Contribution of DNA conformation and topology in right-handed DNA-wrapping by the Bacillus subtilis LrpC protein

show unusual not assigned to proteins or other and that and ie, controls the order of events in protein/DNA leads to polymerization of LrpC and bridging of DNA fragments through protein-protein interactions. D , Interaction of LrpC with a negatively supercoiled plasmid DNA. LrpC creates a positively supercoiled loop which is compensated by a new negatively supercoiled loop. The latter is a target for LrpC. This model explains the formation of successive wrappings in a very close proximity by an invasive mechanism which induces partition of topological domains between LrpC-restrained positive supercoils and free negative ones.


INTRODUCTION
The structural properties of DNA and specific DNA/protein interactions are crucial for the regulation of fundamental cellular processes such as recombination, replication and chromosome organization. In prokaryotes several small DNA binding proteins regulate these processes by local changes in DNA conformation, through the formation of specific nucleoprotein complexes. In Escherichia coli these small DNA binding proteins include the CRP, IHF, Fis, H-NS, Dps, Lrp and HU proteins (1), and in Bacillus subtilis the HU-like protein, HBsu (2). Notably these proteins can regulate DNA transcription. For example, CRP, IHF and Fis facilitate the association of RNA polymerase with upstream DNA sequences or with activator proteins, and can enhance the interactions of activator or repressor proteins at distant sites (3,4). Moreover, HU binding to promoter regions modulates the binding of other transcriptional regulators like CRP (5), LexA (6) and GalR (7). Change of DNA conformation by protein/DNA interaction, however, is not limited to prokaryotic species.
The lrpC gene was identified during the B. subtilis genome sequencing project (11). It encodes a neutral 16.4 kDa protein, LrpC, that forms tetramers in solution. Based on amino acid sequence identity, it has been assigned to the Lrp/AsnC family of transcriptional regulators which in B. subtilis includes seven Lrp/AsnC-like proteins (12)(13)(14). The N-terminal region of LrpC is predicted to form a typical Helix-Turn-Helix DNA binding domain, characteristic of numerous transcriptional regulators (15). Previous experiments have shown that the lrpC gene is autoregulated (16). In addition, phenotypic analysis of a lrpC mutant in B. subtilis has revealed a possible role of LrpC in branched chain amino-acids metabolism, in sporulation, and in long term adaptation to stress (16).

DNAs and protein
The 648-bp fragment encompassing the lrpC promoter region was obtained by a PvuII digestion of the plasmid pUC18prolrpC (16). Shorter, 331-bp DNA fragments containing 5'-lrpC region used in Electron Microscopy (EM) and Atomic Force Microscopy (AFM) experiments were obtained by PCR and purified by an anion exchange MonoQ column using a SMART system (Pharmacia): α and β fragments correspond respectively to -225 to +106 and -270 to +61 with respect to the P1 transcription start site.
The β fragment was biotinylated at its 5'-or 3'-extremity and dimerized using streptavidin. After dimerization of the β fragment the promoters P1/P2 are localized near the extremities when using the 5' biotinylated fragment (5'β dimers) or near the center using the 3' biotinylated fragment (3'β dimers).
Plasmid pBR322 was used in both Electrophoretic Mobility Shift Assays (EMSA) and EM experiments.
Supercoiled pBR322 was from Pharmacia and linear pBR322 DNA was obtained by a SalI digestion of the plasmid. It was then purified using the High Pure PCR product purification kit from Roche Molecular Biochemicals. Relaxed DNA used in relaxation assays was from Lucent Ltd. Topoisomers of plasmid pTZ18R were prepared by a topoisomerase I assay in presence of ethidium bromide to obtain negative topoisomers and in presence of netropsin to obtain positive topoisomers.
The LrpC protein was previously purified (16) and shown to form tetramers in solution (data not shown and 19). Consequently, all the concentrations of protein used in this work correspond to LrpC tetramers. 5 incubation for 10 min at room temperature, the reaction was loaded onto a 6% acrylamide/N,N'methylenebisacrylamide (80:1 final ratio) gel containing 10% glycerol or onto a 0.7% agarose gel in 44.5 mM Tris Borate, 2 mM EDTA (pH 8.3) (TBE 0.5X). Electrophoresis was performed at 10 V/cm and at 4°C.
Radioactive gels were dried, visualized by autoradiography and sometimes quantified with a Phospho Imager (Molecular Dynamics). Non radioactive gels were stained in TBE 0.5X containing 0.2 µg/ml ethidium bromide.

Observation of LrpC/DNA complexes by Electron Microscopy and Atomic Force Microscopy
Complexes were formed as described for EMSA with the following modifications. DTT, PMSF and spermidine were removed from Binding Buffer M to avoid any interference or artifact due to these components. The volume of the assay mixture was 40 µl. After incubation for 10 min at room temperature the complexes were purified by gel filtration (Superose 6B, APBiotech), with a SMART system (APBiotech) to remove unbound protein and to reduce non-specific binding. EM observations were performed as previously described (20). 5 µl of LrpC/DNA complexes, at the concentration of 0.5 µg/ml of DNA, were deposited onto a 600 mesh copper grid covered with a thin carbon film activated by a glow discharge in the presence of pentylamine (21). Grids were washed with aqueous 2% uranyl acetate, dried and observed in annular darkfield, in a Zeiss 902 electron microscope. Using this spreading procedure DNA molecules are rapidly adsorbed onto the carbon film with no major loss in the tridimensional information (22). LrpC-DNA complexes were observed at a final magnification of x 340,000 on a TV screen. Images of LrpC-bound DNA molecules were stored and digitized with a Kontron image processing system as previously described (17). The data were processed in a PC computer and the DNA-protein interactions were mapped from 250 complexes. DNA foreshortening gives an estimation of the length wrapped around the particle.
To analyze LrpC/DNA complexes by AFM, 20 µl of the same solutions used for EM in presence of 5 mM Mg 2+ were deposited onto freshly cleaved mica and then washed with 0.2 % (w/v) aqueous uranyl acetate (23).
The observation was performed in the tapping mode in air specifically available with nanoscope IIIa (Digital instruments-Veeco).

Effect of LrpC on DNA supercoiling in vitro
Different amounts of LrpC ranging from 37.5 nM to 1500 nM (in tetramers) were incubated with 20 nM of pBR322 (supercoiled or relaxed) at room temperature for 15 min in a total volume of 10 µl of buffer containing 6 20 mM Tris HCl pH (7.5), 50 mM NaCl, 0.1 mM EDTA, 1 mM DTT and 20% glycerol. Wheat germ topoisomerase I (2 U) was then added and the incubation continued for 150 min at 37°C. The DNA was deproteinized by adding SDS and NaCl to a final concentration of 1% and 1.7 M respectively, followed by extraction with phenol/chloroform/isoamyl alcohol (25/24/1 v/v) and the DNA precipitated with 100% ethanol.
The DNA pellet was resuspended in 10 µl TE (10 mM Tris HCl (pH 7.5), 1 mM EDTA) and loaded onto a 1% agarose gel. One-dimension electrophoresis was performed for 16 h at 1.5 V/cm in TAE buffer (40 mM Tris HCl, pH 8.3, 25 mM sodium acetate, 1 mM EDTA). Two-dimension electrophoresis was performed as follows: in the first dimension samples were separated in 1 % agarose for 6 h at 3 V/cm in TAE buffer. The gel was then equilibrated for 30 min in TAE containing 10 ng/ml ethidium bromide. The second dimension electrophoresis was performed for 16 h at 1.3 V/cm in the same buffer. Gels were stained with 0.2 µg/ml ethidium bromide.

RESULTS
The effect of LrpC on lrpC promoter architecture -Previous experiments have shown that the LrpC protein binds the upstream region of the lrpC gene in vitro (16; 19). As many transcriptional regulators are known to modify the geometry of their target promoters, we wanted to determine whether LrpC displays such a property.
Therefore, the interaction of LrpC with the lrpC promoter region was visualized by EM. Purified LrpC protein was incubated with a 648-bp DNA fragment digested from plasmid pUC18prolrpC that contains the α fragment in Fig. 2A, this is 331-bp of the 5'-lrpC region (16). Protein/DNA complexes were visualized by EM using an annular dark field mode (24).
The simultaneous presence of free DNA molecules, and of LrpC/DNA complexes, which were either partially or completely condensed (Fig. 1a) confirms the cooperative binding of LrpC to DNA as previously shown (16). Some DNA molecules displayed thickening that was sometimes associated with bending of the DNA ( Fig. 1 and data not shown). Various degrees of organization of the lrpC promoter were observed, ranging from a local binding of LrpC along the DNA to DNA wrapping ( Fig. 1b to e). It is tempting to suggest that these series of micrographs as put in the order b to e actually represent the progressive interaction of LrpC with DNA. Measurements of DNA length in LrpC/DNA complexes indicated that such interaction corresponds either to less than one turn of the DNA molecule around the protein core ( Fig. 1c and d), or to the wrapping of more than one turn of the DNA (Fig. 1a arrow and 1e) which thus appears shorter.
LrpC wraps DNA to form nucleosome-like structures -The conformation of the 648-bp DNA fragment containing the lrpC promoter was significantly altered when bound by the LrpC protein. To further investigate this change in DNA conformation we analyzed the interactions of LrpC with the α fragment itself, which encompasses only the 5'-lrpC region ( Fig. 2A, α fragment, -225 to +106 with respect to the P1 transcription start site). Protein/DNA complexes were allowed to form for different lengths of time using a LrpC/DNA molar ratio of 4:1 (LrpC protein concentration in tetramers, its preferred quaternary structure in solution). The complexes were subsequently visualized by EM at high magnification (x140 000). After 1-10 min the complexes appeared to be localized at various positions of the α fragment, although a preference at or near the extremities was observed (data not shown). After a longer incubation time (15 min) LrpC seems to be nearer to the center of the α fragment, with the DNA molecule clearly tightly wrapped around it ( Fig. 2A). This LrpCmediated DNA wrapping creates spherical structures resembling nucleosomes (Fig. 2Aa). The contour length of the DNA wrapped around LrpC (averaged from measuring 50 LrpC/DNA complexes, see Experimental 8 procedures) was 28 nm +/-4 nm, which corresponds to 80 bp +/-12 bp, and a radius of curvature was 4.5 nm +/-0.2. The presence of intrinsic curvature in the lrpC promoter region presumably promotes the DNA wrapping around LrpC (16). As shown in Fig. 1, several LrpC/DNA complexes resulted from multiple wrappings of the DNA that induced a highly ordered condensation (data not shown).
We also used AFM under air dried conditions to obtain topographic information about the LrpC/DNA complexes. The main results shown in Fig. 2Ab confirmed that the thickenings observed by EM ( Fig. 1 and 7) were really due to the presence of the protein. LrpC covered various lengths of DNA but such complexes were not always associated with DNA bending. LrpC binding therefore appears to progress until stable wrappings are formed (Fig. 2Aa) resulting also in spherical structures.

Localization of the LrpC binding site on the lrpC promoter -To map the location of LrpC binding to the lrpC
promoter DNA fragment we have used the β fragment (Fig. 2B, -270 to +61 with respect to the P1 transcription start site). The β fragment has the same length as the α fragment but the P1 promoter region is much closer to the extremity of the DNA molecule. To obtain an orientation of the β fragments those were bridged by their 5' (5'β dimers) or 3' (3'β dimers) extremities (see Experimental Procedures). In the 5'β dimers the P1 promoter regions localized at the extremities of the dimers. In the 3'β dimers the P1 promoter regions are gathered at the center of the dimers. When LrpC was incubated either with the 5'β dimers ( Fig. 2Ba and b) or the 3'β dimers (Fig. 2Bc), its binding coincided with the position of the P1 promoter as visualized by EM.
Some unbridged β monomers present in the preparation were complexed with LrpC at their extremities. 250 LrpC/5'β complexes were mapped to precise the location of LrpC. 78% of the LrpC/5'β dimers complexes had LrpC bound at the P1 promoter region. Only 10-20% of the complexes had LrpC localized at the P2 promoter region. The average length of DNA complexed with LrpC was 90-bp with a standard deviation of 42.5-bp.
Here also multiple DNA wrappings around LrpC were observed (data not shown).
Binding of LrpC to linear versus supercoiled plasmid DNA -LrpC DNA binding properties described above are not restricted to the lrpC promoter region. This was first demonstrated in an electrophoretic mobility shift assay (EMSA) by using pBR322 DNA as a competitor for the previously shown binding of LrpC to a 32 P-labeled lrpC promoter DNA fragment ( Fig. 3A and C) (16). Increasing concentrations of plasmid DNA were able to disrupt the highly retarded radioactive complexes LrpC/lrpC promoter ( Fig. 3A and C). Moreover, a remarkable difference was observed in the competing ability of the linear and of the supercoiled pBR322 monitored by by guest on March 24, 2020 http://www.jbc.org/ Downloaded from electrophoretic mobility shift assay (EMSA). Up to 0.2 nM the linear form is more able than the supercoiled form to compete with the labeled lrpC promoter region for the LrpC protein. As a result a larger fraction of the LrpC bound DNA is released from the LrpC protein and can move further in the electric field (Fig. 3A). At higher pBR322 concentrations the supercoiled form is more effective to bind the LrpC protein and the totality of the labeled DNA is even free from the LrpC protein (Fig. 3C).
The binding between LrpC and pBR322 DNA was also confirmed directly ( Fig. 3B and D). At low LrpC concentrations a small proportion of linear pBR322 was shifted and led to clearly defined retarded complexes LrpC (data not shown). It contains a curved region previously described as C7 ( Fig. 4A; 17). To identify more precisely the region(s) recognized by LrpC within this fragment, it was cleaved by restriction enzymes into three different sets of DNA fragments (Fig. 4B). Interestingly, a 517-bp fragment that encompasses the C7 curved sequence was preferentially bound by LrpC (Fig. 4C-1). When the curvature or its position within the fragment was altered the preferential binding was lost ( Fig. 4C-2, the 517-bp fragment is cut into 349 and 168bp fragments). Finally, a 361-bp fragment containing only the C7 region was specifically bound by LrpC (Fig.   4C-3). A precise localization of the LrpC binding site was performed using a 1444-bp biotinylated fragment and EM observation (Experimental Procedures, Fig. 5Ac and d). As pBR322 contains three other major curved regions, namely, C4, C6 and C8 (17; Fig. 5B) we sought to investigate the differential affinity of LrpC for these curved regions. To this effect three DNA fragments were amplified from pBR322 by PCR. These contained the C4-C6 region (pc4-6), the C7 region (pc7) and the C8 region (pc8). The three fragments were mixed at equimolar concentration, incubated at a LrpC/DNA molar ratio of 12.5 and 200 complexes were analyzed by EM (Fig. 5Aa). No LrpC/pc4-6 complexes were observed whereas 44% of the pc8 fragments were complexed with LrpC. Consistent with the results presented above 78% of the pc7 fragments were found to be associated with LrpC. LrpC binding to the C7 and C8 regions led to the formation of stable wraps/loops as observed with the lrpC promoter region (Fig. 5Aa to d). Therefore, the presence of curvature favoured the wrapping of DNA around LrpC. Interestingly, the sequence analysis of pc8 revealed that it has two series of oligoA tracts in phase (ie on the same side of the DNA double helix) that create two successive, sharply curved domains that could be potential targets for LrpC. Indeed, double wrappings were frequently observed within LrpC/pc8 complexes as clearly visible in Fig. 5Ab. Indeed, at a LrpC/DNA molar ratio of 75:1 (one tetramer of LrpC per 60-bp) eleven topoisomers could be resolved ( Fig. 6A lane b). These experiments clearly demonstrate that LrpC interaction with pBR322 in presence of topoisomerase I introduces supercoils into a closed circular DNA, consistent with reference (19).
To ascertain whether the supercoils constrained by LrpC were negative or positive, pBR322 samples that were incubated without LrpC (Fig. 6Aa) or with 1500 nM LrpC (Fig. 6Ab) were separated by two-dimensional agarose gel electrophoresis (Fig. 6B). A mixture of negatively and positively supercoiled topoisomers migrates as a biphasic arched pattern of bands. In the presence of LrpC, the arch of topoisomers corresponded exclusively to positively supercoiled topoisomers (Fig. 6Bb), whereas, without LrpC the distribution of DNA topoisomers corresponded to the relaxed state (Fig. 6Ba). Therefore, it can be concluded that most of the DNA bound to LrpC protein is positively supercoiled.
Therefore, binding of LrpC to different forms of pBR322 plasmid DNA was further analyzed by EM. LrpC appearing as small loops were formed ( Fig. 7d and data not shown). In contrast, the assembly of LrpC with negatively supercoiled DNA led to formation of 5-6 homogeneously structured loops (Fig. 7b, c, d). This confirmed the selective affinity of LrpC for supercoiled DNA compared to linear DNA at high LrpC/DNA ratios. Moreover, these LrpC/DNA wrappings were frequently close to each other in a restricted part of the molecule (Fig. 7b). The resulting topological constraints induced by loop formation appeared to be compensated by tight winding in other parts of the DNA, when compared to free DNA molecules (compare Fig. 7b and 7a). Such a partition of DNA structural domains is clearly due to an increase in the free negative supercoiling to compensate for the LrpC-restrained positive supercoils. This clearly demonstrates that the DNA is wrapped around LrpC as a right-handed superhelix, as left-handed wrapping of negatively supercoiled DNA would result in an apparent relaxation of the molecule. Furthermore, when compared to free DNA (Fig. 7a), the part of the DNA exhibiting tight winding displayed thickening that could be due to a local polymerization of LrpC on the DNA (Fig. 7b and c). We also observed that within the same samples several supercoiled DNA molecules were highly compacted by the LrpC protein (Fig. 7d).
Considering the properties of LrpC it was important to monitor its binding to positively supercoiled DNA.
To this effect we used a pBR322-derivative plasmid, pTZ18R, also containing the C7 and C8 regions. With the native negatively supercoiled form of pTZ18R (∆L k = -15) we obtained the same pictures as with pBR322 (LrpC/DNA ratio of 28/1 corresponding to 1 LrpC tetramer per 100-bp, data not shown). However when pTZ18R was artificially positively supercoiled (∆L k = +4) only two loop complexes were observed (Fig. 7e, LrpC/DNA ratio of 8/1). Since positive supercoils are introduced by LrpC binding, the unbound DNA region is relaxed. When LrpC/DNA ratio was increased to 27/1, a mixture of two types of complexes was observed: the two loops complexes already observed at a lower LrpC/DNA ratio and new complexes showing a very organized folding of the pTZ18R (+4) on itself ( Fig. 7f and g

DISCUSSION
We previously identified LrpC as the seventh member of the Lrp/AsnC family of proteins in Bacillus subtilis and we have shown that LrpC positively autoregulates its own gene. In this study, we have analyzed in detail the interactions of LrpC with DNA, with respect to DNA conformation, curvature and topology, using EMSA, EM and AFM. We showed that LrpC progresses unspecifically along DNA, preferentially recognizes a specific type of DNA curvature and wraps DNA in a right-handed superhelix to form looped structures. In addition, we propose that its oligomerization on DNA is not random but is orientated by DNA conformation, mainly its bendability and its topological state. Moreover, LrpC is an unusual bacterial DNA architectural protein due to its capacity to constrain positive supercoiling. We propose a model for dynamic interactions between LrpC and DNA.
An octameric model for LrpC/DNA interactions -We have provided evidence that LrpC wraps DNA and forms stable complexes resembling nucleosomes with various DNA fragments including the lrpC promoter region where its binding coincides with the P1 promoter. Formation of stable complexes between LrpC and DNA results from protein-protein assembly. DNA flexibility or intrinsic curvature favours protein-protein interactions within one DNA fragment to form a stable protein core. This results in a progressive bending of the DNA that leads to loop formation through a complete wrapping of the DNA around the protein (Fig. 8B).
LrpC could interact with DNA through its N-terminal HTH motif and oligomerize through its C-terminal domain (Fig. 8A). The radius of curvature measured in the LrpC/DNA complexes correlates perfectly with the sizes of the octameric model presented for the recently crystallized LrpA protein of Pyrococcus furiosus, in which the four dimerized N-terminal DNA binding domains are diametrically opposed (25). Moreover, as LrpA, LrpC has been shown to form dimers and multimers of dimers, mainly tetramers in solution (19 and data not shown).  (29). Furthermore, all regions stably bound by LrpC contained phased A tracts preceded by a C, ie C(A)n motifs. In contrast, LrpC does not form any complexes with the pBR322 curved C4-C6 region ( Fig. 5) or with the highly curved region located upstream of the -35 box of the P1 lrpC promoter (Fig. 2B;   16). In these fragments, curvature is more related to the wedge model which attributes small deflections of the helix axis at every base-pairs step, with a predominant contribution of the AA dinucleotide (30). Albeit junction and wedge models are comparable in their general predictions of DNA curvature for fragments including phased A tracts they differ for curved fragments without A tracts motifs. Moreover, the wedge model does not take into account co-operativity effects in the stacking of AT base pairs within A tracts (31). Our results clearly show that LrpC discriminates between different types of curvature to form stable complexes within C(A)n phased motifs.

DNA topology and LrpC/DNA interactions -We have shown that LrpC/DNA complex formation is
influenced by DNA topology and moreover that LrpC constrains positive supercoiling. We propose that the formation of a first complex in supercoiled molecules is promoted by one curved region localized at one of the apices as previously observed for the transcription activator NR1 (32) and the Tth 111 glutamine synthetase (33) (Fig. 7 and 8D1). As LrpC induces the formation of a positively supercoiled loop, a new negatively supercoiled loop is then created in the vicinity of the first complex to maintain a constant linking number. The formation of a second complex is favoured by co-operative effects which promote LrpC recruitment to the flanking DNA regions (Fig. 8D2). This will induce the formation of a new positive supercoil and subsequently of a compensatory negative supercoil that will be again targeted by an LrpC oligomer (Fig. 8D3). This model explains the formation of successive wrappings in a very close proximity by an invasive mechanism which induces partition of topological domains between LrpC-restrained positive supercoils and free negative ones (Fig. 8D4). To our knowledge, we present the first visualization of such a partition in negative and positive supercoiled domains within a single DNA molecule and therefore demonstrate that positive supercoiling mediated by LrpC is due to a right-handed DNA wrapping (Fig. 7b and c). The presence of a right-handed DNA superhelix wrapped around a protein core in a negatively supercoiled environment represents a new topological paradox that could be explained by the following considerations.

15
The affinity of LrpC for DNA increases with supercoiling, either positive or negative, because supercoiling favours loops formation. Whatever their chirality, these loops promote protein/protein assembly and the stabilization of the complexes leads to the formation of a right-handed DNA helix. Such DNA transition triggered by LrpC should requires a minimal energy as shown for interaction of (H3-H4) 2 tetramer with supercoiled DNA (34).
Such a topological partition in the plasmid induces an accumulation of negative topological constraints in the free DNA which reduces its flexibility. In consequence, right-handed DNA wrapping around the LrpC protein core is no longer favoured and an alternative mode of protein-protein interaction is adopted: any additional LrpC protein polymerizes along the hyper-negatively supercoiled DNA ( Fig. 7 and 8).
This bimodal assembly of LrpC within nucleoprotein complexes is related to two types of DNA condensation, determined by the topological state of DNA. The first one results from successive DNA wrappings mainly observed with linear fragments and in regions of negatively supercoiled DNA (Fig. 1, 2  archaeal HMf or HTz tetrasomes ( Fig. 2; 38; 39). Eukaryotic dimers of (H 3 -H 4 ) (35) and HMf tetramers (40) are able, under certain conditions, to constrain positive supercoils as is observed for LrpC. Other proteins that activate transcription such as the eukaryotic transcription factors UBF or SWI/SNF, or the B. subtilis PurR regulator, are known to bind upstream of the promoters they regulate and to introduce one positive supercoil (41)(42)(43)(44). Therefore, it is likely that the capacity of LrpC to induce right-handed supercoiling is involved in its regulatory activity. In addition, like the HMf proteins, eukaryotic histones and the HMG proteins, LrpC highly compacts DNA. Such ability has not been described thus far for other members of the Lrp-like family. This work shows that the B. subtilis LrpC protein displays a mosaic of properties present in archaeal and eukaryotic histones, Lrp-like proteins, transcription factors, and eubacterial DNA structuring proteins. Consequently, LrpC is a unique member of the DNA architectural family of proteins.
A fascinating hypothesis is that micro-organisms have developed a DNA overwinding activity to compensate for the DNA underwinding activity displayed by more common nucleoid associated proteins such