Role of centromere sites in activation of ParB proteins for partition complex assembly

The ParB-parS partition complexes that bacterial replicons use to ensure their faithful inheritance also find employment in visualization of DNA loci, as less intrusive alternatives to fluorescent repressor-operator systems. The ability of ParB molecules to interact via their N-terminal domains and to bind to non-specific DNA enables expansion of the initial complex to a size both functional in partition and, via fusion to fluorescent peptides, visible by light microscopy. We have investigated whether it is possible to dispense with the need to insert parS in the genomic locus of interest, by determining whether ParB fused to proteins that bind specifically to natural DNA sequences can still assemble visible complexes. In yeast cells, coproduction of fusions of ParB to a fluorescent peptide and to a TALE protein targeting an endogenous sequence did not yield visible foci; nor did any of several variants of these components. In E.coli, coproduction of fusions of SopB (F plasmid ParB) to fluorescent peptide, and to dCas9 together with specific guide RNAs, likewise yielded no foci. The result of coproducing analogous fusions of SopB proteins with distinct binding specificities was also negative. We conclude that in order to assemble higher order partition complexes, ParB proteins need specific activation through binding to their cognate parS sites.


Introduction
The hub of the mechanism that drives bacterial mitosis, or partition, is a complex formed by binding of a specific ParB protein to a small number of clustered parS binding sites. The parS array functions as a centromere, and the complex serves as a kinetochore by activating the corresponding ParA ATPase to segregate replicas of its own chromosome or plasmid to incipient daughter cells. ParBs of most low copy-number plasmids and of all known chromosomes bind to their cognate parS sites as dimers, primarily via a helix-turn-helix (HTH) motif. They also bind, more weakly, to non-specific DNA. But unlike other proteins that bind through HTH motifs, notably transcription regulators, ParB proteins selfassociate (oligomerize) and so pervade nearby DNA to enlarge their complex, a process termed "spreading". Spreading is not only integral to the partition process but has also enabled use of ParB/S systems as an alternative to fluorescent repressor-operator systems (FROS) for visualization of genetic loci in bacteria [1,2]. We have developed them for use in eucaryote cells and viruses as the ANCHOR system (3)(4)(5). They offer certain advantages over FROS: the weakness of ParB oligomerization and DNA binding interactions allows other chromatin-based processes to disperse the complexes easily, making them less disruptive than FROS, and the small number of integrated parS binding sites involved is less locally intrusive than the hundreds typical of FROS. Nevertheless, dispensing with the need for prior parS integration through direct binding to endogenous sequences would eliminate potential artifacts of even such minor genome modification and would greatly streamline the procedure. Fusion of the ParB and fluorescent peptide (FP) components to proteins whose binding can be tailored to recognize natural genome sequences -TALE and Cas9 -might allow specific tagging of unaltered sites while preserving the advantages of ParB/S systems.
However, this would work only if ParBs can spread without first binding to their parS sites. It was not clear that they can. To assess the feasibility of removing the parS integration step from the ANCHOR system we have aimed in the work reported here to identify the interactions that enable ParB spreading.
HTH-type ParB proteins share a broadly consistent three-domain organization of both structure and function. The intrinsically disordered N-terminal domain interacts with ParA [6] and with itself [7], the central domain contains the motifs responsible for specific parS recognition and for DNA binding [8][9][10][11][12], and the C-terminal domain dimerizes the protein [6,7,9,13] to form the basic active ParB unit; in certain ParBs the latter also plays roles in DNA binding and ParA interaction [14][15][16].
Early studies of mutant ParBs of the P1 and F plasmids implicated the N-terminal domain of these proteins in spreading [8,[17][18][19], as also confirmed later for chromosomal ParBs [14,20], and suggested that spreading is needed for partition. These observations, together with the demonstration that the N-terminal domain oligomerizes P1 ParB dimers [7], the known binding of HTH proteins to non-specific DNA and the small number of foci seen in cells containing fluorescent ParB derivatives [21,22] led to the general view that ParB bound in the core complex recruits further ParB molecules whose weak interactions with themselves and with neighbouring non-specific DNA create a metastable complex large and cohesive enough to activate partition [23]. Spreading was initially envisaged as proceeding laterally from the centromere along adjacent DNA. However, certain observations were inconsistent with this view [24,25], and Bouet et al. [26] proposed that ParB spreads not only in cis from the nucleating complex but also in trans to the nucleoid and to distant sites on the same molecule, like bees round a hive rather than birds on a wire. The transient bridging (trans) and looping (cis) interactions and the indeterminate form of the complex implied by this proposal have since been substantiated and refined by studies of complexes formed by the ParB proteins of several species [27][28][29][30][31], and the idea has recently been extended to the chromatin realm [32].
However, the full role of centromere binding in formation of higher-order partition complexes is not yet understood. Since spreading is not seen to occur spontaneously, in the absence of parS, it would appear to need a specific switch in ParB conformation. Is this induced directly upon binding to parS? Or is it a consequence of the oligomerization interaction of ParB N-terminal domains, with parS binding serving only to focus and anchor the complex ? (Fig. 1) On one hand there are indications that the properties of ParBs do change in response to centromere binding: F SopB and P1 ParB co-repressor activity is stimulated in trans by sopC and parS respectively [33,34], SopB-mediated stimulation of SopA ATP hydrolysis is enhanced by sopC [35] and a newly-discovered CTPase activity exhibited by some ParB proteins, including SopB, is enhanced by centromere binding [36].
On the other hand, Surtees & Funnell [7] observed oligomerization of P1 ParB N-terminal domains in the absence of parS in vitro and in yeast, and Hyde et al. [37] concluded that specific binding of KorB did not reduce intrinsic disorder but rather selected from a population of natural conformers. We report here our attempts to observe spreading of ParBs fused to non-parS DNAbinding proteins, and to distinguish ParB-parS binding from ParB-ParB interaction as the basis for conversion to spreading competence.

Strains
Bacteria: E.coli K12 strains used in microscopy and gene expression experiments were derivatives of W1485, as detailed in Table 1 and schematized in Fig S2. Transformation recipients were DH5 and DH10B [38] except for constructions involving recombinationprone RVD repeats of tal genes where SURE2 (Stratagene) or a derivative cured of the F', D111, was used. The host for recombinational transfer of centromere sequences from plasmids to λ phages [39] was MC1061 [40].

Plasmid constructions
Construction outlines are given here; details are available on request.

ANCHOR system visualization.
Tale constructs were based on the pZHY501 shuttle (S.cereviseae-E.coli) vector, provided by Daniel Voytas (via Addgene) [44], which carries the Nt and Ct (non-RVD) domains of the X.oryzae Tale PthXo1 gene fused to the FokI nuclease coding sequence. pZHY501was modified by site-specific mutagenesis to introduce sites for AvrII and NruI immediately upstream of the Tale Nt sequence, and by deletion of the fokI gene using BamHI, BsaBI and Klenow polymerase, yielding pVR203.
Repeat variable di-residue (RVD) domains specific for URA3 Nter nt 17-32 (U3aL) and Cter nt 632-648 (U3bR) were obtained as BsmBI site-ended PCR products from plasmids kindly provided by Bing Yang [45], and inserted between the BsmBI sites in pVR203 to create the Nt::ura3::Ct fusions in pVR204 (U3aL) and pVR206 (U3bR). A codon-optimized synthetic or3 (parB) gene [4] was obtained as a PCR product with terminal AvrII and NruI sites and inserted between these sites in pVR204 to create the or3s::tal.U3aL fusion in pSA316. The stop>leu-mutated codon of the same or3s gene was fused to codon 2 of the mCherry coding sequence; the or3s::mCh fusion was inserted into a vector then excised with KpnI and NotI and inserted between these sites of pRS424 [46], yielding pSA312.
Plasmids carrying sopB N15 were derived from pZC326 [43], a mini-F -mini-P1 hybrid into which we inserted a high copy-number origin, making pDAG382, to facilitate construction. To place sopB N15 under araC-paraBAD control we substituted it for sopB F in pDAG170 [26]. The araC-para-sopA'-sopB N15 expression unit was joined to the mini-P1 portion of pZC326 to make the binding plasmid pNR189, and sopA'-sopB N15 deleted from pNR189 to make the control plasmid pNR195. The megf gene (see below) was amplified by PCR and fused to the sopB N15 3' end in pNR189 to make pNR197.
Plasmids for centromere transfer to attλ: the sopC sequence was inserted between the kan gene and the pldc promoter in pDAG123 [42] to give pDAG418. A frt-aadA-frt cassette was inserted upstream of the 4xIR unit (see below) in pDAG541 to give pDAG545.

Yeast:
The basic medium was SC: 0.67% yeast nitrogen base (Difco) supplemented with all amino-acids except those used for selection, uracil, adenine and 2% glucose.Strains W303, ySA27 and ySA46 transformed with pSA312 alone or with pSA312 and pSA316 were grown overnight in SC lacking leucine (SC-LEU) or leucine and tryptophan (SC-LEU-TRP) respectively.

Yeast:
Live-cell microscopy was performed as described [52], using an Olympus IX- Silencing assay Transcription reporter strains freshly transformed with plasmids carrying the sopB genes to be tested were grown as for microscopy (above) at 30°C to an optical density of 0.1-0.2 and chilled on ice. Samples were removed for assay of galactosidase and measurement of optical density as described [53].

Results
We begin by describing two of our attempts to observe partition complex assembly primed by specific binding to DNA sites other than the ParB protein's own centromere. One employs a plasmid site inserted in the E.coli xylE gene, the other involves several sites, natural and exogenous, within S.cereviseae chromosomes.

E.coli xylE
The CRISPR-Cas9 system was used. A guide RNA with a 20 nt sequence of xylE (sgRNA-xylE) was co-produced with a polypeptide comprising the F plasmid SopB protein fused at its C-terminus to the enzymatically inactive dCas9 protein. The ability of this fusion to recognise its target was confirmed by the lethality of the SopB fusion to active Cas9 both in xylE + cells producing sgRNA-xylE and in cells with an insertion of the F sopC centromere producing sgRNA-sopC (Table S1). The ability of xylE-bound SopB::dCas9 to prime spreading was tested by observing formation of fluorescent foci in cells also producing SopB::mVenus, a fusion protein known to act normally in complex assembly and plasmid partition [51]. To prevent saturation of the incipient partition complex by SopB::dCas9, its production was kept to a minimal level by allowing transcription from pLtetO only at the basal, uninduced level, while strongly inducing production of SopB::mVenus. Compact foci of normal number and distribution appeared in xylE::sopC cells (Fig 2A); this was due to direct binding of SopB::mVenus to sopC, unmediated by the dCas9 fusion, since the foci also formed when the guide RNA carried a random sequence (Fig. 2C). Cells without sopC expressing the same sgRNA and fusion genes showed no foci, only evenly distributed fluorescence (Fig 2B,D). It is possible that the single xylE binding site used does not form a core complex sufficiently robust to trigger spreading, even though a single SopB binding site does so (Fig.   S3). So we redid the experiment with a tandem-repeat binding sequence that resembles the natural sopC centromere, using a guide RNA sequence corresponding to eight of the ten functional 43bp repeats that make up sopC [54]. The R219A mutant derivative of SopB was used in the fusion proteins to prevent specific binding to sopC while still allowing the nonspecific DNA binding needed for spreading [12]. Fig 2 E,F shows that this modification did not enable focus formation either.

S.cereviseae URA3
Several ParB proteins, when fused to fluorescent peptides, form visible complexes in yeast strains engineered to harbour small arrays of their parS binding sites [3]. We examined the ability of such fusion proteins to form foci in the absence of their cognate parS sites when specific binding was provided by TALE proteins [55]. Fig. 3 shows the results of a representative experiment, employing an experimental format similar to that of Fig. 2. Yeast cells transformed with a plasmid from which the ParB fusion protein Or3::mCherry is produced formed one distinct focus in each nucleus, as expected for cells in G1, provided they have an integrated copy of Or3's cognate centromere site, Anch3 (Fig 3A). No foci were seen in cells of the parental strain, which lacks this site (Fig. 3B, C). When Or3::mCherry was coproduced with a fusion of Or3 to a Tale peptide known to bind specifically to the 5' end of the URA3 gene (Or3::Tale.U3L), it still formed foci in the Anch3 strain ( Fig. 3D) but also still failed to in cells without Anch3, whether or not they contained the Tale.U3L binding site (Fig 3E, F).

Fig 3. Test of spreading by ParB specifically bound via fusion to Tale proteins in yeast.
Or3::mCherry fusion protein production from pSA312 in : A-strain ySA27, with Anch3 inserted, B-ySA46, with no insertion, C-w303, with a deletion in URA3 removing the specific Tal.U3L binding site (blue strip). Or3::mCherry production together with

Spreading of hybrid SopB proteins
It should be possible to distinguish ParB-parS binding from ParB dimer-dimer interaction as the event that enables spreading by using hybrid proteins with centromerebinding and oligomerization domains of distinct specificities. A minimal complex seeded with limiting amounts of one ParB (the binding protein) might be expanded to a large complex upon provision of a second, hybrid ParB (the signalling protein) that shares the Nterminal oligomerization domain but does not bind to the same parS: expansion of the complex could be observed by tagging the second ParB with a fluorescent peptide, and would indicate that direct binding to parS is not needed for spreading. The experimental set-up is similar to that of Fig. 2, but here the initiating complex is natural and known to trigger spreading, and no bulky, potentially interfering foreign protein is involved.
The closely-related Sop partition systems of plasmid F and prophage-plasmid N15 appeared suitable for applying this approach. Their SopB proteins are very similar, at 49% amino acid identity; SopB of F functions only with its cognate binding site (10 tandem copies in the F centromere, sopC [54]), not with those of N15 (IR1-4; [56]) [9]; and many N15:F hybrid proteins are functional, interacting with their SopA and centromere partners with the expected specificity [9]. One of these SopB proteins, SopB N15/F (termed hybrid 10 by Ravin et al., [9]), comprises the N-terminal domain of N15 SopB and the DNA-binding and dimerization domains of F SopB (Fig. 4, top left). It should be able to interact via its Nterminal domain with N15 SopB bound to IR centromere sites, but be unable to bind to these sites itself. The distribution of fluorescent SopB N15/F protein confirms this specificity: discrete foci are seen in cells with sopC integrated as part of a prophage at att (Fig. 4A), whereas in cells with an analogous N15 centromere-site array (4xIR), in which N15 SopB forms normal foci (Fig. 4C), SopB N15/F ::megfp fluorescence diffuses evenly throughout the cell (Fig. 4B).
The SopB N15/F ::megfp fusion is thus sufficiently specific to serve as a signalling protein. Accordingly, we tested whether foci initiated by wt SopB N15 produced at 0.1 and 0.3µM arabinose could be expanded to visible size by spreading of SopB N15/F ::megfp produced at 1nM aTc, the optimal concentration for discrete focus formation on sopC (Fig.   4A), and at 3nM aTc, for a moderate over-production (Fig. 4B) to allow for the possibility that spreading in this heterologous system is less efficient. The results (Fig. 4D) showed no focus formation with any combination of SopB N15 and SopB N15/F ::megfp concentrations.
A variant of this approach, instigation of spreading of a ParB protein unable to bind specifically to an available centromere, is to use a mutant of a natural protein that lacks centromere-specific binding activity but is still capable of the non-specific DNA binding needed for spreading. The R219A derivative of SopB used in the experiment of Fig. 2 is such a mutant. We essentially repeated this experiment using SopB F without the large peptides to which it had been fused to enable specific binding (Fig. 5). After determining the inducer concentrations suitable for producing low levels of SopB (0.1µM arabinose) and SopB F.R219A ::megfp (1nM aTc), using silencing assays (Fig. S1B, C; Fig. 5A), we examined the ability of the former to initiate focus formation by the latter. No foci were seen. These results provide no support for the proposal that interaction of SopB N-terminal domains alone can generate the spreading needed to assemble a functional partition complex.

Discussion
The conclusion drawn from our attempts to prime complex expansion from noncentromere DNA sites -that the ParB proteins need their cognate centromeres to become capable of spreading -appeared clear-cut at first view. But the possibility that the bulky Cas9 and Tale peptides to which they were fused prevented acquisition of spreading competence could not be eliminated. It was therefore important to test centromere-independent spreading without them. The use of the hybrid SopB and the SopB mutant lacking specific binding activity served this purpose. That these proteins also failed to spread when primed by core complexes whose SopB proteins shared their oligomerization domains reinforces the original SopB as readily as it does in the presence of its own centromere (Fig. 4A), which it clearly did not (Fig. 4D) [4]. To reconcile our in vivo conclusion with the SPR and footprinting data it appeared necessary to posit a cellular element needed for rapid release of activated dimers that is absent from the in vitro assays. And indeed, such an element has very recently been identified -cytidine triphosphate (CTP [36]). These authors discerned conserved motifs in the Nter region of several ParBs, including SopB, that enable binding of CTP. The binding was strongly stimulated by centromere DNA. In the case of the B.subtilis protein, binding to parS and to CTP induced interaction between Nter domains to form, as the major product, a dimer ring. Stimulation of ring formation by parS at sub-stoichiometric levels suggested that the rings vacate their binding site rapidly to slide along adjacent DNA, i.e. to spread. If future work shows CTP-SopB-sopC to behave in this way, the discrepancy between our focus-formation and in vitro binding results would disappear.
Given that activation of spreading ability depends on direct contact with the centromere, efforts to bring about ANCHOR visualization without it would now appear futile barring technical innovation. On the other hand, it might be possible to create mutant ParBs predisposed to adopt a spreading-competent conformation independently of their centromeres. The energy barrier to such conformers may well be low; Soh et al. [36] observed CTP to stimulate some formation of B.subtilis ParB dimer rings in the absence of parS, presumably from a subset of suitable conformers normally promoted by parS binding.
A search for suitable mutants is clearly a priority. Superscripts denote previously published strains (see Table 1). Lysogenization by RS phages to integrate promoter-lacZ fusions has been described [39]. pcry denotes a cryptic promoter.

Table S1
Interaction of target sequences with Cas9 guide RNAs. Strains carrying chromosomal xylE and sopC sequences on the chromosome and genes for the corresponding sgRNAs on an expression vector were transformed with plasmids from which production of the SopB::dCas9 fusion or the equivalent active Cas9 fusion protein could be induced.
Viability of transformants on agar medium was scored (see Materials & methods).