The nucleoid as a scaffold for the assembly of bacterial signaling complexes

The FrzCD chemoreceptor from the gliding bacterium Myxococcus xanthus forms cytoplasmic clusters that occupy a large central region of the cell body also occupied by the nucleoid. In this work, we show that FrzCD directly binds to the nucleoid with its N-terminal positively charged tail and recruits active signaling complexes at this location. The FrzCD binding to the nucleoid occur in a DNA-sequence independent manner and leads to the formation of multiple distributed clusters that explore constrained areas. This organization might be required for cooperative interactions between clustered receptors as observed in membrane-bound chemosensory arrays.


Introduction
The bacterial cytoplasm is not a homogeneous solution of macromolecules, but rather a highly organized and compartmentalized space where the clustering and segregation of

Results
The Frz chemosensory system co-localizes with the nucleoid of different bacterial species Fluorescence microscopy has shown that a FrzCD-GFP fusion appears as multiple distributed clusters co-localizing with the nucleoid in M. xanthus cells (Fig 1A-1D and S1 Fig) [27][28][29]. It has been also shown that an inducible FrzE-YFP fusion colocalizes with FrzCD at the nucleoid [29]. To further analyze the co-localization of FrzCD and FrzE with the nucleoid, we first constructed a strain expressing a frzE-mcherry fusion that replaced the frzE locus and is, thus, expressed under its endogenous promoter. The chimeric protein was functional (S2 Fig) and formed clusters very similar to that of FrzCD and also co-localizing with the M. xanthus nucleoid ( Fig 1A-1D and S1 Fig).
To directly show that the nucleoid supports Frz protein localization, we constructed a M. xanthus conditional mutant that lacked ParB, a protein important for nucleoid segregation whose absence causes the presence of anucleated cells, cells with abnormal nucleoid condensation and cells where the division septum is improperly positioned over the nucleoid ("guillotines") [30,31]. When frzCD-gfpand frzE-mCherry were expressed in the parB mutant, we observed that both FrzCD and FrzE clusters always co-localized with the nucleoid and cells lacking the nucleoid also lacked Frz clusters. Similarly, in cells with "guillotines", septa formed in regions occupied by both the nucleoid and Frz clusters instead than in DNA-free regions ( Fig 1E).
As a control, we looked at the cellular localization of another chemoreceptor fusion, DifA-GFP, in the absence of parB. DifA has been recently shown to form membrane bound and uniformly distributed clusters [28]. In the absence of parB, even cells without nucleoid still carried DifA-GFP clusters and these clusters localized similar to the wild type [28] (Fig 1E). These results confirmed that nucleoid-mediated cluster formation is a specific feature of Frz proteins.
To test if FrzCD and FrzE were capable of associating with the nucleoid independently of each other, we expressed frzE-mcherry in a strain lacking frzCD and frzCD-gfp in a strain lacking frzE  and this study). As previously shown, in the absence of FrzCD, FrzE-mCherry was homogeneously dispersed in the cytoplasm, and notably also in the polar regions ( Fig 1A and 1B and S1 Fig) [27,29]. Here we additionally showed that in the absence of frzE, FrzCD lost its typical punctate pattern, but the fluorescent signal was still retained towards the center of the cell body and strictly co-localized with the nucleoid (Fig 1A-1D and S1 Fig). The aberrant localization patterns observed in the deletion mutants were not due to a change in protein levels (S3 Fig). Thus, this result confirms that both FrzCD and FrzE are important for cluster formation, whereas FrzCD is responsible for the recruitment of FrzE to the nucleoid.
To check whether the association between FrzCD and the nucleoid was M. xanthus-specific, we constructed an E. coli strain expressing frzCD-gfp from a plasmid and under the control of an IPTG inducible promoter. Under these conditions, FrzCD also co-localized with the nucleoid and this co-localization was particularly evident in elongated undivided cephalexintreated cells containing multiple segregated nucleoids (Fig 2A-2D).
These observations suggest that FrzCD can associate with the chromosomes of different bacterial species, either directly or by the aid of a docking factor common to M. xanthus and E. coli.

FrzCD directly binds to the DNA
The possibility that FrzCD interacted with the nucleoid was puzzling especially considering that the direct binding between a chemoreceptor and the DNA has not been reported prior to this study. To explore this possibility, we generated a 6His-tagged FrzCD version, purified it from E. coli and tested its ability to form complexes with DNA. FrzCD binds directly to DNA because its presence altered DNA mobility on agarose gels (Fig 3A). Binding did not require extended DNA fragments because it was also observed with oligomers of DNA as small as 69 bp (S4 Fig). Moreover, FrzCD DNA-binding did not seem to depend on the DNA sequence nor on its GC content (S4 Fig) as anticipated by the in vivo results showing that FrzCD is distributed all over the nucleoid in the absence of FrzE (Fig 1A and 1B).
The shift pattern depended on the FrzCD concentration ( Fig 3A). More specifically, the shift of DNA fragments gradually increased as the FrzCD concentration was increased ( Fig 3A  and S4 Fig). Such migration profiles have been previously described for proteins that can nucleate on DNA molecules in a non -specific manner, i.e. some Type Ib ParA-like proteins [32,33]. However, to exclude that such profiles were due to the formation of FrzCD unfolded aggregates, we checked the oligomerization state of our purified 6His-FrzCD. As expected for an MCP, FrzCD forms a homogenous dimer in solution, corresponding to a molecular weight   FrzCD binds DNA to enhance the signaling activity of Che-like clusters While FrzCD does not appear to bind specific DNA motifs in vitro, it could bind to specific sites in vivo (perhaps with the help of additional factors), explaining the formation of clusters. To test this possibility, we performed chromatin-immunoprecipitation (ChIP) experiments using the frzCD-gfp strain and polyclonal GFP antibodies. As expected, FrzCD-GFP but not a FrzCD variant that cannot bind DNA (see below) was able to co-immunoprepicipate significant amounts of DNA. Deep-sequence (ChIP-Seq) [34] of the immunoprecipitated DNA revealed no enrichment in the pool of DNA fragments obtained with the ChIP meaning that FrzCD-GFP can bind any DNA sequence from the M. xanthus genome (S5 Fig; the ChipSeq results have also been deposited on GEO (https://www.ncbi.nlm.nih.gov/geo/, accession number GSE102724)). As a positive control we used a parB-yfp strain [30]. In this case, as expected, the nucleoid region corresponding to positions 9,109 to 9,110 Kb and containing parS [30]  FrzCD binds DNA to enhance the signaling activity of Che-like clusters was highly represented in the DNA pool obtained with ParB-YFP (S5 Fig). We conclude that FrzCD binds DNA in a non-sequence-specific manner to recruit FrzE and thus form clusters containing Frz signaling complexes.

The FrzCD N-terminal region is required for the FrzCD DNA binding
Beside a very conserved C-terminal methylation domain, FrzCD contains a unique 137 residue N-terminal region (Fig 3B). We then asked whether this region corresponded to the FrzCD nucleoid-binding domain and tested its ability to form complexes with the DNA in our gel shift assays. Indeed, the FrzCD N-terminal domain (FrzCD Δ131-417 ) also provoked a mobility shift of DNA fragments of different length, albeit at a lower efficiency (compare Fig 3A and  3C). On the other hand, the FrzCD C-terminal methyl-accepting domain alone (FrzCD Δ1-130 ) did not associate with any DNA fragment ( Fig 3D) showing that the methylation domain does not bind DNA.
To confirm a direct interaction between FrzCD and DNA and also better compare the DNA-binding properties of FrzCD and FrzCD Δ131-417 , we performed DNA-protein interaction experiments using Bio-Layer Interferometry (BLI), a technique previously used to study protein-protein interactions [35]. FrzCD and DNA interaction was also detected in this assay ( Fig 3E). Consistent with the gel-shift experiments, binding appeared complex and could not be fitted to a 1:1 interaction model, precluding precise determination of a K D . Nevertheless, when we compared the DNA binding curves of FrzCD Δ131-417 and FrzCD, the results confirmed that FrzCD Δ131-417 binds the DNA at a lower efficiency than FrzCD (showing slower association and faster dissociation, Fig 3E).
By further analyzing the FrzCD N-terminal domain, we realized that it contains a positively charged peptide of approximately 30 amino acids followed by a more negative region predicted to contain alpha helices (Fig 3B and 3F, S6 Fig). By searching for homologs in the UniProtKB/ SwissProt Data Base, we found that such FrzCD N-terminal basic peptide was similar to the basic tail present at the N terminus of eukaryotic histones (S6 Fig) whose deletion has been shown to substantially affect histone-DNA interactions and decrease nucleosome stability [36,37]. To test whether this sequence had a histone-tail-like function in the binding of FrzCD to DNA in vitro, we generated a 6His-tagged FrzCD version only lacking the basic amino acid sequence from residue 7 to 27 (FrzCD Δ7-27 ), purified it from E. coli and tested its ability to form complexes with the DNA in our gel shift assays. Remarkably, FrzCD Δ7-27 did not shift the migration of DNA fragments on agarose gels, similarly to FrzCD Δ1-130 missing the entire N-terminal domain (Fig 3D and 3G). In the BLI assay, binding of FrzCD Δ7-27 to DNA was still detectable, however it was severely impaired ( Fig 3E). This result suggests that the positively charged motif is required for efficient DNA binding but it may not be the sole determinant.
The different DNA binding efficiencies of the four FrzCD, FrzCD Δ1-130 , FrzCD Δ131-417 and FrzCD Δ7-27 alleles were not due to altered protein stability and folding because all recombinant proteins, except as expected FrzCD Δ131-417 (the signaling domain) were able to activate the autophosphorylation of the FrzE kinase in vitro (S4 Fig).

The binding of FrzCD to the nucleoid is required for FrzCD cluster formation in vivo
To check whether the absence of the N terminus or the basic tail also affected the binding of FrzCD to DNA in vivo, we used M. xanthus strains expressing frzCD Δ6-130 -gfp or frzCD Δ7-27 -gfp at the frzCD locus  and this study). The FrzCD Δ6-130 -GFP and FrzCD Δ7-27 -GFP fluorescence appeared mostly diffused, also occupying the polar regions (Fig 1A-1D). The two protein fusions could only rarely form short-lived clusters that localized anywhere in the cells (not only at the central region). In all cases, the aberrant localization patterns were not due to protein stability (S3 Fig). In addition, when FrzCD Δ6-130 -GFP and FrzCD Δ7-27 -GFP were produced in E. coli they also lost their co-localization with the nucleoid. However, instead of looking dispersed in the cytoplasm as in M. xanthus, FrzCD Δ6-130 -GFP and FrzCD Δ7-27 -GFP were confined at one cell pole in E. coli. It is likely that FrzCD Δ6-130 -GFP and FrzCD Δ7-27 -GFP formed aggregates targeted to the poles due to the absence of the nucleoid anchor (Fig 2).

FrzCD cluster dynamics and organization
To understand how FrzCD clusters are formed along the nucleoid, we analyzed the cluster dynamics at high temporal resolution. Contrarily to previous assumptions based on lower resolution analysis [27], these results show that FrzCD clusters dynamics are limited to constrained areas of the nucleoid and are featured by undirected Brownian-like motions (Fig 4A-4C). This mobility decreased with the increase of cluster intensity, suggesting that clusters containing more molecules might be more tightly anchored to the chromosome and, hence, more fixed ( Fig 4D). To test whether the signaling state of FrzCD also affects the cluster mobility, we tested whether clusters were also constrained in strains carrying point mutations in methylation sites either generating FrzCD loss of function (frzCD E202A-E203A -gfp) or, oppositely, FrzCD gain of function (frzCD E168A-G169A -gfp) [27]. There were no notable differences between the tested conditions, suggesting that signaling does not affect the nucleoid dynamics of Frz signaling complexes along the nucleoid (Fig 4E).
Given that FrzCD clusters are not very dynamic and form in a limited number of sites per chromosome, we wondered how these clusters were segregated to daughter cells following cell division. Analysis of the number of FrzCD clusters as a function of chromosome length revealed a linear correlation (Fig 4F), suggesting that more clusters are assembled as the DNA is being replicated. A demograph representation revealed that the clusters appear distributed similarly on replicating and segregated chromosomes and indeed their number increases with chromosome length (Fig 4F and 4G). This distribution might ensure equal partitioning in daughter cells.
The nucleoid-dependent assembly is important for cooperative signaldependent response Similar to transmembrane chemosensory clusters [6,[9][10][11], the formation of Frz nucleoidassociated clusters is not strictly required for signaling. In fact, it has been previously shown that a frzCD Δ6-130 strain, where FrzCD molecules can no longer bind DNA and are, thus, diffused (Fig 1) can still produce reversals (Bustamante et al., 2004). In E. coli chemosensory cluster formation is also not critical for signal transduction but it confers properties such as the amplification of signal, a direct consequence of the cooperative interactions between clustered chemoreceptors [11][12][13][14][15]. Thus, we decided to check if the formation of DNA-bound Frz clusters can also promote cooperativity in the signaling activity of the Frz chemosensory system. For this, we took advantage of a newly developed microfluidic single cell assay where the frequency of reversals can be measured as a function of increasing concentrations of an artificial Frz-signal activator, isoamyl alcohol (IAA) [25]. Consistent with previous observations [25], in cells where FrzCD formed nucleoid bound clusters, IAA induced a dose-dependent response with a sigmoidal shaped curve that could be fitted by the Hill equation with a coefficient n = 3.017 +/-0.2 (P = 0.0007), thus revealing possible cooperativity in IAA-dependent Frz activation (Fig 5A). Such response is FrzCD-dependent because a ΔfrzCD strain does not reverse at any IAA dose (Fig 5A). In the frzCD Δ6-130 mutant, the cells still responded to IAA in a dose-dependent manner but in this case, the calculated Hill coefficient was n = 1.15 +/-0.01 (P = 0.008), revealing that cooperation is lost in this mutant. These results suggest that FrzCD clusters have signal amplification properties that are absent in the frzCD Δ6-130 mutant. These properties are advantageous to M. xanthus social behaviors because swarming and predation are defective (although moderately) in the frzCD Δ6-130 mutant compared to the wild type ( Fig  5B and 5C).

Discussion
In bacteria, transmembrane chemoreceptors use the bacterial inner membrane as a platform to form the well described arrays of trimers of dimers [38]. Remarkably, the M. xanthus Frz system forms clusters not on the membrane, but directly on the bacterial chromosome. Cluster assembly is directed by the chemoreceptor FrzCD, which binds to the DNA by a N-terminal domain carrying a positively charged tail similar to that found in eukaryotic histones [36,37]. While the binding of FrzCD to DNA is essential to target the Frz chemosensory system to the nucleoid, it is not sufficient for Frz cluster formation, as it requires downstream interactions with the FrzE kinase. Because FrzCD appears to bind DNA in a non-sequence specific manner, DNA-bound clusters do not occupy fixed localization sites and move across small areas on the nucleoid surface. Analogous to trans-membrane proteins diffusing in the bacterial membrane, the FrzCD cluster dynamic behavior may be affected by the size of the complex (and thus the number of interactions with DNA), explaining why bright clusters show Brownian-like motions that only explore constrained nucleoid areas.
Several lines of evidence suggest that the Frz cluster formation on the nucleoid occurs in a stochastic manner similarly to the assembly of the E. coli Che lattices in the membrane. First, the initial binding of FrzCD to DNA might take place anywhere on the nucleoid as such binding is not DNA-sequence specific. Once recruited to the nucleoid, small FrzCD foci diffuse, non-directionally, across confined small areas until they might nucleate large fixed clusters by attracting more FrzCD molecules. "Newborn" FrzCD foci might also, at one point, be incorporated by existing neighboring clusters. Thus, the areas explored by FrzCD clusters might represent the minimal critical distance from other clusters at which foci can exist. The existence of such minimal critical distance is supported by the fact that the number of Frz clusters increases linearly with the nucleoid size, suggesting that more clusters can form when more surface becomes available. Thus, like for transmembrane chemoreceptors, FrzCD molecules might either nucleate new dynamic foci if they are far enough from existing clusters, or encounter and join neighboring clusters. This simple mechanism might ensure correct partitioning of the clusters in daughter cells without the need for a faithful partitioning system. Such partitioning system appear, in fact, to be required in the presence of a single cluster like in R. sphaeroides where PpfA ensures the faithful segregation of Tlp clusters [18,20]. FrzCD binds DNA to enhance the signaling activity of Che-like clusters FrzCD binds DNA to enhance the signaling activity of Che-like clusters Cluster formation is unlikely to be uniquely linked to partitioning constraints during cell division. Indeed, the FrzCD variants that do not bind to the nucleoid form diffused signaling complexes that can be easily partitioned between daughter cells via diffusion. We therefore favor that nucleoid binding confers particular signaling properties to Frz complexes. In support of this, chromosome-bound complexes respond to signals in a cooperative manner, while soluble complexes do not. This could be explained if Frz clusters were arranged on the chromosome similar to the transmembrane chemoreceptors [39], forming ordered signaling arrays with signal amplification properties.
But, what signals are sensed by the chromosomal complexes? Frz clusters have been previously associated to the regulation of the cell reversal frequency in response to cell-cell contact. In fact, FrzCD clusters align in adjacent M. xanthus cells, a behavior that also seemed to induce cell reversals [27]. In support of this hypothesis, adjacent cells of a frzE strain did not produce cluster alignment. We now find that what seemed to be more diffused FrzCD clusters in the frzE strain at the time, are, in fact, FrzCD molecules dispersed on the nucleoid. Moreover, we now show that the Frz cluster dynamics are independent on the Frz signaling activity. Thus, the observed FrzCD cluster alignment might be more likely determined by similar dynamic rearrangements of the nucleoid of adjacent cells rather than by an active regulated mechanism. Therefore, we propose that chromosome assembly allows the formation of ordered chemosensory complexes that sense a yet unidentified intracellular signal. Finally, it cannot be excluded that the Frz complex responds to or even regulates chromosomal processes, although there is currently no evidence for such activity of the Frz system.
Finally, the analysis of the FrzCD sequences from some related species of δ-proteobacteria shows that while the FrzCD C-terminal region is very conserved, its N-terminus largely varies. Nevertheless, the FrzCD N-terminus always shows a positively charged sequence (S7 Fig) suggesting that the non-sequence-specific recruitment of Frz proteins to the nucleoid essentially requires the presence of a positively charged protein domain rather than a specific amino acid sequence.
This type of cellular organization may be common to other bacterial macromolecular complexes to provide important regulatory functions. In this sense, the Frz example provides new perspectives to the role of the bacterial nucleoid as a scaffold for the spatial control of cellular functions.

Bacterial strains, plasmids and growth
Strains and plasmids are listed in S1 and S2 Tables. M. xanthus strains were grown at 32˚C in CYE rich media as previously described. Pcuo::parB-ΔparB cells were grown at 32˚C in CTT supplied with 300 μM CuSO 4 .
Plasmids were introduced into M. xanthus cells by electroporation. Deletions and GFP fusions were inserted in frame to avoid polar effects on the downstream gene expression. These strains were obtained by homologous recombination based on a previously reported method using the pBJ113 or pBJ114 vectors [26,28]. To generate strains expressing GFP or mCherry fusion proteins, we constructed DNA cassettes including the last approximately 800 bp of each gene, with the exception of the stop codon; the gene encoding the egfp or mcherry genes from pEGFP-N1 (Invitrogen) or pEM147 [40] excluding the start codon and including the stop codon; the intergenic region between the gene of interest and its immediately downstream gene, if any; the first 800 bp of the downstream gene.
To construct strains carrying a FrzS-YFP fusion, we used the pEFrzSY plasmid [25]. Escherichia coli cells were grown under standard laboratory conditions in Luria-Bertani broth supplemented with antibiotics if necessary.
For swarming assays, cells (5 μl) at a concentration of 5×10 9 cfu ml −1 were spotted on CYE agar plates and incubated at 32˚C and photographed after respectively 48h with an Olympus SZ61 binocular stereoscope. For predation assays, E. coli (3μl at a concentration of 5×10 9 cfu ml −1 ) and M. xanthus cells (3μl at a concentration of 5×10 9 cfu ml −1 ) were spotted at the average 0.7 mm distance from each other on CF agar plates, incubated at 32˚C and photographed after 72 h.
Protein purification BL21(DE3) [F − ompT hsdSB(rB − mB -) gal dcm (DE3)] cells were grown in Luria-Bertani broth supplemented 100 μg/ml ampicillin to mid-exponential phase at 37˚C. For the experiments shown on Fig 3 and phosphorylation assays on S4 Fig, overexpression was induced by adding 0,1 mM IPTG to cells containing plasmid pEM414 and 0,5mM for pEM415 and pEM433. Cells were then grown at 16˚C over night. Cells were washed and resuspended in lysis buffer (50 mM TrisHCl, pH 8; 300 mM NaCl; 100 μg/ml PMSF; 30 U/mL Benzonase) and lysed at the French press. The cell lysates were centrifuged at 4˚C for 30 min at 13000× rpm. Soluble tagged His 6 -proteins were purified on 1ml HisTrap TM FF columns (GE Healthcare) and desalted with PD-10 columns (GE Healthcare). Ultimately, purified proteins were eluted in 50 mM TrisHCl, pH 8 and 300 mM NaCl.
For the experiments shown on S4A- S4E Table), in buffer (10 mM of TrisHCl at pH 8; 60 mM of NaCl; 10% glycerol). Reactions were incubated for 40 min at 4˚C before being loaded on 1% agarose gels. Gel migration was performed in 1X TBE at 4˚C for 55 min. Gels were, then, stained with ethidium bromide and revealed at the UV light.
EMSAs for shorter oligos (S4 Fig and S4 Table) were carried out in buffer containing 50 mM Tris, 50 mM NaCl, 1 mM dithiothreitol and 10% glycerol, pH 7.4. All the oligonucleotide fragments were PCR-amplification products, except the 69 bp DNA that was generated by annealing the custom-synthesized oligonucleotides (Sigma) corresponding to the 5' to 3' sequence and its complementary strand (S4 Table). The samples were incubated at 25˚C for 20 min and loaded onto the appropriate % of agarose gel (1% for 1.3 kbp, and electrophoresed for 60-90 minutes in 1x TAE (40 mM Tris-Acetate and 1 mM EDTA).
Oligomerization study using size exclusion chromatography coupled with multi-angle light scattering (SEC-MALS) The expected elongated structure of an MCP precludes estimation of the oligomeric status by size exclusion chromatography alone, and hence we carried out SEC coupled with multi-angle light scattering to estimate the molecular mass in solution. The mass was determined using a Wyatt Dawn Heleos II equipped with light scattering detectors at 18 angles and an Optilab TrEX differential refractive index detector. Protein sample (100 μl of 2 mg/ml solution or 45 μM) was injected into the size exclusion column (BioRad EnRich650) equilibrated with the buffer 50 mM Tris, 50 mM NaCl, pH 8.0, and the run carried out at a flow rate of 0.4 ml/min. The observed molecular masses at various points along the peak in the elution curve were calculated using the protein concentration estimated from the differential refractive index (dRI), and measured scattered intensities. The Debye model in the ASTRA software provided with the equipment was used for fitting the data.

Biolayer interferometry
Protein-DNA interaction experiments were conducted at 25˚C with the BLItz instrument from ForteBio (Menlo Park, CA, USA). The BLI consists in a real time optical biosensing technique exploits the interference pattern of white light reflected from two surfaces to measure biomolecular interactions [41]. Purified 6His-FrzCD, 6His-FrzCD Δ131-417 , 6His-FrzCD Δ6-130 and 6His-FrzCD Δ7-27 protein ligands were immobilized onto two different Ni-NTA biosensors (ForteBio) in duplicate at 1μM concentrations. A PCR amplified DNA fragment (474bp) with primers AGACCCCCGCACCCACGGAG and TCACGCGGGCTCCGGCTC (Eurogentec) was used as the analyte throughout the study at the 38nM. The assay was conducted in PBS pH 7.5, 0.001% tween-20. The binding reactions were performed with an initial baseline during 30 seconds, an association step at 120 seconds and a dissociation step of 120 seconds with lateral shaking at 2200rpm. A double reference subtraction (sensor reference and 6His-FrzCD Δ1-130 ) was applied to account for non-specific binding, background, and signal drift to minimize sensor variability.

Protein sequence analyses
In order to search for homologs of the FrzCD N-terminal domain, the first 130 aminoacids of the FrzCD sequence were BLAST into the UniProtKB/SwissProt Data Base (http://blast.ncbi. nlm.nih.gov). Predictions of secondary structures and protein sequence alignments were obtained with Jpred [42] and Clustal Omega [43], respectively. To analyze the FrzCD N-terminal region protein charge, "Sliding window" analyses were performed with Microsoft Excel.

Fluorescence microscopy and image analysis
For fluorescence microscopy analyses, 5 μl of cells from 4 x 10 8 cfu ml -1 vegetative CYE cultures were spotted on a thin fresh TPM agar pad at the top of a slide (Mignot et al., 2005). A cover slip was added immediately on the top of the pad, and the obtained slide was analyzed by microscopy using a Nikon Eclipse TE2000 E PFS inverted epifluorescence microscope (100 x oil objective NA 1.3 Phase Contrast) [44].
To study the colocalization with the DNA, the TPM agar pads were supplied with 1μg/ml DAPI stain and 1 mM IPTG. Prior to imaging, E. coli cells were grown in 1 mM IPTG for one hour, then spotted on agar pads containing or not 10μg/ml cephalexin and incubated for 1 hour. Cell fluorescence profiles were obtained with the "plot profile" function of FIJI [45]. FrzCD cluster numbers and distances, nucleoid areas and cell areas were automatically determined and verified manually with the "MicrobeJ" Fiji/ImageJ plugin created by A. Ducret, Brun Lab (http://www.indiana.edu/~microbej/index.html). All data plots and statistical tests were obtained with the R software (https://www.r-project.org/).
To study the dynamics of FrzCD-GFP clusters we automatically tracked clusters (imaged every second) by MicrobeJ and recorded parameters such as the mean square displacement (MSD), the confinement radius and the fluorescence intensity along the time.
For manual image analysis we chose a sample size that allowed an error on the mean (sem) lower than 10%. For results generated by automated analyses, we analyzed all the available samples. We chose all cells where clusters and nucleoids were tractable by our image analysis tools (Fiji and MicrobeJ). For the cluster dynamic determinations we analyzed all clusters that were tractable for at least 5 consecutive frames.

Reversal frequencies
These assays were performed as previously described [25] by using homemade PDMS glass microfluidic chambers [46] treated with 0.015% carboxymethylcellulose after extensive washing of the glass slide with water. For each experiment, 1mL of a CYE grown culture of OD = 0.5 was injected directly into the chamber and the cells were allowed to settle for 5 min. Motility was assayed after the chamber was washed with TPM 1mM CaCl 2 buffer. For IAA injections, IAA solutions made in TPM 1mM CaCl 2 buffer at appropriate concentrations were injected directly into the channels and motility was assayed directly under the microscope. Time-lapse movies of strains carrying a FrzS-YFP fusion were shot for 20 minutes with frames captured every 15 seconds.
To discriminate bona fide reversal events from stick-slip motions [25], the fluorescence intensity of FrzS-YFP was measured at cell poles over time. In fact, this protein has been shown to switch from the leading cell pole to the lagging pole when M. xanthus cells reverse their movement direction [47]. For each cell that was tracked, the fluorescence intensity and reversal profiles were correlated to distinguish bona fide reversals from stick-slip events. About one hundred cells for the wild type and frzCD Δ1-130 strains and fifty for the ΔfrzCD strain were analyzed (refer to S3 Table for the exact number of cells analyzed for each strain and IAA dose). The number of reversals was plotted against time. The best fits for the reversal frequencies values at the different IAA doses were obtained with the following Hill equation: where the N rev /ΔT is the number of reversal events per hour; K d is the apparent affinity constant; [IAA] is the IAA dose; B is the plateau; A is the basal reversal frequency and n is the Hill coefficient describing cooperativity. Reversal frequency values for each IAA dose and each strain are the results of two independent biological triplicates. We chose a sample size that allowed an error on the mean (sem) lower than 10%. We used all cells that moved, remained isolated for at least 20 consecutive frames and where FrzS-YFP foci were detectable by FIJI.  Table) Table. DNA sequences used for EMSA in Fig 3 and S4 Fig.  (DOCX)