SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • androgen receptor;
  • androgen receptor coregulators;
  • androgen-independent prostate cancer

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

Androgen depletion in combination with antiandrogenic agents is initially highly effective for treating prostate cancer, and is the recommended treatment for more advanced or higher-grade tumours. However, many tumours eventually become insensitive to androgens, even though the androgen receptor (AR) continues to be expressed. Computational chemistry combined with structural analysis of nuclear receptors and determination of binding affinities of natural and designed coregulators (coactivators and corepressors) provides the theoretical framework for the rational design of novel therapeutic agents directed at the AR. Adding alternative groups to various sites throughout the receptor can alter the conformation of the molecule and its functional binding with coactivators or corepressors. Possible molecules can be identified thoroughly and systematically using intelligent high-throughput screening and FASTrack chemistry (three-dimensional crystallography). Applying these techniques should eventually result in therapeutic agents for androgen-independent prostate cancer that can block binding of AR coactivators while simultaneously increasing binding of AR corepressors.


Abbreviations:
AR(A)

androgen receptor (-associated)

LBD

ligand-binding domain

DBD

DNA-binding domain

NTD

N-terminus domain

NCoR

nuclear receptor corepressor

PPAR

peroxisome proliferator–activated receptors

SRC

steroid receptor coactivator

LXXLL

leucine-X-X-leucine-leucine

FXXLF

phenylalanine-X-X-leucine-phenylalanine

AF

activation functions

3D

three-dimensional

TR

thyroid hormone receptor.

INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

Despite the extensive use of orchidectomy, androgen ablation, or hormone therapy with antiandrogens, prostate cancer remains second only to lung cancer as the leading cause of cancer-related death in men. As prostate cancers are unique in requiring androgens to maintain growth and avoid apoptosis [1], androgen depletion combined with anti-androgenic agents is initially highly effective and is the recommended treatment for more advanced or higher-grade tumours [2]. However, many tumours eventually become insensitive to androgens, even though the androgen receptor (AR) continues to be expressed.

The mechanisms for this loss of sensitivity to androgens are unclear. Gain-of-function mutations might increase sensitivity of the AR to androgens, resulting in a paradoxical insensitivity to androgen loss [3]. Other studies suggest that amplification of the AR gene [4] or decreased expression of AR coregulators such as PIAS1 and SRC1[5] can result in androgen independence of tumour cells. An alternative hypothesis is that mutations in the AR eventually cause anti-androgenic agents such as flutamide or bicalutamide to function as agonists rather than as antagonists, acting as ‘super ARs’ that can respond either to lower concentrations or a wider variety of ligands [6].

Since currently available antiandrogenic agents become ineffective in treating androgen-independent prostate cancer, an alternative approach might involve the rational development of agents that interfere with the function of the AR itself. One molecular target for this approach is the ligand-binding domain (LBD), or hormone-binding domain, for the AR.

STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

The AR is a nuclear receptor that acts as a transcription factor, and belongs to the steroid hormone receptor family. It is expressed in females as well as in males, but occurs in much higher concentration in males and is absolutely required for normal male development. The AR also plays a role in germ cell development by interaction with SRY (the male sex–determining protein encoded in the Y chromosome) [7]. As is characteristic of proteins in the nuclear receptor superfamily, the AR binds to DNA through a highly conserved central DNA-binding domain (DBD) and to its ligand through the C-terminus LBD, which is also highly conserved. The N-terminus domain (NTD) of the molecule is not as well conserved, and contains a transcription activation domain [7–9]. The AR is held inactive when bound to heat-shock proteins, which are presumed to act as chaperone proteins [10,11]. Activation of the AR occurs either through a ligand-independent mechanism, or through binding of dihydrotestosterone. Testosterone may also bind to the AR, but with much lower affinity. Activation through this ligand-dependent pathway causes dissociation of the heat-shock proteins and a conformational change in the AR itself, after which the AR interacts with a number of coactivators. The fully activated AR then activates transcription of androgen-responsive genes.

Like virtually all nuclear receptor proteins, activation of the AR is enhanced upon interaction with a variety of coactivators and diminished upon interaction with corepressors. Many coactivators have been identified, including proteins of the p160 family [12], AR-associated (ARA) proteins ARA70, ARA55, and ARA54 [13], and cyclic adenosine monophosphate (cAMP) response element-binding, protein–binding protein. These and other coactivators enhance the transcription of genes dependent on the AR. Repressors of the AR have also been identified. The nuclear receptor corepressor (NCoR) regulates transcription of the AR induced by dihydrotestosterone binding [10].

RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

The presence of three domains in the AR as well as the possibility for interaction with coactivators raises several questions, including which coactivators bind to the LBD, the function and linkage of NTD–LBD interactions, the competition between coactivators and the NTD for LBD binding, and the role of LBD in hormone-independent activation. The observed interaction with corepressor molecules raises additional questions, including whether the LBD binds corepressors, how it might act in repression, how currently-used androgen antagonists such as hydroxyflutamide promote repression, and whether repression by NTD–NCoR interaction increases when the ligand is unavailable or when the LBD has bound an appropriate antagonist. Finally, the question arises as to whether the AR leaves the nucleus after active repression. Although answers are not available to most of these questions, they raise the possibility of rational interference with the coactivators that work with the AR.

Intervention with nuclear receptor function is possible, as shown in studies of progesterone [14] and oestrogen binding [15]. Binding of oestrogen [16] or the agonist diethylstilbestrol occurs deep inside the oestrogen receptor protein, leading to interaction of the mobile helix 12 with other immobile structural elements to form a binding site for a coactivator [17]. Binding of an antagonist such as tamoxifen perturbs the structure of the nuclear receptor, perhaps moving the mobile helix 12 to block the coactivator site rather than forming a binding site for the coactivator (Fig. 1). Coactivators compete with corepressors at this site. In peroxisome proliferator–activated receptors (PPAR), the corepressor SMRT (small mediator of retinoic acid and thyroid hormone receptor activity) competes effectively with the coactivator by forming a three-turn α-helix that prevents the activation helix of the receptor from assuming an active conformation. Binding of the antagonist GW6471 reinforces binding of the corepressor [18].

image

Figure 1. Nuclear receptors can be inhibited indirectly by blocking coactivation. Helix 12 is represented by light blue.

Download figure to PowerPoint

These data suggest that interfering with the AR-coactivator protein interactions might be a useful strategy for down-regulating transcription activated by the AR. Drug designs that target protein–protein interactions rather than large surfaces or mixed motifs could generate drugs of mass < 600 Da, which tend to be more effective than drugs that target large interaction surfaces.

An alternative to targeting a protein–protein interaction would be to target biomotifs, subsets of which are highly conserved. Design of novel therapeutic agents directed at these biomotifs would ideally combine computational chemistry, structural biology, and synthetic chemistry to create small molecules that mimic binding faces of protein complexes. Several biomotifs that might interact with the AR have been identified (Table 1). Steroid receptor coactivator (SRC) family proteins, which act as housekeeping coactivator proteins for nuclear receptors, can serve as the basis for constructing peptides of only 12 amino acids. These peptides are fully equivalent to the entire coactivator and bind tightly to the leucine-X-X-leucine-leucine (LXXLL) motifs in the NTD of the AR [19]. These motifs are small, and could be potential targets for therapeutic intervention. However, the AR has three unique groups of biomotifs. First, the biomotif phenylalanine-X-X-leucine-phenylalanine (FXXLF) causes interaction between the NTD and the LBD. It also mediates the interaction between the hormone-dependent and hormone-independent activation functions (AF), known as AF1 and AF2 respectively. For some transcribed genes, AF1, within the NTD, is the primary contributor for AR functional activity, whereas AF2 makes a lesser contribution. Second, the two families of housekeeping corepressors, SRCs and NCoRs, have complex motifs that provide other potential targets. Third, the AR itself has an additional motif in the NTD.

Table 1.  Biomotifs with potential for interaction with the AR
BIOMOTIFMOLECULE AND FUNCTION
LXXLLSRCs, p160
FXXL(F/Y)ARA70, ARA55, ARA54, FHL2, AR
LXXXIXXX(I/L)SMRT, NCor
WHTLFAR

Androgen ablation and blockage of androgen action through the AR produces an initial response in most patients with cancer. However, in some cases prostate cancer cells are induced to proliferate by antiandrogens exerting an agonist effect, and androgen dependence is eventually lost during treatment. This has been shown in LNCaP, a prostate cancer cell line, which responded to antiandrogen as an agonist but became androgen-unresponsive. Thus, new approaches to inhibition of AR-mediated prostate cancer growth are required [20,21]. ARA70 is an example of a coregulator that acts via a biomotif, creating potential for therapeutic intervention. This protein in oligomeric form enhances the transcriptional activity of the AR and mediates the agonist activity of antiandrogens [22]. When a dominant-negative ARA70 was added to the culture, AR transcription was inhibited and androgen-independent growth of the tumour cells was halted. Dominant-negative ARA70 or RNA interference–mediated silencing of ARA70 reduced the agonist activity of hydroxyflutamide and rescued the normal function of the antiandrogen [23]. These data suggest that, in at least some prostate cancers, ARA70 is a required factor for disease progression. ARA55, another AR coregulator, also seems to function in the acquired agonist activity of antiandrogens [24].

Identification of ARA70 as a required factor for prostate cancer progression suggests that blocking ARA70 binding could reduce AR activity. A rational design of an agent to block ARA70 binding to the AR would target the interaction either with the NTD or with the ARA70 domain itself. Interaction would be hypothesized to occur at the helix 12 coactivator site, as is common for other nuclear receptors.

How might such a drug be designed? Although seemingly paradoxical, rational drug design might lead to the development of inactivators for activators, and activators for inactivators (Fig. 2). It is obvious that helix 12 must be in the correct position to bind the coactivators necessary for the androgen response. However, although this hormone-binding site is completely buried inside the protein (Fig. 3), in the absence of hormone binding, helix 12 moves in response to corepressor binding. The corepressor is larger, and binds at the same site as the hormone, but binding pushes helix 12 aside [18]. One approach would be to create a biomotif inhibitor that prevents binding of the coactivator, thus causing active repression or active inactivation. Flutamide and other antiandrogens bind somewhere in the LBD, but they bind inappropriately so that the location of helix 12 is perturbed and the coactivator cannot bind. An ideally designed version of flutamide would support active recruitment to allow the protein to fold, and would help corepressor binding. Bicalutamide, an alternative antiandrogen that stabilizes interaction of the AR with heat-shock protein, does not prevent the AR binding to DNA, but rather stimulates assembly of a transcriptionally inactive receptor on DNA [25].

image

Figure 2. Corepressors (red) compete effectively with coactivators (green) to block transcription.

Download figure to PowerPoint

image

Figure 3. Hormone analogues can be designed to inactivate coactivators or to enhance binding of corepressors. Cylinder + sphere: helix 12; small cylinder: hormone; pentagon: coactivator; large cylinder: corepressor; lined sphere: biomotif inhibitor; rectangle: hormone analogue activator; pyramid: hormone analogue inactivator.

Download figure to PowerPoint

Imaging the AR at the highest possible resolution (1.4 Å) gave data that provides a foundation for a structural approach to the problem. Creating a three-dimensional (3D) structure from this map (Fig. 4A) showed binding of dihydrotestosterone to the binding site within the AR. Designing an antagonist to this binding might involve either adding a bulky group to the hormone analogue to alter the coregulator binding, or blocking the protein interface with a biomotif that prevents the coregulator from binding (Fig. 4B).

image

Figure 4. (A) Structural data set for the AR bound to dihydrotestosterone; (B) antagonist designs for AR.

Download figure to PowerPoint

Eight biomotifs in the SRC family have been well described. The AR binds only one of these biomotifs, the SRC2 Box 3 LXXLL domain, a motif distinct from that to which the oestrogen receptor binds (Fig. 5A). Binding of the ARA70 FKLLF domain appears similar (Fig. 5B), but in reality it does not bind to the same residues. This domain has more articulation and bigger hydrophobic side chains than the LXXLL motif. Mutations in glutamic acid 897 on helix 12, a site that would affect binding of coactivators, have not been found. However, a mutation at arginine 855 in the LBD led to near normal binding properties but thermolability and decreased AR N/C terminal interactions resulted in partial androgen insensitivity, whereas a different mutation at the same site resulted in very low binding and no transactivation, and a phenotype of complete androgen insensitivity [26]. It therefore appears that a different coactivator might be responsible for androgen-insensitivity syndromes. Some mutations have been found in subsets of prostate cancer, suggesting that a coactivator may not always be required for activation (Fig. 6).

image

Figure 5. (A) SRC Box 3 LXXL domain and AR; (B) ARA70 domain bound to the AR.

Download figure to PowerPoint

image

Figure 6. Mutations in the ARA70 binding site seen in nonpathologic human mutations, in partial androgen insensitivity syndrome (PAIS), and in prostate cancer (PC).

Download figure to PowerPoint

The two principal families of coactivators, ARA and SRC, bind differently to the AR (Fig. 7), an observation that has not been made for any other nuclear receptor. Helix 12 moves, but the rest of the protein structure does not. If binding of the LXXLL motif to the oestrogen receptor is taken as the standard configuration of the receptor and coactivator domain, the LXXLL motif is translated about 1.5 Å whereas the FXXLF motif binds like the standard. The SRC helix is translated towards H3 (Fig. 8A). These data, combined with observations of the biomotif from within the protein (Fig. 8B), show a hydrophobic cluster of side chains interacting with three leucines presented by SRC, the oestrogen receptor, and by glucocorticoid receptors.

image

Figure 7. ARA70 (RED) and SRC3 bind differently to the AR.

Download figure to PowerPoint

image

Figure 8. (A) The SRC helix is translated 1.5 Å toward helix 3; (B) SRC motifs in the oestrogen receptor and glucocorticoid receptors show leucine residues at the AF2 interface; (C) isoleucine residues in the AR opens groove to accommodate hydrophobic motifs.

Download figure to PowerPoint

Substituting isoleucine at residue 737 in the AR provides a possible target site for rationally designed antiandrogens (Fig. 8C). The leucine in the oestrogen and glucocorticoid receptors projects more deeply into the pocket than does the isoleucine in the AR. The groove is therefore relatively small and shallow, and the retraction of the isoleucine in the pocket allows two very deep phenylalanine rings to be inserted into this site in the receptor.

To design a drug based on the parameters described above, the association of the AR with its coregulators (both coactivators and corepressors) was investigated. All of the peptides from a phage-display library that bound to the AR LBD were identified, and the 3D structures of these peptides and of the natural coregulator domains were determined. The bound structures of the natural coactivators were compared with the 3D structures of the binding peptides from the phage-display library. Binding affinity was then measured and compared with binding affinity of the identified structures. The data showed that the pocket not only fits phenylalanine as expected, but also fits leucine and tryptophan (Table 2). The SRC1 Box 2 peptide domain did not bind to the AR as a coactivator, but did bind to other steroid receptors. The precise mechanism behind the differential binding of coactivators is not clear, but the specificity might reside in a highly charged set of three aspartic acids beyond the biomotif.

Table 2.  AR biomotif structures at 1.4–2.5 Å resolution
  1. Specificity derives in part from flanking amino acids

• SRC1 BOX3KENALLRYLLDKDD 
• SRC1 BOX2KHKILHRLLQDSSAll amino acids in domain resolved
• ARA70RETSEKFKLLFQSYN 
• Rac3HKKLLQLLT 
• FXXLFSSRFESLFAGEKESR 
• FXXLWSSKFAALWDPPKLSR 
• FXXFFSRFADFFRNEGLSGSRCore amino acids in peptide resolved
• WXXLFSRWQALFDDGTDTSR 
• LXXLLSSRGLLWDLLTKDSR 
• FXXYFSSNTPRFKEYFMQSR 

Binding of the FXXFL biomotif dramatically changed the coregulator binding site. After the receptor was presented with the coactivator, the pocket opened, methionine, lysine, and valine moved, and the ARA70 binding set was accommodated (Fig. 9). The SRC housekeeping family did not fill the site as completely as did ARA70, but both appeared to open the same site on the receptor.

image

Figure 9. Binding of the FXXFL biomotif to the coregulator binding site of the AR alters the conformation of the receptor. The coloured regions are surfaces of side chains that move on binding the biomotif.

Download figure to PowerPoint

Surface plasmon resonance analysis of the binding driven by hydrophobic interactions was used to determine the differences in binding affinity between natural coactivators and the AR compared with that of the designed peptides and the AR. The apparent affinity as measured by kdiss was ≈ 1 µm for the SRC family of natural coactivators, ≈ 2 µm for tryptophan, and ≈ 1 µm for phenylalanine (Fig. 10). The structural changes did carry a cost in entropy, but the binding constant remained in the micromolar range.

image

Figure 10. Surface plasmon resonance analysis of binding of peptides to AR.

Download figure to PowerPoint

LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

Fluorescence polarization assays of a library of natural activators of nuclear receptors showed that, unlike other nuclear receptors, SRC2 Box 1 and SRC 2 Box 3 bound, while SRC 2 Box 2 did not (Fig. 11). The data also suggest that, with regard to its two biomotifs, ARA70 binds with FXXLF but not with LXXLL. Finally, a structural profile of the AR AF2 showed two well-defined hydrophobic sites close together, ideal locations for the generation of binding energy by burying hydrophobic amino acids.

image

Figure 11. Fluorescence polarization assays of a library of natural activators of nuclear receptors

Download figure to PowerPoint

The data from structural analysis of the AR and its coactivators led to identification of three possible classes of antiandrogens, all inhibitory small molecules targeted to the AR LBD. Class 1 antiandrogens were designed to block binding of dihydrotestosterone at the LBD. Intelligent high-throughput screening and FASTrack chemistry (3D crystallography of single or mixtures of chemicals) was used to investigate Class 2 molecules, where coactivators block at the coactivator site. Class 3 molecules were also LBD site-directed, but involved re-engineering the AR to preferentially bind corepressor rather than coactivator.

Preparation of antagonists to the thyroid hormone receptor (TR) illustrated an approach to generating nuclear receptor antagonists [27]. One compound synthesized was identical to the TR, but had a phenyl group rather than an iodine. The resulting molecule did not bind to the alpha receptor (TRα), but bound to the beta receptor (TRβ) as tightly as did natural thyroid hormone (Fig. 12). This suggested that an agonist could significantly change the 3D structure of a receptor without affecting activation, and might actually enhance activation. The converse proposal is that there should also be antagonists that destabilize the receptor and degrade it, causing it to lose its repression function. This hypothesis was tested by adding a large hydrophobic group, which abducted two helices by 4–5 Å. Helix 12 was not affected, resulting in a functional receptor that bound the coactivator molecule completely normally. Assays of transcription (not shown) proved normal activation despite the large distortions of the receptor. Where is the source of stabilization energy that changes the receptor but allows it to function normally? Adding a bulky group to the TR ligand was shown to have added new hydrophobic stabilization energy between existing hydrophobic side chains buried within the receptor, near the hormone site, and the added phenyl group. There were four hydrophobic side chains in the cluster with normal hormone, and seven with the hormone analogue. Many side chains rearranged, clustering into a solvent-shielded hydrophobic island, leading to stabilization of helix 12. With helix 12 stable and in the correct place, the receptor binds coactivator and functions normally in cells.

image

Figure 12. Addition of large hydrophobic groups enhances hormone binding to TR.

Download figure to PowerPoint

This approach could be put into practice either by identifying possible compounds with computational chemistry followed by synthesis of the compounds, or by high-throughput 3D screening with X-ray crystallography. The latter was applied; one 3D X-ray diffraction data set was measured from AR crystals soaked in a series of compounds with properties like those of the side chains binding to the coactivator site on the receptor. Crystals were soaked in simple organic molecules such as 2-methylindole. The electron density showed that 2-methylindole appeared to fit the coactivator binding pocket in the AR. In a first screen, 10% of the organic molecules that were added to AR crystals attached to the nuclear receptor coactivator binding site (unpublished data). In future experiments, after identifying the groups that bound well, compounds will be altered to increase affinity. Repeated iterations of these steps should eventually yield molecules that can block the AR coactivator.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES

Computational chemistry combined with structural analysis of nuclear receptors and determination of binding affinities of natural and designed coactivators and corepressors provides the theoretical framework for rational design. Adding alternative groups to various sites throughout the receptor can rebuild the conformation of the molecule and alter its functional binding with coactivators or corepressors. Possible molecules can be identified thoroughly and systematically with intelligent high-throughput screening and FASTrack chemistry. These techniques should eventually result in therapeutic agents for androgen-independent prostate cancer that can prevent binding of AR coactivators while increasing binding of AR corepressors.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. STRUCTURE AND FUNCTION OF THE AR LENDS ITSELF TO RATIONAL DRUG DESIGN
  5. RATIONAL DRUG DESIGN TARGETS PROTEIN–PROTEIN INTERACTIONS AT THE MOLECULAR LEVEL
  6. LESSONS LEARNED FROM THE THYROID RECEPTOR CAN BE APPLIED TO HORMONE ANALOGUE DESIGN
  7. CONCLUSION
  8. REFERENCES