|Home | About | Journals | Submit | Contact Us | Français|
The DNA binding domain (DBD) of nuclear hormone receptors contains a highly conserved globular domain and a less conserved carboxyl-terminal extension (CTE). Despite previous observations that the CTEs of some classes of nuclear receptors are structured and interact with DNA outside of the hexanucleotide hormone response element (HRE), there has been no evidence for such a CTE among the steroid receptors. We have determined the structure of the progesterone receptor (PR)-DBD-CTE DNA complex at a resolution of 2.5 Å, which revealed binding of the CTE to the minor groove flanking the HREs. Alanine substitutions of the interacting CTE residues reduced affinity for inverted repeat HREs separated by three nucleotides, and essentially abrogated binding to a single HRE. A highly compressed minor groove of the trinucleotide spacer and a novel dimerization interface were also observed. A PR binding site selection experiment revealed sequence preferences in the trinucleotide spacer and flanking DNA. These results, taken together, support the notion that sequences outside of the HREs influence the DNA binding affinity and specificity of steroid receptors.
The nuclear receptor superfamily is composed of ligand-dependent transcription factors that regulate a variety of cellular processes including metabolism, development, growth, differentiation, and reproduction. Nuclear receptors are modular proteins containing a conserved DNA binding domain (DBD), a C-terminal ligand binding domain (LBD), and a diverse amino-terminal region (1, 2). Transcriptional regulation by nuclear receptors primarily occurs through direct binding to specific hormone response elements (HREs) of target genes. The nuclear receptor superfamily can be subdivided based on distinct DNA binding mechanisms. Class I receptors for steroid hormones interact optimally as head-to-head homodimers with hexanucleotide HREs arranged as inverted repeats with a trinucleotide spacer (3N) between the half-sites (2–4). There are two hexanucleotide HREs for the steroid receptors: AGGTCA is preferred by estrogen receptor (ER), as well as by the class II and orphan receptors, and AGAACA is preferred by the glucocorticoid receptor (GR) subfamily, which also includes androgen receptor (AR), mineralocorticoid receptor (MR), and progesterone receptor (PR). Class II receptors for the nonsteroidal hormones and dietary lipids such as retinoic acid receptor (RAR), thyroid hormone receptor (TR), vitamin D3 receptor (VDR), and peroxisome proliferator-activated receptor (PPAR), bind primarily as heterodimers together with retinoic acid X receptor (RXR) to HREs arranged as direct repeats with variable spacing between the half-sites (1). Orphan receptors, for which physiological ligands are unidentified, interact through various configurations including monomer binding to extended HRE half-sites (1, 4–9).
The DNA binding regions of the nuclear receptors share a core DBD, consisting of two α helices and two asymmetric zinc binding modules coordinated by eight conserved cysteine residues, followed by a less conserved region termed the C-terminal extension (CTE). The first helix of the core DBD inserts into the major groove of the HRE, making base-specific contacts, and helix 2 crosses over the top of helix 1 and the major groove leading into the CTE. The dimerization domain (D box) resides within the C-terminal zinc-binding module and is important for mediating DNA binding-dependent dimerization of receptors (8, 10–16).
Unlike the core DBD, the CTE is not well conserved in sequence or structure. Structural studies of class II and orphan receptors have shown that the CTE, which has a highly variable length of up to 40 amino acids beyond the conserved Gly-Met sequence at the C terminus of the core DBD (see Fig. 3D), provides additional contacts with the DNA minor groove outside of the canonical HREs. The CTEs of TR and VDR form a third α-helix that projects across the minor groove to make extensive contacts with the phosphate backbone of spacer nucleotides. The CTEs of TR and VDR are also important for DNA-dependent homo- and heterodimerization, as well as for proper spacing of receptor subunits between direct-repeat response-element half-sites (12, 17). The CTEs of the orphan receptors [RevErb, nerve growth factor-I-β, liver receptor homolog-1, and estrogen-related receptor-2 (ERR2)] form an extended loop that occupies the minor groove of DNA sites flanking the HRE (8, 9, 18, 19). In addition to the extended loop the liver receptor homolog-1 DBD contains a non-DNA binding α-helix termed the Ftz-F1 motif (19). Conformational changes in the CTE have been observed in response to DNA binding. The RXR CTE is an alpha helix in solution but forms an extended loop conformation when bound to DNA, whereas the CTE of ERR2 is unstructured in solution and becomes structured when bound to DNA (9, 11, 13–15). Biochemical and mutagenesis studies have also shown that the CTE is required for high affinity DNA binding by class II and orphan nuclear receptors (5, 11, 20–24). In particular, the CTE is important for the DNA binding of orphan receptors that interact as monomers with a half-site HRE. The CTE effectively extends the protein-DNA interface, increases the stability of monomer binding to HRE half-sites, and creates a preference for 5′ trinucleotide sequences flanking half-site HREs.
In contrast to class II and orphan receptors, much less is known about the steroid receptor CTEs. In previous studies, we showed that the CTE of ER and PR is required for interaction with the high mobility group protein-1 and -2 (HMGB-1/-2) that function as coregulatory proteins for all members of the steroid class of receptors by facilitating receptor binding to specific DNA target sites (25–32). Similar to the orphan receptors, the CTE of ER and PR was also required for efficient binding of ER and PR to HRE half-sites (31–33). However, structural studies of steroid receptors DBDs including GR, AR, and ER, either lacked sufficient CTE sequence in the protein expression constructs, or the CTE was disordered in the crystal structures (34–36). Furthermore, there are no reports of structures of a PR-DBD/DNA complex.
To understand the role of the CTE in PR-DNA binding, we determined the crystal structure of the PR DBD including the CTE bound to consensus inverted-repeat progesterone response elements (PREs). Arg637 and Lys638 of the CTE contact the DNA minor groove at sites flanking the PREs. Alanine substitutions of Arg637 and Lys638 reduced binding affinity of PR to the inverted repeat PREs and abrogated binding of PR to a half-site PRE. Other novel structural features of the PR-DBD DNA complex included interactions of the subunits with the PRE consensus sequence bound by the opposite subunit, which coincided with a narrow minor groove in the 3N spacer DNA. These studies further our understanding of steroid receptor-DNA recognition mechanism and suggest that minor groove DNA interactions outside the HREs contribute to binding affinity and sequence selectivity.
To examine the role of the CTE in PR-DNA binding, the PR DBD including the sequence for the CTE was expressed, purified, and cocrystallized with PRE DNA. The PR-DBD670 [amino acids (aa) 562–670] was designed to have a CTE of the same length as TR in the crystal structure of TR-DBD-CTE (12). However, purification of the bacterially expressed PR-DBD670 polypeptide resulted in a mixture of two distinct species, the intact PR-DBD670 and a C-terminal proteolytic degradation product determined to be predominantly aa 562–648 (Fig. 1). Double-stranded oligonucleotides that were tested in crystallization trials contained two inverted repeat consensus PREs separated by a variable 3N spacer sequence and variable 5′ and 3′ flanking sequences. The stoichiometric ratio, at which a maximum amount of PR-DNA complex formed with a minimum of free protein, was determined by stoichiometric EMSA using concentrations of protein and DNA equivalent to crystallization conditions. Crystals of the protein-DNA complex were obtained at a molar protein/DNA ratio of 1.95:1 with the smaller PR DBD proteolytic product and a double stranded DNA fragment of 18 nucleotides (designated PRE_2C7A), which contained a 3N spacer derived from the mouse mammary tumor virus promoter and a one base overhang on each side of the PREs (Fig. 1).
The structure of PR-DBD-DNA complex was determined by x-ray crystallography to a resolution of 2.5 Å. The structure was solved by molecular replacement using a homology model of the PR-DNA complex (Donham, D. C., and M. E. A. Churchill, unpublished results) that was based on the ER DBD-estrogen response element (ERE) complex structure (35). Data collection and refinement statistics are presented in Table 1. The final model contains one homodimer of PR-DBD (subunit A includes aa 562–640 and subunit B includes aa 562–638), associated zinc ions and solvent molecules, and one DNA duplex, which forms a pseudocontinuous helix with a base triple at the junction between adjacent asymmetric units. The base triples create a discontinuous phosphate backbone and slight inclination of the bases, which appears to promote additional hydrogen bonding between adjacent base pairs and a narrowing of the minor groove at the 5′ and 3′ ends of the DNA.
The overall structure of the PR-core DBD was similar to that of other nuclear receptors. The PR-DBD binds to the response element as a head-to-head dimer (Fig. 2), and the recognition helix (helix 1) of each monomer lies within the major groove of the PRE half-site where it makes sequence-specific interactions. Helix 2 crosses over and is perpendicular to helix 1, and there is also a short third helix just C-terminal to helix 2, which we have termed helix 2′. Although the structures of the individual DBD subunits are similar, they do not interact with the DNA equivalently. Solvent accessibility and the atomic packing density at each PR monomer-DNA interface, as determined by the program FADE (37) and CCP4 (38), revealed that subunit A of the DBD dimer (shown in Fig. 2 as the PR DBD subunit on the right) associates more closely with the DNA than the B subunit. The same analysis applied to the structures of other steroid receptor-DNA complexes showed that monomer binding of the GR subfamily is also not equivalent, with subunit A (as defined in the PDB entries) associating more closely with the DNA than the B subunits (34–36). This asymmetric DNA interaction suggests that the GR subfamily members have a dominant monomer of the two involved in binding to the inverted repeat HREs.
To more precisely determine DBD structural relationships, the best fit of the local protein backbone atoms of each subunit of PR DBD (A and B) was compared with each subunit of other steroid receptors using root mean squared deviation analyses (55). PR was most closely related to ERα, and the most closely related nonsteroid receptor was found to be class II receptor RXR (1.01 Å). Interestingly, all of the B subunits (as defined in the PDB entries) among the receptors were more closely related structurally than the A subunits. The subunit B of PR is most structurally comparable to the subunit B of ERα (0.81 Å) and AR (0.94 Å), than GR (1.44 Å). The finding that subunit B is the most closely related subunit was not unexpected considering that it is the subunit more loosely associated with the DNA and therefore under fewer structural restraints.
In addition to the core DBD, clear electron density was visible for the first eight residues of the CTE projecting from the A subunit (aa 632–640) and up to six residues (aa 632–638) from the subunit B (Figs. 1 and and3A).3A). Thus, for the first time, the structure of a steroid receptor CTE and its interaction with DNA has been observed. The CTE interacts with the minor groove of DNA flanking the two PRE half-sites (Fig. 2).
Although there are slight differences in subunit orientation, the protein and DNA contacts are virtually indistinguishable in both subunits of the PR-DNA complex (Fig. 2). The sequence-specific interactions between residues in helix 1 and sites in the major groove of the PRE are nearly identical with those previously reported for GR and AR (34, 36). The side-chain atoms Lys588-Nζ and Arg593-NH1 and -NH2 make sequence-specific hydrogen bonds with Gua6-N7 and Gua3-N7 and -O6, respectively, and Val589 has significant van der Waals interactions with Thy4-C5. These base-specific contacts are further stabilized by numerous direct and water-mediated phosphate backbone interactions summarized in the schematic (Fig. 1) and the DNA ladder diagram (Fig. 2; and the supplemental figure published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org) showing the A dimer subunit interactions with a single PRE, 3N spacer, and flanking sequence.
The PR-CTE contacts the DNA outside of the canonical PRE, and effectively extends the protein-DNA interface made by the DBD (Figs. 2 and and3A).3A). In both the A and B PR DBD subunits, Arg637 of the CTE inserts into the minor groove. There it forms hydrogen bonds to Cyt8-O2 and -O4′, which are part of the base triple at the junction of adjacent DNA molecules, and is further stabilized by an apparent hydrogen bond between the NH1 and the phosphate of Gua8 in the adjacent asymmetric unit. In subunit A, the carbonyl oxygen of the next residue (Lys638) hydrogen bonds to Arg637-NH1, which positions the side chain atom Lys638-Nζ for a phosphate contact with Ade7. Phe639 is the final CTE residue to have distinguishable electron density for both the main chain and side-chain in the structure of subunit A. Although, in this structure no further side chains of the CTE were seen, the nuclear magnetic resonance (NMR) structure of the ERR2-DNA complex showed that a region of the ERR2 CTE, C-terminal to the minor groove interacting region, looped back to interact with a hydrophobic pocket on the core DBD (9). This region of the core PR DBD also has a hydrophobic pocket (Val581 and Val633) in the same position as the contact region of ERR2 (Val117 and Leu169), as well as a unique surface patch of residues (Val581, Leu634, and Met595) that have the potential for additional interactions with hydrophobic residues of the PR CTE (Fig. 3B).
To determine whether the CTE-DNA interactions observed in the crystal structure were functionally important, DNA binding studies were conducted with PR-DBD648 mutants. Arg637 and Lys638 were substituted with alanine either as single (R637A) or double residue substitutions (R637A/K638A) in the context of the PR-DBD648 construct. The wild-type and mutant proteins were expressed and purified, and relative DNA binding affinities for a DNA fragment containing palindromic PREs were determined by EMSA (Fig. 3, B and C). The single mutant (R637A) and wild type proteins had similar DNA binding affinities (Wt Kd = 104 nM and R637A Kd = 121 nM), whereas the double mutant bound DNA with a 2.5-fold lower affinity (Kd = 302 nM), and also had a decreased maximal binding at saturation (0.7 vs. 0.9), indicating a weaker and more labile interaction with the PRE_2C7A DNA.
This binding study also showed that a fraction of the PR DBD bound to the DNA as a monomer, which was detected as a faster mobility DNA complex than the PR DBD dimer (Fig. 3B). Interestingly, no monomer DNA binding was observed for the double mutant R637A/K638A (Fig. 3B), which suggested an important role for the CTE in binding to PRE half-sites. Therefore, to examine this possibility more closely, we compared the binding affinities of wild-type and mutant PR DBDs for a DNA fragment containing only a single PRE or half-site (Fig. 3C). The single mutant R637A exhibited a small 1.5-fold reduction in DNA binding affinity (Kd = 205 nM) compared with wild-type DBD (Kd = 133 nM), whereas the binding of the double mutant (R637A/K638A) was essentially not detectable (Fig. 3C). These findings confirmed, in solution without the influence of any crystal packing contacts, that the CTE-DNA interactions observed in the crystal structure are important for PR binding to a half-site PRE and potentially also to imperfect inverted repeats.
CTE-DNA interactions are known to play an important role in class II and orphan nuclear receptor DNA binding (4). The DNA interactions of the PR-CTE were most comparable to those seen in the ERR2 orphan receptor in the NMR structure of the ERR2-DNA complex (Fig. 3D) (9). ERR2 belongs to the type III group of monomeric orphan receptors, which preferentially bind extended half-site response elements with the consensus sequence TNAAGGTCA (where TNA is the preferred flanking sequence, and the consensus sequence is underlined). The CTE of ERR2 has a Val-Arg-Gly-Gly-Arg sequence (a Grip box-like sequence) that interacts directly with the flanking TNA sequence by inserting arginine residues into the minor groove to make base-specific contacts (9, 18). The PR CTE sequence (Val-Leu-Gly-Gly-Arg-Lys) that inserts into the minor groove is similar to that of ERR2, except that it is positioned within the CTE closer to the GM boundary (Fig. 3D). A similar Gly-Gly-Arg sequence is also present in the CTE of ERα, and the ERα CTE was shown previously by us and others to be important for binding to ERE (estrogen response elements) half-sites (32, 33). In addition, all of the steroid class of nuclear receptors except ERβ have a conserved Arg-Lys sequence in a similar position in the CTE as the Arg-Lys residues of PR that are seen interacting with the minor groove (Fig. 3D), and mutations that disrupt only this region of GR are functionally deficient (39). These data, taken together, suggest that a motif in the CTE for minor groove DNA interaction similar to that of orphan receptors may extend to members of the steroid class of nuclear receptors.
The PR-DBD dimer interface consists of several direct and water-mediated inter subunit interactions (Figs. 2 and and4A).4A). The dimerization boxes (D-boxes, aa 604–608) have interactions between Arg 606-Asp608 and Ala604-Ile610, which are conserved with those seen in the ERα, GR, and AR structures (34–36). Outside of the D-box an additional conserved interaction was observed between the main-chain carbonyl oxygen of Leu602 and the side-chain atom Nε of Arg615 in helix 2′ (aa 616–618), as well as a unique interaction between the main-chain carbonyl oxygen Cys603 and the NH2 of Arg615. Helix 2′also maintains water-mediated contacts between opposing complementary Lys617 residues. Asymmetric interactions with water molecules are coordinated by the A subunit Lys617 main-chain carbonyl oxygen and the B subunit side chain amine of Lys617, and vice versa. A symmetric interaction via a water molecule is coordinated by the side-chain amine of Lys617 in both subunits. Solvent accessibility and packing density between the subunits A and B were determined for all the steroid receptor DBD-DNA structures by the programs FADE and CCP4 (37, 38). PR-DBD was found to have a looser packing density between the A and B subunits than the other steroid receptor, suggesting that it has a more relaxed dimer interface. The tighter association between the PR-DBD subunits and the PRE (discussed in Overall Structure) potentially could compensate for a weaker inter-subunit interaction.
Several protein-DNA interactions are made by H2′ residues (Fig. 2). These include a direct phosphate interaction between Arg616-Nε and Gua3, a water-mediated interaction with the phosphate of Thy4 via the Arg616-NH2, and the carbonyl oxygen of Lys617 also makes water-mediated contacts with the phosphate of Thy2. An interesting and unique interaction was also made by the side chain of Lys617, which extends across the 3N spacer to interact directly with the phosphate backbone (Thy2) of the PRE half-site bound by the opposite PR DBD subunit. In other steroid receptor structures, as shown here with ER-DBD as an example, the side chain of the equivalent lysine to PR Lys617 bends back toward the DNA of the response element to which the subunit is bound (Fig. 4B).
The PRE_2C7A DNA bound to PR (Table 2) is similar to standard B-form DNA (40), with the exception of distortions at base pairs AT(−5), AT (0), TA (5), and GC(−6) as well as the width of the minor groove. The width of the minor groove, 3.6 Å, is well below the mean value of 9.6 Å reported for B-DNA (Table 2 and Fig. 5A). Comparison of the DNA in the steroid receptor DBD-DNA structures using CURVES (Fig. 5A) showed that all of the complexes have compressed minor grooves but revealed that the PR-DNA complex had the most compressed minor groove on average and in the 3N spacer region (41). The ERE in the ER DBD complex also showed compression with an average minor groove width of 6.23 Å, whereas the AR and GR structures with one subunit bound to a perfect half-site and one associated with an incorrectly oriented half-site (designated AR and GR4S), had more greatly compressed minor groove widths on the side of the response element that contained the correct consensus half-site. Even the minor groove width of the DNA bound to GR, which had identical hexanucleotide HRE sequences and a similar crystal-packing environment (designated GR3S) had a wider (4.8 Å) minor groove than that of the PRE_2C7A DNA. Whether these differences in minor groove width represent intrinsic abilities of different steroid receptors to compress the minor groove or reflect differential flexibility or curvature of DNA due to the sequence of the 3N spacer and/or flanking region cannot be determined from crystallographic analysis.
The observed novel interactions of the PR-DBD-CTE with the minor groove outside of the HREs raised the question of whether there is an intrinsic sequence preference in the 3N spacer and/or sequences flanking the hexanucleotide HREs for optimal PR binding. To address this question, we conducted a binding site selection experiment. The selectable DNA contained fixed canonical PRE sequences with a randomized 3N spacer and four randomized nucleotides immediately flanking either side of the PREs (N4AGAACAN3TGTTCTN4). Purified PR DBD was bound to the randomized DNA, complexes were immunoprecipitated, the purified DNA was amplified, and after five rounds of selection, the selected DNA was subcloned and sequenced (42).
A slight strand bias and nonrandom distribution of bases in the 3N spacer and flanking DNA was detected using the MEME program (Fig. 5B) (43). The occurrence of a 3N spacer itself generates asymmetry in an otherwise palindromic binding site, which the alignment programs are sensitive to. However, when these sequences were aligned by MEME, an orientation was selected where position 0 was predominantly a pyrimidine, and the bases at −1 and +1 appeared to have a greater propensity also to be pyrimidines. This also corresponded to the A and B subunit orientation seen in the PR-DNA crystal structure. Interestingly, the PRE_2C7A DNA has a T-track in the 3N spacer, which is a sequence consistent with a narrow minor groove (reviewed in Ref. 44). The predicted curvature for the 3N spacer of the selected elements, as determined by the program BendIt (45) was increased across the 3N spacer similar to that predicted for the PRE_2C7A DNA used in the structure (Fig. 5A). Furthermore, it has been observed by several investigators that the steroid receptors have higher affinity for HREs with certain nucleotides within the 3N spacer (46–49).
The selection experiments also revealed a preference for sequences flanking the PRE, with stronger selection for the bases flanking half-site B, which is the subunit with the looser interactions with DNA. The sequence preference seen in the bases flanking the HREs is difficult to explain unless the CTE, which binds to these sites, actually contributes to sequence selectivity. Arg637 forms hydrogen bonds directly with the bases in the minor groove, and this may result in slight sequence bias in these regions, similar to what has been observed for the CTEs of the class II and orphan receptors (9, 18). Taken together, these results support the model that receptors have a dominant subunit (A), which selects for a pyrimidine-rich 3N spacer 3′, and which creates the opportunity for the for weaker half-site (B) to be more greatly influenced by the flanking sequence and the binding of the CTE.
The crystal structure of PR-DBD bound to DNA has provided insight into steroid receptor DNA interaction. The observed CTE in the structure demonstrates for the first time that a steroid receptor can make additional DNA contacts outside of the consensus HREs. These CTE interactions, analogous to the CTE interactions of class II and orphan receptors, are crucial for PR binding to PRE half-sites. This role for the CTE may be biologically relevant because many natural target genes for PR contain multiple weak half-sites or palindromic PREs that deviate from consensus sequences, and may require the extended protein-DNA interface provided by the CTE to assemble a stable PR-DNA complex needed for gene activation (48). The unique cross PRE half-site interactions correlated with a particularly narrow minor groove width, and the CTE interactions flanking the PREs are consistent with the observed subtle preference of PR for sequences outside of the consensus PREs. In particular, different 3N spacer sequence may have propensities to conform to different DNA structural features, and the CTE interaction may also confer some sequence selectivity to the PRE-flanking DNA. It is well known that the GR subgroup of steroid receptors can recognize the same consensus HRE, yet each receptor clearly regulates different sets of target genes in vivo (48). Although coregulatory proteins no doubt play a role in target site selection, the studies herein suggest that compression of the minor groove and sequence preferences in the three-nucleotide spacer and flanking DNA could also contribute to the differential ability of closely related receptors in the GR subgroup to bind and activate specific target genes.
The expression vector used to produce the for the PR-DBD670 glutathione-S-transferase (GST)-fusion protein for crystallization was previously described (31). Based on the polypeptide observed in the structure, a similar, but truncated, DBD termed PR-DBD648 (aa 562–648) was subcloned into the pGEX2T expression vector (Amersham Biosciences, Piscataway, NJ) via BamH1 and EcoR1 restriction enzyme sites that were introduced using standard PCR protocols. The N-terminal primer was previously described for PR-DBD670 the C-terminal primer is 5′-GATCGAATTCACTACACAACTCTGACTTTATTGAAC-3′. For mutagenesis, complementary synthetic oligonucleotides (Macromolecular Resources, Ft. Collins, CO) of PR-DBD648, containing a StyI site and the codon change for R637A (5′-G-ATCCCTTGGAGGTgccAAATTTAAAAAGTTCAATAAAGTCAG-AGTTGTGTAGTGAATTCCGATCA-3′ and 5′-GATCGGAATTC-ACTACACAACTCTGACTTTATTGAACTTTTTAAATTTggcACC TCCAAGGGATCA-3′), or the double mutant R637A/K638A underlined (5′-GATCCCTTGGAGGTgccgccTTTAAAAAGTTCA-ATAAAGTCAGAGTTGTGTAGTGAATTCCGATCA-3′ and 5′-GATCGGAATTCACTACACAACTCTGACTTTATTGAACTTTTT-AAAggcggcACCTCCAAGGGATCA-3′). Annealed oligonucleotides were inserted into the pGEX2T vector containing PR-DB-D670 via the StyI and EcoRI sites yielding the PR-DBD648, which contained either the point mutant R637A or the double mutant R637A/K638A. The PR coding regions of plasmids were confirmed by DNA sequencing.
GST-tagged PR-DBD670, PR-DBD648, and PR-DBD648 mutants were overexpressed in Escherichia coli strain BL21 (DE3) and purified by GST affinity chromatography (Amersham Biosciences). The proteins were cleaved from the glutathione-Sepharose resin with thrombin (Sigma, St. Louis, MO) at room temperature for 12–16 h. PR-DBD670 was further purified using size-exclusion chromatography in 20 mM Tris (pH 7.5), 100 mM NaCl, 10% glycerol, 1 mM EDTA, and 1 mM dithiothreitol (Superdex 30; Amersham Biosciences) and concentrated. PR-DBD648 and mutant proteins cleaved from the GST resin were further purified with a cation exchange column (Source 15S; Amersham Biosciences). The final concentration of the proteins was determined by UV absorbance spectroscopy (ε = 2580 at 280 nm), and the homogeneity of the DBD was examined using SDS-PAGE, native gel analysis, and matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Tufts Protein Chemistry Facility, Boston, MA).
Oligonucleotides (Macromolecular Resources) for crystallography (dCCAGAACAAACTGTTCTG and dCCAGAACAGTTTGT-TCTG) were purified by HPLC using an ion-pair reverse-phase chromatography with XTERRA (Waters, Milford, MA) oligonucleotide column (50). Fractions were pooled, lyophilized, and resuspended in water. The concentration of each of the DNA strands was determined by UV absorbance spectroscopy, so that equal amounts of complementary strands were annealed by boiling and slow cooling. Correct annealing was assessed by non-denaturing PAGE, followed by staining with Stainsall (Sigma) or Vistra Green (Amersham Biosciences).
The stoichiometric gel shifts were carried out with high concentrations of protein (0–45 μM) and DNA (10–15 μM) to simulate crystallization conditions. Increasing concentrations of PR-DBD were incubated with PRE in 0.5× TAE and 30% glycerol on ice. Reactions were electrophoresed on a cooled 6% nondenaturing polyacrylamide gel (running buffer 0.5× TAE) and stained with Vistra Green (Amersham Biosciences) for visualization via phosphorimaging (GE Healthcare, Piscataway, NJ). The ratio of DBD to PRE, at which 100% complex formation was observed, was decreased by 2.5% in crystallization trials, providing a slight excess of DNA needed for maximum crystallization.
The quantitative EMSA was carried out as described previously (31–33). PR-DBD648 or PR-DBD648 mutants were incubated with 10 mM Tris-HCl (pH 7.5), 5 mM NaCl, 5% glycerol, 2 mM MgCl2, 1 mM EDTA, and 5 mM dithiothreitol in the presence of 0.1 μg polydA-polydT and 1 μg ovalbumin as a carrier protein, in 25 μl total volume, for 30 min. The DNA used for the palindromic PRE and half-site PRE were (5′-GATCTTTGAGAACAAACTGTTCTTAAAACGAGGATC-3′) and (5′-GATCAAGTTTATAAAAGAACATGCTCAATTT-3′), respectively, with the half-sites underlined. The 32P end-labeled DNA was added to reactions to a final concentration of 0.6 nM and incubated for a further 30 min on ice. Reactions were electrophoresed on 6% nondenaturing polyacrylamide gels (37.5:1 acrylamide/bisacrylamide ratio) in 0.5× TAE buffer [0.02 M Tris-acetate (pH 8.0), and 0.5 mM EDTA] at 4 C. Gels were dried under vacuum at 80 C and the fraction of bound to free 32P-labeled probe was determined with a series 400 Molecular Dynamics PhosphorImager. The fraction of DNA bound as a function of protein concentration was determined from the average of at least three independent experiments and graphed as a best-fit binding curves using the equation y = (x/Kd)/1+(x/Kd), where y is the fraction of PRE bound and x is the concentration of DBD, and the error bars indicate the standard error.
The binding site selection method described in Pollock and Treisman (42) was modified to identify PR sequence preferences outside of the hexanucleotide response elements. A polyclonal antiserum against PR-DBD was used to immunoprecipitate complexes of PR-DBD648 bound to DNA fragments containing a single inverted repeat PRE with a randomized 3N spacer and four randomized nucleotides on either side of the response element and flanked by fixed ends for PCR priming and subsequent subcloning. Complex formation was carried out as described earlier. The immunoprecipitated DNA was washed with 10 mM Tris-HCl (pH 7.8), 1 M NaCl, 5% glycerol, 1 mM EDTA, 2 mM MgCl, 1 μM ZnCl, 0.1% Nonidet P-40, amplified using PCR, and purified using QIA-quick Nucleotide Removal Kit (QIAGEN, Valencia, CA). This process was repeated five times using a decreasing concentration of PR-DBD648 (50, 25, 10, 5, and 2.5 nM), and we noted that the amount of DNA precipitated during each cycle increased steadily. After the fifth amplification, the selected DNA was electrophoresed in the presence of 2.5 nM PR-DBD648 on a 8% polyacrylamide gel (37.5:1 acrylamide:bisacrylamide) for 90 min at 20 mA in 0.5× TAE (Tris-acetate-EDTA). The shifted DNA was visualized by Vistra Green-staining (Amersham Biosciences), gel-purified using the crush and soak procedure (51), subcloned into the BamHI-EcoRI site of a pUC18 vector and sequenced using standard methods. The sequences were aligned using MEME (43). The alignment used 27 sequences with the option to select the best alignment of between 15 and 23 nucleotides in length. The program selected 18 as the optimal length of the alignment. Preference for the sequences shown is indicated by the value of the information content, where a value of 0.21–0.42 represents a bias of the sequences such that approximately 70% of the selected sequences are the bases shown.
PR-DBD654 (140 μM) and PRE_2C7A DNA (71.75 μM) were incubated together on ice for 20 min to form the complex. Crystals were grown by sitting drop vapor diffusion in 96-well plates at 4 C with the addition of 1 μl of 70 μM DBD654-PRE complex to an equal volume of reservoir solution [0.1 M 2-(N-morpholino)ethanesulfonic acid (pH 6.0), 10 mM NaCl, 2.5 mM spermine, and 5% hexanediol]. In 1 wk, the crystals grew to approximately 300 μm × 50 μm × 20 μm in space group P212121. Cryopreservation was achieved with the addition of 20% glycerol before flash freezing in liquid nitrogen for data collection at −180 C. Images were collected for 30 min each at 0.5° oscillation over a range of 74°, at the University of Colorado at Denver and Health Sciences Center Biomolecular X-Ray Crystallography Center, using a Rigaku/MSC (The Woodlands, TX) Ru-H3R generator (λ= 1.5418 Å) equipped with an Raxis IV++ area detector, confocal optics and XTREAM system. Data were indexed and scaled with HKL2000 (HKL Research, Inc., Charlottesville, VA).
A PR-DNA homology model, based on the ER-ERE complex structure (35), was used to obtain a molecular replacement solution using AmoRe within the Crystallography and NMR Software suite (CNS) (52). Model building was performed with O (53). Initial refinement was performed with CNS and completed with the program Refmac 5 from CCP4 (38). The program CONTACT (CCP4: Supported Program (38) was used to determine interatomic distances and hydrogen bonds within the structure. GETAREA (54), a web-based server, was used to determine solvent accessible surface area with a 1.4 Å probe. DNA conformation and groove width was determined using the program CURVES (41). Images in the figures were made using the program Pymol (DeLano Scientific LLC, San Carlos, CA). The coordinates have the PDB ID 2C7A.
We acknowledge the support of the Howard Hughes Medical Institute, University of Colorado Cancer Center (Core Grant P30-CA46934) for the University of Colorado Health Sciences Center Biomolecular X-ray Crystallography Center. This work was also supported by grants from the National Institutes of Health (CA46938 to D.P.E. and GM56881 to M.E.A.C.).
Disclosure Statement: The authors have nothing to disclose.