Introduction

Prostate cancer (PCa) is the most common male malignancy and second leading cause of cancer deaths among men, with 30 to 40% of patients diagnosed with localized disease eventually develop metastatic disease [1,2,3,4]. The androgen receptor (AR), a ligand inducible steroid hormone receptor, is involved in regulation of prostate growth, spermatogenesis, and bone and muscle mass; however, it is the key determinant in the initiation and progression of PCa [5]. Most patients with PCa respond dramatically to antiandrogen therapy; unfortunately, the median duration of response to hormone therapy is usually only 12 to 18 months, with the majority of patients eventually progressing to castration resistant PCa (CRPC) [6, 7].

Members of nuclear receptor (NR) superfamily, including the AR, act as ligand-activated, DNA-binding, transcription regulatory factors. NRs contain three main functional domains: the C-terminal ligand-binding domain (LBD) receives the signal, the central DNA-binding domain (DBD) binds to a response element near the target gene, and the N-terminal domain (NTD) modulates gene transcription. All NR LBDs share a common three-dimensional structure encompassing 12 α-helices (numbered 1–12) and two small β-sheets [8,9,10]. NR ligand specificity is achieved mainly by specific interactions between the ligand and the amino acids lining the ligand-binding pocket (LBP) [11]. In response to ligand-binding, NR LBDs undergo conformational changes and helix 12 moves from an extended (open) position to a closed position, thereby sealing the LBP [12]. Repositioning of helix 12 completes the assembly of the activation function (AF-2) domain, which contains a conserved hydrophobic cleft flanked by opposing-charge residues (formed by specific amino acids from helices 3, 4, 5, and 12) [8, 13, 14].

Therapeutic resistance and the realization that prostatic neoplasms could outwit hormonal therapeutics through gain-of-function somatic mutations of the AR is the biggest shortcoming of current androgen/AR-directed therapeutic modalities [15]. Recent documentation of the mutational profile of CRPC, has identified the AR as the most commonly mutated gene [16]. Moreover, the intervention with androgen hormonal therapy in indolent disease has resulted in selection of diagnosis of patients presenting for aggressive disease [17]. Possible mechanisms of antiandrogen resistance in advanced PCa includes the presence of hypersensitive AR activity through AR amplification, overexpression of co-factors, or specific promiscuous somatic mutations. Somatic AR mutations, allow for a variety of other steroid hormones and classical antiandrogens to replace androgens as active ligand and allow continued activation of AR [18,19,20]. Somatic mutations of the AR, such as Thr877Ala (T877A), Trp741Cys (W741C), and His874Tyr, and their subversive activity in the presence of antiandrogens have been well documented [19, 21, 22]. Furthermore, the somatic mutation, Phe876Leu (F876L) has been identified and found to circumvent the actions of Enzalutamide (MDV-3100), the compound currently used in the treatment of CRPC [23].

Therefore, the acceptance of “watchful waiting” and non-intervention in early disease has led to a greater improvement in deferring more advanced PCa [24,25,26]. This has resulted in a real need to devise new therapeutics, where the onus is placed on treating early disease to prevent more advanced disease and limit the progression to therapeutic resistance disease. In this paper, we report the design and synthesis of a number of bivalent androgen compounds for efficient targeting of AR. By creating these bivalent androgen-peptide conjugates with a much-enhanced affinity for the AR-AF2 motif as a targeting arm, where effector moieties, such as radioactive agents, cytotoxic drugs, and small de-couplers which can act as prodrugs, can be delivered with no or limited systemic toxicity.

Experimental

Molecular Modeling

The crystal structure of the ligand-binding domain of AR (PDB code 1t7r) was used as the starting point for the modeling study. Structure manipulation and visualization were done in SYBYL 8.0 (Tripos, Inc., St. Louis, MO). In the crystal structure, dihydrotestosterone (DHT) is in the ligand-binding pocket of AR and the FxxLF motif-containing co-activator peptide is bound in the AF-2 pocket. Both N- and C-termini of the AR were modeled in the ionized state. The C-terminus of the co-activator peptide fragment was also modeled in the ionized state. All histidine residues were protonated. Hydrogen atoms were added and the complex was subjected to conjugate gradient energy minimization using a distance-dependent dielectric function (ε = 4r) and an 8 Å non-bonded cut-off down to an rms gradient of 0.01 kcal mol−1 Å−1 using the AMBER ff99SB force field [27, 28]. This complex served as the template for the initial design of our bivalent compounds. The resulting complex and other protein-ligand complexes were energy-minimized with the AMBER and GAFF [29] force fields.

Molecular dynamics simulations were carried out using the AMBER suite of programs. Each system was solvated in a truncated octahedron TIP3P water box. The distance between the wall of the box and the closest atom of the solute was 10.0 Å. Na+ and Cl− ions were added to produce an ionic strength of 0.15 M and to neutralize the net charge of the complex. Each system was energy-minimized, applying harmonic restraints with force constants of 1 kcal/mol Å2 to all non-hydrogen solute atoms, followed by heating from 10 to 150 K over 30 ps in the canonical ensemble (NVT). Heating was continued with harmonic restraints to bring the temperature from 150 to 300 K over 100 ps in the isothermal-isobaric ensemble (NPT) simulation and to adjust the solvent density under 1 atm pressure. The harmonic restraints were then gradually reduced to zero with four rounds of 250 ps NPT simulations. A 50-ns production run was then carried out with snapshots collected every 10 ps. For all simulations, 2 fs time-step and 8 Å non-bonded cutoff were used. The Particle Mesh Ewald (PME) method [30] was used to treat long-range electrostatics, and bond lengths involving bonds to hydrogen atoms were constrained by SHAKE [31]. For simulations containing peptidomimetic linkers, the AMBER force field had to be supplemented with parameters for the unnatural amino acids present.

Chemistry of SPEP Compounds

Preliminary modeling studies suggested the feasibility of extending the N-terminal end of an AF-2 peptide to connect to the 17β-hydroxy group of DHT through a carbamate linkage. Although substitutions at 17α [32] and 17β [33] could be found in the literature, we designed and successfully carried out a novel synthetic scheme for introducing such a linker at the 17β-OH position of DHT. AF-2 peptides were synthesized by Fmoc solid-phase chemistry using manual coupling (Fmoc amino acid, 4 equiv.; 2-(H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate (TBTU), 4 equiv; N,N-diisopropylethylamine (DIPEA), 6 equiv) in N-methylpyrilidone. Synthesized peptides were added to the linker in a subsequent reaction. A five-step synthetic procedure of the bivalent androgen-peptide conjugates has been developed. The first step of the five-step synthetic procedure adopted was the formation of the steroidal-chloroformate (A) by reacting 1 equiv (350 μM) dihydrotestosterone (DHT) with 1.5 equiv triphosgene, and 1 equiv anhydrous pyridine in anhydrous dichloromethane (DCM) for 5 h at room temperature. Subsequently, coupling of 1 equiv (570 μM) compound A with 1 equiv t-butyl-glycine in the presence of 2.5 equiv triethylamine (Et3N) in acetonitrile (MeCN) was carried out to form a glycine steroid carbamate (B). Finally, the t-butyl group was removed from compound C by trifluoroacetic acid (TFA) in dichloromethane. The glycine steroid carbamate was coupled with the peptide attached to the AF-2 motif peptide on solid phase to form compound D. The final step of the synthetic route was the cleavage of compound D from the attached resin, which generates compound E. Spectral analyses (mass spectroscopy, analytical HPLC) were consistent with the desired products (Scheme 1).

Scheme 1
scheme 1

A five-step synthetic procedure of the bivalent androgen-peptide conjugates

Full size image

Synthesis of the Stapled Peptide

We procured stapled AF-2 peptide from Biopeptide Co. Inc., (San Diego, CA). Synthesis carried out by Fmoc solid-phase chemistry using unnatural amino acid (S)-2-(2_-pentenyl) alanine to replace i and i + 4 position alanine (A) and glutamine (Q) of AF-2 peptide sequence respectively. The side chains of this unnatural alanine are then cross-linked with an olefinic bond [34].

In Vitro Binding Assays

We have developed an AR expression system for measure binding properties of bivalent androgen derivatives to androgen receptor. We isolated full-length His-tagged human AR from lysed transfected COS-1 cells by using Ni2+-affinity (Ni2+-NTA) chromatography [35]. Briefly, we performed competition binding assays by pre-incubating immobilized His-AR with specific androgen-peptide conjugates, or the DHT control, overnight at 4 °C. The following day, 5 nM [3H]-Mibolerone (MB, a synthetic nonmetabolizable androgen) and an additional aliquot of each compound are added. Specific [3H] MB-AR binding is measured after further 2.5 h incubation at 4 °C. The concentrations of compounds required to inhibit 50% of MB binding (IC50) are determined for these androgen peptide conjugates.

Cell Culture Competition Assays

Competition binding assay was performed by pre-incubating normal genital skin fibroblast (GSF) expressing androgen receptor with synthesized androgen peptide conjugate for 2 h at 37 °C. After the pre-incubation, this compound was incubated with [3H]-MB at room temperature for 20 min. These cells were then harvested and lysed and the % [3H]-MB binding in presence of compounds was determined. Finally, we determined percent inhibition of MB binding in genital skin fibroblast (GSF) in presence of specific concentration of compounds (conjugate concentration was 100 nM, except SPEP-24 which was 50 nM).

Results and Discussion

Bi-Valent Androgen Design Strategy

Among all-natural androgens, 5α-dihydrotestosterone (DHT) was selected as a moiety to occupy the androgen-binding pocket. DHT is more selective than testosterone for AR in prostate and has approximately three times greater affinity for the AR than testosterone. DHT acts as the primary androgen in the prostate in adults and like testosterone, it cannot be converted by the enzyme aromatase to estradiol. However, it is almost universal that any steroid manipulation leads to dramatic loss of affinity to the ligand-binding domain (LBD) [36]. To circumvent this problem, we sought to create a bivalent androgen-peptide conjugate where the utilization of two binding sites of AR in the LBD and AF-2 domains results in recuperating the loss of binding affinity associated with the manipulated steroid. A bivalent targeting strategy can exhibit as much as 20- to 50-fold increased affinity [37,38,39,40]. Non-steroidal bivalent compounds have also been synthesized for the Estrogen receptor and do show comparable binding to Estradiol, with some compounds’ similar effects on breast cancer cell line growth as raloxifene [41, 42]. The AF-2 surface provides binding sites for the recruitment of several co-activators through the co-activators’ conserved α-helical LxxLL or FxxLF motifs [43] (Fig. 1a). The AR AF-2 regions exhibit a distinct preference for co-activators with aromatic-rich motifs [8, 44, 45]. Such motifs have been found in N-terminal domain of AR and in an AR cognate family of co-activators that include AR-associated protein (ARA) 54, 55, and 70 [46,47,48]. The AF-2 region of AR-LBD mediates an androgen-dependent N-terminal/C-terminal (N/C) interaction through an FxxLF motif of the N-terminal AR. This interaction is thought to facilitate and stabilize androgen binding. Each motif forms a two-turn α-helix residing between charged glutamate and lysine residues of AF-2 of AR-LBD. The strategy then was to covalently link the androgen hormone with AF-2-binding peptides, via a spacer arm, to form our bivalent ligand (Fig. 1b). Aside from contributing to binding affinity, the AF-2 peptide moiety provides a much more versatile surface-exposed point of attachment than the steroid for a therapeutic abiotic moiety.

Fig. 1
figure 1

Rationale for the androgen-AF-2 bivalent ligand. a The androgen is completely buried. Attaching a payload directly to the androgen perturbs the AR structure and androgen binding, resulting in significantly reduced affinity. Linking the androgen to an AF-2 peptide, which binds to the AR surface, recuperates the binding affinity lost resulting from the derivatization of the androgen. The bivalent ligand has an AR binding affinity comparable to that of the androgen and attachment of a payload to the surface-exposed AF-2 peptide is relatively easy without affecting affinity. b Components of the androgen-AF-2 bivalent ligand

Full size image

Linker and AF-2 Peptide Design

The AR-LBD co-crystal structure shows that the binding sites of the DHT and SSRFESLFAG co-activator peptide (pdb code 1t7r) are 13 Å apart. The androgen-binding pocket is completely buried while the AF-2 binding site is on the surface with no obvious path to connect them. However, inspection of the crystal structure suggested the existence of a nascent channel between the two sites. Molecular modeling suggested that a 5-glycine linker (SPEP-1, Table 1) spanning this channel and conjugating DHT with a co-activator peptide could be accommodated with a modest protein conformational change after energy minimization of the complex (Fig. 2).

Table 1 Composition of bivalent ligands
Full size table
Fig. 2
figure 2

Energy refined structure of steroid-peptide conjugate 1 (SPEP-1) bound to AR LBD. The steroid and glycine linker are represented as capped sticks and AF-2 peptide as a cyan ribbon. The AR LBD is represented as an ash colored ribbon. To help orient the view, helix 12 of AR is also labeled

Full size image

To explore the optimal linker length, two other compounds SPEP-2 (Gly6 linker) and SPEP-3 (Gly4 linker) were made and tested. All three ligands had an AF-2 FxxLF peptide SSRFGSLFAG [extracted from the crystal structure (1t7r)] with one substitution from FESLF to FGSLF. Of these, SPEP-1 with a Gly5 linker showed good binding to AR (IC50 500 nM in an in vitro assay) while SPEP-2 (Gly4) and SPEP-3 (Gly6) failed to show any affinity for AR (Table 2). This suggests that Gly4 is too short to bridge the two sites while the additional two bonds in Gly6 relative to Gly5 creates a geometrical constraint on positioning the AF-2 peptide that is incompatible with proper binding.

Table 2 In vitro cell-free binding assays
Full size table

Using the optimal Gly5 linker, three other variants containing the FxxLF motif taken from the literature were explored (SPEP-4, SPEP-5, and SPEP-6; Table 1). The three flanking residues on either side of the motif were taken from ARA70 [50]. The SPEP-4 FQNLF motif was derived from the N-terminal domain of AR. SPEP-4 had an IC50 of 300 nM, a modest improvement over SPEP-1 (Table 2). With FAALF (SPEP-5) and FKLLF (SPEP-6), the IC50 improved further to 70 nM and 100 nM, respectively. We have also calculated the inhibitor constant (Ki) for all compounds.

SPEP-9 is a variant of SPEP-5 with an extra-long Gly10 linker. The extra five glycines allow the linker to extend out of the channel with enough flexibility to position the AF-2 peptide with minimal bias from the geometry of exit of the linker from the channel. SPEP-9 has an IC50 of 100 nM, comparable to SPEP-5 (Table 2). This suggests that the Gly5 linker exits the channel with a geometry that is near optimal for connecting to the preferred binding mode of the AF-2 peptide.

To reduce the number of peptide bonds and the polarity of the linker, we also used γ-Abu-Ahx (γ-4-aminobutanoic acid and 6-aminohexanoic acid) to replace four glycine residues in the linker. G-γ-Abu-Ahx is of similar length to Gly5 and modeling suggested that it could be accommodated well. SPEP-10 and SPEP-27 are analogs of SPEP-5 and SPEP-6, respectively, with their Gly5 linkers replaced by Gly-γ-Abu-Ahx. SPEP-10 and SPEP-27 have IC50s of 25 nM (3-fold improvement) and 50 nM (2-fold improvement), respectively. This shows a clear advantage to using Gly-γ-Abu-Ahx versus Gly5 as the linker.

Unnatural Amino Acids

To mitigate potential proteolytic degradation of androgen-peptide conjugates, we replaced one or two phenylalanines in the FxxLF motifs of SPEP-5 and SPEP-10 with unnatural amino acids napthylalanine (NAL) and cyclohexylalanine (CHA) (SPEP-11, SPEP-12, SPEP-13, SPEP-19, SPEP-21, and SPEP-22, Table 1). None of these exhibited improved affinities over their parent compounds (Table 2).

LxxLL Motif

Another series of compounds were designed using sequences from the members of the steroid receptor co-activator p160 SRC family. The p160 SRC family includes SRC-3, SRC-1 and TIF2/GRIP1/SRC-2 co-activators. A conserved feature of these co-activators is an α-helical LxxLL motif or NR box. The androgen receptor recruits SRCs to its ligand-binding domain using these short leucine-rich hydrophobic motifs (LxxLL). We therefore explored replacing phenylalanines with leucines in the AF-2 peptide. SPEP-26 incorporates the LVQLL sequence from SRC-1. The LxxLF motif has not been shown to be a bona fide AF-2 binding motif, thus SPEP-23 and SPEP-24 replace one or both phenylalanines in SPEP-10. Significant improvement in binding affinity was observed for SPEP-24 upon replacing the FxxLF motif with the LxxLF motif. SPEP-24 has an IC50 of 5 nM, 5-fold better than SPEP-10, the most potent compound in the FxxLF series. The LxxLL motifs of SPEP-23 and SPEP-26 have IC50s of 20 and 30 nM, respectively, comparable to SPEP-10.

Derivatization for Enhanced Cell Permeability

The presence of charged residues in SPEP-10 (SEKFAALFQSY) sequence may hinder the compound’s penetration through the plasma membrane, as the AR is located first in the cytoplasm and then translocates to the nucleus upon ligand binding. Therefore, a number of substitutions were made in AF-2 peptide sequence, aimed to enhance membrane permeability of the compounds. We generated four less polar analogs (SPEP-25, SPEP-29, SPEP-35, and SPEP-36) of SPEP-10 by replacing either the glutamate or lysine residue with an uncharged one (Table 1). SPEP-20, SPEP-25, SPEP-29, SPEP-35, and SPEP-36 did not show equivalent or better potency (IC50s of 200, 800, 500, and 75 nM, respectively) of binding vs. the parent compound SPEP-10. However, it was observed that solubility of the compounds could be an issue in these series of compounds.

Another strategy to improve cell permeability is to derivatize the peptides with fatty acids. Fatty acids with different carbon chain lengths were attached to SPEP-10 and SPEP-24 through lysine side chain amino groups by peptide linkage (Table 1). It is well known that medium chain fatty acids can enhance the cell permeability of certain drugs [51,52,53]. Molecular modeling suggested that incorporation of these fatty acids would not interfere with peptide binding. Therefore, hexanoic acid (SPEP-41), octanoic (SPEP-42, and decanoic acid (SPEP-43) derivatives of SPEP-24 were made and tested. These have in vitro IC50 values of 40, 50 and 25 nM, respectively. These are of lower affinity relative to SPEP-24 but still have respectable potency. SPEP-38 is an octanoic acid derivative of SPEP-10 and has an IC50 of 200 nM, an 8-fold decrease in potency, similar to the 5- to 10-fold decrease seen for the SPEP-24 derivatives.

Peptide Stapling

A number of structural analyses of AR-cofactor interactions shows that the FxxLF motif binds as a short helix located between a charge clamp at a fixed position of the AR AF-2 [45]. This helical structure is crucial to properly position the AF-2 peptide side chains for optimal binding. Our AF-2 peptide moiety is based on a short 11-residue fragment extracted from a larger domain. This short peptide, in the absence of the larger domain, may have a reduced helical stability, compromising its binding ability and also making it more prone to proteolytic digestion. Peptide stapling is a technique that locks a peptide in its rigid α-helical form, helping preserve its structural integrity and intermolecular interactions [54,55,56,57,58,59] (Fig. 4). Furthermore, preserving the intramolecular hydrogen bonding in the helix reduces the polar exposure of the amide backbone, thereby reducing the barrier to membrane penetration and increasing its resistance to protease cleavage. Typically, the amino acids to be linked are replaced by unnatural amino acids such as (S)-2-(2_-pentenyl) alanine whose side chains are then cross-linked with an olefinic bond. The AF-2 peptides are two-turn amphipathic α-helices. Molecular modeling suggested that an i, i + 4 crosslink for the olefinic alanines replacing the underlined A and Q amino acids in the AF-2-binding peptides, SEKFAALFQSY and SEKLAALFQSY, are not expected to interfere with AR binding. The olefinic amino acids are on the solvent-exposed face of the α-helix. The two stapled peptides (HGSF-1 and HGSF-2) were made and tested. HGSF-1 is a stapled version of SPEP-10 and has an in vitro IC50 of 5 nM, a 5-fold improvement over SPEP-10. HGSF-2 is a stapled version of SPEP-24 and has an IC50 of 2 nM, a 2.5-fold improvement. These results are supportive of the ability to design constrained fragments of known structural motifs that can affect critical protein-protein binding interactions as AR and co-activator peptide interactions, and possibly why they are considerably more potent than the unconstrained LxxLF motif.

In Vivo Binding

Eleven compounds with good in vitro binding (less than 100 nM) to AR were checked for their in vivo affinity for AR (Table 3) by competing for binding against synthetic [3H]-labeled androgen mibolerone in a genital skin fibroblast assay. The most potent ligands were SPEP-24 and SPEP-36, with 90 and 91% inhibition, respectively. SPEP-36 is 15-fold weaker than SPEP-24 in vitro but has comparable potency in vivo. This may be due to greater cell permeability of SPEP-36 resulting from mutating a charged glutamate to a neutral asparagine residue. Despite the good in vitro potency of the stapled peptides, in vivo potency was relatively weak. It is possible that the stapled peptides may have reduced cell permeability.

Table 3 Cell-based binding assay
Full size table

Molecular Dynamics Simulation

Figure 2 shows the energy-minimized structure of the complex of AR and SPEP-1. The introduction of the Gly5 linker is accommodated with relatively small adjustments in the AR structure. To further assess whether the designed bivalent conjugates would seriously perturb the structural integrity of the AR protein a 50-ns molecular dynamics simulation was carried out on SPEP-24, one of the designed conjugates that exhibited good binding in both in vitro and cellular assays. A stereoscopic snapshot of the structure of the complex of AR with SPEP-24 taken towards the end of the 50-ns MD simulation is shown in Fig. 3a. Superposition onto the crystal structure of AR with DHT and the AF-2 peptide (pdb code 1t7r) shows that the AR structure is not significantly perturbed, and the AF-2 peptide binding mode and helical structure is preserved. The integrity of helix 12 in AR is also maintained, albeit translated somewhat. Figure 3b shows a more detailed view of the interactions of the AF-2 motif with its binding site on AR. We see that tethering the AF-2 peptide to the steroid does not compromise its ability to bind to AR and can thus contribute to the overall binding affinity of the bivalent ligand. The hydrophobic groups in the LxxLF motif (labeled + 1, + 4 and + 5 in Fig. 3b) all point towards the AR and are buried. The leucine side chain at position + 1 interacts with I737 at the bottom of the binding site groove and is sandwiched by L712 and M734 at the sides of the groove. The leucine side chain at + 4 packs against L712, V713, and V716 in a relatively shallow binding region. The phenylalanine side chain at + 5 is buried deeply at the other end of the binding groove and is sandwiched between the aliphatic carbons of the K720 and the side chain of Q733 as well as interacting with F725 at the bottom of the groove. The modeled structure also shows that the + 2 and + 6 positions point to solvent and are compatible with peptide stapling in HGSF-1 and HGSF-2 (Table 1, Fig. 4). The DHT-bivalent SPEP-24 compound interaction complex appears to have equilibrated from 30 ns onwards to a structure with a backbone root-mean-square deviation (RMSD) of 1.7 Å from the crystal structure of AR (Fig. 5). This suggests that the introduction of the bivalent compound does not significantly perturb the structure of the AR LBD. As a control, a molecular dynamics simulation was also carried out for crystal structure of AR with DHT alone and a similar RMSD of 1.6 Å from the initial structure was observed.

Fig. 3
figure 3

Complex of AR and SPEP-24. a A stereoscopic snapshot of the complex of AR and SPEP-24 taken towards the end of the 50-ns MD simulation. The orientation of the complex is similar to that in Fig. 2. SPEP-24 is in teal with the steroid and linker represented as sticks. All other residues are represented as ribbons. Overlaid in orange is the crystal structure of a complex of AR, DHT, and an AF-2 peptide. b Detailed view of the interactions of the AF-2 peptide with AR. The ligand is represented with capped sticks and the AR LBD as a molecular surface. The LxxLF motif is labeled from + 1 to + 5

Full size image
Fig. 4
figure 4

Strategy adopted for stapled AF-2 peptides. a Cross-linked pentenylalanines replacing the + 2 A (n) and + 6 Q (n + 4) in SEKFAALFQAY. b. Interaction of HGSF-1 with the AF-2 of AR. The stapled peptide HGSF-1 are displayed as sticks (cyan) and AF-2 surface displayed as molecular surface (green). The stapled side chains are all solvent-exposed

Full size image
Fig. 5
figure 5

Structural dynamics of the bound AR-LBD and SPEP-24 complex. In black is the root-mean-square deviation (RMSD) along the molecular dynamics trajectory of all backbone atoms of AR LBD in complex with SPEP-24 with respect to the crystal structure of DHT and an AF-2 peptide bound to AR-LBD (pdb code 1t7r). In red is the corresponding plot for a molecular dynamics simulation for AR-LBD with DHT alone

Full size image

Conclusions

We have designed bivalent ligands of AR that utilize both the steroid and AF-2 binding sites. This was made possible by linking the two sites through a hitherto unrecognized nascent channel that is readily enlarged to accommodate a Gly5 linker (or the isosteric Gly-γ-Abu-Ahx peptidomimetic) with minimal perturbation of the protein structure. The most active of these ligands exhibited single-digit nanomolar IC50s in vitro. Moreover, some of these ligands retained good activity in cellular assays with up to 90% inhibition of mibolerone binding. The goal of this work was not to create androgen antagonists per se but rather to create potent and specific targeting agents that could eventually carry a cytotoxic payload. For example, the C-terminal Tyr in the ligand facilitates the introduction of radioactive iodine-131. The AF-2 binding moiety provides other solvent-exposed side chain positions that could be derivatized to incorporate photodynamic agents or other cytotoxic agents. Previous attempts at using the steroid alone as the targeting moiety to which a cytotoxic agent is attached have had limited success [36]. For example, an androgen-mustard conjugate, where 2-choroethyl aniline is connected to an 8-carbon atom linker at the substitution-tolerant C11 position of the androgen methyltrienolone (R1881), shows extensive cytotoxic properties, but with one-tenth the binding affinity of natural androgens, resulting in an unacceptable level of systemic toxicity [36]. The greater affinity of the bivalent ligand developed in this work could alleviate the problem of systemic toxicity for delivering cytotoxic agents. Therefore, the functionality of the modified compound-AR complex (i.e., agonist or antagonist activity) is not relevant; the vital parameter is its ability to deliver toxic agents or sensitizers to the tumorigenic cells. One of the advantages of choosing the AR and its ligands as a target system is the large amount of structural information available for the AR ligand-binding domain. All modalities listed above would incorporate active moieties that are extremely small, allowing for good access to cytoplasmic ARs.

The use of any present-day cancer therapeutics is likely to falter because of the uncanny ability of cancers to become resistant to such treatments. We propose that a different approach is needed, i.e., delivery of toxic physical modalities or sensitizers (drug/pro-drugs) not susceptible to biological resistance. Such de novo therapeutic agents require much improved targeting techniques. Prostate tissue and PCa have relatively high AR content (approximately 400 fmol/mg protein) [60]. Indeed, although the AR is expressed in other male reproductive organs, kidney, liver, and certain areas of the central nervous system, 18F-labeled-DHT given systemically targets prostate tissue [61,62,63,64]; similarly, these bivalent compounds should do the same Therefore, the bivalent ligands described here were developed specifically to target AR, with a similar strategy that could be pursued with other members of the steroid receptor family [41, 42]. Thus, the potential for toxicity in other tissues expressing AR is possible if the targeted active agents are directly toxic, but not so if the non-biological targeted agents are delivered as a pro-drug or sensitizer, i.e., a second modality that can also be targeted to the prostate is required to activate the pro-drug. Classical antagonists certainly can be seen to be ineffective in such cases, because they are required to “operate” within the cellular environment, as are most chemotherapeutics. Thus, cells with early somatic changes, rendering them resistant to the proposed therapy, or cells where AR expression is lost, can still be exposed to toxins delivered to the near-by environment. Cells devoid of the AR are unlikely to be affected. However, non-biological agents typically do not require a cellular environment to be effective and can be effective beyond the single host cell. A “radius effect” is commonly seen. The best example would be radioactive iodine-131 (131I), used in the treatment of thyroid cancer, where beta emission has been well documented to kill cells in a microenvironment, even though not all cells take up iodine. Moreover, this work may have much broader implications, as any neoplastic tissue harboring a sufficient concentration of a steroid receptor may be amendable to effective treatment employing compounds targeted to their respective receptors.