Expression, purification and primary crystallographic study of human androgen receptor in complex with DNA and coactivator motifs

Abstract

The androgen receptor (AR) is a DNA-binding and hormone-activated transcription factor that plays critical roles in the development and progression of prostate cancer. The transcriptional function of AR is modulated by intermolecular interactions with DNA elements and coactivator proteins, as well as intramolecular interactions between AR domains; thus, the structural information from the full-length AR or a multi-domain fragment is essential for understanding the molecular basis of AR functions. Here we report the expression and purification of full-length AR protein and of a fragment containing its DNA-binding and ligand-binding domains connected by the hinge region in the presence of its natural ligand, dihydrotestosterone. Crystals of ligand-bound full-length AR and of the AR fragment in complex with DNA elements and coactivator motifs have been obtained and diffracted to low resolutions. These results help establish a foundation for pursuing further crystallographic studies of an AR/DNA complex.

Introduction

The androgen receptor is a DNA-binding transcription factor that belongs to a large superfamily of evolutionarily related nuclear hormone receptors [1, 2]. Like all members of this family, AR contains a modular domain arrangement comprising an N-terminal, constitutively active activation function domain (AF-1, or A/B region); a central zinc-finger DNA-binding domain (DBD, or C); a C-terminal ligand-binding domain (LBD or E) containing the C-terminal flanking with a hormone-dependent activation function region (AF-2); and a hinge region (D) connecting the DBD and LBD [3] (Figure 1A).

Figure 1.

A, schematic diagram of the androgen receptor amino acid sequence, including the N-terminal domain with AF-1 (A/B), the DNA binding domain (C), the hinge region (D), and the ligand binding domain with the C-terminal AF-2 motif (E). B, upon ligand binding, AR translocates to the nucleus and binds to the androgen response elements (ARE); coactivators and the transcriptional machinery are then recruited to this complex to initiate gene transcription.

In the cytoplasm, AR is stabilized by the binding of chaperone proteins, including heat shock protein 90 (hsp90) and p23 [4, 5]. Upon a ligand binding to its LBD, conformational changes allow AR to dimerize, translocate to the nuclei, and then bind to specific DNA sequences (termed androgen response elements, AREs) through its DBD. Coactivators and the transcriptional machinery are then recruited to this protein/DNA complex to initiate gene transcription [6–8] (Figure 1B).

The androgen receptor regulates the expression of genes required for the differentiation and function of the male reproductive system, including the prostate gland. AR has been extensively investigated in the last few decades because of its roles in the development and progression of prostate cancer [9, 10]. Prostate cancer development is closely associated with gain-of-function mutations that convert AR from a normal gene regulator to a hyperactive transcriptional activator in prostate cancer cells [11, 12]. In later-stage prostate cancer, AR undergoes additional molecular changes that allow it to activate genomic or non-genomic signaling events even in the absence of (or at very low circulating levels of) androgen [13]. Although small-molecule AR antagonists have been successfully used in the treatment of early-stage prostate cancer, no therapy has been really successful in treatment of late-stage hormone-independent prostate cancer. The development of new therapies and the design of new drugs for treating late-stage prostate cancer has been a major challenge for medical science.

AR is the major target for prostate cancer drug discovery and the focus of many efforts to solve the atomic structures and to elucidate the structure-function relationships of AR. The crystal structures of an agonist-bound ligand-binding domain [14, 15] and the DNA-binding domain in complex with AREs have been published [15]. The crystal structures of the LBD in complex with agonists or antagonists provide useful information that facilitates drug design targeting at the AR LBD and its ligand-binding pocket. It is clear, however, that AR activation and function are modulated by a range of interactions between AR, co-regulators, and DNA elements involving different domains throughout the full-length receptor. A range of intramolecular interactions between the multiple domains of AR is also important for the proper function of the receptor. Understanding the interdomain and intermolecular interactions of the receptor is crucial for elucidating the mechanism of AR transcriptional function and for better design of AR-targeted drugs against prostate cancer.

Here we report the expression, purification, and crystallization of a DHT-bound full-length AR protein in complex with DNA oligomers derived from the ARE sequences, as well as a fragment of human AR containing the DBD, the hinge region, and the LBD in complex with DNA elements and coactivator motifs.

Materials and methods

Gene cloning and plasmid constructs

The construction of the recombinant shuttle vector pDW464/hAR and recombinant baculovirus for expression of full-length AR in Sf9 insect cells has been described by Juzumiene, Chang et al. [16]. The AR fragment (amino acids 556–919) that contains DBD, the hinge region, and the LBD (regions C, D, and E, Fig. 1) was cloned into pET24a between NdeI and XhoI restriction sites with a His₆-GST tag at the N-terminus, connected by a 6-residue thrombin digestion sequence.

Expression and purification of full length AR

The full-length AR protein (ARFL) was expressed in Sf9 insect cells and purified using a streptavidin mutein column as described by Juzumiene, Chang et al. [16]. Briefly, Sf9 cells were grown in serum-free media to a density of 2 ×10⁶ cells/ml and were infected with baculovirus containing an AR expression construct at a multiplicity of 0.5 to 1. After shaking at 27 °C for 24 h, 5α-dihydrotestosterone (DHT) and zinc acetate were added to final concentrations of 50 μM and 10 μM, respectively. Protein expression was allowed to continue for another 48 h. A weight of 44 grams of wet Sf9 cells was obtained from 1000 ml cell culture. The Sf9 cells were re-suspended in 100 ml of ice-cooled hypotonic buffer (10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 1.5 mM MgCl₂, 10 mM β-mercaptoethanol, 1 μM zinc acetate, 1 μM DHT and 1/200 mL protease inhibitor cocktail set III (Calbiochem)) and transferred into a Dounce homogenizer and lysed with 25 strokes. After centrifugation at 7000Xg for 10 min, the cytoplasmic supernatant was discarded and the nuclear pellet was soaked by slow shaking for 30 min at 4°C with 20 ml of nuclear extract buffer (10 mM Tris-HCl, pH 8.0, 0.3 M NaCl, 10 mM β-mercaptoethanol, 50 mM β-glycerolphosphate, 50 mM NaF, 1 μM zinc acetate, 1 μM DHT and 1/200 (volume) protease inhibitor cocktail set III), and centrifuged at 50,000Xg for 30 min. The nuclear extract supernatant was loaded to a streptavidin mutein column (Roche Applied Science), washed with 50 ml of washing buffer (10 mM Tris-HCl, pH 8.0, 0.15 M NaCl, 10% glycerol, 10 mM β-mercaptoethanol, 1 μM zinc acetate and 1 μM DHT), and eluted with 30 ml of elution buffer (5 mM biotin with 50 mM Tris-HCl, pH 8.0, 0.2 M NaCl, 10% glycerol, 4 mM DTT, 1 μM zinc acetate and 1 μM DHT). The elution fractions were pooled and further purified using a 300 ml Superdex S200 gel filtration column in a buffer of 20 mM Tris-HCl, pH 8.0, 0.2 M NaCl, 10% glycerol, 10 mM DTT, 10 μM zinc acetate and 10 μM DHT. The yield after the final polishing gel filtration column was 0.7 mg from 1000 ml of Sf9 cell culture.

Expression and purification of AR CDE

The AR CDE fragment (amino acids 556 to 919, ARCDE) was expressed as a His₆-GST fusion protein from the expression vector pET24a in BL21 (DE3) E. coli cells. A concentration of 10 μM of both DHT and zinc acetate was used during protein expression and purification. For a typical purification, 20 grams of wet cells from 6 liters of cell culture were resuspended in 200 ml of lysis buffer (20 mM Tris-HCl pH 7.3, 1.5 M urea, 0.2 M NaCl, 10% glycerol, 10 μM zinc acetate, and 10 μM DHT) and lysed using French press and centrifuged at 20000Xg. The supernatant of the E. coli cell lysate was loaded on a 40 ml Ni-NTA column saturated with the lysis buffer, washed with 500 ml of wash buffer (20 mM Tris-HCl pH 8.0, 0.2 M NaCl, 10% glycerol, 10 μM zinc acetate, 10 μM DHT, and 20 mM imidazole), and then eluted with 200 ml elution buffer (wash buffer with 250 mM imidazole). The pooled peak fractions were dialyzed against buffer A (20 mM Tris-HCl pH 8.0, 0.2 M NaCl, 10% glycerol, 10 μM zinc acetate and 10 μM DHT) with 3 changes to remove imidazole and were digested in cold room using thrombin at 1/1000 weight ratio of thrombin to ARCDE fusion protein to cleave the His₆-GST tag. The digestion mixture was loaded on a 10 ml Ni-NTA column, washed with buffer A and eluted using a shallow gradient of 10 mM to 50 mM imidazole in buffer A. The His6-tag free ARCDE protein was eluted at a low concentration (about 30 mM) of imidazole because of its low affinity to Ni-NTA and was separated from the His6-GST tag that was strongly bound to the Ni-NTA resins and needed 100–250 mM imidazole to elute. The protein sample was further purified using a 300 ml Superdex S200 gel filtration column in a buffer of 20 mM Tris-HCl, pH 8.0, 0.2 M NaCl, 10% glycerol, 5 mM DTT, 10 μM zinc acetate and 10 μM DHT and then was concentrated to about 5 mg/ml for crystallization.

Design and synthesis of AR response element-containing DNA

DNA fragments containing the AR response elements were designed (Table 1) and synthesized using an AB3400 DNA synthesizer (Applied Biosystems). The crude oligos were de-protected by concentrated ammonium hydroxide at 55 °C overnight, lyophilized to remove the ammonium hydroxide, and resuspended in 10 mM triethylammonium acetate buffer at pH 7.0. The resuspended crude oligo was loaded on a Dynamax-300A reverse phase column, washed with 10% acetonitrile in water to remove failure oligos, and detritylated by washing with 0.125% trifluoroacetic acid. The target oligo was eluted using 8–9% acetonitrile in water, lyophilized overnight and resuspended in 10 mM triethylammonium bicarbonate buffer at pH=8.0. The complementary oligos were annealed in the same buffer to form a DNA duplex by keeping the oligos in a 95 °C water bath for 5 min and slowly cooling to room temperature.

Table 1.

Sequences of ARE-derived DNA duplexes used for crystallization of the ARFL and ARCDE complexes, and their crystallization results from the initial screen

DNA	Length	Sequence	Crystals of complexes
			ARFL	ARCDE
1	16mer	AGA ACA TGA TGT TCT G	No	Yes
		TCT TGT ACT ACA AGA C
2	18mer	GAG AAC ATG ATG TTC TTG	Not determined	Yes
		CTC TTG TAC TAC AAG AAC
3	20mer	TTG GGT ACA GAG TGT TCT AG	Yes	Yes
		C CCA TGT CTC ACA AGA TC AA
4	21mer	GGA AGA ACA CTG TGT TCT TGT	Yes	Yes
		T TCT TGT GAC ACA AGA ACA CC
5	26mer	AAT GAG AGA ACA TAG TGT TCT TGG GT	Yes	Yes
		A CTC TCT TGT ATC ACA AGA ACC CA TT
6	29mer	AAG TTG AGA GTA CAT AGT GTT CTT GGG TA	Yes	Yes
		C AAC TCT CAT GTA TCA CAA GAA CCC ATT T

Alpha Screen peptide profiling

The binding of various peptide motifs to AR was determined by Alpha Screen assay using a hexahistidine detection kit from Perkin-Elmer. AR CDE proteins were prepared as His6-GST fusion proteins for the assays. AR coactivator motif peptides were designed to have 15 residues with a central fragment of LxxLL or FxxLF (x is for any amino acid) flanked by six and four residues at the N-terminus and the C-terminus, respectively (Table 2), and synthesized by SynBioSci Corporation. The experiments were conducted with 20 nM ARCDE in complex with R1881 and 20 nM of biotinylated SRC2–3 peptide (Table 2) in the presence of 5 μg/ml donor and acceptor beads in a buffer containing 50 mM MOPS, 50 mM NaF, 50 μM CHAPS, and 0.1 mg/ml bovine serum albumin, all adjusted to a pH of 7.4.

Table 2.

Coactivator peptides used for the AR binding screen

1	SMRT-2 (corepressor)	ASTNMG LEAII RKALMGKYDQ
2	SHP1	ASHPTI LYTLL SPGP
3	SHP2	APVPSI LKKIL LEEPNS
4	SHP3	ASQGR LARIL LMAST
5	DAX1	QWQGSI LYNML MSAK
6	DAX2	PRQGSI LYSML TSAK
7	DAX3	PRQGSI LYSLL TSSK
8	SRC1–1	SQTSHK LVQLL TTTA
9	SRC1–2	TERHKI LHRLL QESS
10	SRC1–3	SKDHQL LRYLL DKDE
11	SRC1–4	AQQKSL LQQLL TE
12	SRC2–1	SKGQTK LLQLL TTKS
13	SRC2–2	KEKHKI LHRLL QDSS
14	SRC2–3	KENAL LRYLL DKDD
15	SRC3–1	SKGHKK LLQLL TCSS
16	SRC3–2	QEKHRI LHKLL QNGN
17	SRC3–3	KENNAL LRYLL DRDD
18	TRAP220–1	VSQNPI LTSLL QITG
19	TRAP220–2	KNIHPM LMNLL KDNP
20	CBP-1	ASKHKQ LSELL RGGS
21	PGC1A-1	AEEPSL LKKLL LAPA
22	PGC1A-2	RRPCSE LLKYL TTND
23	PGC1B-1	VDELSL LQKLL LATS
24	PGC1B-2	WAEFSI LRELL AQDV
25	PRC	PREGSS LHKLL TLSR
26	ARA70–1	QQQAQQ LYSLL GQFN
27	ARA70–2	RETSEK FKLLF QSYN
28	ASC2–1	TLTSPL LVNLL QSDI
29	ASC2–2	REAPTS LSQLL DNSG
30	RIP140–2	KQDSTL LASLL QSFS
31	RIP140–9	SKSFNV LKQLL LSEN
32	PRIC285–1	NADDAI LRELL DESQ
33	PRIC285–2	NLPPAA LRKLL RAEP
34	PRIC285–3	FAGDEV LVQLL SGDK
35	ARN1	YRGA FQNLF QSVR
36	ARN2	ASSS WHTLF TAEE
37	4–1	QPKH FTELY FKS

Crystallization and crystallographic analysis of ARFL and ARCDE complexes

The purified DHT-bound ARFL was complexed with DNA oligos containing AR response elements (shown in Table 1) at a protein/DNA ratio of 1:1.2. DHT-bound ARCDE was complexed with both DNA oligos and the first LxxLL motif from coactivator SRC3 (SRC3-1, Table 2) at molar ratios of 1:1.2 and 1:1.5, respectively, before setting up crystallization trays. Crystallization was performed with 24-well crystallization trays using the hanging drop vapor diffusion technique at 20 °C. Crystallization conditions were initially obtained from screening using Hampton kits and home-made recipes in the presence of DHT, zinc acetate, and spermine, and were optimized from the initial conditions. ARFL and ARCDE crystals were flash-frozen in liquid nitrogen. As the reservoir solutions contained sufficient polyethylene glycol and 2-methyl-2,4-pentanediol, additional cryoprotectants were not used. Crystals were first analyzed with both an in-house source at 93 K using Cu K radiation from a Rigaku MicroMax-007 rotating-anode X-ray generator with a Mar 345 detector, and then with synchrotron X-ray beams at the DND-CAT and LS-CAT at the Advanced Photon Source (APS), Argonne National Laboratory, Chicago, IL.

Results

Expression, purification, DNA complexes, and crystallization of the full-length AR

The androgen receptor is a multi-domain protein of 919 amino acid residues and is usually post-translationally modified for proper function. Bacterial expression of AR has disadvantages, including low solubility and stability because of the lack of post-translational modification [17]. Juzumiene et al. [16] successfully expressed full-length AR containing an N-terminal biotin tag in Sf9 insect cells and purified the protein using a streptavidin mutein column [16], making functional and structural studies possible from the large amount of homogenous protein produced. Further purification by gel filtration produced a higher quality protein sample which benefited protein crystallization (Figure 2A).

Figure 2.

Protein expression, purification and crystallization. A, Coomassie-stained SDS gel of purified ARFL. The label on the SDS gel lanes: M, molecular weight marker; 1, crude extract from Sf9 cell nuclei; 2–4, elution fractions from the streptavidin mutein column. B and C, ARFL crystals. D, Coomassie-stained SDS gels for purification of ARCDE. The labels of the SDS gel lanes: M, molecular weight marker; 1, ARCDE His₆-GST fusion protein eluted from the first Ni-NTA column; 2, fusion protein digested with thrombin; 3–5, ARCDE eluted from the second Ni-NTA column with His₆-GST removed. E and F, ARCDE crystals.

AR protein, once induced by androgen, forms homodimers by N/C interaction (the interaction between the N-terminal FxxLF motif and the coactivator binding cleft at the LBD), translocates to the nucleus and binds to DNA sequences containing AR response elements GGTACAXXXTGTTCT (X is for any nucleotide) or variants with the first three nucleotides replaced by AGT or AGA. [18]. As shown by the crystal structures of nuclear receptor DBD-DNA complexes from PDB bank, the DNA length is critical for successful crystallization and structure determination. We designed and synthesized DNA duplexes containing the ARE and zero to eight nucleotides flanking both ends, which made the length of the DNA duplexes from 15 to 31 nucleotides (Table 1).

The purified, DHT-bound full-length AR was complexed with the DNA duplexes listed in Table 1 before crystallization. Crystallization screen was performed by using commercially available screen kits and laboratory-developed recipes. Microcrystals of DHT-bound full-length AR in complex with DNA fragments 3, 4, 5, and 6 were obtained (Table 1). After much refinement and optimization of the initial conditions, the complex crystals with dimensions of about 150 μm X 50 μm X 50 μm appeared in 2–3 weeks under conditions of 30% PEG 2000, 100 mM TrisHCl, pH 7.0, 20 μM zinc acetate, and 10 mM spermine (Figure 2B&C).

Construct design, expression, and purification of the ARCDE

A crystal structure of the ligand-binding domain together with the DNA-binding domain and the hinge region will provide more molecular information than a crystal structure of any single domain, because AR transactivation is modulated by intramolecular (domain-domain) and intermolecular interactions and recruitment of co-regulators. The amino acid sequence of the AR CDE region (Figure 3A) is more conserved among different species than that of the N-terminal domain; it has a stable secondary structure except for the hinge region, which is predicted to be dominantly a loop region (Figure 3B). As an alternative to the full-length AR, crystallization and structure determination of ARCDE are expected to be more successful. ARCDE was bacterially expressed as a fusion protein with a His₆-GST tag at amino terminus connected by a 6-amino acid linker, LVPRGS, which contains the thrombin digestion site. The His6-GST tag is for correct folding and stabilization of the ARCDE protein, rather than for purification purpose. Protein expression was induced at a low concentration of isopropyl β-D-1-thiogalactopyranoside (IPTG, 0.1 mM) and low temperature (16 °C) in order to reduce aggregation and to increase the solubility and stability of the protein. The addition of the AR ligand DHT and zinc acetate to the cell culture was to further stabilize the folding of the ligand-binding domain and the DNA-binding zinc fingers, respectively. Including 1.5 M urea in the buffer for bacteria lysis is for stripping bacteria chaperones from the target protein and making the downstream purification easier. A total of 400 mg of His6-GST tagged ARCDE was recovered from 6 L of cell culture after a Ni-NTA chromatographic column as the first step of purification (Figure 2D). After an overnight dialysis against a imidazole-free buffer and digestion with thrombin and passage through a second Ni-NTA chromatographic column, about 40 mg of ARCDE protein was obtained (Figure 2D). A widely used strategy for removing His or His-GST tag is to run the digested sample through a Ni-NTA column thus the tag is hold up on the column and the target protein goes out in the flow-through. Our approach is to load the digested and dialyzed (imidazole free) sample to a second Ni-NTA column, wash the column with a imidazole-free buffer and elute the protein with low concentration (about 20–40 mM) of imidazole. The second Ni-NTA column further removed contaminant proteins and other impurities and concentrated the protein sample. We have not found any published similar example, but this strategy is suitable for most protein purification because most proteins have a few histidine or other nickel binding amino acid residues on surface, which bind to Ni-NTA with an affinity much less than that of a His tag. About 85% of ARCDE protein was lost upon dialysis, thrombin digestion and the second Ni-NTA column due to protein truncation and non-specific thrombin digestion. Further purification by gel filtration removed aggregated, truncated or mis-folded protein. The final yield of high quality protein after the gel filtration was about 1.5 mg of ARCDE per liter of bacterial culture. The whole process of the purification needs to be done as soon as possible, typically in two days, to reduce the loss of protein by degradation. The newly purified ARCDE was protected from further degradation by adding coactivator peptides and frozen with liquid nitrogen and stored at −80° for characterization and crystallization. Although the yield of crystallization quality protein remains low, it is enough for us to performed crystallographic studies and functional assays.

Figure 3.

A, Multiple sequence alignment of the AR CDE region from various species. The color code for residues: blue, positively charged; red, negatively charged; green, hydrophobic and grey, polar residues. B, secondary structure assignment (a strand marked as an arrow, a helix as a bar, and a loop or turn as a line) of human ARCDE based on the published AR DBD and LBD structures; the region highlighted in yellow is the hinge region between the DBD and LBD, which has no structural information and is predicted as a loop region.

Identification of AR coactivator motifs and crystallization of the ARCDE/DNA/coactivator complexes

AR coactivators, including the members of the steroid receptor coactivator family (SRC), are proteins containing multiple LxxLL or FxxLF motifs that are able to bind to the coactivator-binding cleft at the AR LBD. More than sixteen coactivators have been reported to be active in AR transcription activation [19]. In order to identify high-affinity AR binding motifs, we performed an Alpha Screen peptide profiling [20] and found that SRC3-1, ARA70-2, and ARN-1 have the highest affinity to the AR LBD among a total of 23 motifs from previously reported AR coactivators (Figure 4). The peptides used for the AR binding screen are listed in Table 2 and the result is shown in Figure 4.

Figure 4.

Binding profile of peptide motifs to the AR CDE in complex with AR agonist R1881 in Alpha Screen assay. Various unlabeled peptides at identical concentrations of 500 nM were used to compete off the binding of biotinylated SRC2-3 LXXLL motif to AR. No peptide (NONE) and 50 μM SRC3-2 were used for negative and positive controls, respectively. The result shown is the average of triplicate experiments, with error bars showing standard deviations. The SRC3-1 and the ARA70-2 motifs show the highest affinity for the AR CDE/R1881 complex as determined by the peptide competitions.

The SRC3-1 peptide is the first nuclear receptor-binding motif (LLQLL) of the steroid receptor coactivator 3 (SRC3), flanked on one side with 6 and on the other with 4 amino acid residues. DHT-bound ARCDE was first complexed with coactivator motif SRC3–1, and then with DNA duplexes from Table 1 in the same way as for full-length AR. DHT-bound ARCDE, upon addition of SRC3-1, can remain without obvious degradation for more than one week in a solution at 4 °C (data not shown). The protein-DNA complex alone lose about 80% by degradation in two to three days at 4 °C, implying that coactivator binding largely stabilizes the conformation of the DHT-bound AR LBD.

The ARCDE/DHT complex, upon binding to both the coactivator motif and DNA, is ready for crystallization screen. The initial screen was performed by using commercial screen kits and laboratory-developed recipes. Tiny crystals were obtained from DHT- and SRC3-1-bound ARCDE in complex with any of the 6 DNA duplexes under various crystallization conditions. But complexes using the DNA 1, 3, or 6 fragments proved to produce bigger crystals under optimized conditions. The best ARCDE complex crystals were obtained in about one week when 2 μL of ARCDE/DHT/SRC3-1/DNA6 complex at a concentration of 5mg/ml of ARCDE, was mixed with 1 μL of reservoir solution containing 0.1 M MES pH 5.8–6.0, 25% 2-methyl-2,4-pentandiol, 20 mM spermine, 10 μM zinc acetate, and 5 mM DTT, and was equilibrated against 500 μL of reservoir solution (Figure 2E–F).

Initial X-ray crystallographic analysis of crystals of full-length AR and ARCDE

Both full-length AR and AR CDE crystals diffracted poorly. The best crystal of the AR CDE/DHT/SRC3-1/DNA6 complex diffracted to about 6 Å with in-house rotating anode X-ray source and about 5 Å with synchrotron X-ray beam along one axis and worse in other directions. The crystals of full-length AR in complex with DNA6 had diffraction patterns of about 12 Å resolution with the in-house X-ray source and about 8 Å with synchrotron source in one direction and did not have signal in other directions (data not shown). It seems these crystals were heavily anisotropic with one direction packed nicely and another badly because the C-terminal ligand binding domain of the protein is nicely folded and the N-terminal domain is largely unstructured.

Discussion

The major challenge for expressing a multi-domain protein like full length AR or AR CDE in a bacteria system is to stabilize the protein folding and keep it soluble and functional during expression. Our experiment showed that the expression level of soluble protein was greatly increased when the hormone DHT was included in the bacteria culture. Same situation happens to the expression and purification of many other nuclear hormone receptors, such as glucocorticoid receptor [21], mineralocorticoid receptor [22] and vitamin D receptor [23]. In in vivo expression, nuclear receptor proteins are stabilized by protein chaperones when they are first synthesized, and stabilized by ligand binding, coactivator interaction, and/or DNA binding when they are translocated to nucleus. Including ligand in the expression system and the purification steps is an effective strategy for improving protein solubility and increasing soluble protein production for many ligand binding proteins including enzymes and nuclear receptors [24]. Using this strategy, we significantly increased our protein yield and obtained a sufficient amount of AR proteins for characterization and crystallization trials.

Binding to coactivator peptides is another strategy for stabilizing AR CDE protein during purification and storage. ARCDE is less stable than the full length AR because of its lack of the N/C interaction. It is demonstrated that AR forms N/C interaction at the absence of coactivator until a coactivator competitively binds to AR and releases the N-terminal hydrophobic motif of AR from binding to its own coactivator binding cleft. Coactivator binding induces conformational change in the receptor thus facilitates the formation of transcription machinery by networking AR to basal transcription factors. Coactivator peptides containing a nuclear receptor binding motifs (typically LxxLL) are widely used for stabilizing nuclear receptor LBDs for structure determination [25–27]. We demonstrated that coactivator peptides with high binding affinity to AR worked better than those with relatively low binding affinity for receptor stabilization. In addition, our experiments showed that the coactivator motif not only stabilizes the LBD domain but the multi-domain fragment AR CDE as well, implying that domain-domain interactions may occur among the CDE fragment upon coactivator binding.

Much effort went into optimizing the crystallization conditions in order to obtain diffraction-quality crystals, but the quality of both ARCDE and ARFL crystals remains unsatisfactory. Further work is clearly needed on the crystallization conditions. Successful crystallization of signaling proteins and their complexes, including full-length nuclear receptors, is a major challenge for protein crystallographers. Most signalling proteins, especially those from mammals (including human), contain highly flexible or even unfolded fragments that dramatically interfere with crystallization. Modifications, including surface mutagenesis to reduce surface entropy and remove flexible loops, are usually needed for such proteins to be successfully crystallized, but those changes may lead to loss of some (perhaps crucial) biological information. A recent report on the crystallization and structure determination of a near full length nuclear receptor heterodimer, PPARγ/RXRα, with DNA elements [28] is a breakthrough and encourages those working in this area.