Breaking and Restoring the Hydrophobic Core of a Centromere-binding Protein

Background: Accurate segregation of antibiotic resistance plasmids requires dedicated centromere-binding proteins. Results: Assembly of the hydrophobic core of the ParG centromere-binding protein encoded by multiresistance plasmid TP228 is governed by a triad of key amino acids. Conclusion: ParG retains functionality with certain substitutions of the core triad. Significance: The study provides valuable insights into how multiresistance plasmids are maintained in bacteria. The ribbon-helix-helix (RHH) superfamily of DNA-binding proteins is dispersed widely in procaryotes. The dimeric RHH fold is generated by interlocking of two monomers into a 2-fold symmetrical structure that comprises four α-helices enwrapping a pair of antiparallel β-strands (ribbon). Residues in the ribbon region are the principal determinants of DNA binding, whereas the RHH hydrophobic core is assembled from amino acids in both the α-helices and ribbon element. The ParG protein encoded by multiresistance plasmid TP228 is a RHH protein that functions dually as a centromere binding factor during segrosome assembly and as a transcriptional repressor. Here we identify residues in the α-helices of ParG that are critical for DNA segregation and in organization of the protein hydrophobic core. A key hydrophobic aromatic amino acid at one position was functionally substitutable by other aromatic residues, but not by non-aromatic hydrophobic amino acids. Nevertheless, intramolecular suppression of the latter by complementary change of a residue that approaches nearby from the partner monomer fully restored activity in vivo and in vitro. The interactions involved in assembling the ParG core may be highly malleable and suggest that RHH proteins are tractable platforms for the rational design of diverse DNA binding factors useful for synthetic biology and other purposes.

The mechanisms of bacterial DNA segregation are best described for low copy number plasmids. Three principal segregation complexes have been characterized. Each comprises a cis-acting centromere site that is bound by a centromere-binding protein, which in turn interacts with a polymerizing NTPase that drives plasmid movement (1)(2)(3)(4). In the case of the actin-like ParM protein, ATP-mediated polymerization separates and pushes plasmid pairs to opposite cell halves (5). In contrast, the tubulin homologue TubZ assembles into filaments upon GTP binding. Concomitant growth and disassembly of opposite ends of these filaments (treadmilling) induces the directional displacement of attached plasmids (6 -8). However, the most widely distributed plasmid partition NTPases are those of the ParA superfamily. Numerous ParA homologues have been shown to polymerize in response to ATP binding. Polymerization is modulated by the partner centromere-binding protein and/or by DNA (9 -25). Significantly, mutation of residues at the polymer interface disrupts both polymerization and segregation (26). Nevertheless, unlike with actin-and tubulin-based complexes, the mechanism that underpins ParA-mediated plasmid partitioning remains elusive, although the nucleoid may act as a substratum on which segregation occurs (17,(27)(28)(29)(30)(31).
The partition complex of multidrug resistance plasmid TP228 comprises the ParA homologue, ParF; the ParG centromere-binding factor; and, the parH centromere site (32,33). The N-terminal tail of ParG is intrinsically unstructured whereas the C-terminal domain adopts a ribbon-helix-helix (RHH) 3 fold (34). The RHH superfamily of DNA-binding proteins is dispersed widely in bacteria and archaea. The dimeric RHH fold involves the interweaving of two monomers into a 2-fold symmetrical structure composed of four ␣-helices surrounding a pair of antiparallel ␤-strands (ribbon) (Fig. 1A). Amino acids that comprise the ribbon region are the major determinants of DNA binding, with three hydrophilic residues in particular making a set of base-specific contacts in the major groove (35,36). Extensive sequence variability in the ribbon element permits diverse RHH proteins to interact with discrete DNA sequences (37). In the case of ParG, the ribbon recognizes an array of degenerate tetramer boxes interspersed by AT-rich spacers both in the parH centromere, and at the operator site during transcriptional repression of the parF-parG genes (33,38).
The assembly of a linear polypeptide into a folded, compact structure is governed principally by hydrophobic interactions. These interactions sequester hydrophobic surface area from external solvating water thereby providing an increase in entropy and tight packing of the protein interior (39). The hydrophobic core of dimeric RHH proteins is formed by residues located in both the ␣-helices and ␤-strand elements. In particular, seven conserved branched hydrophobic residues, four in the ␣-helices and three in the ␤-strand, in both monomers are implicated in organizing the RHH core (37). In view of the important role of ParG in stable maintenance of a multiresistance plasmid, here we characterize positions in the ␣-helices that are especially crucial for ParG function in segregation and in assembly of the protein core. A hydrophobic aromatic amino acid at one of these positions was functionally replaceable by other aromatic residues, but not by non-aromatic hydrophobic amino acids. Nevertheless, intramolecular suppression of the latter by a compensatory change to a residue that approaches nearby from the partner monomer restored activity in vivo and in vitro. The data suggest that the hydrophobic interactions involved in assembling the ParG core are highly tractable. Combined with recent observations that the ribbon element that determines DNA binding by ParG is functionally interchangeable with ribbons from other RHH factors (40), ParG and other RHH proteins may be potent platforms for the rational design of compact transcription factors valuable for synthetic biology and other purposes.
Plasmids-Plasmid pFH547 comprises the parFGH cassette of multiresistance plasmid TP228 cloned in the plasmid partition vector, pFH450 (32). Site-specific mutations in parG were constructed in pFH547 by cassette mutagenesis (45) or by overlap extension PCR (46). Mutated genes additionally were amplified from pFH547-based plasmids and cloned in pT18 and pT25 two-hybrid vectors (47) and in the pET22-22b(ϩ) expression vector (Novagen) as described previously (48). Mutations in all cases were verified by sequencing.
Plasmid Segregation Assays-Segregation assays were conducted using pFH547 derivatives that replicate at low copy number in strain BR825 as detailed elsewhere (32). Briefly plasmid-bearing strains were grown for ϳ25 generations without chloramphenicol selective pressure. Plasmid retention was determined by replica plating colonies to LB agar medium with and without antibiotic. Results are means of at least three independent tests with typical standard deviations of ϳ10%.
Two-hybrid Assays-Two-hybrid assays were based on reconstitution of two subunits, T18 and T25, that confer adenylate cyclase activity in E. coli (47). Plasmid pairs producing fusions of these subunits with either wild-type or mutated ParG were cotransformed in strain SP850. Interactions were monitored by ␤-galactosidase assays on cultures grown at 30°C for ϳ16 h. Results are means of at least three independent tests with typical standard deviations Ͻ10%.
Protein Purification and Analysis-Wild-type and mutant ParG were purified as His-tagged proteins as described previously using pET22b(ϩ) expression plasmids (48). Cross-linking experiments with dimethyl pimelimidate (DMP; 10 mM) were performed as outlined elsewhere (48). Size exclusion chromatography coupled to multi-angle light scattering (SEC-MALLS)  Fig. 6, A and D were made using PyMol (70). B, primary sequence of ParG with secondary structure elements shown beneath. C, plasmid partition assays with ParG derivatives carrying alanine mutations in ␣1 and ␣2. Results are means of at least three independent tests with typical standard deviations of ϳ10%. Blue and gray bars indicate amino acids located in ␣-helical and non-␣ helical regions, respectively. Residues Phe-49, Trp-71, and Leu-72 are highlighted with arrows in panels B and C. was used to determine the average mass of ParG proteins eluting from a Superdex-75 10/300 column (GE Life Sciences). Proteins (200 l) were loaded onto the column in 10 mM Tris, 50 mM KCl, pH 7.5 typically at concentrations Ͼ150 M. The samples passed through a Wyatt Helios multi-angle detector, rEX differential refractive index detector, QELS (for the simultaneous analysis of diffusion) and UV detectors before being collected. Data were analyzed using a modified Zimm plot using a 2nd-degree polynomial fit and an estimated dn/dc of 0.18 ml/g within the program Astra version 5.3.4. Sedimentation equilibrium analysis was performed in a Beckman XLA analytical ultracentrifuge. Mid-peak fractions from SEC-MALLS were pooled and diluted in buffer (10 mM Tris-HCl, pH 7.5, 50 mM KCl) to give three concentrations typically in the range of 10, 25, and 100 M. Samples were loaded into 6-sector Epon centerpieces and centrifuged at running speeds of 20,000, 30,000, and 45,000 rpm and then scanned at 280 nm after 14 h. The molar extinction coefficients used were 5000 and 16800, respectively, and a vbar of 0.72 ml/g was used in all cases. Data were analyzed globally with the Sedfit/Sedphat suite of programs (49,50) using a monomer-dimer equilibrium model.
Gel Retardation Assays-The parH centromere site (33) was amplified from pFH547 using one primer bearing a 5Ј-biotin label and a second unlabeled primer. Conditions for DNA binding, gel electrophoresis, and detection of labeled DNA using the LightShift chemiluminescent EMSA kit (Pierce) were described previously (48).

RESULTS
A Triad of ␣-Helix Residues Are Critical for ParG-mediated Plasmid Segregation-The folded C-terminal domain of ParG comprises a single hydrophobic core formed by the symmetrical intertwining of a pair of monomers into a RHH fold (34). Molecular hydrophobicity potential (MHP) previously was used to calculate the strength of each inter-residue contact in ParG. First, by comparing the MHP contact plots and contribution of individual amino acid side-chains of ParG, interprotomer contacts were revealed to be stronger than intraprotomer contacts. Second, the analysis suggested that the strongest hydrophobic contacts within the ␣-helical region that hold the ParG dimer together potentially are formed by the side-chains of amino acid residues Lys-45, Arg-48, and Phe-49 in ␣-helix 1 (␣1) and Val-64, Leu-67, Val-68, Trp-71, and Leu-72 in ␣-helix 2 (␣2) ( Fig. 2A) (34).
The contribution of each side-chain in the ␣-helices to the surface area change on dimerization was calculated ( Fig. 2A). Trp-71 is most prominent, followed by non-polar sidechains Leu-72, Leu-67, and Val-68. Certain charged side-chains, e.g. Lys-45 and Arg-48, also contributed to the surface area change. However, charged side-chains are likely to be more flexible than  (34). B, contribution of charge-charge interactions to ParG dimerization. Calculations of charge-charge interactions were made with pK a calculations based on Debye-Hückel modeling of interactions between ionizable group sites, in water dielectric, at ionic strength 0.15 M (71). The calculations were done with the dimer structure in which each monomer is taken from the dimer. Unfavorable contributions to dimerization that occur upon intertwining partly account for the addition of charge-charge and buried non-polar surface in the dimer summing to more than the observed association. There also is scope for charges to re-arrange when in the monomer state such that favorable interactions are made that are not necessarily apparent in the dimer.
non-polar side-chains, which makes it difficult to assess their importance to dimerization solely from a surface area change plot. Therefore, the contribution of charge-charge interactions to dimerization was computed at neutral pH when aspartic acid and glutamate are negatively charged and lysine and arginine are positive (Fig. 2B). The value of ϳ30 kJ/mol suggests that charge interactions in the ␣-helices indeed play a role in ParG dimerization. However, this calculation of the electrostatic contribution to dimerization is likely to be an overestimate. Since dimerization is coupled to folding in the RHH superfamily (51), a calculation based on monomers with the same structure exhibited in the dimer lacks the (likely favorable) re-organization of charge-charge interactions in the unfolded monomers, in turn over-estimating the dimerization energy. Because of uncertainty in the monomer state structure, calculations were made with a simple Debye-Hückel model for charge-charge interactions, neglecting the desolvation penalties that are associated with partial charge burial at interfaces. This contribution would also lower the contribution of charge-charge interactions to dimerization. Nevertheless, some charged amino acids that interact across the dimer interface do exhibit a fall in activity (e.g. Lys-73), but not to the same degree as some amino acids bearing non-polar side-chains. This general picture of non-polar groups contributing the bulk of binding energy to protein-protein interactions is well known (52). Furthermore, it has become clear that a relatively small number of mostly non-polar groups, termed hot spots, typically are dominant (53,54), and that it is possible to predict such residues (55,56). In the current work, a relatively simple analysis of changes in solvent accessible surface area (for non-polar groups) is indicative of binding hot spots, even in a system where monomer folding is coupled to dimerization, and thus the relevant surface areas for the monomer form are unclear. It appears sufficient for predicting their importance to dimerization that the key residues make extensive non-polar interactions in the dimer, and will contribute substantially to the experimental K d (ϳ1.0 M) previously determined for the protein (48). With respect to charge-charge interactions at the dimer interface, while they do not figure in binding hot spots, it is possible that modulation of charge patterns could contribute to specificity in homo-and hetero-interactions in the RHH family, as seen for the Arc repressor (57).
Amino acids 43-to-56 and 60-to-74 comprise ␣1 and ␣2, respectively, in ParG (34; Fig. 1B). These residues, the intervening three residue loop, and the two C-terminal amino acids were subjected to alanine-scanning mutagenesis (58) to assess further their contribution to ParG function and organization of the hydrophobic core. Alanine residues at positions 51, 52, and 54 were converted to glycine. Mutations were constructed in plasmid pFH547 that contains the complete parFGH cassette in a partition probe vector (32) to permit assessment of the effect of the alterations on plasmid segregation in E. coli. Repeated attempts to make the K50A mutation in this context using different mutagenesis strategies were unsuccessful, although this alteration was introduced straightforwardly into parG cloned in the absence of parF and parH. ParG-K50A may cause cellular toxicity when produced in conjunction with ParF and parH.
Partition tests revealed that certain alanine substitutions in ParG, including of Thr-47, Ala-52, and Asn-70, exerted partial defects in segregation activity (Fig. 1C). However, the most deleterious effects on segregation were induced by the F49A mutation in ␣1 and the W71A and L72A mutations in ␣2 which reduced plasmid retention to the level of the empty vector (Fig.  1C). Thus, residues Phe-49, Trp-71, and Leu-72 are crucial for ParG function in segregation. Interestingly, the outcomes of the mutational studies are congruent with the MHP analysis and surface area change calculations outlined above that implicated residues Phe-49, Trp-71, and Leu-72, among others, in strong contacts that maintain the ParG hydrophobic core. Phe-49 shows less change in solvent accessible surface area upon dimerization than Trp-71 and Leu-72 (Fig. 2) and is integral to the monomer core, whereas Trp-71 and Leu-72 instead are crucial for the dimer core.
The Contribution of Phe-49, Trp-71, and Leu-72 to ParG Dimerization in Vivo-ParG dimerization can be monitored in two-hybrid assays based on reconstitution of adenylate cyclase activity in E. coli (48,59). The interaction efficiency in this assay can be assessed semi-quantitatively by ␤-galactosidase production (47). Fusion of wild-type ParG to the T18 and T25 subunits of adenylate cyclase generated ϳ4000 ␤-galactosidase units indicative of ParG dimerization. Introduction of the F49A, W71A, or L72A mutations into the fusions reduced ␤-galactosidase values ϳ2-fold indicating that the mutations perturbed ParG self-association (Fig. 3A, filled bars).
The N-terminal mobile tails in ParG are crucial for partition function, for stimulation of ATPase activity in the ParF partner protein, for modulating ParF polymerization dynamics, and for correct binding both to the parH centromere and during transcriptional autoregulation of the parF-parG cassette (13,33,59). Residues 20 -30 of the N-terminal tail of ParG have a propensity to form a ␤-strand structure that interacts transiently with the ␤-sheet DNA recognition element of the folded domain to give an extended ␤-sheet, providing two additional short-lived antiparallel ␤-strands per ParG dimer (59). Although the flexible tail is not necessary for ParG dimerization (59), the ⌬30ParG derivative that lacks 30 amino acids from the N terminus (Fig. 1B) showed a modest reduction in ␤-galactosidase values in the two-hybrid assay compared with full-length ParG (Fig. 3A, open bars). This may reflect that the RHH fold is destabilized in the absence of the N-terminal tail that includes the transient ␤-strand element. Moreover, the F49A, W71A, and L72A mutations in the ⌬30ParG context reduced ␤-galactosidase activity almost to basal levels confirming the importance of these residues for ParG dimerization (Fig. 3A, open  bars).
The contribution of Phe-49, Trp-71, and Leu-72 to ParG Dimerization in Vitro-The self-association properties of purified ParG-F49A, -W71A, and -L72A proteins were characterized and compared with those of wild-type ParG. First, dimers of wild-type ParG in solution are captured efficiently by the cross-linking reagent DMP and subsequently migrate as intense ϳ20 kDa species in denaturing gels (48). The formation of detectable cross-linked species was abolished entirely by the W71A and L72A mutations, whereas very weak dimerization was evident with ParG-F49A (Fig.  4A). Second, size exclusion chromatography coupled to multi-an-gle light scattering (SEC-MALLS) was performed for wild-type and mutant ParG proteins to determine their oligomeric state (Fig.  4B). Wild-type ParG has an average molecular mass of ϳ21 kDa with the mass across the peak seemingly homogeneous which is indicative of a single species, in this case a dimer (9,933 Da monomer). The mutant proteins eluted significantly later, with the exception of ParG-F49A. The most destabilized mutants, ParG-W71A and ParG-L72A, exhibited both a later elution and a lower average mass across the peak representative of their destabilizing effect.
Third, sedimentation equilibrium analysis of wild-type and mutant ParG proteins was performed. Experiments were accomplished at three rotor speeds (20,000, 30,000, and 45,000 rpm) scanning at 280 nm after 14 h. Three different protein concentrations were tested in each case, typically in the range of 10, 25, and 100 M. Absorbance values were plotted as functions of radial position. Representative sedimentation profiles of wild-type ParG and mutant proteins are shown in Fig. 4C. ParG had an average molecular mass of 20,910 Ϯ 200 Da, which agrees well with the expected mass (19,866 Da). ParG-F49A, -W71A, and -L72A had average molecular masses of 19,500 Ϯ 1,600, 16,370 Ϯ 2,600, and 16,200 Ϯ 700 Da, respectively, that display a progressive tendency toward monomer values. Previous sedimentation equilibrium analysis of wild-type ParG showed best fitting to a monomer-dimer model with a K d of ϳ1.0 M (48). This agrees closely with the current analysis that indicates a K d of 0.52 M for ParG. In contrast, ParG-F49A, -W71A, and -L72A had K d values of 15, 39, and 140 M, respectively, confirming that the mutations markedly attenuated dimerization. In summary, two-hybrid, chemical cross-linking, SEC-MALLS and sedimentation equilibrium studies unequivocally demonstrated that residues Phe-49, Trp-71, and Leu-72 play crucial roles in assembly of the hydrophobic core that mediates ParG dimerization.
The parH centromere (ϳ100-bp) consists of an array of degenerate tetramer boxes interspersed by AT-rich spacers. The ParG protein at high concentrations assembles the site into a single, major complex in gel retardation assays (33). The effects of the F49A, W71A, and L72A mutations on centromere binding by ParG were examined. Purified ParG-F49A and ParG-W71A did not bind detectably to parH and ParG-L72A bound the site only weakly at the highest protein concentrations tested (Fig. 3B). Thus, mutations that disrupted critical hydrophobic contacts in ␣1 and ␣2 impaired the ParG-centromere interaction. These observations correlate with previous studies, which demonstrated that the integrity of the ␣-helical backbone is fundamental for DNA binding by RHH proteins (37,60).

Restoration of ParG Dimerization and Function-Numerous
ParG homologues from diverse bacterial species are identifiable in protein databases. Alignment of 282 of these proteins revealed considerable diversity in the amino acids at positions 49, 71, and 72 that were shown above to be crucial for maintenance of the ParG hydrophobic core. Nevertheless, positions 49 and 72 favor hydrophobic residues. In particular, leucine and phenylalanine occur frequently at both positions 49 (36.2 and 27.3%, respectively) and 72 (45.4 and 20.6%, respectively). Hydrophobic amino acids also dominate at position 71 (31.2% tyrosine, 11.3% tryptophan, 8.9% leucine, 6.4% alanine, and 3.5% phenylalanine) with the exception of the polar, negatively charged glutamic acid, which is also present at high frequency (22.7%) (Fig. 5A). Although hydrophobic amino acids commonly occupy position 71 in ParG homologues, one of the most prevalent residues, alanine, was shown above to be unable to substitute functionally for Trp-71 in the prototypical ParG under study here. The capacity of the other most abundant hydrophobic residues at position 71, leucine, phenylalanine, and tyrosine, to functionally replace tryptophan at this position also was tested. Plasmids producing ParG-W71F, -W71L, or -W71Y were examined both in segregation and two-hybrid assays. The W71L change reduced plasmid retention to the level of the vector lacking the parFGH cassette (Fig. 5B), and also exerted a major impact on ParG self-association in vivo (Fig. 5C). In contrast, plasmids encoding ParG derivatives in which residue Trp-71 was replaced by other hydrophobic aro- matic residues, i.e. tyrosine or phenylalanine, were partitioned as efficiently as the plasmid producing the wild-type protein (Fig. 5B). Values for homodimerization of either ParG-W71F or ParG-W71Y assessed in two-hybrid assays were intermediate between those for self-association of wild-type ParG and ParG-W71L (Fig. 5C). In summary, hydrophobic aromatic residues (tyrosine and phenylalanine), but not other hydrophobic amino acids (alanine and leucine), can functionally replace Trp-71 for ParG-mediated plasmid segregation.
Intramolecular Suppression of the ParG-W71L Mutation-The Trp-71 side-chain in ␣2 of one monomer in ParG approaches within 3.47 Å of the Ala-52 side-chain in ␣1 of the second monomer (Fig. 6A). Twenty-six of 32 homologues that possess Trp-71 (Fig. 5A) contain Ala-52, including the protoypical ParG under study here. Interestingly, the A52G change in ParG reduced segregation activity (Fig. 1C) further emphasizing that the Trp-71-Ala-52 interaction is vital for maintenance of the protein hydrophobic core. Although the W71L alteration abolished ParG-mediated partition function (Fig.  5B), Leu-71 is not uncommon among ParG homologues (Fig.  5A). Intriguingly, 22 of the 25 homologues with Leu-71 possess tyrosine at position 52. Thus, the Ala-52-Trp-71 and Tyr-52-Leu-71 combinations found in numerous ParG homologues comprise a hydrophobic aromatic residue paired with a non- To ascertain whether Tyr-52-Leu-71 could substitute functionally for Ala-52-Trp-71 and thereby suppress the W71L change, the A52Y-W71L double mutation was introduced into the canonical ParG under study here. ParG-A52Y-W71L selfassociated strongly in two-hybrid (Fig. 5C) and cross-linking assays (Fig. 6B). Moreover, the mutant protein eluted with an average molecular weight of 20,430 Ϯ 800 Da in SEC-MALLS analysis (Fig. 4B), which is very similar to the value described above for wild-type dimeric ParG (20,910 Ϯ 200 Da) indicating no major shift in molar mass for the double mutant protein. The K d for ParG-A52Y-W71L determined from sedimentation equilibrium studies (Fig. 4C) was 10 M compared with 0.52 M for the wild-type protein. Notwithstanding the higher K d value, ParG-A52Y-W71L supported plasmid partition as effectively as wild-type ParG (Fig. 5B). Furthermore, whereas ParG-W71L displayed impaired binding to the parH centromere, ParG-A52Y-W71L showed a similar gel retardation pattern as the wild-type protein (Fig. 6C). In summary, the A52Y mutation suppresses the dimerization, plasmid segregation and centromere binding defects elicited by the W71L mutation in ParG. Thus, the Ala-52-Trp-71 and Tyr-52-Leu-71 combinations that comprise a hydrophobic aromatic residue paired with a non-polar residue in opposing positions are functionally interchangeable in ParG suggesting that the hydrophobic core that forms the monomer-monomer interface in the protein is highly malleable.

DISCUSSION
A vital early step in the stable segregation of multiresistance, virulence, and other low copy number plasmids involves centromere recognition by a sequence-specific DNA-binding protein. Centromere-binding factors that possess either helixturn-helix or RHH motifs have been defined (4). ParG is among the most well studied of the latter type and is composed of a C-terminal RHH-folded domain attached to a pair of flexible N-terminal tails (34). The parH centromere that is recognized by ParG comprises a set of degenerate tetramer motifs separated by AT-rich spacers (33). ParG also is a transcriptional repressor of the parF-parG cassette which is achieved by binding to a second array of tetramer boxes upstream of the genes (38). Like other RHH proteins, ParG recognizes DNA via the antiparallel ␤-strands that comprise the ribbon element. This element can be replaced by ribbons from other RHH factors producing hybrid proteins that possess the DNA binding characteristics associated with the swapped ribbon (40). Moreover, ParG recruits the ParF protein to the segrosome and additionally modulates its nucleotide hydrolysis and polymerization properties (9,13,14,25,26,48,59).
In view of the multifunctionality of ParG and its vital role in the stable maintenance of a multiresistance plasmid, the residues within the ␣-helix backbone that are key for segregation activity of ParG were defined here. Mutations in the three-residue loop that connects ␣1 with ␣2 and in the two C-terminal amino acids also were tested. Among these thirty-three residues, only Phe-49 in ␣1 and Trp-71 and Leu-72 in ␣2 were functionally intolerant of alanine substitutions and thus crucial for full segregation activity. The K d values associated with these mutations were 30 -280-fold higher than the value for the wildtype protein indicating that mutation of Phe-49, Trp-71, or Leu-72 disrupts folding and/or stability of ParG. Phe-49, Trp-71, and Leu-72 cluster together: the side-chain of Leu-72 is within 3.94 Å of the side-chain of Trp-71, which in turn is 3.81 Å from Phe-49 (Fig. 1A). This triad may act as a clamp that maintains the integrity of the RHH hydrophobic core. Interestingly, ParG derivatives in which residue Trp-71 was replaced by other hydrophobic aromatic residues were fully active in plasmid segregation. These substitutions are not expected to perturb the interactions that maintain the conformation of the ParG core (Fig. 6D).  A and B) each of which contributes a ␤-strand and two ␣-helices (Fig. 1A). The C-terminal end ␣1 of chain A is in blue, and the central portion of the two ␣-helices of chain B are in yellow. The side-chains of residues Ala-52, Gln-66, and Trp-71 are drawn as sticks. The interaction shown between Trp-71 of chain A and Ala-52 of chain B is necessary for packing of the ParG hydrophobic core. The side-chain of Trp-71 projects from ␣2 of chain A toward the center of the dimer. The side-chain of Ala-52 that projects from ␣1 of chain B is also oriented toward the dimer center, whereas the side-chain of Gln-66 from ␣2 of the same chain is close to Trp-71 but not toward the dimer center. Elements in each residue are colored in green (carbon), blue (nitrogen), gray (hydrogen), and red (oxygen). B, cross-linking of ParG-W71L and ParG-A52Y-W71L was performed without (Ϫ) and with (ϩ) DMP (10 mM) for 60 min at 37°C. Reactions were analyzed by SDS-PAGE as outlined elsewhere (48). The Arc protein of Salmonella bacteriophage P22 is a transcriptional autorepressor and also regulates expression of the gene for the antirepressor protein, Ant. The folding kinetics and hydrophobic core organization of Arc are the best characterized among RHH proteins (61). The root mean square deviation of C ␣ atom positions for the folded domains between ParG and the Arc repressor is 1.67 Å indicating that the structures are closely related (34). Comprehensive alanine scanning mutagenesis of Arc revealed that approximately half of the mutations exerted significant effects on the protein's equilibrium stability, unfolding, or refolding (60,62,63). Phe-49, Trp-71, and Leu-72 that were shown here to be crucial for plasmid segregation and dimerization by ParG correspond to positions Val-22, Ser-44, and Phe-45, respectively, in Arc (34). Mutations of Val-22 and Phe-45 are among the most destabilizing of all the alanine changes in Arc. The side-chains of these two residues, along with those of Glu-36, Ile-37, and Val-41, pack together tightly into a cluster from which the Arc protein is assembled (62). Accordingly, the E36A, I37A, and V41A mutations also are highly destabilizing (61). Val-63, Val-64, and Val-68 are the equivalent residues in ParG. MHP analysis previously suggested that the side-chains of Val-64 and Val-68 form some of the strongest hydrophobic contacts within the ␣-helical region that hold the ParG dimer together (34). However, only the V63A alteration among these three residues exerted any detectable effect on ParG-mediated plasmid segregation (Fig. 1C). This observation does not preclude that alanine mutation of residues Val-63, Val-64, and Val-68 causes destabilizing, but tolerable, effects on ParG function.
The protein is a RHH factor involved in segregation of plasmid pSM19035 (64,65) and thus, like Arc, is a structural homologue of ParG (34). Double alanine mutations of the residues equivalent to Arg-48 and Phe-49 or Val-68 and Leu-72 in ParG reduced, but did not abolish, heterodimerization of with the wild-type protein in two-hybrid assays in E. coli (66). Although double mutations were analyzed, the data suggest that residues equivalent to Phe-49 and Leu-72 in ParG also are important for assembly of the hydrophobic core and dimerization of (66).
Residue Trp-71 in ParG was functionally replaceable by other hydrophobic aromatic residues, i.e. phenylalanine or tyrosine, but not by non-aromatic hydrophobic amino acids, i.e. alanine or leucine. ParG-W71L was severely impaired in dimerization, segregation activity, and centromere binding. However, these defects associated with the W71L change were suppressed fully by an alanine-to-tyrosine change at position 52. The A52Y-W71L compensatory change was engineered based, first, on the observation that Tyr-52-Leu-71 pairs occur in numerous ParG homologues and, second, that the side-chain at position 52 in ␣1 of one ParG monomer closely approaches that of residue 71 in ␣2 of the second monomer. The K d values suggest a 4-fold increase in tightness of the ParG-A52Y-W71L dimer compared with ParG-W71L. Although the K d of the ParG-A52Y-W71L variant was noticeably higher than that of the wild-type protein, the double mutation is expected to maintain the organization of the ParG hydrophobic core (Fig. 6D). Thus, the ParG core may be able to accommodate a variety of appropriate hydrophobic aromatic residues paired with non-polar residues at positions 52 and 71 (Fig. 5A). Interestingly, numerous double and multiple mutations in Arc also retained partial or full stability and/or biological functionality (67)(68)(69). Moreover, the ribbon element that dictates centromere recognition by ParG can be functionally replaced by ribbons from other RHH proteins (40). These observations highlight that RHH proteins may be tractable scaffolds for the rational design of DNA-binding factors with altered specificity and dimerization properties. Finally, improved understanding of the folding and assembly of ParG and other proteins that are necessary for the segregation of multiresistance plasmids will provide insights into whether the partition apparatus is a viable target for novel anti-plasmid compounds.