The earliest example of small molecules known to alter the actions of NRs and that bind outside the LBD were described for ERα. These where pyrimidines that block the interaction of purified estradiol (E2)-activated ERα with labeled SRC1 box 2 peptide in fluorescence polarization assays (Rodriguez et al. 2004) (Table 1). Posterior optimization of these inhibitors led to a pyramidine based inhibitor with an inhibition constant (K_i) value of 2–3 μM using time resolved fluorescence resonance energy transfer (FRET). From then on, many examples of different inhibitors have shown to have effects in vitro and in vivo against ER (Becerril and Hamilton 2007; Zhou et al. 2007). Examples are amphiphatic benzene inhibitors that proved to inhibit mediated transcription of ERα in the endometrial adenocarcinoma HEC-1 cells (Gunther et al. 2009a), or the guanylhydrazone ERI-05 discovered by high throughput screening and that inhibits endogenous expression of the ERα regulated gene pS2 in the breast cancer MCF-7 cell line at a 20 μM concentration (LaFrate et al. 2008; LaFrate et al. 2009) (Table 1). The first crystal structure of ER co-crystallized with one of these coactivator binding site inhibitors was that of 4-hydroxy-tamoxifen (4-OHT) (Fig. 2) (Kojetin et al. 2008; Wang et al. 2006a). Tamoxifen, which has been used for more than 30 years, is one of the most effective treatments for ER-positive breast cancer. This inhibitor, which has been shown to bind to the LBP, had previously hinted that its total binding capacity to ERα was 2 times that of estradiol, the crystal structure ERβ-4-OHT surprised the scientific community by demonstrating that 4-OHT binds both to the AF-2 groove and the expected LBP (Kojetin et al. 2008, Jensen and Khan 2004). The binding of this 4-OHT to the surface appears to be mostly hydrophobic with numerous van der Waals interactions helping to accommodate this inhibitor perfectly in the hydrophobic patches of the coactivator-binding groove (Kojetin et al. 2008; Wang et al. 2006a; Wang 2006b). The binding of 4-OHT to the ER AF-2 pocket may account for the mixed agonist/antagonist activity of type I antiestrogens; nevertheless this discovery opens potential therapeutic avenues for improved antiestrogens.

As for ER, the function of most NRs has been modulated through the modulator binding to LBP. This was the case for TR where alterations in function was achieved mostly through analogues of the thyroid hormone (T3), for instance in the treatment of hypothyrodism. The first description of a non-LBP binding inhibitor for TR was that of a β-amino ketone. Arnold et al . carried out a high through put screen (HTPS) utilizing SRC2 box 2 peptide as a fluorescent probe for fluorescence polarization (FP) assays with the TRβ isoform. This HTPS led to a group of α/β unsaturated ketones that were shown to irreversibly bind to an exposed cysteine on the surface of the AF-2 groove of TR, with the best 3-(dibutylamino)-1-(4-hexylphenyl)-propan-1-one (DHPP) having an IC₅₀ of 2 µM (Table 1). Several lines of evidence including TRβ mutants and mass spectroscopy (MS) analysis have led to direct evidence of a covalent adduct formation of these inhibitors to Cys298 in TRβ. This reaction has been shown to occur in two steps, initial beta elimination of the molecule DHPPA to form 1-(4-hexylphenyl)-prop-2-en-1-one (HPPE) followed by a covalent bond formation between HPEE and TRβ. This is also consistent with the fact that a deamination reaction of DHPPA would not be feasible at a physiological pH without the aid of another active molecule. For DHPPA this mechanism is catalyzed directly on the protein surface. X-ray crystallographic studies of DHPPA with TRβ showed the deaminated active principle in the AF-2 groove supporting the specific mechanism of alkylation and deamination in situ catalyzed by Cys298 (Fig. 2b). These compounds were extensively studied and optimized using a structure-activity-relationship (SAR) approach. Though potency measured through FP did not seem to improve other important qualities such as solubility, permeability, cytotoxicity and selectivity towards TR over AR, ERα and PPAR were greatly improved (Arnold et al. 2007; Estébanez-Perpiñá et al. 2007c d; Arnold et al. 2006). Other HTPS have led to the discovery of methylsulfonylnitrobenzoates (MSNB) inhibitors (Hwang et al. 2011). MS and mutational studies demonstrated that these inhibitors also bind to TR irreversibly by targeting the Cys298 residue in the AF-2 pocket. Selectivity towards TRβ over AR and PPARγ as well as good solubility and cytotoxicity profiles made these inhibitors highly interesting. The most potent MSNB inhibitor was MI151 with a potency of 1.8 µM in the FP-SRC2 interaction assay with TRβ showed less off target activity than DPPAH (Hwang et al. 2011). These MSNB inhibitors were further optimized by the removal of the ester linker between the methylsulfonylnitro and the rest of the molecule to a cyclic bioisostere, such a thiazole. These inhibitors were therefore termed sulfonylnitrophenylthiazoles (SNPTs) and showed improved in vivo properties as well as the classical irreversible activity through Cys298 binding. Two inhibitors showed low µM potency, that of 2{4,1,5} and 2{3,1,2} showed by FP-SRC-2 binding assays to have an IC₅₀ of 0.3 and 2.4 µM respectively however the less active of these two presented additionally good transcriptional inhibitory activity (5 µM in the T3-mediated luciferase assay) (Hwang et al. 2012) (Table 1).

Inhibition of NRs through their AF-2 binding pocket has only been minimally attempted in other NRs. Two AF-2 small molecule inhibitors have been described for the vitamin D receptor (VDR/NR1I1) and although there is no in vivo data of their inhibition, they have proven to bind to this NR in vitro by a TR-FRET assay. The first non-LBP binding inhibitors for the VDR were the benzodiazepines derivatives (Mita et al. 2010) (Table 1). These inhibitors proved to affect the interaction between VDR and LxxLL coactivator peptides, whereby the most potent of the group showed an IC₅₀ of 20 µM in a VDR TR-FRET based assay (Mita et al. 2010). Monitoring of the effect of benzodiazepines in VDR mediated transcription, in the presence of increased concentration of its natural ligand (1,25 dihydrozyvitamin D3), showed no shift in the dose response curve, confirming the non-LBP binding nature (Mita et al. 2010). It was then shown that this inhibitor acts as an α-helix peptido-mimetic inhibitor, whereby the three alkyl side chains presented in this inhibitor mimic the three leucines in the LxxLL motif. The tetrahydrobenxodiazepinamine therefore acts as a rigid structure that mimics the α-helical arrangement and the methylbutene and the two isopentane groups mimic the leucines in the “NR box” motif of coactivators (Mita 2010). In a separate study, small molecule 3-indolyl-methanamides were identified as alternative site binders for the VDR LBD alternative binding was that of (Nandhikonda et al. 2012) (Table 1). These molecules block the interaction of VDR with SRC-2 in FP-based assays as well as inhibit VDR mediated transcriptions at a cellular level and presented with good selectivity towards VDR over AR, ERβ, TRα, TRβ and PPARγ (Nandhikonda et al. 2012). Surprisingly this indol-methanamine proved to bind to VDR irreversibly involving a reaction mechanism, whereby an azafulvenium salt is produced as an intermediate compound forming a positive like species, which can act with any nucleophile in the surface of VDR (Nandhikonda et al. 2012). However this promising inhibitor which has an IC₅₀ in the low micromolar range proved to have high cytotoxicity that is yet to be determined if it is due to the irreversible mechanism observed (Nandhikonda et al. 2012). More in vivo data is required in order to see if this inhibitor is a viable option for clinical inhibition of VDR.

Another example of non-LBP bound inhibitors was shown for the NR pregnane X receptor (PXR/NR1I2). PXR, as well as the constitutive androstate receptor (CAR/NR1I3), are two orphan NRs, meaning that no endogenous ligand has been identified as of yet (Willson and Kliewer 2002). However, it has been shown that they can function as xenobiotic sensors as they interact with many foreign chemicals (Biswas et al. 2009). Antagonists for PXR are of great interest due to their impact on the therapeutic efficiency of drugs and their ability to prevent harmful drug-drug interactions. Nevertheless, this is a highly complicated task as their LBP is highly flexible and can accommodate many ligands as has been repeatedly shown in the many structures of PXR co-crystallized with different ligands of different sizes (Xue et al. 2007; Watkins et al. 2001; Chrencik et al. 2005; Cheng and Redinbo 2011). The first non-LBP ligand described for PXR is ketoconazole, a known antifungal drug that is also used as an antiandrogen for metastatic PCa (Peer et al. 2014; Huang et al. 2007) (Table 1). This molecule antagonizes PXR in the presence of the antibiotic rifampicin, a known agonist of PXR and was also shown to disrupt the interaction of activated PXR with SRC-1 in a non-competitive fashion (Huang et al. 2007). Furthermore, the non-LBP binding nature of ketoconazole to PXR was confirmed through scintillation proximity assays, which excluded a potential LBP competitive mechanism. This was further confirmed by site-directed mutagenesis. Other inhibitors have been shown to bind to AF-2 such as the phytochemical coumestrol, found to bind to the surface of PXR in functional gene reporter assays and to antagonize PXR ligand dependent SRC-1 recruitment to the same extent regardless of the concentration of rifampicin (Want et al. 2008). FP-competition experiments with fluorescently labeled SRC-1 LxxLL ligands confirmed its binding to AF-2 (Huang et al. 2007;Wang et al. 2007; Ekins et al. 2007). This inhibitor was found to be highly selective for PXR except for some cross-reactivity with ER. The drug leflunomide is used for rheumathoid arthritis and is another inhibitor which was found to show inhibition of the PXR/SRC-1interaction and proved by site-directed mutagenesis, to directly bind to AF-2 (Ekins et al. 2008). However, more structural studies are needed to view the binding positioning of these inhibitors and additional in vivo data of these inhibitors has not been obtained to date (Ekins et al. 2007 2008).

ER, TR, PXR and VDR are all NRs whose AF-2 function has been shown to be a viable druggable non-LBP target. Nonetheless following ER, AR is the second NR whereby its non-LBP binding pockets have been extensively exploited for rational drug design. The first non-LBP binding inhibitors for AR were based on the pyrimidines inhibitors, initially designed for ER (Gunther et al. 2009b). The authors rationally designed tri-substituted peptidomimetic pyrimidines with bulkier groups than those initially utilized for ER emulating the AR preference for bulkier side chains (i.e FxxLF) in the NR boxes of co-regulatory proteins (Gunther et al. 2009b). This classical peptidomimetic approach achieved pyrimidines that were AR selective over ER and that had IC50 values ranging 1.5–6.6 μM as determined by AR inhibition in a transcriptional activated cellular assay (Gunther et al. 2009b).

Axerio-Cilies et al . carried out a structure-based drug design approach using all the structural information available on AR and its binding to different coactivators (Axerio-Cilies et al. 2011). This study led to a set of molecules with a rigid linker and two aromatic features that mimicked an FxxLF NR box: the diarylhydrazide inhibitors (Axerio-Cilies et al. 2011; Caboni et al. 2012). With high selectivity for AR over GR (NR3C1), ERα/β, but still showing some inhibition of PR, these inhibitors were shown to inhibit AR coactivator binding around 13–26 μM IC₅₀ in a TR-FRET assay. Furthermore, they induce prostate specific antigen (PSA) expression in the absence of androgens and to inhibit DHT and cyproterone acetate stimulated PSA (Axerio-Cilies et al. 2011; Caboni et al. 2012) (Table 1). A further example of the utility of peptidomimetics for AR targeting was published recently by Ravindranathan et al. (2013) . This group evaluated the X-ray structure of peptide coactivator sequence LxxLL/FxxLF bound to AR and designed two compounds mimicking these coactivator motifs. The main structural scaffold was that of a bis-benzamide group where two benzyl groups of compound D1 mimicked FxxLF motif and in the analogous compound D2, two isobutyl groups were used to mimick the LxxLL motif (Ravindranathan et al. 2013). Surprisingly, D1 did not show any inhibitory effect in AR-coactivator binding, however, D2 was shown to block the interaction of many coregulator proteins in co-immunoprecipitation assays using the LNCaP cell line (Ravindranathan et al. 2013). D2 also proved to block AR dependent transcriptional activity and prevent AR nuclear translocation as well as inhibiting AR-mediated PCa cell proliferation. Furthermore, it was reported to inhibit AR-dependent growth of xenograft tumors in vivo and AR expression in human tumors cultured ex vivo (Ravindranathan et al. 2013). These findings provide evidence that targeting the AF-2 binding site of AR using peptidomimetics may be a viable therapeutic approach for patients with advanced PCa.

Other functional FP screens with the aim of identifying potential PCa treatment molecules binding to the AF-2 function identified the off-patent nonsteroidal antiinflamatory drugs (NSAIDs) and thyroid hormones as potential antagonists of AR (Estébanez-Perpiñá et al. 2007a; Buzón et al. 2012). This screen led to the identification of: (i) two kinase inhibitors, 1-tert-butyl-3-(2,5-dimethyl-benzyl-)-1 H-pyrazolo[3,4,-D]pyrimidin-4-ylamine (K10) and 3-((1-tert-butyl-4-amino-1 H-pyrazolo[3,4D]pyrimidin-3-yl)methyl)phenol (RB-1); (ii) five small molecules, flufenamic acid (FLF), tolfenamic acid (TOL), meclofenamic acid (MEL), 2-methylindole and indole-3-carboxylic acid; and (iii) two thyroid hormones, triiodothyronine (T3) and 3,5,3’-triiodothyroacetic acid (TRIAC) (Table 1). The complex structure of the two kinase inhibitors with AR proved their binding preference to the AF-2 binding site. These inhibitors form a plethora of hydrophobic interactions and are additionally stabilized by H-bonding to the residue M734. RB-1 was shown to form an additional H-bond with the residue K720, a key residue that forms part of the “charge clamp” residues that normally stabilize the LxxLL motifs in their binding to the AF-2 site. Some small molecules such as 2-methylindole also showed by crystallographic data to be located in the AF-2 occupying the same subpocket of the AF-2 groove as the pyrimidine moiety of RB-1 and K10. The two thyroid hormones T3 and TRIAC also proved to bind to this coregulator binding groove, however these two inhibitors seem to not form any hydrogen bonds and be stabilized uniquely by week hydrophobic interactions (Fig. 2c). These two thyroid hormone compounds and some other small molecules in this screen proved to additionally bind to the exposed BF-3 pocket (Fig. 2e).