|Home | About | Journals | Submit | Contact Us | Français|
Protein quality and stability are critical during protein purification for X-ray crystallography. A target protein that is easy to manipulate and crystallize becomes a valuable product useful for high throughput crystallography for drug design and discovery. In this work, a single surface mutation, D355R, was shown to be crucial for converting the modestly stable monomeric ligand binding domain of the human thyroid hormone receptor (TR LBD) into a stable dimer. The structure of D335R TR LBD mutant was solved using X-ray crystallography and refined to 2.2 Å resolution with Rfree/R values of 24.5/21.7. The crystal asymmetric unit reveals the TR dimer with two molecules of the hormone-bound LBD related by a two-fold symmetry. The ionic interface between the two LBDs comprises residues within loop H10-H11 and loop H6-H7 as well as the C-terminal halves of helices 8 of both protomers. Direct intermolecular contacts formed between the introduced residue Arg 355 of one TR molecule and Glu 324 of the second molecule become a part of the extended dimerization interface of 1330 Å2 characteristic for a strong complex assembly that is additionally strengthened by buffer solutes.
Thyroid hormone receptor (TR) belongs to the family of nuclear transcriptional regulators, proteins intensively studied for development of new pharmaceuticals (1, 2). TR is a drug target for diabetes, obesity and other thyroid dysfunctions as it regulates numerous metabolic processes through transcriptional control of its target genes. TR is required for normal function of nearly all tissues and organs, and it is vital for regulation of oxygen consumption and metabolic rate. Multiple functions of TR are explained in part by its ability to form homodimers or heterodimers with retinoid X receptor (RXR) (3,4). The dimerization event is controlled by different ligands in different cellular contexts or may not require ligands for either receptor (5,6). Additional diversity in function comes from four existing TR isoforms that are products of two distinct genes and alternate splicing (7). The TRα1, TRβ1 and TRβ2 isoforms are expressed in most tissues; they bind thyroid hormones and function as ligand-regulated transcription factors. In contrast, the TRα2 isoform does not bind hormone and is prevalent in pituitary and central nervous system (8,9). Selective targeting of the tissue specific TR isoforms is a major goal of drug development.
Similar to other nuclear receptors (NRs), TR has a modular structure consisting of a ligand-independent N-terminal domain (AF1), which associates with transcriptional regulators, a highly conserved DNA-binding domain (DBD), and a C-terminal ligand-binding domain (LBD) (9). TR LBD mediates transcriptional activation or repression through conformational changes induced by hormone binding (10,11). Several structures of TRα and TRβ LBD bound by activating ligands have been determined to date, all of which are of the TR monomer in the stable agonist-bound conformation (12,13). Similar to other NRs, the TR LBD comprises a three-layered sandwich structure formed by 12 α-helices and two short β-strands. The bound ligand agonist completes the hydrophobic core of the LBD.
For rational design of TR specific drugs, high-resolution structural data revealing the molecular basis of ligand selectivity are particularly useful (2, 14). However, the TR LBD has proven to be a delicate protein and a challenging crystallization target. For difficult to crystallize proteins, improving protein behavior through protein engineering has shown good results. The most common approaches include mutating surface cysteine residues (15) or removing unstructured loops or hinge regions (16) that might cause structural heterogeneity. One report described a rational possibility of improving crystallizability of proteins through designing symmetrical covalent oligomers stabilized by synthetic Cys-Cys contacts (17). Our study demonstrates a unique approach when the monomeric protein has been stabilized via formation of a symmetrical non-covalent homodimer. This designed homodimer is characterized by an extended dimerization interface built by combination of physical forces: hydrophobic interactions, hydrogen bonds and electrostatic contacts.
A cDNA fragment encoding human TRβ LBD (a.a. 209-461) was made using PCR and cloned into pETDuet-1 expression vector (Novagen) using BamHI/HindIII restriction sites. The D355R mutation was introduced using QuickChange kit (Stratagene) according to the manufacturer's protocol. The resulting recombinant plasmid was verified by sequencing. The recombinant His6-tagged D355R TRβ LBD was expressed in E. coli BL21(DE3) host (Invitrogen). Cells were grown at 37°C in LB medium supplemented with 100 μg/ml ampicillin until OD600~0.6. Expression was induced by addition of isopropyl-β-thiogalactoside (IPTG) to a final concentration of 0.5 μM, cells were grown for 12 hours at 15° C and harvested by centrifugation at 5,000 g for 15 min. The cell pellets from 1 L of culture were resuspended in 20 ml of buffer A (20 mM Tris-HCl pH 8, 300 mM NaCl, 50mM lithium sulfate, 25 mM imidazole, 3mM β-OG, 0.2 mM PMSF, 10% glycerol, 200 μg/ml lysozyme, 30 μM TRIAC), cells were lysed by sonication, and cell lysates centrifuged at 15,000 g for 30 min at 4°C. The supernatant was loaded onto a 5 ml Ni2+-affinity column (HiTrap chelating column, Amersham Biosciences) equilibrated with buffer A using the Äkta FPLC system (Amersham Biosciences). The column was washed with 20 volumes of buffer A. Bound proteins were eluted with buffer A containing 500 mM imidazole. Fractions containing eluted proteins were pooled and dialysed overnight against buffer B (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 50mM lithium sulfate, 10% glycerol, 5 mM DTT, 1mM EDTA) with two fold molar excess of TRIAC (Sigma). The protein was then further purified using a Superdex 75 16/60 size exclusion column (Amersham Biosciences) calibrated with globular standard proteins (gel-filtration calibration kit, Amersham Biosciences) and equilibrated with buffer B. The protein eluted in two distinct fractions corresponding to the TR homodimer and monomer. The main protein fraction corresponding to the TR homodimer (~80% of total protein) was concentrated and rechromatographed using the same size exclusion column.
The purified TRβ was concentrated to ~ 20 mg/ml, incubated with a two fold molar excess of TRIAC, and centrifuged for 20 min at 14,000 g prior to crystallization. Crystallization trials were performed by the hanging-drop vapor diffusion method at 17 °C using 24-well plates (Qiagen). Different commercial crystallization screens (Hampton Research, Emerald BioSystems, Qiagen) were used to determine the initial crystallization conditions. The protein solution (1 μl) was mixed with an equal volume of the reservoir solution and placed on the cover slips over the well containing 1 ml of reservoir solution. Several high ionic strength conditions containing ammonium sulfate (1.5 - 2 M) or sodium citrate (1.4 - 1.6 M) produced small needle-like crystals. The largest cluster of crystals was grown using 25% ethylene glycol as precipitant (condition No. 3, Crystal Screen 2, Hampton Research). These crystals appeared within 3-5 days and grew to their full size within a week at room temperature. The cluster was made of large thin blades that were separated into individual crystals of ~ 550 × 400 × 10 μm. These individual crystals were used for data collection.
Clustered crystals were reproducibly grown from either freshly prepared or frozen protein. Comparable quality data sets were collected using crystals obtained from either preparation.
Protein crystals were mounted directly from the drop containing mother liquor into cryo loops (Hampton Research) and flash-frozen in liquid nitrogen. X-ray diffraction data were measured at −180 °C and collected to 2.2 Å at the Advanced Light Source (Lawrence Berkeley National Laboratory) Beamline 8.3.1 (λ= 1.1 Å) using a single crystal. The crystal was rotated through 180° with an oscillation angle of 1° per frame and an exposure time of 3 s per image. Diffraction data were processed using DENZO and scaled with SCALEPACK (18). The crystal was of the monoclinic space group P21, with one homodimer in the asymmetric unit and cell dimensions of a=55.6 Å, b=93.0 Å, c=58.8 Å, and β=110.00. The solvent content of the crystal was ~46.3%, with Matthew's coefficient of ~2.4 Å3/Da.
The structure of the TR homodimer was determined by the molecular replacement method using the program package CNS and the structure of TRβ LBD (PDB ID 1NAX) as a search model. All water and ligand molecules as well as the C-terminal helix H12, which could adopt different conformations depending on the presence of a ligand, were removed from the search model. The cross-rotation search algorithm implemented in the CNS package produced two independent solutions, consistent with the presence of two LBDs in the asymmetric unit. The final solution for the LBD dimer had a monitor value of 0.555 and a packing coefficient of 0.45. The presence of a clear electron density for the two TRIAC molecules in the difference Fourier (Fo-Fc) maps for the protein model confirmed the molecular replacement solution. Additionally, the presence of continuous and well-defined electron density for helices H12 in the expected agonist position became apparent at the early stages of the structure refinement. The tetrahedrally shaped density observed at the dimer interface was attributed unambiguously to SO4− ion, due to the presence of sulfate in the buffer solutions.
The current model for the TRβ homodimer is refined to 2.2 Å resolution with Rfree/R values of 24.5/21.7 and contains 494 amino acid residues, 2 TRIAC molecules, 1 sulfate ion and 123 water molecules. Analysis of the stereochemistry of the model using PROCHECK (19) indicates that the main chain dihedral angles for all residues are located in the permitted regions of the Ramachandran diagram and that the root mean square deviations from ideal values are distributed within the expected range for a well refined structure. Data collection and refinement statistics are summarized in Table I. The structure has been deposited with the Protein Data Bank (PDB) and assigned the following ID Number: RCSB ID code rcsb047606 and PDB ID code 3D57.
In our experience, TR LBD is a delicate protein that crystallizes only when carefully prepared fresh. Because of the protein instability, TR requires many iterative protein preparations for obtaining and optimizing crystals suitable for high-resolution structural analysis. In addition, TR is unstable without a ligand, which is absolutely required during protein purification at all stages. When stabilized by a bound ligand, TR LBD is a monomer in solution, although it is shown to form dimers and trimers on DNA in vivo (4, 20). To date, there is no crystal structure of any TR self-assembly, however extensive mutational analyses and homology modeling suggested that TR LBD might employ helix 11 to form dimers (21). Independent studies showed that clustered charged surface residues in helices 7 and 8 might influence stability of the TR LBD and affect its dimerization on DNA and function on TR response elements (22). In particular, a single surface mutation D355R in helix 8 was shown to enhance TR homodimer formation on DNA in the presence of hormone (22). This observation was noteworthy because it revealed a curious improvement of protein stability without any obvious explanation. This mutation is remote from the predicted dimerization helix 11, therefore, the molecular mechanism of the assembly enhancement could not be explained by similarities with other NR homodimers (23, 24, 25).
Any novel protein assemblies provide useful information of higher order complex architecture (26). Moreover, homodimeric complexes are generally more stable and, therefore, amenable for protein crystallization (17, 27). In this work, we performed a detailed structural and functional analysis of D355R TR LBD and its dimeric assembly. Specifically, we determined the three-dimensional structure of this protein to visualize the mechanism of its dimerization. We complemented this structural study with analyses of the protein stability, behavior in solution and ultimately its utility for high throughput crystallographic techniques.
We found that in contrast to the wild type TR monomer, the mutant receptor LBD undergoes a ligand-dependent molecular assembly, forming a homodimer in addition to the monomer in certain preparative conditions. In our experiments, a typical size exclusion chromatography run for the purified D355R TR mutant protein resulted in separation of 60 kDa from 30kDa protein species, which correspond to the dimeric and monomeric states of TR LBD (Fig. 1A). Repeated chromatography of the main protein fraction corresponding to the TR homodimer (~80% of total protein) demonstrated that the formed dimer is stable under the experimental conditions (Fig. 1A). Altering the buffer composition such as removal of sulfate ions, resulted in loss of dimers, with only monomeric protein present in the chromatographic fractions (Fig.1B). In contrast, the TR homodimer formed very efficiently if sulfate ions were present in preparative buffers (Fig.1B). Other variations in protein buffers such as ionic strength, pH and the presence of detergents were also found to influence formation of the TR homodimer and the dimer/monomer partition in solution (data not presented). Based on these experiments, we found the protein preparation conditions that allow purification of the TR homodimer as a major protein fraction with high yield (specified in Experimental procedures). The ultimate step of protein purification employing FPLC was critical as it resulted in efficient separation of the dimeric protein from the TR LBD monomer, thus ensuring protein purity and homogeneity. After this final purification step, the TR homodimer was more than 95% pure as judged by gel electrophoresis. Furthermore, the D355R TR dimer was highly soluble and maintained its monodisperse dimeric state both at high (40mg/ml) and low (5mg/ml) concentrations as judged by native gels and size exclusion chromatography analyses. Notably, the TR LBD retained its dimeric state not only throughout protein preparation, but also after several freeze-thaw cycles, as judged by size exclusion chromatography. Based on these observations, we conclude that the dimeric form of the mutant increased stability of the TR LBD, which could tolerate 2-3 cycles of freezing – thawing and produce high quality crystals that diffracted to 2.0 - 2.5 Å resolution. The increased stability of TR LBD is a crucial improvement as wild type protein could not produce crystals upon storage or freezing. Furthermore, these improved protein and crystal properties are compliant with the demands of a high throughput crystallography (28, 29).
The structure of the TR homodimer was determined by the molecular replacement method and refined to 2.2 Å resolution with Rfree/R values of 24.5/21.7. The current model includes residues 210-460 and 211-460 for the first and the second TRIAC bound TR LBDs (Fig. 2A). The unstructured loop residues 255-260 (chain A) and residues 256 and 257 (chain B) are omitted from the model due to the lack of electron density. There are 123 water molecules and one coordinated sulfate ion in the TR homodimer structure. Data collection and refinement statistics are summarized in Table I.
The crystal structure of TR homodimer showed that D355R mutation affected neither the tertiary structure of the nuclear receptor LBD nor its AF-2 surface as judged by comparison of the two LBDs from the complex with available structure of monomeric TR (PDB ID 1NAX). The structure of TR mutant revealed one homodimer in the crystal asymmetric unit (Fig.2). The two monomers in the complex are related by two-fold non-crystallographic symmetry and are structurally very similar to each other, with r.m.s. deviation of only 0.23 Å for 169 Cα atoms. The dimer interface is extensive (~ 1330 å2) and is formed between mostly polar or charged residues (Table II). Hydrogen bonds link the side chains of the interfacial residues through six water molecules that are shared at the interface. Although mostly composed of polar and charged residues, the dimer interface is additionally stabilized by hydrophobic interactions (Table II). The residues forming contacts between monomers are mostly contributed from helix 8 and loops H6-H7 and H10-H11. There are two sites at the interface that are evidently crucial for the dimer formation. The first site includes a four-histidine cluster composed of H412 and H413 from the two TR LBDs. A single sulfate ion links the four His side chains within the cluster; His412 requires a water molecule to bridge to the sulfate ion (Fig.2B). The presence of the coordinated sulfate ion at the dimer interface is consistent with our biochemical data showing the necessity of sulfate for dimerization of the mutant TR LBD (Fig. 1B). The second site is composed of residue R338 and mutated residue R355 of each monomer that form hydrogen bonds to the oppositely charged E324 and E326 of the partner LBD (Fig.2B). In wild type TR, side chain of D355 forms salt bridge with the side chain of R338, which is also hydrogen bonded with the side chain of D351. Thus R338, D351 and D355 are charged residues of a solvated 3 side chain intramolecular cluster. In the dimer, the polar interaction network is intermolecular.
Based on the data described here, we conclude that with the strategically placed single surface mutation D355R we engineered a symmetrical dimer of TR LBD with an interface composed mostly of polar amino acid side chains. The polar side chains are partly solvated, the dimer interface contains 6 water molecules (Fig.2C). The interface has almost no significant hydrophobic interactions, with only 10% of the contacts formed between paired nonpolar atoms.
The mutant TR is characterized by the increased protein stability and crystallizability and thus is a good tool for crystal based screenings employing high throughput soakings of surface binding small molecules. We can envision other nuclear receptors being amenable for dimer engineering, making this approach a valuable technique for obtaining stable protein tools for high resolution crystallography. Predicting single site mutants that would form stable protein dimers is recognized to be challenging. The results presented here suggest that appropriate stabilizing residues for the protein of interest are better found by experiment with a biochemical assay for stability.
We thank James Holton and George Meigs for assistance with data collection at the Advanced Light Source 8.3.1. Beamline (University of California Berkeley).
Funding: project was supported by NIH R01 DK51281 and DK41842 (J. Baxter), and RO1 DK058080 (R. Fletterick and R. Guy).