AKR1C3 as a Target in Castrate Resistant Prostate Cancer

4.1. N-Phenylanthranilate Based (N-PA) Inhibitors

These compounds are based on the non-steroidal anti-inflammatory drug (NSAID), flufenamic acid (FLU) (Figure 3). Our lab discovered that NSAIDs, which are used clinically for their ability to inhibit cyclooxygenases (COX) are non-selective AKR1C inhibitors at therapeutic concentrations required for COX inhibition [50]. The inhibitory activity of NSAIDs on AKR1C enzymes is believed to be due to the presence of carboxylic acid or ketone group which can interact with the conserved catalytic site/oxyanion hole of these enzymes [54]. Since, these compounds are relatively well tolerated and have been extensively studied; they can be used as lead compounds for the development of selective AKR1C3 inhibitors. Also, the relative simple structure of the N-PA scaffold is amenable to a large range of structural modification. However, to avoid adverse effects related to chronic COX inhibition, desirable compounds will have to be stripped of COX inhibitory activity and in addition lack inhibitory activity on AKR1C1 and AKR1C2. Using the known structure activity relationship (SAR) for COX, we were able to design compounds that were devoid of COX inhibitory activity; however, the compounds still had inhibitory activity for AKR1C1 and AKR1C2 at concentrations that will inhibit AKR1C3 [51].

Subsequent SAR studies on FLU that explored the effect of movement of the carboxylic acid group of the N-benzoic acid A-ring (Figure 3) and the introduction of various substituents on the phenylamino B-ring revealed key structural requirements for optimal AKR1C3 potency and selectivity [58, 59]. A fluorimetric assay based on the NADP⁺ dependent oxidation of S-(+)-1,2,3,4-tetrahydro-1-naphthol (S-tetralol) catalyzed by AKR1C3 at 37 °C was used to determine the inhibitory potencies of these compounds. Concentrations of S-tetralol used in the assays were equal to the K_M (5 μM, 22.5 μM, 165 μM and 25 μM for AKR1C1-4, respectively) for the respective enzymes determined under the same experimental conditions enabling direct comparison of their IC₅₀ values. The presence of the carboxylic acid group on the meta position of the benzoic acid ring and the introduction of an electron withdrawing group (EWG) at the para position of the phenylamino ring were critical for optimal AKR1C3 potency and selectivity. This led to the discovery of substituted 3-(phenylamino)benzoic acids as potent and selective AKR1C3 inhibitors [58]. These compounds represented the first reported compounds that inhibit AKR1C3 with nanomolar potency and display > 100 fold selectivity for AKR1C3 over AKR1C1-2. Unlike FLU, its isomer, compound 1 was 250 fold selective for AKR1C3. Of the 3-phenylaminobenzoic acids analogs synthesized and tested, the p-NO₂ analog compound 2, with an IC₅₀ value of 36 nM was the most potent AKR1C3 inhibitor, while the acetyl analog compound 3, with 360 fold selectivity for AKR1C3 over AKR1C1-2 was the most selective, Table 1. The AKR1C3 potency of these compounds displayed a statistically significant correlation with the electronic properties of the substituent on the phenylamino ring [58]. With the exception of 2, all the other compounds were devoid of COX 1 and 2 inhibitory activities at 100 μM concentrations. The compounds also did not display inhibitory activities on AKR1C4, AKR1B1 and AKR1B10 [59]. Importantly, lead compounds (10 μM) were effective in a LNCaP-AKR1C3 prostate cancer cell line where they produced a near complete inhibition of testosterone formation when the cells were treated with the AKR1C3 substrate Δ⁴-AD.

4.2. N-Benzoylanthranilic Acid Based Inhibitors

Due to the close structural similarity to N-phenylanthranilates, a series of N-benzoylanthranilic acids that were initially designed as inhibitors of the penicillin binding protein (PBP) were screened for AKR1C3 inhibitory activity and selectivity [60]. Of the 16 compounds that were tested, two compounds 4 and 5 inhibited AKR1C3 in a potent and selective manner. In the S-tetralol oxidation assay as described above, 4 inhibited AKR1C3 with an IC₅₀ value of 310 nM and displayed 164-, 162- and 65- fold selectivity for AKR1C3 over AKR1C1, AKR1C2 and AKR1C4, respectively. Compound 5 likewise inhibited AKR1C3 with IC₅₀ value of 350 nM and displayed 286-, 180- and 86- fold selectivity versus AKR1C1, AKR1C2 and AKR1C4, respectively (Table 1). SAR studies on these compounds indicate a need for a 3-OH group on the N- benzoylamino B- ring and a 5′-halogen on the benzoic acid A-ring. Movement of the carboxylic acid to other ring positions as well as substitution on other positions on the N-benzoylamino ring was not explored with these compounds. Hence, further modification of these compounds to increase potency as well as selectivity should be possible. It will be necessary to evaluate the efficacy of compounds 4 and 5 in cellular assays of AKR1C3 activity. Also, since the compounds are structurally similar to N-PA, it is essential to determine their effect on COX 1 and 2.

4.3. 3-(3,4-Dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids

Using a high throughput screening (HTS) approach to identify putative AKR1C3 inhibitors, Jamieson et al [61] screened a total of 99000 compounds. Filtering of hits for undesirable properties and a potency cut off value of < 1 μM led to 187 compounds out of which 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acid (compound 6) was deemed to be the most promising lead AKR1C3 inhibitor. The authors used a fluorimetric assay that monitored the AKR1C catalyzed formation of a fluorescent alcohol from a non-fluorescent ketone substrate (probe 5) [62] at 37 °C and pH 7.2 with NADPH as the cofactor. The authors also utilized a cell-based assay that monitored product formation when HCT-116 cells stably transfected with AKR1C3 were treated with the AKR1C3 substrate, PR-104H in the presence of inhibitors [63]. PR-104A is a bioreductive prodrug that is activated by AKR1C3 into the DNA cross linking metabolites, PR-104H and PR-104M [63]. Compound 6 was a significantly more potent AKR1C3 inhibitor than FLU and other marketed NSAIDs. It inhibited AKR1C3 with an IC₅₀ value of 13 nM and displayed > 1500 fold selectivity for AKR1C3 relative to the other AKR1C isoforms, Table 1. Compound 6 was likewise devoid of any inhibitory activity on COX 1 and 2 at 10 μM concentration. Subsequent SAR studies indicated that the orientation of the dihydroisoquinoline ring in compound 6 was optimal for AKR1C3 inhibition. AKR1C3 potency and selectivity was also found to be tolerant of small functional group substitutions on various positions of the dihydroisoquinoline ring. Relatively modest changes in AKR1C3 potency were observed as a result of the introduction of these substituents. These findings led to the identification of some of the most potent AKR1C3 inhibitors with several compounds having nanomolar potency and remarkable selectivity for AKR1C3 over other AKR1C isoforms. Compound 7, with a 6-Br group on the dihydroisoquinoline ring was the most potent and most selective inhibitor AKR1C3 inhibitor from the study. It inhibited AKR1C3 with an IC₅₀ value of 6 nM and was > 5000 fold selective over other AKR1C isoforms. Compounds 6 – 9 displayed relatively similar inhibitory potency in the cell based assay. Table 1.

4.4. Indomethacin Analogs

New indomethacin based AKR1C3 inhibitors have been reported by us [68]. These include three classes of compounds which are (i) indomethacin analogs; (ii) 2′-des-methyl-indomethacin analogs in which the 2′-methyl group has been removed; and (iii) 3′alkyl indomethacin analogs in which the position of the 2′-methyl group and 3′ acetic side chain have been reversed, exemplified by compounds 13–15. Each of these compounds have mid-nanomolar potency for inhibition of AKR1C3, are selective for AKR1C3 over AKR1C isoforms, have weak inhibitory potency against COX isozymes, and inhibit AKR1C3 mediated testosterone production in LNCaP-AKR1C3 cells. A crystal structure of 2-des-methylindomethacin, 14 bound to AKR1C3 was also obtained to probe the binding modes of these compounds. A full description of these analogs has been recently published [68].

4.5. Natural Products

Phytochemicals such as flavonoids and cinnamic acids have long been known to display anticancer properties. Their anticancer properties is believed to be as a result of inhibition of one or more of the following processes including cell adhesion, matrix metalloproteinase (MMP) secretion, angiogenesis, protein kinase activities and tumor cell invasion (reviewed by Kanadaswami et al. 2005) [69]. However, some of these compounds also inhibit AKR1C3 at nanomolar to low micromolar concentrations suggesting that AKR1C3 inhibition is a potential mechanism for their antineoplastic effects [48, 49, 69–71]. Endo et al [67] recently studied the potential anticancer mechanism of two compounds, baccharin (3-prenyl-4-(dihydrocinnamoyloxy)cinnamic acid) and drupanin ((E)-3-(4-hydroxy-3-(3-methylbut-2-en-1-yl)phenyl)acrylic acid) which are components of Brazilian honey bee propolis, Figure 3. Both compounds are naturally occurring prenylated cinnamic acid derivatives that display growth inhibitory effects by inducing apoptosis in human cancer cell lines [72, 73]. Using a variation of the S-tetralol oxidation assay at 25 °C and pH 7.4 (S-tetralol concentrations were 100 μM for AKR1C1 and 1mM for AKR1C2-AKR1C4), baccharin inhibited AKR1C3 with an IC₅₀ value of 110 nM (K_i = 56 nM) with no measurable inhibitory activity on AKR1C1 and AKR1C2 at 100 μM. This translates to > 1000 fold selectivity for AKR1C3. Drupanin on the other hand was about 150 fold less potent as a AKR1C3 inhibitor with little or no selectivity compared to baccharin. This disparity in AKR1C3 potency and selectivity was attributed to the importance of the dihydrocinnamoyloxy moiety of baccharin for AKR1C3 potency and selectivity. Baccharin was effective and selective for AKR1C3 inhibition in cellular assays where it inhibited the reduction of farnesal in MCF-7 cells and androsterone in A549 cells, with an IC₅₀ of 30 μM.

4. 6. Bifunctional AKR1C3 Inhibitors and AR Antagonists

Due to the challenges encountered with Abiraterone and MDV3100 and the adaptive response that often results in development of resistance to these compounds, combinations of both drugs has been proposed for the treatment of CRPC [17, 74]. Concurrent inhibition of androgen biosynthesis and AR signaling-the two pathways implicated in the development of progression of CRPC should produce maximal blockade of the AR axis, increase therapeutic efficacy and reduce the incidence of resistance. Additionally, the use of a bifunctional compound may minimize the risk of adverse effects that may occur with using two separate drugs. The new P450 17A1 inhibitor TOK001 satisfies some of this need since it not only inhibits the enzyme but also causes degradation of the AR. Unfortunately, due to targeting P450 17A1 the issue of drug induced adrenal insufficiency will remain and will require the co-administration of prednisone.

Novel FLU analogs were recently described as AR antagonists [75]. This suggests that AR and AKR1C3 ligands share a common pharmacophore which could be exploited to develop bifunctional agents that target both proteins. We recently reported the discovery of one such compound, 3-((4-nitronaphthalen-1-yl)amino)benzoic acid (compound 10), a “first-in-class”, dual acting AKR1C3 inhibitor and AR antagonist [64]. In the S-tetralol oxidation assay, compound 10 inhibited AKR1C3 with IC₅₀ value of 85 nM and was over 100 fold selective for AKR1C3 over other AKR1C isoforms. Unlike its phenyl analog 3, it did not display any significant effect on COX-1 and COX-2 with IC₅₀ values > 100 μM. It also produced a robust reduction in testosterone formation in LNCaP-AKR1C3 cells at 10 μM concentrations. Notably, compound 10 competitively inhibited DHT induced AR transcriptional activity in an AR-luciferase reporter gene assay with an IC₅₀ value of 4.7 μM and produced a concentration dependent reduction in cellular AR levels in the presence and absence of DHT. Dual inhibition of AKR1C3 and AR represents a novel concept with immense potential for the treatment of CRPC. It will be critical to understand the functional implication of using bifunctional agents like compound 10 in prostate cancer cell lines that express either one or both targets.

4.7. In- silico Screening

Using a fragment based virtual screening approach, a set of 143000 compounds were docked into the active site of AKR1C3 [65]. The highest ranked 33 compounds were then tested for AKR1C3 inhibitory activity. Of the new set of compounds, 11 was the most promising. In the S-tetralol oxidation assay, it inhibited AKR1C3 with an IC₅₀ value of 200 nM, was about a 1000 fold selective versus AKR1C2 and was even more selective for AKR1C3 relative to AKR1C1/AKR1C4. This compound has yet to be evaluated in a cell based assay. Schuster et al also identified a novel AKR1C3 inhibitor from a pharmacophore –based virtual screening [66]. One of the most interesting compounds from their study is compound 12. Compound 12 displayed an IC₅₀ value of 290 nM for AKR1C3 in a radiometric assay that measured the ability of the compound to inhibit the NADPH dependent reduction of [³H]-Δ⁴-AD by AKR1C3. The assay was conducted at pH 6.6 with a substrate concentration of 21 nM. The authors did not evaluate the compounds selectivity for AKR1C3 over other AKR1C enzymes but the compound was shown to preferentially inhibit AKR1C3 over 17β-HSD Types, 1, 2, 3, 4 and 7.

5.1. N-PA based inhibitors and 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids

Both N-PA based inhibitors and 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids utilize the SP1 pocket and interact with the oxyanion site (Figure 6). This similarity in binding conformations is not too surprising since these compounds share a similar A-ring -linker- B-ring scaffold. The A-ring provides the basic anchorage of the inhibitor by hydrogen bonding between the carboxylate/carboxamide substituent and residues Tyr55 and His117. The B-ring optimizes inhibitor selectivity by targeting specific properties of the AKR1C3 SP1 pocket. The bulky bicyclic B-ring of the 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids offers superior selectivity by taking advantage of the size of the AKR1C3 SP1 pocket,[61] whereas the smaller monocyclic B-ring of the N-PA compounds greatly benefits from strong electron withdrawing substitutions that may form dipole interactions or hydrogen bonding within the SP1 pocket [59, 77, 78]. Besides the B-ring, inhibitory selectivity is also affected by the linker region that directs B-ring into the SP1 pocket. Optimal selectivity is obtained when a single bridge atom connects the meta-carboxylate substituted A-ring and the substituted B ring [59, 61]. Choice of the bridge atom is not limited to nitrogen as in N-PA or sulfur as in the 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids. 3-Phenoxybenzoic acid (3-PBA), which isosterically replaces the amine linker of the N-PA with an oxygen atom, adopts a similar binding pose in AKR1C3 as Compound 1 [76]. Not surprisingly, 3-PBA likewise displays similar AKR1C3 inhibitory potency and selectivity as the unsubstituted 3-phenylaminobenzoic acid ([58, 59] and Unpublished observation Adeniji.A.O). The rigidity of the linker is proposed to enhance selectivity once the B-ring is correctly positioned [61]. It should be noted that some of the 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids carrying extraordinary large A-ring substitutions instead of the carboxylic acid are well accepted by AKR1C3. These substitutions apparently could not be accommodated in the oxyanion site, suggesting that the anchorage to the oxyanion site may be beneficial but not obligated for high binding affinity. These compounds likely adopt an alternative binding pose resembling the one used by indomethacin at lower pH as described below (Figure 7A). This conformation utilizes the SP3 pocket instead of the oxyanion site.

5.2. Indomethacin Based Inhibitors

Indomethacin analogs are the class of AKR1C3 inhibitors that show the most diversity in binding positions (Figure 7). At pH 6.0 (PDB ID 1S2A) [78], indomethacin occupies the SP3 pocket with the indole ring and leaves its p-chorobenzoyl ring at the entrance to the SP1 pocket. The carboxylate group helps to anchor the indole ring by hydrogen bonding to the diphosphate bridge of the cofactor and Gln222. With a pH increase to 7.5 (PDB ID 3UG8), [77] the weakened interaction in the SP3 pocket allows the carboxylate group to occupy the oxyanion site, so that the p-chorobenzoyl ring extends into the SP1 pocket and the indole ring partially stays in the SP3 pocket (Figure 7A). The pH-dependency of the binding conformations is likely caused by the different protonation state of the phosphate bridge, O1N/O2N phosphate oxygens in particular, on the cofactor, have a pK_a around 7 [77]. Since most of the inhibitor studies are performed around pH 7, it is likely that both binding poses contribute to the inhibitory potency of indomethacin analogs [53, 61]. Removing the methyl group on the indole ring gives 2′-des-methyl-indomethacin which binds differently. In this instance the indole ring enters the steroid binding channel while the carboxylate group remains tethered to the oxyanion site [79]. This binding pose also permits a second molecule of 2′-des-methyl-indomethacin to enter the active site. But kinetic data suggest the second molecule is unlikely to contribute to the inhibition potency. In contrast, replacing the indole ring with a pyrrole ring as in zomepirac elicits a minimal effect on the binding pose as zomepirac maintains the indomethacin binding pose at pH 7.5 (Figure 7C)(PDB ID 3R8H, pH undefined) [77]. Sulindac possesses minor structural changes compared to indomethacin. The inhibitor exhibits a novel binding pose not seen with the other NSAID based AKR1C3 inhibitors (PDB ID 3R7M, pH undefined) [77] (Figure 7D). The carboxylate group anchors sulindac to the oxyanion site while the p-chorobenzoyl ring extends to the entrance to the SP2 pocket, a location known to be occupied by substrate prostaglandin D₂ (PDB ID 1RY0) and the inhibitor rutin (PDB ID 1RY8) [80]. Flanagan et al [77] stated that the sulindac conformation causes significant movement of the Ser129-Phe139 loop (part of Loop A), but the loop region was not modeled in the published structure. Sulindac has a bulkier sulfoxide substitution on the benzoic ring in place of the chorine in indomethacin, which likely precludes the compound from adopting the indomethacin binding poses. However, sulindac is a prodrug that is activated in the gastrointestinal tract by the reduction of the sulfoxide to a sulfide. The small size of the sulfide may allow the compound to act as a potent AKR1C3 inhibitor in vivo.

5.3. Arylpropionic acids

The arylpropionic acids cocrystallized with AKR1C3 ((R)-flurbiprofen, PDB ID 3R94; (R)-ibuprofen, PDB ID 3R8G; (R)-naproxen, PDB ID 3UFY; (S)-naproxen, PDB ID 3R58) are less potent and less selective Inhibitors. The inhibitors occupy the SP1 pocket and the oxyanion site and even extend into the SP1 pocket to a similar depth to that of potent N-PA compounds (Figure 8), however, it is not apparent from the structures why the compounds are not as selective or potent. Besides, different studies have reported discrepant IC₅₀ values for these compounds using racemic mixtures and different assay methods, [61, 81] making it hard to extrapolate useful information.

5.4. Bifunctional inhibitors

In the course of developing AKR1C3 inhibitors, we identified a bifunctional compound 10 and recently determined its crystal structure in complex with AKR1C3 (PDB ID 4DBS) [64]. Compound 10 is designed based on the FLU scaffold and its binding pose showed significant resemblance to FLU and the other N-PA compounds. However, unlike the N-PA compounds, there are two molecules of inhibitor bound per enzyme monomer (Figure 9). Both molecules project their naphthyl rings into the SP1 pocket, forming a double decker arrangement and both molecules are deemed necessary for AKR1C3 inhibition as suggested by kinetic data [64]. The 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids also possess bicyclic rings but bind only one molecule per AKR1C3 active site (Figure 6). The size and position of the sulfonyl linker probably prevents the binding of two molecules of the inhibitor per enzyme monomer. Future optimization of compound 10 may result in a single molecule of inhibitor bound at the active site. This may be accomplished by replacing the 1-naphthyl group of compound 10 with a 2-naphthyl group to mimic the structure of the 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids. The mechanism by which compound 10 acts as an AR antagonist has not been fully elucidated. The compound could down regulate AR expression or directly interact with AR. N-PA compounds have been reported to bind to the AR surface and block coactivator recruitment [82]. Docking simulations demonstrate that compound 10 could bind to the AR ligand binding domain with high affinity (unpublished data).

Despite the fruitful structural studies on AKR1C3-inhibitor interaction, molecular simulations to predict future inhibitor design have been hampered by the size and flexibility of the AKR1C3 subpockets. Inhibitors distinct in size, shape, and binding conformation could exhibit similar affinity for AKR1C3. Almost all the highly selective AKR1C3 inhibitors elicit significant residue/loop movement of the enzyme. The pH-dependency of the binding conformations could also be a setback for computer-based drug discovery. Docking studies have only succeeded in predicting the correct binding conformation on limited occasions [77]. For future in silico screening for AKR1C3 inhibitors, it will be necessary to allow highly flexible residues to assume multiple rotamer conformations and carefully examine the protonation state of the enzyme, cofactor, and ligand upon binding.