CdbA is a DNA-binding protein and c-di-GMP receptor important for nucleoid organization and segregation in Myxococcus xanthus

Cyclic di-GMP (c-di-GMP) is a second messenger that modulates multiple responses to environmental and cellular signals in bacteria. Here we identify CdbA, a DNA-binding protein of the ribbon-helix-helix family that binds c-di-GMP in Myxococcus xanthus. CdbA is essential for viability, and its depletion causes defects in chromosome organization and segregation leading to a block in cell division. The protein binds to the M. xanthus genome at multiple sites, with moderate sequence specificity; however, its depletion causes only modest changes in transcription. The interactions of CdbA with c-di-GMP and DNA appear to be mutually exclusive and residue substitutions in CdbA regions important for c-di-GMP binding abolish binding to both c-di-GMP and DNA, rendering these protein variants non-functional in vivo. We propose that CdbA acts as a nucleoid-associated protein that contributes to chromosome organization and is modulated by c-di-GMP, thus revealing a link between c-di-GMP signaling and chromosome biology.

To gain insights into the mechanism of CdbA and CdbB, we determined the CdbA crystal structure (Supplementary Table 1). We obtained two crystal forms of full-length CdbA-His 6 . The first structure (traced residues  verified that CdbA is a RHH superfamily member, with crystallographic symmetry generating a classical RHH dimer (Fig. 2a). This observation was confirmed by DALI analysis 38 , which suggested the streptococcal RHH repressor CopG as the closest structural homolog (PDB: 2CPG, r.m.s.d. of 1.9 Å over 45 residues 39 ). In CdbA, the exposed face of the β-strand projects residues K8, S10 and Y12, while L11 and F13 are facing the hydrophobic core of the dimer. The β-strand is terminated by P14, leading into α-helix1 (residues [16][17][18][19][20][21][22][23][24][25][26][27][28], followed by a loop and then α-helix2 (residues [32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48], with a strongly kinked region at A43/ R44, and a C-terminal tail traceable to A52 (Fig. 2a). The two α2 helices run antiparallel, with kink residues I42 and A43 packing against their counterparts in the opposing chain (Fig. 2a). Successful model building of the first form allowed us to solve the second form via molecular replacement. This form has three copies of the protein in the asymmetric unit, one of which forms a tetramer via crystallographic symmetry (four copies of same chain) and two of which form a dimer within the asymmetric unit, which then tetramerizes over the final crystallographic axis. These two tetramers are independent yet essentially identical and we interpret these to represent the physiological tetramer identified by SEC. The tetramer centers on residue F50 of the two innermost chains, and the C-terminal tails can be traced five residues further than in form one, and with the C-terminal tail of the innermost chains in a dimer contacting an opposing dimer (Fig. 2b). Protein-protein contacts mediated by the tail region include a hydrophobic component from F50 and V53 in one chain projecting towards a pocket in the opposing dimer formed by W34/I42/I47, a second hydrophobic interaction between V56 and P14/M17, and a more polar interface between the free carbonyl groups of A52/V53/D55 and the side chains of Q38 and K41 (Fig. 2b). Upon solving this crystal form, we realized that a similar, but more loosely-packed, tetramer can be observed when considering the crystal symmetry of form one. The arrangement of the two dimers in the tetramer places their β-sheets~35 Å apart (measured between Cα atoms of equivalent Y12 residues).
Using the RHH protein Arc, which binds DNA as a tetramer and with the two β-sheets separated by 26Å 40,41 , as a template, we created a model of how CdbA could bind DNA (Fig. 2c, d).
Consistent with the idea that the two β-sheets in the tetramer make DNA contacts, a surface electrostatic map of CdbA revealed that this region is positively charged (Fig. 2c). Because the βsheets are~35 Å apart in CdbA, this model suggests that CdbA bends DNA upon binding (Fig. 2d).
c-di-GMP binding to CdbA induces conformational changes. Because we were unsuccessful in obtaining the crystal form of cdi-GMP-bound CdbA, we performed hydrogen-deuterium exchange (HDX) mass-spectrometry (MS) in the presence and absence of c-di-GMP to identify regions in CdbA involved in cdi-GMP binding or undergoing conformational changes upon cdi-GMP binding. We obtained a total of 68 peptides in both states that cover the entire CdbA sequence with~11-fold redundancy per amino acid (Supplementary Data 1, Supplementary Fig. 2a). The HDX profile in the absence of c-di-GMP is in agreement with the crystal structure, i.e., the RHH fold (amino acids  exchanges hydrogen to deuterium more slowly than the disordered C-terminal region ( Supplementary Fig. 2a). The difference in HDX between c-di-GMP-bound and apo-CdbA revealed changes that were restricted to the RHH fold (Fig. 3a, Supplementary Fig. 2b). Specifically, the N-terminus, the β-strand, a portion of α-helix1, the loop between α-helices1 and 2, and the Nterminal tip of α-helix2 had reduced HDX in the presence of c-di-GMP (Fig. 3a-c). When illustrating the location of three representative peptides that exhibit substantial differences in deuterium incorporation (Fig. 3b) onto the crystal structure of a CdbA dimer or tetramer (Fig. 3c), two of the three peptides (3-13 and 29-34) converge providing a polar pocket formed by residues Y12 and N54 from the two innermost chains and K6, K8, D29, R30, and S31 from the two outermost chains (Fig. 3d). Because HDX-MS cannot discriminate between peptides originating from the inward-and outward-facing chains, we cannot dissect whether the copies of the two peptides that do not engage in formation of this pocket have altered HDX profiles upon c-di-GMP binding. Nevertheless, given the proximity of the β-strand (peptide 3-13) and the helix α1-α2 loop including the N-terminal tip of α-helix2 (peptide 29-34), we hypothesized that they might represent the c-di-GMP binding site. From hereon, we refer to these two regions as interface −1 and −2, respectively (Fig. 3e).
To probe the participation of interface −1 and −2 in c-di-GMP binding, we substituted residues in both interfaces that show 100% conservation in CdbA homologs ( Supplementary  Fig. 1c, Fig. 3e), to generate CdbA K8A/S10A and CdbA R27A/R30A (Fig. 3e). In SEC, both variants, eluted with an apparent molecular mass of~35 kDa (Fig. 1d) indicating correct tetramer formation; however, they did not detectably bind c-di-GMP as estimated by ITC (Fig. 1c) and SEC (Fig. 1d) suggesting involvement of both interface-1 and -2 in c-di-GMP binding. Of note, the two interfaces important for c-di-GMP binding overlap with the regions of CdbA predicted to be involved in DNA binding (Fig. 2d) Fig. 3 HDX assay for CdbA in the presence and absence of c-di-GMP. a Representative peptides of CdbA are colored according to the difference in HDX between the c-di-GMP-bound and the apo-state. The secondary structure of CdbA is shown below. b HDX for three representative peptides of CdbA (marked red in panel a) in the c-di-GMP-bound (blue) and apo-state (red). Data represent mean ± SD (n = 3 separate reactions). c Location of the three representative peptides in b in the crystal structure of the CdbA dimer. d Location of the three representative peptides from b in the crystal structure of the CdbA tetramer with residues proposed to be involved in formation of the polar pocket for c-di-GMP binding indicated as sticks. e Crystal structure of CdbA dimer (left) and tetramer (right) with interface 1 and -2 of each monomer colored in red, green, yellow, and blue. Amino acids selected for site directed mutagenesis are indicated. CdbA is essential for viability while CdbB is dispensable. We attempted to generate in-frame deletions in cdbA and cdbB. We readily obtained the in-frame deletion of cdbB and the mutant was indistinguishable from WT with respect to growth, cell morphology, motility and development ( Fig. 4a-c, Supplementary  Fig. 3a). By contrast, we were unable to generate an in-frame deletion of cdbA (ΔcdbA). We, therefore, constructed two merodiploid derivatives of WT in which mCherry (mCh)-tagged a b c d CdbA was expressed ectopically. In one strain, cdbA-mCh was expressed from the native cdbA promoter in a single copy from the Mx8 attB site (WT/Pnat_cdbA-mCh) and in the second, cdbA-mCh was expressed from the vanillate-inducible promoter in a single copy from the MXAN_18/19 locus (WT/Pvan_cdbA-mCh). In immunoblots, CdbA-mCh accumulated in both strains; however, even at the highest concentration of vanillate (500 µM), the level of CdbA-mCh was~9-fold lower when expressed from the vanillate-inducible promoter than when expressed from the native promoter (Fig. 4d). In the absence of vanillate, CdbA-mCh was undetectable by immunoblotting in the WT/Pvan_cdbA-mCh strain verifying that Pvan tightly regulates cdbA-mCh expression (Fig. 4d). Using the WT/Pvan_cdbA-mCh strain, we successfully generated an in-frame deletion of cdbA at the native locus in the presence of 500 µM vanillate (ΔcdbA/Pvan_cdbA-mCh). This strain grew in the presence of 500 µM vanillate but at a reduced rate compared to WT (Fig. 4a) while a derivative of this strain in which the Pnat_cdbA-mCh construct was also integrated at the Mx8 attB site grew like WT in the absence of vanillate (Supplementary Fig. 3b, c). We conclude that the CdbA-mCh protein is fully active and that the reduced growth rate of the ΔcdbA/ Pvan_cdbA-mCh strain is likely caused by the lower accumulation of CdbA-mCh. Removal of vanillate caused growth arrest of ΔcdbA/ Pvan_cdbA-mCh cells after 24 h correlating with the earliest time-point at which CdbA-mCh was no longer detectable by immunoblotting and also caused a four-log defect in plating efficiency (Fig. 4a, e). Cells of the CdbA-mCh depletion strain were slightly longer than WT cells in the presence of vanillate, while cells grown under CdbA-mCh depleting conditions for 24 h were highly filamentous (Fig. 4b, c). These cells still formed constrictions at midcell but failed to complete division and eventually lysed (Fig. 4b). We conclude that CdbA is essential for viability and that lack of CdbA causes a cell division defect.
To investigate the relationship between CdbA and CdbB, we generated a ΔcdbAB double deletion in the strain containing the Pvan_cdbA-mCh construct (ΔcdbAB/Pvan_cdbA-mCh). In the presence as well as in the absence of vanillate, cells of this strain had an even more significant growth and cell division defect than the ΔcdbA/Pvan_cdbA-mCh cells (Fig. 4a-c) suggesting that CdbB can partially substitute CdbA. Interestingly, CdbA and CdbB interact in a bacterial two-hybrid assay suggesting that they can form heterooligomers ( Supplementary Fig. 3d). Altogether, these observations indicate that CdbA and CdbB have similar functions. However, CdbA alone is required and sufficient for viability while CdbB is neither required nor sufficient for viability.
CdbA binds the nucleoid globally and sequence-specifically. Because CdbA is a member of the RHH family of DNA binding proteins, we asked whether CdbA binds DNA. In fluorescence microscopy analyses of WT/Pnat_cdbA-mCh cells, CdbA-mCh colocalized with the DAPI-stained nucleoid in M. xanthus (Fig. 5a). Similarly, CdbA-mCh colocalized with the nucleoid in E. coli (Fig. 5b) supporting that CdbA binds DNA in vivo.
We used ChIP-seq to identify CdbA binding sites at a genomewide scale in M. xanthus using a strain in which cdbA at the native locus was replaced with an allele encoding a C-terminally 3×FLAG-tagged CdbA variant. This strain grew like WT, indicating that the protein was fully functional ( Supplementary  Fig. 4a). As a positive control, we used a strain expressing 3×FLAG-tagged ParB from the native locus and as a negative control, WT without a FLAG-tagged protein. In the two replicates, the WT control only showed a few peaks enriched ≥4-fold over input (Fig. 5c), while the ParB-3×FLAG control showed a broad peak in the region containing the parS sites ( Fig. 5c). The ChIP-seq peaks for CdbA-3×FLAG revealed broad occupancy over the M. xanthus chromosome (Fig. 5c). For both 3×FLAG-tagged proteins, the ChIP-seq peaks were strongly correlated between the two replicates while no correlation was observed for the WT control (Fig. 5c).
We detected 569 peaks with ≥4-fold enrichment over input for CdbA in intergenic regions and within structural genes (Fig. 5c, Supplementary Data 2). To experimentally verify the ChIP-seq results, we performed electrophoretic mobility shift assays (EMSAs) with purified CdbA-His 6 and 250 bp fragments corresponding to peak rank #1, #8, #161, and a fragment with no peak as a control. CdbA bound to the peak-fragments with an affinity that decreased with decreasing peak rank (K d from~0.4 tõ 0.9 µM) and the fragment with no peak had significantly lower affinity (K d of~4 µM) (Fig. 6a) suggesting that CdbA binds with some sequence specificity and non-specifically at high concentration. Consistently, at the highest concentration of CdbA-His 6 (4 µM) the shifted complexes tended to run higher in the gels than at lower concentrations supporting that more than one CdbA molecule would bind to a DNA fragment (Fig. 6a).
To identify the CdbA consensus DNA binding sequence, we searched the 100-bp sequences centered around each peak summit using the MEME-ChIP web tool. Using the top 100 peak sequences, we identified a direct repeat motif separated by four bp, close to the peaks summits ( Fig. 6b), while only half of this motif was detected when we used the top 500 peak sequences (Fig. 6b).
To analyze the importance of the direct repeat, we performed EMSAs with a DNA fragment from peak rank #1 containing the WT sequence or fragments in which the two most conserved bp had been mutated in one or both repeats (Fig. 6c). While the WT sequence had a K d of~0.4 µM, the fragment with one repeat mutated had a K d of~1.1 µM and the fragment with two repeats mutated had a K d of~2.5 µM, close to that for unspecific binding by CdbA (K d~4 µM; Fig. 6a, c). These observations support that the identified motif is important for sequence-specific DNA binding by CdbA and that its conservation dictates CdbA affinity for DNA. The repeat fits well to the model of CdbA DNA binding in which the two DNA binding regions are separated by one helical turn. Using a hidden Markov model, we searched the M. xanthus genome for the presence of the direct repeat motif and detected 8917 sequences with a p-value < 0.00077. Out of the 569 ChIP-seq peaks, 296 (52%) have at least one identifiable direct repeat motif present in a distance ±50bp from the peak summit ( Supplementary Fig. 4c, Supplementary Data 2).
The model of DNA bound tetrameric CdbA, predicts that CdbA bends DNA when bound to the direct repeat (Fig. 2d). To test this prediction, we performed a circular permutation analysis using 250 bp fragments from peak rank #1 in which the direct repeat was systematically placed at different locations along the 250 bp. All DNA fragments showed the same migration with no CdbA bound (Supplementary Fig. 4d) while CdbA binding induced a more pronounced shift in the DNA fragment containing the motif towards the center of the fragment in comparison to fragments in which the binding site was positioned towards the end, demonstrating that CdbA bends DNA upon binding (Supplementary Fig. 4d).
Using quantitative immunoblot analysis, we determined the number of CdbA monomers per cell to~7000 corresponding to a concentration of 2.2 µM for the tetramer supporting that CdbA is able to broadly occupy the M. xanthus genome ( Supplementary  Fig. 4e). We conclude that CdbA binds to >500 sites along the M. xanthus chromosome, that CdbA likely binds DNA as a tetramer with a consensus sequence consisting of a direct repeat and causes DNA bending upon binding.

CdbA depletion causes only minor changes in transcription.
198 of the CdbA ChIP-seq peaks (~35%) mapped to intergenic regions, and among the top 100 peaks, 47% were intergenic (Supplementary Data 2) while less than 10% of the M. xanthus genome is intergenic 42 . Genes downstream from the 198 intergenic peaks encode proteins annotated as hypothetical or involved in various cellular functions and no specific functional category was enriched (Supplementary Data 2). Cell division defects resulting in cell filamentation can be caused by defects in divisome assembly 29,30,43,44 , peptidoglycan synthesis 29 , DNA replication as well as DNA damage [45][46][47][48][49] , chromosome segregation or chromosome organization 50,51 . Because CdbA-depletion caused cell filamentation with incomplete cell division, we searched all 569 peaks for genes known to be involved in such functions and with a peak in the predicted promoter region but did not identify any (Supplementary Data 2).
To determine whether CdbA functions as a transcriptional regulator, we performed RT-qPCR analysis on total RNA isolated from WT and CdbA-mCh-depleted cells. Among the nine genes tested, which all had a ChIP-seq peak in the predicted promoter WT was used as a negative control. The log2 enrichment ratio was calculated from IP DNA and input DNA and plotted against location on the 9.28 Mb M. xanthus chromosome for one replicate. Grey line indicates no difference between sample and input. Red line indicates 4-fold difference (log2 = 2) set as a significance threshold for peaks. Right panels, scatter plots show correlation between two replicates for each strain. The log2 enrichment over input for replicate 1 is plotted against log2 enrichment over input for replicate 2 for each genomic position. The correlation coefficient, R 2 , is indicated for each strain.
region, three showed minor but significant decreases in transcription upon CdbA-mCh depletion, but this effect did not correlate with peak rank; four were unaffected by CdbA depletion and two were not detectably expressed under the conditions tested (which were identical to those under which the ChIP-seq experiment was performed) (Fig. 6d). Thus, even though CdbA binds in the promoter regions of genes, it does not function as a classical transcriptional regulator consistent with the observation that most ChIP-seq peaks are within structural genes.
c-di-GMP inhibits DNA binding by CdbA. As noted, the regions in CdbA involved in c-di-GMP binding overlap with the regions predicted to be important for DNA binding (Figs. 2d and 3e). We analyzed whether c-di-GMP affects DNA binding by CdbA using a DNA fragment that covers 250 bp upstream of the cdbA translational start codon. CdbA-His 6 binds this fragment with a K d of 0.9 µM (Fig. 6a). Increasing c-di-GMP concentrations reduced CdbA DNA binding and at 16 µM, DNA binding was completely abolished (Fig. 7a). Because CdbA binds c-di-GMP with a K d of 83 nM while c-di-GMP only reduced CdbA DNA binding in the µM range, we performed an order of addition experiments in which we first added c-di-GMP to CdbA followed by DNA or vice versa. Independently of the order of addition, c-di-GMP had the same inhibitory effect on CdbA DNA binding ( Supplementary Fig. 5). This effect was specific to c-di-GMP as none of five other tested nucleotides had an effect on a b c d CdbA DNA binding (Fig. 7b). Similarly, c-di-GMP abolished binding to the other DNA fragments tested in EMSAs (Fig. 6a). CdbA K8A/S10A -His 6 and CdbA R27A/R30A -His 6 , which do not detectably bind c-di-GMP (Fig. 1c, d), also failed to bind the cdbAB promoter in EMSAs (Fig. 7c). These observations are in agreement with a model in which the same regions in CdbA are important for DNA binding and c-di-GMP binding and that binding of these two ligands is mutually exclusive.
To determine whether DNA binding and/or c-di-GMP binding by CdbA is important in vivo, we generated three strains in which (A) the native cdbA gene was deleted (ΔcdbA), (B) CdbA WT -mCh was expressed from the vanillate-inducible promoter, and (C) CdbA WT -mCh, CdbA K8A/S10A -mCh or CdbA R27A/R30A -mCh were expressed from the native promoter from the Mx8 attB site. Upon removal of vanillate, these three strains accumulate CdbA WT -mCh, CdbA K8A/S10A -mCh or CdbA R27A/R30A -mCh, respectively (Fig. 7d). While the strain accumulating CdbA WT -mCh grew as WT (Fig. 7e), the two strains accumulating CdbA K8A/S10A -mCh or CdbA R27A/R30A -mCh had a growth defect (Fig. 7e) and became filamentous (Fig. 7f) demonstrating that the variants are inactive in vivo and suggesting that DNA and/or cdi-GMP binding is essential for CdbA function in vivo. The two variants also failed to localize over the nucleoid in M. xanthus (Fig. 7f), confirming the in vitro result that they have a DNA binding defect (Fig. 7c).
CdbA depletion affects chromosome organization and segregation. Having ruled out that CdbA functions as a classical transcription factor and inspired by its abundance and global occupancy over the M. xanthus chromosome and essentiality, we hypothesized that CdbA could be a nucleoid-associated protein (NAP). We, therefore, analyzed nucleoid organization and segregation in CdbA-mCh-depleted cells using DAPI staining to assess nucleoid organization and a ParB-YFP fluorescent fusion as a marker for the origin and to assess chromosome organization and segregation 31 .
In the presence of vanillate, WT and the CdbA-mCh depletion strain showed similar DAPI staining patterns with a single nucleoid in short cells and two fully replicated and segregated nucleoids in longer cells (Fig. 8a). As previously observed 31-33 , most WT cells had two ParB-YFP signals at 25 and 75% of the cell length while slightly more cells of the CdbA-mCh depletion strain had more than two ParB signals (Fig. 8a, b). Also, the distance from the ParB-YFP clusters to the nearest cell pole was slightly but significantly longer than in WT (2.2 ± 0.8 vs 1.8 ± 0.6 μm) (Fig. 8c). As a control for the CdbA-mCh depletion strain in the absence of vanillate, WT cells were artificially elongated with 8 h of cephalexin treatment, which inhibits cell division without affecting chromosome replication, organization and segregation 29 ( Fig. 8a, b). As reported 29 , cephalexin treated WT cells had a regular distribution of nucleoids and ParB-YFP signals along the cell length. By contrast, in the CdbA-mCh depleted cells, nucleoids appeared more condensed, less well separated and mostly localizing in the center of the long cells ( Fig. 8a, b, Supplementary Fig. 6a). Also, ParB-YFP foci were not regularly distributed along the cell length but clustered in the center of cells paralleling the localization of the nucleoid (Fig. 8a, b, Supplementary Fig. 6a). Consistently, in these cells, the distance from the ParB-YFP clusters to the nearest cell pole was significantly larger than in cephalexin treated cells (5.3 ± 2.4 vs 3.7 ± 2.5 μm) (Fig. 8c, Supplementary Fig. 6b). This phenotype is different from ΔbacNOP and ΔpadC strains where localization of ParB-YFP foci was also affected but no cellular filamentation was observed 33 .
In time-lapse recordings of non-motile cells, we observed that segregation of ParB clusters was significantly slower and more erratic after CdbA-mCh depletion compared to that in WT [30][31][32][33] ( Supplementary Fig. 6c). Despite the abnormal localization of ParB-YFP clusters and the abnormal morphology and localization of the nucleoid, the number of ParB-YFP clusters per cell length increased similarly for cephalexin treated WT and the CdbA-mCh depletion strain (Fig. 8a) suggesting that DNA replication is unaffected by depletion of CdbA-mCh. These observations demonstrate that CdbA is important for chromosome organization and segregation in agreement with the idea that CdbA is a NAP.
Because DNA and c-di-GMP binding by CdbA is mutually exclusive in vitro, we speculated that if the essential function of CdbA lies in its NAP activity, then changing the cellular concentration of c-di-GMP would affect CdbA DNA binding and, therefore, chromosome organization and segregation. To test this hypothesis, we analyzed cell length distribution and nucleoid morphology in a strain that accumulates the heterologous DGC DgcA of Caulobacter crescentus, and which accumulates c-di-GMP at a 7-fold higher level than WT and in a strain that accumulates the heterologous PDE PA5295 of Pseudomonas aeruginosa and which accumulates c-di-GMP at a 2-fold lower level than WT 26 . These strains grow as WT and we neither observed an effect on cell length nor on nucleoid morphology ( Supplementary Fig. 7a, b).

Discussion
Here, we report the identification of the DNA binding RHH protein CdbA in M. xanthus, a previously undescribed type of cdi-GMP receptor, that globally binds the M. xanthus chromosome and is essential for viability. Cells depleted for CdbA have severe defects in chromosome organization and segregation and are impaired in cell division and highly filamentous. Taken together our data support a model whereby CdbA is an essential Fig. 6 CdbA binds DNA in a moderately sequence-specific manner. a CdbA binds DNA with sequence specificity. Upper panels, zoom in on ChIP-seq results for selected DNA fragments from different peaks sorted by peak rank and shown as log2 enrichment over input for one replicate. ORFs are shown as black arrows. Horizontal red line indicates 250 bp fragments used for EMSA analysis with position of peak summit indicated with a vertical red line. Lower panels, EMSA analysis of CdbA binding to DNA fragments shown in upper panels. Concentration of CdbA added is indicated. c-di-GMP was added to a final concentration of 100 µM in the last lane, 10 min prior addition of DNA probe. K d was calculated based on two replicates. Source data are provided as a Source Data file. b MEME-ChIP results for CdbA binding motif search. Left panels, motif identified based on top 100 peaks and the probability distribution of the identified motif occurrence across the input sequences in which position 0 corresponds to peak summits; right panels, as in the left panels except that the top 500 peaks were analyzed. For both analysis 100 bp (peak summit ±50 bp) were used as input sequences for the MEME-ChIP analysis. c EMSA analysis of CdbA binding to WT and mutant DNA fragments of MXAN_4058 promoter. Below each image, the sequence of the WT binding motif and the two mutant variants are indicated with substituted bp indicated in red. Experiments were done as in (a) and K d calculated based on two replicates. Source data are provided as a Source Data file. d RT-qPCR analysis of gene expression in WT and a strain depleted for CdbA-mCh for 24 h. Peak rank is indicated. Transcript levels are shown as mean ± SD from three biological replicates with each three technical replicates relatively to WT. *p < 0.05 in two-sided Student's t-test, between WT and CdbA-mCh depletion. Individual data points are in green-gray. Exact p-values and source data are provided in the Source data file. transcription in enterobacteria, shows sequence-specific DNA binding to~500 AT-rich regions and affects chromosome organization and transcription by forming DNA loops 56 . GapR in C. crescentus is an essential NAP that recognizes AT-rich sequences, binds to~600 regions across the genome preferentially associating with intergenic regions. GapR depletion causes formation of filamentous-and minicells but does not cause major changes in transcription 57,58 . By comparison CdbA binds DNA with some sequence specificity and a K d of 0.4-4.0 µM, which is significantly lower than for other RHH proteins that have K d s in the range of 10-200 nM [59][60][61] . Forty-seven percent of the top 100 ChIP-seq peaks are in intergenic regions; however, RT-qPCR experiments comparing transcription in WT and CdbA-depleted cells either revealed modest effect, no effect or no expression of genes with a CdbA binding site in the promoter region suggesting that CdbA does not function as a classical transcriptional regulator. Also, the putative functions of proteins encoded by genes with a ChIP-seq peak in the promoter region did not show enrichment of functional categories as would be expected for a classical transcriptional regulator. Moreover, these genes did not include those encoding proteins involved in cell growth, cell division, chromosome organization and/or segregation that could explain the chromosome and cell division defects in cells depleted of CdbA. Additionally, CdbA binds the 9.28 Mb M. xanthus genome at >500 sites, bends DNA upon binding, is highly abundant with 1700 tetramers per cell, and is essential. Altogether, these observations are consistent with the notion that CdbA is a NAP with an essential function in chromosome organization and segregation. According to this model, the primary function of CdbA is that of a NAP to assist in organizing the chromosome and support chromosome segregation while the effects on transcription and cell division are likely secondary to the defect in chromosome organization. In contrast to other NAPs, the CdbA consensus binding site is GC-rich. We speculate that this might be explained not only by the GC-rich nature of M. xanthus genome (69% GC content) but also by the mutually exclusive binding of DNA and c-di-GMP potentially allowing the guanine bases of c-di-GMP and GC-rich DNA fragment to compete for the same binding site on CdbA (see below). Other NAPs are typically regulated by abundance 52,53 ; whether CdbA is also regulated by abundance is not known; however, our data support the notion that CdbA's DNA binding activity is regulated by c-di-GMP, i.e., CdbA is a ligand-regulated NAP.
In vitro c-di-GMP binding by the CdbA tetramer induces conformational changes. C-di-GMP and DNA binding are mutually exclusive and the two ligands appear to compete for the same binding interfaces on CdbA. In other RHH proteins the βsheet provides specificity in DNA binding, while the loop region between α-helix1 and α-helix2 together with the N-terminus of α-helix2 make DNA backbone contacts and increases the affinity for DNA 37 . Precisely these two regions show a modified response upon addition of c-di-GMP in the HDX experiment. The 1:2 ratio of c-di-GMP:CdbA, HDX data of binding interfaces, and our structure of a tetramer formed from two RHH dimers allow the postulation of two models for c-di-GMP binding. The first model would place one c-di-GMP molecule directly over each β-sheet in the tetramer; this singular dinucleotide would adopt either the extended conformation observed in EAL PDEs 62 or the clamplike conformations adopted for example in the HD-GYP domain protein PmGH 63 . In this model, the alterations in HDX of the α-helix1-2 loop could be explained by a change in the relative orientation of the two CdbA dimers in the tetramer upon c-di-GMP binding. A second, intriguing possibility is that the unique dimer:dimer pocket formed at the center of the tetramer acts to bind the commonly encountered intercalated dimer form of c-di-GMP 64 , and the two β-sheets flex inward to explain their implication in the HDX data. Flexation would also allow the helical turn (residues 29-34) in the center of the tetramer to recognize the c-di-GMP phosphate group in the same manner that RHH proteins use to bind the DNA backbone. The tetramer interface displays different hydrophobic packing between the two crystal forms we observed here, suggesting that it may have inherent plasticity that allows it to undergo this change. In total, the first model would rely on steric blockage to enable c-di-GMP to license CdbA release from DNA, whereas the second model could modulate DNA-binding via c-di-GMP alteration of the β-sheet:βsheet distance. RHH domains are found as stand-alone RHH domains or together with other domains 37 . Ligand binding and modulation of DNA binding by other RHH proteins has only been demonstrated for RHH proteins with additional domains and the ligands bind to these extra domains 65,66 . Thus, CdbA is the first RHH protein that binds a ligand (other than DNA) via the RHH domain. In P. aeruginosa AmrZ, one of the best studied RHH proteins, is involved in regulating the c-di-GMP level 67 . However, we did not observe c-di-GMP binding to AmrZ (Supplementary Fig. 8) documenting that c-di-GMP binding is not a general feature of RHH proteins.
In contrast to CdbA, other c-di-GMP binding transcriptional regulators use distinct domain for c-di-GMP and DNA binding. C-di-GMP may stimulate DNA binding by stimulating oligomer formation as is the case for VpsT of Vibrio cholerae 18 , BldD of Streptomyces coelicolor 19 and CuxR of Sinorhizobium meliloti 20 . Alternatively, upon c-di-GMP binding, FleQ of Pseudomonas aeruginosa undergoes a conformational change causing reduced DNA binding 21 .
In the absence of DNA, CdbA binds c-di-GMP with a K d8 3 nM; and CdbA binds DNA fragments with K d~0 .4-4 µM (depending on the fragment used). Nevertheless, in vitro c-di-GMP only reduced CdbA DNA binding when added in the µM range indicating the complexity of this interplay; there is a possibility that DNA-binding changes CdbA structure/dynamics such that c-di-GMP binding is less favorable than when DNA is absent (i.e., this is not represented by a simple competition model, and can be enacted by an extensive interface for DNA binding Fig. 8 Lack of CdbA affects chromosome organization and segregation. a CdbA is important for chromosome organization and segregation. First row, demographs with overlay of DAPI-stained nucleoids and ParB-YFP localization (red circles) in strains of indicated genotypes and in the presence and absence of vanillate and cephalexin as indicated. In the absence of vanillate, cells were imaged 24 h after vanillate removal; cephalexin was added for 8 h. Second row, box plot showing number of ParB-YFP foci per cell as a function of cell length. In the boxplots, boxes enclose the 25th and 75th percentile with the black line representing the mean, whiskers indicate the 10th and 90th percentile, and "+" indicates outliers. Numbers in bars indicate the percentage of analyzed cells having the indicated number of ParB-YFP foci. Third row, scatter plots of position of ParB-YFP clusters along the long axis of cell in different strains as a function of cell length. n = 180 for each strain and for each analysis. Source data are provided as a Source Data file. b Images of representative cells used for the analysis in a. Scale bar, 10 μm. c Box plot showing distance between ParB-YFP clusters and the closest pole as a function of cell length for cells with at least two ParB-YFP clusters. The same 180 cells as in a were analyzed and only the ParB-YFP clusters closest to a pole were used in the analysis. Box plot as in a. Numbers above each box indicate number of cells used for quantification (n) and mean ± SD. **p < 0.001 in two-sided Student's ttest. Exact p-values and source data are provided in the Source data file.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15628-8 ARTICLE NATURE COMMUNICATIONS | (2020) 11:1791 | https://doi.org/10.1038/s41467-020-15628-8 | www.nature.com/naturecommunications that is more than just the two β-sheets). The estimated c-di-GMP concentration in M. xanthus cells in rich medium is~1.4 ± 0.5 µM 26 , suggesting that c-di-GMP binding by CdbA should be relevant in vivo. However, strains in which the c-di-GMP level was artificially increased or decreased 7-or 2-fold, respectively, have no defects in growth, chromosome organization/segregation or cell division. Based on these observations and the observation that inactivation of CdbA has a dramatic effect on viability, we speculate that alteration in c-di-GMP may not elicit an all-ornone response with respect to CdbA DNA binding but rather a graded response in which c-di-GMP binding to CdbA modulates or fine-tunes CdbA DNA binding in vivo. Alternatively, exposure of CdbA to c-di-GMP is controlled by complex formation with specific DGCs/PDEs creating a local c-di-GMP pool. In this scenario, manipulating the global c-di-GMP level would not affect CdbA-DNA binding. Of note, because the~10-fold increase in cdi-GMP during development is essential for fruiting body formation and sporulation 27 and spores are diploid 68 it is interesting to speculate that CdbA could participate in development specific chromosome reorganization and cell division inhibition. In the future, it will be of interest to identify other conditions that cause significant changes in the level of c-di-GMP.
In conclusion, we have identified the NAP CdbA in M. xanthus and our data support that its activity is modulated by the second messenger c-di-GMP, thus, revealing a link between c-di-GMP signaling and chromosome biology. Motility and development assays. For motility assays, cells were grown in CTT medium, harvested, and resuspended in 1% CTT to a calculated density of 7 × 10 9 cells ml −1 . Five-microliter aliquots of cell suspensions were placed on 0.5% and 1.5% agar supplemented with 0.5% CTT and incubated at 32°C. After 24 h, colony edges were observed using a Leica MZ8 stereomicroscope or a Leica IMB/E inverted microscope and visualized using Leica DFC280 and DFC350FX charge-coupleddevice cameras, respectively. For development assay, cells were harvested and resuspended in MC7 buffer (10 mM MOPS pH 7.0, 1 mM CaCl 2 ) to a calculated density of 7 × 10 9 cells ml −1 . Twenty microliter aliquots of cells were placed on TPM agar (10 mM Tris-HCl pH 7.6, 1 mM K 2 HPO 4 /KH 2 PO 4 pH 7.6, 8 mM MgSO 4 ); for development in submerged culture, 50 μl of the cell suspension were mixed with 350 μl MC7 buffer and placed in an 18 mm diameter microtiter dish. Cells were visualized at the indicated time points using a Leica MZ8 stereomicroscope or a Leica IMB/E inverted microscope and imaged using Leica DFC280 and DFC350FX CCD cameras, respectively. Sporulation levels were determined after development for 120 h in submerged culture as the number of sonication-and heat-resistant spores relative to WT.
Fluorescence microscopy and live cell imaging. Exponentially growing cells were transferred to slides containing a thin pad of 1.5% agarose (Cambrex) with TPM buffer (10 mM Tris-HCl pH 7.6, 1 mM KH 2 PO 4 pH 7.6, 8 mM MgSO 4 , 0.2% CTT, covered with a coverslip and imaged with a temperature-controlled Leica DMi8 inverted microscope. Phase contrast and fluorescence images were acquired using a Hamamatsu ORCA-flash V2 Digital CMOS camera. Cells in phase contrast images were automatically detected using Oufti software 72 . Fluorescence signals were identified and analyzed using a custom-made Matlab v2016b (MathWorks) script. E. coli cells were induced with 0.05 mM isopropyl-β-D-thiogalactopyranosid (IPTG) for 2 h and treated with 30 μg ml −1 chloramphenicol for 30 min before DAPI staining. For DAPI staining, cells were incubated with 1 mg ml −1 DAPI for 10 min at 32°C prior to start of microscopy. Image processing was performed using Metamorph_ v 7.5 (Molecular Devices).
Bacterial two-hybrid assay (BACTH). BACTH experiments were performed as described 73 . Briefly, full-length cdbA and cdbB were cloned into the appropriate vectors to construct N-terminal and C-terminal fusions with the 25-kDa Nterminal or the 18-kDa C-terminal adenylate cyclase fragments. cAMP production was observed by the formation of blue color on LB agar supplemented with 80 µg ml −1 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) and 0.25 mM IPTG.
Operon mapping. Total RNA was isolated using a Trizol (Sigma) extraction method. RNA was treated with DNase I (Invitrogene) and purified with the RNeasy kit (Qiagen). PCR analysis was used to confirm that the RNA was DNA free. One microgram of RNA was used to synthesize cDNA with the High capacity cDNA Archive kit (Applied Biosystems) using random hexamer primers. For the operon mapping, following primer pairs were used: "4361 qPCR forw"/"4361 qPCR rev" (fragment 1), "4362 qPCR forw"/"4362 qPCR rev" (fragment 2) and "4361 qPCR forw"/"4362 qPCR rev" (fragment 3). Genomic DNA and RNA were used as positive and negative controls, respectively.
Real-time PCR. Total RNA was isolated and cDNA was synthetized as described for operon mapping. qRT-PCR was performed in 25 μl reaction volume using SYBR green PCR master mix (Applied Biosystems) and 0.1 μM primers specific to the target gene in a 7500 Real Time PCR System (Applied Biosystems). Relative gene expression levels were calculated using the comparative Ct method. All experiments were done with three biological replicates each with three technical replicates.
Determination of protein copy number per cell. Number of CdbA molecules per cell was determined using a quantitative immunoblot analysis. Exponentially growing WT culture was used to prepare cell lysates for immunoblot analysis. Different amounts of cell lysates and purified protein were separated on a 15% SDS-gel, transferred to a nitrocellulose membrane and probed with α-FLAG antibodies. Signal intensities of the lysate bands were quantified using Fiji 75 and compared against a standard curve generated from known amounts of His 6 _Cd-bA_3×FLAG from the same Immunoblot. Experiment was performed in two independent biological replicates.
Electrophoretic mobility shift assay (EMSA). The HEX-labelled DNA fragments were generated by PCR. Reactions were carried out in 10× EMSA buffer (380 mM HEPES, 380 mM NaCl, 50 mM MgCl 2 , 50% glycerol, 10 mM DTT, 15 ng µl −1 poly (dI-dC) (Sigma-Aldrich) in the presence of indicated concentrations of CdbA_His 6 and 50 ng DNA probe. C-di-GMP was added at indicated concentrations and order of addition of c-di-GMP and DNA is indicated in figure legends. Reactions were incubated at 30°C for 30 min. Samples were loaded on a 10% polyacrylamide gel, and electrophoresed in 1× TBE at 4°C. The gel was pre-run in 1× TBE for 0.5 h. Gels were scanned for HEX signals on a Typhoon Scanner (GE Healthcare).
Circular permutation analysis. Circular permutated DNA fragments were generated with PCR. Gel shift was performed as for EMSA assay but instead of using fluorescence, signal was detected using GelRed dye.
Structure determination. Concentrated CdbA (28 mg ml −1 ) was used in sitting drop crystallization screening experiments. Protein (1.8 µl) was mixed with an equal volume of crystallization condition. Initial protein crystal needles, unsuitable for X-ray diffraction experiments, were obtained in 0.8 M Na Formate, 25% w/v PEG 2000 MME, 0.1 M tri-sodium citrate, pH 6.5. This condition was selected to seed for novel crystallization conditions via a micromatrix seeding protocol 78 . Briefly, the crystals from this sitting drop were harvested and vortexed in 500 µl of the condition mother liquor to generate a crystal seed stock. New sitting drop screening was performed with drops comprised of 1.8 µl protein and 1.2 µl crystallization condition which were then supplemented with 0.6 µl of seed stock. This screening approach yielded a dramatic improvement on the initial screening with a large number of three dimensional protein crystals. Crystals of the dimeric form of CdbA were grown in 0.1 M MgCl 2 , 0.1 M HEPES-NaOH, pH 7.0, 15% w/v PEG 4000 supplemented with seed stock. Crystals grown in this condition were cryoprotected by sequential addition of mother liquor supplemented with 25% (v/v) glycerol prior to flash freezing in liquid nitrogen. A heavy atom derivative of this protein crystal was prepared by sequential addition of mother liquor supplemented with 5 mM K 2 PtCl 4 . After 70 min, the derivatized crystals were cryoprotected and harvested as described for the native form. The tetrameric form of CdbA crystallized in 0.1 M sodium malonate dibasic monohydrate, 0.1 M HEPES-NaOH, pH 7.0, 30% w/v polyacrylic acid sodium salt 2100 and were cryoprotected in mother liquor supplemented with 20% w/v ethylene glycol. Diffraction data were collected at the Diamond Light Source in Oxford, UK. Data reduction and processing was achieved using Xia2 79 . The SAD and molecular replacement phases were calculated using scripts in the PHENIX suite of programs 80 . Protein structures were built/ modified using COOT 81 , with cycles of refinement in both PHENIX and PDB-REDO 82 . The low sequence identity (15-20%) with known RHH structures enforced us to use of de novo phasing, which was provided by SAD data collected at the LIII edge of a Platinum derivative of form one. The derivative gave a clear result (Pt modification of residue M17, FOM 0.36) and allowed tracing of residues 5-52.
c-di-GMP capture compound experiments. Experiments were performed as described 19,35 . Briefly, M. xanthus cultures were grown in liquid CTT medium at 32°C until exponential phase and harvested by centrifugation for 10 min at 4150 × g. The cell pellet was resuspended in lysis buffer (6.7 mM 2-(N-morpholino)ethanesulfonic acid (MES), 6.7 mM (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 200 mM NaCl, 6.7 mM potassium acetate (KAc), pH 7.5) containing cOmplete™, Mini Protease Inhibitor Cocktail (Roche), resulting in 250 times concentrated cells. Cells were lysed three times through a French Press and then centrifuged at 109,760 × g rpm for 1 h at 4°C. The protein concentration of the supernatant was determined using a Bradford protein assay and 233 μg protein were mixed with 20 μl 5× capture buffer (100 mM HEPES, 250 mM KAc, 50 mM magnesium acetate (MgAc), 50% glycerol, pH 7.5) and 10 μM c-di-GMP capture compound. The reaction volume was adjusted with H 2 O to 100 μl. In the negative control, c-di-GMP capture compound was replaced by 1× capture buffer. In the second competition control, 1 mM c-di-GMP (BioLog) was added and incubated for 30 min prior to capture compound addition. Reactions were incubated for 2 h at 4°C in the dark on a rotary wheel. After UV irradiation for 4 min at 4°C in a caproBox, 50 μl magnetic streptavidin beads (Thermo Fisher Scientific) and 25 μl 5× wash buffer (250 mM Tris HCl pH 7.5, 5 M NaCl, 0.1% n-octyl-β-glucopyranoside) were added to the reaction and the mixture was incubated for 1 h at 4°C on a rotary wheel. The beads were then collected with a magnet and washed six times with 200 μl 1× wash buffer, one time with 200 μl H 2 O and six times with 80% acetonitrile and resuspended in 200 μl H 2 O. 7 µl 1 M ammonium bicarbonate, 17 µl 100 mM DTT and 5.1 µl 0.1 μg μl −1 Trypsin (Promega) were added and the beads incubated for 12 h at 25°C. Beads were removed with a magnet, 3 μl of AcOH was added and samples dried in a speedvac for 50 min. Captured proteins were identified by MALDI-TOF-TOF mass-spectrometry analysis as described 83 .
In vitro c-di-GMP binding assays. In the DRaCALA assay 36,84 [α-32 P]-c-di-GMP was mixed with 20 µM of the relevant protein and incubated for 10 min at 30°C in binding buffer (10 mM TrisHCl pH 8.0, 100 mM NaCl, 5 mM MgCl 2 ). Ten microliter of this mixture was transferred to a nitrocellulose filter (GE Healthcare), allowed to dry and imaged using a STORM 840 Scanner (Amersham Biosciences). Hydrogen-deuterium exchange mass-spectrometry (HDX-MS). Sample preparation for HDX-MS was aided by a two-arm robotic autosampler (LEAP Technologies). CdbA (7.5 µl, 50 µM) with or without 500 µM c-di-GMP (BioLog) was mixed with 67.5 µl D 2 O-containing SEC buffer (20 mM of HEPES-Na, pH 7.5, 200 mM NaCl 20 mM MgCl 2 , 20 mM KCl). After incubation for 10/30/95/1000/ 10000 s at 25°C, 55 µl of the HDX reaction were added to 55 µl quench solution (400 mM KH 2 PO 4 /H 3 PO 4 , 2 M guanidine-HCl, pH 2.2) kept at 1°C and 95 µl of the mixture injected into an ACQUITY UPLC M-class system with HDX technology (Waters) 88 . Peptides were generated online using immobilized pepsin at 12°C and 100 µl/min flow rate of water + 0.1% (v/v) formic acid and the resulting peptic peptides trapped on a C18 column (Waters) kept at 0.5°C. After 3 min, the trap column was placed in line with an ACQUITY UPLC BEH C18 1.7 µm 1.0 × 100 mm column (Waters) and the peptides eluted at 0.5°C using a gradient of water + 0.1% (v/v) formic acid (eluent A) and acetonitrile + 0.1% (v/v) formic acid (eluent B) at 30 µl/min flow rate: 0-7 min/95-65% A, 7-8 min/65-15% A, 8-10 min/ 15% A, 10-11 min/5% A, 11-16 min/95% A. Undeuterated samples were obtained by similar procedure through dilution in H 2 O-containing SEC buffer. Mass spectra were recorded in positive ion mode using a Synapt G2-Si HDMS mass spectrometer equipped with an ESI source (Waters) in HDMS (High Definition-MS) or HDMS E (Enhanced High Definition MS) 89,90 mode for deuterated and undeuterated samples, respectively. [Glu1]-Fibrinopeptide B standard (Waters) was used for lock mass correction. The pepsin column was washed three times with 80 µl of 4% (v/v) acetonitrile and 0.5 M guanidine-HCl during each run and additionally blank runs were performed between each sample to avoid peptide carry-over. All measurements were performed in triplicates. Peptide identification and assignment of deuterium incorporation was carried out as described previously [91][92][93] aided by PLGS and DynamX 3.0 softwares (Waters). A total of 68 peptides were obtained that covered the entire sequence of CdbA with 11-fold redundancy per amino acid. HDX-MS data can be found in the Supplementary Data 1 94,95 .
ChIP-seq data analysis. CLC workbench 12.0 (Qiagen, Hilden, Germany) was used for computational processing of sequencing data. The functions 'Trim Reads', 'Map Reads to Reference', 'Duplicate Mapped Reads Removal', 'Transcription Factor ChIP-Seq' & 'Annotate with Nearby Gene Information' were applied with default settings. ChIP-Seq peak calling in CLC workbench is based on a shape learning algorithm 97 .
Here every two samples were compared individually to their respective controls where the genome assembly of Myxococcus xanthus DK1622 (Accession: NC_008095) obtained from NCBI was used as reference. Sequence coverage of peaks was inferred via samtools 98 (option 'mpileup', Version 1.9). Peaks were considered significant if the enrichment in the sample over input was ≥4-fold. The DNA binding consensus of CdbA was identified using sequences from the top 100 or top 500 peaks (peak summit ±50 bp) and the MEME-ChIP web tool 99 . The probability matrix from the MEME-ChIP analysis using the top 100 peaks served as an input to create the Hidden Markov Model used for the genome-wide search for putative CdbA binding sites with a given score. P-value of identified sequences was determined based on in silico simulation. In this simulation 1,000,000 of 18 bp random sequences with the same GC content as M. xanthus genome were screened for the presence of the same probability matrix with a given score. A motif was assigned to the peak if it was found within a range of a peak summit ±50 bp.
Bioinformatics. Homology based structure prediction for CdbA and CdbB was performed using HHPred 100 . 16 s rRNA and protein sequences were aligned with ClustalW using MEGA7 101 and the 16 s rRNA phylogenetic tree was generated using the Maximum Likelihood method. CdbA/B homologs were identified using BLASTP 102 analysis and the sequence identity/similarity was calculated using EMBOSS Needle software 103 (pairwise sequence alignment).
Statistics. All statistics were performed using a two-tailed Student's t-test for samples with unequal variances using Sigmaplot 12.5 and Microsoft Excel 2013.
Plasmid construction. Primers used in this study are listed in Supplementary  Table 4. pDJS 83 (for overexpression and purification of CdbB), cdbB was amplified from DK1622 genomic DNA with the primer pair "4362 F"/"4362 -stop R", digested with NdeI and HindIII, cloned into pET24b+ digested with the same enzymes and sequenced. pDJS 85 (for generation of in-frame deletion of cdbA-B), up-and downstream fragments were amplified from DK1622 genomic DNA using the primer pairs "4361-2_A"/"4361-2_B (5aa)" and "4361-2_C (5aa)"/"4361-2_D", as described 104 . The AB and CD fragments were used for overlapping PCR with the primer pair "4361-2_A"/"4361-2_D" to generate the AD fragment. The AD fragment was digested with KpnI and XbaI, cloned into pBJ114 digested with the same enzymes and sequenced. pDJS 86 (for overexpression and purification of CdbA), cdbA was amplified from DK1622 genomic DNA with the primer pair "4361 Fw"/"4361 -stop Rev", digested with NdeI and HindIII, cloned into pET24b+ digested with the same enzymes and sequenced. pDJS 97 (for generation of in-frame deletion of cdbB), up-and downstream fragments were amplified from DK1622 genomic DNA using the primer pairs "4362_A"/"4362_B 5aa" and "4362_C 5aa"/"4361-2_D", as described 104 . The AB and CD fragments were used for overlapping PCR with the primer pair "4362_A"/ "4361-2_D" to generate the AD fragment. The AD fragment was digested with KpnI and XbaI, cloned into pBJ114 digested with the same enzymes and sequenced. pDJS 99 (for generation of in-frame deletion of cdbA), up-and downstream fragments were amplified from DK1622 genomic DNA using the primer pairs "4361-2_A"/"4361_B 5aa" and "4361_C 5aa"/"4361_D 5aa" as described 104 . The AB and CD KpnI"/"4361 -st R BamHI", digested with KpnI and BamHI, cloned into pNG62 digested with the same enzymes and sequenced.
pDJS 177 (for overexpression and purification of His 6 _CdbA_3xFLAG), the cdbA_3xFLAG fragment was amplified from pDJS 170 with the primer pair "4361 no start NdeI"/"4361 3xFLAG stop BamHI", digested with NdeI and BamHI, cloned into pET28a+ digested with the same enzymes and sequenced.
pDJS 179 (for replacement of parB with parB-3xFLAG at native site), up-and downstream fragments were amplified from DK1622 genomic DNA using the primer pairs "ParB 3xFLAG A"/"ParB 3xFLAG B" and "ParB 3xFLAG C"/"ParB 3xFLAG D". Resulting fragments were used as a template for overlapping PCR with the primer pair "ParB 3xFLAG A"/"ParB 3xFLAG D". Resulting fragment was digested with EcoRI and XbaI, cloned into pBJ114 digested with the same enzymes and sequenced.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Co-ordinates and structure factors have been deposited with the RCSB: accession codes of 6SBW (form one) and 6SBX (form two). The ArrayExpress accession number for the ChIP-seq experiment is E-MTAB-8535. The HDX-MS data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD018028. The source data underlying Figs. 1b, d, 4a, c-e, 6a, c, d, 7a-e, 8a, c and Supplementary Figs. 3b-d, 4a, d, e, 5, 6b, 7a, 8a, b are provided as a Source Data file.

Code availability
Custom MATLAB scripts used for ChIP-seq data analysis and for microscopy analysis are available from the corresponding author upon request.