| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Department of Pathology (H.J.D., R.H., H.A.G.M.v.d.K., J.v.T., A.C.J.Z.-v.d.M., J.T.), Josephine Nefkens Institute, and Department of Reproduction and Development (C.A.B., A.O.B.), Erasmus Medical Center, 3000 DR Rotterdam, The Netherlands; and Structural Biology Laboratory (C.S.V., A.C.W.P.), Department of Chemistry, University of York, York YO10 5DD, United Kingdom
Address all correspondence and requests for reprints to: Hendrikus J. Dubbink, Department of Pathology, Josephine Nefkens Institute, Erasmus Medical Center, PO Box 1738, 3000 DR Rotterdam, The Netherlands. E-mail: h.dubbink{at}erasmusmc.nl.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-LBD interactions of the wild-type AR FXXLF motif and coactivator transcriptional intermediary factor 2 LXXLL motifs and variants of these motifs. Random mutagenesis revealed a key role for the F residues in FXXLF motifs in high-affinity and selective AR LBD interaction. The FXXLF motif in full-length AR and transcriptional intermediary factor 2 LXXLL motifs competed for an overlapping binding site. A computer model of the AR LBD/AR FXXLF complex showed that the bulky F residues are buried in a deep coactivator-binding groove. The corresponding groove in estrogen receptor LBD is considerably shallower, explaining lack of binding of any of the FXXLF motifs tested. FXXLF and LXXLL motif interaction depended on different charged amino acid residues in the AR LBD present at opposite ends of the coactivator groove. In conclusion, our data demonstrate the importance of a deep hydrophobic groove and alternative usage of charged amino acids in specifying peptide binding to the AR LBD. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-dihydrotestosterone (DHT). AR function is crucial in the development and the maintenance of the male phenotype (1, 2, 3). The AR belongs to the nuclear receptor (NR) family. Like other NRs, the AR is composed of an N-terminal domain (NTD) containing activation function 1 (AF1), a central DNA-binding domain (DBD), and a C-terminal ligand-binding domain (LBD), which harbors activation function 2 (AF2). Transcriptional activation by NRs is mediated by AF1 and AF2 after ligand-induced conformational changes (4). In current models of transcriptional initiation, specific transcription factors recruit coactivators that facilitate association with chromatin remodeling complexes, general transcription factors, and RNA polymerase II. Among the best-characterized NR coactivators are the p160 family members, steroid receptor coactivator 1, transcriptional intermediary factor 2 (TIF2)/glucocorticoid receptor (GR)-interacting protein 1 (GRIP-1), and ACTR/RAC3 (4, 5, 6). P160 coactivators contain binding sites for secondary coactivators such as cAMP response element binding protein-binding protein (CBP) and p300, which possess strong intrinsic histone acetyltransferase activity, and the histone methyltransferases coactivator-associated arginine methyltransferase 1 and protein arginine methyltransferase 1 (5, 7, 8). These enzymatic activities are important for local chromatin remodeling to facilitate recruitment of chromatin remodeling complexes and basal transcription factors.
P160 coactivators can associate both with NR NTDs and LBDs. The specific requirements for coactivator-NTD interactions are not well established. In contrast, ample knowledge exists about coactivator-NR LBD interactions. These interactions are mediated by short -helical signature sequences denoted LXXLL motifs (X is any amino acid residue) or NR boxes, which bind directly and in a ligand-dependent manner to NR LBDs (9). The NR interaction domain (NID), located in the central part of the p160 coactivators, contains three such LXXLL motifs (10, 11). LXXLL motifs have been found not only in p160 coactivators but also in variable numbers in the cofactors receptor interacting protein 140, CBP, thyroid hormone receptor (TR)-associated protein 220/vitamin D receptor (VDR) interacting protein 205, PGC-1 [peroxisome proliferator-activated receptor 
(PPAR) coactivator 1], NR-activating protein 250/TR-binding protein, and TIP60. Distinct LXXLL motifs display different specificities to individual nuclear receptors due to different flanking amino acid residues (10, 12, 13, 14).
Crystal structures of unliganded, agonist-bound, and antagonist-bound NR LBDs, including the AR LBD, share a grossly similar overall fold of 10–12 -helices, which adopt a sandwich-like structure (15). Agonist-bound NR LBDs are well equipped for the interaction with coactivators as has been demonstrated by the ternary structures of several liganded NR LBDs with distinct p160 coactivator LXXLL peptides (12, 15, 16, 17, 18, 19, 20, 21). The LXXLL motifs bind as amphipathic -helices to a hydrophobic groove, composed of amino acid residues in helices H3, H4, H5, and H12, which is formed by the repositioning of helix H12 upon ligand binding. Stabilization of the LXXLL motifs in this coactivator groove is the result of hydrophobic interactions of its leucine residues with amino acid residues lining the groove and hydrogen bonding with glutamate and lysine residues present at the opposite ends of the groove, also referred to as "charge clamp."
Many protein partners, which can bind to distinct AR domains, have been described, yet the role of most of these cofactors in AR function remains unclear (for reviews see Refs. 3 and 22). Overexpression of the p160 coactivators, steroid receptor coactivator 1 and TIF2, stimulates the transactivation function of AR (10, 23, 24, 25, 26, 27, 28). Protein-protein interaction assays have demonstrated LXXLL-dependent physical interaction of coactivators with the AR LBD and LXXLL-independent interaction with the AR NTD. Recent chromatin immunoprecipitation assays provided evidence for in vivo recruitment of these factors to androgen-responsive promoters (29, 30).
In addition to cofactor binding, hormone-dependent intra- or intermolecular interaction between the NTD and LBD (N/C interaction) has been demonstrated (31, 32, 33). AR N/C interaction might play a role in slowing the androgen dissociation rate and in selective gene activation (34, 35, 36). Ligand-dependent N/C interactions have also been described for ER and progesterone receptor (PR) (37, 38). Unique for the AR N/C interaction is the involvement of an amphipathic -helix containing an FXXLF motif in the NTD, with similar structural features as coactivator LXXLL motifs (39, 40). In this study we use the structural resemblance of the AR FXXLF motif and the TIF2 NID LXXLL motifs to investigate the role of F residues in NR LBD recognition and to perform a computer model- and mutation-based analysis of the interaction of the AR LBD with these peptide motifs. We provide evidence explaining the preference of the AR coactivator groove for FXXLF motifs and describe a differential dependence of the AR FXXLF motif and TIF2 coactivator LXXLL motifs on charged amino acid residues flanking the coactivator-binding pocket.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
) are important for AR LBD binding (39, 40). To gain insight into the contribution of the F residues in AR LBD binding, we compared interaction of the AR FXXLF motif and p160 coactivator TIF2 NR boxes, LXXLL motifs, in a yeast two-hybrid readout system utilizing GAL4DBD-AR.LBD as a bait and Gal4AD fusions with peptides derived from AR FXXLF and TIF2 LXXLL motifs (Fig. 1, B and C). Figure 1D shows interaction of AR LBD with wild-type AR FXXLF and the three TIF2 LXXLL motifs. The AR FXXLF motif and TIF2 box I and III interactions were comparable. No binding was observed for TIF2 box II. Next, we compared the effect of F to L and L to F substitutions at both positions +1 and +5 of the AR and TIF2 motifs, respectively (Fig. 1E). In agreement with a previous study (39), using the entire AR NTD, F to L replacement in the AR peptide completely abolished the interaction with the AR LBD. In contrast, L to F substitutions in the context of the TIF2 boxes did not affect the interaction of TIF2 box III and even led to an enhanced interaction with TIF2 box I. In addition, a small increase in TIF2 box II interaction was observed. Western blot analysis of the yeast transformants showed a lower expression level of the Gal4AD-AR peptide fusion proteins (Fig. 1F). Therefore, interaction of the AR FXXLF motif with AR LBD might be better, in fact, than that of each of the TIF2 boxes. Together, these results demonstrated that F residues at positions +1 and +5 of an interacting LXXLL motif did not affect, or even increased, the affinity for the AR LBD, whereas L substitutions disrupted binding of the interacting AR FXXLF motif. Single F to L or L to F substitutions in the AR motif or TIF2 box III, respectively, led to a complete absence of interaction with the AR LBD (data not shown). These results indicated that for optimal peptide binding, either two L residues (TIF2) or F residues (AR/TIF2) are required at positions +1 and +5, respectively.
|
|
). Expression constructs for Gal4DBD-peptide fusion proteins and full-length AR and an upstream activating sequence (UAS) 4 TATA luciferase reporter construct were cotransfected in Hep3B cells. The same peptide panel (Fig. 1C
) as used in the yeast assay was studied for AR binding in the absence and presence of DHT in the mammalian test. In this assay only the AR FXXLF motif showed a clear hormone-dependent interaction with full-length wild-type AR (Fig. 2B). To study whether TIF2 motif binding was prevented due to competition with the AR FXXLF motif in the full-length AR, the experiments were repeated with AR containing F to L substitutions at positions +1 and +5 in the FXXLF motif (F23L/F27L-AR). Figure 2C shows that this resulted in higher absolute activity and an approximately 10-fold higher hormone-induced binding in the case of TIF2 NR box I (from 4- to 48-fold) and III (from 3- to 24-fold). Also the AR FXXLF motif showed increased interaction with the mutated AR (from 56- to 129-fold). AR and F23L/F27L-AR were expressed at comparable levels (Fig. 2, B and C). We obtained similar increased interactions by F to R substitutions in the FXXLF motif in the full-length AR or complete deletion of the amino-terminal 36 amino acids (data not shown). In agreement with the yeast two-hybrid experiments, TIF2 NR box II interacted with neither wild-type nor F23L/F27L-AR. In conclusion, in mammalian cells the AR FXXLF motif had a higher affinity than the TIF2 boxes for an overlapping binding site in the AR LBD. As a consequence, interaction of TIF2 boxes with full-length AR was repressed by the FXXLF motif in the AR NTD.
|
|
, D and E). No substantial interaction of full-length AR with the AR LXXLL variant motif could be observed. However, the mutant TIF2 FXXLF variants displayed similar (box III) or increased AR LBD binding (boxes I and II) compared with the wild-type TIF2 LXXLL motifs (Fig. 2, B–E). Only the affinity of TIF2 box I FXXLF was sufficient to compete with the internal AR FXXLF motif (Fig. 2D). Overall, the mammalian one-hybrid results with F23L/F27L-AR are very similar to the yeast two-hybrid results. The differences between binding of AR and TIF2 FXXLF and LXXLL peptides and their variants to AR LBD might be explained by the influence of amino acid residues adjacent to F and L residues at positions +1 and +5.
Because interaction of the AR FXXLF motif with the AR LBD may affect ligand binding parameters, we measured R1881 ligand binding affinities and dissociation rates of wild-type AR and F23L/F27L-AR transiently transfected in Hep3B cells. Table 1 shows that the F to L substitutions in F23L/F27L-AR did not affect ligand binding affinity. In accordance with previous data (39), an approximately 2-fold drop in dissociation half-time compared with wild-type AR (47 vs. 83 min) was observed (Fig. 3 and Table 1).
|
|
). In this system FXXLF and LXXLL motifs were fused to the AR DBD. Direct activation of the promoter by full-length AR was prevented by introduction of the previously described A573D substitution in the DBD of full-length AR (41, 42). To prevent AR N/C interaction, the F23L/F27L substitutions were introduced in A573D-AR. The overall pattern of interaction of the wild-type (Fig. 4B) and variant (Fig. 4C) FXXLF and LXXLL motifs was very similar to the results obtained in the Gal4DBD-UAS readout system (Fig. 2, C and E), indicating that binding of these motifs with AR is hardly affected by promoter context or properties of the DBD.
|
and is described in detail in Materials and Methods. Mutagenesis was performed in a peptide encompassing AR amino acids 18–30, because this peptide displayed a slightly higher affinity for the AR LBD than the AR 17–32 peptide used above. Interacting peptides were selected based on colony growth in selective medium and blue staining. Screening of the library yielded 33 blue colonies after 3 d of culture. After 4 d hundreds of blue colonies became visible. Of the 33 colonies visible at d 3, 28 expressed the FXXLF motif. The other five clones expressed three FXXLM and two FXXLW peptides, respectively. Analysis of a selection of d 4 colonies showed the same interacting peptides. Interactions were confirmed in a quantitative ß-galactosidase assay after retransformation of the rescued plasmids (Fig. 5A). Both the FXXLM and FXXLW peptides displayed an approximately 70% weaker AR LBD interaction than the FXXLF motif. No interaction was observed with control peptides containing M and W residues at position +1 and M residues at both positions +1 and +5. However, the WXXLW variant could interact, although with an even lower affinity than the FXXLM and FXXLW peptides (Fig. 5A). All peptides were expressed to an equal level (Fig. 5B).
|
F Residues Prevent Peptide Interaction with the ER LBD
To address AR specificity of FXXLF peptides, ER LBD interaction studies were performed. Yeast two-hybrid experiments were done in the absence and presence of 17ß-estradiol (E2) utilizing the same AR and TIF2 peptide panel as above (Fig. 1, B and C). Neither the AR FXXLF motif nor the AR LXXLL peptide interacted with the ER LBD (Fig. 6), demonstrating that the AR FXXLF motif selectively bound to the AR LBD. However, because the AR LXXLL peptide was also unable to interact, it was not possible to decide whether this was due to the F residues, the flanking amino acids, or both. In contrast to the AR LBD, ER LBD showed a preference for TIF2 box II and exhibited virtually no binding to box III (Fig. 6). However, none of the TIF2 FXXLF variants showed binding to ER LBD, indicating that F side chains are not compatible with binding to the coactivator groove in the ER LBD.
|
LBD in complex with a LXXLL motif from the coactivator TIF2 (see Fig. 7, A and B) (20, 43). The potential binding mode of the FXXLF motif (residues G21–S29) was explored using molecular dynamics and conformational free energy analysis (see Materials and Methods).
|
-helical conformation and bind along the coactivator-binding groove in the AR LBD in a manner similar to that observed for the related coactivator LXXLL motifs to ER LBD (Fig. 7). In the model, the -helix was orientated in such a way that the bulky F side chains project into the groove formed by helices H3, H4, H5, and H12 (Fig. 7B). L+4 rested on the periphery of the coactivator-binding groove. The N- and C-terminal ends of the helix were capped by the predicted charge clamp residues, E897 and K720, respectively. E897, located in H12, interacted with the backbone amide of the F+1 residue of the FXXLF motif (Fig. 6C). K720 interacted with the main-chain carbonyl groups of the L+4, F+5, Q+6, and S+7 residues. Two other basic amino acids (K717 and R726) also interacted with the polypeptide backbone of the C-terminal portion of the FXXLF motif. K717 (H3) and R726 (H4) made hydrogen bonds with the main chain carbonyls of the L+4 and F+5 residues, respectively. An additional hydrogen bond was predicted between the main chain of the N-terminal (–2) residue and the side chain of Q738 (H5).
We examined the AR LBD/DHT crystal structure (43) and observed that the cofactor-binding groove had a deep, well-defined pocket adjacent to E897 and a slightly shallower depression adjacent to K720 (Fig. 7E). Both sites were much deeper than the corresponding regions in the ER LBD crystal structure (Fig. 7D) and appeared to be able to bury effectively the entire F side chains of the FXXLF motif (Fig. 7E). The side chain of F+1 was tightly bound in a site adjacent to E897 formed by L712 (H3), V716 (H3), M734 (H5), I737 (H5), and M894 (H12). The F+5 bound at the other end of the cofactor-binding groove beneath K720 and R726. The +5 binding site was lined by A719 (H3), K720 (H3), F725 (H4), Q733 (H5), and I737 (H5). L+4 made contact with V716 (H3). Additional contacts were made between Q+2 and residues in H5 (Fig. 7C).
Comparison of the AR LBD crystal structure (43) and the AR LBD/FXXLF model demonstrated that only minor changes were needed in the orientations of residues that line the coactivator-binding cleft to accommodate the bulky motif. The largest adjustments were at the F+5 binding site where the side chains of F725 (H4) and I737 (H5) moved in toward the body of the LBD to facilitate binding of this motif.
Differential Influence of AR K720 and E897 on FXXLF- and LXXLL-Peptide Interaction
The AR LBD/FXXLF model was used as a guide to examine the contribution of amino acid residues in the AR LBD in FXXLF- and LXXLL-peptide recognition. In these experiments TIF2 box II was excluded, because it showed hardly any binding to wild-type AR LBD. First, we studied the importance of the classic charge clamp amino acid residues K720 and E897 (Fig. 7) in the yeast two-hybrid assay (Fig. 1B). K720A substitution almost completely abolished interaction with TIF2 boxes I and III, but did not decrease AR FXXLF motif interaction (Fig. 8A). E897A substitution resulted in an opposite pattern: AR FXXLF motif interaction, but not binding to the TIF2 boxes, was diminished. Neither the K720A nor the E897A mutant showed significant binding to the AR LXXLL variant. The interactions of the TIF2 FXXLF variants were very similar to those of the wild-type peptides (Fig. 8B). Although K720A expression was slightly lower, all Gal4DBD-AR.LBD fusion proteins were well expressed (Fig. 8C). Essentially identical results were obtained in the mammalian one-hybrid system using full-length F23L/F27L-AR (Figs. 2A and 8, D and E), although TIF2 box III and its FXXLF variant showed a 2-fold decreased binding to the E897A mutant. The K720A and E897A mutants were equally expressed (Fig. 8F). Together these results can be explained by differential hydrogen bonding between the K720 and E897 side chains and the individual peptide backbones. In addition, the data stress the important overall contribution of the amino acid residues flanking the core F+1 and F+5 residues, because L to F substitutions at positions +1 and +5 in the two TIF2 motifs did not change their interaction pattern with K720A or E897A.
|
|
, B and E). Lack of effect of the K720 mutation on FXXLF binding might be due to these interactions. To address this issue, interactions of the AR FXXLF motif with AR LBDs harboring single, double, or triple A substitutions at positions K717, K720, and R726 were investigated. Yeast two-hybrid experiments showed that these substitutions did not affect or even increased interaction with the AR FXXLF motif for different mutants (Fig. 9A). In the mammalian one-hybrid system no effect or an up to 2- to 3-fold decreased interaction was found for individual mutants (Fig. 9B). Next we studied the influence of the LBD mutations on interaction with TIF2 boxes I and III. Our findings emphasized the dominant role of K720 in interaction with both TIF2 LXXLL motifs, because all AR LBDs containing this mutation showed a sharply decreased binding both in yeast two-hybrid assays (Fig. 9, C and E) and in mammalian one-hybrid assays (Fig. 9, D and F). Furthermore, binding of both motifs was sensitive to K717A substitution, but differentially affected by R726A substitutions. R726A mutation strongly decreased interaction with TIF2 box III (Fig. 9, E and F), but had no effect on TIF2 box I binding (Fig. 9, C and D). All AR variants expressed at similar levels (Fig. 9, G and H).
|
and Table 1 the binding affinities and dissociation half-times of all AR LBD mutants were in the same range as that of F23L/F27L-AR from which they were derived.
In conclusion, our results, as demonstrated in Fig. 9, showed that K717 and R726 are not essential for AR FXXLF motif binding; however, K717 and R726 are selectively involved in binding of TIF2 boxes I and III.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-helix, binding to AR LBD strongly depends on the hydrophobic residues at positions +1, +4, and +5, and this interaction is modulated by flanking sequences (39, 40). In this study, we explored the binding properties of the AR FXXLF motif and TIF2 LXXLL modules to demonstrate distinct peptide binding modes to the AR LBD coactivator groove. Additionally, evidence was provided for the structural basis of selective FXXLF binding to AR LBD and not to ER LBD.
Specific activation of individual NRs depends on selective coactivator recruitment due to differential affinities of LBDs for these cofactors (45). AR N/C interaction can compete with selective coactivator recruitment. We showed that N/C interaction, mediated via binding of the FXXLF motif in the AR NTD to the coactivator groove in the AR LBD, completely abolished interaction of the AR LBD with TIF2 NR boxes. Disruption of N/C interaction leads to a strong increased interaction capability of the TIF2 LXXLL motifs (Fig. 2). AR N/C interaction may thus affect FXXLF- or LXXLL-mediated coactivator interaction. Furthermore, these results imply that AR LBD/AR FXXLF interaction occurs in full-length wild-type AR in vivo. Our data are not conclusive regarding whether AR N/C interaction is preferentially an intramolecular interaction or involved in AR homodimerization. However, interaction of the AR FXXLF motif with full-length wild-type AR indicates that intermolecular interaction is possible (Fig. 2).
In the model of the AR FXXLF/AR LBD complex that we describe, the docking mode of the AR FXXLF peptide and the overall structure of the AR LBD coactivator binding groove is similar to that seen in crystal structures of coactivator LXXLL motifs with ER LBD, and LBDs for glucocorticoids (GR), peroxisome proliferators (PPAR), retinoids (retinoid X receptor-), and thyroid hormones (TRß) (12, 16, 17, 18, 19, 20, 21). The FXXLF peptide adopts an amphipathic -helical structure, as predicted from the primary amino acid sequence (40), and a deep AR coactivator groove provides enough space for extensive hydrophobic interactions with the F side chains of the peptide (Fig. 7). Additional in silico data support the model. Attempts to generate a computer model of an ER LBD/AR FXXLF complex failed (data not shown), in agreement with the incapability of FXXLF peptide binding to ER LBD (Fig. 6). An in silico control experiment to generate an ER LBD/TIF2 NR box II complex resulted in a virtually identical model as the experimentally determined crystal structure (Ref. 45 and data not shown).
The FXXLF variants of the TIF2 LXXLL peptides show that F residues add another level of specificity to binding in coactivator grooves (Fig. 6). The AR specificity of the distinct FXXLF modules is apparently due to the inability of the F residues to dock into the less deep ER coactivator groove (compare panels D and E of Fig. 7). Distinct shapes of NR coactivator grooves may thus contribute to selectivity of peptide binding.
Several lines of experimental evidence indicate the importance of the depth of the AR LBD coactivator groove for optimal peptide binding. Random mutagenesis of the AR FXXLF motif at positions +1 and +5 shows that F residues are indispensable for high affinity to AR LBD. Weaker interactions are observed with variant AR FXXLM and FXXLW motifs (Fig. 4). Single F to L substitutions in the AR FXXLF motif were not found in the random screen and prevented binding in interaction assays (data not shown). We speculate that this is due to lack of sufficient hydrophobic interactions with amino acid residues lining the coactivator binding groove. The TIF2 FXXLF variants show that L to F exchange in an interacting peptide is allowed and might even increase its affinity, probably because of a better fit in the AR coactivator groove (Fig. 1, D and E, and Fig. 2, B–E). These results strongly suggest that bulky side chains are needed to form optimal hydrophobic interactions in the AR coactivator groove. In line with this hypothesis are the high affinity of FXXLF motifs in the AR cofactors ARA54 and ARA70 (46) and the two high-affinity FXXLF peptides and the FXXLY peptide identified in a recent random screen for AR LBD interacting peptides (47).
However, the deep AR groove does not exclude high-affinity binding of LXXLL peptides like TIF2 boxes I and III (this study) and the D11 and D30 peptides (44). Binding of these LXXLL peptides to the AR LBD surface might not predominantly depend on extensive hydrophobic interactions of L residues, but to a large extent on the properties of the amino acid residues flanking L+1 and L+5. Recently, it has been published that amino acid residues flanking the LXXLL core determine specific electrostatic interactions with the AR LBD (48). This might be complementary to the role of hydrogen bonds and hydrophobic interactions proposed in the present study.
According to our model, the FXXLF peptide is positioned between AR K720 and E897 (Fig. 7). These K and E residues correspond to the classic charge clamp in other NRs, which dictate a common binding mode of LXXLL motifs. Mutation of either of these residues in ER, PPAR, TRß, and VDR severely reduced LXXLL-mediated coactivator interaction due to disruption of hydrogen bonds with the charge clamp (16, 49, 50, 51). Our results indicate that interaction of the charge clamp in the AR LBD with TIF2 LXXLL motifs mainly depends on K720 and not on E897 (Fig. 8). Additionally, K720 and E897 display a differential selectivity for the AR FXXLF motif and TIF2 peptides, because E897 is the major determinant for interaction with the AR FXXLF motif.
Interaction between the entire AR NTD or TIF2 NID with AR LBD shows a similar dependency on K720 as for the corresponding peptides (23, 24, 26). Substitution of E897 by a K or Q residue severely diminished interaction with both AR NTD and TIF2 NID (23, 26). In the present peptide study, we presented data on E897A substitution. In contrast to E897A substitution, E897Q mutation differentially affected peptide binding, whereas E897K mutation inhibited FXXLF- and LXXLL-peptide interaction completely (data not shown). It is most likely, therefore, that K and Q substitutions actively interfere with peptide interaction.
The AR LBD/FXXLF model predicts hydrogen bonding between K717, K720, and R726 and the FXXLF peptide main chain; nevertheless, mutational analysis showed that these residues are not essential for binding of the AR FXXLF motif. We propose that the hydrophobic interactions of the F residues with the groove amino acid residues compensate for the lack of hydrogen bond formation. Regarding TIF2 boxes, K720 is most important and K717 plays a modulating role. R726 discriminates between TIF2 boxes I and III and is involved only in box III recognition, demonstrating that these motifs have a different binding mode (Fig. 9, C–F). The apparent discrepancies in the data obtained for AR FXXLF interaction with the LBD mutants, as observed in the yeast and mammalian readout systems, cannot be easily explained (Fig. 9, A and B). Because the difference between yeast and Hep3B cells was much less pronounced for the TIF2 peptides (Fig. 9, C–F), we expect that it depends on the peptide structure and not on the AR LBD. It is possible that the AR FXXLF motif adopts a slightly different conformation in yeast and mammalian cells.
The three basic amino acid residues K717, K720, and R726 are conserved in mammals, Xenopus, and fish AR LBD, supporting their functional importance. In addition, GR, PR and mineralocorticoid receptor (MR) have three positively charged residues at positions corresponding to K717, K720, and R726 in the AR. GR and PR contain the same K-K-R composition, MR contains a K-K-K motif. Therefore, these three residues may also be involved in selective recognition of coactivator LXXLL motifs by GR, PR, and MR. This hypothesis is supported by the recently solved crystal structure of the GR LBD/TIF2 box III complex (21). This complex contained, in addition to the classical K579 and E755 charge clamp, a second clamp comprised of GR residues R585 and D590. Experimental evidence indicated that this second clamp is involved in preferential binding of TIF2 box III to the GR LBD. The GR residues R585 and D590 correspond to R726 and D731, respectively, in the AR. Because our model predicts that D731 is located in the groove (Fig. 7C), we also addressed its involvement in TIF2 box recognition. We examined the effect of single D731A substitution or D731A combined with either K717A, K720A, or R726A on recognition of TIF2 boxes. In contrast to AR R726, we observed no substantial contribution of D731 to interaction with or selectivity for TIF2 boxes (data not shown), indicating that interactions of TIF2 NR box III with AR LBD and GR LBD are not completely identical.
In contrast to other NRs, AR AF2 function is weak (52, 53) and becomes only clearly manifest upon p160 coactivator overexpression (23, 24, 25). Charge clamp mutations in ER, PPAR, TRß1, and VDR drastically reduce transactivation capacity, without affecting ligand binding affinity and DNA binding, due to abrogation of LXXLL-mediated coactivator interaction (16, 49, 50, 51, 54). The charge clamp mutations reported here maximally reduce full-length AR transcriptional activity by approximately 50% (data not shown). Therefore, AR coactivator groove function differs from that in other NRs.
The most obvious function of the AR groove is AR N/C interaction. AR N/C interaction prolongs androgen binding by slowing the dissociation rate of AR-bound androgen without altering androgen affinity, as previously shown for separate F23L/F27L and LBD mutants (26, 39) and in this study for F23L/F27L-LBD double mutants (Table 1 and Fig. 3). Prolonged androgen binding may result from stabilization of H12 through interaction of the FXXLF motif with E897. Two recent studies point toward the involvement of AR N/C interaction in selective transactivation of target genes (35, 36). Although both studies describe that disruption of N/C interaction results in promoter-dependent differences in AR transactivation, apparent conflicting conclusions were drawn as to which promoter types are affected. Therefore, these observations require further investigation. It is possible that AR N/C interaction regulates coactivator recruitment in a promoter-dependent manner. In addition, the depth of the AR coactivator groove itself provides an alternative way for the AR LBD to discriminate between coactivators or selectively exclude certain LXXLL-containing coactivators. Therefore the AR coactivator groove may have a modulator function for several aspects of AR functioning including consecutive binding of the AR NTD and specific coactivators, resulting in cell-type and promoter-dependent protein complexes.
AR LBD mutations are associated with two human diseases, androgen insensitivity syndrome and prostate cancer (http://ww2.mcgill.ca/androgendb). Some of the mutations are in amino acids lining the coactivator groove (see Fig. 7C). K720E, R726L, and V730M represent groove mutations found in prostate cancer patients. Two of these involve charged amino acid residues studied in this paper. So far, however, their functional association with prostate cancer is not well understood. Three mutations of amino acid residues lining the AR LBD coactivator groove associated with partial androgen insensitivity syndrome, L712F, F725L, and I737T, have been partially characterized (1, 55). These mutations do not, or hardly, affect androgen binding. In transient transfections defective transcriptional activation is restricted to low androgen concentrations and can be restored by increased androgen levels (26, 55, 56). The F725L and I737T mutants display impaired N/C and TIF2 interaction, which also can be restored by increased hormone concentrations. Importantly, patients carrying the L712F mutation benefit from high-dose androgen treatment (55). These naturally occurring mutations underscore the importance of the coactivator groove in AR function.
In conclusion, our results show that the mode of AR LBD interaction with LXXLL- and FXXLF peptides differs from the classic charge clamp model. In AR LBD/peptide interaction, a combination of hydrogen bonds at one side of the charge clamp, along with extensive hydrophobic interactions within the deep coactivator groove, appears essential for high affinity. Interactions of the core F+1 and F+5 residues with amino acids lining the deep coactivator groove can be dominant over interactions of flanking sequences with AR LBD, resulting in preferential binding of FXXLF motifs. Furthermore, we provided evidence that the shape of an NR coactivator groove can be an important determinant in selective peptide interaction, adding a new level of specificity to NR LBD-peptide interaction.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Yeast construct pGalDBD-AR.LBD (AR661–919) has been described previously (32). AR expression plasmid pCMVAR0 was constructed by subcloning a SalI fragment from pSVAR0 (57), encoding full-length human AR, into the XhoI site of pcDNA3.1. pAR DBD was constructed by exchange of the BglII-BamHI Gal4DBD cDNA fragment from pM-B/E with AR DBD cDNA encoding amino acid residues 540–669. PCR was performed on pCMVAR0 with the primers 5'-AATTGAGATCTAGGATGGAAACTGCCAGGGACCATGTTTTG-3' (BglII site in bold, start codon in bold, italic) and 5'-AATTGGGGATCCGACATTCATAGCCTTCAATGTGTGAC-3' (BamHI in bold). The AR DBD cDNA sequence was verified. Yeast Gal4DBD-ER.LBD expression construct pGBT9-HBDER was generously provided by Michael Stallcup (58). The (UAS)4 TATA and (ARE)2 TATA luciferase reporter constructs were kindly provided by Magda Meesters and Guido Jenster, respectively. All mutant constructs used in the mammalian and yeast protein-protein interaction assays were prepared using QuikChange Site-Directed Mutagenesis (Stratagene, La Jolla, CA). As templates pGalDBD-AR.LBD and pCMVAR0 were used. All mutations were confirmed by sequencing.
Yeast Culture, Transformation, and ß-Galactosidase Assay
Y190 yeast culture, transformation, and liquid culture ß-galactosidase assays to quantify protein-protein interactions were performed as described previously (32, 40). Yeast cells were grown in the presence of 1 µM 5-DHT (Steraloids, Wilton, NH), 100 nM estradiol (E2, Steraloids), or vehicle.
Library Construction and Screening
Random mutagenesis at positions 23 and 27 of the AR FXXLF motif was performed by QuikChange Site-Directed Mutagenesis (Stratagene) using plasmid pACT2-AR 18–30 F23R/F27R as template. This construct expresses a Gal4AD-AR18–30 peptide, with R substitutions at positions 23 and 27, which cannot interact with the AR LBD. Used oligonucleotides were 5'-CGAACCTACCGAGGAGCTNNNCAGAATCTGNNNCAGAGCGTGGAATTC-3' and its complementary strand. The resulting plasmids were transformed to Escherichia coli strain XL-1 blue, and 20,000 independent colonies were harvested from which the library DNA was prepared. Mutagenesis was verified by sequencing randomly selected clones showing that the template construct was present in less then 5% of the cases.
The library DNA was transformed to yeast strain Y190 containing pGalDBD-AR.LBD, as described above. A total of 150,000 transformants was plated on selective medium lacking leucine, tryptophan, and histidine and containing X--gal (CLONTECH), 20 mM 3-amino-1,2,4-triazole, and 1 µM DHT. Library plasmids were recovered from colonies, according to the Yeast Protocol Handbook (CLONTECH).
Mammalian Cell Culture, Transient Transfections, and Luciferase Assay
Hep3B cells were cultured and transfected as described previously (59). Transfections were performed with 150 ng (UAS)4 TATA or (ARE)2 TATA luciferase reporter construct, 50 ng Gal4DBD-peptide or AR DBD-peptide expression construct, and 50 ng AR expression construct, in the presence of either 0.1 µM DHT or vehicle. Luciferase activity was assayed according to Steketee et al. (59), except that luciferase activity was measured in a Fluoroscan Ascent FL (Labsystems Oy, Helsinki, Finland). Light emission was recorded during 5 sec after a delay of 2 sec.
Western Blot Analysis
Preparation of yeast protein extracts and Western blot analysis for detection of Gal4AD and Gal4DBD fusion proteins were performed as described previously (32, 40) using monoclonal antibodies against Gal4AD (CLONTECH) and GAL4DBD (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). For monitoring AR expression, Hep3B cells were transfected as described above. Cells were lysed in SDS-PAGE loading buffer. AR was visualized with monoclonal antibody F39.4.1 as described previously (42).
Scatchard Analysis and Dissociation Rate Measurements
Hormone binding characteristics of wild-type and mutant ARs were measured in Hep3B cells transfected with 500 ng of vector expressing the appropriate AR mutant as described above. After 24 h, culture medium was substituted by medium without serum, and cells were incubated in this medium overnight.
In the case of Scatchard analysis the cells were subsequently incubated for 2 h at 37 C with increasing concentrations (0.03–10 nM) of 3H-labeled R1881 (New England Nuclear, Boston, MA). Nonspecific binding was measured in parallel experiments performed in the presence of a 200-fold excess of cold R1881.
For dissociation rate measurements, the cells were next incubated with 5 nM 3H-labeled R1881 for 2 h at 37 C and subsequently incubated with a 2000-fold excess of cold R1881 for different time periods (0–150 min). Nonspecific binding was measured in parallel experiments in which the cells obtained a 200-fold excess of cold R1881 during the entire 3H-R1881 incubation period.
After these incubations, the cells were washed four times in ice-cold PBS and lysed in 0.5 M NaOH, and radioactivity was measured in a scintillation counter.
Construction of the Molecular Model of the AR LBD and FXXLF Motif Complex
The crystal structure of ER LBD in complex with TIF2 NR box II [Protein Data Bank: 1GWQ (20)] was used as a template for modeling of the interaction between the AR FXXLF motif and the AR LBD. The model was constructed in three stages: 1) Preparation of a composite AR/ER model comprising the AR LBD along with the LXXLL NR box II peptide; 2) Mutation of the LXXLL motif to the corresponding AR FXXLF sequence; 3) Subjecting the resultant AR/FXXLF model to conformational free energy analysis.
Briefly, the structures of ER and AR (Protein Data Bank: 1I37) were superimposed and the LXXLL peptide segment from the ER NR box complex was incorporated into the AR structure. The starting AR/peptide motif model comprised the entire LBD along with DHT, 17 buried water molecules present in the original AR crystal structure, and the LXXLL NR-box II peptide. The CHARMM force field (60), employing the polar hydrogen topology and associated parameters, was used to represent the protein whereas the ligand was modeled using an all-hydrogen potential. The initial minimized structure for the AR/peptide complex was prepared as outlined elsewhere (61).
The LXXLL peptide component of the initial model was then mutated so that the sequence corresponded to that of the AR FXXLF motif (residues G21–S29). The side chains were initially positioned based on standard values in CHARMM. The conformational space spanned by the side chains of F+1, L+4, and F+5 in the FXXLF sequence (729 conformations) were explored as follows: The whole system was subject to minimizations under constraints that were periodically reduced until the whole system was free. This was followed by fixing all protein atoms that were outside of a 14-Å radius of the peptide. The remainder of the system was subject to rounds of minimization until the change in potential energy was less than or equal to 10–5 kcal/mol. This system was then subjected to normal mode analysis, and the entropies were computed using the standard statistical mechanical treatment (62). The free energies of interaction between the LBD and peptide were then computed. The AR LBD/FXXLF model described here had the second lowest free energy of interaction. The lowest energy structure was discarded as the FXXLF motif exhibited a conformation in which both F side chains were pointing away from the body of the LBD. All model building was carried out in the molecular graphics package QUANTA (Accelrys, San Diego, CA).
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Present address for C.S.V.: Bioinformatics Institute, 30 Biopolis Way, 07-01, Matrix, Singapore 138671.
Abbreviations: AF, Activation function; AR, androgen receptor; ARE, androgen response element; DBD, DNA-binding domain; DHT, 5-dihydrotestosterone; E2, estradiol; ER, estrogen receptor ; GR, glucocorticoid receptor; LBD, ligand-binding domain; MR, mineralocorticoid receptor; N/C interaction, interaction between NTD and LBD; NID, NR interaction domain; NR, nuclear receptor; NTD, N-terminal domain; PPAR, peroxisome proliferator-activated receptor ; PR, progesterone receptor; TIF2, transcriptional intermediary factor 2; TR, thyroid hormone receptor; UAS, upstream activating sequence; VDR, vitamin D receptor.
Received for publication September 26, 2003. Accepted for publication May 28, 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
. Nature 395:137–143[CrossRef][Medline]
/RXR crystal structure reveals the molecular basis of heterodimerization among nuclear receptors. Mol Cell 5:545–555[CrossRef][Medline]
. J Biol Chem 277:21862–21868
