Ss-LrpB from Sulfolobus solfataricus Condenses about 100 Base Pairs of Its Own Operator DNA into Globular Nucleoprotein Complexes*

Ss-LrpB from the hyperthermoacidophilic crenarchaeote Sulfolobus solfataricus P2 is a member of the Lrp-like family of Bacterial/Archaeal transcription regulators that binds its own control region at three regularly spaced and partially conserved 15-bp-long imperfect palindromes. We have used atomic force microscopy to analyze the architecture of Ss-LrpB·DNA complexes with a different stoichiometry formed with the wild type operator and with an operator mutant. Binding of dimeric Ss-LrpB to all three target sites is accompanied by the formation of globular complexes, in which the protein induces strong DNA deformations. Furthermore, DNA contour length foreshortening of these complexes indicates DNA wrapping, with about 100 bp being condensed. The average bending angle is 260°. The establishment of protein-protein contacts between Ss-LrpB dimers in these globular complexes will contribute to the cooperativity of the binding. The profound remodeling of the control region is expected to have a strong impact on gene expression and might constitute the key element in the autoregulatory process.

Archaeal transcription is a chimera of eukaryotic and bacterial features (for a recent review, see Ref. 1). The basal transcription apparatus resembles that of eukaryotes. The archaeal polymerase is a homologue of RNA polymerase II and initiation requires the basal transcription factors TATA Box-binding protein and transcription factor B. In contrast, archaeal transcription regulators are mainly of the bacterial type. This is a surprising finding and the question arises as to how these bacterial type regulators (mainly helix-turn-helix proteins) interact with this multicomponent eukaryal type basal transcription machinery. To date, little information is available on archaeal regulators. Most of the best characterized proteins belong to the leucine-responsive regulatory protein (Lrp) 2 family (2). At present, the crystallographic structure of only two members of this family is known: LrpA from Pyrococcus furiosus (3) and FL11 from Pyrococcus OT3 (4). Sulfolobus solfataricus contains at least five Lrp-like regulators (5)(6)(7)(8), including Ss-LrpB (9). Their physiological role is still unclear, except for LysM that is involved in the regulation of lysine metabolism (7).
DNA-binding proteins, especially architectural nucleoid-associated proteins, polymerases, general transcription factors such as TATA Boxbinding protein, and transcription regulators, often change the DNA conformation upon binding as part of their mode of action. This can result in DNA bending or even wrapping, DNA looping, DNA stiffening, or other forms of DNA remodeling. Intrinsic bending, proteininduced remodeling of the DNA, and the conformability of operator segments, even if they constitute linker regions that are not directly contacted by the regulatory protein(s), may play an important role in the proceeding of a regulatory response. The determination of the architecture of the regulatory nucleoprotein complex(es) is therefore a crucial step in the analysis of gene regulation.
The structural analysis of nucleoprotein complexes can be performed by biochemical analyses, crystallography, or microscopy techniques such as atomic force microscopy (AFM) or electron microscopy. AFM is becoming a widespread and very valuable technique in the study of protein⅐DNA complexes at nanometer resolution (10). This technique is straightforward, and sample preparation and analyses occur under near native conditions, thereby limiting the risk of artifacts. AFM allows the simultaneous analysis of many aspects of the complexes (e.g. morphology, contour length, and bending angle), resulting in both qualitative and quantitative information. As a consequence, the architecture of several protein⅐DNA complexes formed by bacterial and, to a lesser extent, eukaryal regulators and RNA polymerases have been studied (for a review on bacterial studies, see Ref. 11). To our knowledge, this has not yet been done for archaeal transcription regulators. Here, we use the high resolution imaging capacity of AFM to study the architecture of various Ss-LrpB⅐DNA complexes.
Previously, we have shown that Ss-LrpB binds to its own control region at three specific and regularly spaced binding sites called Box1, Box2, and Box3 (9). This Ss-LrpB-DNA interaction occurs with an apparent dissociation constant K D of ϳ10 nM, and is cooperative. Each binding site is 15 bp long, semipalindromic, and exhibits a variable degree of sequence identity with the consensus sequence 5Ј-TTGYAW-WWWWTRCAA-3Ј (Y ϭ pyrimidine, R ϭ purine, W ϭ weak bp). Box1, the most promoter proximal target, shows an overlap of 1 bp with the BRE promoter element. In-gel footprinting demonstrated that the two outermost binding sites that show a high degree of sequence identity with the consensus sequence are occupied before the middle site, Box2, is bound. Box2 shows less sequence identity with the consensus and is a low affinity site. Protein-protein interactions and DNA deformations are supposed to play an important role in this cooperative interaction (9).
Here, we determine the oligomeric state of Ss-LrpB in solution and present a detailed analysis of the three-dimensional structure of protein⅐DNA complexes exhibiting different stoichiometries, formed by binding of Ss-LrpB on the own control region (wild type and mutant forms). Measurements of contour length and bending angle of the complexes indicate that Ss-LrpB binding to the three targets accompanies pronounced DNA bending and condensation. This leads us to the hypothesis that the regulator wraps the control region DNA. These observations provide further insights into the autoregulatory mechanism that is employed by Ss-LrpB.

EXPERIMENTAL PROCEDURES
Protein Purification-Recombinant Ss-LrpB protein was obtained by a combination of heat treatment and ion exchange chromatography. Ss-LrpB was purified from a 300-ml culture of Escherichia coli BL21(DE3) containing pET24Ss-lrpBNde-null (9). The culture was grown in rich medium containing kanamycin (30 g ml Ϫ1 ) at 30°C. Expression was induced at a cell density of 13.5 ϫ 10 8 cells ml Ϫ1 by adding 1 mM isopropyl ␤-D-thiogalactopyranoside followed by overnight incubation. Cells were collected by centrifugation at 7000 rev min Ϫ1 for 10 min (Sorvall RC5B Plus, rotor SLA-1500) and resuspended in 6 ml of extraction buffer (20 mM piperidine buffer (pH 10.7)). Cells were broken by sonication for 6 min at 20% of the maximal amplitude (Vibracell, Bioblock Scientific) in a cell cooled at 4°C. After centrifugation of the disrupted cells at 10,000 rev min Ϫ1 for 10 min (microcentrifuge), the supernatant was incubated at 80°C for 10 min. Denatured proteins were removed by centrifugation at 13,000 rev min Ϫ1 for 3 min (microcentrifuge). The extract was then loaded on a 6-ml ResourceQ anion exchange column (Amersham Biosciences), equilibrated with extraction buffer. Ss-LrpB was eluted by applying a linear gradient of 0 to 1.0 M NaCl. Fractions containing Ss-LrpB were identified by SDS-PAGE and electrophoretic mobility shift acssay (EMSA) and pooled. This resulted in ϳ4 mg of electrophoretically pure Ss-LrpB. All protein concentrations were determined by a MicroBCA assay (Pierce) and are expressed in Ss-LrpB monomer equivalents.
An aliquot of purified Ss-LrpB was analyzed by gel filtration chromatography on a Superdex 75 16/60 column (Amersham Biosciences). The column was equilibrated with 20 mM phosphate buffer (pH 7.4) and calibrated with RNase A (13.7 kDa), chymotrypsin A (25 kDa), ovalbumine (43 kDa), and albumin (67 kDa). In total, 185 g of the Ss-LrpB protein was dialyzed against this buffer and loaded onto the column.
Chemical Cross-linking-Chemical cross-linking was performed as described before (12), with the following modifications. Purified Ss-LrpB was diluted in 0.5 M triethanolamine HCl (pH 8.5) (TEA buffer) to a final concentration of 30 M. Dimethyl suberimidate (DMSI) was freshly prepared in TEA buffer and added at a final concentration ranging from 0 to 10 mg/ml. Higher DSMI concentrations resulted in smearing and the disappearance of higher molecular mass bands (data not shown). The final volume of each reaction mixture was 16 l. After a 15-min incubation at room temperature, the reaction was quenched by adding 4 l of 250 mM Tris (pH 6.8). Fifteen l of each reaction mixture were analyzed by polyacrylamide gel electrophoresis in denaturing conditions (SDS-PAGE).
DNA Manipulations-The DNA fragments used in the AFM experiments were prepared by a PCR approach. To obtain the wild type operator fragment, we started from the template DNA pBendBox1 ϩ Box2 ϩ Box3 (9) and amplified a 579-bp region containing the three boxes. This was done by using ReadyMix TaqPCR reaction mix (Sigma) and the oligonucleotides 5Ј-GGTTCCGCGCACATTTCCCCG-3Ј and 5Ј-CGGCATAACCAAG-CCTATGCC-3Ј as primers. The 583-bp Box2 mutant operator fragment was obtained by a similar approach with pUC18 p/oSs-lrpBBox2TTT as template DNA (9) and the oligonucleotides 5Ј-CAGGAAACAGCTAT-GACCATG-3Ј and 5Ј-GAGAGTGCACCATATGCGGTGTG-3Ј as primers. Following the PCR, the DNA fragments were separated from parasite DNA on a 1.5% agarose gel, excised, and eluted using a GenElute gel extraction Kit (Sigma). All oligonucleotides used in this work were purchased from Sigma Genosys.
Atomic Force Microscopy-Prior to deposition, DNA molecules (ϳ130 ng) were diluted in LrpB binding buffer (20 mM Tris-HCl (pH 8.0), 1 mM MgCl 2 , 0.1 mM dithiothreitol, 12.5% glycerol, 50 mM NaCl, 0.4 mM EDTA) in a total volume of 15 l. The mixture was then diluted 2-fold in adsorption buffer (40 mM Hepes (pH 6.87), 10 mM NiCl 2 ⅐6H 2 O) and 15 l were deposited on freshly cleaved mica. Allowing 5 min of adsorption, the sample was rinsed with deionized ultrapure water, and excess water was blotted off with absorbing paper. Finally, the mica disc was blown dry in a nitrogen gas stream. Ss-LrpB⅐DNA complexes were formed by incubating 63 ng of Ss-LrpB-protein with ϳ200 ng of DNA (either the WT fragment or the mutant operator fragment) in LrpB binding buffer in a total volume of 15 l at 37°C for 20 min. This corresponds to an Ss-LrpB concentration of 180 nM. The mixture was deposited on mica following the same procedure as described above.
Immediately after deposition, 512 ϫ 512 pixel images were acquired with a Nanoscope IIIa atomic force microscope (Digital Instruments/ Veeco) operating in tapping mode at room temperature. We used Nanoprobe SPM tips, type TESP (Veeco), with 125-m cantilevers with a nominal spring constant of 50 N m Ϫ1 and resonant frequencies in the range from 279 to 362 kHz. The scan rate was 2 Hz and the scan size was 1.5 ϫ 1.5 m. All images in one analysis were obtained with the same tip and deposition.
Image Analysis-In the chemical cross-linking experiment, the band densities from the gel electrophoresis analysis were quantified using the Intelligent Quantifier software (Bio Image).
All AFM images were flattened prior to analysis using the NanoScope 6.11r1 software (Digital Instruments/Veeco). To measure the contour length and end-to-end distance, the molecules were manually traced using ImageJ (Ref. 13; available at rsb.info.nih.gov/ij/). All contour length histograms were fitted to a normal distribution (Statistical Package for the Social Sciences). Direct measurement of the bending angle was done using ImageJ, and bending angle analysis was performed by applying the Bending Analysis software (Ref. 14; available at www.nat. vu.nl/compl/bendinganalysis). The optimal bin size was determined by dividing the range of measurements by the square root of the number of measurements. Volume analysis was done by performing a section analysis with the NanoScope 6.11r1 software. This allowed measurement of the height and width of each complexed region. The width was measured at a height of 0.5 nm, and 0.5 nm was also subtracted from the measured height as a correction for the height of the DNA molecules and the background. It was hypothesized that each complexed region can be approached by an oblate spheroid with the minor axis corresponding to the measured width and the major axis corresponding to the measured height. The apparent volume was calculated as follows: V app ϭ ⅐height⅐(width) 2 /6. The apparent volume of the complexed region deviates significantly from the real volume; therefore it was only used for comparisons between molecules of the same deposition, which were imaged with a single tip.

Ss-LrpB Behaves
Mainly as a Dimer in Solution-DMSI cross-linking is a classical method used for the determination of the oligomeric state(s) of a protein (12). DMSI forms covalent cross-links between lysyl residues by aminidation of the primary amino groups. Fig. 1A shows a DMSI cross-linking experiment carried out with purified Ss-LrpB (theoretical monomer molecular mass of 17.5 kDa). One major band of ϳ32.2 kDa and four minor bands of ϳ52.2, 67.3, 78.2, and 94.1 kDa appeared after cross-linking. These correspond to cross-linked dimers and cross-linked trimers, tetramers, pentamers, and hexamers, respectively (Fig. 1A). Even at the highest DMSI concentrations used, the Ss-LrpB dimer constitutes by far the predominant cross-linked species (Ͼ85%). Tetramers and trimers made up between 5 and 10% of the cross-linked protein at DMSI concentrations of at least 1.25 mg/ml. Cross-linked hexamers and pentamers only appeared at the highest DMSI concentrations used. Therefore, in solution Ss-LrpB occurs mainly as a dimer. In parallel, we performed similar experiments with two other Lrp-like proteins, Ss-Lrp and Sa-Lrp from S. solfataricus and Sulfolobus acidocaldarius, respectively, which have a characterized oligomeric state (8). DMSI cross-linking confirmed that Ss-Lrp exists primarily as a tetramer in solution and that Sa-Lrp forms different oligomeric species up to a dodecamer (data not shown).
The oligomeric state of Ss-LrpB as determined by cross-linking was supported by size exclusion chromatography, in the sense that no Ss-LrpB oligomeric forms higher than a dimer were detected. Purified Ss-LrpB was applied to the column at a concentration of 5.2 M. Only one peak, eluting with a maximum corresponding to 22.4 kDa (Fig. 1B), contained Ss-LrpB perceptible by SDS-PAGE. This molecular mass lies closer to that of an Ss-LrpB monomer (17.5 kDa) than to that of a dimer (35 kDa). Due to possible interactions with the column beads or a nonspherical architecture of the dimeric protein, the peak can more likely be explained by a dimer eluting later than a monomer eluting earlier. In any case, no higher molecular weight Ss-LrpB peaks were observed. Therefore, it can be concluded that no significant amount of Ss-LrpB exist as oligomeric forms higher than a dimer at the concentrations used.
Visualization of DNA Molecules by AFM-Purified 579-bp-long PCR amplicons were used containing the three regularly spaced binding sites of the Ss-lrpB control region near the center of the fragment ( Fig. 2A). Each Box is 15 bp long. The center-to-center distance between Box1 and Box2, and Box2 and Box3, is 32 and 31 bp, respectively (Fig. 2B). Tapping mode AFM in air allowed visualization of these DNA molecules (Fig. 2C). The contour length (L) of 355 DNA molecules was measured by tracing the molecules from one end to the other using ImageJ. Overlapping DNA molecules or molecules that contained visible anomalies were omitted from the analysis. This resulted in an average length L of 173 Ϯ 16 nm (Fig. 3A). The calculated axial base pair rise is 0.30 nm/bp (L/579), which is a value lower than the rise of regular B-form DNA (0.34 nm/bp) but in agreement with the bp rise determined in previous AFM studies. The difference with the theoretical bp rise can be explained by the smoothing procedure that rounds sharp bends and the limited resolution of the microscope, incapable of resolving bends within a small range (15).
A primordial concern in AFM studies is whether the DNA molecules and the protein⅐DNA complexes are able to freely equilibrate on the mica surface before capture. This is in contrast to kinetic trapping, a process that results in the projection of the three-dimensional conformation of the molecules (16). The latter would result in unrelevant data on the conformation of the complexes. This can be assessed by analyzing DNA persistence length (P). P describes the bendability of a DNA molecule and is a measure of the average length at which thermal energy causes the DNA molecule to bend in another direction. The mean square end-to-end distance (͗R 2 ͘) can be expressed as a function of L and P for both cases (͗R 2 ͘ 2D or ͗R 2 ͘ proj , respectively) (16). Assuming a P of 53 nm, which is the value for DNA molecules in solution (independent of the length of the molecules), and a L of 173 nm for the operator DNA, ͗R 2 ͘ 2D would correspond to 18,597 nm 2 and ͗R 2 ͘ proj to 8623 nm 2 . The experimentally obtained ͗R 2 ͘ of this set of DNA molecules is 18,494 nm 2 which is a value consistent with the two-dimensional (2D) model. P was also determined when applying simulation-based bending analysis software (see below; Ref. 14). The normalized R distribution of these molecules was fitted using distributions obtained by simulations of DNA molecules exhibiting no protein-induced bending. This resulted in a L/p value of 3.2, corresponding to a P of 54 nm. Therefore, free equilibration can be assumed.
Visualization of Distinct Ss-LrpB⅐Operator Complexes-AFM allowed visualization of the conformational change of the operator DNA molecules caused by binding of Ss-LrpB (Fig. 2). To obtain a stoichiometrically uniform population representing full occupation of the operator, Ss-LrpB⅐DNA complexes were allowed to be formed at a Ss-LrpB concentration (180 nM) that favorizes binding to all three Boxes. Many complexes were observed, most of them being "single" complexes, containing only one DNA molecule (Fig. 2D). Single complexes with different morphologies were observed (Fig. 2, E-H). They have in common that their complexed region is typically globular shaped with two DNA arms of nearly identical length. This globular shape and the sharp angle formed between the in-and out-going DNA arms are indicative of DNA wrapping.
Besides the single complexes, "multiple" complexes were observed, containing several DNA molecules linked by higher oligomeric Ss-LrpB forms (Fig. 2, D and J). Their number might be underestimated because larger complexes might not attach as firmly to the mica surface as smaller ones. In a single case, a complex was formed with two adjacent globular regions (Fig. 2, D and I). This morphology is an indication of the existence of rare complexes (at this protein concentration) with a different stoichiometry. Likely, this molecule represents binding of two Ss-LrpB oligomers to two binding sites. The existence of these types of complexes was confirmed when analyzing complexes with a mutant operator fragment (see below). For the remainder of the analysis with WT operator complexes, only single complexes are considered.
Ss-LrpB Shortens the DNA Contour Length upon Full Occupation of the Operator-Measuring the contour length of protein⅐DNA complexes can provide more information on protein-induced conformational changes. In case of DNA wrapping a foreshortening can be expected. The contour length of AFM-visualized complexes can be measured in different ways (Fig. 3, B and C). The visible contour length corresponds to the total length of the two naked DNA arms. This might result in an underestimation of the length because the "shadow" of the protein causes a partial occlusion of the naked DNA (17). The readthrough contour length corresponds to the visible length increased with the length of the shortest path through the complexed region. This measurement could result in an overestimation when the entry and exit points of the DNA are distant; the DNA measured through the com-plexed region could then be part of the DNA that is condensed by the protein.
We measured both visible and read-through contour length for 353 Ss-LrpB⅐DNA complexes with the WT operator fragment (Fig. 3, B and  C). It can be assumed that in most complexes all three Boxes are bound (due to the high Ss-LrpB concentration). The visible contour length was on average 130 Ϯ 19 nm, which is a difference of 43 nm or 143 bp with the contour length of unbound DNA molecules. The read-through contour length was 158 Ϯ 18 nm, which corresponds to a difference of 15 nm or 50 bp. Therefore, it can be concluded that Ss-LrpB considerably condenses the DNA upon full occupation of the operator, a very strong indication of DNA wrapping. The exact amount of bps condensed by the protein will probably be situated between 50 and 143 bp. The three Boxes and linkers correspond to 78 bp.
Ss-LrpB Bends the DNA with an Average Bending Angle of 260°-Circular permutation assays (9) and the AFM images described above suggest that Ss-LrpB might cause a pronounced bending or even wrapping of the operator region. We used the tangent method to determine the apparent bending angle of the Ss-LrpB⅐DNA complexes by direct measurement of the angle between the in-and out-going DNA tails of the 353 complexes, which were also used for contour length measurements. The bending angle was calculated by subtracting the measured angle from 180° (Fig. 4A). This resulted in an apparent bending angle of 88 Ϯ 45° (Fig. 4B). This is a very broad distribution that likely reflects the effect of thermal fluctuations and the flexibility of the complexes, rather than the existence of several "microstates" exhibiting different bending angles. The latter would result in separate populations with different mobilities in EMSAs, which is not the case (9). This broad distribution is consistent with other AFM studies of protein-induced DNA bending (15,18).
Recently, a method was developed to allow bending angle analysis of AFM images based on the end-to-end distance (R) distribution of the nucleoprotein complexes (14). The R distribution, normalized by L, is fitted to histograms based on simulations using least squares minimization (14). The bin size was chosen to be optimal. The application of the simulation-based bending analysis software (Bending Analysis) indicated a best fit apparent bending angle of 100° (Fig. 4, C and D). This value is higher than the apparent angle obtained by direct measurement (tangent method) and will probably be a better approximation of the real bending angle. If the DNA is being wrapped, an apparent bending angle of 100°would correspond to a 260°bend in the opposite direction (Fig. 4A).
Ss-LrpB Binding to the Box2 Mutant Operator-To further analyze the structural identity of the globular nucleoprotein complexes observed with the WT operator, we have analyzed complexes formed with an operator mutant carrying a three bp substitution (CAA to TTT) in the downstream half-site of Box2 (Fig. 5A). As shown previously by EMSA, this mutation inhibits but does not completely abolish the for-mation of complexes with all three sites bound (9). In the DNA fragment used for AFM the operator has an asymmetric position on the fragment (Fig. 5A). We observed DNA molecules with one (Fig. 5, B and E), two (Fig. 5C), or in rare instances even three (Fig. 5D)    not always evident by means of visual inspection; the distinction was made after volume analysis of the complexes (see below). Previously we have shown by circular permutation assay that the binding of Ss-LrpB to a single Box or to the two outer Boxes results in a DNA bending angle of about 59 or 80°, respectively (9). Although this is not evident based upon the examples shown in Fig. 5, A and B, it was obvious for other complexes that Ss-LrpB also induced DNA bending upon binding one or two Boxes. Their small number does, however, not allow a correct calculation of the bending angles.
Three complexes were observed with three adjacently bound Ss-LrpB oligomers and no extensive DNA deformation (Fig. 5D). These types of complexes were only captured with the mutant operator, and they were never observed with the WT operator fragment. It is hypothesized that these complexes represent a transition state between complexes bound at two Boxes and complexes bound at three Boxes with DNA wrapping. In EMSAs, these complexes would probably be unstable (due to weak interactions with Box2 and a lack of stabilization by protein-protein contacts) and dissociate during electrophoresis, resulting in a smearing out of the third complex, as observed previously (9).
Volume Analysis of DNA-bound Ss-LrpB Oligomers-Neither the height nor the surface of a molecule measured by AFM is accurate, due to interactions between the tip and the surface, feedback settings, and tip convolution. Therefore, when comparing AFM-determined volumes, this has to be done for molecules from a single deposition and measured with a single tip. Furthermore, the results have to be inter-preted with caution. Taking this into account, several authors could establish a linear relationship between volume measured by AFM and molecular mass (19 -22).
We estimated the apparent volume of all complexed regions present on Ss-LrpB⅐Box2 mutant operator complexes exhibiting one, two, or three complexed regions observed from a single deposition (Fig. 5F). We measured the apparent height and width of these regions by applying section analysis. For complexes with two or three separate Ss-LrpB oligomers bound (13 and 3 complexes, respectively; corresponding to 35 globular regions), we obtained an average apparent volume of 45 Ϯ 23 nm 3 for the globular structure of separately DNA-bound Ss-LrpB molecules. This analysis was repeated for complexes with a single condensed region (39 complexes). The apparent volume distribution for this type of complexes appeared to be bimodal (Fig. 5F), with one maximum located near the average apparent volume of the above-mentioned complexes (40 -60 nm 3 ) and another maximum located at a much higher value (140 -160 nm 3 ). This suggests that complexes with a single globular region are composed of two species, which consist of DNA molecules bound at one Box and molecules bound at all three Boxes. An artificial limit was set to distinguish between these two types at 107 nm 3 , based on the highest apparent volume measured in the first set of complexes with two or three distinct globular regions per DNA molecule. The average apparent volume for condensed regions formed by protein molecules bound at one Box is 55 Ϯ 22 nm 3 (17 complexes). The volume for these regions in complexes bound at three Boxes, with the DNA wrapped around Ss-LrpB, is 167 Ϯ 42 nm 3 (22 complexes). Hence, the apparent volume of the complexed region of complexes with all three sites bound is on average about three times larger than the apparent volume of the distinct globular regions of one-, two-, and three-site bound complexes, when no DNA wrapping occurs. This result was confirmed with other data sets from other depositions and measured with different tips. Therefore, these data sets were not merged. The result is consistent with the binding of three interacting Ss-LrpB oligomers (likely dimers).

DISCUSSION
In this study, we present a detailed analysis of the architecture of the nucleoprotein complexes formed by binding of Ss-LrpB to the operator region of its own gene. To the best of our knowledge this is the first case of an archaeal transcription regulator system studied by AFM. With the WT operator fragment and at a high protein concentration, both single and multiple complexes were observed. The latter are complexes containing more than one DNA molecule. The existence of this type of complexes provides a possible explanation for the supershifting observed previously in EMSAs at high protein concentrations (9). This is most likely caused by protein aggregation, which is a characteristic feature of Ss-LrpB, rather than by cumulative binding at nonspecific sites on the DNA.
Previously, we have shown by EMSA and in-gel footprinting that the two outer Boxes are both bound before the weaker middle Box2 (9). AFM experiments with a Box2 mutant operator fragment allowed distinct visualization of complexes bound at one (Box1 or Box3), two (Box1 and Box3), or three Boxes. The ability to observe the individual protein oligomers bound to adjacent sites reflects the high resolution of AFM which is in this case at least 4.8 nm (corresponding to the 16-bp linker). Based on these observations suggestions can be made regarding the stoichiometry of the complexes. Ss-LrpB behaves mainly as a dimer in solution and it might be hypothesized that each Box is bound by one dimer. The semipalindromic nature of the binding sites also reinforces the idea of recognition by protein dimers. When all three Boxes are bound, close associations occur between the DNA-bound dimers. This is supported by volume analysis of the globular regions of these complexes. Combined, the results presented here indicate that the globular nucleoprotein complexes formed with the WT operator consist of DNA wrapped around three interacting Ss-LrpB dimers bound to three regularly spaced binding sites. DNA wrapping and the establishment of protein-protein contacts appear to be tightly linked in this system and might be responsible for the apparent cooperativity in the binding (9).
Ss-LrpB induces extensive DNA curvature upon binding. Previously, circular permutation assays indicated that the operator is increasingly deformed by Ss-LrpB binding to one, two, and three boxes, culminating in an apparent bending angle of 148° (9). However, this technique has its drawbacks and is not valid for bending angles above 120° (23). It is clear that Ss-LrpB-induced DNA bending exceeds this angle, given the observation that the DNA is condensed by wrapping around the protein. The DNA bending was quantified using two independent methods: the tangent method and bending analysis based on the EED/L distribution. It has been shown that the tangent method leads to an underestimation of the bending angle, especially when the DNA is strongly bent (14), and this was indeed the case. Bending analysis based on the end-to-end distance distribution revealed an average apparent bending angle of 100°, which corresponds to 260°when assuming DNA wrapping (see Fig. 4A).
Ss-LrpB binding to all three binding sites resulted in a DNA foreshortening. The exact amount of bps that are being condensed is bordered by a lower limit of 50 bp (possibly underestimated when measuring the read-through contour length) and an upper limit of 143 bp (possibly overestimated when measuring visible contour length). Previously, it was shown that DNase I footprinting resulted in a global region of protection of 86 bp (9). Therefore, we propose that the real amount of bps being condensed by Ss-LrpB is near the average between the readthrough and visible contour length which is about 100 bp. This clearly indicates DNA wrapping at full occupation of the Boxes resulting in the formation of globular nucleoprotein structures. DNA wrapping around Ss-LrpB maximizes the contact area between the DNA and the protein while allowing the Ss-LrpB dimers to establish intermolecular interactions in order to stabilize the complex (see below). This observation contributes to the aforementioned hypothesis that structurally very different Ss-LrpB⅐operator complexes exhibiting different stoichiometries would exert different autoregulatory effects (9). At low Ss-LrpB concentrations, when only one Box is bound, a positive regulation might occur. At higher concentrations, when all three Boxes are bound, severe alterations in DNA conformation, caused by the DNA wrapping, would result in a negative autoregulation. Binding to the three Boxes is cooperative (9). Therefore, small changes in Ss-LrpB concentration result in large changes in binding site occupancy and, accordingly, DNA conformation. This can lead to a fine-tuned "switch" between positive and negative autoregulation, and wrapping might be the key component of this switch. Direct confirmation of this hypothesis will be sought by in vitro transcription.
Assuming 100 bp to be wrapped around a protein core, this would correspond to a globular structure with a diameter of 95 Å. It seems likely that these molecular dimensions agree with three Ss-LrpB dimers having the DNA wrapped around them. Similarly, a model was built of an FL11 octamer with the DNA wrapped around exhibiting a diameter of 90 Å (4). These dimensions are also comparable with the molecular size of an LrpA octamer (96 ϫ 96 ϫ 110 Å) (3). Besides, these globular structures are reminiscent of the structure of eukaryotic (H 3 -H 4 ) 2 nucleosomes that have a diameter of ϳ100 Å and of archaeal Hmf or HTz tetrasomes (24,25).
If the DNA molecule behaves as a worm-like chain, the energy required to bend the DNA is dependent on the persistence length, temperature, bending angle, and the length over which the bend is extended (15). To bend the Ss-LrpB operator with 260°over a length of 100 bp, 47 kJ/mol are required (taking a complex equilibration temperature of 37°C into account). Assuming a similar nucleoprotein complex conformation at higher temperatures, 54 kJ/mol would be required at a temperature of 80°C, which is the optimal growth temperature of S. solfataricus P2 (26). This energy would mainly be compensated by the formation of additional favorable Ss-LrpB-DNA interactions and especially by protein-protein interactions between the three distinct Ss-LrpB dimers. Only a few weak interactions would suffice to compensate for 40 -60 kJ/mol. Also, intrinsic bending of the operator region might reduce this energetic cost, which is the case for the Ss-lrpB operator. Indeed, fragments containing the three Boxes show an intrinsic bending with an average bending angle of 35°, as demonstrated by a circular permutation assay (9).
The formation of higher order nucleoprotein structures with the DNA wrapped around multiple protein molecules appears to be a general propensity of regulators belonging to the Lrp family. All cases reported so far can be divided in two classes. The first class consists of Lrp-like proteins that specifically recognize DNA sequences and bind these sites cooperatively (on the same face of the DNA helix). Each binding site undergoes deformation upon binding. In some cases, this will eventually lead to a structure in which the DNA is wrapped around multiple Lrp units. It might be speculated that each Lrp dimer establishes interactions with one binding site. This is the case for the binding of the global regulator Lrp from E. coli to the operator regions of certain regulon members (e.g. ilvIH; Ref. 27), for the abovementioned FL11 (4), and is also suggested for LrpA from P. furiosus based on its three-dimensional octameric structure (3). In contrast to FL11 and LrpA, Ss-LrpB is proposed to form an array of three instead of four dimers and interacts with three regularly spaced sites. Furthermore, it is striking how well conserved these binding sites are. In other cases, the Lrp-like regulators often recognize clusters of strongly degenerated binding sites (E. coli Lrp, Ref. 28; Ptr2 from Methanocaldococcus jannaschii, Ref. 29). The second class consists of Lrp-like proteins that seem to bind specific structures instead of an array of base-specific groups or that seem to bind nonspecifically. Examples are Smj12 from S. solfataricus (nonspecific binding; Ref. 6) and LrpC from Bacillus subtilis (structure-specific binding; Ref. 30), which both introduce positive supercoils by righthanded wrapping. Besides being a global regulator, E. coli Lrp is also suggested to be a less-specific DNA-wrapping protein (31). Often, these proteins are suggested to play a role in the global organization of the nucleosome.