Right-handed DNA Supercoiling by an Octameric Form of Histone-like Protein HU

In bacteria, the contribution of global nucleoid organization in determining cellular transcription programs is unclear. Using a mutant form of the most abundant nucleoid-associated protein HU, HUαE38K,V42L, we previously showed that nucleoid remodeling by the mutant protein re-organizes the global transcription pattern. Here, we demonstrate that, unlike the dimeric wild-type HU, HUαE38K,V42L is an octamer and wraps DNA around its surface. The formation of wrapped nucleoprotein complexes by HUαE38K,V42L leads to a high degree of DNA condensation. The DNA wrapping is right-handed, which restrains positive supercoils. In vivo, HUαE38K,V42L shows altered association and distribution patterns with the genetic loci whose transcription are differentially affected in the mutant strain.

tions in DNA transactions like replication, transcription, recombination, and DNA repair (6 -10). Earlier observations of compact nucleosome-like structures in DNA induced by HU binding (11,12) were recently challenged by studies that suggested that HU might in fact counteract DNA compaction by antagonizing other DNA-condensing proteins like H-NS (13). There has also been conflicting reports about the exact contribution of HU toward global chromosomal superhelicity (14). Thus, despite HU being one of the most abundant nucleoidassociated proteins, the exact molecular mechanisms of HU action with regard to its role in chromosome architecture and function remain largely obscure and controversial.
We have recently identified a gain-of-function mutant form of HU␣, HU␣ E38K,V42L , which caused a reconfiguration of the E. coli nucleoid from a loosely packed, dispersed structure into a densely condensed, globular conformation (15). The mutant HU␣ contains the amino acid changes E38K and V42L. The nucleoid compaction was accompanied by dramatic changes in the morphology, growth, and physiology of the hupA mutant; transcription of many inducible as well as constitutive genes expressed under laboratory conditions was silenced, whereas many cryptic virulence genes were activated. Thus, HU␣ E38K,V42L represented the de novo generation of a unique form of HU with characteristics and consequences distinctively diverse from that of the wild-type protein and offered a unique opportunity to explore the nexus between bacterial nucleoid organization and global gene expression. In this study, we analyzed the biochemical properties of HU␣ E38K,V42L to elucidate the molecular mechanism of its action with respect to DNA condensation and transcription control. The results show that the wild-type and mutant HU have contrasting functional and biochemical behaviors that are translated to their differential effects on global transcription pattern.

EXPERIMENTAL PROCEDURES
Purification of Wild-type and Mutant HU␣-The wild-type and mutant hupA genes were cloned in expression vector pET15b (Novagen) and transformed into BL21(DE3)/pLysS strains. The proteins were purified essentially as described previously (16). After nickel-nitrilotriacetic acid-agarose column purification, fractions containing pure HU␣ proteins were pooled and dialyzed against 2 mM HEPES (pH 7.9), 50 mM NaCl.
Electrophoretic Mobility Shift Assays-Mixtures of pUC19 DNA (6 nM) and increasing amounts (60, 120, 300, 600, and 1200 nM) of wild-type and mutant HU␣ proteins were incu-bated for 25 min at 25°C in 10 mM Tris (pH 8.0), 75 mM KCl. The products were separated by electrophoresis at 0.7 V/cm through an 0.8% agarose gel run in 0.5ϫ TAE buffer and visualized by ethidium bromide staining. For proteinase K, EcoRI and DNase I digestions, the protein:DNA ratio used was 100:1.
Atomic Force Microscopy-Solutions of supercoiled pUC19 DNA with and without wild-type HU␣ and HU␣ E38K,V42L were diluted with dilution buffer (10 mM Tris (pH 8.0), 75 mM KCl). Solutions of M13mp18 RF1 DNA (linearized with EcoRI) with and without wild-type HU␣ and HU␣ E38K,V42L were diluted with ligation buffer (10 mM Tris-HCl, pH 8.0, 50 mM KCl). Five l of fresh solutions were adsorbed onto 11-mm diameter freshly cleaved mica disks (Ted Pella, Inc.) for 5-10 min. Each sample was washed with 200 l of ultrapure water and dried with argon gas. In the case of the linear 444-bp DNA, all samples were prepared in 10 mM Tris (pH 8.0) and 50 mM KCl before being deposited on aminopropyl-silatrane-treated mica. The AP mica obviates the need of Mg 2ϩ for DNA adsorption. Both PicoForce Multimode Atomic Force Microscopy (AFM) 2 with Nanoscope IV controller and type E scanner head and the Bioscope with Nanoscope IIIa controller (Veeco/Digital Instruments) were used for imaging. All imaging was performed at room temperature and in air with tapping mode AFM. Oxidesharpened silicon microcantilevers (Olympus America, Inc.) with nominal spring constants of 42 newton/m were used.
Analytical Ultracentrifugation-Equilibrium analytical ultracentrifugation was performed in a Beckman XL-A analytical ultracentrifuge. Self-associations of the wild-type HU␣ and HU␣ E38K,V42L were studied using 4-hole rotors at rotor speed of 25,000 rpm. All measurements were performed at 20°C when binding equilibrium was reached. The buffers used for all samples were 50 mM KCl, 10 mM Tris (pH 8.0). Two sets of experiments were performed, with the wild-type HU␣ in the first and the mutant HU␣ E38K,V42L homodimer in the second set. For the wild-type HU␣, two cells were loaded with 180 l of 9 and 18 M solutions of the wild-type protein. In the second experiment, two cells were loaded with 180 l of 8 and 15 M solutions of the HU␣ E38K,V42L protein. The second channel of each cell was filled with 185 l of the buffer. Transmitted light intensities were collected at 230 nm because HU does not contain tryptophan, tyrosine, or cysteine residues. The mathematical model used to fit the data were,  37°C for 30 min with wild-type HU␣ and HU␣ E38K,V42L at  increasing concentrations of 3.125, 6.25, 25, 75, 100, 125 M, and 150 M in a 10-l reaction mixture containing 35 mM Tris-HCl (pH 8.0), 20 mM NaCl, and 5 mM dithiothreitol. Then calf thymus topoisomerase I (Invitrogen) was added and incubated at 37°C for 2 h. Proteinase K (10 g) was added and incubation continued for another 30 min. The DNA samples were analyzed by two-dimensional gel electrophoresis. The first dimension electrophoresis was run in 0.5ϫ TBE at a constant voltage of 69 V for 16 h. The second dimension electrophoresis was run in 0.5ϫ TBE containing 1 g/ml chloroquine for 4 h at 77 V. For the analysis of in vivo plasmid topology, pGFPuv (Invitrogen) plasmid was transformed into MG1655 and SK3842 strains. The plasmids were separated on 1% agarose gel as well as twodimensional agarose gel, as described above.
Chromatin Immunoprecipitation-Mid-log phase MG1655 and SK3842 cells were brought to equivalent colony forming units/ml (5 ϫ 10 9 cells/ml) and treated with formaldehyde to a final concentration of 1%. After 30 min, cross-linking was quenched by the addition of glycine (final concentration 0.5 M). Cells were harvested by centrifugation and resuspended in 0.1ϫ original volume of lysis buffer (10 mM Tris-HCl, pH 8.0, 20% sucrose, 50 mM NaCl, 10 mM EDTA, and 50 mg/ml lysozyme) for 10 min, followed by the addition of an equal volume of 2ϫ immunoprecipitation buffer (100 mM Tris-HCl, pH 7.0, 300 mM NaCl, 2% Triton X-100, 0.2% sodium deoxycholate). Cellular DNA was sonicated to an average size of 500 bp. After centrifugation, the cleared supernatant was used as the "input" fraction for the chromatin immunoprecipitation (ChIP) assay. 5 g of anti-HU antibody or 8 g of anti-HNS antibody were added per ml of input fraction and incubated overnight at 4°C. Complexes were incubated for 1 h with 30-l bed volume of pre-equilibrated protein A-Sepharose CL-4B beads (Amersham Biosciences). Samples were then washed twice with IP buffer, once with IP buffer plus 500 mM NaCl, once with wash buffer (10 mM Tris-HCl, pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% Nonidet P-40, 0.5% sodium deoxycholate), and once with 10 mM Tris-HCl (pH 7.5), 1 mM EDTA. Immunoprecipitated complexes were eluted twice by incubation of beads with 30 l of elution buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, 1% SDS) at 65°C for 10 min. The two eluates were combined as the output fraction for PCR analysis. 3% of input fraction and 50% of output fraction were diluted with Tris-HCl (pH 8.0), containing 10 mg/ml RNase A, and heated overnight at 65°C to reverse cross-linking. This was followed by treatment with 2 mg/ml Proteinase K for 2 h at 50°C. Samples were ethanol precipitated and suspended in 30 l of dH 2 O. 5 l of the DNA samples were used in a 50-l PCR reaction mixture. PCR was performed for 25 to 28 cycles and the PCR products were separated on 2% agarose gel. After staining with ethidium bromide, reverse images were photographed.

RESULTS
Mutant HU␣ Condenses Naked DNA in Vitro-Because the HU␣ E38K,V42L protein induced intense nucleoid condensation in cells harboring the mutant hupA gene (15), we investigated the potential of HU␣ E38K,V42L to mediate similar DNA condensation in vitro. We used electrophoretic mobility shift assays to elucidate and compare the DNA-protein complexes formed by wild-type HU␣ and HU␣ E38K,V4L . A 2,686-bp supercoiled plasmid was incubated with increasing amounts of wild-type HU␣ and HU␣ E38K,V42L proteins and the resulting nucleoprotein complexes were separated by agarose gel electrophoresis. Wildtype HU␣, with increasing concentrations, produced progressively retarded DNA bands until saturation was reached at a (HU␣) 2 :DNA molar ratio of 100:1. Beyond that, there was no more change in the mobility of the DNA-protein complexes (Fig. 1A). At lower protein:DNA ratios, the HU␣ E38K,V42L protein generated similarly retarded complexes, indicating that the affinity of the mutant protein for DNA was not significantly altered. However, when the molar ratio of (HU␣ E38K,V4L ) 2 : DNA was 100:1 or higher, the DNA-protein complexes were completely retained in the wells. This indicated that the HU␣ E38K,V42L protein induced a structural conformation of the DNA that was different from the one generated by the wild-type protein. Proteinase K digestion of the immobilized HU␣-DNA complex restored normal mobility, indicating that the structural alterations induced by the mutant protein was reversible (Fig. 1B). If the immobilization of the DNA-HU␣ E38K,V4L complexes within the wells of the gel was a result of DNA condensation, DNA in such complexes was likely to be resistant to nucleases. Digestion of the immobile HU␣ E38K,V42L -DNA com-plexes by restriction enzyme EcoRI showed that the HU␣ E38K,V42L protein impeded the cleavage of the bound DNA (Fig. 1C). Similarly, timed digestions of the HU␣ E38K,V42L -DNA complexes with DNase I showed that DNase I was inaccessible to the HU␣ E38K,V42L -induced DNA conformation (Fig. 1D). This demonstrated that the HU␣ E38K,V42L induced an overall sterical or conformational change in the DNA that hindered the action of both specific and nonspecific nucleases. In contrast, complexes formed with wild-type HU␣ gave regular restriction enzyme and DNase I digestion patterns. This result showed that HU␣ E38K,V42L is capable of inducing a high level of DNA condensation, which is reflected in the changes in the electrophoretic mobility and resistance to DNA-cleaving enzymes of the HU␣ E38K,V42L -DNA complexes.
HU␣ E38K,V42L -DNA Condensates Have Globular Structures-Next, we analyzed the nature of DNA-protein complexes formed by wild-type HU␣ and HU␣ E38K,V42L by AFM imaging. AFM images of a 2,686-bp supercoiled plasmid DNA showed uniform structures with intertwined strands and smooth contours ( Fig. 2A). The phase shift caused by the supercoiled DNA ( Fig. 2A, inset) was smaller than that by the substrate mica itself.
In the presence of wild-type HU␣, at a protein:DNA molar ratio of 100:1, the DNA adopted a more relaxed appearance with protein molecules distributed randomly throughout the circle (Fig. 2B). Increasing protein concentrations did not generate any further change in the overall conformation of the HU␣bound DNA. This behavior of HU␣ homodimer is consistent with previous observations with wild-type HU␣/HU␤ heterodimers (13) and reflects the ability of HU to restrain negative supercoils locally, without leading to any global DNA compaction. In the case of HU␣ E38K,V42L homodimers, the conformational changes in the DNA at a lower protein:DNA ratio (20:1) were analogous to those generated by the wild-type protein (Fig. 2C), with portions of the intertwined DNA duplexes becoming more open. But, with increasing protein concentrations, some of the DNA molecules progressed to a different configuration with the formation of patches of dense, amorphous condensate having folds of DNA emanating from them ( Fig. 2D). Finally, at a (HU␣) 2 :DNA molar ratio of 100:1, all the DNA molecules collapsed into compact globular particles (Fig.  2E). The condensates were 30 -60 nm in diameter and 3-4 nm in height. In comparison, the mutant HU␣ protein itself was globular, but much smaller, appearing under the AFM as a particle with Ͼ10 nm diameter and a 1-1.5-nm height (Fig. 2F). This condensation of extended, unconstrained DNA molecules into globular particles is comparable with the coil-globule transition induced by neutral polymers and polycationic lipids. Supercoiled pUC12 DNA was shown to be transformed to 39 -45-nm particles by cobalt hexamine chloride (17), which is very similar to the size of the plasmid DNA condensates that was obtained with HU␣ E38K,V42L . Another striking characteristic of the DNA condensates was the significant phase shift in the AFM cantilever oscillation compared with that observed on the background mica or the non-condensed DNA (Fig. 2E, inset), indicating that the exterior of the condensed DNA has a significant charge redistribution compared with its free form. The minimum concentration of the mutant HU␣ required for complete condensation of the 2,686-bp supercoiled plasmid DNA was around 100 molecules of mutant HU␣ dimers per DNA molecule, with no more visible changes at higher protein concentrations. With a linear 7,249-bp DNA (Fig. 2G), wild-type HU␣ caused thickening of the DNA, as demonstrated by the considerable increase in thickness of the DNA strands along with some kinks at different positions of the DNA (Fig. 2H), similar to what has been reported previously (18). At lower concentrations of HU␣ E38K,V42L , some intermediate forms of DNA condensates were observed, where DNA strands projected out from the edges of partially condensed DNA mole-cules ( Fig. 2I). At higher concentrations (protein:DNA molar ratio of 120:1), DNA was once again condensed into dense globular particles that were 50 -150 nm in diameter and 10 nm in height (Fig. 2J). The phase shift observed with the supercoiled plasmids (Fig. 2E, inset) was also present here (Fig. 2J, inset). It appears that the compacting properties of HU␣ E38K,V42L are independent of size, sequence, and supercoiled status of the target DNA.
Transmission electron micrographs of plasmid DNA-HU␣ E38K,V42L condensates revealed that the DNA was condensed into tightly coiled spheres with apparently little space in between (Fig. 2K).
HU␣ E38K,V42L Wraps DNA Around Its Core Surface-Next, we took a closer look at the architecture of individual complexes formed by HU␣ E38K,V42L on supercoiled and linear DNA. At a low protein:DNA ratio (10:1), there appeared small globular nucleoprotein structures at positions where the proteins were located. The HU␣ E38K,V42L -DNA complexes (Figs. 3A, i-ii) were quite uniform in dimensions, with heights that were 2-4 times the height of the naked DNA (height of naked DNA was ϳ0.6 nm, whereas heights of the DNA-protein looped complexes were 1.5-2.2 nm). The two simplest forms of HU␣ E38K,V42L binding to DNA would be: (a) HU␣ E38K,V42L binds to the DNA minor grooves by its ␤-sheet arms and induces local bends like the wild-type HU (19) or (b) HU␣ E38K,V42L wraps the DNA around itself in a nucleosome-like structure. A wrapping mode of binding would cause a significant reduction in the contour length of the DNA, whereas a simple DNA minor-groove binding mode would produce kinks but not much length reduction. Although the exact contour length of a supercoiled DNA is hard to measure accurately, formation of a single HU␣ E38K,V42L -mediated wrapped complex caused a significant reduction in the apparent length of the DNA. The contour length of unbound plasmid DNA was around 835 Ϯ 20 nm. A reduction of roughly 50 -60 nm was observed for a single wrapped complex (Fig. 3A, i and  ii). The shortening of the DNA can most easily be explained by DNA wrapping around the protein core by more than 360°, namely, more than one full turn. It is also possible that some of the reduction in length is due to increased helix angle of the DNA supercoil in which case HU binding induces tighter (or more positive) superhelical DNA conformations. The height of the protein core in single-wrapped complexes (Fig. 3A, i and ii) was slightly greater than the height of free HU␣ E38K,V42L molecules present in the same sample (ϳ1.5-2 versus ϳ1.2-1.6 nm). At low protein: DNA ratios, individual wrapped complexes were clearly distinguishable. With increase in HU␣ E38K,V42L concentration, the number of wrapped protein-bound complexes per DNA molecule increased (Fig.  3A, iii and iv). Although, in a few cases the new complexes were formed at well separated locations, the assembly of new DNA-protein wrapped structures was favored predominantly in the vicinity of an existing wrapped complex. This implied that the formation of wrapped DNA-protein complexes on supercoiled DNA was cooperative in nature. The assembly of successive wrapped complexes gave the appearance of an overlap of contiguous loops, resulting in thick fibrillike segments of DNA. The diameter of these fibrils was consistent with successive 10-nm protein particles wrapped with DNA, as seen in the ϳ12 nm periodicity in the length-wise section profile in Fig.  3A, iv, inset. These closely spaced, multiple wrapped structures aggregated to form three-dimensional higher-order packing that appeared as patches of dense condensation on DNA molecules seen in Fig. 2, D and I. With an increase in protein concentration, in addition to a broadening and thickening of the DNA strands corresponding to the position of the wrapped DNA complexes, the plasmid assumed a more densely wound structure (Fig. 3A,  v). To further confirm the DNA wrapping ability of HU␣ E38KV42L , we analyzed the contour length shortening of a linear PCR-generated 444-bp DNA fragment upon complex formation with HU␣ E38K,V42L . Representative images of the linear DNA complexed with wild-type and mutant HU␣ are shown in Fig. 3B. The protein-free 444-bp DNA had a contour length of 150 Ϯ 4 nm (Fig. 3B, i). Wild-type HU␣ were found to form local bead-like structures on the DNA but the contour length of the HU␣bound DNA (Fig. 3B, ii and iii) was essentially the same as that of naked DNA. Surprisingly, we did not observe any sharp kinking or bending of the DNA upon HU␣ binding. HU␣ E38K,V42L formed large spherical structures and was bound both at the ends and internal positions along the DNA (Fig. 3B, iv-viii). Analysis of DNA contour lengths upon HU␣ E38K,V42L binding showed that the average length of the DNA with a single nucleoprotein complex was 132 Ϯ 4 nm, amounting to a shortening by about 18 Ϯ 4 nm (Fig. 3B,  iv-vii). This represents an average DNA contour length reduction of 54 Ϯ 11 bp as a result of HU␣ E38K,V42L binding. Although the path of DNA cannot be followed accurately in these images, the loss of this relatively large amount of DNA is consistent with the wrapping of the DNA around the HU␣ E38K,V42L protein core surface. With two nucleoprotein complexes, the length of the DNA was further reduced to 110 Ϯ 5 nm (Fig. 3B, viii). This represented a contour length shortening by about 119 Ϯ 14 bp that was roughly twice the reduction made by a single wrapped complex. Because HU is such a small protein, for a significant length of DNA to be wrapped around the protein core surface it is likely that a multimeric form of HU␣ E38K,V42L was involved (as shown in a later section). Thus, we conclude that HU␣ E38K,V42L wraps DNA in spherical, nucleosome-like structures.
HU␣ E38K,V42L Generates Positive Supercoiling in DNA-The DNA wrapping by HU␣ E38K,V42L is expected to change the writhe of a supercoiled plasmid DNA molecule. Intuitively, an initial unwinding of the supercoiled plasmid DNA by HU␣ E38K,V42L , as seen in the AFM images, followed by enhanced supercoiling of the plasmid DNA with increasing protein concentrations would suggest that HU␣ E38K,V42L introduces positive supercoils in a negatively supercoiled DNA. We investigated the topological changes generated in vitro in a closed circular DNA upon interaction with HU␣ E38K,V42L in comparison to wild-type HU␣. Because HU␣ has relatively very low affinity for relaxed DNA (20), we used partially relaxed 2,686-bp plasmid DNA as substrate and incubated with increasing concentrations of the HU␣ and HU␣ E38K,V42L . After subsequent incubation with eukaryotic topoisomerase I to remove all unconstrained supercoils in the plasmid DNA followed by deproteinization, the reaction products were analyzed by two-dimensional agarose gel electrophoresis (Fig. 4). Increasing amounts of HU␣ E38K,V42L led to a progressive conversion of the negative topoisomers to completely positively supercoiled DNA species (Fig. 4A, lane 9). Protein-free DNA had an average of 8 -9 negative topoisomers. With increase in HU␣ E38K,V42L concentrations, the average linking number changed from slightly negative (lane 2) to relaxed (lane 7) to positive (lane 9). The net change in average linking number was from Ϫ3 to ϩ1. This change is most likely due to introduction of positive supercoils, which first neutralizes the existing negative supercoils and then appears as exclusive positive topoisomers. In the presence of calf thymus DNA topoisomerase 1 alone, under identical conditions, more than 90% of the negative topoisomers were converted to the relaxed form (Fig. 4B), indicating that the positive topoisomers were stabilized by HU␣ E38K,V42L in the DNA-protein complexes. Wild-type HU␣ at lower protein concentration increased the average linking number from Ϫ3 to Ϫ4 (Fig. 4C, lane 2). At higher protein concentrations, HU␣ seems to cause a slight reduction in the negative superhelicity of the plasmid DNA. We conclude that HU␣ E38K,V42L , at high concentrations, alters the topology of closed, circular DNA molecules by inducing positive supercoiling. To check whether HU␣ E38K,V42L can also generate positive supercoils in vivo, we analyzed the topology of the plasmid isolated from wild-type and hupA mutant strains. In agarose gels, plasmid isolated from the mutant strain showed two distinct populations: a faster-moving band that corresponded with the plasmid band isolated from the wild-type strain and a set of slightly slower moving bands, which were absent in the wildtype strain (Fig. 4D). In two-dimensional gel electrophoresis, the plasmid from the mutant strain showed that it indeed had two distinct populations of plasmid topoisomers. There was a band corresponding to negatively supercoiled topoisomers as well as a set of distinct positively supercoiled topoisomers (Fig.  4E). In contrast, the plasmid isolated from the wild-type strain was almost exclusively negatively supercoiled. This confirms that HU␣ E38K,V42L can generate positive supercoils in plasmid DNA both in vivo and in vitro. The reason why there are two populations of plasmid topoisomers would need further detailed investigations.

DNA Wrapping by Octameric HU
DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 HU␣ E38K,V42L Self-associates to an Octamer Both in Vivo and in Vitro-To better correlate the DNA wrapping and condensing property of HU␣ E38K,V42L with its molecular structure, we investigated the oligomerization state of the mutant protein.
Samples of His 6 -tagged wild-type HU␣ and HU␣ E38K,V42L were separated on 4 -20% BisTris gels using non-denaturing sample buffer and omitting the sample boiling step prior to loading on the gels. There appeared to be clear differences in the oligomeric status of the wild-type and mutant HU␣ protein (Fig. 5A). In the case of wild-type HU␣, there appeared to be two distinct bands, corresponding to the monomeric and dimeric forms of the protein. However, for the HU␣ E38K,V42L protein, there was a distinct higher molecular weight band in addition to the monomeric and dimeric forms. The molecular weight of this species corresponded to an octameric form of the mutant protein. The octamer formation was not sensitive to salt concentration. To eliminate the possibility that the observed self-association of HU␣ E38K,V42L was influenced by the histidine tags and establish that the octameric form was present under physiological conditions inside the cell, we performed Western blot analysis of the cell lysates from strains harboring the wild-type hupA gene and the mutant hupA gene under the same electrophoretic conditions as described above. As evident from Fig. 5B, in both log phase and stationary phase cultures, the mutant HU␣ E38K,V42L contained a strong octameric species in contrast to the wild-type HU␣ that had no detectable higherorder oligomers.
To further confirm the existence of the octameric state of HU␣ E38K,V42L , we used analytical equilibrium centrifugation to quantify the self-associations of wildtype HU␣ and HU␣ E38K,V42L proteins in solution, by fitting the absorbance data to assumed interaction models. We considered the following models. 1) No self-association of dimers. 2) Equilibrium association of dimers and tetramers.
3) Equilibrium association of dimers, tetramers, and octamers. 4) Equilibrium association of dimers, tetramers, and hexamers. 5) Equilibrium association of dimers and octamers. Wild-type HU␣ homodimers were found to associate weakly into tetramers (model 2) as shown in Fig. 5C. There were no detectable higher oligomers present. Global fit of the data to Equation 1 (described under "Experimental Procedures") gave an association constant of ln(K 1-2 ) ϭ 7.65 for wild-type HU␣, which corresponds to a dissociation constant of K d ϭ 0.47 mM (dimer to tetramer). The free energy of association, ⌬G 0 , was about Ϫ4.5 kcal/mol. The mutant HU␣ E38K,V42L homodimers were also best fit by using Equation 1. However, at equilibrium, no tetramers were detected (Fig. 5D). Instead, it appeared that the protein formed a rather strong octamer (model 5), with an association constant of ln(K 1-4 ) ϭ 29.83 and with free energy of association ⌬G 0 ϭ Ϫ17.4 kcal/mol of tetramer. The value of K d , which was about 16 M for dimer to octamer formation, indicates a much stronger association than the wild-type HU␣. This was consistent with the PAGE analysis described earlier, where HU␣ E38K,V42L octamers were clearly detectable under conditions where no visible tetramers of the wild-type HU␣ were present. These experiments clearly established that HU␣ E38K,V42L is present in simple dimer-octamer equilibrium and there are no additional heterodisperse, oligomeric species present. This self-association of the protein is independent of DNA binding.
HU␣ E38K,V42L Has Different Modes of Interaction with Different Genetic Loci-As mentioned in the Introduction, HU␣ E38K,V42L caused a profound shift in the global transcription pattern in the cell. Two of the representative genetic loci demonstrating this reprogramming of the transcription plan are gal (encoding enzymes for galactose metabolism) and hlyE (encoding the cytolytic protein hemolysin). The gal locus is super-repressed in the mutant hupA gene and the locus is tran-  MG1655 (lanes a and b) and hupA mutant SK3842 (lanes c and d) with anti-HU antibody. Lanes a and c were log phase culture samples, whereas lanes c and d were stationary phase culture samples. C and D, experimental data and modeling of absorbance equilibrium data for HU␣ and HU␣ E38K,V42L . C, data for wild-type HU collected at 233 nm were best fitted with a (HU␣) 2 7 (HU␣) 4 equilibrium model, and D, data for the mutant HU␣ E38K,V42L collected at 234 nm were best fitted with a (HU␣) 2 7 (HU␣) 8 model. Every other data point is shown for clarity purposes.
scriptionally silent even in the presence of the inducer, D-galactose (15). Conversely, the cryptic hlyE gene, which remains quiescent in wild-type E. coli under normal laboratory conditions, is constitutively expressed in the mutant hupA strain (15). To understand the role of HU␣ E38K,V42L in the reversal of the basal transcription status in these two genetic loci, we used the ChIP technique to determine the degree and extent of association of HU␣ E38K,V42L with the regulatory regions gal and hlyE.
ChIP experiments define DNA domains that are associated with a particular protein, with the location of the peak corresponding to the actual binding site of the protein and the spread of signal on either side determining the level of association. The width of the distribution pattern is proportional to the number of protein molecules bound at the locus. We used primer pairs to probe 200 bp of the promoter regions of gal and hlyE and two upstream and downstream regions whose distances were 500 bp from the 5Ј and 3Ј ends, respectively, of the promoter-specific fragment (Fig. 6D). Immunoprecipitation with anti-HU antibody showed that, in the presence of galactose, wild-type HU␣ had little or no association with the promoter, the promoter-upstream or the promoter-downstream regions probed in the gal regulatory region (Fig. 6A). Conversely, under the same conditions, HU␣ E38K,V42L associated robustly not only with the gal promoter region but showed almost identical levels of association with both the upstream and downstream region probes. Such a spread of the mutant protein over a large DNA segment indicated that multiple molecules of HU␣ E38K,V42L were bound to this region, probably in the form of nucleoprotein filaments, to turn off transcription. On the other hand, the hlyE locus presented a much different picture (Fig. 6B). Here, HU␣ E38K,V42L was found to be associated with the promoter region but the association dropped sharply on either side of this region, showing that, unlike in gal, HU␣ E38K,V42L binding to the hlyE promoter region did not spread. The wild-type HU␣ was sometimes found to be weakly associated with the promoterupstream region but was not consistently detectable in all samples. The hlyE locus is under the repressive control of H-NS and has two H-NS binding sites, located on either side of the transcription start site (21). We examined the association of H-NS with the hlyE locus in the presence of HU␣ E38K,V42L (Fig. 6C). Chromatin immunoprecipitation with anti-H-NS antibody in the wild-type background revealed that H-NS was bound strongly to the promoter region. However, the signal was sharply diminished in the mutant stain expressing HU␣ E38K,V42L . These results revealed that binding of HU␣ E38K,V42L interfered with the formation of the H-NS nucleoprotein complex at the hlyE promoter.
If gal and hlyE represented the new targets of HU␣ E38K,V42L for which wild-type HU␣ had little or no affinity, the hupA gene itself represented a location where the normal association of wild-type HU␣ was reversed for HU␣ E38K,V42L . Fig. 6E shows that HU␣ was strongly associated with the hupA regulatory region, as has been reported before. However, there was almost no immunoprecipitable hupA DNA in the case of HU␣ E38K,V42L . Taken together, these experiments demonstrate that HU␣ E38K,V42L has a altered map of distribution on the bacterial chromosome and has at least two different modes of interaction with the regulatory loci of the genes whose expression are altered by this protein.

DISCUSSION
The bacterial nucleoid is a dynamic entity whose structure is maintained by a delicate balance between the histone-like proteins, condensins like MukB, global superhelicity, and general transcription status of the cell. An alteration in any one of the key players in this inter-connected web is expected to shift the balance to a different state of nucleoid equilibrium. These kinds of changes can be extremely detrimental to cell physiology and result in an unsustainable condition. We have isolated a mutant form of bacterial HU protein, HU␣ E38K,V42L , which caused major changes in the nucleoid architecture and activity by virtue of its radically different biochemical and functional properties. Wild-type HU is functionally a dimer and has been well characterized as a DNA-bending protein that constrains negative supercoils. HU␣ E38K,V42L forms octamers in solution and is capable of inducing extremely high levels of DNA condensation. It wraps DNA around itself to form nucleosome-like structures that restrain positive supercoils. Finally, the mutant HU␣ protein shows a major change in the distribution pattern within the chromosome and has at least two different kinds of association patterns in the regulatory regions of genes whose expression was modulated by the mutant protein. These properties are all starkly different from the well studied characteristics of wild-type HU and forms the foundation for the sweeping morphological and physiological changes encountered in cells harboring the mutant hupA gene in the chromosome (15).
Unlike H-NS, where genuine evidence for its DNA-condensing properties are abundant, the exact role of HU in nucleoid compaction is murky at best. Overproduction of HU does not Chromatin immunoprecipitation assays with hupA mutant SK3842 and wild-type E. coli MG1655 cells. ChIP assays were performed with anti-HU antibody (A, B, and E) and anti-H-NS antibody (C). DNA samples isolated from immunoprecipitation (ChIP) and prior to immunoprecipitation (Input) were amplified by PCR with primer pairs specific to the regions indicated by the diagram in D. For A-C, primer pairs for regions 1, 2, and 3 were used, whereas for E, the primer pair for only region 1 was used.
increase chromosome condensation or impede cellular transcription (22). AFM results show that HU-bound DNA assumes a more open conformation (13). One of the current theories about the role of HU in nucleoid organization is that HU, in fact, acts as a foil for H-NS action by counteracting DNA condensation and the nucleoid structure is determined by the opposing roles of these two histone-like proteins. Given the high degree of DNA condensation by HU␣ E38K,V42L , both in vivo and in vitro, it initially appears extraordinary that the bacterial nucleoid is not only able to withstand this degree of powerful compaction but exhibit a robust, albeit a very different, transcription profile. Interestingly, there have been reports of HU from two different thermophilic bacteria that can induce strong DNA condensation in biochemical experiments (23,24), similar to what was observed with HU␣ E38K,V42L . There have also been reports of HU-like proteins from chloroplasts and mitochondria that caused a high degree of nucleoid compaction in E. coli without being detrimental to cell growth and survival (25)(26)(27). The in vivo concentration of HU␣ E38K,V42L molecules (1 dimer per 100 bp), which has been shown to be similar to that of wild-type HU, would be far less than the amount needed for complete condensation of the nucleoid (1 dimer per 26 bp). Using large bacterial artificial chromosomal DNA at a protein:DNA ratio of 1 dimer per 100 bp, we can show that HU␣ E38K,V42L causes localized patches of dense condensation on the chromosome, interspersed with regions of HU␣ E38K,V42L -free DNA. 3 The DNA condensation mechanism of HU␣ E38K,V42L is probably related primarily to its DNA wrapping property. In this scenario, successive toroidal DNA wrappings by HU␣ E38K,V42L would help promote DNA condensation by cooperative protein-protein interaction between adjacent wrapped DNA-bound proteins, similar to eukaryotic and archaeal histones. HU␣ E38K,V42L , at high concentrations in vitro, introduced positive supercoils in plasmid DNA in the presence of topoisomerase I. Plasmid isolated from the hupA mutant strain also had a population of positively supercoiled topoisomers along with the negatively supercoiled plasmid species. From these results, we propose that the mutant HU generates positive supercoils at chromosomal segments where it binds as octamers, creating regions of positively supercoiled domains in the chromosome. The generation of positive supercoils would suggest that the HU␣ E38K,V42L wraps the DNA in a right-handed fashion. The supercoiling status in the cell is a principal determinant of both gene expression patterns as well as the compaction state of the nucleoid (28). Negative supercoiling provides the energy for DNA melting required for most DNA transactions, whereas positive supercoiling stabilizes the double helix. Intuitively, the DNA regions bound and condensed by multiple, closely spaced HU␣ E38K,V42L would be expected to be transcriptionally inactive due to steric hindrance. This was supported by the chromatin immunoprecipitation assays that showed that inside the cell, under inducing conditions, the HU␣ E38K,V42L occupied a very large segment of the gal regulatory region while the wild-type HU␣ had almost no association with this region. The formation of this kind of repressive nucleoprotein complex is similar to those formed by H-NS at the loci that under its control. But, whereas H-NS functions exclusively as a global repressor by forming these silencing complexes, HU␣ E38K,V42L displays a locus-specific binding pattern distinction that is correlated to the transcription status of that locus. The hlyE locus, which is a phenotypically silent gene in E. coli, is activated in the presence of HU␣ E38K,V42L . ChIP assays showed, unlike in the gal region, the HU␣ E38K,V42L has only limited association with the hlyE regulatory region, covering a discrete part of the region that was probed. This suggests that, at active transcription loci, binding of HU␣ E38K,V42L in the promoter region is restricted. The binding of HU␣ E38K,V42L to the upstream region of hlyE also negatively affected the binding of H-NS that is the typical regulator bound to this region. We hypothesize that the binding of a restricted number of HU␣ E38K,V42L to the hlyE promoter modulates the DNA topology unfavorably for H-NS binding and favorably for RNA polymerase recruitment. Interestingly, many of the genes activated by HU␣ E38K,V42L like hlyE, proU, and csg, are under the negative regulation of H-NS. Factors that relieve H-NS-mediated gene silencing are usually sequence-specific transcription activators like IHF, CRP, Cfa-D, and VirF. If a promiscuous DNA-binding protein like HU is transformed into a dominant H-NS antagonist, the effect on the basal gene expression program would be far more widespread. We also showed that, not only does HU␣ E38K,V42L have non-conventional DNA binding targets, it is also absent from some of its traditional sites like the hupA regulatory region. The DNA structural specificity and molecular mechanism that determines the mode of interaction that HU␣ E38K,V42L would follow at any particular genetic locus would require further detailed analysis.
The technical inability to detect an ordered structure of bacterial nucleoid, the promiscuous ground state of global transcription, and the association of nucleoid condensation to general shut-down of cellular functions have long fostered the belief that bacteria do not have the sophisticated hierarchy of chromosome organization like that found in Archaea and Eukarya. The finding that a mutant form of HU␣ can assemble toroidal-wrapped nucleoprotein complexes locally, which is translated to a global change in nucleoid architecture and transcription profile, indicates that the correlation between chromosome organization and transcription pattern is more complex in bacteria than commonly believed. Any factor that has the potential to alter the level of nucleoid compaction must have an inherent effect of regulating DNA accessibility and consequently, global transcription. So far, almost all studies have linked increased chromosomal compaction in bacteria to general downshifts in genomic activity. HU␣ E38K,V42L reveals unexpected functional links between bacterial chromosome condensation and defined transcription pattern changes. Like in eukaryotes, bacterial chromosome re-organization may provide the structural basis for whole genome transcription changes necessary for adaptation under certain stressful environmental conditions. Additional biochemical and physiological studies of HU␣ E38K,V42L can provide valuable insights into the mechanism and dynamics of nucleoid organization and its functional outputs.