Article

“F plasmid partition depends on interaction of SopA with non-specific DNA”.

Cet article démontre le rôle majeur que joue l’ADN du nucléoïde dans le processus de partition du plasmide F.

F plasmid partition depends on interaction of SopA with

non-specific DNA

Jean-Philippe Castaing,1_{Jean-Yves Bouet}1,2_{* and}

David Lane2

1_{Laboratoire de Microbiologie et Génétique}

Moléculaires, Université Paul Sabatier, F31000 Toulouse, France.

2_{LMGM, Centre National de Recherche Scientifique,}

F31000 Toulouse, France.

Summary

Bacterial ATPases belonging to the ParA family assure partition of their replicons by forming dynamic assemblies which move replicon copies into the new cell-halves. The mechanism underlying partition is not understood for the Walker-box ATPase class, which includes most plasmid and all chromosomal ParAs. The ATPases studied both polymerize and interact with non-specific DNA in an ATP-dependent manner. Previous work showed that in vitro, polymer- ization of one such ATPase, SopA of plasmid F, is inhibited by DNA, suggesting that interaction of SopA with the host nucleoid could regulate partition. In an attempt to identify amino acids in SopA that are needed for interaction with non-specific DNA, we have found that mutation of codon 340 (lysine to

alanine) reduces ATP-dependent DNA binding>100-

fold and correspondingly diminishes SopA activities that depend on it: inhibition of polymer formation and persistence, stimulation of basal-level ATP hydrolysis and localization over the nucleoid. The K340A mutant retained all other SopA properties tested except plasmid stabilization; substitution of the mutant SopA for wild-type nearly abolished mini-F partition. The behaviour of this mutant indicates a causal link between interaction with the cell’s non-specific DNA and promotion of the dynamic behaviour that ensures F plasmid partition.

Introduction

The main drivers of bacterial plasmid mitosis (partition) are ATPases with intrinsic filament-forming properties reminis-

cent of eukaryotic cytoskeletal proteins. Of the two main types known, the actin-like ParM proteins have so far yielded the clearest mechanistic insights. Binding of ATP enables ParM proteins to polymerize to form proto- filaments which are stabilized by interaction with cognate partition complexes and so lengthen, pushing the attached plasmids polewards, before finally disassembling (Moller- Jensen et al., 2002; Garner et al., 2004). The other type of partition ATPase (ParA/Soj) is characterized by Walker- box motifs and is more widespread, its genes being found in most known low-copy number plasmids and in the majority of sequenced chromosomes. ParA/Soj proteins also undergo ATP-dependent polymerization, but how this serves their partition function has been harder to discern.

In vivo and in the presence of their partition complexes they

are seen not as lengthy filaments but as assemblies which often appear to move to and fro through the cell (Marston and Errington, 1999; Quisel et al., 1999; Ebersbach and Gerdes, 2001; Lim et al., 2005; Hatano et al., 2007). Despite recent attempts to relate this oscillatory behaviour to segregation of plasmid copies (Ebersbach et al., 2006; Fogel and Waldor, 2006), we do not yet understand the molecular basis of either the dynamics of Walker-box partition ATPases or the partition mechanism they animate. This paper reports our exploration of the role played by a property that these proteins share, affinity for DNA.

There have long been reasons for thinking that interac- tion of ParA proteins with DNA contributes to the partition mechanism. Early studies of the purified SopA protein of plasmid F and ParA protein of prophage P1 revealed that their intrinsically weak hydrolytic activity could be moder- ately stimulated in vitro by either the cognate centromere- binding protein (SopB/ParB) or non-specific duplex DNA and more strongly by both in combination (Davis et al., 1992; Watanabe et al., 1992). Such stimulation has proved to be a general property of ParA/Soj proteins (Barilla et al., 2005; Leonard et al., 2005; Lee et al., 2006). Visualization of ParA proteins using immuno- fluorescence or fluorescent peptide fusions has shown that these proteins are associated with the nucleoid rather than spread through the cytosol (Hirano et al., 1998; Marston and Errington, 1999; Ebersbach and Gerdes, 2001), indicating an affinity of ParA proteins for DNA

in vivo. Moreover, where ParA proteins have been seen to

migrate, the mobility tends to be restricted to the nucleoid

Accepted 15 September, 2008. *For correspondence. E-mail jean-yves.bouet@ibcg.biotoul.fr; Tel. (+33) 561 335 906; Fax (+33) 561 335 886.

Molecular Microbiology (2008) 70(4), 1000–1011 _䊏 doi:10.1111/j.1365-2958.2008.06465.x First published online 7 October 2008

region (Marston and Errington, 1999; Quisel et al., 1999; Ebersbach and Gerdes, 2001; Hatano et al., 2007). These observations are suggestive but do not tell us how or at what point DNA intervenes in the process.

The view that non-specific DNA plays a role in partition took a more concrete form recently, when we found that such DNA counteracts polymerization of purified SopA protein (Bouet et al., 2007). At concentrations well below those likely to be encountered in vivo, DNA prevented SopA-ATP from forming polymers and provoked disas- sembly of pre-existing ones. SopB protein neutralized this effect by binding non-specifically to the DNA, thus masking it; in addition SopB stimulated polymerization through direct interaction with SopA. These reactions may well make up the key nexus of the partition mechanism. If so, it should be possible to locate on SopA a DNA inter- action domain that both mediates SopA polymer break down and plays an essential role in partition. We report here evidence for the existence of such a domain.

Results

Partition ATPases of the subclass to which SopA belongs (Type Ia – Gerdes et al., 2000) are distinguished from chromosomal and many plasmid homologues by an addi- tional N-terminal segment that enables the protein to repress transcription from its own promoter. As affinity for a specific binding site endows DNA-binding proteins with some general affinity for DNA regardless of sequence (Wang et al., 1977), it is possible that the N-terminal helix– turn–helix (HTH) motif of SopA that mediates binding to the

sop operators also contributes to the protein’s non-specific

DNA-binding activity, and perhaps to the partition mecha- nism. However, Walker-box partition ATPases having no N-terminal extension bind to non-specific DNA in vitro

(SojTth – Leonard et al., 2005) and associate with the

nucleoid in vivo (ParApB171– Ebersbach and Gerdes, 2001).

In the case of SopA, ATP is not needed for, and may actually inhibit, specific binding to operator DNA via the N-terminal HTH whereas it stimulates binding to non- specific DNA (Bouet et al., 2007), suggesting the existence of a second binding domain. Moreover, SopA proteins with mutated HTH motifs which fail to bind specifically to sop promoter region DNA still show ATP-dependent binding to non-specific DNA (our unpublished data). These observa- tions suggest that the non-specific DNA-binding surface is located in a more C-terminal region of SopA.

Identification of a putative DNA binding domain

Our approach to finding a potential DNA-binding domain was to search the only partition ATPase whose 3D struc- ture has been determined, the Soj protein of Thermus

thermophilus (Leonard et al., 2005), for surfaces likely to

interact with DNA. Leonard et al. (2005) showed that the

Soj monomer must attach ATP to dimerize, and must dimerize to fasten to DNA. By using a hydrolysis-deficient mutant, they were able to determine the structure of the dimer in the ATP-bound state. Inspection of the model shows one border of the dimer interface to be lined with positively charged amino acids (Fig. 1A), making it the most obvious candidate for interaction with phosphates of the DNA backbone.

Alignment with the corresponding region of diverse ParA proteins revealed an approximate matching of arginine and lysine residues with those of Soj (Fig. 1B), suggesting that such a surface might be similarly located in many proteins

of this group. Although SojTth shows little amino acid

sequence correspondence with SopA outside the ATP- binding motifs, the secondary structure predictions for the two proteins imply overall similarity (the SopA N-terminal

segment apart). In particular, theb-sheet and two major

helices which correspond in both the predicted and crystal

structures of SojTthare conserved in SopA. Mutation of the

arginine and lysine residues should enable testing of the involvement of this region in DNA binding.

Properties of putative DNA-binding mutants

The codons of five selected residues (Fig. 1B, green; Fig. 1C, bold) were mutated, and the mutant proteins subjected to two preliminary in vivo quality screens (Fig. 2A and B). One was measurement of intracellular SopA concentration, to monitor increased exposure to proteases, an indicator of misfolding, and changes in sopA expression. The other is based on the assumption that in a correctly folded protein, C-terminal mutations should not much affect the ability of the N-terminal domain to repress the sop operon promoter, provided interaction with SopB co-repressor is not perturbed. One of the mutant SopAs, R320G, failed on both counts and was not further analy- sed. Of the four others, steady-state concentrations of K303A and R346G were somewhat reduced while those of R333G and K340A were comparable to wild-type (Fig. 2A).

Only SopAK340A _{and SopA}R346G _{behaved like wild-type}

repressor, albeit with slightly lower efficiency in the latter

case (Fig. 2B). SopAK303A _{was effective only when pro-}

duced at a threefold higher rate from a pZS*21-based plasmid; a deficiency in SopB interaction might contribute here as K303 is within a segment that contains SopB interaction determinants (residues 206–313, Ravin et al.,

2003). Repression by SopAR333G_{was efficient although it}

did not respond to higher synthesis rates.

The four active mutant SopA proteins were then purified in C-terminal His-tagged form and their DNA-binding properties tested. Specific complexes formed by SopA on linear sopOP fragments in the presence of non-specific competitor DNA do not appear in mobility shift gels as discrete species; rather, the retardation is progressive, as SopA interaction with non-specific DNA 1001

© 2008 The Authors

if inherently unstable complexes shed their SopA after a time proportional to SopA concentration (Bouet et al., 2007). This is seen for wild-type SopA in Fig. 2C. By this test, the affinity of the K303A, R333G and K340A proteins for a linear sopOP DNA fragment was about the same as

that of wild-type SopA, while SopAR346G_{did not retard the}

fragment at all. This contrast with the relatively minor deficit in repression implies that the mutation interferes with operator binding by the HTH of the purified protein but is relatively unimportant in vivo, where interaction with SopB co-repressor may override the defect.

Binding of wild-type SopA to non-specific DNA, mea- sured by filter retention of protein-DNA complexes (Fig. 2D), was strongly stimulated by ATP but scarcely raised above no-nucleotide background levels by ADP. The K303A and R333G mutant SopAs had undiminished, indeed three- to fivefold greater, affinity for DNA. In con- trast, mutant K340A retained little ATP-dependent DNA binding capacity, and R346G none. As the N-terminal HTH motif of the K340A mutant still binds normally to

sopOP in vitro, this result is direct evidence for the exist-

ence of a distinct, non-specific DNA binding domain. Loss rates of mini-F plasmids carrying the sopA muta- tions were compared with that of a mini-F (pDAG115) whose partition has been disabled by deletion of the sopC centromere (Fig. 2E). The plasmids were derivatives of an

otherwise wild-type mini-F (pDAG173) that carries a con- stitutive promoter in place of sopOP, to avoid the compli-

cation of variable SopA repressor activity. The sopAK303A

mutant plasmid (pJP33) was even less stable than pDAG115, possibly through an effect on SopB interaction

(Ravin et al., 2003), while the sopAR333G_{plasmid (pJP35)}

was as stable as wild-type. More importantly, plasmids

with the DNA-binding alleles sopAK340A _{(pJP36) and}

sopAR346G_{(pJP37) were lost at 3.2% and 4.4% per gen-} eration respectively, representing 150- and 220-fold loss of stability, rates comparable to that of pDAG115 (5.5%); the slightly lower loss rate of pJP36 might reflect residual DNA binding of its SopAK340A (Fig. 2D). Hence, muta- tions that prevent non-specific DNA binding also result in partition deficiency.

Although the virtually complete loss by the R346G mutant of ATP-dependent DNA binding and partition capa- bility corresponds to the most clear-cut phenotype of the

kind sought, deficiencies of SopAR346G_{in operator binding}

(Fig. 2C) and other in vitro properties (see below) implied a

general loss of activity of the purified SopA::his6protein

that renders it unreliable as a basis for relating DNA binding to biological function. On the other hand, the properties of SopAK340A _{other than ATP-dependent DNA binding and} partition closely resembled wild-type. Further characteriza- tion was therefore focused primarily on this mutant.

Fig. 1. Identification of a candidate DNA binding interface in SopA.

A. Structure of T. thermophilus SojD44A.ATP (Leonard et al., 2005) in monomer (left) and dimer (right) forms. The yellow region is the dimer interface edge rich in accessible lysines and arginines, which are shown with side-chains in red and with residue numbers on the monomer; blue,b-sheets; green, ATP. Images were constructed by importing co-ordinates 2BEK from PDB entry into Chimera.

B. Conservation of lysines and arginines in C-terminal segments of partition ATPases of chromosomal (Soj of T. thermophilus, ParA of

C. crescentus and S. coelicolor) and plasmid

(SopA of F, ParA of P1, ParF of pTP228) origin. The portions shown are extracted from an alignment of the entire ParA/Soj family. Numbered SopA residues (green) are the presumed equivalents of the numbered Soj residues (red); the choice of equivalent residues (highlighted) in SopA and other proteins was made on the basis of closest alignment with the known surface residues of SojTth, and so is to some degree arbitrary. C. Secondary structures of the Soj and SopA segments shown in B, predicted using PSIPRED. Basic residues corresponding to those highlighted in A and B are in bold. Note that the secondary structure of Soj derived from the 3D model in A, shown at the bottom, differs from that generated by the prediction programme.

SopA oligomerization is unaltered by K340A

Leonard et al. (2005) reported that ATP-dependent dimer-

ization of SojTthis needed for the protein to bind to DNA,

raising the possibility that the inability of SopAK340A_{to bind}

DNA might result from failure to form dimers. Our attempts to use size-exclusion chromatography to analyse di- oligomerization in SopA-ATP mixtures were unsuccessful: Fig. 2. Properties of mutant SopA proteins.

A. Concentration of SopA produced in exponentially growing cells from pLtetO-1on mini-F plasmids derived from pDAG173, measured by Western blotting of PAGE-fractionated extracts.

B. Repressor activity of SopA produced from the above mini-F plasmids, which include sopB and sopC, or from pLtetO-1on pZS*21-derived plasmids carrying sopB; values are percentages of the unrepressed psop::lacZ values and are the averages of at least three independent assays, except for R320G and R333G (two and one assays respectively).

C. EMSA of SopA::his6(0, 0.1, 0.25, 0.8mM) binding to a 168 bp32P-labelled sopOP fragment with a 128 bp32P-repE fragment as non-specific control.

D. Fraction of non-specific DNA (RepE DNA) bound (ordinate) as a function of SopA concentration (mM, abscissa); error bars are standard deviations of at least three independent assays. Assays with no nucleotide yielded values equal to or less than those for ADP, and have been omitted for clarity.

E. Partition function, measured as the rate of mini-F loss from exponentially growing cells per generation (ordinate) and expressed relative to wild-type mini-F loss rate as the destabilization factor (figures below). The dotted line represents random loss of a presumably freely diffusing mini-F lacking the sopC centromere. Error bars in (A) and (E) are standard deviations derived from at least three independent measurements.

SopA interaction with non-specific DNA 1003

© 2008 The Authors

a small fraction of loaded SopA eluted as a dimer when elution buffer without ATP was used, whereas inclusion of ATP resulted in wholesale polymerization. To compare oligomerization of the K340A mutant and wild-type pro- teins, we turned to cross-linking using the lysine-bridging

reagent bis-(sulphosuccinimidyl) suberate (BS3_{). Cova-}

lent joining of monomers by this method requires not only that oligomers are present but also that their subunits are suitably oriented. The electrophoretic profiles of the reac- tion products (Fig. 3) showed that in the presence of ATP both proteins were cross-linked to a major species migrat- ing as a SopA dimer, with apparent trimers also constitut- ing a significant fraction. This result was unaffected by addition of DNA to the reaction mixture (not shown). Inclu- sion of EDTA (only the K340A result shown) prevented dimer formation, as expected for removal of the magne- sium needed for stable ATP binding. For both proteins,

reactions in which ATP was replaced by ADP or ATPgS

generated small amounts of uncharacterized products,

one of ~85 kDa migrating close to but behind the ATP

dimer and several migrating at~100–120 KDa; with AMP-

PNP or no nucleotide, trace amounts of the ~85 kDa

species were seen. The essential point is that the same products in approximately the same proportions were formed from both wild-type and mutant proteins, imply- ing that monomer composition and orientation in the SopAK340A_{oligomers is unaltered.}

Reaction of SopAK340A_{polymers to DNA and SopB}

Our previous study identified formation and maintenance of polymers as the point at which non-specific DNA had its most marked effect on SopA behaviour. Accordingly, we first ascertained that the K340A mutant could polymerize (as could K303A and R333 but not R346G; not shown), and then examined the sensitivity of polymer formation

and persistence to DNA, using dynamic light scattering (DLS). Following addition of ATP, the K340A mutant and

wild-type SopA::his6 proteins polymerized with similar

kinetics and to about the same extent (Fig. 4A). Polymer- ization was strongly stimulated by SopB in both cases (Fig. 4B), as reported previously for wild-type SopA (Bouet et al., 2007); SopB alone did not increase light scattering above background in these conditions (data not shown, and Bouet et al., 2007). This stimulation indicates that the K340A mutant interacts normally with SopB in

vitro. But whereas DNA completely prevented polymeriza-

tion of wild-type SopA, it had no effect on that of SopAK340A

(Fig. 4C). Likewise, addition of DNA to preformed wild- type polymers caused their immediate disassembly but

did not detectably alter polymers of SopAK340A_{(Fig. 4D).}

This failure of the mutant polymer to react to DNA is consistent with the failure to bind it seen earlier (Fig. 2D).

Response of SopAK340A_{-catalysed ATP hydrolysis}

to DNA

Hydrolysis of ATP by SopA, which appears to be neces- sary for partition (Libante et al., 2001), is inherently weak but is stimulated by non-specific DNA in synergy with SopB (Watanabe et al., 1992). Inability of SopA to interact with DNA might thus limit its hydrolysis of ATP in vivo to the point where partition is compromised. We compared the effects of DNA and of SopB on ATP hydrolysis by the

DNA-binding deficient mutant SopAK340A_{and by wild-type}

SopA. The intrinsic ATPase activities of the two SopAs are similar (Fig. 6, left bar of each panel). Addition of SopB resulted in an approximately twofold increase in ATP hydrolysis by both proteins. Although addition of DNA appeared to cause a slight stimulation of ATP hydrolysis by wild-type SopA not seen with the K340A mutant, this was not outside the range of experimental error. A clear

Fig. 3. Cross-linking analysis of SopA oligomerization. The concentration of SopA protein in the reaction mixtures was 0.4mM monomer. BS3_{denotes the presence of 2 mM} cross-linking reagent (see Experimental

procedures). At left are molecular weights in

kDa of protein size standards. At right are