Low-Resolution Molecular Models Reveal the Oligomeric State of the PPAR and the Conformational Organization of Its Domains in Solution

The peroxisome proliferator-activated receptors (PPARs) regulate genes involved in lipid and carbohydrate metabolism, and are targets of drugs approved for human use. Whereas the crystallographic structure of the complex of full length PPARγ and RXRα is known, structural alterations induced by heterodimer formation and DNA contacts are not well understood. Herein, we report a small-angle X-ray scattering analysis of the oligomeric state of hPPARγ alone and in the presence of retinoid X receptor (RXR). The results reveal that, in contrast with other studied nuclear receptors, which predominantly form dimers in solution, hPPARγ remains in the monomeric form by itself but forms heterodimers with hRXRα. The low-resolution models of hPPARγ/RXRα complexes predict significant changes in opening angle between heterodimerization partners (LBD) and extended and asymmetric shape of the dimer (LBD-DBD) as compared with X-ray structure of the full-length receptor bound to DNA. These differences between our SAXS models and the high-resolution crystallographic structure might suggest that there are different conformations of functional heterodimer complex in solution. Accordingly, hydrogen/deuterium exchange experiments reveal that the heterodimer binding to DNA promotes more compact and less solvent-accessible conformation of the receptor complex.


Introduction
Peroxisome proliferators activated receptors (PPARs) are members of the nuclear receptor (NR) family, acting as liganddependent transcription factors and modulating the activation of cognate genes. There are three different PPAR isotypes: PPARa, PPARb/d and PPARc, which exhibit considerable amino acid sequence conservation. PPARc plays a central role in the glucose regulation, lipid homeostasis and in the control of the energy balance. Because of this, it has been extensively studied as a molecular target in type II diabetes treatment [1]. PPARc also stimulates adipose tissue differentiation and functional maintenance [2] and has considerable anti-inflammatory activity [3].
PPARs, like other nuclear receptors, are modular proteins composed of several separable domains [4]. Their N-terminal region (A/B) harbors a ligand-independent activation function 1 (AF-1). The conserved C region corresponds to the DNA binding domain (DBD) and is responsible for sequence-specific DNA recognition. A highly structured E region, or ligand-binding domain (LBD), is responsible for ligand specificity and co-factors recruitment. Hinge or D region is located between C and E domains and is the target of functionally relevant post-translational modifications like phosphorylation and sumoylation [4] (Figure 1).
To understand the function of nuclear receptors at a molecular level, the structural features that mediate heterodimer formation, ligand binding, sequence-specific DNA recognition, and the molecular events underlying the switch from inactive to active receptors must be understood. PPARs activate target-gene transcription upon agonist binding. In this process, PPAR DBDs recognize and bind to specific DNA core motifs known as responsive elements (PPREs), which are direct repeats of two halfsites of the consensus sequence AGGTCA, spaced by one nucleotide. The PPREs are recognized by heterodimers of PPAR with RXR, whereas PPARs alone are unable to bind these DNA response elements [5]. Dimerization is a frequent process in DNA recognition of many eukaryotic transcription factor families [6] and is common within nuclear receptor superfamily, where functional DNA interactions frequently involve homodimers or heterodimers with RXR [7,8]. PPARc/RXR heterodimers specifically regulate transcription of genes involved in insulin action, adipocyte differentiation, lipid metabolism and inflammation [9].
The crystallographic structure of intact PPARc/RXRa heterodimer bound to DNA has recently become available [10]. Overall architectures of the DBD and LBD receptor domains are very similar to the crystallographic structures of the separate domains [11,12]. However, the full-length structure of nuclear receptor heterodimer bound to DNA PPRE made it possible to study the interactions between functional domains. The two receptors (PPARc and RXRa) are asymmetrically positioned, with PPARc and RXRa interactions mediated by well-known interfaces formed by the two LBDs [12] and DBDs. The structures also revealed a third heterodimerization interface between the PPARc LBD and the DBD and hinge region of RXRa. This interface seems to be modulated by the interactions with DNA, through positioning of both receptors in a unique polarity and spatial arrangement [10].
Although protein crystallography reveals detailed and precise information about tertiary structure of macromolecules, the proteins can adopt other functional conformations. For example, protein conformation is thought to be regulated by DNA contact and chromatin context. The overall shape of a macromolecule and/or its more dynamic quaternary structure in solution can be more reliably accessed by small-angle X-ray scattering (SAXS) [13]. This technique only provides low-resolution structural information relative to X-ray diffraction data, but can reveal overall structure and oligomeric states of native proteins in nearly physiological aqueous conditions, thus permitting analysis of structural changes in response to variations in experimental parameters.
More recently, SAXS and cryo-electron microscopy models of NRs heterodimers revealed alternative conformations for LBD and DBD positions in solution, indicating possible conformational differences in heterodimer arrangements. In addition, the cognate DNA sequences and coactivator presence in the heterodimer seem to result in a more open conformation of the complex. These different conformational states (more closed X-ray structure and more open solution models) might originate from the inherent NRs flexibility [14]. In other words, the NRs crystal structure may reveal only one of the multiple conformational states explored by the receptors.
In order to gather more information about NRs conformations and mobility, here we present systematic analysis of oligomeric state of hPPARc LBD and LBD-DBD constructs with and without heterodimerization partner hRXRa in the absence of cognate DNA. Furthermore, we also conducted analysis of PPAR solvent accessibility in its monomer and heterodimer forms (with and without DNA) using hydrogen/deuterium exchange (H/D-Ex) monitored by mass spectrometry, which provide additional information about the macromolecular interfaces and the mobility of the complex.

Characterization of hPPARc Monomers and hPPARc-hRXRa Heterodimers in Solution
We subjected purified preparations of hPPARc LBD, hRXRa LBD, hPPARc DBD-LBD and hRXRa DBD-LBD to size exclusion chromatography (SEC). The hPPARc (LBD and DBD-LBD) showed elution profiles with a single predominant peak (Figures 2A and 2B), corresponding to a hydrodynamic radius (R H ) of 28.6 Å and 35.3 Å , respectively, consistent with hPPARc LBD and DBD-LBD monomers (apparent molecular weight of approximately 30 kDa and 42 kDa, respectively [15]). After analytical gel filtration, the proteins were submitted to SDS-PAGE ( Figure S1), native electrophoresis ( Figure S2) and dynamic light scattering experiments ( Figure S3) confirming the previous . The N-terminal region (A/B) represented by a light gray bar is absent in the structure because of it high flexibility. The conserved C region, which corresponds to the DBD, is given in black; the LBD, or region E, is shown in gray; and located between C and E domains, the hinge given here in dark gray. doi:10.1371/journal.pone.0031852.g001 values found to R H and apparent molecular weight ( Table 1). The experimentally determined hPPARc DBD-LBD R H value is close to that of thyroid hormone receptor (TR) LBD-DBD monomers [16]. Aiming to examine the influence of the concentration on the R H values to hPPARc LBD, the protein, at different concentrations, was submitted to native gel electrophoresis and dynamic light scattering experiments. Both methods of analysis gave the same result, confirming that hPPARc LBD remains monomeric over a range of protein concentrations from 1 to 20 mg/mL ( Figure S2B).
Since the active form of hPPARc is a heterodimer with RXR [17], we performed similar studies with the hPPARc/hRXRa LBD and DBD-LBD heterodimers. The addition of RXR to PPAR (DBD-LBD and LBD) changed the SEC profiles to larger oligomeric forms (Figures 2A and 2B). In this case, complexes were eluted with R H of 39.0 Å and 47.8 Å , respectively, for LBD and LBD-DBD constructs. The experimentally determined R H for heterodimer are consistent with values found for hRXRa LBD and NGFI-B LBD dimers, which R H s are 36.0 Å and 38.5 Å , respectively [18]. The hPPARc/hRXRa DBD-LBD R H is also in agreement with R H s of other NR dimers, such as hTRb DBD-LBD and hRXRa DBD-LBD that are equal to 42.0 Å [16] and 44.0 Å [19]. Therefore, experimental R H values indicate that hPPARc LBDs and hPPARc DBD-LBDs readily form heterodimers with hRXRa. After analytical gel filtration hPPARc and hPPARc/hRXRa were submitted to SDS-PAGE and native electrophoresis to verify complex formation and the stoichiometry of the complexes ( Figure S1A and S2).
Small Angle X-ray Scattering Studies of hPPARc LBD and Its Heterodimerization with RXR The X-ray scattering curves obtained for hPPARc LBD and hPPARc/hRXRa LBD were practically identical at different concentrations, thus indicating the absence of spatial correlation effects over the applied concentration range (Table S1). Therefore, subsequent analysis steps were performed at 3 mg/mL for both hPPARc LBD and for hPPARc/hRXRa LBD ( Figure 3A and 3B). The Guinier plots gave radius of gyration (R g ) values, which were consistent with monomers hPPARc LBD and dimers hPPARc/hRXRa LBD complex ( Figures 3A and 3B, inset). Furthermore, these R g obtained by Guinier analysis showed a good correlation with R g obtained by the p(r) analysis (Figures 3C & 3D  and Table 2).
SAXS data are consistent with the results of SEC, native gel electrophoresis and dynamic light scattering analysis. The obtained structural parameters are also similar to the SAXS studies of NGFI-B LBD dimers (R g = 28.9 Å and a D max = 90.0 Å ) [20].
The three SAXS-based methodologies used to calculate the molecular weights, which included absolute scattering intensity using water and BSA as standards [21,22] and SAXS MoW web tool [23], consistently reveal monomers of hPPARc and heterodimers of hPPARc/hRXRa in solution ( Table 2). It is interesting to note, that the molecular weights predicted by SAXS MoW for the hPPARc/hRXRa LBD and hPPARc/hRXRa LBD-DBD heterodimers are somewhat overestimated. Since SAXS MoW algorithm is based on the assumption of the fixed protein density per volume occupied by the molecular envelope [23], this might be a consequence of conformational mobility of the heterodimers (see Discussion sections).
Ten independent ab initio simulations were performed with Gasbor package [24] without any symmetry restrictions and of those 252 dummy atoms were attributed to the final model of the monomer and 611 dummy atoms for the heterodimer (Figure 4). The dummy atoms models, PPARc LBD monomer and hPPARc/ hRXRa LBD heterodimer have a maximum diameter (D max ) of 65.0 Å 85.0 Å , respectively. The models generated for the protein monomer showed globular shape, as expected according to the   crystallographic structure for this domain, while the generated heterodimer model showed a more elongated shape, quite different from the former models. Overall, the computed scattering and p(r) curve based on the crystallographic structure of monomeric hPPARc LBD exhibited reasonable fit to the experimental scattering curve ( Figure 3 and Table 2). Our lowresolution hPPARc LBD DAM is also in a good agreement with the crystallographic structure of a single ligand-binding domain of hPPARc (Table 2 and Figure 4A). Conversely, SAXS scattering data for hPPARc/hRXRa LBD complex are not fully compatible with the simulated scattering data computed from the hPPARc/hRXRa LBD heterodimer crystallographic structure (PDB id 1FM6) [12] (Table 2). Essentially, the heterodimer model needs to be more open than the crystallographic structure to fit experimental SAXS data. Selecting and keeping the fundamental contacts to maintain the known heterodimer interface [12], we performed the rigid body adjustments of the crystallographic model based on our SAXS curves ( Table 2). The resulting rigid body model shows a more open heterodimer, with an opening angle between LBDs of about 47 degrees, whereas the opening angle of the crystallographic structure is close to 30 degrees ( Figure 4C). This means that the solution dimer interface is likely to be considerably smaller than that observed in the crystal structure (PDB id 1FM6). Numerically, the crystal structure heterodimer interface has an area of 1054.9 Å , while the interface of body rigid model generated has an area of 480 Å , according to ''Protein interfaces, surfaces and assemblies service'' (PISA) at European Bioinformatics Institute (http:// www.ebi.ac.uk/pdbe/prot_int/pistart.html) [25]. The rigid body adjustments of the hPPARc/hRXRa LBD resulted in a considerably better fit to SAXS experimental data ( Table 2).
The superposition of the high-resolution structure monomer hPPARc LBD with the ab initio DAM, performed with the program Supcomb, is shown in Figure 4A. The same approach was taken for the superposition of the heterodimer hPPARc/hRXRa LBD rigid body model with its ab initio The presence of DBD does not influence hPPARc oligomeric state SAXS studies of a PPARc construct consisting of both DBD and LBD (hPPARc DBD-LBD) were conducted to study how the DBD influences hPPARc oligomeric state. The X-ray scattering curves obtained for protein solutions at the different concentrations did not show any spatial correlation effects (Table S1). Typical scattering curves obtained for hPPARc DBD-LBD monomer and hPPARc/hRXRa DBD-LBD heterodimer are shown, respectively in Figures 5A and 5B. The structural parameters derived from these curves are given in Table 3. The R g values are approximately 30.0 Å and 35.0 Å for hPPARc DBD-LBD monomer and hPPARc/hRXRa DBD-LBD heterodimer, respectively. These values are compatible with the estimates obtained from the Guinier analysis (Table 3; Figures 5A and 5B, inset), and they are consistent with expected for respective monomers for hPPARc and heterodimers for their complexes with hRXRa. Moreover, they can be confirmed by the curve obtained on the basis of distances distributions (p(r)) ( Figure 5C and 5D).
The particle shapes (DAMs), computed using Dammin package [26], reveals that one of the molecular envelopes is consistent with monomeric protein (this is the case for hPPARc DBD-LBD) and another one with the heterodimer (hPPARc DBD-LBD in the presence of hRXRa DBD-LBD). The molecular DAM for hPPARc DBD-LBD monomer has a packing radius of about r a = 2.8 Å , with a maximum diameter D max = 110.6 Å , whereas molecular envelope for the hPPARc-hRXRa DBD-LBD heterodimers has a packing radius r a = 3.3 Å , with a maximum diameter D max = 129.1 Å , respectively ( Figure 6). The experimental SAXS curves and scattering curves computed from the DAMs show good agreement (Table 3). Molecular weights computations using three different methods based on SAXS analysis also confirmed the oligomeric states of hPPARc DBD-LBD and hPPARc/ hRXRa DBD-LBD as being monomer and dimer, respectively (Table 3).

Dummy Atom Model Reveals More Open Conformation of hPPARc/hRXRa in Solution as Compared to High
Resolution X-ray Structure of the Complex DAM generated by Dammin package for hPPARc/hRXRa DBD-LBD is prolate, elongated and has an asymmetric form. This asymmetry was partly expected based on the arrangement of the domains in the crystallographic heterodimer formed by hPPARc and hRXRa, which is non-symmetric, allowing several contacts of hPPARc LBD with other domains of both proteins of the complex with LBD and DBD of hPPARc closely positioned, and hRXRa LBD and DBD far apart with the space between them filled by the hPPARc LBD [10].
To compare our low resolution SAXS data with the crystal structure, we computed the theoretical SAXS curves and the pairdistance distribution function for the crystal structures of hPPARc DBD-LBD monomer and hPPARc/hRXRa DBD-LBD heterodimer ( Figure 5). The crystallographic models do not fit well to the DAMs derived from our SAXS experiments. The profiles of the distance distribution functions p(r) corresponding to DAMs and generated for crystallography structures are typical for elongated particles. Nevertheless, the D max of the DAMs are larger than the crystallographic structure, which indicates that the protein in solution is more elongated than in the crystal.
Rigid body models were generated to minimize discrepancy between crystallographic and experimental models. For the PPAR LBD-DBD monomer rigid body model, the hinge was maintained and the protein domains were separated into two rigid bodies.
Discrepancy between our SAXS data and crystallographic model for the heterodimer could stem from the absence of DNA in our samples and/or from the fact that the SAXS measurements were performed in solution, conditions under which the protein did not have restrictions imposed by the crystalline environment. Thus, the rigid body model generated with the Sasref package [27] was introduced to improve the quality of the fits of experimental SAXS curves to the generated model. This was done by separating their relative domain positions and orientations determined to minimize the differences between the experimental data and the model predictions. The hinge was excluded from computations since the resolution of the SAXS model is not sufficient to define its position and conformations. As mentioned in the Introduction, there was an unexpected intramolecular interface in the crystallographic structure of intact hPPARc/hRXRa complex [10], which allows interaction of the DBD of hPPARc with the hinge of hRXRa. In our rigid body model, this interaction could not be observed. This interaction was also not observed in the SAXS experiments performed by another group that studied the envelopes of this complex in the presence of DNA (hPPARc/ hRXRa DBD-LBD+DR-1) [14]. The rigid body model obtained in these studies reflects distant and dissociated positions of DNA and ligand binding domains. This contrasts with the crystal structure [10] which shows a compact conformation of full-length nuclear receptors complex, but it is very consistent with our SAXS measurements, providing envelopes of the same complex but in the absence of DNA. Our solution SAXS measurements performed with the complex at the absence of DNA reveals that: 1) hPPARc DBD-LBD forms heterodimers with hRXRa DBD-LBD; 2) the heterodimer is asymmetric; and 3) it has a more extended and elongated shape induced by further separation of hPPARc/hRXRa LBD and DBD. These structural differences can be observed in the p(r) function ( Figure 5D), for which the value of D max for the SAXS model exceeds the value of the crystallographic structure, ensuring a less globular form of the sample in solution. As a result of the rigid body modeling, the two LBDs were positioned in the most bulky part of envelope and the DBDs were positioned in an asymmetric way along the envelope ( Figure 6C). This model predicts that the third dimerization interface created by LBD of PPAR and DBD of RXR will not be maintained. Additionally, there are marked differences in the spacing of DBDs and LBDs in the crystallographic structure and the generated rigid body model, which reveals the widely separated domains ( Figure 6C). To comply with the SAXS data, the DBD of the PPAR and RXR were translated from initial model (PDB id 3DZU), respectively, on 46.9 Å and 47.6 Å . Our rigid body model describes the smallangle X-ray scattering curves well, and has highly improved the fitting as compared to the X-ray crystallographic structure of the complex ( Figure 5 and Table 3).

Structural Dynamics and Molecular Interfaces of PPAR, PPAR/RXR and PPAR/RXR+DR-1 as Analyzed by Mass Spectrometry
The dynamic behavior and the interface-protected regions of PPAR/RXR heterodimer in solution were analyzed by hydrogendeuterium exchange experiments analyzed by mass-spectrometry (H/D-Ex MS). In H/D-Ex MS of hPPARc/hRXRa complex, we identified 51 peptides for hPPARc, covering 92% of its amino acid sequence ( Figure S4A). The deuterium uptake rate was higher for hPPARc alone, intermediate for the hPPARc/hRXRa heterodimer and very low for hPPARc/hRXRa+DNA complex ( Figure  S4B). Specifically, the uptake rates were 30%, 22% and 10% of D 2 0 incorporation, respectively. The differences in deuterium uptake between the preparations reflect increased compactness and lower flexibility of the more structured complexes with cognate DNA and/or heterodimerization partner, in comparison to the hPPARc alone. In addition, through measures of different deuterium incubation times, the kinetics of deuterium incorporation seems to be fast, since 15 and 30 minutes of incubation experienced no expressive variation ( Figure S4B).
The differences in the D 2 O uptake behavior of hPPARc, hPPARc/hRXRa and hPPARc/hRXRa with cognate DNA response element (DR-1) were observed, and as expected, the deuterium incorporation profiles for hPPARc monomer show that it is more flexible and solvent-expose than the other complexes ( Figure 7). The DBD are subject to a high degree of H/D  123-143). Surprisingly, the hinge domain appears more protected than expected. Perhaps, it might be because of its position close to the receptor's body, as revealed by our SAXS model ( Figure 8A). The LBD is by far the most structured and rigid domain, showing low overall H/D exchange (deuterium incorporation below 40%) and the main core (H1, H3, H5, H6 and H9) very well protected.
Overall, the hPPARc/hRXRa heterodimer is more protected than hPPARc alone ( Figure 8B). The DBD protections show the footprint of DBD dimerization interface (between H9 and H11), which is in accordance with direct repeat array. The hinge is more flexible, disordered or exposed to the solvent, when compared to the hPPARc alone, suggestive of local protein unfolding, which could be necessary for interaction between the domains of the complexes. The main differences between the hPPARc/hRXRa heterodimer and hPPARc monomer are located in the LBD. The LBD core (H1, H3, H5, H6) becomes more structured and compact, with many protected areas. The dimerization interface has a medium level of deuterium incorporation (31% to 50%), H7 is strongly protected and H11 is more accessible, indicating asymmetry of this interface, in compliance with our hPPARc/ hRXRa LBD SAXS model (Figure 7 and 8). Coupled with SAXS analysis which suggests that hRXRa DBD and hPPARc LBD are far apart and unable to form the interface, this finding represents further evidence that the third interface does not form in solution and in the absence of DNA, and the hPPARc/hRXRa heterodimer adopts an intermediate state, more compact than that found in the separate proteins, but less packed together than the crystallographic complex (hPPARc/hRXRa+DR-1 DNA element).
The presence of DNA induces an even more solvent protected conformation in the heterodimer hPPARc/hRXRa ( Figure 8C), which is more consistent with the crystallographic structure (PDB id 3DZU). The DBDs and hinges of both subunits of the heterodimers become more protected. Further example is the hinge region (residues 154-195), which display lower mobility, presumably because of its possible interactions with DNA.
The LBD protein core is also significantly more protected from solvent. Significantly lower dynamic exchange as compared to other samples (less than 10% of D 2 O incorporation) was observed for the H9 and H11 (mainly responsible for dimerization interface). This might indicate that the interface becomes larger and more symmetric. Consequently, the hPPARc/hRXRa heterodimer bound to DNA seems to be more compact and further stabilized by the DNA addition.
In addition to the protected regions belonging to the domains core, the region comprising loops formed by residues 110-120 of DBD and the 378-385 and 422-431 of LBD showed higher protection to H/D exchange. Analyses of the crystallographic structure the hPPARc/hRXRa complex reveals that these parts of the structure become more internalized in the presence of DNA. This does not happen in the absence of DNA, because of the extended conformation of the heterodimer. These observations are in agreement with the hypothesis that the presence of the DNA will trigger the rearrangements of the hPPARc/hRXRa dimer conformation toward a more compact state.
The third heterodimerization interface also displays stronger protection as compared to a complex without DNA, as can be seen for H7, for example, as well as some parts of the LBD surrounding this interface (H6 and H3). Together, these findings suggest that there is an increase in overall compactness and reorganization in the protein complex when the DNA is added. Furthermore, the dimerization interfaces of hPPARc/hRXRa heterodimer in solution, is different from the interfaces of hPPARc/ hRXRa+DR-1 complex in the crystalline state.

Discussion
PPARc has a central role in the regulation of glucose and lipid homeostasis and is involved in inflammatory processes and is an important drug target for treatment of Type 2 Diabetes and Table 3. Structural Parameters Derived from SAXS for hPPARc DBD-LBD (monomer) and hPPARc/hRXRa DBD-LBD (heterodimer).   Although PPARs associate with RXR in the presence of ligand in living cells [30][31][32], its oligomeric state in solution had not been explored. Our SAXS-derived structural parameters, supported by SEC, native electrophoresis and DLS are consistent with the monomeric form of both, hPPARc LBD and DBD-LBD, constructs in solution, even at high protein concentrations required for SAXS experiments. This is highly unusual since other nuclear receptors, studied to date in solution by SAXS and other techniques, form dimers and higher oligomeric forms [18][19][20]33]. Nevertheless, our SAXS experiments reveal that in the presence of hRXRa, both hPPARc LBD and DBD-LBD protein constructs readily form heterodimers. This suggests that our hPPARc preparations are comprised of functional protein, which retains the capacity to heterodimerize with RXR and to bind to DNA, essential steps in eliciting its functional activity, and confirms that hPPARc is a constitutive monomer with a high capacity for heterodimerization.
Fitting of high-resolution X-ray structural models into our lowresolution SAXS models revealed unexpected differences between organization of the heterodimer in the crystal and in solution. The SAXS-based rigid body model constructed for LBDs render the hPPARc/hRXRa heterodimer considerably about 17 degree more open relative to high-resolution hPPARc/hRXRa LBD crystallographic structure [12]. In addition, our SAXS experiments performed on hPPARc/hRXRa DBD-LBD complex reveals that this heterodimer becomes asymmetric and adopts a more extended and elongated shape as compared to the conformation found in the crystal structure [10], which are in agreement with SAXS envelopes of these proteins in complex with DNA [14]. This elongated form of the hPPARc/hRXRa DBD-LBD heterodimer in solution is induced by further separation of hPPARc LBD and DBD with respect to one another.
Our H/D-Ex experiments also revealed differences in hPPARc, hPPARc/hRXRa and hPPARc/hRXRa+DR-1 species in terms of solvent accessibility. Our results indicate that hPPARc/hRXRa heterodimer alone, in the absence of DNA, is an intermediately condensed form, which is stabilized by the cognate DNA binding. Essentially, the asymmetric dimerization interface between hPPARc/hRXRa LBDs, became more protected after DR1 binding. Finally, our data predicts that the third dimerization interface, between DBD of hRXRa and LBD of hPPARc, could be formed only in the presence of DR1 DNA, as predicted from analysis hPPARc/hRXRa+DR1 tridimensional structure [10], but the open and close conformation of the complex remain in a dynamic equilibrium.
Our results shed more light on the functionally relevant heterodimer hPPARc/hRXRa formation and the hPPARc behavior in solution. Based on our studies, we purpose a following model of PPAR activation (Figure 9). According to this model, ligand-bound PPAR recruits RXR and forms an intermediary heterodimer, more stable then the PPAR alone, but with LBD heterodimer surfaces relatively open as compared to the crystallographic model. The DBDs show extended conformations, separated from the LBDs, as revealed by our SAXS model. After DNA binding, this intermediary heterodimer undergoes additional conformational changes, caused by the interactions between the receptors and DNA, and becomes more compact, able to adopt the conformation similar to the one revealed by the crystallographic structure [10]. Thus, our data confirms that DNA could induce significant changes in the interactions of DBDs, LBDs and hinge organization of the PPARc/RXRa complex, consistent with predictions that DNA acts as an allosteric ligand, inducing widespread reorganizations in receptor conformation [34,35]. Furthermore, the mass experiments showed changes in the deuterium incorporation pattern, after the DNA addition. It will be interesting to understand how these structural alterations may affect PPARc function on DNA elements versus their actions at alternate elements where direct DNA interaction is not required [36].
The same protocol of protein expression and purification was used for all protein studied in this work. Protein expressions were conducted in LB culture and were induced with 1 mM IPTG (isopropyl b-D-1-thiogalactopyranoside), under incubation at 20uC for 3 h. 5 mM of zinc sulfate was added to the culture during expression of the constructs with DBD domains. Cells were collected by centrifugation and the pellets were resuspended in 50 mM sodium phosphate, pH 7.5, 300 mM NaCl, 10% glycerol, 2 mM 2-mercaptoethanol, and 10 mM imidazole (buffer A). Phenylmethylsulfonyl fluoride (PMSF) and lysozyme were present at 10 mM and 250 mg/mL, respectively. The lysate was sonicated, clarified by centrifugation and the supernatant loaded onto a Talon Superflow Metal Affinity Resin (BD Biosciences Clontech, Palo Alto, CA) pre-equilibrated in Buffer A. The bound hPPARc was eluted with 50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 10% glycerol, 2 mM b-mercaptoethanol, and 300 mM imidazole (buffer B), in a single step. The eluted pool was collected, and the His-tag was subsequently removed (except to hRXRa) by incubation with thrombin at 10 U/mg for 12 h at 18uC. After, as an additional purification step, hPPARc LBD was loaded into the gel filtration HL Superdex 75 26/60 column and hRXRa LBD and DBD-LBD, and hPPARc DBD-LBD, into HL Superdex 200 16/60 column (GE Healthcare) equilibrated with 20 mM Hepes-Na buffer (pH 8.0), 3 mM dithiothreitol, 200 mM NaCl, and 5% glycerol.
Protein content and purity were confirmed by coomassie bluestained sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Protein concentrations were determined using the Bradford dye assay (Bio-Rad, Hercules, CA).

Heterodimer Preparation
The purified protein pairs: hPPARc DBD-LBD and hRXRa DBD-LBD, or hPPARc LBD and hRXRa LBD at a concentration of 20 mg/mL each, were incubated in a molar proportion of 1:1 for 1 h at 4uC. After, each complex was purified by loading Figure 9. Cartoon schematically representing the mechanism of heterodimerization and binding to the DNA. When the PPAR is activated, it recruits RXR, forming an intermediary heterodimer, which has the LBDs and DBDs domains in extended and open conformation. Following to DNA binding, the PPAR/RXR heterodimers suffer additional conformational changes, becoming more condensed and less solventexposed. doi:10.1371/journal.pone.0031852.g009 Figure 8. PPAR DBD-LBD models colored according to H/D Ex-data. Protections and solvent exposure are colored according deuteration level, from blue (0-10% D 2 O incorporation), green (11-30% D 2 O incorporation), yellow (31-50% D 2 O incorporation) to red (more that 50% of D 2 0 incorporation). A) hPPARc monomer; B) hPPARc/hRXRa heterodimer. The box shows in details the dimerization interface, with H10-11 being not very strongly protected (yellow -11 to 49% D 2 O incorporation). C) hPPARc/hRXRa+DR1 complex, the boxes show dimerization interface (top box, framed in black), which presents H10-11 and H7 more protected than that in hPPARc/hRXRa heterodimer alone; and the third heterodimerization interface (bottom box -orange) indicating higher degree of protection. doi:10.1371/journal.pone.0031852.g008 onto a Superdex 75 HR 10/30 (GE Healthcare) for LBD constructions, and Superdex 200 HR 10/30 (GE Healthcare) for DBD-LBD constructs. The size exclusion chromatography (SEC) was also used to evaluate the oligomeric species present in solution (Text S1). The column was standardized with the gel filtration calibration kit (GE Healthcare), thyroglobulin, ferritin, catalase, aldolase, albumin, ovoalbumin, chymotrypsinogen, and ribonuclease A (hydrodynamic radii (R H ) of 85.0, 61.0, 52.2, 48. 1, 35.5, 30.5, 20.9, and 16.4 Å , respectively), utilized as calibration standards. The elution volumes of these proteins were used to calculate the K av values according to columns calibration as described [16]. All the eluted samples were checked by SDS-PAGE 15%. Other methodologies were applied to assist in the oligomeric states evaluation (Text S2 and S3).
Small-Angle X-ray Scattering SAXS experiments. SAXS data for hPPARc LBD and hPPARc/hRXRa LBD complex at 1, 3 and 6 mg/mL, as well as, hPPARc DBD-LBD and hPPARc/hRXRa DBD-LBD at 1, 3 and 6 mg/mL, were performed at the D02A-SAXS2 beamline of the Synchrotron Light National Laboratory (Campinas, Brazil) (Text S4). Measurements were done with a monochromatic X-ray beam with a wavelength of l = 1.488 Å and the X-ray patterns were recorded using a two-dimensional CCD detector (MarResearch, USA). The sample-to-detector distance was set at 955.3 mm, resulting in a scattering vector range of 0.015 to 0.35 Å 21 , where q is the magnitude of the q-vector defined by q = 4psinh/l (2h is the scattering angle). The samples diluted in a gel filtration buffer were centrifuged at 23,500 g for 30 minutes, at 4uC to remove any aggregates or particles and then placed on ice. For SAXS measurements, protein samples were introduced into a 1 mm path length cell with mica windows at 20uC. Two successive frames of 300 s each were recorded for each sample to monitor radiation damage and beam stability. Buffer scattering was recorded before the sample scattering. The SAXS patterns were individually corrected for the detector response and scaled by the incident beam intensity and the sample absorption. The buffer scattering (parasitic scattering from windows, narrows, etc.) was subtracted from the corresponding sample scattering. The integration of SAXS patterns were performed using Fit2D software [37], and the curves were scaled by protein concentration.
SAXS data analysis. The radius of gyration, R g is a global measure of the size and shape of the molecular complex which is related to hydrodynamic radius (R H ) by R H = R g 61.3 [38] and was approximated using two independent procedures, by Guinier equation [39] and by indirect Fourier transform method using Gnom package [40]. The distance distribution functions p(r) also was evaluated by Gnom and the maximum diameter, D max was obtained. Molecular weights (MW) were estimated by three methods: (1) by determining the absolute scattering intensity using water scattering (primary standard) (Text S5) [21], (2) by comparison of the forward-scattered intensity with the secondary protein standard, bovine serum albumin (BSA) (Text S6) [22] and (3) using a novel procedure implemented as a web tool SAXS MoW (www.ifsc.usp.br/,saxs/saxsmow.html) [23]. The later procedure does not require the measurement of SAXS intensity on an absolute scale and does not involve a comparison with another SAXS curve determined from a known standard protein.
To calculate the forward scattering I(0) in the absolute scale, the known scattering of water equal to 1.632610 22 cm 21 at 288 K was used [21].
SAXS ab initio modeling. Dummy atom models (DAMs) were calculated from the experimental SAXS data using ab initio procedure implemented in either Dammin [26] and Gasbor packages [24]. Several runs of ab initio shape determination with different starting conditions led to consistent results as judged by the structural similarity of the output models, yielding nearly identical scattering patterns and fitting statistics in a stable and selfconsistent process. Crysol package was used to generate the simulated scattering curves from DAMs [40]. The evaluation of R g and D max , were performed with the same package.
Fitting of DAMs with crystallographic structures. The crystallographic structures of hPPARc LBD monomer (PPAR monomer part from the PDB id 1FM6) [12], hPPARc/hRXRa LBD complex (PDB id 1FM6), hPPARc DBD-LBD monomer (PPAR monomer part from the PDB id 3DZU) and hPPARc/ hRXRa DBD-LBD complex (PDB id 3DZU) [10] were used to generate the simulated scattering curves by Crysol package [40] and to determine the R g and D max . Some of the simulated curves based on the crystallographic structures had good agreement with the experimental SAXS data. The correspondent threedimensional structures were superimposed with ab initio DAMs using the Supcomb package [24]. Figures of the superpositions were generated by the program PyMOL [41].
Rigid body modeling. Rigid body modeling was performed for the hPPARc/hRXRa LBD complex using Sasref package [27]. The two monomers from the crystallographic structure (PDB id 1FM6) were separated and their relative position and orientation were minimized. Based on the known classic dimerization interface between hPPARc LBD and hRXRa LBD, the intermolecular contacts RXR F415-A433 PPAR and RXR L420-L436 PPAR [12] were maintained during the minimization procedure. In order to improve the quality of fits, the protein domains were allowed to separate and thus their relative positions and orientations were determined by rigid body modeling. For the PPAR LBD-DBD monomer rigid body model, the hinge was maintained and the protein was separated into two rigid bodies maintaining the primary sequence of amino acid residues P206-E207. The rigid body refinement allowed better adjustment of the structure inside the DAM. To perform rigid body modeling with the heterodimer DBD-LBD, we separated the complex into two rigid bodies, one containing the LBDs and the other with the DBDs, since limited structural information of SAXS data did not allowed us to use too many independent domains and more degrees of freedom. The hinge domains (for PPAR, a fragment between A172-P206 and for RXR, the fragment between E203-N227) have been excised from the structural templates. The dimerization interface of LBD was maintained, as it had been described previously for structure with separate LDB domains [12] and also observed in the structure of the fulllength receptor [10]. For position of DBDs, we used the dimerization interface described for the DBDs of the estrogen receptor (ER) (PDB 1HCQ), which shows a complementarity of shape as well as a number of direct contacts between domains [42]. Crysol package was used to generate the simulated scattering curves.
The H/D exchange mass spectrometry experiments started by diluting hPPARc/hRXRa complexes at high concentration (26 mg/mL) 10 times in D 2 O buffer (final buffer: 2 mM Hepes, 15 mM NaCl, 0.5% Glycerol, 0.3 mM DTT and 60% v/v D 2 O). These mixtures (100 mL) were incubated for 3, 10 and 30 minutes at a room temperature, in order to have mild conditions of hydrogens exchange by deuteriums at the protein surfaces. The kinetics of deuterium incorporation is fast and after 30 minutes of incubation, the H/D exchange becomes stabilized (Text S7). After incubation, the proteins were immediately submitted to pepsin cleavage at a ratio of 1:50 enzime to protein by mass, for 10 minutes, with addition of 60 mL of 100 mM of sodium phosphate buffer pH 2.5, on ice to avoid H/D back-exchange. After addition of 30% acetonitrile, the samples contained the peptic fragments was immediately applied, to avoid back-exchange with solvent hydrogen, by direct injection onto a Quattro II triplequadrupole mass spectrometer (Micromass, UK), equipped with a standard ESI source. By the analysis of the displacement in peptide peaks, the fragments of the protein undergoing H/D exchange were identified. The software MS-Digest (The Regents of the University of California) was used to identify the sequence of the peptic peptide ions, generated by pepsin cleavage. The deuterium incorporation level for each peptide was determined from differences in mass centroids between the deuterated and nondeuterated fragments using Masslinx software (Micromass, UK).