thickness with 200-nm z-step spacing and 50-ms exposure time at only 10% laser power (488 nm laser line and EM525/50 filter). To generate one image per time point, z-slices were combined in ImageJ (46) using the z-projection “maximum intensity” function. Subsequently, time series were corrected for x-y-drift based on brightfield channel micrographs employing the ImageJ “multi-stack reg” plugin (Brad Busse).

Image analysis.

All images were processed using the most recent version of ImageJ Fiji software (46). Automatized detection of cell contours and associated CheW₁-GFP fluorescent maxima was performed with MicrobeJ 5.13l (47). Cell outlines were detected using slightly defocused brightfield micrographs (resulting in a phase contrast-like appearance) that were corrected for background inhomogeneity using the NIS-Elements shading correction tool (Nikon). Analysis regarding fluorescent maxima was conducted based on z-series maximum intensity projection micrographs to reliably detect all fluorescent maxima in helical cells. An analysis of CheA₁ localization was conducted manually. Pearson’s correlation coefficient was calculated based on dual-color 3D-SIM images and outlined cells defined as region of interest, employing the NIS-Elements colocalization analysis tool (Nikon).

Statistical analysis.

Statistical analysis was performed using Prism 7.04 (GraphPad) as described in the respective legend of each figure. Data sets were tested for normality using the D’Agostino and Pearson, Shapiro-Wilk, and Kolmogorov-Smirnov tests.

Data availability.

The genetic constructs and the originals of the images made for this study are available upon request from the corresponding author.