Self-Organization of parS Centromeres by the ParB CTP Hydrolase

Abstract

ParABS systems facilitate chromosome segregation and plasmid partitioning in bacteria and archaea. ParB protein binds centromeric parS DNA sequences and spreads to flanking DNA. We show that ParB is an enzyme that hydrolyzes cytidine triphosphate (CTP) to diphosphate (CDP). parS DNA stimulates cooperative CTP binding by ParB and CTP hydrolysis. A nucleotide co-crystal structure elucidates the catalytic center of the dimerization-dependent ParB CTPase. Single-molecule imaging and biochemical assays recapitulate features of ParB spreading from parS in the presence but not absence of CTP. The findings suggest that centromeres assemble by self-loading of ParB DNA sliding clamps at parS. ParB CTPase is not related to known nucleotide hydrolases and might be a promising target for developing new classes of antibiotics.

Sister chromosomes separate from one another and distribute within bacterial cells by means of the ParABS system and the SMC complex. Bacterial centromeres – assembled from ParB proteins and 16 bp palindromic parS sites – initiate chromosome segregation soon after replication initiation by recruiting SMC condensin and by moving along ParA ATPase gradients (1–5). Multiple ParB proteins localize to a given parS site thereby forming distinctive clusters in the cell (6, 7). ParB recognizes parS and enriches in ~15 kb wide DNA regions flanking parS (6, 8–10). ParB spreading is hindered by DNA-binding protein ‘roadblocks’ engineered next to parS (10–12). Several models for ParB spreading have been proposed (6, 13). However, despite being critical for chromosome segregation, spreading has not yet been faithfully reconstituted in vitro, indicating the lack of an essential component.

ParB proteins harbor three globular domains (Fig. 1). A helix-turn-helix (HTH) motif in the middle domain (M) recognizes parS DNA. The C-terminal domain (C) homodimerizes and in chromosomally encoded ParB also supports sequence-non-specific DNA binding (14). The N-terminal domain (N) harbors conserved residues that produce strong phenotypes when mutated (5–7, 9, 15) but its function is unclear. Domain N shares marginal sequence similarity with a functionally unrelated eukaryotic enzyme, called sulfiredoxin (Srx) (fig. S1A) (16). Srx catalyzes the repair of peroxiredoxin by transferring ATP γ-phosphate groups onto hyper-oxidized cysteine moieties. A conserved GxxRxxA motif forms the ATP binding pocket in Srx (16, 17). Mutations in corresponding ParB residues (G77S and R80A) lead to loss of function and subcellular localization defects in Bacillus subtilis (Bsu) (5–7). Those subtle similarities invoke the possibility that ParB proteins are not merely DNA-binding proteins but enzymes.

Fig. 1. ParB CTP binding.

(A) Membrane-spotting assay using radiolabeled nucleotides. (B) ITC measurements with full-length ParB_Bsu in the presence of Ca²⁺. The Kd obtained in a typical experiment is given. The interval indicates deviations of data points from the fit. (C) Crystal structure of a single chain of ParB_Bsu(21-218)/with CDP/Ca²⁺. (D) Nucleotide binding pocket in ParB_Bsu/CDP/Ca²⁺. GxxRxxA residues marked in bold. (E) Superimposition of ParB_Bsu(21-218)/CDP/Ca²⁺ with human Srx/ATP/Mg²⁺ (PDB: 3CYI).

Membrane-spotting assays with radiolabeled NTPs (18) surprisingly indicated that full-length ParB_Bsu and the fragment ParB_Bsu(21-218) bind CTP rather than ATP (Fig. 1A and fig. S1B). CTP binding was abolished in the ParB_Bsu(R80A) mutant (fig. S1C). Isothermal titration calorimetry (ITC) confirmed ParB_Bsu binding to CTP yielding dissociation constants (K_d) in the range of 10-50 μM. In contrast, other nucleotides produced little response in ITC (Fig. 1B and fig. S2A-B). CTP was also bound by plasmid-encoded ParB proteins (F plasmid ParB_F and P1 prophage ParB_P1) implying broad evolutionary conservation (fig. S1B-D) (10, 19).

We then solved the atomic structure of ParB_Bsu(21-218) (comprising domains N and M) using crystals grown with CTP and Ca²⁺ ions. The unit cell contained two ParB_Bsu(21-218) dimers. Each chain associated with a CDP molecule and two coordinated Ca²⁺ ions (Fig. 1C). CTP must thus have been hydrolyzed during crystallization. In the structure, the CDP β phosphate group contacts G77 and R80, explaining the strict conservation of the GxxRxxA motif (Fig. 1D) and implying that phenotypes associated with GxxRxxA mutations are caused by defective nucleotide binding (or hydrolysis). Individual domains N and M superimposed well with published ParB structures of bacterial and archaeal origin (with few exceptions) (fig. S2C) (20–22). Domain N also superimposed fairly well with human Srx/ATP/Mg²⁺ (17) yielding a close overlap of α and β phosphate groups (Fig. 1E). This striking resemblance implies that Srx ATP phosphotransferases and ParB CTP phosphohydrolases evolved from a common ancestor.

We next investigated ParB CTPase activity by measuring the accumulation of inorganic phosphate (Fig. 2A and fig. S3A). ParB_Bsu dimers hydrolyzed about five CTP molecules per hour. CTP hydrolysis was abolished in the CTP-binding mutant R80A and also in the CTP-binding proficient mutant N112S (fig. S2B, S3B, S4A) (5, 23). Importantly, addition of 40-mer parS duplex (^parSDNA40) strongly increased CTP turn-over to about 36 molecules per hour (Fig. 2A). Sub-stoichiometric amounts of parSDNA40 (ratio ~40:1) were sufficient to saturate enzymatic activity (Fig. 2B), implying that parS only transiently associates with ParB. Noncognate DNA sequences did not alter the activity, even at elevated concentrations (Fig. 2A). ParB_F and the paralogue Noc_Bsu (24) also displayed CTPase activity which was stimulated by cognate DNA (fig. S3A).

Fig. 2. ParB CTP hydrolysis.

(A) CTP hydrolysis measured by colorimetric detection of inorganic phosphate using malachite green. Mean values from three repeat measurements. Data points are shown as dots. (B) CTP hydrolysis by ParB_Bsu dimers with increasing DNA concentrations. Data points and fits of three repeat measurements. (C) Membrane-spotting showing parS-stimulated cooperative CTP binding. (D, E) Crystal structure of a ParB_Bsu(21-218)/CDP/Ca²⁺ dimer. (F) Catalytic center in ParB_Bsu(21-218)/CDP/Ca²⁺.

Addition of ^parSDNA₄₀ promoted cooperative CTP binding by ParB_Bsu in membrane-spotting assays, suggesting that the functional unit harbors at least two CTP molecules (Fig. 2C and fig S4C). Indeed, two nucleotides are sandwiched by a ParB dimer in the ParB/nucleotide structure (Fig. 2D-E) being analogous to dimerization-activated NTPases of the large family of P-loop proteins (e.g. SIMIBI and ABC proteins including ParA and SMC, respectively) (25, 26). Residues E111 to E116 are located next to the nucleotide at the N dimer interface and appear critical for N engagement (Fig. 2F). They are conserved in ParB proteins but absent from monomeric Srx (fig. S1A). Consistent with a role in N engagement, N112S impairs parS-stimulated cooperative but not basal CTP binding (Fig. 2C). To detect N engagement in solution, we employed chemical cross-linking of purified ParB(T22C). T22 is ideally positioned at the symmetry axis of the ParB/nucleotide structure to support dimer cross-linking (Fig. 3D). ParB_Bsu(T22C) cross-linking was robust only in the presence of both CTP and ^parSDNA₄₀ (Fig. 3A). As expected for a N engagement-defective mutant, N112S abolished parS-stimulated T22C cross-linking (fig. S5C). Stimulation by ^parSDNA₄₀ occurred at sub-stoichiometric concentrations again implying that ^parS does not stably bind ParB and indicating that parS catalyzes the N engagement step (fig. S5D). The non-hydrolysable analog CTPγS also promoted efficient T22C cross-linking with ^parSDNA₄₀ but CDP failed to do so (Fig. 3A). In contrast to CTP, CTPγS supported robust cross-linking even without parS after extensive incubation periods (Fig. 3B and fig. S6B). This demonstrates that N engagement is energetically favorable, but the transition to the engaged state is very slow without parS. We conclude that parS catalyzes CTP-dependent engagement of domain N in ParB.

Fig. 3. ParB gate closure.

(A) Gel analysis of cross-linking products of purified ParB_Bsu(T22C). X denotes cross-linked species. CBB, Coomassie Brilliant Blue. (B) Time-course of ParB_Bsu(T22C) cross-linking with CTPγS without parS. Aliquots were taken at the indicated time points and mixed with BMOE. Quantification of cross-linked fraction also shown for CTP (see fig. S6B). (C) As in A using ParB_Bsu(T22C, S278C). X-T22C and X-S278C denote single cross-linked dimer species. XX marks double cross-linked ParB rings. Other species are oligomeric forms. (D) Model of ParB built from ParB_Bsu(21-218)/CDP/Ca²⁺ and ParB_Bsu(229-282) (PDB: 5NOC) dimers. Linker at the M-C junction is manually modelled as fully extended peptide. (E) Superimposition of domain M in ParB_Bsu(21-218)/CDP/Ca²⁺ and a ParB/parS structure (PDB: 4UMK). Only DNA is shown for the latter.

To determine whether N engagement occurs within or between ParB dimers, we purified ParB_Bsu(T22C, S278C) (Fig. 3C). In the presence of CTP and ^parSDNA₄₀, double cross-linked ParB dimers emerged as predominant species, demonstrating that N engagement produces mostly ring-shaped ParB dimers (and few ParB oligomers). This invokes the possibility that domain N serves as a nucleotide-operated DNA gate in ParB rings which closes at parS.

Comparing the ParB/nucleotide structure with a ParB/parS DNA co-crystal structure, it became apparent that parS DNA associating with one HTH motif clashes with the opposing HTH motif (Fig. 3E and fig. S6C). This indicates that parS DNA duplexes may detach from the HTH motif upon N engagement (or CTP hydrolysis) possibly explaining the transient nature of ParB/parS association. ParB dimers may, however, remain associated with chromosomal DNA even after release from parS if they were to entrap the DNA backbone. To test this, we next used circular DNA for in vitro DNA loading experiments (Fig. 4A). ParB_Bsu(T22C, S278C) was incubated with plasmid DNA (1873 bp) or linear DNA40 in the presence of CTP and cross-linked with BMOE to preserve topological interactions under protein-denaturing conditions (5). Control samples were nuclease-treated to release proteins from DNA. As expected, ParB_Bsu(T22C, S278C) generated large fractions of covalently closed ParB rings with parS DNA only (Fig. 4A). When nuclease digestion was omitted, the ParB ring species was retained in the ^parSDNA₄₀ sample but was eliminated from the ^parSplasmid sample containing CTP or CTPγS. Instead a slowly migrating species was observed near the top of the gel, presumably corresponding to ^parSplasmid interlocked with multiple covalent ParB rings. DNA staining of polyacrylamide and agarose gels supported the notion of multiple ParB rings entrapping ^parSplasmid (Fig. 4A and fig. S7). The efficient stimulation of ParB-DNA entrapment by CTPγS implies that CTP hydrolysis is dispensable for all steps of DNA loading. CTP hydrolysis might instead play a role in unloading and recycling of ParB thereby determining the residence time of ParB on chromosomes (and the extent of ParB spreading).

Fig. 4. ParB DNA loading and sliding.

(A) Polyacrylamide gel analysis of ParB_Bsu(T22C, S278C) protein species (CBB) and DNA species (SYBR) from BMOE-cross-linked DNA loading reactions. A presumed ParB-XX/parSplasmid adduct is marked by asterisk. CDP and scrplasmid (scrambled) only support limited ParB loading. (B) Single-molecule imaging (SMI) of DNA-bound TMRParB in the presence of HaloTagEcoRIE111Q roadblocks. Mean and standard deviation were calculated from three repeat measurements of a sample. (C) Residue contact maps by coevolutionary analysis (DCA; blue and yellow dots) and residue proximities (grey dots) in monomeric (below diagonal) and dimeric chains (above diagonal) of ParB_Bsu(21-218)/nucleotide and ParB_Bsu(229-282) (PDB: 5NOC). (D) Model for ParB spreading by self-loading of DNA sliding clamps at parS.

We next used single-molecule imaging to elucidate how CTP binding may relate to ParB spreading (27). The ends of linear 26 kb DNA molecules containing a single parS sequence were immobilized in flow chambers. Increased DNA occupancy of labeled ParB (^TMRParB) was detected in presence of CTP (fig. S8A). However, no local enrichment of ^TMRParB on DNA was observed, potentially due to ParB spreading from parS. We therefore used a labeled and catalytically inactive variant of the EcoRI restriction enzyme to test if this ^HaloTagEcoRI^E111Q protein could function as a roadblock, which might constrain ParB spreading. Intriguingly, under these conditions ^TMRParB became locally enriched on DNA between two EcoRI sites flanking parS (Fig. 4B), indicating that ^HaloTagEcoRI^E111Q had indeed restrained ParB spreading. TMRParB localized more broadly when these two parS flanking EcoRI sites were removed (Fig. 4B). Local enrichment of ^TMRParB was also lost when ATP was used instead of CTP or when ^HaloTagEcoRI^E111Q was added at the end of the experiment, despite normal localization of ^HaloTagEcoRI^E111Q (Fig. 4B, S8B). These results demonstrate that ParB enrichment at parS DNA is enhanced by CTP and indicate that ParB translocates onto parS flanking DNA, unless it is restrained by roadblocks. These findings support the notion that ParB forms self-loading DNA clamps whose sliding on chromosomal DNA is hampered by protein roadblocks. On the chromosome, ParB may frequently encounter roadblocks formed by nucleoid associated proteins (NAPs) and RNA polymerase. ParB diffusion might be locally restricted in a stochastic manner thus producing the characteristic gradient distribution patterns observed by ChIP-Seq (6, 13).

ParB_Bsu(T22C) cross-linking was rather inefficient in vivo (fig. S9), possibly indicating that levels of N engagement are low in the cell. Alternatively, cross-linking might be hampered for technical reasons or by steric blockage by other proteins such as ParA or SMC. To obtain information on ParB architecture by independent means, we extracted residue co-evolution signatures from sequence data by direct coupling analysis (DCA) (28). We obtained excellent agreement between the top 900 DCA contacts predicted from an alignment of 157776 ParB sequences and residue proximities observed in the ParB/nucleotide dimer (Fig. 4C). In contrast, the ParB/nucleotide monomer and published ParB structures failed to explain numerous DCA contacts or produced proximities not observed by DCA (fig. S10). These results strongly suggest that the dimer in the ParB/nucleotide structure closely resembles a state of ParB that is relevant in many bacteria.

In summary, our work shows that the widespread family of ParB proteins are enzymes that bind the atypical nucleotide CTP to promote spreading from cognate DNA sequences. CTP utilization enables robust enrichment of ParB at the rare recognition sites (~1 cognate per 10⁶ noncognate sites) even with little or no sequence-specificity in DNA binding (14). The CTPase domain is found in a large variety of protein sequences with diverse domain organizations (> 100; PFAM: PF02195), implying that it also plays prominent roles in other cellular processes.