Aldo-Keto Reductase (AKR) 1C3 inhibitors: a patent review.

Aldo-keto reductase (AKR) 1C3 (type 5 17β-hydroxysteroid dehydrogenase; prostaglandin [PG] F 2α synthase; and dihydrodiol dehydrogenase X) belongs to the AKR superfamily of proteins [ 1 – 4 ]. AKR1C3 is a drug target due to its involvement in intratumoral androgen biosynthesis in prostate and breast cancer. In breast cancer, AKR1C3 is a principal source of testosterone (T), the substrate for aromatase. The ability of AKR1C3 to regulate ligand access to the androgen receptor (AR) and estrogen receptor (ER) in a tumor-specific fashion makes it a superior drug target than either AR or ER antagonists [ 5 – 8 ]. By acting as a PGF 2α synthase [ 9 , 10 ], it also deprives peroxisome proliferator activating receptor γ(PPARγ) of its putative ligand Δ 15 -PGJ 2 and generates ligands for the PG FP receptor to stimulate the mitogen activated protein kinase cascade to promote cell proliferation [ 11 , 12 ]. AKR1C3 inhibitors thus offer promise for the treatment of hormonal and hormonal-independent malignancies [ 11 , 12 ]. AKR1C3 inhibitors have been used in clinical trials of castration-resistant prostate cancer (CPRC) [ 13 ] and acute myeloid leukemia (AML) [ 14 ]. CRPC is the fatal form of prostate cancer. This disease remains androgen dependent despite castrate levels of circulating T. Androgen dependency remains since the tumor undergoes adaptive responses to sustain AR signaling. One mechanism involves adaptive intratumoral androgen biosynthesis and a second mechanism involves changes in the AR itself [ 15 , 16 ]. Intratumoral androgen biosynthesis is targeted by P45017A1 (17α-hydroxylase/17,20-lyase) inhibitors to prevent the conversion of pregnenolone to dehydroepiandrosterone (DHEA) in the adrenal and hence deprive the tumor of its source of androgens. Abiraterone acetate (Abi) 1 is the P45017A1 inhibitor in clinical use and is approved by the US FDA [ 17 – 19 ]. The second mechanism, involving AR, is targeted with the super AR antagonist enzalutamide (ENZ) 2 [ 20 ]. To surmount the CNS side effects seen with this drug, a second-generation compound AN-509 (Apalutamide) 3 has been developed ( Figure 1 ) [ 21 ]. These antagonists not only bind to the AR but prevent its nuclear translocation and binding to chromatin. Both Abi and ENZ increase median survival time in CRPC patients by only 3–4 months before drug resistance occurs [ 17 – 19 , 22 , 23 ]. Thus, a critical clinical unmet need is for better agents. While multiple mechanisms can contribute to drug resistance, both Abi and ENZ resistance can be surmounted in xenograft models using the AKR1C3 inhibitor indomethacin [ 24 , 25 ] first identified by Byrns et al. [ 26 ]. AKR1C3 is overexpressed in prostate cancer as part of the adaptive response to androgen deprivation therapy (ADT). AKR1C3 is overexpressed in cell lines deprived of androgens [ 27 , 28 ], in prostate cancer xenografts in castrate mice [ 24 , 25 , 28 ] and in CRPC patients [ 29 – 31 ]. AKR1C3 is involved in all pathways to T and 5α-dihydrotestosterone (DHT) in the prostate due its 17-ketosteroid reductase activity; it reduces 4-androstene-3,17-dione to T (by the canonical pathway) [ 32 ]; it reduces 5α-andros-tane-3,17-dione to DHT (by the alternative pathway) [ 33 ]; it reduces androsterone to 5α-androstane-3α,17β-diol (3α-diol) which is then oxidized by 17BHSD6 to DHT (by the backdoor pathway) [ 34 – 36 ]; and it reduces DHEA to 5-androstene-3β,17β-diol which is converted by 3HSDB1 to T. AKR1C3 inhibitors would block all pathways to T and DHT within the tumor and could surmount drug resistance to Abi and ENZ ( Figure 2 ). AKR1C3 is also overexpressed in ductal carcinoma in situ of the breast [ 37 ]; its overexpression is correlated with the expression of ERα [ 38 ], and with breast cancer relapse [ 39 ]. By acting as a peripheral 17β-hydroxysteroid dehydrogenase that converts 4-androstene-3,17-dione to T, AKR1C3 becomes a peripheral source of T so that aromatase can synthesize 17β-estradiol in the breast [ 6 ]. Thus, AKR1C3 inhibitors have a place in the treatment of ERα positive breast cancer and offer an advantage over aromatase inhibitors that would block estrogen biosynthesis systemically. AKR1C3 inhibitors have been exploited in AML to alter PG signaling. In combination with PPARγ agonists, e.g. bezafibrate (BZF), the AKR1C3 inhibitor 6-medroxyprogesterone acetate (6MPA) 4 gave a superior response than was achieved by either agent alone [ 15 ]. In this treatment, BZF could stimulate PPARγ signaling and 6MPA would block the formation of PGs of the F series that would bind to the FP receptor ( Figure 3 ). This is the first clinical example of the use of AKR1C3 inhibitors in a nonhormone-dependent malignancy. The development of AKR1C3 inhibitors that are potent and selective is challenging since it is highly related to AKR1C1, AKR1C2, and AKR1C4 that share more than 86% sequence identity and their inhibition in the context of prostate cancer would be deleterious. For example, AKR1C1 converts DHT to 5α-androstane-3β,17β-diol (3β-diol) a proapoptotic ligand for ERβ and its inhibition should be avoided [ 40 ]. Similarly, AKR1C2 inactivates DHT by forming 3α-diol and its inhibition should be avoided [ 41 , 42 ]. By contrast, AKR1C4 is liver specific and is required for the synthesis of bile-acids and its inhibition would lead to bile-acid deficiency [ 43 ]. Despite this challenge, both academic and industrial groups have filed patents on AKR1C3 inhibitors ( Table 1 ).

Potent and selective AKR1C3 inhibitors that are based on non-steroidal, steroidal, and natural product scaffolds have been disclosed. Compounds reported but not claimed in patent applications include the N -benzoylanthranilates 5 [ 44 ], the 2,3-arylpropenic acids 6 – 9 [ 45 ], and the natural products stylopne 10 (an isoquinoline alkaloid) [ 46 ] and 2′-hydroxyflavone 11 [ 47 ]. Compound 5 was synthesized by coupling 3-hydroxybenzoate with 4-bromoaniline and the 2,3-diarylpropenic acids were synthesized by a Perkin reaction between substituted benzal-dehydes and functionalized aryl acetic acids in base ( Figure 1 ). The synthesis of the lead agent SN33638 12 , a potent and selective inhibitor of AKR1C3 (IC 50 = 13 nM), is based on an N -phenylsulfonyl-piperidine [ 48 ]. SN336381 is also related to the N -phenylsulfonyl indoles 13 , developed by Astellas and claimed in patent W02007100066 ( Table 1 ). Building on this lead, a series of N -(benzimidazolyl)-, N -(indolyl)-, and N -(-pyridinelpyroylyl)-carbonyl piperidines 14 – 16 were claimed in JP201202018, W02010101127, and W02010101128, respectively, by Astellas. The N -(indolyl)-carbonyl piperdines include ASP9521 15 , which was taken through to a Phase 1/11b clinical trial [ 14 , 49 ]. A related series of N -benzimidazole or N -indole benzoic acids 17, 18 were claimed in W02009014150, WO2010087319 ( Figure 4 ). A screen of existing drugs identified nonsteroidal antiinflammatory drugs (NSAIDs) as selective inhibitors of different AKR1C isoforms [ 26 ]. Compounds of interest include indomethacin 19 which was selective for AKR1C3; the N- phenylanthranilates (e.g. meclofenamic acid 23 ) were pan-AKR1C inhibitors; and aryl pro-pionic acids (e.g. naproxen 25 ) displayed equal AKR1C2 and AKR1C3 selectivity and potency. This led to the repurposing of these NSAIDs for AKR1C3 inhibition while eliminating inhibition of the COX-isoforms (PGH synthase I and II). Patent WO2012122208 was issued for N -phenylaminobenzoates represented by 24 and patent WO2013059245 was issued for indo-methacin analogs, represented by 20 – 22 [ 50 , 51 ]. This was subsequently followed by patent W02017070448 for β-naphthylacetic acids ( R -naproxen analogs), 26 [ 52 ]. In each case, the NSAID analogs were subjected to medicinal chemistry optimization to remove structural features required for inhibition of COX-1 and COX-2 while inhibition of AKR1C3 was retained. Indomethacin gave rise to three classes of analogs: Class I analogs (retain the core structure of indomethacin, 20 ); Class II analogs (are des-methyl indomethacin compounds in which the 2′-methyl group has been removed, 21 ); and Class III analogs (are 3′-alkyl derivatives where the acetic acid side chain has been substituted with an alkyl group and the carboxylic acid side chain has been moved to the 2′-position 22 or the 3′ and 2′ positions are cyclized to yield a cyclic carboxylic acid or sulfonamide) [ 53 , 54 ]. Indomethacin analogs with conserved 5′-methoxy groups and p -chlorobenzoyl groups at the indole N1 position were synthesized by the method of Yamamoto [ 55 , 56 ] . The key reagent for the underlying Fischer indolization is 4-chloro- N -(4-methoxyphenyl)-benzohydrazide hydrochloride. Target compounds containing a 3′-propionic acid group or a 2′-des methyl-group were readily obtained from the benzohydrazide hydrochloride by using either a slight excess of 5-oxohexanoic acid (here, R 1 = Me, n = 2) or 4-oxobutanoic acid (R 1 = H, n = 1) in acetic acid, respectively, to give, 20 – 21 . Use of 4-oxohexanoic acid (R 1 = Me, n = 1) quantitatively yielded the reverse 2′-pro-pionic acid/3′-alkyl indole derivative, 22 [ 57 ] ( Figure 4 ). Following the issuance of patent WO2013059245 for these indomethacin analogs, a patent claiming the use of indomethacin for CRPC was filed, WO2015065919. For the N -phenylaminobenzoates, simple coupling chemistry involving a Buchwald–Hartwig C–N coupling reaction followed by saponification of the formed methyl ester produced an extensive library of compounds that are claimed in patent WO2012122208 and US 20140107085 [ 50 , 51 ]. Compounds in which the arrangement of the amine and carboxylic acid was changed from ortho- to meta- position followed by introduction of a para- electron withdrawing group on the B-ring gave compounds of mid-nanomolar potency and selectivity for AKR1C3 ( Figure 4 ). For the aryl propionic acids 25 , β-naphthylacetic acids in which the stereochemistry at the alkyl substituent at the alpha carbon was changed from S- to R- were sufficient to abolish COX-1 and COX-2 inhibition but retain AKR1C3 inhibition; compounds such as 26 are disclosed in WO2017070448 ( Figure 4 ) [ 52 ]. Bifunctional AKR1C3 nonsteroidal inhibitors have also been disclosed ( Figure 4 ). Isoquinolines represented by the lead compound GTX-560 27 not only act as competitive inhibitors of AKR1C3 but also block its AR coactivator function which was previously unknown [ 58 ]. The isoquinolines were claimed in patents WO2013142390 and WO2014039820A1 filed by GTx-Therapeutics. BMT4-158 28 , which is a N -naphthylaminobenzoate, is covered by the patent on the N -phenylaminobenzoates and acts as a bifunctional AKR1C3 competitive inhibitor and direct acting AR antagonist [ 59 ]. Attempts have been made to develop steroidal-based inhibitors for AKR1C3 as it reduces 17-ketosteroids. Extensive nitrogen and sulfur-substituted estrenes with the core structure 30 have been claimed by Bayer in four patents WO201345407, WO2014009274, WO2014128108, and WO02016037956 ( Figure 5 ). The key features of these steroids are the presence of either an amide, amine, or sulfone at the C3 position of the steroid coupled with the presence of either a nitrogen heterocycle or a trifluorosulfone at the C17 position. Natural products such as baccharin analogs 29 (from the Brazilian propolis) have also be claimed as AKR1C3 inhibitors, and these derivatives contain a phenolic cinammic acid substituted with an isopropyl group and a phenylpropionic ester [ 60 , 61 ]. However, these compounds are likely to hydrolyze in vivo to the corresponding alcohol and acid.

Thirty-five crystal structures of AKR1C3·NADP + ·inhibitor complexes exist in the PDB. Inspection of these structures shows that if the inhibitor contains a carboxylic acid, it can often form hydrogen bonds with the catalytic tetrad members Tyr55 and His117. Other portions of the inhibitor can occupy one of several subpockets (SP), e.g. SP1 Ser118, Asn167, Phe306, Phe311, and Tyr319 (e.g. occupied by the B-ring of N -phenylaminobenzoates). The SP2 sub-pocket refers to Ser129, W227, and F311 (e.g. occupied by the side-chain of PGs), and the SP3 sub-pocket which contains Y24, E192, S217, S221, Q222, Y305, and F306 [ 62 ]. While the presence of these sub-pockets can be rationalized to determine binding mode and can be used as the basis of docking studies, some important caveats exist as illustrated by the binding of indomethacin. Two different binding poses for indomethacin exist in the AKR1C3·NADP + ·indomethacin depending on pH. In the AKR1C3·NADP + ·indomethacin complex at pH 6.0 (PDB ID 1S2A), where indomethacin is fully protonated, the carboxylate is anchored by Q222 and Y24 in SP3, the bridge carbonyl forms a hydrogen bond with Tyr55 through an intervening water molecule, and there is no occupancy of SP1. However, in the AKR1C3·NADP + ·indomethacin complex at pH 7.5 (PDB ID 3UG8), where indomethacin is deprotonated, the drug rotates so that the carboxylic acid now forms a hydrogen bond with Tyr55, the SP1 pocket is now occupied by the p -chlorobenzoyl ring, and there is interaction between W227 with the methoxyindole in the SP2 pocket ( Figure 6 ) [ 54 ]. These structures illustrate the difficulty in performing structure-based inhibitor design for AKR1C3.

Tiered screening has been conducted to support patent claims. Tier 1 screening includes in vitro inhibition assays on recombinant AKR1C3 to claim compounds with mid-nanomolar affinity. Counterscreens have been performed in many instances versus either AKR1C1 or AKR1C2, to claim compounds that are 40–500-fold selective for the target (see Table 1 ). Many compounds have cleared this screen, but often only IC 50 values are reported and the pattern of inhibition is not given. Since AKR1C3 catalyzes an ordered bi-bi mechanism, in the reduction direction, two inhibitor complexes can form e.g. E·NADPH·I (competitive complex) and E·NADP + ·I (uncompetitive complex) [ 26 ]. Thus, depending on the mode of inhibition, the IC 50 values may not be directly comparable. Tier 2 screening for repurposed NSAIDs includes a subsequent counter screen against all the human AKRs, and a counter screen against COX-1 and COX-2. This level of screening was conducted for patents WO2012122208 and US 20140107085 and patents WO2013059245 and US 20160303082. In other patents, specificity was assessed by demonstrating the inability of leads to inhibit HSD17B3, the major androgenic 17β-HSD found in the testis and a member of the short-chain dehydrogenase/reductase superfamily [ 63 ]. Tier 3 screening includes cell-based assays to determine whether compounds inhibit the conversion of 4-andros-tene-3,17-dione to T in LNCaP-AKR1C3 cells or another prostate cancer cell model in which AKR1C3 is overexpressed. Often HEK-293 cells expressing AKR1C3 have been used as a substitute. These screens determine whether the inhibitor has cell bioavailability and retains potency. Claimed compounds have been shown to be effective in these models, albeit with some loss of potency. Cell-based assays using AR-reporter gene assays and AR-ligand binding assays have also been performed to determine whether compounds act as AR-antagonists or inhibit the co-activator function of AKR1C3, as is the case for GTx-560 [ 58 ]. The AR coactivator domain of AKR1C3 was located to amino-acid residues 171–237 by deletion mutagenesis [ 58 ], which is distal to the enzyme active site. This region contains a coactivator peptide consensus peptide LXXLL (LEMIL). This suggests that some small molecule competitive inhibitors may have an allosteric effect that radiates to distal portions of the protein to affect AKR1C3–AR interaction. Interestingly, indomethacin does not have this property [ 58 ]. Tier 4 screening determines whether AKR1C3 inhibitors are effective in vivo and cause a reduction in tumor volume or tumor incidence in either xenograft or patient-derived xenografts of prostate cancer. ASP9521 and indomethacin have been shown to inhibit tumor growth in xenografts ex-vivo and in vivo , respectively [ 24 , 25 , 49 ]. Similar results have been obtained with GTx-560 [ 58 ]. Some attention to the xenograft model is required. Demonstration of reduced tumor incidence and volume in SCID mice transplanted with prostate cancer tumors is not a model of CRPC. CRPC can be modeled if the recipient mouse is castrated after the transplant and the tumor then regrows under castrate conditions. This model has been rarely used. No experiments have been performed with AKR1C3 inhibitors in patient-derived xenografts. Nevertheless, proof-of-principle xenograft data indicate that AKR1C3 inhibitors are effective antitumor agents in animal models [ 24 , 25 , 49 ]. Based on preclinical data, ASP9521 15 was advanced to a Phase I/IIb clinical trial by Astellas. ASP9521 was found to be well tolerated but without efficacy [ 14 ]. In this small trial, 7/13 patients completed the regimen. Serum levels of ASP9521 reached levels that would be sufficient to inhibit AKR1C3. However, decreases in serum PSA and serum steroid hormone levels were not achieved. However, inclusion criteria did not screen for AKR1C3 expression and the authors concluded that drug failure may have been due to the exclusion of patients who had been on prior ADT, which is known to upregulate AKR1C3. The steroid-based estrenes with substitutions at C3 and C17 have been shown to be potent competitive inhibitors in vitro using recombinant AKR1C3 and in HEK-293 cells over-expressing AKR1C3. However, counter screens against other human AKRs have not been reported. The presence of the nitrogen heterocycle at C17 is reminiscent of the heterocycle found in Abi and raises issues as to whether they inhibit P45017A1 or other steroid metabolizing P450 isoforms.

As AKR1C3 is a major peripheral 17β-HSD required for the synthesis of T and DHT, inhibitors of the enzyme may have a place for treating endocrinological disorders associated with androgen excess in males and females, e.g. prostate cancer, benign prostatic hyperplasia, alopecia, pattern baldness, hirsutism, polycystic ovarian syndrome, etc. As T synthesized locally is also a substrate for aromatase, AKR1C3 inhibition may also be desirable in breast cancer, endometriosis, and endometrial cancer. However, the major focus has been on prostate cancer. The majority of AKR1C3 inhibitors claimed are mono-functional agents and act downstream from Abi. Since they do not inhibit P45017A1, they do not have to be coadministered with prednisone to prevent adrenal insufficiency. The monofunctional AKR1C3 inhibitors would be superior to other P45017A1 inhibitors (orterenol and galeterone) since they target an enzyme involved in intratumoral androgen biosynthesis that is overexpressed upon ADT. Even though P45017A1 inhibitors decrease serum DHEA-SO 4 and DHEA by more than 90%, the amount of DHEA-SO 4 that remains leaves a substantial reservoir for intra-crine androgen biosynthesis by AKR1C3 [ 64 , 65 ]. Mechanisms of drug resistance to P45017A1 inhibitors also include HSD3B1 allelic variants that stabilize the enzyme responsible for the conversion of DHEA to 4-androstene-3,17-dione [ 66 ]. The properties of AKR1C3 inhibitors versus other agents that target the AR axis in prostate cancer are presented in Table 2 . AKR1C3 is overexpressed in prostate cancer cells, in xenografts, and in tumors of patients that over undergone ADT [ 24 , 25 , 28 – 31 , 67 ]. But it is likely that the use of these mono-functional AKR1C3 inhibitors will require precision medicine to ensure that the target is expressed in the patient. Interestingly, steroid 5α-reductase inhibitors e.g. finasteride and dutasteride are not approved by the FDA for the treatment of prostate cancer since these may cause the appearance of a more aggressive disease ( Table 2 ). Few compounds claimed in the patents have undergone a complete counter screen and for many, DMPK studies have yet to be performed limiting their effective use in animal xenograft and human studies. Here, the repurposed NSAIDs hold promise since they are anticipated to retain the properties of the parent drug from which they were derived [ 5 , 68 ]. Monofunctional AKR1C3 inhibitors are ultimately predicted to fail in the clinic due to the issue of drug resistance. However, they could be added to existing regimens, e.g. Abi or ENZ with the prospect of achieving a synergistic effect and more durable drug response. A starting point would be to add indomethacin to patients who progress on Abi or ENZ. Both Abi and ENZ resistance are likely to involve the overexpression of AKR1C3 as a component of drug resistance [ 24 , 25 ]. Mechanisms of drug resistance include AR gene amplification [ 69 ], the selection of AR mutants that make it ligand promiscuous [ 70 , 71 ], and the appearance of AR splice variants (AR-SV) that have lost their ligand binding domain and are constitutively active [ 72 – 74 ]. The bifunctional AKR1C3 inhibitors, e.g. GTX-560 27 , offers promise since it blocks the coactivator function of AKR1C3 on full-length AR ( Table 2 ). Whether AKR1C3 can act as a coactivator of AR-SVs is unknown. The other bifunctional agent claimed is BMT4-158 28 , which acts as a competitive inhibitor of AKR1C3 and as a direct acting AR antagonist, suggesting that single agents that target intra-tumoral androgen biosynthesis and AR signaling can be developed. Whether these single agents would be superior to a combination treatment of Abi plus prednisone plus ENZ remains to be determined.