Stability of the Retinoid X Receptor-α Homodimer in the Presence and Absence of Rexinoid and Coactivator Peptide

Abstract

Differential scanning calorimetry and differential scanning fluorimetry were used to measure the thermal stability of human retinoid X receptor-α ligand binding domain (RXRα LBD) homodimer in the absence or presence of rexinoid and coactivator peptide, GRIP-1. The apo-RXRα LBD homodimer displayed a single thermal unfolding transition with a T_m of 58.7 °C and an unfolding enthalpy (ΔH) of 673 kJ/mol (12.5 J/g), much lower than average value (35 J/g) of small globular proteins. Using a heat capacity change (ΔC_p) of 15 kJ/(mol K) determined by measurements at different pH values, the free energy of unfolding (ΔG) of the native state was 33 kJ/mol at 37 °C. Rexinoid binding to the apo-homodimer increased T_m by 5 to 9 °C and increased the ΔG of the native homodimer by 12 to 20 kJ/mol at 37 °C, consistent with the nanomolar dissociation constant (K_d) of the rexinoids. GRIP-1 binding to holo-homodimers containing rexinoid resulted in additional increases in ΔG of 14 kJ/mol, a value that was the same for all three rexinoids. Binding of rexinoid and GRIP-1 resulted in a combined 50% increase in unfolding enthalpy, consistent with reduced structural fluidity and more compact folding observed in other published structural studies. The complexes of UAB110 and UAB111 are each more stable than the UAB30 complex by 8 kJ/mol due to enhanced hydrophobic interactions in the binding pocket because of their larger end groups. This increase in thermodynamic stability positively correlates with their improved RXR activation potency. Thermodynamic measurements are thus valuable in predicting agonist potency.

Graphical Abstract

Human retinoid X receptors (RXRα, RXRβ, and RXRγ) play a key role in many signaling processes which control cellular proliferation, differentiation, and growth in epithelial tissue as well as maintain proper lipid and glucose homeostasis.¹ The three RXR isotypes each display different functions in modulating gene transcription of target genes.² The proteins are approximately 51 kDa and contain two folded domains and three flexible, largely unstructured regions. The DNA binding domain (DBD) binds to specific DNA sequences at the start of target genes³ using two Zn-finger motifs. The ligand binding domain (LBD) at the c-terminal end of RXR is responsible for ligand recognition as well as the associated conformational changes that promote recruitment of coactivator proteins (with replacement of corepressor proteins) important for stabilizing RNA polymerase II to start transcription. The LBD of RXR is connected to the DBD by a short unstructured linker region largely separating the two functions of RXR: specific recognition of target genes and ligand-induced activation of the target gene. Both DBD and LBD contribute to the dimerization interface of the receptor.

The X-ray crystal structures of RXR LBD homodimers alone (apo-RXR LBD) or bound to a variety of ligand agonists (holo-RXR LBD) and coactivator peptides revealed the significant conformational changes that allow for agonist-induced transcriptional activation.^4–14 In apo-RXRα LBD, helix 12 extends away from the core of the domain (Figure 1D) whereas, in the complexes, helix 12 reorients and folds back toward the ligand binding pocket (LBP). Molecular interactions, which occur between the LXXLL¹⁵ motif of the coactivator peptide and helices 3, 4, and 12, stabilize the active conformation only observed in the presence of agonists (Figure 1D). The tertiary structural changes induced by coactivator peptide binding also extend to helix 11 residues that interact directly with the agonists. Hydrogen–deuterium exchange mass spectrometry (HDX-MS) and fluorescence polarization studies strongly support that helix 12 is dynamic in both apo- and holo-homodimers and only stabilized when both the coactivator peptide and the agonist are present.^8,9,16–18 Recent X-ray and HDX-MS studies^19–23 of nuclear receptor heterodimers containing the entire RXR bound to double stranded DNA with a target recognition site support the validity of those structural and dynamical conclusions drawn from studies on RXRα LBD homodimers.

Figure 1.

Structures of RXRα LBD, rexinoids, and coactivator peptide. (A) Crystal structure of RXRα LBD homodimer in complex with UAB30 and GRIP-1 (PDB ID: 4K4J). Secondary elements of each dimer are colored according to their positions in the three-layer helical sandwich: helices 1 and 3 are blue; helices 4, 5, 8, and 9 and the β-sheets orange; and helices 6, 7, 10, and 11 teal. Helices 12, which change conformation upon GRIP-1 binding, are red. GRIP-1 peptides are pink. UAB30 rexinoids are displayed as green sticks. (B) Chemical structures of UAB30, UAB110, and UAB111. (C) Peptide sequence of GRIP-1 NR-Box II coactivator peptide. (D) Conformations of helices 3 (blue), 11 (teal), and 12 (red) in the complex, in comparison with their positions in apo-RXRα LBD (gray, PDB ID: 6HN6). (E) Conformations of UAB30 (green), UAB110 (orange), and UAB111 (deep pink) in the ligand binding pocket of RXRα LBD. The rexinoids are displayed as sticks. The amino acid residues lining the pocket are displayed as thin lines.

Our group has designed and studied many rexinoids for the use in cancer treatment and prevention.^{13,14,24–27} UAB30 (Figure 1B) is a novel rexinoid developed by our group currently in phase II clinical trials. Compared to Targretin, the only clinically used rexinoid approved by the FDA,²⁸ UAB30 exhibits high efficacy in cancer prevention without inducing hyperlipidemia which is the dose-limiting toxicity of Targretin.^25,29–32 UAB110 and UAB111 (Figure 1B) are two rexinoids belonging to another structural class of UAB rexinoids other than UAB30. Each of these three UAB rexinoids bind the RXRα LBD with nanomolar affinities. UAB110 and UAB111 exhibit potencies approximately 40-fold higher than UAB30 in RXR transcriptional activation.¹⁴ However, UAB111 induces hyperlipidemia at its effective dose, while UAB110 does not increase triglyceride levels.

To design more potent and nontoxic rexinoids with greater clinical potential and selectivity, we need to fully understand the interactions between the rexinoids and RXRα LBD, and their effects on the structure and dynamics of the RXR homodimers and heterodimers with other nuclear receptors. X-ray crystal structures of RXRα LBD in complex with each UAB rexinoid and a coactivator peptide, GRIP-1 (e.g., RXRα LBD:UAB30:GRIP-1 complex in Figure 1), revealed that only subtle structural differences occurred in the LBP, but there was essentially no difference in the overall backbone fold of the homodimer (Figure 1E).^8,9,13,14,33 Using isothermal titration calorimetry (ITC), we showed that GRIP-1 binds to holo-RXR LBD homodimers in complex with each UAB rexinoid with nearly identical affinity and similar enthalpic and entropic signatures.^8,9,13,14 The binding thermodynamics of rexinoids to RXR LBD in the absence of coactivator peptide have not been determined, because of limitations to the applicability of ITC in such systems of high-affinity, hydrophobic ligands with small binding enthalpy changes.^34,35 Differential scanning calorimetry (DSC) provides an alternative method to obtain thermodynamic data on ligand binding. From the differences in the unfolding temperature and enthalpy with and without a ligand, the thermodynamic parameters of binding of the ligand are determined.^36,37

Using DSC, we conducted in this study a complete thermodynamic analysis of the thermal unfolding of human RXRα LBD homodimer in the absence or presence of rexinoid and the coactivator peptide, GRIP-1. The goal was to understand how rexinoids and coactivator peptides modulate the energetics of this protein domain, and to gain knowledge on the thermodynamic parameters of interaction between rexinoids and apo-RXRα LBD. Differential scanning fluorimetry (DSF), a complementary thermal unfolding method to DSC, was used to inform on key aspects of the unfolding of the domain and to rapidly scan unfolding conditions. We obtained the full panel of unfolding thermodynamic parameters on apo-RXRα LBD, including unfolding temperature (T_m), enthalpy (ΔH), entropy (ΔS), and heat capacity change (ΔC_p^u). These parameters are essential for analyzing binding energetics of agonists and coactivators to RXR but have never been reported before. From the DSC data obtained in the presence of UAB rexinoids and/or GRIP-1, we calculated the thermodynamic parameters of UAB rexinoids binding to apo-RXRα LBD homodimer, which clearly demonstrated for the first time that rexinoid binding was an entropically driven process and correlated well with rexinoid potency as agonists. We also calculated thermodynamic parameters of GRIP-1 binding to holo-RXRα LBD homodimers containing the rexinoids. They corresponded closely to those determined by ITC, which demonstrated the validity of this approach. Thermal unfolding of RXRα LBD homodimer was not completely reversible, and we provided a general guideline on how to deal with this complication to gather meaningful thermodynamic values of ligand binding.

MATERIALS AND METHODS

Protein Purification.

Overexpression and purification of human RXRα LBD (amino acid residues T223–T462) were performed as described by Xia et al.⁹ Briefly, the His₆-tagged RXRα LBD fusion protein was purified using a HiTrap Ni-chelating column (GE Healthcare, Piscataway, NJ), and the His₆-tag was removed by incubating with human α-thrombin overnight. The complete removal of His₆-tag was confirmed by MALDI-TOF mass spectrometry. The tagless RXRα LBD was further purified on a HiLoad Superdex 75 size exclusion column (SEC, GE Healthcare, Piscataway, NJ). The fractions containing RXRα LBD homodimers or tetramers (Figure S1A) were pooled separately. SDS-PAGE and MALDI were used to establish a purity of >95% and a mass of the monomers at 27 283 Da.

Fluorescence-Based Binding Affinity Assay.

The dissociation constant (K_d) of the rexinoids binding to apo-RXRα LBD was determined in a fluorescence-based binding assay as previously described.²⁴ The fluorescence quenching data were plotted in Origin 7 (OriginLab, LLC.), and nonlinear least-squares curve fitting was performed to obtain an apparent K_d based on the following equation:³⁸

$$\%\text{quenching}=100(1-Z)(2R_{\text{t}})/\{(N_{\text{t}}+R_{\text{t}}+K_{\text{d}})+{[{(N_{\text{t}}+R_{\text{t}}+K_{\text{d}})}^2-4N_{\text{t}}{R}_{\text{t}}]}^{1/2}\}$$ (1)

where R_t was the total concentration of the test rexinoid, N_t was the total concentration of apo-RXRα LBD in monomer unit, and Z represented the residual unquenched % fluorescence after the binding site was completely saturated. N_t was determined from the binding curve of UAB111 which had the highest binding affinity. This value was fixed for fitting of UAB30 and UAB110 binding curves.

Differential Scanning Calorimetry.

Calorimetry experiments were performed using the VP-Capillary DSC system (Malvern Instruments, Westborough, MA) in 0.130 mL cells at a heating rate that varied between of 0.5 and 4 °C/min. An external pressure of 2.0 atm was maintained to prevent possible degassing of the solutions upon heating. Purified apo-RXRα LBD homodimer freshly eluted from SEC was dialyzed in 2 L of DSC buffer (10 mM sodium phosphate, pH 7.0, 50 mM NaCl, 0.5 mM EDTA, and 1 mM TCEP; sodium phosphate was replaced by sodium borate for pH 8.5–9.5) for >4 h and used within 24 h. Holo-RXRα LBD homodimers were prepared by adding rexinoids from a 100-fold concentrated stock in DMSO into apo-RXRα LBD homodimer. The samples were incubated at 22 °C for 10 min and protected from light. GRIP-1 was added into the holo-RXRα LBD homodimers from a 10 mM stock solution in water. To prepare the buffer control, identical amounts of rexinoid and GRIP-1 were added into the dialysis buffer. DSC data analyses were performed using the built-in analysis module in Origin 7 provided by the DSC manufacturer.³⁹ Briefly, after subtraction of the buffer scan, and normalization to molar heat capacity (C_p), a cubic baseline was subtracted from the molar C_p curve to set the pre- and post-transition baselines to zero (Figure S2). The thermal unfolding temperature (T_m) was determined from the maximum of the C_p curve. The calorimetric enthalpy (ΔH_c) was obtained by integrating the C_p curve. The apparent van’t Hoff enthalpy (ΔH_v) was obtained by fitting the curve to a built-in model, MN2STATE, which fit ΔH_v independently of ΔH_c (Figure S3). Equations for the calculation of thermodynamic parameters are listed in Appendix I of the Supporting Information.

Differential Scanning Fluorimetry.

Using the Prometheus NT.48 instrument (NanoTemper Technologies, LLC, South San Francisco, CA) with 48 capillary chambers, each sample was excited at 290 nm, and emission was detected simultaneously at 330 and 350 nm. The first derivative of the fluorescence signal at 350 nm relative to 330 nm (F₃₅₀/F₃₃₀) versus temperature produced a DSC-like thermogram. The first derivative of fluorescence intensity at each wavelength exhibited the same thermal unfolding profiles. Because the UV–vis absorption of the rexinoids interfered with the detection of unfolding using F₃₅₀/F₃₃₀, DSF data of fluorescence intensity at 350 nm were used to obtain T_m. Sample preparation for DSF was the same as for DSC. Each capillary required 10 μL to load. Duplicate measurements were performed on each sample. T_m of each sample was automatically determined by the built-in analysis software and tabulated in an Excel output file. Global curve fitting of the apo-RXRα LBD DSF data obtained at different scan rates was performed using previously published equations⁴⁰ written in Origin C.

Circular Dichroism.

CD spectra were recorded in an OLIS CD spectrophotometer (OLIS, Athens, GA). Quartz cuvettes with a path-length of 0.02 cm were used. The protein concentration was 12 μM dimer. The CD cell holder was heated using an external water bath to the desired temperatures. CD spectra were recorded from 260 to 190 nm. Buffer baselines were recorded at the same temperatures and subtracted from the protein spectra. Total data collection time was ~5 min for each spectrum. The secondary structural content was determined by the CDNN program (CDNN: CD Spectra Deconvolution, Version 2.1, Universität—Böhm, 1997). For the completely denatured sample in guanidine–HCl, the sample was prepared by adding guanidine–HCl as a solid into the thermally unfolded protein solution to reach a final concentration of 5 M. The sample was then cooled in the cell holder, and the CD spectra were collected at 25 °C. The final protein concentration was adjusted based on the volume change after the addition of the denaturant. Optical interference from the denaturant prevented scans below 210 nm.

Isothermal Titration Calorimetry.

ITC experiments were performed on an Auto-iTC200 system (Malvern Instruments, Westborough, MA). The buffer used in ITC was the DSC buffer (pH 7.0) with 1% w/v DMSO. For titration of UAB30 with apo-RXRα LBD, the ITC sample cell contained 5 μM UAB30 (titrand), and the syringe contained 20 μM apo-RXRα LBD homodimer (titrant). For titration of RXRα LBD:rexinoid with GRIP-1, the titrand was 12.5 μM RXRα LBD dimer and 35 μM test rexinoid, and the titrant was 350 μM GRIP-1. Each titration experiment consisted of 16 injections of 2.5 μL of titrant into titrand at 10, 20, or 30 °C. Background mixing heat was determined from injections of titrant into the same buffer without titrand. Data analysis was performed using the built-in analysis module in Origin 7 provided by the ITC manufacturer. Monomeric protein concentration was used for analysis to obtain the stoichiometry of binding to each RXRα LBD monomer.

RESULTS AND DISCUSSION

I. Thermal Unfolding of Apo-RXRα LBD Homodimers.

The ΔG of apo-RXRα LBD unfolding at 30 °C was previously determined by Harder et al.⁴¹ using isothermal chemical denaturation. However, the enthalpic and entropic changes of the unfolding process were not obtained, nor was the unfolding heat capacity change which contributes significantly to the temperature dependence of ΔG.

1.1. Partial Reversibility of Transition.

To determine the stability of apo-RXRα LBD homodimer at 37 °C and other temperatures, thermal unfolding of this homodimer was measured by DSC from 35 to 70 °C using a scan rate (v) of 4 °C/min. A single endotherm was observed centered at the maximum thermal unfolding temperature (T_m) of 58.2 ± 0.1 °C, with a calorimetric unfolding enthalpy (ΔH_c) of 665 ± 4 kJ/mol (Figure 2, trace 1). Upon rescan, no thermal unfolding endotherm was observed, indicating that the unfolding of the homodimer was irreversible when heated to 70 °C (Figure 2, trace 2). To determine whether partial reversibility occurred during the unfolding process, the extent of reversibility at various points through the DSC endotherm was examined by consecutive scans of the same homodimer sample from 35 °C to various ending temperatures (Figure S4). DSC endotherms were also obtained on homodimers systematically heated for 1 min at various temperatures and then rapidly cooled in an ice bath. When the incubation temperature was 57.3 °C, which corresponded to 25% unfolding [the extent of unfolding at temperature T was determined by the ratio of integrated area from 35 °C to T, divided by the total area under the DSC curve], an endotherm was observed with the same T_m, but with approximately 80% the intensity of the first scan (Figure 2, trace 3). As the incubation temperature was increased to 58.2 °C (50% unfolding), the ΔH_c was decreased to 50% (Figure 2, trace 4). The ΔH_c was further decreased to 30% when the incubation temperature was increased to 60.0 °C (75% unfolding, Figure 2, trace 5). These data support that the unfolding of the homodimer was partially reversible through the endotherm.

Figure 2.

Thermal unfolding of apo-RXRα LBD is partially reversible. DSC molar heat capacity profile of 1.5 μM apo-RXRα LBD homodimer. (1) First scan; (2) rescan; (3) after heating at 57.3 °C for 1 min; (4) after heating at 58.6 °C for 1 min; and (5) after heating at 60.0 °C for 1 min.

The scan rate dependence of the thermal unfolding of apo-RXRa LBD was next examined. The scan rate, v, of the DSC measurement was decreased from 4.0 to 0.5 °C/min. As displayed in Figure 3A and summarized in Table 1, T_m decreased systematically as v decreased, reaching 55.6 ± 0.1 °C when v was 0.5 °C/min. ΔH_c remained nearly constant at 670 kJ/mol regardless of v. An apparent van’t Hoff enthalpy (ΔH_v) was determined by nonlinear least-squares curve fitting to a two-state model, whereby a natively folded protein, N, unfolds reversibly to U: N ↔ U.⁴² ΔH_v systematically increased from 815 ± 13 kJ/mol when v was 4 °C/min to 1208 ± 4 kJ/mol when v was 0.5 °C/min.

Figure 3.

Equilibrium unfolding parameters of apo-RXRα LBD obtained by extrapolation to infinite scan rate. (A) DSC molar heat capacity profiles for 1.5 μM apo-RXRα LBD homodimer at different scan rates, v (from left to right): 0.5, 1.0, 2.0, 3.0, and 4.0 °C/min. (B) T_m, ΔH_c, and ΔH_v as a function of v⁻¹. The T_m and ΔH_v values were fitted linearly to v⁻¹. The equilibrium unfolding parameters were obtained by extrapolation to zero v⁻¹.

Table 1.

Scan Rate Dependence of the Thermal Unfolding Parameters of apo-RXRα LBD

scan rate	DSC			DSFa
v (°C/min)	Tm (°C)	ΔHc (kJ/mol)	ΔHv (kJ/mol)	Tm (°C)
0.5	55.6 ± 0.1	681 ± 4	1208 ± 4	55.2 ± 0.1
1	56.1 ± 0.1	669 ± 1	1133 ± 4	55.6 ± 0.1
2	57.2 ± 0.1	677 ± 13	966 ± 8	56.4 ± 0.1
3	57.8 ± 0.02	677 ± 1	874 ± 13	56.7 ± 0.1
4	58.2 ± 0.1	665 ± 4	815 ± 13	57.0 ± 0.1
infiniteb	58.7 ± 0.2	673 ± 8	736 ± 25	57.4 ± 0.1

^aT_m measured from DSF monitored at 350 nm.

^bExtrapolation to v⁻¹ is 0 yielded the equilibrium unfolding parameters.

The thermal unfolding of apo-RXRα LBD was also monitored by measuring the changes in intrinsic fluorescence at 330 and 350 nm (Figure S5). The changes in tryptophan (Trp) emission intensity at 330 or 350 nm reflected the change in Trp environments during unfolding. The differential scanning fluorimetry (DSF) curves of apo-RXRα LBD, which contains two Trp residues, displayed only one unfolding transition that was similar to the DSC endotherms. The trends in the T_m at different scan rates matched those determined by DSC (Table 1), suggesting that both DSC and DSF detected the same global unfolding event. The T_m values determined by DSF were systematically lower than the T_m measured by DSC by ~1 °C. The lower T_m measured by DSF relative to those by DSC was observed for other protein unfolding processes.^43–45 Trp residues in folded states may sense changes in their local environments at a slightly lower temperature than onset of the global unfolding measured calorimetrically.⁴⁶

The lack of reversible unfolding transition in DSC rescans and the dependence of T_m on v indicated that the thermal unfolding of RXRα LBD was not in complete equilibrium throughout the heating process. DSC irreversibility is a common phenomenon, especially for multidomain or oligomeric proteins.⁴⁷ Often protein aggregation occurs at higher temperatures, which prevents reversibility.⁴⁸ However, these irreversible DSC transitions yield equilibrium data if the unfolding follows the Lumry–Eyring model: whereby N unfolds reversibly to U, which converts to a denatured state (D) slowly and irreversibly: N ↔ U → D.^49–51 If the irreversible step occurs slower than the rate of protein unfolding and refolding, then thermodynamic data for the reversible step are obtained by extrapolation of measured data to infinite scan rate.^47,50–54 For apo-RXRα LBD, the T_m and ΔH_v were plotted versus v⁻¹ (Figure 3B); each thermal parameter was linearly dependent on v⁻¹ for v faster than 0.5 °C/min. The extrapolated equilibrium unfolding parameters (Table 1) were as follows: T_m^eq = 58.7 ± 0.2 °C, ΔH_v^eq = 736 ± 25 kJ/mol, and ΔH_c^eq = 673 ± 8 kJ/mol (using the average of the apparent ΔH_c at different v values).

1.2. Two-State Unfolding without Dimer Dissociation.

The ratio of the van’t Hoff enthalpy to the calorimetric enthalpy, ΔH_v^eq/ΔH_c^eq, was essentially 1 using the extrapolated parameters. A cooperative unit of 1 when both ΔH_c^eq and ΔH_v^eq were determined based on per mole of dimer indicated that the native dimer was the cooperative unfolding unit. The ratio of 1 for ΔH_v/ΔH_c was not expected if the dimeric protein dissociated during unfolding: N₂ ↔ 2U. For this unfolding equilibrium, the unfolding endotherm is asymmetrical about its midpoint, and the apparent ΔH_v/ΔH_c is about 0.7⁵⁴ (see Figure S6 for a simulated DSC transition based on the N₂ ↔ 2U model, in comparison to the N ↔ U model). The asymmetry and low ΔH_v/ΔH_c were not observed in any of the DSC traces obtained in this study, suggesting that the native dimer unfolded without significant dissociation to monomers.^55–57

To examine this further, native gels were used to determine the aggregation state of the homodimer at three different temperatures: 57 °C which was below T_m, 58.7 °C which was the extrapolated equilibrium T_m, and 65 °C which was above the completion of the endotherm. As displayed in Figure 4, apo-RXRα LBD homodimer migrated near 50 kDa, consistent with a dimer, whereas the tetramer migrated near 100 kDa. When the dimer was heated at 57 °C for 1, 2, and 5 min and rapidly cooled in an ice bath, the band corresponding to the dimer was observed, but the intensity decreased as heating time increased. In addition, a species that migrated significantly above that of the unheated dimer emerged and increased in intensity, consistent with an aggregate of much higher molecular weight. When the heating temperature increased to 58.7 °C, less dimer and more aggregated species were observed. When heated to 65 °C for these times, the dimer disappeared, and only a highly aggregated species was present in the native gels. A band consistent with monomer was not observed in any lane.

Figure 4.

Changes in apo-RXRα LBD aggregation state during thermal unfolding. Native PAGE of 5 μM apo-RXRα LBD homodimer incubated at different temperatures for 1, 2, and 5 min, and then rapidly cooled in an ice bath. Lane 1: Fraction from SEC containing a mixture of tetramer and dimer that was not heated. Lane 2: Fraction from SEC containing only dimer that was not heated. Lanes 3–5: dimer heated at 57 °C (~25% unfolding) for 1, 2, and 5 min. Lanes 6–8: dimer heated at 58.7 °C (50% unfolding) for 1, 2, and 5 min. Lanes 9–11: dimer heated at 65 °C (past 100% unfolding) for 1, 2, and 5 min.

Circular dichroism (CD) and fluorescence spectroscopy were measured at several temperatures to inform on changes in secondary and tertiary structure during the endotherm. The X-ray crystal structure of apo-RXRα LBD contained 12 helices composing about 66% of its secondary structure.^4,7 The 222 and 208 nm negative CD signals (Figure 5A), which are significant in helical protein structures, were present in CD spectra at temperatures up to 56 °C.⁹ The magnitude of the 208 and 222 nm negative bands decreased ~20% when heated to temperatures near to T_m measured by DSC. These signals decreased ~40% when the dimer was heated to 75 °C, slightly higher than the temperature of the end of the DSC endotherm. No further decrease in CD signals was observed when the sample was incubated at 75 °C for 10 more minutes. This indicated that the secondary structure of apo-RXRα LBD was not completely unfolded throughout the DSC endotherm. Based on these CD spectra, the helical content of the native homodimer was 57%, and 33% remained after heating to 75 °C. In contrast, the 222 and 208 nm signals were completely lost for the homodimer in the presence of 5 M guanidine–HCl at 25 °C, which completely denatured and dissociated the dimer into monomers.⁴¹

Figure 5.

Spectroscopic changes in apo-RXRα LBD during thermal unfolding. (A) Circular dichroic spectra of 12 μM apo-RXRα LBD homodimer at different temperatures and in the presence of 5 M guanidine–HCl at 25 °C. (B) Fluorescence spectra of 0.05 μM apo-RXRα LBD homodimer at different temperatures and in the presence of 5 M guanidine–HCl at 25 °C. The excitation wavelength was 280 nm.

Each monomer of apo-RXRα LBD contains two Trp: W282 in helix 2 and W305 in helix 4, which is near the rexinoid binding site. An intense fluorescence band centered at 335 nm was observed at 35 °C, consistent with the hydrophobic environment of the indole group of the two Trp in a folded state (Figure 5B). Upon heating the protein to higher temperatures, the Trp fluorescence intensity decreased due to thermal quenching,⁵⁸ and the emission maximum wavelength (λ_max) gradually red-shifted to 344 nm at 70 °C. A red-shift of λ_max to 355 nm was observed for the completely denatured homodimer in 5 M guanidine–HCl. The much less red-shifted signals for the thermally denatured apo-RXRα LBD indicated that either one or both of the Trp were still in a partially hydrophobic environment when heated to 70 °C. Taken together, these thermodynamic and spectral data are most consistent with the native homodimer being converted to a partially unfolded and aggregated apo-RXRα LBD species above T_m.

1.3. Thermodynamic Unfolding Parameters at 37 °C.

Using a v of 4 °C/min for DSC measurements, the time needed to unfold the protein was less than 3 min, which minimized the effects of the irreversible processes during the unfolding transition. This is reflected in the fact that the unfolding parameters obtained at 4 °C/min were very similar to the extrapolated equilibrium parameters (Table 1 and Figure S3). Therefore, all subsequent DSC experiments were conducted at 4 °C/min. To modulate unfolding T_m and enthalpy for determining the unfolding heat capacity change (ΔC_p^u), the pH of the buffer was changed. Due to the rapidity of the measurement and lower sample amounts, DSF was first used to survey a range of pH values between 5 and 9 (Figure S7). DSF showed that apo-RXRα LBD was most stable at pH 7.0, and it required an increase of 2 pH units to induce significant decreases in T_m.

DSC was performed on apo-RXRα LBD homodimer at four different pH values: 7.0, 8.6, 9.0, and 9.5 (Figure 6A). To ensure that buffer ionization enthalpy did not contribute to the observed ΔH_c, inorganic buffers with low ionization enthalpies were used. Additional DSC experiments were performed using buffers with higher ionization enthalpies. The observed ΔH_c displayed no dependence on buffering components (Figure S8). ΔH_c obtained in inorganic buffers was fitted linearly to T_m (Figure 6B). From the slope of the line, ΔC_p^u was determined to be 15 ± 1 kJ/(mol K). Using this ΔC_p^u and the extrapolated values of T_m^eq and ΔH_c^eq from the DSC measurements (Table 1), ΔH of unfolding for the apo-RXRα LBD homodimer was calculated to be 347 ± 21 kJ/mol at 37 °C. ΔS of unfolding was 1.01 ± 0.06 kJ/(mol K); −TΔS was −314 ± 17 kJ/mol, and ΔG of unfolding was 33 ± 3 kJ/mol at 37 °C. A ΔG of 32 ± 3 kJ/mol was calculated using parameters obtained at a v of 4 °C/min, which was within experimental error of the ΔG calculated from the extrapolated parameters. Apo-RXRα LBD homodimer was most stable at 20 °C, with a ΔG of 43 ± 3 kJ/mol.

Figure 6.

Determination of the unfolding heat capacity change (ΔC_p^u) of apo-RXRα LBD homodimers. (A) DSC molar heat capacity profiles for 1.5 μM apo-RXR-LBD homodimer using a scan rate of 4 °C/min at different pH values (from left to right): 9.5, 9.0,8.6, and 7.0. (B) ΔC_p^u was determined by a linear regression of ΔH_c versus T_m. The R value for the linear fit was 0.99.

According to Harder et al.,⁴¹ the chemical denaturation of apo-RXRα LBD in guanidine–HCl is reversible and follows a three-state mechanism: N₂ ↔ 2I ↔ 2U, wherein “I” is a monomeric, partially unfolded, intermediate. Although the ΔG change of the first step accounts for ~80% of the total unfolding ΔG, spectroscopy data support that the monomeric intermediate retains significant native secondary structure, and the two Trp residues remain partially buried. It is during the second step of unfolding where this unfolding intermediate, I, loses all secondary structures, and the Trp are completely solvent exposed. The authors estimate that the ΔG of chemical denaturation for the first step is 35 ± 0.8 kJ/mol at 30 °C. From our study, the estimated thermal unfolding ΔG for apo-RXRα LBD homodimers at 30 °C was 39 ± 2 kJ/mol. Likewise, the spectral data on the thermally unfolded protein (Figure 5) are similar to those estimated for the monomeric intermediate in guanidine–HCl. Guanidine–HCl is well-known for its capability to prevent aggregation of partially or fully unfolded proteins because of its chaotropic and ionic properties.^59–61 It is possible that, in our study, DSC detected the unfolding of the native dimer to a partially unfolded intermediate very similar to that detected by Harder et al. However, because of the lack of chaotropes such as guanidine–HCl to stabilize a monomer, the unfolded state was thermally aggregated without a significant heat signature.

The data presented strongly support the model N₂ ↔ I₂ → D, where I₂ is a partially unfolded, dimeric intermediate. This model⁴⁰ was used to globally fit a set of four DSF data at different scan rates (1 to 4 °C/min, Figure S5). The best fitted parameters for the reversible first step were T_m = 57.5 °C and ΔH_v = 710 kJ/mol. These values were very similar to those obtained by extrapolation to infinite scan rate (Table 1). The second, irreversible step was best fitted with a single rate-constant, k, of 2.1 min^–1. To gain more understanding on how the rate of the irreversible step affects the extrapolated thermodynamic parameters obtained by fitting the DSC to a reversible model, a series of simulations were performed using k values of 0.5, 2.1, and 8 min⁻¹ (Figure S9A). The simulated curves were fitted to the two-state reversible model, N₂ ↔ I₂, and the resulting thermodynamic parameters, T_m, ΔH_v, and ΔH_c, were plotted against v⁻¹ (Figure S9B). When k was 0.5 min⁻¹, which was relatively slow compared to the scan rates, the extrapolated T_m^eq and ΔH_v^eq were essentially identical to the input values for the simulation (Figure S9B, green circles on the axes). When k was 8 min⁻¹, which was relatively fast compared to the scan rates, T_m^eq and ΔH_v^eq were different from the input values. T_m^eq was 1.5 °C lower, and ΔH_v^eq was 23% higher. In contrast, the observed ΔH_c value was constant and equal to the input value. When k was 2.1 min⁻¹, which was our fitted value using the Lumry–Eyring model, T_m^eq was lower by only 0.5 °C, and ΔH_v^eq was higher by 9%. Again, ΔH_c was scan-rate-independent and equal to the input value. In short, if partial reversibility was established, and the rate of the irreversible step was no faster than the highest scan rate, then T_m^eq obtained by extrapolation to infinity v was within 1% of the input value. ΔH_v^eq could be different from the input value, but ΔH_c^eq was equal to the input value. Therefore, ΔH_c^eq became the logical choice to use for the calculation of other thermodynamic parameters, such as ΔG.

The ΔG of unfolding for apo-RXRα LBD was unusually low compared to other dimeric proteins, which may suggest a weak dimeric interface ready to dissociate into monomers at sub-micromolar concentrations to facilitate the formation of heterodimers with other receptors.⁴¹ However, our thermal unfolding data indicated that the dimer interface was relatively stable as it did not dissociate throughout the unfolding endotherm. This was more consistent with the DNA binding properties and transcriptional activities observed for the full-length RXR homodimer.^3,62 Moreover, the thermal unfolding of apo-RXRα LBD homodimer was accompanied by a relatively low unfolding ΔH_c of 673 kJ/mol at T_m, i.e., a low specific heat of 12.5 J/g, which was substantially lower than the specific heat of a typical soluble globular protein at the same T_m (29–38 J/g according to Murphy and Freire⁵⁵). The low specific heat may be caused by three factors. First, it may be due to incomplete unfolding as demonstrated by the CD and fluorescence spectroscopy data on the thermally unfolded protein. Second, it may be due to homodimers generally unfolding with lower specific heats than single domain proteins.⁶³ Third, it may indicate that the native apo-homodimer is less compactly folded and more dynamic than typical, well-folded globular proteins. Such structural fluidity likely relates to the lack of rexinoid bound to its hydrophobic pocket and the dynamic helix 12 region of the domain.^8,9

II. Thermal Unfolding of Holo-RXRα LBD Homodimers Bound with Rexinoids.

The stability of RXRα LBD bound with rexinoid (holo-RXRα LBD) was examined next to determine the relationship between the rexinoid structure and holo-RXRα LBD unfolding energetics. Rexinoids UAB30, UAB110, and UAB111 quenched more than 90% of the protein fluorescence signal of apo-RXRα-LBD at 337 nm at 25 °C when the ratio of the protein to the rexinoids reached 1:1. The dissociation constants (K_d) determined from the binding isotherm based on fluorescence quenching was 38 ± 14 nM for UAB30, 22 ± 6 nM for UAB110, and 2.4 ± 0.4 nM for UAB111.

DSF was first used to determine the rexinoid concentration required to saturate the binding sites at T_m because K_d was expected to be dependent on temperature. For UAB30, the DSF unfolding curve continuously shifted to higher T_m (Figure S10A) when the ligand concentration was increased from 1.25 to 25 μM using a dimer protein concentration of 2.5 μM. There was no measurable increase in T_m when the rexinoid concentration was doubled to 50 μM, indicating that effective saturation was reached at rexinoid concentrations above 25 μM (5:1 molar ratio, based on one binding site/monomer). An increase in T_m of 4.4 ± 0.1 °C in the presence of UAB30 at a 2:1 rexinoid/monomer molar ratio was consistent with UAB30 strongly binding to the native homodimer at 25 °C and at elevated temperatures. For UAB110 and UAB111, the shifts in T_m in the presence of rexinoids at a 2:1 rexinoid/monomer molar ratio were 7 and 7.5 °C, respectively, which were substantially higher than the shifts caused by UAB30 (Figure S10B) and consistent with their higher binding affinities.

DSF was also used to examine the effect of scan rate on the T_m of the holo-RXRα LBD homodimer bound with each rexinoid. The linear regression of T_m values of each complex with respect to v⁻¹ yielded slopes almost identical to that of apo-RXRα LBD (Figure 7). This suggested that the rate of the irreversible step in the Lumry–Eyring unfolding model was similar for both holo-homodimers and the apo-homodimer.

Figure 7.

DSF T_m values as a function of v⁻¹ for RXRα LBD homodimers with and without rexinoids or coactivator peptide. DSF T_m values of 2.5 μM RXRα LBD homodimer at pH 7.0, in the presence of no rexinoid or GRIP-1 (Apo), 0.4 mM GRIP-1 (Apo + GRIP), 30 μM UAB30 (UAB30), 30 μM UAB30 and 0.4 mM GRIP-1 (UAB30 + GRIP), 10 μM UAB110 (UAB110), 10 μM UAB110 and 0.4 mM GRIP-1 (UAB110 + GRIP), 10 μM UAB111 (UAB111), or 10 μM UAB111 and 0.4 mM GRIP-1 (UAB111 + GRIP). The scan rate, v, was 4.0 °C/min. The lines are linear regressions of T_m values with respect to v⁻¹. All samples contained 1% DMSO.

To gather thermodynamic data on holo-RXRα LBD dimers, DSC was performed. The effect of DMSO, a cosolvent necessary for solubilizing the rexinoids, was first examined. Using a v of 4 °C/min, the T_m, ΔH_c, and ΔH_v values of apo-RXRα LBD homodimer determined by DSC in the presence of 1% w/v DMSO (Table 2) were within experimental errors of those values obtained in the absence of DMSO (Table 1). This sample was used as the point of reference (DMSO control) for comparison with holo-RXRα LBD dimers containing rexinoids.

Table 2.

Thermodynamic Parameters of Unfolding of Apo-RXRα LBD and Holo-RXRα LBD with and without GRIP-1

	Tm (°C)	ΔHc (kJ/mol)	ΔHv (kJ/mol)	ΔCpu (kJ/(mol K))	ΔH (37 °C) (kJ/mol)	−TΔS (37 °C) (kJ/mol)	ΔG (37 °C) (kJ/mol)
apo-RXRα LBDa	58.7 ± 0.2	673 ± 8	736 ± 25	15 ± 1	347 ± 21	−314 ± 17	33 ± 3
DMSO controlb	57.9 ± 0.4	690 ± 8	803 ± 4	15 ± 1	376 ± 17	−342 ± 17	33 ± 2
+GRIP-1	59.3 ± 0.2	690 ± 29	811 ± 13	14 ± 2	389 ± 52	−353 ± 48	36 ± 7
+UAB30	63.6 ± 0.1	803 ± 4	836 ± 13	17 ± 1	371 ± 26	−326 ± 23	45 ± 4
+UAB110	66.6 ± 0.5	849 ± 13	811 ± 17	16 ± 2	362 ± 37	−310 ± 32	52 ± 8
+UAB111	66.9 ± 0.3	840 ± 21	823 ± 21	17 ± 1	362 ± 24	−309 ± 20	53 ± 5
+UAB30/GRIP-1	67.5 ± 0.3	932 ± 33	924 ± 25	17 ± 1	374 ± 35	−317 ± 30	58 ± 8
+UAB110/GRIP-1	69.9 ± 0.5	995 ± 17	924 ± 8	18 ± 1	393 ± 35	−327 ± 30	66 ± 8
+UAB111/GRIP-1	70.4 ± 0.2	991 ± 29	928 ± 8	20 ± 2	380 ± 35	−314 ± 30	66 ± 9

^aEquilibrium unfolding parameters obtained by extrapolating to v⁻¹ is zero, in the absence of DMSO.

^bDMSO control was apo-RXRα LBD containing 1% w/v DMSO.

Data for the DMSO control and all others were obtained at a v of 4 °C/min and in the presence of 1% DMSO.

Rexinoid concentrations that caused maximum increases in T_m determined by DSF (30 μM for UAB30, and 10 μM for either UAB110 or UAB111) were used in DSC to minimize unwanted effects of the rexinoids at high concentrations (Figure S11). Similar to apo-RXRα LBD homodimer, only one DSC unfolding transition was observed in the presence of rexinoids (Figure 8), and the T_m of each holo-RXRα LBD increased due to the shift of unfolding equilibrium toward the native state caused by rexinoid binding to the native apo-dimer. To verify that the unfolding parameters obtained at a v of 4 °C/min for the holo-RXRα LBD homodimers were good approximations of the equilibrium parameters, the scan-rate dependence of the DSC endotherms of RXRα LBD:UAB30 (holo-RXRα LBD homodimer bound with UAB30) was examined (Figure S12 and Table S1). Similar to apo-RXRα LBD, a linear relationship existed between each unfolding parameter and v⁻¹ (Figure S12B). The extrapolated equilibrium unfolding parameters for RXRα LBD:UAB30 were T_m^eq = 63.5 ± 0.1 °C, ΔH_v^eq = 761 ± 38 kJ/mol, and ΔH_c^eq = 798 ± 8 kJ/mol (Table S1). The unfolding parameters obtained at a v of 4 °C/min were very similar to these extrapolated values. The ratio of ΔH_v^eq to ΔH_c^eq was close to 1, as found for the apo-homodimer.

Figure 8.

RXRα LBD is strongly stabilized by rexinoid binding and coactivator peptide GRIP-1. DSC molar heat capacity profiles of 1.5 μM RXRα LBD at pH 7.0, in the presence of no rexinoid or GRIP-1 (Apo), 0.4 mM GRIP-1 (+GRIP), 30 μM UAB30 (UAB30), 30 μM UAB30 and 0.4 mM GRIP-1 (UAB30 + GRIP), 10 μM UAB110 (UAB110), 10 μM UAB110 and 0.4 mM GRIP-1 (UAB110 + GRIP), 10 μM UAB111 (UAB111), or 10 μM UAB111 and 0.4 mM GRIP-1 (UAB111 + GRIP). The scan rate, v, was 4.0 °C/min. All samples contained 1% DMSO.

An unfolding T_m of 66.6 ± 0.5 °C was obtained for RXRα LBD:UAB110 at a v of 4 °C/min, which was 3 °C higher than the T_m of RXRα LBD:UAB30 (Figure 8 and Table 2). The ΔH_c was 849 ± 13 kJ/mol, which was approximately 260 kJ/mol higher than the apo-homodimer in 1% DMSO. RXRα LBD:UAB111 displayed similar unfolding parameters as UAB110:RXRα LBD, with a T_m of 66.9 ± 0.3 °C and a ΔH_c of 840 ± 21 kJ/mol. Both RXRα LBD:UAB110 and RXRα LBD:UAB111 also exhibited a ΔH_v/ΔH_c ratio close to unity, consistent with the holo-RXRα LBD homodimers being the cooperative unfolding unit.

To determine the ΔC_p^u values of the holo-RXRα LBD homodimers, T_m and ΔH_c were measured at pH 8.8 and 9.5. In a similar manner to apo-RXRα LBD homodimers, DSF was first used to evaluate whether pH modulates the degrees of stabilization caused by the presence of rexinoids (Figure S13 and Table S2). The shifts in T_m caused by rexinoid were within experimental errors at all three pH values. Using DSC, the measured T_m values and enthalpies were lower as pH decreased, but the relative positions and magnitudes of all DSC endotherm remained the same (Figure S14). The ΔH_c value of each holo-homodimer was fitted linearly with respect to T_m to determine ΔC_p^u values (Table 2).

Using ΔC_p^u, T_m, and ΔH_c from DSC measurements at a v of 4 °C/min, ΔH, ΔS, and ΔG of unfolding at 37 °C for each holo-RXRα LBD homodimer were calculated (Table 2). Compared to the apo-RXRα LBD homodimer DMSO control, the holo-RXRα LBD homodimers had significantly higher ΔG values consistent with tight binding and strong stabilization of the native state by the rexinoids. The degrees that UAB110 and UAB111 stabilized the native apo-RXRα LBD homodimer (ΔΔG = 19 and 20 kJ/mol, respectively) were notably higher than UAB30 (ΔΔG = 12 kJ/mol), which was consistent with the improved binding affinities of UAB110 and UAB111 over that of UAB30.

The difference in ΔH between the DMSO control and holo-RXRα LBD homodimers (ΔΔH) at 37 °C was essentially zero for each of the three rexinoids. A near-zero value for ΔΔH suggested that the binding of rexinoid to apo-RXRα LBD at physiological temperatures was accompanied by a small binding enthalpy. In order to gather information on the thermodynamics of rexinoid binding to apo-RXRα LBD, ITC measurements were performed on UAB30 between 10 and 30 °C. No ITC isotherms were obtained due to low solubility of the rexinoid in water. When reverse titrations were performed at these temperatures, the measured heat changes were within the noise levels (Figure S15), which suggested a low-enthalpy binding reaction at these temperatures, consistent with the DSC results.

The ligand binding pocket (LBP) of RXRα LBD containing UAB30 is lined with 16 hydrophobic residues contributed by 4 protein helices that surround the UAB rexinoids.^8,9,13,14 The main forces that stabilize the interactions between the rexinoids and LBD residues are the ionic interaction between the carboxylate group and Arg316, and numerous hydrophobic contacts throughout the binding pocket. Burying nonpolar atoms upon ligand binding leads to a negative heat capacity change (ΔC_p^a).^9,37 Exposing these buried nonpolar surface areas upon protein unfolding and ligand dissociation increases the unfolding ΔC_p^u. The ΔC_p^u of each of the three holo-RXRα LBD homodimers was slightly higher than that of the apo-RXRα LBD homodimer in 1% DMSO (+1 to 2 kJ/(mol K)). The − TΔS values of the holo-RXRα LBD homodimers were each less negative than that of the apo-RXRα LBD homodimer in DMSO (+17 to 31 kJ/mol), which contributed favorably to the ΔG values. Therefore, rexinoid binding to apo-RXRα LBD appeared to be entropically driven at physiological temperatures. Crystallographic data indicate that, upon binding to UAB110 or UAB111, the size of the LBP expanded nearly 20% (compared to LBP in the presence of UAB30) to accommodate the two larger rexinoids (Figure 1E).¹⁴ The contact surface areas of these two rexinoids were about 100 Ų larger than those observed for UAB30. Burying more hydrophobic surfaces likely resulted in more favorable entropy changes during the binding process and, hence, the more favorable ΔG of folding for holo-RXRα LBD homodimers bound to UAB110 or UAB111. In addition, the increase in stability of the holo-RXRα LBD homodimers could also arise from reduced dynamics due to rexinoid binding. HDX MS studies have revealed that the dynamics of helices 3 and 11 are significantly decreased when UAB30 is bound to the homodimer⁸ and reduced even further in the presence of UAB110 and UAB111 whose binding affinities were higher than that of UAB30 (unpublished data). In summary, the LBP was expanded but more thermodynamically stable when bound with UAB110 or UAB111. How these conformational changes affect coactivator binding was examined next.

III. Thermal Unfolding of Holo-RXRα LBD Homodimers Bound with Rexinoids and a Coactivator Peptide.

It was previously shown by using ITC measurements that the 13-mer coactivator peptide, GRIP-1 (Figure 1C), binds to holo-RXRα LBD complexes with a micromolar K_d at 25 °C in an exothermic reaction with binding enthalpies in the range from −36 to −42 kJ/mol per monomer.^8,14 Because the binding affinity is expected to be weaker at T_m, DSF was first used to determine the GRIP-1 concentration required to saturate the peptide binding site at T_m (Figure S16). The T_m of UAB30:RXRα-LBD:GRIP-1 unfolding incrementally increased when GRIP-1 concentration was increased from 12.5 to 800 μM using a dimer protein concentration of 2.5 μM and 15 μM UAB30. A concentration of 400 μM GRIP-1 was chosen for the collection of thermodynamic data on RXRα LBD homodimer bound with rexinoid and GRIP-1.

DSF was used to examine the dependence of T_m on v (Figure 7). In the absence of rexinoids, GRIP-1 increased the T_m of apo-RXRα LBD by ~1 °C, consistent with its low affinity to the apo-homodimer at 25 °C.⁹ A much larger increase in T_m was observed in the presence of each rexinoid. All T_m values appeared linearly dependent on v⁻¹, and the slopes of the linear fits were similar to those obtained in the absence of GRIP-1. This again suggested that the rates of the irreversible step of the Lumry–Eyring unfolding mechanism were similar in the presence or absence of rexinoids or GRIP-1. The RXRα LBD:UAB111:GRIP-1 complex displayed the highest extrapolated T_m of 68.7 ± 0.1 °C, nearly 11 °C higher than that of the apo-homodimer. The RXRα LBD:UAB110:GRIP-1 complex had an extrapolated T_m of 67.9 ± 0.2 °C, and the RXRα LBD:UAB30:GRIP-1 complex had an extrapolated T_m of 66.4 ± 0.1 °C. Using DSC (Figure 8 and Table 2), similar increases in T_m were observed for the ternary complexes.

The ΔC_p^u values of holo-RXRα LBD homodimers in complex with GRIP-1 were determined from the linear dependence of ΔH_c on T_m and listed in Table 2. The unfolding parameters of apo-RXRα LBD homodimer in the presence of 400 μM GRIP-1 were similar to the apo-homodimer with only slightly increased T_m and ΔH_c, and almost identical ΔC_p^u. This indicated that significant dissociation of GRIP-1 likely occurred before the apo-homodimer unfolded because of low binding affinity at T_m. For two of the three RXRα LBD:rexinoid:GRIP-1 complexes, ΔC_p^u increased with respect to their corresponding holo-homodimer without GRIP-1, while there was little observed change in ΔC_p^u for RXRα LBD:UAB30:GRIP-1. The largest increase in ΔC_p^u was 3 kJ/(mol K) for RXRα LBD:UAB111:GRIP-1, which was within the experimental errors of the ΔC_p^u measurement. Therefore, an average ΔC_p^u value of 18 ± 2 kJ/(mol K) was calculated from the experimentally determined values for the three ternary complexes.

Using the above average ΔC_p^u value and the T_m and ΔH_c values from DSC measurements at 4 °C/min, ΔH, ΔS, and ΔG of unfolding at 37 °C for the holo-RXRα LBD homodimers bound with GRIP-1 peptide were calculated (Table 2). Binding of GRIP-1 increased ΔG compared to measurements obtained in the absence of the coactivator peptide: 13 ± 4 kJ/mol for RXRα LBD:UAB30, 14 ± 3 kJ/mol for RXRα LBD:UAB110, and 13 ± 3 kJ/mol for RXRα LBD:UAB111. Using eq 5 in Appendix I (Supporting Information) based on the shifts in T_m, the estimated K_d of GRIP-1 binding at T_m was 22 μM for RXRα LBD:UAB30, 26 μM for RXRα LBD:UAB110, and 24 μM for RXRα LBD:UAB111.

In our previous analysis using ITC,⁸ the binding enthalpy of GRIP-1 to RXRα LBD:UAB30 was determined to be −44.6 ±0.1 kJ/mol per monomer at 30 °C, with a binding heat capacity change (ΔC_p^a) of −1.5 ± 0.1 kJ/(mol K). Using these values, binding enthalpy at 67.5 °C (T_m of the RXRα LBD:UAB30:-GRIP-1 ternary complex) was calculated to be −100 kJ/mol per monomer or −200 kJ/mol per dimer, which was similar to the measured value of −129 ± 35 kJ/mol. Based on the van’t Hoff equation (eq 6 in Appendix I of Supporting Information), and a K_d of 26 μM at T_m, the K_d at 25 °C was calculated to be 0.74 μM, in close agreement with the K_d determined by ITC at this temperature.⁸

To obtain an accurate measurement of the ΔC_p^a values for GRIP-1 binding to RXRα LBD:UAB110 or RXRα LBD:UAB111, ITC was performed between 10 and 30 °C (Table S3). As observed for the RXRα LBD:UAB30 complex, the binding stoichiometry was nearly 1:1 (coactivator peptide/RXRα LBD monomer unit). The free energy change of GRIP-1 binding to RXRα LBD:rexinoid complexes was driven strongly by a large negative enthalpy change. The ΔC_p^a was determined to be 1.19 ± 0.01 kJ/(mol K) for GRIP-1 binding to RXRα LBD:UAB110, and 1.11 ± 0.07 kJ/(mol K) for GRIP-1 binding to RXRα LBD:UAB111 (per monomer unit). The K_d at 25 °C was calculated to be 1.5 μM for RXRα LBD:UAB110 and 1.8 μM for RXRα LBD:UAB110, also in close agreement with the K_d determined by ITC at this temperature.¹⁴

In summary, we report here the thermal unfolding of three species of RXRα LBD homodimers: (1) the apo-homodimer alone, N₂, which reversibly unfolded to a dimeric intermediate, I₂, that irreversibly denatured to D with a rate constant, k; (2) the holo-homodimer bound with a rexinoid to each monomer, N₂R₂; and (3) the ternary complex of the homodimer bound with a rexinoid and a coactivator peptide GRIP-1 to each monomer, N₂R₂P₂.

$$\begin{gathered} {\text{N}}_2\leftrightarrow {\text{I}}_2\to \text{D} \\ \Delta G(37°\text{C})=33\text{kJ}/\text{mol};k({\text{I}}_2\to \text{D})=2.1{\text{min}}^{-1} \end{gathered}$$ (2)

$$\begin{gathered} {\text{N}}_2{\text{R}}_2\leftrightarrow {\text{N}}_2+2\text{R} \\ \Delta \Delta G(37°\text{C})=12-20\text{kJ}/\text{mol}(\text{rexinoid-dependent}) \end{gathered}$$ (3)

$$\begin{gathered} {\text{N}}_2{\text{R}}_2{\text{P}}_2\leftrightarrow {\text{N}}_2{\text{R}}_2+2\text{P} \\ \Delta \Delta G(37°\text{C})=14\text{kJ}/\text{mol}(\text{rexinoid-independent}) \end{gathered}$$ (4)

From the analyses of DSC endotherms of N₂ at different scan rates, we estimated that N₂ was 33 kJ/mol more stable than I₂ at 37 °C. Relative to this equilibrium, the analyses of DSC endotherms of N₂R₂ bound with three different UAB rexinoids indicated that N₂ was stabilized by an additional 12–20 kJ/mol, depending on the structure of the rexinoid. This stabilization was consistent with our reported 9–35 nM binding affinity of UAB rexinoids determined by fluorescence quenching assays.¹⁴ The ΔΔG of stabilization caused by each rexinoid positively correlated with their binding affinities, with UAB111 being the tightest binder. The DSC analyses of N₂R₂P₂ revealed that GRIP-1 binding further stabilized the complex by an additional 14 kJ/mol at 37 °C. Unlike rexinoid binding, this additional stabilization was nearly identical for the homodimers bound with three different rexinoids, which was in close agreement with the K_d values of GRIP-1 release obtained by ITC.¹⁴ Even though GRIP-1 binding stabilized N₂R₂ the same, N₂R₂P₂ bound with UAB110 or UAB111 was 8 kJ/mol more stable than this complex with UAB30.

The equilibrium ΔΔG of binding was determined in this study even with the apo-homodimer unfolding irreversibly. Rexinoid binding to each monomer of N₂ shifted the population of N₂ to N₂R₂, causing the equilibrium N₂ ↔ I₂ to require higher temperature to reach 50% unfolding. GRIP-1 binding further shifted the population to N₂R₂P₂, thus requiring even higher temperature to produce unfolding. As long as the rexinoid/coactivator binding processes are in equilibrium, the thermodynamic parameters of binding determined from ΔΔG (T_m) and extrapolated to 37 °C using measured ΔC_p will be reasonable. These measurements could be compromised if the ligand-bound species, N₂R₂ and/or N₂R₂P₂, irreversibly unfold to new denatured species with different rates. This does not appear to be the case for this study since, as shown in Figure 7, the T_m vs v⁻¹ plots display linear relationships with nearly identical slopes, suggesting that the rate constant and the activation energy for the irreversible step are similar for each native species.

CONCLUSIONS

In this study we demonstrated that apo-RXRα LBD homodimer is favored in the folded state by at least 33 kJ/mol in free energy change at 37 °C, driven by a favorable enthalpy change. On a per residue basis, the enthalpy value is almost 50% less than other small globular proteins, most likely due to two factors: (1) thermal unfolding which produced an incompletely unfolded state and (2) the less compact nature of the apo-homodimer that contains a large unfilled hydrophobic rexinoid binding pocket and a dynamic c-terminal end. Upon rexinoid binding, the hydrophobic pocket is filled, stabilized by numerous hydrophobic and hydrophilic interactions between the rexinoids and binding pocket residues. These interactions also help bridge the helices interacting with the rexinoids to those at the terminal end where the coactivator peptide binds. We clearly demonstrate here for the first time that rexinoid binding to apo-RXRα LBD is entropically driven at physiological temperatures with negligible enthalpy change. The rexinoids enhanced the stability of the homodimer complex by 12–20 kJ/mol, depending on their structure. The ΔΔG of stabilization is 8 kJ/mol less for UAB30 than the larger and more hydrophobic UAB110 or UAB111. The binding of coactivator peptide results in an additional stabilization to the homodimer complex of approximately 14 kJ/mol. The incremental increase in free energy is the same for each of the three rexinoids studied. Our data on GRIP-1 binding to apo-RXRα LBD also indicate that a bound rexinoid is required for coactivator peptide binding. Structurally, the homodimer complexes with GRIP-1 are essentially identical. It seems as if GRIP-1 does not sense the subtle differences in the LBP conformation caused by different rexinoids as long as the LBP is occupied by an agonist of sufficient size and hydrophobicity to promote GRIP-1 binding to its site on the surface of the domain. These results suggest that, for future studies on RXR heterodimers, we would want to investigate RXR heterodimer interactions in addition to differential coactivator affinity. Alternatively, we may find that different RXR coactivator LXXLL motifs have unique binding characteristics with unique RXR ligands.

Finally, even though the GRIP-1 binding thermodynamics are the same for each of the three holo-homodimers, the complexes of UAB110 and UAB111 are each more stable than the UAB30 complex (8 kJ/mol) not due to enhanced coactivator interactions but due to hydrophobic interactions of the rexinoid in its interior binding pocket. The enhanced stability of UAB110 and UAB111 correlates well with in vitro transcriptional assays¹⁴ that clearly demonstrate that UAB111 is one of the most potent RXRα agonists discovered to date (40-fold more potent than UAB30). As reported here, despite the essentially identical X-ray crystal structures of the homodimer complexes, the two homodimer complexes with UAB110 or UAB111 are clearly more stable than the complex with UAB30. This demonstrates the value in thermodynamic measurements in predicting agonist potency. The gathering of this information was made more complicated by an irreversible step in the thermal unfolding of apo-RXRα LBD dimers. The comprehensive analyses presented here not only yield specific thermodynamic information on the rexinoids studied but also provide a general method of how to deal with this complication. These studies lay the groundwork for future thermodynamic studies on RXR agonists and antagonists, as well as RXR heterodimers and other coactivators.

ABBREVIATIONS

RXRα LBD

retinoid X receptor-alpha ligand binding domain

LBP

ligand binding pocket

9cRA

9-cis retinoic acid

GRIP-1

glucocorticoid receptor interacting protein-1

RXRα LBD:UAB30

RXRα LBD bound to UAB30

RXRα LBD:UAB30:GRIP-1

RXRα LBD bound to UAB30 and GRIP-1

RXRα LBD:UAB110

RXRα LBD bound to UAB110

RXRα LBD:UAB110:GRIP-1

RXRα LBD bound to UAB110 and GRIP-1

RXRα LBD:UAB111

RXRα LBD bound to UAB111

RXRα LBD:UAB111:GRIP-1

RXRα LBD bound to UAB111 and GRIP-1

SRC

steroid receptor coactivator

DSC

differential scanning calorimetry

v

scan or heating rate

T _m

thermal unfolding temperature

ΔH_v

van’t Hoff enthalpy of unfolding

ΔH_c

calorimetric enthalpy

ΔG

Gibbs free energy of unfolding

ΔS

entropy of unfolding

ΔC_p^u

unfolding heat capacity change

K _u

unfolding equilibrium constant

DSF

differential scanning fluorimetry

F₃₅₀/F₃₃₀

ratio of fluorescence intensity at 350 nm divided by fluorescence intensity at 330 nm

λ _ex

excitation wavelength for fluorescence measurements

CD

circular dichroism

ITC

isothermal titration calorimetry

ΔH_a

binding enthalpy

ΔC_p^a

binding heat capacity change

K _a

ligand binding constant

K _d

ligand dissociation constant

HDX-MS

hydrogen–deuterium exchange mass spectrometry

N

native folded protein

U

reversibly unfolded protein

D

irreversibly denatured protein

I

partially unfolded intermediate

k

rate constant of irreversible step

R

rexinoid

P

coactivator peptide GRIP-1

N₂

native folded apo-RXRα LBD homodimer protein

I₂

partially unfolded apo-RXRα LBD dimeric intermediate

N₂R₂

native folded holo-RXRα LBD homodimer bound with rexinoids

N₂R₂P₂

native folded holo-RXRα LBD homodimer bound with rexinoids and GRIP-1