Relief of ParB autoinhibition by parS DNA catalysis and ParB recycling by CTP hydrolysis promote bacterial centromere assembly

Three-component ParABS systems are widely distributed factors for plasmid partitioning and chromosome segregation in bacteria. ParB protein acts as an adaptor between the 16 bp centromeric parS DNA sequences and the DNA segregation ATPase ParA. It accumulates at high concentrations at and near a parS site by assembling a partition complex. ParB dimers form a DNA sliding clamp whose closure at parS requires CTP binding. The mechanism underlying ParB loading and the role of CTP hydrolysis however remain unclear. We show that CTP hydrolysis is dispensable for Smc recruitment to parS sites in Bacillus subtilis but is essential for chromosome segregation by ParABS in the absence of Smc. Our results suggest that CTP hydrolysis contributes to partition complex assembly via two mechanisms. It recycles off-target ParB clamps to allow for new attempts at parS targeting and it limits the extent of spreading from parS by promoting DNA unloading. We also propose a model for how parS DNA catalyzes ParB clamp closure involving a steric clash between ParB protomers binding to opposing parS half sites.


Introduction
Faithful transmission of genetic material from one cell generation to the next is a prerequisite for survival and propagation. Chromosome segregation is spatially and temporally coordinated with DNA replication, cell division, and cell growth to maintain genome integrity and cell architecture over multiple generations. In bacteria, three-component ParABS systems promote faithful and efficient chromosome segregation (1-3). They are also important regulators of cell physiology with mutations leading to diverse and pleiotropic phenotypes in addition to chromosome partitioning errors. These phenotypes include defects in the control of DNA replication, chromosome organization, cell division, gene expression, motility, sporulation, and competence development (1, [4][5][6][7][8][9][10][11][12][13][14]. ParABS systems are present on most bacterial genomes and are also frequently encoded by low-copy number plasmids (15). They comprise 16 bp DNA sequence elements, called parS sites, the DNA-binding protein ParB, and the ATP-hydrolysing protein ParA. ParB proteins and parS sites together form the partition complex that serves as chromosome organizing centre ('parS centromere') (3,16). Partition complexes are thought to follow ParA protein gradients on the bacterial chromosome thus becoming equidistantly positioned within the cell (12,17,18). They stimulate ATP hydrolysis by DNAbound ParA dimers converting them into ParA monomers that dissociate from chromosomal DNA (5,(19)(20)(21). Using the same 'diffusion-ratchet' mechanism, ParABS is thought to promote plasmid partitioning (18,22). The parS centromere also instructs another active DNA segregation mechanism, the Smc DNA loop extrusion motor. Starting from parS centromeres, the Smc complex aligns the two chromosome arms, helping to segregate nascent sister chromosomes (6,(23)(24)(25)(26). ParABS furthermore regulates the initiation of DNA replication via the initiator protein DnaA in B. subtilis. Inability to convert ParA ATP-dimers into monomers (e.g. in a ΔparB mutant) leads to over-initiation of DNA replication (4) and as a consequence also blocks sporulation (27).
parS sequences are positioned near the replication origin, often in multiple copies scattered over a more or less wide replication origin region (<1 Mb) (15). ParB proteins accumulate in high numbers near a given parS site, leading to the formation of distinctive protein clusters in the cell (28)(29)(30). A related and essential feature of ParB is the ability to occupy not only the parS recognition sequence itself but also flanking DNA sequences. The spreading of ParB was first observed indirectly by its effects on plasmid supercoiling and silencing of parS proximal genes (31,32). Chromosomal ParB is enriched in regions ranging from few kb up to ~15 kb around parS sites in chromatin immunoprecipitation profiles (28,33,34). We and others have recently discovered that ParB spreading requires the binding of the unusual cofactor CTP by ParB (16,(35)(36)(37). Based on a nucleotide-ParB co-structure, we have proposed that ParB dimers form DNA sliding clamps that entrap chromosomal DNA in a parS-catalysed closure reaction to then slide onto parS-flanking DNA (37).
ParB proteins comprise three globular domains. The aminoterminal N domain forms a conserved binding pocket for the ribonucleotide CTP (36,37). CTP binding (Kd ~10 µM) (35,37) is expected to be (nearly) saturated under standard physiological conditions (~100-200 µM CTP). Upon contact with parS DNA, open ParB dimers convert into closed clamps by CTP-bound N domains forming interlocking dimers. N-gate closure is slow in the absence of parS DNA sequences. The parS-specific recognition of DNA originates from a helix-turnhelix (HTH) motif located in the ParB M domain (38). The carboxy-terminal C domain supports ligand-independent ParB dimerization as well as sequence-unspecific DNA binding (39,40). CTP hydrolysis does not seem to be required for any step of ParB targeting and sliding in vitro (35,37).
Here, we aimed to get a better understanding of the physiological role of CTP hydrolysis by generating ParB mutants in B. subtilis. We identified two CTP hydrolysis-defective mutants of ParB, which retained the ability to bind CTP, to close the N-gate, and to load onto parS DNA in vitro. We showed that CTP hydrolysis is essential for normal ParB focus formation in vivo. It serves two main functions by recovering CTP-locked ParB clamps trapped off the chromosomes and by restricting the extent of ParB spreading through DNA unloading. Despite the aberrant localization, CTP-hydrolysis-defective ParB mutants supported normal sporulation and Smc recruitment, but the mutant ParABS systems were unable to support chromosome segregation in the absence of Smc. Moreover, we present a model for parS-catalyzed ParB DNA loading based on parSmediated relief of ParB autoinhibition.
Relief of ParB autoinhibition by parS DNA catalysis and ParB recycling by CTP hydrolysis promote bacterial centromere assembly.

Results
Efficient chromosomal loading of ParB clamps requires the cofactor CTP and the catalyst parS DNA. CTP hydrolysis is thought to promote the reverse reaction, ParB clamp unloading. Here, we investigate the molecular mechanisms underlying these reactions as well as the physiological consequences of blocking CTP hydrolysis.

Intra-and intermolecular tethering of N and M domains
In the crystal structure of a CDP-bound B. subtilis ParB dimer, the N domain of a given ParB chain closely associates with the M domain of the partner chain ( Fig. 1A) (Fig. S1A) (PDB: 6SDK) (37). An equivalent domain-swap organization has recently been reported for a ParB paralogue, the B. subtilis protein Noc (41) (PDB: 7NG0), and has previously been noted in the CTPbound dimer of Myxococcus xanthus PadC (PDB: 6RYK) (36). This suggests that domain swapping is not a crystal packing artefact but a conserved feature of ParB and ParB-like proteins.  To test if ParB adopts the domain-swap organization in vivo, we performed site-specific chemical cross-linking at the N-M interface using strains with a halotag (HT) fused to parB at the endogenous gene locus for the detection and quantification of cross-linked species (Fig. 1A) (37). Residues on helix α4 in the N domain (A102C and R105C) and on helix α5 in the M domain (H133C and L134C) were selected for cysteine mutagenesis (Fig. 1A). One of the four cysteine pairs (A102C/L134C) was omitted from the analysis as it exhibited a sporulation defect, being indicative of non-functional protein (42). The three other pairs did not grossly affect ParB function as indicated by the absence of a noticeable sporulation defect and by normal growth of the double cysteine mutants in a sensitized background harboring the smc-pk3 allele (Fig. S1B) (24). If intermolecular N-M tethering does occur, cross-linking with the thiol-specific compound BMOE would generate a dimeric ParB-HT species that is expected to migrate more slowly through an SDS-PAGE gel. We indeed observed such a slowly migrating species (Inter XL ) with all tested cysteine pairs ( Fig.  1A) (Fig. S1C). The efficiency of intermolecular cross-linking was low and slightly higher only with one of the three cysteine pairs (A102C/H133C). Importantly, we also detected a more prominent, faster migrating cross-linked ParB-HT product, which we interpreted as a species (Intra XL ) derived from intramolecular cross-linking of N and M domains (Fig. 1A).
This experiment suggested that N-M tethering occurs in two ways: within a protomer ('ParB Intra' ) and between two protomers ('ParB Inter ') ( Fig. 1, A S2A). We conclude that most or all ParB protein displayed the ParB Intra configuration prior to ligand binding. Addition of CTP or parS DNA 40 alone did not alter the cross-linking pattern significantly, while addition of both strongly increased the abundance of the two dimeric species and in turn almost eliminated the intramolecular cross-linking. Similar results were obtained with ParB(A102C, H133C) (Fig. S2B). The requirement for both ligands mimicked what was observed for N-gate closure by T22C cross-linking ( Fig. 1B) (37). N-gate closure thus goes along with the conversion of ParB Intra into ParB Inter .

A mechanism for parS-catalyzed ParB loading based on ParB autoinhibition
The closed ParB Inter state is thermodynamically favorable in the presence of CTP (and CTPγS but not CDP) (37). The conversion of CTP-bound ParB Intra to ParB Inter however is slow without parS DNA (i.e. a kinetically inhibited reaction) ( Fig.  S2C) (37). What inhibits the reaction in the absence of parS DNA is unclear. Disengagement of intramolecular N-M tethers in both ParB Intra protomers is presumably a prerequisite for forming the closed N-gate built from two ParB Inter protomers. We thus wondered whether the intramolecular N-M tether inhibits N-gate closure. If so, ParB Intra would correspond to an autoinhibited form of ParB. Stochastic N-M disengagement is likely inefficient due to rapid reengagement of the juxtaposed N and M domains, even in case of limited N-M tether stability.
To uncover how such ParB autoinhibition might be overcome by parS DNA binding, we generated a structural model of ParB Intra by manually reconnecting the chains at the N to M junction of the CDP-bound ParB dimer (PDB: 6SDK) (Fig.  S3A). We also computationally predicted ParB structure (de novo) using trRosetta (43). One of the top five hits closely resembled the ParB Inter structure, while the other four matched well with the manually created ParB Intra model, providing unbiased and independent support for the existence of these states (Fig. 1C). Next, we superimposed two ParB Intra chains with opposing halves of a parS site (PDB: 4UMK), using the helix-turn-helix motif in the M domain as guide for structural alignment. We found that a single ParB Intra protomer can readily accommodate a parS half site, but two ParB Intra protomers clash with one another when bound to opposing half sites, indicating that a ParB Intra dimer fails to strongly bind parS DNA ( Fig. 2A). However, when the N domain was manually detached from the M domain (ParB Untethered ) in at least one of the two protomers, then binding of both HTH motifs to the half sites of a given parS DNA sequence becomes feasible without steric clash ( Fig. 2A). Thus, we propose that parS DNA catalyzes N-gate closure by preventing N-M reengagement and selecting and stabilizing a ParB Untethered state (Fig. 2B) A prediction from this Brownian-ratchet reaction scheme is that the steric clash between two ParB Intra protomers at parS DNA is crucial for catalysis. To artificially eliminate the steric clash, we inserted an increasing number of spacer nucleotides (adding 1 to 8 bp) between the two parS half sites in parS DNA 40 . We then tested for the ability of the modified parS DNA to stimulate CTP hydrolysis-an indirect readout for the efficiency of N-gate closure. We found that all altered parS sites failed to stimulate CTP hydrolysis. Intriguingly, DNA molecules with intermediate spacer lengths (2-5 bp) also inhibited the reaction when present at stoichiometric amounts ( Fig. S3B), presumably by allowing unhindered ParB Intra binding to both parS half sites. These observations are consistent with the idea of a steric clash between parS-bound ParB Intra protomers being the basis of parS-catalyzed ParB DNA loading. Notably, ParB binding to isolated parS half sites has been observed in ChIP-Seq profiles for Pseudomonas aeruginosa ParB, particularly upon ParB over-expression (44), and possibly also in other organisms (45).

CTP hydrolysis-defective ParB mutants
We next focused on the reverse reaction, i.e. ParB clamp opening, which is presumably supported by CTP hydrolysis. The ParB CTPase is a recently discovered member of a larger family of proteins with diverse enzymatic activities and cofactors (37,46). Little is yet known about the mechanism of ParB CTP hydrolysis. In ATP and GTP hydrolysis, a water molecule attacks the γ-phosphate moiety to hydrolyze the scissile bond between β-and γ-phosphates. The water molecule is usually activated for nucleophilic attack by an acidic residue, as for example in the Walker B motif. To identify residues with equivalent functions in ParB, we biochemically characterized ParB proteins harboring mutations in selected active site residues. The desired mutants were expected to be defective in CTP hydrolysis but retain their ability to bind CTP, engage the N domains, and entrap parS DNA. The ParB-CDP co-crystal structure (PDB: 6SDK) highlighted two glutamate residues (at position 78 and 111) at the CTP binding pocket (Fig.  3A) (37). These residues belong to widely conserved sequence motifs (GE 78 RRY/F and E 111 NLQR) but they are absent from M. xanthus PadC protein which binds CTP but does not hydrolyze it (with F348 in PadC corresponding to B. subtilis ParB E78 and PadC lacking the ENLQR motif altogether) (Fig. S5A) (36). Examining the PadC-CTP interaction map confirmed that residue F348 is suitably positioned next to the CTP γ-phosphate (Fig. S5B). The glutamate residues were chosen as candidates for further analysis and replaced separately by glutamine (Q). Substitutions for alanine or histidine produced similar initial results but were deemed more intrusive and thus omitted from further investigations (Fig. S7C). The E78Q and E111Q mutant proteins were recombinantly expressed and purified. We assayed for CTP binding affinity, for the rate of CTP hydrolysis, for the efficiency of N-gate closure, and for ParB loading onto parS DNA. Our results showed that the E78Q and E111Q mutants failed to display appreciable levels of CTP hydrolysis as measured by the release of inorganic phosphate, while a wild-type control protein hydrolyzed CTP with a basal rate of approximately 0.15 min -1 , which was stimulated by the presence of parS DNA to about 1 min -1 , comparable with published results (Fig. 3C) (37). All three proteins bound CTP with similar affinity (Kd ~5-8 µM) as judged by ITC (Fig. 3B), and they had roughly comparable efficiencies of N-gate closure with CTP and parS DNA as measured by T22C cross-linking ( Fig. 3D). They also loaded onto parS DNA in the presence of CTP based on biolayer interferometry (BLI) (Fig. 3E). In brief, BLI allowed for the immobilization of a double biotin labeled 169 bp parS DNA fragment on a streptavidin-coated biosensor tip (35). Binding of ligands, including ParB protein, is inferred from a wavelength shift in the reflected light. Association with parS DNA appeared normal or even slightly improved with the EQ mutants in the presence of CTP or the non-hydrolysable analog CTPγS. When shifting the biosensor from the CTP-containing loading buffer to a CTP-free dissociation buffer, wild-type ParB protein was released from DNA with an estimated apparent rate of about 0.4 min -1 , while the two EQ mutants displayed a significantly longer residence time on DNA. These differences in the dissociation rates between wild-type and EQ proteins were eliminated when CTPγS was used instead of CTP during loading. Of note, we observed that the presence of CTP in the dissociation buffer significantly reduced the off-rate of wild-type ParB (Fig. 3F), suggesting that B. subtilis ParB can efficiently exchange CDP for CTP without dissociating from DNA as previously reported for C. crescentus ParB (35). We determined a CTP concentration for half-maximal ParB retention of about 5-10 µM, which is in agreement with the CTP affinity measured by ITC (Fig. 3F).
CTPγS promotes slow but robust N-gate closure even in the absence of parS DNA (37). CTP is unable to do so in wildtype ParB protein, presumably due to its hydrolysis to CDP with ParB Intra being the most populated CDP-state (and apostate). As expected, we found that the EQ mutants supported N-gate closure equally well with CTP and CTPγS (Fig. S4), thus providing further support for the notion that the EQ mutants are defective in CTP hydrolysis and that CTP hydrolysis counteracts N-gate closure. Notably, the N-gate closure reaction was somewhat slower in E78Q and more so in E111Q when compared to wild type, possibly implying that the mutant proteins are either slightly more strongly autoinhibited (more stable ParB Intra ) or less stably closed (less stable ParB Inter ) (Fig.  S4). Altogether, we conclude that the EQ mutants are defective in CTP hydrolysis but support all other biochemical functions normally or near normally in vitro.

CTP hydrolysis is dispensable for SMC recruitment but not for ParABS function.
To elucidate the physiological consequences of defective CTP hydrolysis, we transferred the ParB(EQ) mutations into B. subtilis by allelic replacement. We observed that neither the E78Q nor the E111Q mutation resulted in a noticeable sporulation defect as judged after extended periods of incubation on nutrient-rich agar plates (Fig. 4A), implying that the regulation of DNA replication is only mildly or not at all perturbed in the mutants in contrast to the parB in-frame deletion mutant (ΔparB). Next, we combined the EQ mutations with a hypomorphic allele of the smc gene, smc-pk3, to sensitize cells for defects in chromosome organization and segregation (24). Again, unlike the ΔparB mutant, the EQ mutants supported robust growth of the smc-pk3 strain (Fig. 4B), demonstrating that the parB(EQ) mutants supported chromosome segregation well, presumably by promoting the loading of Smc-ScpAB onto the chromosome. To test this more directly, we next performed chromatin immunoprecipitation (ChIP) using antiserum raised against the B. subtilis Smc protein. ChIP-Seq showed that the chromosomal distribution of the Smc protein was similar in the EQ mutants and wild type, and distinct from the ΔparB mutant (Fig. 4C). ChIP-qPCR analysis of the same samples suggested that the levels of enrichment of the Smc protein were slightly reduced at the parS-359 site and the replication origin ('dnaA') in the EQ mutants, but clearly not as much reduced as in ΔparB, together providing further support for the notion that CTP hydrolysis is largely dispensable for Smc recruitment and loading (Fig. S6B).
Next, we investigated the role of ParB CTP hydrolysis in the     core function of the ParABS system. To completely rule out effects of ParB-Smc interactions on chromosome organization and segregation, we constructed the parB(EQ) mutants in a Δsmc strain. We note that the isolation of such double mutant strains was difficult, similar to the mutant of both Δsmc and ΔparB. The resulting strains failed to grow on nutrient-rich medium and also displayed very poor growth characteristics under nutrient limiting conditions and at reduced temperatures, thus being markedly sicker than the Δsmc strain (Fig. 4D).  (B) Growth assay by dilution spotting of wild-type B. subtilis and strains carrying parB alleles in an smc-pk3 background (hypomorphic mutant). 9 2 and 9 5 -dilutions were spotted on minimal medium agar plates (SMG) and rich medium agar plates (ONA) and imaged after 24 and 16 hours, respectively.

(C)
Chromatini m m u n o p r e c i p i t a t i o n coupled to deep sequencing (ChIP-Seq) using α-Smc serum. Panels show ratio plots of Smc enrichment against that of Δsmc. All ChIP-Seq profiles were split into 1 kb bins with the origin of replication placed at position 0 Mb. The ratio was calculated by dividing the higher value by the lower, if the condition/Δsmc ratio was > 1, it was plotted above the genome position axis (purple colour). If the Δsmc/condition ratio was > 1 the ratio was plotted below the axis (grey colour). Black dashed lines and arrows represent the 8 prominent parS sites. Of note, the peak at position +1.8 Mb (green dashed line) marked by asterisk represents enrichment of the smc gene presumed to be a contamination. (D) Growth assay by dilution spotting of wild type B. subtilis and strains carrying parB alleles in a smc deletion background ('Δsmc'). 9 2 and 9 5 -fold dilutions were spotted on minimal (SMG) and rich medium (ONA) and imaged after 24 and 16 hours, respectively. (E) Quantification of the fraction of anucleate cells using ImageJ with the microbeJ plugin (Ducret et., al 2016) in parB and parB, Δsmc mutants. Five fields of view were captured randomly for each strain containing roughly 100 cells per field of view. All cells in the field of view were counted (unless they're too close to the border of the field of view). Cell outline and focus detection were reviewed manually to ensure accurate segmentation and true focus (maxima) detection. The percentage of anucleate cells is reported as the number of cells without maxima divided by the total cell count and corrected to 100. Mean value of anucleate cells per field of view is reported for each strain with the standard deviation.

CTP hydrolysis prevents excessive ParB spreading and off-target accumulation of closed ParB
To reveal how CTP hydrolysis may promote ParABS function while being apparently dispensable for Smc recruitment, we studied the cellular localization and chromosomal distribution of ParB protein in B. subtilis. We performed ChIP-qPCR using antiserum raised against full-length B. subtilis ParB protein.
The enrichment of ParB(E78Q) was clearly reduced at the parS-359 site and the neighboring gene parA (soj) (Fig. 5A). This reduction was even more pronounced for ParB(E111Q) (Fig. 5A). Deep sequencing of the ChIP eluates resulted in reduced read counts in the EQ mutants at parS-359 and at other parS sites, again with the E111Q variant displaying a more drastic phenotype. Moreover, the shape of the ParB distribution at parS sites was clearly altered in both EQ mutants. The enriched region was significantly expanded by extending further onto one side of parS-away from the replication origin-but not onto the other (Fig. 5B). The width of the peak at parS-359 increased from approximately 15 kb for wild type to roughly 40 kb for the two EQ mutants. The peak position was also slightly shifted away from the parS site in the direction of the extended shoulder. This suggested that the ParB(EQ) mutants exhibited excessive spreading, presumably by an increased chromosome residence time in the absence of CTP hydrolysis. This excessive spreading of ParB(EQ) was highly asymmetric, putatively due to the impediment of spreading in one direction by other chromosomal processes such as head-on transcription or DNA replication. If so, then CTP hydrolysis might reduce the frequency or impact of such encounters in wild-type cells. How this might be accomplished is unclear.
We then investigated the cellular localization of wild-type and mutant ParB-mScarlet fusion proteins by live-cell imaging (Fig.  5C). These strains also expressed a mTurquoise-tagged ectopic copy of hbs, encoding for an abundant nucleoid associated protein with sequence-unspecific DNA-binding properties. Wild-type ParB-mScarlet formed the characteristically bright foci that are known to colocalize with parS sites. In contrast, the two hydrolysis mutants displayed more diffusive signal with only faint foci being detected. Of note, the mScarlet fusion mildly reduced the enrichment of ParB at two tested loci as judged by ChIP-qPCR (Fig. S7B).
In summary, we observed defects in the accumulation and the distribution of ParB(EQ) proteins near parS sites. The total level of protein in ParB-mScarlet foci was decreased in the two EQ mutants, mildly in E78Q and strongly in E111Q. The occupancy of the parS-proximal regions was also reduced, while the occupancy of more distal positions on one of the two parSflanking regions was increased.

CTP hydrolysis-defective ParB mutants accumulate in an alternate state on the chromosome
To elucidate the state of ParB(EQ) proteins in the cell, we next employed in vivo cysteine cross-linking using ParB-HT fusion proteins. Presumably due to CTP hydrolysis dominating over N-gate closure in vivo, T22C cross-linking of otherwise wildtype ParB-HT protein was barely detectable (Fig. 1A, 6A), as reported previously (37). Cross-linking of R105C/H133C also showed mostly the ParB Intra configuration (Fig. 1A, 6A). The EQ mutants, however, robustly generated ParB N-N cross-links with T22C and ParB Inter N-M cross-links with R105C/H133C (Fig. 6A). Thus, in contrast to wild-type ParB, the EQ mutants accumulated predominantly in the closed form in vivo. This finding provided further support for the notion of defective CTP hydrolysis in the EQ proteins.
The fact that EQ proteins showed more robust N-gate closure compared to wild type (Fig. 6A) implies that a large fraction of wild-type ParB in partition complexes harbors an opened N-gate. Moreover, the poor localization of the EQ mutants (and in particular of E111Q) to partition complexes indicated that they may have undergone N-gate closure without parS stimulation or have dissociated from parS DNA after N-gate closure. To discriminate between these two possibilities, we have repeated the T22C cross-linking experiment in a strain lacking 8 strong parS sites (47). We found that T22C crosslinking was only mildly reduced in both EQ proteins (Fig. 6B), suggesting that a significant fraction of closed ParB clamps are formed without parS stimulation. CTP hydrolysis likely recovers such futile clamps in wild-type ParB.
As a measure for chromosome association irrespective of chromosome localization, we finally performed chromosome entrapment assays. Briefly, we used cysteine cross-linking to generate covalently closed circular species of ParB-HT in vivo (Fig. 6C) to then detect their co-isolation with chromosomal DNA in agarose beads under protein denaturing conditions (48,49). Only cross-linked species that entrap the chromosomal DNA double helix are retained in the agarose matrix. Using R105C/ H133C at the N-M interface in combination with S278C at the C-C interface, we found that significant levels of cross-linked ParB(E78Q) species were retained in agarose beads (Fig. 6D)  wild-type equivalents do not show robust ParB Inter cross-linking with R105C/H133C, no information on hydrolysis-proficient ParB was obtained in this experiment. Together our results indicate that ParB and ParB(EQ) occupy distinct states in the cell and when loaded onto the chromosome, which may explain at least in part the defects in ParABS function in the EQ mutants. A model figure summarizes these findings (Fig. 7A).

CTP hydrolysis promotes ParB/parS ultra-affinity
The detailed model (Fig. 7A) could be further coarse grained to a simple scheme for parS DNA association (Fig. 7B) by means of a few reasonable approximations. Only the open form of ParB can engage the substrate, whereas closed ParB cannot bind or release DNA substrates, because steric clashes make these processes extremely unlikely, if not impossible. Furthermore, ParB without nucleotides is a transient, poorly populated state, as could be inferred from the longer persistence of already bound ParB on DNA in the presence of CTP in BLI experiments (Fig. 3F). Thus, we captured the transition from CDP-bound to CTP-bound ParB by means of a simple transition. The resulting model can be analytically solved (see Materials and Methods), giving an estimate of the "apparent", observed, dissociation constant ( Kd obs ) of ParB for DNA: The parS sequence greatly accelerates ParB closure (k k oc parS h & ), thus decreasing the value of the observed dissociation constant. In the case of non-specific DNA, however, the rate of ParB closure is unaffected and likely slower than the hydrolysis rate, resulting in a dissociation constant which is the same as the one of open ParB. The same observed dissociation constant also applies to hydrolysis deficient mutants at equilibrium (see Materials and Methods).

Discussion
The partition complex is critical for ParABS function. The assembly has been intensely studied over several decades. The recently discovered requirement for CTP binding by ParB in partition complex assembly opened new avenues for investigation. Here, we focused on the mechanistic basis of parS DNA catalysis of ParB clamp loading and on the physiological role of ParB CTP hydrolysis in partition complex assembly and chromosome segregation.

parS DNA: A B-form DNA double helix as catalyst
Ribozymes are naturally occurring nucleic acid catalysts that generally utilize folded single-stranded RNA to catalyze a variety of chemical reactions. Unlike ribozymes, parS DNA appears to work as catalyst in the standard B-form of doublestranded DNA. Based on structural information and chemical cross-linking, we here propose a molecular mechanism for parS catalysis. We found that ParB dimers are unable to engage with both halves of the palindromic parS site when adopting the predominant conformation with tethered N and M domains, ParB Intra and ParB Inter . Partial unfolding of ParB appears to be needed for full engagement of parS DNA. This idea explains how parS DNA selectively stabilizes an intermediate of the reaction, while also binding to the reactant and product of the reaction, albeit with reduced affinity. Available co-crystal structures are consistent with the notion of a steric clash between parS-bound ParB chains. The structure of H. pylori ParB (lacking the C domain) bound to parS (PDB: 4UMK) (38) shows partially unfolded N domains (in four variations). Similarly, a C-terminally truncated variant of C. crescentus ParB (PDB: 6T1F) displays partial unfolding in one of the two parS-bound chains. These structures imply that parS DNA binding sufficiently compensates for the energetic investment in the partial unfolding of the N domain. The fact that a cysteine pair at the N-M interface (A102C/L134C) hinders sporulation and like another cysteine pair (A102C/ H133C) leads to elevated levels of the ParB Inter conformation (Fig. S1C) supports the notion that this interface is indeed important for function and conformational control. We also demonstrated that the apposition of parS half sites is critical for catalysis. Spacer addition not only eliminated catalysis but lead to inhibition of the reaction, presumably due to improved binding of ParB Intra dimers to the modified parS sites (Fig. 2B). This paradigm puts forward a strategy for engineering selfloading clamps with altered targeting specificity by repurposing sequence-specific or sequence-guided DNA binding proteins, eventually allowing for flexible chromosome labelling and DNA detection for diagnostic purposes.

ParB CTP hydrolysis in partition complex assembly
We showed here that CTP hydrolysis by ParB is dispensable for some functions of ParABS in B. subtilis and critical for others. Sporulation and Smc recruitment were largely unperturbed in the CTP hydrolysis-deficient mutants E78Q and E111Q (Fig. 4). In contrast, ParB mutants interfering with CTP binding (G77S and R80A) are deficient in Smc recruitment (37, 50) as well as sporulation (28,42). The robustness of Smc recruitment was surprising, particularly when considering the mediocre enrichment of ParB(E111Q) at parS sites. Conceivably, the poor enrichment is compensated by ParB(EQ) being trapped in a state that efficiently supports Smc recruitment. It is thus tempting to speculate that the CTP-ParB Inter state is proficient in Smc loading and conceivably even solely responsible for it. In case of sporulation, any closed ParB clamps accumulating off the chromosome may also contribute to the conversion of ParA ATP-dimers to monomers-which is a prerequisite for sporulation-since a HTH mutant appears to support sporulation quite well (24). In contrast, chromosome segregation in the absence of smc was strongly compromised in either CTP-hydrolysis mutant, indicating defective ParABS function (Fig. 4). This defect likely explains why the CTPase activity is highly conserved and maintained over extended periods of time during evolution.
Explanations for the poor function of ParABS in the absence of CTP hydrolysis include (1) an inadequate assembly of the partition complex and (2) the accumulation of ParB in a state that is incompatible with proper ParABS function. We present evidence for either scenario. The notion of altered states of ParB in the EQ mutants is supported by our cysteine crosslinking (Fig. 6). While CTP hydrolysis-proficient wild-type ParB shows only low levels of ParB Inter clamp formation in vivo (Fig.  1A, 6A), the EQ mutant variants predominantly adopt this state. It is also consistent with a recent report of CTP-controlled ParA-ATPase activation in the F plasmid ParABS system (51). Future work will have to establish, which form of ParB supports Smc loading and which one promotes ParA ATP hydrolysis. This may potentially establish temporal or spatial control of chromosome organization, segregation, and DNA replication.
Several observations point to defects in concentrating ParB in partition complexes in E78Q and E111Q. ChIP and fluorescence imaging revealed a reduced enrichment and broadened distribution of ParB(EQ) proteins at and near parS sites. CTP hydrolysis conceivably supports partition complex assembly by recycling ParB Inter species. We envision two pools of ParB Inter that may require recycling: ParB clamps that have excessively spread away from the parS loading site and clamps that have undergone N-gate closure without parS DNA stimulation, either becoming locked off the chromosome or trapped on the chromosome at large distances from parS sites (Fig. 7A). Such clamps would be permanently lost from the pool of productive ParB protein without the ability to hydrolyze CTP. parS-independent clamp closure readily occurs in vitro at least with B. subtilis ParB as indicated by the low but appreciable rates of CTP hydrolysis without parS DNA stimulation and by the slow but robust closure of the N-gate with the help of CTPγS in the absence of parS DNA. Similarly, we detected closed ParB(E78Q) clamps in cells lacking parS sites (Fig. 6B). Nevertheless, we cannot exclude the possibility that the poor enrichment at parS sites is caused at least in part by indirect consequences of the E78Q and E111Q mutations on ParB activity, for example by reducing N-gate stability or ParB autoinhibition.
We also described ParB behaviour by a simple reaction scheme (Fig. 7B). Thanks to CTP hydrolysis, ParB exhibits an enhanced affinity for parS sequences, beyond the one that would be possible at equilibrium (e.g. by hydrolysis deficient mutants). This ultra-affinity (52) corresponds to a roughly 50fold reduction of the dissociation constant, from an estimate of the hydrolysis and parS-catalyzed ParB closure rates (K h ~1/ min; k k oc parS h & ~1/sec; (37)). The enrichment of ParB on the parS flanking regions is a consequence of this effect because the increased amount of ParB bound to parS can subsequently diffuse away, thus populating the nearby non-specific DNA sequences.
Wild-type ParB barely adopts the clamp state in vivo despite being mostly localized in ParB foci (9,53,54). Mechanisms other than DNA entrapment may thus contribute to the retention of wild-type ParB protein in the partition complex, such as contacts between (open) ParB dimers (3,28,30,55). DNA unloading upon CTP hydrolysis is prevented in vitro by the presence of CTP in case of B. subtilis ParB (Fig. 3F) as well as C. crescentus ParB (35). This implies that CDP   (1) and (2). The N domain spontaneously dissociates from the M domain to generate ParB Untethered (3). This transition is thermodynamically unfavorable and slow but promoted by parS DNA binding (for details see Fig. 2B). Subsequent clamp closure is rapid off DNA (left side) (4) and on parS DNA (right side) (4'). Due to a steric clash between parS DNA and the two M domains (Fig. 2B), the DNA is released from parS DNA and located in a lumen formed by the is readily substituted for CTP without ParB unloading from the chromosome (Fig. 7B). It will be interesting to establish whether this nucleotide exchange reaction requires localization of ParB to the partition complex or DNA entrapment by ParB.
If so, this feature would selectively eliminate off-target clamps or DNA-free clamps, respectively, thus further enhancing parSproximal enrichment of ParB.
We conclude that parS DNA catalysis and ParB CTP hydrolysis lead to ParB/parS ultra-affinity and super-concentrate ParB within the partition complex. The unusually high concentration of ParB (56) is likely required to optimally support chromosome and plasmid segregation by the ParABS diffusion-ratchet mechanism but might not be essential for other cellular functions of ParABS such as Smc recruitment.

Bacillus subtilis strains and growth
The B. subtilis 1A700 isolate was used for all experiments. B. subtilis was transformed via natural competence and homologous recombination as described in (57) and grown on SMG-agar plates with appropriate antibiotic selection. Transformants were next checked by PCR and Sanger sequencing. Tagging of ParB (with HaloTag or mScarlet) or HBsu (with mTurquoise) proteins was done at the carboxy terminus. Genotypes of strains used in this study are listed in Table S1.
For spotting assays, the cells were cultured in SMG medium at 37 °C to stationary phase and 9 2 and 9 5 -fold dilutions were spotted onto ONA (~16 h incubation) and SMG (~24 h incubation) agar plates.

Expression and purification full-length B. subtilis ParB protein
Expression constructs were prepared in pET- 28

CTPγS
CTPγS was custom-synthesized and purified by reversedphase chromatography (RP-HPFL) to a final purity of 90.6 % by Jena Biosciences (Jena, Germany). A stock solution at a concentration of 100 mM (pH-adjusted to 8.0 by addition of NaOH) was aliquoted and stored at -80 °C.

Isothermal titration calorimetry (ITC)
The measurement was done using MicroCal iTC200 (GE Healthcare Life Sciences). The instrument was pre-cooled to 4 °C. All measurements were made in buffer containing 150 mM NaCl, 50 mM Tris/HCl pH 7.5, and 5 mM MgCl2. Both measurement cell and injection syringe were subjected to a series of washes with buffer. 280 μL of the protein solution at 80-120 μM monomer concentration was added to the measurement cell, and the injection syringe was filled with buffer containing 2 mM of CTP or buffer only. Measurements were taken with an initial delay of 180 seconds, and the settings of the instrument were adjusted to: reference power of 5 μcal/ sec, a stirring velocity of 1000 RPM, and a "high feedback" mode. Raw data were integrated to kcal/mol, presented as a Wiseman plot, and wherever possible, regression curves were calculated according to a 1:1 nucleotide to ParB monomer binding model. Origin (GE Healthcare) was used for fitting results from the measurements by using the following equation: is the heat released from the i th injection which is in turn calculated using the following equation: K is the binding constant, ΔH is the molar heat of ligand binding, Xt is bulk concentration of nucleotide, and Mt is the bulk concentration of ParB (moles/liter) in Vo. K and ΔH were estimated by Origin, and ΔQ(i) for each injection was calculated and compared to the measured heat. K and ΔH estimates were then improved using standard Marquardt methods. Several iterations were performed until the fit could no longer be improved.

Measurement of CTP hydrolysis by Malachite Green colorimetric detection
Mixtures of CTP (2x) with or without parS DNA 40 (2x) in reaction buffer (150 mM NaCl, 50 mM Tris pH 7.5, 5 mM MgCl2) were prepared on ice. Protein solutions (2x) in reaction buffer were also prepared on ice. CTP/ parS DNA 40 pre-mix (10 μL) was added to protein solution (10 μL) using BenchSmart 96 (Rainin) dispenser robot and mixed by pipetting. After mixing, samples (containing CTP 1 mM, parS DNA 40 1 μM, protein 10 μM) were placed in a PCR machine and set to incubate at 25 °C for 1 hour. In parallel, phosphate blanks were prepared for each experiment. Samples were diluted 4-fold by the addition of 60 μL water, then mixed with 20 μL working reagent (Sigma) and transferred to a flat bottom 96 well plate. The plate was left to incubate for 30 minutes at 25 °C, and the absorbance was then read at a wavelength of 620 nm. Absorbance values from the phosphate standard samples were used to plot an OD 620 versus phosphate concentration standard curve. Raw values were converted to rate values using the standard curve. Absolute rates were calculated by normalizing for protein concentration. Mean values and standard deviation were calculated from three repeat experiments. Graphs were plotted on GraphPad Prism for presentation.

40-nucleotide, double stranded parS DNA 40 preparation
Complementary strands of oligonucleotides (at 100 μM each) (for sequences see Supplementary Table S2) were mixed 1:1 and heated to 95 °C for 10 minutes and then left to cool down to 25 °C.

In vitro cross-linking
Mixtures (2x) of NTP with or without parS DNA 40 were prepared in reaction buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 5 mM MgCl2) and kept at room temperature (RT) for 5 minutes. Protein solution (2x in 300mM NaCl, 50mM Tris-HCl pH 7.5) was added to a final concentration of 10 μM and the mixture (containing 1 mM NTP with or without 1 μM parS DNA 40 was incubated for an additional 5 minutes at room temperature. 1 mM BMOE (from a 20 mM stock solution) was added to the samples. After 5 minutes at RT, samples were quenched with β-mercaptoethanol (23 mM final), loading dye was added, and samples were incubated at 70°C for 5 minutes and loaded onto Bis-Tris 4-12 % gradient gels (ThermoFisher). Bands were stained by CBB and relative band intensity was quantified by scanning and semi-automated analysis in ImageQuant (GE Healthcare).

In vivo cross-linking
Culture, grown overnight in 0.5 % (w/v) glucose containing LB, was used to inoculate fresh LB medium to an OD 600 of 0.005 and were grown at 37°C to mid-exponential phase (OD 600 of 0.3). Around 70g of Ice cubes were added directly to the culture flask, and the cells were harvested by centrifugation for 5 minutes at 14,000 g at 4°C. The cell pellet was washed in cold PBSG (PBS with 0.1 % (v/v) glycerol). Cells were pelleted again and resuspended in 1 mL cold PBSG. 1.25 OD units were collected, and resuspended in 30 μL of cold PBSG. 0.5 mM BMOE was then added to the samples, mixed by pipetting, and placed on ice for 15 minutes. 15 mM β-mercaptoethanol was added and the samples were incubated on ice for 3 minutes to quench the reaction. The following reagents were added to the given final concentrations: 750 U/mL Benzonase (Sigma), 5 μM HaloTag-TMR (Promega), 47 U/μL Ready Lyse Lysozyme (Epicentre), and 1x PIC (Sigma). The cells were then lysed at 37°C in the dark for 30 minutes, loading dye containing 5 % (v/v) β-mercaptoethanol was added, and the samples were incubated at 70 °C for 5 minutes. 10 μL of each sample was loaded onto a 4-12 % Bis-Tris Gel (ThermoFisher). Following SDS PAGE, gels were imaged on an Amersham Typhoon (GE Healthcare) with Cy3 DIGE filter setup. Bands were quantified using ImageQuant (GE Healthcare).

Chromosome Entrapment Assay
The assay was done as described in (48,49) using agarose beads. In brief, cells were grown, harvested, and 3.75 OD units was cross-linked as described above (in vivo cross-linking). The sample was split into two aliquots (2:3 for the output/beads and 1:3 for input). To each input sample, 5 µL of Mastermix-1 was added. Mastermix-1 contains (per one input sample): 400 units of Ready-Lyse Lysozyme, 12.5 units Benzonase, 1 µM TMR HaloTag, 0.5 µL protease inhibitor cocktail (PIC), and 0.9 µL 1X PBS). The input samples are then incubated for 25 min at 37 °C covered from light, and an equal volume of 2X protein loading dye was added and the samples were stored at -20 °C.
For agarose bead preparation, 1 µL PIC and 9 µL of Dynabeads™ M-280 Streptavidin (ThermoFisher Scientific) were added to each output ('beads') sample. 100 µL of preheated 2 % low melt agarose was then added, and immediately followed by 700 µL of mineral oil (preheated to 4 °C). The sample was then vortexed vigorously for 2 min on ice. Each sample was treated quickly and separately.
The output (beads) samples were then spun down at 10,000 g at RT, and the supernatant (oil) was removed as much as possible. The beads were then resuspended in 1 mL PBSG (0.1 % glycerol) very gently to prevent the release of cells from the agarose beads.
To each output sample, 4U/µL of Ready-Lyse Lysozyme, 1 µM of TMR-HaloTag, 5 µL PIC, and 1 mM EDTA were added (same as Mastermix-1 but without Benzonase). The output samples are then incubated for 25 min at 37 °C covered from light. Next, the beads were washed twice with 1 mL RT PBSG, and with TES buffer (50 mM Tris-HCL pH 7.5, 10 mM EDTA, 1 % SDS) three times, with the first time incubated for 1 hr at RT with moderate shaking (500 rpm) and the following two incubations were done for 30 minutes. The beads were resuspended in 1 mL TES and incubated overnight at 4 °C on a rotating wheel and covered from light.
The next day, the beads were washed twice with PBS 1X at RT to remove EDTA and SDS and resuspended in 100 µL of PBS. 1µL of SmDnase (750 U) was added to each output sample and incubated for 1 hr at 37 °C with moderate shaking and covered from light. The samples were next incubated for 1 min at 70 °C at 14000 rpm shaking to melt the agarose plugs, and then transferred onto ice and incubated for 5 min. The samples were spun down at maximum speed for 15 min at 4 °C and the supernatant was transferred to an acetate spin column and spun down for additional 5 min at RT at maximum speed to recover as much liquid as possible. The approximate recovered volume is 100 µL which was then diluted to 1 mL with water and 10 µL of 2 % sodium-deoxycholate and 3.3 mg/mL BSA was added and incubated for 30 minutes at 4 °C protected from light. After incubation, 120 µL of 80 % trichloroacetic acid (TCA) solution was added and incubated for 2 hours on ice and covered from light. The samples are then spun down for 15 minutes at maximum speed at 4 °C. The precipitate was collected and resuspended in 10 µL of 1x loading dye. The input samples were thawed and boiled with the output samples for 5 min at 95 °C. 5 % of the input and all of the output volume were loaded on an SDS-PAGE using 8-12 % Novex Wedge Well Tris-Glycine gel (Thermo Scientific). The gel was then imaged on Amersham Typhoon (GE Healthcare) with Cy3 filter.

Biolayer interferometry (BLI)
Measurements were done in a buffer containing 150 mM NaCl, 50 mM Tris-HCl pH 7.5, and 5 mM MgCl2 on BLItz machine (FortéBio Sartorius). Streptavidin coated biosensors were used in all the measurements and were hydrated in the reaction buffer for 10 minutes before loading. A baseline was first recorded by equilibrating the biosensor in 250 µL reaction buffer in a black 0.5 mL Eppendorf tube for 30 seconds. 4 µL of 100 nM biotin labelled double stranded parS or mut-parS DNA 169bp was then loaded on the biosensor for 5 minutes. After the DNA loading phase, the biosensor was washed once with the reaction buffer and once with the reaction buffer containing 1 mM NTP. Next, 2X ParB solution and 2x NTP solution were mixed 1:1 (final concentration of 1 µM parB and 1 mM NTP) and 4 µl of the mixture was loaded immediately on the biosensor for 2 minutes. The dissociation phase is then carried for 5 minutes in 250 µL protein-free reaction buffer with or without NTP. All measurements were analyzed on the BLItz analysis software and replotted on GraphPad Prism for presentation.

Chromatin Immunoprecipitation (ChIP)
ChIP samples were prepared as described previously ( For ChIP-qPCR, each pellet was resuspended in 2 mL of buffer L (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA pH 8.0, 1 % (v/v) Triton X-100, 0.1 % (w/v) Sodium-deoxycholate, 0.1 mg/mL RNaseA and PIC (Sigma)) and transferred to 5 mL round-bottom tubes. Cells were sonicated three times for 20 sec on a Bandelin Sonoplus with a MS72 tip (settings: 90 % pulse and 35 % power output). The lysates were next transferred into 2 mL tubes and centrifuged for 10 min at 21000 g at 4°C. 800 µL of the supernatant was used as input (IP) and 200 μL was kept as whole cell extract (WCE). The WCE tubes are frozen at -20°C for later.
For the IP, antibody serum was pre-incubated with Protein G coupled dynabeads (Invitrogen) in 1:1 ratio for 2h at 4°C on a rotating wheel. Next, the beads were washed in buffer L and 50 μL were aliquoted to each sample tube. Samples were incubated with the beads for 2h at 4°C with rotation. After incubation, all samples were subjected to a series of washes with buffer L, buffer L5 (buffer L containing 500 mM NaCl), buffer W (

Processing of ChIP-Seq reads
Reads were mapped to Bacillus subtilis genome NC_000964.3 with bowtie2 using the default mode. Subsequent data analysis was performed using Seqmonk http://www.bioinformatics. babraham.ac.uk/projects/seqmonk/. A bin size of 1 kb was used.

Mathematical model of ParB-DNA association
The steady state solution of the model in Fig. 7B 1 1 1 H   ) p a r B ( E 1 1 1 A ) α-ParB ChIP-qPCR α-ParB ChIP-qPCR  Figure S7. (A) Chromatin-immunoprecipitation coupled to quantitative PCR (ChIP-qPCR) using α-ParB serum. Same as in Fig. 5A in addition to a strain carrying the parB(T22C) allele. (B) Chromatin-immunoprecipitation coupled to quantitative PCR (ChIP-qPCR) using α-ParB serum for strains with mScarlet tagged ParB. As in (A). (C) Chromatin-immunoprecipitation coupled to quantitative PCR (ChIP-qPCR) using α-ParB serum. Same as in Fig. 5A in addition to strains carrying histidine (H) and alanine (A) substitutions of E78 and E111 residues. (D) Chromatin-immunoprecipitation coupled with deep sequencing (ChIP-Seq) using α-ParB serum. Same as in Fig. 5B but with the addition of a ΔparB strain and showing genome wide distribution. Of note, the peak at position +1.8 Mb represents enrichment at the smc gene presumed to be a contamination. (E) Ratio plot of sequencing reads found in E78Q, E111Q, and ΔparB divided by read number in wild type. The region shown is 1 Mb wide surrounding the origin of replication. Peaks correspond to regions with higher enrichment of ParB(EQ) near parS sites (higher spreading efficiency). This work