Order and Disorder in the Domain Organization of the Plasmid Partition Protein KorB

The plasmid partition protein KorB has a dual role: it is essential for the correct segregation of the low copy number broad host range RK2 plasmid while also being an important regulator of transcription. KorB belongs to the ParB family of proteins, and partitioning in RK2 has been studied as a simplified model of bacterial chromosome segregation. Structural information on full-length ParB proteins is limited, mainly due to the inability to grow crystals suitable for diffraction studies. We show, using CD and NMR, that KorB has regions of significant intrinsic disorder and hence it adopts a multiplicity of conformations in solution. The biophysical data are consistent with bioinformatic predictions based on the amino acid sequence that the N-terminal region and also the region between the central DNA-binding domain and the C-terminal dimerization domain are intrinsically disordered. We have used small angle x-ray scattering data to determine the ensemble of solution conformations for KorB and selected deletion mutants, based on models of the known domain structures. This conformational range of KorB is likely to be biologically required for DNA partitioning and for binding to a diverse set of partner proteins.

The partitioning of DNA between daughter cells upon cell division is critical for the survival of all organisms. Partitioning in most bacteria is dependent on two proteins from the ParA/ParB families (1). ParB binds to a specific centromerelike sequence (parS) on the chromosome forming a higher order nucleoprotein complex that is thought to pair the sister chromosomes/plasmids. ParA, an ATPase, binds to ParB and is thought to act as a motor, pulling or pushing two ParB-bound chromosomes apart to different poles of the bacterial cell (2)(3)(4).
One of the best characterized ParB proteins is KorB, encoded by the low copy number broad host range plasmid RK2, from the IncP-1 family (2)(3)(4)(5)(6)(7)(8)(9)(10). RK2 has been studied for several years as a small model genome and is also important as a carrier of multiple antibiotic resistance genes across a range of bacteria (11)(12)(13)(14)(15). KorB acts in conjunction with IncC (7,16), a protein of the ParA family also encoded on the plasmid, and is more similar to chromosomal ParB homologues than those from P1 or F plasmids. This makes KorB a good candidate for structural studies to understand partitioning of bacterial chromosomes. In addition to its role in partitioning, KorB acts as a global repressor of transcription for at least six operons in RK2, where it also interacts with a number of other proteins (7). KorB binds simultaneously with RNA polymerase at promoters and interacts with it to prevent open complex formation (17). KorB can act upstream or downstream of the promoters and also at a distance (2)(3)(4)9). At all promoters, KorB acts cooperatively with a second repressor, either KorA, TrbA, or KorC, forming a regulatory network that coordinates expression of the operons on the plasmid (3,6,7,16). Therefore, KorB is of interest for understanding not only partitioning, but also for understanding cooperative protein-protein and protein-DNA interactions involved in gene regulation. KorB consists of three distinct segments (7) (Fig. 1) of which only two have been crystallized and their structures determined to atomic resolution. Firstly, the C-terminal 64 amino acid residue dimerization domain (residues 294 -358) (18), which is structurally similar to SH3 domains, and consists of ␤-strands. Secondly, the central domain (residues 137-252), which interacts with operator DNA (19) and is entirely ␣-helical. The structure of the N-terminal 137 amino acids of KorB is unknown, however the corresponding region of a homologous protein, SpoOJ from Thermos thermophilus, together with its central domain, have been determined (20). It is of mixed secondary structure, and the N-terminal 20 amino acid residues of SpoOJ are not seen in this structure. KorB has an additional N-terminal 20 amino acids compared with SpoOJ; it also contains two additional predicted helices at the N terminus of the central domain, after the helix-turnhelix motif, which are not found in SpoOJ (Fig. 1B) (1). In KorB these helices contact DNA and are postulated to lead to sequence-specific binding. Thus, despite the sequence homology, the overall structure and DNA binding of SpoOJ may differ from that of KorB. To date there is no structural information about a full-length ParB protein.
Using a range of biophysical techniques we show the protein is modular in its domain organization and that the orientations of the domains of KorB within the protein are highly flexible due to zones of intrinsic disorder along the length of the protein. As the first structural description of a full-length ParB protein, we have reconstructed KorB from the known crystal structures using small angle x-ray scattering (SAXS) 5 and calculated an ensemble of conformations that the protein adopts in solution. This flexibility will be important for the protein's functional role in the cell enabling it to bind at different distances along DNA and to different binding partners (3)(4)(5)(6)9).

EXPERIMENTAL PROCEDURES
Proteins KorB-Deletion mutants were made by PCR amplification of the appropriate segment of the wild-type korB gene and inserted into the modified pET28a plasmid previously described (7). This yields proteins with an N-terminal 23amino acid His tag. Proteins were overexpressed and purified on a nickel-agarose column as described (7), followed by gel filtration on a S200 column in 10 mM Tris HCl, pH 7.0, 100 mM NaCl, 0.1 mM EDTA, 10 mg/liter phenylmethylsulfonyl fluoride. They were concentrated to 5-10 mg/ml by ultrafiltration.
CD-CD spectra were obtained on a JASCO 715 spectrometer using 2-mm path length cuvettes. The proteins (ϳ0.02 mg/ml final concentration) were diluted into 10 mM sodium phosphate buffer, pH 7.0, 20 mM NaClO 4 . This allowed the observation of shorter wavelengths, as chloride ions used previously absorb highly below 210 nm. Spectra from 180 -260 nm were scanned at 25°C at a rate of 50 nm/min with a resolution of 0.2 nm. Spectra of buffer in the same cuvettes were taken under identical conditions and subtracted from those of the protein. Thermal denaturation was monitored at 220 nm, using a temperature ramp between 10°and 80°C over ϳ1 h. Secondary structure content was extracted from CD spectra using the program CDSSTR (21,22) within DICHROWEB (23,24), using reference sets 4 and 7.
NMR Spectroscopy-One-dimensional 1 H NMR spectra were taken at 25°C on a Bruker 500-MHz NMR machine, using ϳ0.5 mM protein in 10 mM sodium phosphate buffer, pH 7.0, 100 mM NaCl, 0.1 mM EDTA. The solvent water peak was attenuated using pulsed field gradients or by presaturation. Typically 128 scans were taken, with a relaxation delay of 1.5 s. Two-dimensional 15 N-1 H heteronuclear single quantum coherence spectra of 15 N-labeled (N⌬31-C⌬221)KorB and (N⌬297)KorB were taken at 30°C on a Bruker 500-MHz NMR machine. WT KorB and (N⌬150)KorB were both 13 C/ 15 N/ 2 Dlabeled, and transverse relaxation-optimized spectroscopy 15 N-1 H heteronuclear single quantum coherence spectra were taken on a Varian 900-MHz machine.
Analytical Ultracentrifugation-All analytical ultracentrifugation data were obtained on a Beckman XL-I using absorbance optics and utilizing a protocol previously defined (25). Proteins (0.1-1 mg/ml) were in 10 mM Tris-HCl, pH 7.0, 100 mM NaCl, 0.1 mM EDTA. Sedimentation velocity experiments were carried out at 40,000 rpm using an AnTi60 rotor, at 20°C, and at three loading concentrations to check for self-association. Cells were scanned every 10 min at 280 nm. All data were analyzed using SEDFIT (26).
SEC-MALLS-Analytical fractionation was carried out using a series of SEC columns TSK G6000PW and TSK G4000PW protected by a similarly packed guard column (Tosoh Bioscience, Tokyo, Japan) with on-line MALLS (Dawn HELIOS II, Wyatt Technology, Santa Barbara, CA) and refractive index (Optilab rEX, Wyatt Technology) detectors. The eluant was Tris-HCl, pH 7.0, 100 mM NaCl, 0.1 mM EDTA, 10 mg/liter phenylmethylsulfonyl fluoride pumped at 0.8 ml min Ϫ1 (PU-1580, Jasco, Great Dunmow, UK), and the injected volume was 100 l (ϳ1.0 ϫ 10 Ϫ3 g ml Ϫ1 ) for each sample. Absolute weight average molecular weights (M w ) were calculated using ASTRA (Version 5.1.9.1) software (Wyatt Technology). Online viscometry measurements were made using the Viscostar differential pressure viscometer (Wyatt Technology), and data were transformed using the software provided.
Model of the N-terminal Region of KorB-The model of the N-terminal region 54 -154 of KorB was built using the program CPHmodels 2.0. (27) based on homology to SpoOJ (20).
SAXS-Small angle scattering data were collected at Beamline X33 at the Deutsches Elektronen-Synchrotron (Hamburg, Germany). Proteins (3-15 mg/ml) were dialyzed into 10 mM Tris-HCl, pH 7.0, 100 mM NaCl, 0.1 mM EDTA. Data were collected at an x-ray wavelength of 1.54 Å, and the SAXS cell was equilibrated to 20°C. Data were taken in 1-min frames, and typically 2 frames were taken for each cell. Low angle portions of successive frames were compared for evidence for aggregation of the sample due to photo-damage. Data were reduced to one dimension and displayed using the program PRIMUS (28). Molecular mass was determined from the forward scattering I(0) calculated from the Guinier analysis using the scattering from 4 mg/ml bovine serum albumin as a standard. Data from three concentrations of each protein were compared with examine for aggregation, and, where necessary, the data sets were merged so the low angle scattering came from the most dilute sample. SAXS data is of the form, where r is the distance between scattering centers in the molecule, P(r) is the distance distribution function, and s ϭ 2sin For potentially folded domains, data were reconstructed using the program Gasbor (30). Here, a simulated annealing method is used to minimize against X 2 discrepancy, where c is a scaling function, I DR is the intensity from a bead representing an amino acid, and is the standard deviation. The process is repeated 10 -12 times, with resulting structures averaged. Conformers for the intrinsically disordered domains were calculated using a bead for each amino acid, and the vector between the center of each bead was weighted by the known vectors for each dipeptide C ␣ -C ␣ from the Protein Data Bank (31). Pools of 10,000 conformers were generated using the program RANCH (31). For constructs containing the N-terminal domain, this was modeled either completely unfolded or with residues 51-137 modeled using SpoOJ as a template. In the latter case, the structures were aligned with the central domain across residues 137-154. For WT and (N⌬150)KorB, dimers were constructed using the known structure of the dimerization domain as a template. Random conformers were selected from the pool of 10,000 generated monomer conformers and aligned using the CCP4 module LSQKAB (32). Each constructed dimer was checked for steric clashes; dimers without clashes were then collated until 10,000 models, each with nonsymmetry-related monomers, had been constructed. The theoretical intensity of each conformer was calculated using CRYSOL (33). Selected ensembles of curves forming distributions of conformations that represent the raw scattering data were generated with the genetic algorithm GAJOE13 (31). Here, the pool of theoretical scattering curves is divided into "chromosomes": ensembles of curves that represent the scattering data. Cycles of mutation and crossing are carried out and the goodness of fit of each ensemble compared with the theoretical scattering data using the discrepancy in Equation 5.
This process was repeated five times, and distributions were compared. Fig. 1A shows the domain organization in KorB. Bioinformatic analysis of the amino acid sequence using the program DISPROT (34) predicts that KorB protein has a zone of disorder for the first 64 amino acids, as well as between the DNA-binding domain and C-terminal dimerization domain (supplemental Fig. S1). These regions contain many charged residues, with only small hydrophobic residues that cannot form a compact hydrophobic core. Such disordered regions are predicted to be common in eukaryotic genomes and occur in several proteins with key functions in cell signaling and gene regulation, but few examples have been found in bacteria (35).

KorB Exhibits Zones of Intrinsic Disorder along Its Length-
Predictions of disorder in these regions of KorB correspond well with the known structural data (Fig. 1B). For instance the known structure of the DNA-binding domain (PDB code: 1R71) was determined from a KorB derivative that also included the linker region (residues 252-294), although none of the additional residues in this region were visible in the crystal structure (19). The C-terminal dimerization domain (residues 297-358), which has also been crystallized (18) (PDB code: 1IGU) was initially obtained on attempting to crystallize the full-length protein. This suggests that the wild-type protein is readily proteolyzed at residue 296, at the end of the linker region.
To characterize the structure of the protein in solution, we have studied wild-type KorB and a series of deletion mutants of KorB, which contain different combinations of predicted ordered and disordered domains ( Light Scattering and Analytical Ultracentrifugation Show Solution Homogeneity and Shape Asymmetry-To determine sample homogeneities and solution molecular weights, the four largest proteins were analyzed by both SEC-MALLS (size exclusion chromatography coupled to multi-angle laser light scattering) and analytical ultracentrifugation. As expected, full-length KorB elutes first from the SEC column, and (C⌬105)KorB last (supplemental Fig. S2), each giving a single peak. The constructs (N⌬297)KorB and (N⌬31-C⌬221)KorB are below the size that can be accurately determined by SEC-MALLS and, as such, were not subjected to this method of analysis. Analysis of the multiangle-dependent light scattering of these peaks showed that all the proteins were monodisperse; as judged by the ratio of the Z-averaged and the number averaged molecular masses, M z /M n ( Table 1). The weight M w values were determined for each of the protein samples from solution light scattering and are also shown in Table 1. The WT KorB and (N⌬150)KorB proteins are dimeric as expected, because they contain the C-terminal dimerization domain, whereas (C⌬105)KorB and (C⌬60)KorB are monomeric, due to the deletion of this domain.
All sedimentation velocity profiles ( Fig. 2A) showed a single symmetric peak, which did not vary with loading concentration of the sample (data not shown). This is also indicative of a single monodisperse species and complements the SEC-MALLS analysis. Molecular weights determined from the ratio of sedimentation and diffusion coefficient were close to those determined from SEC-MALLS (Table 1), again indicative of monodispersity. For the smaller proteins, (N⌬297)KorB and (N⌬31-C⌬221)KorB, analytical ultracentrifugation shows the appropriate masses and the frictional ratios derived from the analysis of the sedimentation velocity data are reasonable for globular proteins (25).
In contrast, the frictional ratios for the four largest proteins range from 1.6 to 1.9 (Table 1) are larger than expected for a "typical" globular protein (1.01-1.60) (25). This indicates that the molecules are quite extended in solution. The viscosity increments derived for the four largest proteins were found to fall between values of 11.0 and 14.1 ml/g. Intrinsic viscosity has a weak dependence on molecular mass but a strong dependence on shape and flexibility (36). These values fall closer to values in the literature for flexible and unfolded proteins (typically 13-15 ml/g), than for globular proteins (typically 2.5-5.0 ml/g). This situation is consistent with previous determinations of intrinsic viscosity on unstructured proteins (36,38) and in-line with our bioinformatics predictions.
All Constructs Are Predominately Folded but Contain a Significant Unfolded Component-NMR and CD spectroscopy values were used to examine the secondary structure of the proteins. The one-dimensional NMR spectra of the proteins (supplemental Fig. S3A) showed some signatures of folded protein, with dispersed signals, including high field methyl groups (supplemental Fig. S3B) and low field amino protons. However, all the proteins also had a number of sharp, intense peaks corresponding to highly flexible segments. Fig. 2B shows CD spectra for fulllength KorB and the deletion mutants. All of the constructs showed some folded character; WT KorB, (N⌬150)KorB, and (C⌬60)KorB have similar CD spectra with a peak at 222 nm, consistent with some ␣-helical component, but the component at 210 nm is higher than that at 222 nm, concordant with a measurable component of random coil. (N⌬31-C⌬221)KorB shows a similar spectrum, but much less intense suggesting it contains less secondary structure. For (C⌬105)KorB, the negative maximum is shifted to 206 nm, concordant with the protein containing a more random coil, whereas for (N⌬297)KorB the spectrum is consistent with a ␤-sheet with a maximum at 200 nm and only a small peak at 217 nm. The thermal denaturation of the proteins was observed by monitoring the CD spectrum at 222 nm; all the constructs melted with a midpoint at ϳ40°C (see supplemental Fig. S4), although the transitions did not appear to be cooperative, suggesting that the proteins are not folded as single globular domains.  (20). Secondary structure elements are from crystal structures for the DNA-binding domain (PDB code: 17R1) and the C-terminal dimerization domain (PDB code: 1IGU) are shown above KorB. The structure of the N-terminal region of SpoOJ (PDB code: 1VZ0) is shown below its sequence. The numbering is for KorB. The helix-turn-helix DNA binding motif is highlighted in orange for both structures. MAY 14, 2010 • VOLUME 285 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 15443
Analysis of the CD spectra (23, 24) using the CDSSTR method (21,22) showed that each of the constructs contains 30 -40% random coil. WT KorB, (N⌬150)KorB, and (C⌬60)KorB contain around 35% ␣-helix with ϳ15% ␤-sheet, (C⌬105)KorB contains less ␣-helix and more ␤-sheet, while (N⌬31-C⌬221)KorB and (N⌬297)KorB contain 30 and 45% beta sheet, respectively (see Table 2). From all these spectra, it is clear that each of the proteins contains some folded regions but also a significant amount of random coil structure, in line with the prediction of segments of intrinsic disorder.
The secondary structure content determined from the CD analysis for (N⌬297)KorB agrees well with the crystal structure of this domain, whereas that of (N⌬150)KorB agrees with that predicted from the two known crystal structures of KorB with random coil at residues 251-297. The other constructs contain the N-terminal region of KorB, the structure of which is currently not available. The secondary structural elements for these KorB constructs were estimated in two ways: either solely using the two crystal structures of KorB, with residues 1-137 and 258 -297 of KorB unstructured ( Table 2,  KorB Domain Structure Exhibits Modularity-WT KorB, (N⌬150)KorB, (N⌬31-C⌬221)KorB, and (N⌬297)KorB were 15 N-labeled, and the environments of the amide NH-groups in these proteins were examined by two-dimensional 1 H-15 N heteronuclear single quantum coherence NMR. Despite its 84-kDa mass, WT KorB shows relatively sharp lines in the NMR spectrum (Fig. 3), probably due to independent motion between domains. There was also a large number of intense peaks at 8.0 -8.5 ppm, characteristic of a random coil. Less overlap is seen in the spectrum of (N⌬150)KorB, although it also has a number of sharp peaks in this region of the spectrum. The smaller domains show well resolved spectra. All of the peaks in the spectrum of (N⌬297)KorB are observed in similar positions in the spectrum of (N⌬150)KorB. Similarly, the resolved peaks observed in the spectrum of (N⌬150)KorB are also seen in identical positions in the spectrum of WT KorB, whereas some of the additional peaks in the latter are observed in the spectrum of (N⌬31-C⌬121)KorB. One example is the group of peaks at 9.0 -9.5 ppm in the 1 H dimension and 127-130 ppm in the 15    and similarly the central domain is unaffected by the N-terminal amino acids. From this we can deduce that the free protein has a modular structure, whereby there is little or no proteinprotein interaction between domains, so that the domains show independent orientations in solution.

SAXS Reveals the Extent of Multiple Conformers in Solution-
To construct a model of the whole of KorB from the high resolution domain structures available, we used SAXS. Tradition-ally, this technique provides excellent low resolution data of the overall shape of proteins into which higher resolution structures can be placed (39). SAXS also has been proposed as a technique that is capable of characterizing intrinsically disordered domain structure due to its appropriate solution-based measurements of intermediate length scales (5-100 Å) (40,41). SAXS data (Fig. 4A) were obtained at DESY in Hamburg and processed to produce Guinier plots (Fig. 4B), Kratky plots (Fig.

TABLE 2 Structure prediction from (i) CD and (ii) from crystallography and homology model of domains
The first four columns show the predictions from the CDSSTR algorithm contained within the DICHROWEB server. The differences in prediction reflect the use of different reference sets.

Construct
Predicted % ␣-helical  4C), and distributions of distances (Fig. 4D). Guinier plots of the proteins were linear in Q 2 in the low angle region, showing that the proteins are monodisperse (Fig. 4B), in line with the measurements by light scattering and sedimentation velocity. Analysis of the Guinier region (R g Q Ͻ 1.3) yielded the radii of gyration shown in Table 3, consistent with that found by second moment analysis of the distance distribution functions (Fig.  4D). The calculated molecular masses from the Guinier plots are similar to those from sedimentation analysis, again demonstrating the monodispersity of the samples. These are also consistent with the excluded (Porod) volumes from the distance distribution function (Table 3). WT KorB, (N⌬150)KorB, (C⌬60)KorB, and (C⌬105)KorB all exhibited a definite peak in the Kratky plots, a signature of globular structure, although the data did not go to zero at high Q value. This is indicative of some degree of random coil component (40), again in line with the predictions of intrinsic disorder and the CD (Fig. 2B) and NMR (Fig. 3) measurements. We initially derived a low resolution model of the overall shape of the dimeric C-terminal domain, (N⌬297)KorB, from the SAXS data using Gasbor (30) and showed that it was consistent in shape with the crystal structure (PDB code: 1IGU). This shows that the solution structure is similar to that in the crystal structure (see supplemental Fig. S5A). The model of the N-terminal of KorB fits inside the envelope derived from the scattering data (see supplemental Fig. S5B). However, the previous CD data ( Table 2) suggest that the long ␣-helix is largely absent, casting doubt as to whether this region is fully folded. Therefore we constructed models of WT KorB, (C⌬60)KorB, and (C⌬105)KorB with region 31-136 folded and unfolded, so as to compare both of these models with our scattering data. The WT KorB, (N⌬150)KorB, (C⌬60)KorB, and (C⌬105)KorB constructs containing regions of intrinsic disorder do not adopt a single conformation in solution and, as such, the above method of structure determination from SAXS is not appropriate. We therefore adopted the ensemble optimization method of Svergun and co-workers (31). Firstly, two sets of 10,000 conformers, based on the known domain structures with flexible amino acid chains between the domains, were generated (see "Experimental Procedures"). In the first set the N-terminal domain up to residue 137 is modeled as a flexible chain, whereas in the second residues 54 -137 are assumed to have the structure based on the SpoOJ homology model. From each pool of structures, we selected distributions of 15-20 conformers that together fit the raw x-ray scattering data, using the genetic algorithm optimization method contained in the ensemble optimization method software (31) (supplemental Fig. S6 and Table 3). Fig. 5 shows the distributions of the radii of gyration (left-hand column) and maximum distance (D max ) of each construct (right-hand column) for the con- Both data sets for the (N⌬31-C⌬221)KorB and (N⌬297)KorB constructs have been omitted for clarity, because both of these data sets were considerably more noisy in the high angle region, thus obscuring the other data sets in the plot. As can clearly be seen, all constructs exhibit a characteristic Gaussian shape due to folded domains of KorB, but also a significant rise at higher angle regions indicating a random coil component. D, the distance distribution functions calculated using GNOM for each of the KorB constructs.  5A) and with residues 54 -137 folded (Fig. 5B). In black are the data for the pool of all calculated 10,000 random conformers and in red are the data for the distributions selected by the ensemble optimization method algorithm for each of the constructs (supplemental Fig. S7 shows the 20 selected conformers selected).
Although there is conformational heterogeneity, similar structural themes emerge. For the wild-type protein, a dimer has two flexible regions per monomer; there appears to be virtually no selection of preferred conformers from the generated random pool with the N-terminal domain fully unfolded (the red and black distributions overlay). With calculations based on a partially folded N-terminal domain, the peak for the selected distributions (red) is shifted toward the longer conformers in the pool of random conformers (black).
To assess this further, we took 9,920 conformers of the WT KorB from the initial pool of 10,000 conformers with residues 54 -137 folded and only 80 conformers from the unfolded pool and repeated the selection procedure. Despite the unselected pool containing only 0.8% unfolded conformers, the selected pool contained 14.5% of these conformers, indicating that the unfolded proteins describe the conformation in solution better than the folded N termini. This result is in concert with the CD data shown previously where there is a loss of expected second-ary structure with respect to that predicted from the homology model for the N-terminal domain. The NMR data for the N-terminal domain does show that there is some structure present (Fig. 3), however, the peaks between 8.0 and 8.5 ppm could arise from amide protons in either unstructured or ␣-helical regions. This domain may exists in an equilibrium between a folded and unfolded state, and the folded and unfolded conformations of this domain we have generated represent limiting cases.
The (N⌬150)KorB mutant, a dimer with one flexible region in each monomer but lacking the N-terminal domain, exhibits a small bias of conformers toward compact structures, as judged by both the R g and D max distributions. The deletion of the N-terminal region is enough to reduce the available number of conformers compared with the wild-type protein from a random to a slightly more defined set. The linker region is slightly more compact than the random coil calculations. The (C⌬60)KorB mutant, a monomer with two flexible regions, shows a similar biasing of selected conformers toward compact structures; irrespective of whether the N-terminal domain is modeled as folded or unfolded; compared with the wild-type protein. Finally, for (C⌬105)KorB mutant, a monomer with only one intrinsically disordered domain, there is a strong bias in the selection toward conformers with smaller R g values and D max than the random pool, indicating that the protein has a more compact structure than a random selection of conform- ers, However, in both of these constructs the distributions calculated for the unstructured N-terminal region fit better than for the partially folded N-terminal region of KorB (Table 3,  column 9).

DISCUSSION
The solution data obtained on KorB are consistent with sequence predictions of zones of order and intrinsic disorder along the length of protein with disordered regions at the N terminus of the protein and between the central DNA-binding domain and the C-terminal dimerization domain. The CD and NMR data are all consistent with the proteins containing both folded and disordered domains that do not interact. The sedimentation and SEC-MALLS data show that the proteins are monodisperse and extended in solution (as judged by the measured f/f o ), with large Stokes radii. This result explains the previous difference between the stoichiometry of KorB estimated from gel filtration (17) and from analytical ultracentrifugation ( Fig. 2A): an extended conformation will sweep out a large hydrodynamic volume and as such will elute earlier on the column and be assigned an anomalously larger molecular weight. The sedimentation velocity profiles of all of the proteins were symmetrical and did not show either bimodality or anomalous frictional ratios, both previously postulated to be signatures of intrinsic disorder (40,43). This indicates that analysis of the sedimentation coefficient may not be a sensitive enough probe of solution conformational heterogeneity for intrinsically disordered systems; viscosity increment may be a more sensitive probe (36).
For WT KorB and the deletion mutants containing more than one domain, the regions of intrinsic disorder mean that a structural model at atomic resolution is presently beyond our technical grasp (44). However, SAXS, allied with other hydrodynamic techniques, is well placed for answering questions about the gross shape and extent of conformational flexibility adopted by the protein (40). As such, we have shown that the flexible linker region in (N⌬150)KorB adopts a slightly more compact conformations rather than the full range of random conformers available. In contrast the N-terminal region is quite compact, as seen with the (C⌬105)KorB construct. It is, however, somewhat difficult to judge the degree of folding of the N-terminal domain, and the folded and unfolded conformers we have generated most probably represent limiting cases. Overall, it is intuitive that the greater the number of intrinsically disordered domains, the greater the number of conformers the protein adopts.
Interestingly, the ordered and previously crystallized portions of KorB are associated with a single function (i.e. DNA binding for the central domain and dimerization for the C-terminal domain). In contrast, the disordered regions are implicated in many functions of KorB. The N-terminal region of KorB modulates DNA-binding strength and site selectivity, repression at a distance, as well as the localization of KorB in the cell. The region between the C-terminal domain and the central domain (residues 252-294) is implicated in several functions of KorB, including repression of transcription and cooperativity with other proteins. Such multifunctionality is a hallmark of intrinsically disordered domains (45).
We propose that the intrinsically disordered domain provides KorB a wide set of possible conformers available as templates for binding to a diverse set of proteins such as KorA, IncC, and RNA polymerase. Such binding events will have an effect on the ensemble of conformations KorB can adopt in solution and thus mediate the set of conformers available for interaction to a second binding partner, e.g. DNA or a third protein. The flexibility allows binding to occur at different distances between KorB and a second repressor while the conformational changes give rise to the cooperative effects.
The disordered regions are also likely to be important for DNA partitioning, as many of the partition proteins need to span two DNA duplexes (or several proteins) to carry out their function. Flexible interactions between the central and C-terminal domains of a ParB protein have been seen in the crystal structures of central and C-terminal domains of ParB from P1 plasmid of Escherichia coli bound to different target sequences (37,42). In ParB, there is a 4-amino acid linker between the central and C-terminal domains, and the domains are entirely distinct and rotate 60 -160 degrees relative to each other in the asymmetric ParB dimer, allowing it to contact different DNA duplexes. Although KorB and ParB from P1 show only limited structural homology and appear to bind DNA differently, flexibility arising from intrinsic disorder appears to be an essential requirement for partitioning.
The multifunctional role of KorB, involving multiple interacting partners as well as roles in both gene regulation, and partition, makes it an important network hub in the plasmid maintenance and partition process, and an excellent model for understanding other proteins with such domains.