Steroidogenic factor-1 (SF-1) has been implicated in sex determination during development and in formation of steroidogenic tissues.^iv SF-1 is involved in endocrine function throughout life with expression in the pituitary, testes, ovaries, and adrenal gland.^v Knockout mice exhibit male to female sex reversal and impaired development of adrenals and gonads.^vi Due to the potential role SF-1 plays in regulation of steroid hormone synthesis including adrenal androgen and gonadal testosterone synthesis, selective control of this receptor could result in therapeutic treatment of metastatic prostate cancer.^vii Additionally the involvement of SF-1 in energy metabolism suggests relevancy in controlling obesity.^viii

Thus, the development of selective small-molecule biological probes of SF-1 is an important objective. Phospholipids have been found in the ligand binding domain of human SF-1^ix and recently the first small molecules with the ability to modulate the activity of this transcription factor were described.^x

Approximately 65,000 compounds were screened for SF-1 inhibition by the Molecular Library Screening Centers Network (MLSCN) at The Scripps Research Institute.³,^xi All initial hits were counter screened against the retinoic acid receptor-related orphan receptor α (RORα), a phylogenetically distant nuclear receptor,² in order to eliminate promiscuous as well as non-selective compounds. This led to the identification of two mid-nanomolar SF-1 selective inhibitors 1 (PubChem SID 7970631) and 2 (SID 7969543; Figure 1).^3,11 Accordingly, isoquinolinones 1 and 2 were selected as starting points for the development of SF-1 small molecule probes. We have developed and report herein three routes for the synthesis of 1 and 2 that enable different aspects of the SAR of this SF-1 inhibitor series to be examined. Among the analogs reported here, 31 and 32 have improved SF-1 inhibitor potency, lower cellular toxicity, and possess improved selectivity compared to the initial leads 1 and 2.

Our initial strategy for synthesis of analogs of 1 and 2 focused on sequential alkylations of an isoquinolinone core (Scheme 1). Treatment of commercially available 5-hydroxyquinoline with peroxyacetic acid provided N-oxide 4. Heating of the N-oxide 4 in acetic anhydride gave a diacyl derivative that was not isolated but converted directly to 5-hydroxyisoquinolinone 5.^xii Selective alkylation of 5 at the phenolic position by using ethyl (±)-2-bromopropionate (NaH, DMF, 25 °C, 80%) gave propionic ester 6. Subsequent N-alkylation of 6 with tert-butyl bromoacetate (Cs₂CO₃, DMF, 25 °C, 80%) provided diester 7. Acidic deprotection of the tert-butyl ester and subsequent coupling (EDC, DMAP, CH₂Cl₂) of the derived carboxylic acid with either benzyl amine 8 or aniline 9 provided authentic samples of 1 and 2, which were used to confirm the results of the MLSCN SF-1 inhibitor screen.

While the sequence summarized in Scheme 1 permitted us to synthesize a number of analogs, this route proved less than ideal especially for synthesis of isoquinolinone ether analogs owing to its linear nature. Especially problematic is that the chemoselectivity of alkylations of 5 with agents less reactive than α-halopropionate was poor, with N-alkylation of 5 competing with phenol O-alkylation with alkyl halides. This led us to develop the more convergent second generation synthesis summarized for 14 in Scheme 2.

Aldehyde 10, available from the Claisen rearrangement of m-salicaldehyde allyl ether,^xiii was alkylated in high yield with a variety of electrophiles (use of iodobutane is illustrated in Scheme 2). The resulting product was then oxidized using chromic acid to give acid 11.^xiv Standard peptide coupling of 11 with glycinamide 12 (EDC, DMAP, CH₂Cl₂, 83%) provided diamide 13, which was then converted to SF-1 inhibitor analog 14 via ozonolysis of the vinyl group and dehydration of the hemiaminal that is formed in situ. In some cases, the targeted isoquinolinone was obtained directly from standard workup of the ozonolysis reaction. However, in other cases the dehydration of the hemiaminal did not occur spontaneously. Therefore, as a standard procedure, a catalytic amount of I₂ iodine was added to a solution of the crude hemiaminal in CH₂Cl₂ to promote dehydration and aromatization.

In order to probe the SAR of the glycine spacer connected to the isoquinolinone nitrogen, we demonstrated that condensation of hydroxylactone 16 (accessible from ozonolysis of acid 15) and glycine tert-butyl ester 17a provides ester 7 (Scheme 3). Deprotection of 7 provides carboxylic acid 18, which is an intermediate in the sequence summarized in Scheme 1. Alternatively, amino acids such as glycine (17b), alanine and phenylalanine can be coupled directly with 16 to give 18, thereby obviating the need for use of the tert-butyl ester protecting group. Subsequent coupling of carboxylic acid 18 with a range of amines and anilines, using the conditions summarized in Scheme 1, greatly facilitated SAR studies of this amide substituent.

Finally, in order to probe the substitution pattern of the isoquinolinone nucleus of 1 and 2, a third synthetic strategy was developed starting from commercially available hydroxyindanones (Scheme 4); the synthesis of 22 starting from 19 is illustrative. Thus, after alkylation of 5-hydroxyindanone with ethyl (±)-2-bromopropionate (K₂CO₃, acetone, 60 °C, 4 h), indanone 19 was treated with TBS-OTf and Et₃N^xv to give the silyl enol ether which, without isolation, was treated with N-methylmorpholine N-oxide (NMO) and catalytic osmium tetroxide.^xvi The resulting α-hydroxyindanone 20 was cleaved using NaIO₄ to give the hydroxylactone, which was condensed with glycine tert-butyl ester under conditions summarized in Scheme 3. Subsequent coupling of the carboxylic acid derived from deprotection of 21 with aniline 9 provided the regioisomeric isoquinolinone analog 22.

SF-1 inhibition activity of analogs of 1 and 2 are summarized in Table 1, along with comparative inhibition data vs. RORα. All assays were performed as described in ref. ³. Briefly, luciferase-format functional assays were used to measure inhibition of constitutively active SF1(LBD)-Gal4 or RORα (LBD)-Gal4 chimeric constructs. Compounds 1, 2, 22−35 and 40−43 were studied as racemates, or as mixtures of diastereomers (41, 42).

Initial efforts focused on replacing the oxygenated benzylic amide or anilide units of 1 and 2, but introduction of alkyl amides as in 23 and 24 led to significant loss of SF-1 activity. Replacement of the highly oxygenated benzylic amide unit of 1 with potentially more metabolically stable benzyl (25) or p-fluorobenzyl (26) amides led to 5−50 fold loss of SF-1 activity. However, introduction of a 2-phenylethyl amide (28) yielded a 300 nM SF-1 inhibitor. Still better results were obtained by replacing the potentially metabolically sensitive anilide unit of 2 with other aniline amides, such as the p-methylphenyl (30), p-methoxyphenyl (31), p-trifluoromethoxyphenyl (32), and p-ethoxyethyl (33) amides. These analogs displayed SF-1 inhibitor activity in the 100−200 nM range, a three to six-fold improvement over 2.

All substitutions of the R” unit spanning the isoquinolinone and amide units led to substantial or complete loss of SF-1 activity (analogs 40−43), as did moving the alkoxy substituent from the meta position in 1−2 to the para position in 22. However, clear evidence of a productive SAR for the isoquinolinone alkoxy substituent has been demonstrated by analogs 14 and 36−39. The stereochemistry of the ether substituent in 1 and 2 does not appear to be critical as 38, with a –C(Me₂)CO₂Et ether at this position, is more potent than 2. In addition, an analog of 2 with a –CH₂CO₂Et aryl ether (39, not shown), was deemed to be “inactive” in the original HTS campaign.³ These data suggest that hydrophobicity at this position is important for SF-1 inhibitor activity.

All analogs were further assessed in another transactivation assay (Gal4-VP16)^xvii to confirm that any SF-1 activity measured was specific to that receptor and not due to transactivation assay artifacts (e.g., promiscuous inhibition of the reporter system); cellular cytotoxicity was also determined (CellTiterGlo assay;^xviiisee Table 1).³ The titration curves for compound 32 are shown in Figure 1. This compound showed significant activity only in the SF-1 assay; no activity in the RORα, Gal4-VP16 and cytotoxicity assays were detected at concentrations up to 99 μM.

01

02