Identification of Mineralocorticoid Receptor Modulators with Low

Subscriber access provided by Iowa State University | Library

Article

Identification of Mineralocorticoid Receptor Modulators with Low Impact on Electrolyte Homeostasis but Maintained Organ Protection Kenneth Granberg, Zhong-Qing Yuan, Bo Lindmark, Karl Edman, johan kajanus, Anders Hogner, Marcus Malmgren, Gavin O'Mahony, Anneli Nordqvist, Jan Lindberg, Stefan Tångefjord, Michael Kossenjans, Christian Löfberg, Jonas Brånalt, Dongmei Liu, Nidhal Selmi, Grigorios Nikitidis, Peter Nordberg, Ahlke Hayen, Anna Aagaard, Eva Hansson, Majlis Hermansson, Ida Ivarsson, Rasmus Jansson Löfmark, Ulla Karlsson, Ulrika Johansson, Lena William-Olsson, Judith Hartleib-Geschwindner, and Krister Bamberg J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01523 • Publication Date (Web): 31 Dec 2018 Downloaded from http://pubs.acs.org on January 3, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Medicinal Chemistry

Identification of Mineralocorticoid Receptor Modulators with Low Impact on Electrolyte Homeostasis but Maintained Organ Protection Kenneth L. Granberg,*† Zhong-Qing Yuan,† Bo Lindmark,† Karl Edman,‡ Johan Kajanus,† Anders Hogner,† Marcus Malmgren,‖ Gavin O’Mahony,† Anneli Nordqvist,† Jan Lindberg,† Stefan Tångefjord,‡ Michael Kossenjans,‡ Christian Löfberg,† Jonas Brånalt,† Dongmei Liu,┴ Nidhal Selmi,‡ Grigorios Nikitidis,‖ Peter Nordberg,† Ahlke Hayen,† Anna Aagaard,‡ Eva Hansson,‡ Majlis Hermansson,† Ida Ivarsson,‡ Rasmus Jansson-Löfmark,† Ulla Karlsson,‡ Ulrika Johansson,† Lena William-Olsson,† Judith Hartleib-Geschwindner,† and Krister Bamberg† †Cardiovascular,

Renal and Metabolism, ‡Discovery Sciences, ‖Pharmaceutical Sciences,

IMED Biotech Unit, AstraZeneca, SE-431 83 Gothenburg, Sweden ┴Pharmaron

Beijing Co., Ltd., No. 6 Taihe Road, BDA, Beijing 100176, P. R. China

KEYWORDS: MR, modulator, nuclear hormone receptor, hyperkalemia, X-ray, structure– based drug design, early dose to man prediction, eD2M, AZD9977. 1

ACS Paragon Plus Environment

Journal of Medicinal Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT: The mechanism-based risk for hyperkalemia has limited the use of mineralocorticoid receptor antagonists (MRAs) like eplerenone in cardio-renal diseases. Here we describe the structure and property driven lead generation and optimization, which resulted in identification of MR modulators (S)-1 and (S)-33. Both compounds were partial MRAs but still demonstrated equally efficacious organ protection as eplerenone after four weeks treatment in uninephrectomized rats on high salt diet and aldosterone infusion. Importantly, and in sharp contrast to eplerenone, this was achieved without substantial changes to the urine Na+/K+ ratio after acute treatment in rat, which predicts a reduced risk for hyperkalemia. This work led to selection of (S)-1 (AZD9977) as the clinical candidate for treating MR-mediated cardiorenal diseases including CKD and HF. Based on our findings we propose an empirical model for prediction of compounds with low risk of affecting the urinary Na+/K+ ratio in vivo.

INTRODUCTION The marketed steroidal mineralocorticoid receptor antagonists (MRAs) spironolactone and eplerenone (Figure 1) are used for effective treatment of heart failure with reduced ejection fraction (HFrEF) and resistant hypertension.1 Increasing evidence also suggests that MRAs are efficacious agents for treatment of chronic kidney disease (CKD) and heart failure with preserved ejection fraction (HFpEF).

2


Page 2 of 72

Page 3 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


N O

O

OH

O O

H

H H

O

O

HO H

H O

O

O H S

H H

O O

Aldosterone Endogenous MR agonist

O

N H

Eplerenone Cl

O O S F

Finerenone BAY 94-8862

N

N N

O HN

N

N H2

N

O

Spironolactone

O O

O

F

O

O

N

O O NH O S

Apararenone MT-3995

F3 C

O

O

N N

N H

O O

N

OH

Esaxerenone CS3150 (XL550)

NH OH

KBP-5074

O

AZD9977 (S)-1

Figure 1. Aldosterone, the marketed MR antagonists (spironolactone and eplerenone) and five clinical candidates in development. Clinical trials have shown that addition of MRAs to the current CKD standard of care (angiotensin converting enzyme inhibitor (ACEi) and angiotensin receptor blocker (ARB) treatment) further reduces proteinurea.2,

3

There is also data suggesting that long-term MRA

treatment may stabilize declining glomerular filtration rates (GFR).4 Evidence for the efficacy of MRAs in HFpEF patients is primarily derived from the TOPCAT outcome trial with spironolactone.5, 6 The beneficial effects of MRAs on top of ACEi and ARBs, which are used as standard of care for treating both heart failure and CKD, may in part be attributable to the fact that prolonged treatment with ACEi/ARB results in elevated aldosterone concentrations in up to 50% of the patient population, so called aldosterone breakthrough. This suggests that improved 3



treatment options specifically targeting the aldosterone component of the renin-angiotensinaldosterone system (RAAS) should be explored.7-10 A serious adverse effect of both marketed MRAs is the development of hyperkalemia, which may lead to cardiac arrest and death.11 Fear of hyperkalemia significantly restricts the use of MRAs with the majority of patients on MRA treatment receiving suboptimal MRA doses and >30% of eligible patients do not receive MRA treatment.12-14. Aldosterone is the main activating hormone (Figure 1) of the mineralocorticoid receptor (MR, NR3C2) in epithelial tissues and is a key regulator of electrolyte homeostasis and blood pressure.15 The classical pharmacological response of MR activation is upregulation and activation of the epithelial sodium channel (ENaC) and the Na+-K+-ATPase in the collecting duct.16 This causes an increase of Na+-reabsorption from urine and as a consequence water retention and K+-excretion. More recently, a more multifaceted role for MR has emerged involving both genomic and nongenomic effects.17 This makes MR a target with the potential to modulate physiological processes in several tissues, as MR is expressed in several organs including kidney, brain, colon and heart.18 Although aldosterone is the main activating hormone for MR, the receptor may also be activated in non-epithelial tissues by cortisol, particularly in situations of increased oxidative stress.18, 19 Pathological activation of MR in target organs promotes oxidative stress, inflammation, fibrosis and thereby target organ dysfunction. Therefore, MR antagonism constitutes an important and possibly superior alternative compared to inhibition of aldosterone synthesis using an aldosterone synthase inhibitor as well as being an attractive treatment paradigm downstream of ACEi and ARB intervention of the RAAS system. Beside MR, the oxosteroid nuclear receptor family also includes the glucocorticoid receptor (GR, NR3C1), the progesterone receptor (PR, NR3C3) and the androgen receptor (AR, NR3C4).20 4


Page 4 of 72

Page 5 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The oxosteroid receptors have a common architecture comprising an N-terminal activation domain, a central DNA binding domain and a C-terminal ligand binding domain (LBD) and share many of the key amino acid residues in the first shell of the ligand binding pockets making compound design for selectivity a challenge.21 Clinical use of spironolactone is not only restricted due to the risk of hyperkalemia, but also because of its sexual side effects such as menstrual disturbances in women and painful gynecomastia in men, which are due to poor selectivity over PR and AR.22 Eplerenone was developed to have improved selectivity over the other oxosteroid receptors, but is a less potent MRA and is more expensive to manufacture compared to spironolactone.23 Several novel non-steroidal, third generation MRAs (Figure 1) are being progressed in clinical trials with hopes to deliver an MRA with an improved safety profile for treatment of patients suffering from hypertension, HF, CKD and nonalcoholic steatohepatitis (NASH).24-32 Amongst these is AZD9977 ((S)-1) and we recently disclosed its preclinical pharmacology.33 Based on the high unmet medical need in both HFpEF and CKD populations and encouraged by the promising effects observed with steroidal MRAs, we set out to search for a clinical drug candidate with retained or improved efficacy over eplerenone, but without the risk of hyperkalemia. Herein we report on the lead generation and optimization towards “MR modulators”, i.e. compounds with MR antagonist efficacy providing organ protection at least equal to classical MRAs, but with minimal or no effect on electrolyte homeostasis. We describe how compound selection for synthesis was guided by a strong element of structure–based drug design supported by X-ray crystallography of novel compounds in complex with the MR LBD and how we assessed compound quality and steered compound progression using early dose to man (eD2M) predictions in parallel with ligand lipophilic efficiency (LLE). We also report on our work to 5



establish a correlation between effect of MRAs on aldosterone mediated nuclear translocation of MR in vitro and their effects on urinary Na+/K+ ratio in an acute in vivo rat model. A subset of partial MRAs were shown to retain their kidney protective efficacy after chronic treatment in a proteinuric rat model and we labelled these ‘MR modulators’. These MR modulators have the potential to deliver organ protection like MRAs with reduced risk of hyperkalemia. RESULTS AND DISCUSSION Structure–based hit identification and lead generation The interactions formed between a ligand and a nuclear hormone receptor form the basis of ligand affinity, selectivity and allosteric modulation of auxiliary surfaces. The structural state of the receptor surface governs the communication between the receptor, DNA and co-regulator proteins and ultimately determines the functional response. A number of X-ray structures of MR LBD in complex with steroid ligands provided an initial overview of the key amino acids and functional areas within the ligand binding pocket.34-37 We integrated this information through docking of known non-steroidal MR antagonists and a selection of near neighbor compounds from the AstraZeneca corporate collection to build our understanding of the structure–activity relationship (SAR) and to drive the design of novel MR modulators (Figure 2). Predicted binding modes for the published MRAs 238, 39 and 340, 41 were generated by docking into the rigid X-ray structure of MR-LBD in complex with spironolactone (PDB ID 2oax).35 However, docking of 442 was not successful due to clashes with the receptor, highlighting the intrinsic conformational flexibility within the ligand binding pocket, so an induced fit protocol was used to generate a model of the MR-4 complex.

6


Page 6 of 72

Page 7 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2. Structural insights to build and execute the lead generation strategy. Data including statistics are reported in Supporting information (Table S6). AZLogD7.4 was predicted by a support-vector machine model based on AstraZeneca experimental logD7.4 data. Data for compounds 2 and 4 were extracted from the respective publications.39, 43 (A) X-ray structure of spironolactone in complex with MR (2oax).35 The steroidal A, B, C and D rings of spironolactone and position C7 are marked together with helices three, five, seven and eleven (solid blue circles). The three key areas identified as important for design are circled. (B) Predicted binding modes of compounds 2,38, 39 3,40, 41 and 4, 42, 43 in comparison to X-ray structure of MR:spironolactone (PDB ID 2oax). (C) The chemical structures, in vitro data and predicted binding mode of initial hit rac5 in comparison with a 3D model of 4 and the follow up compound, rac-6, in comparison with predicted binding mode of compound 2.

7



The predicted binding mode for compound 2 was later confirmed by X-ray (PDB ID 3vhv39, ligand heavy atom RMSD = 0.48 Å) and similarly the predicted binding mode for compound 4 was supported by X-ray structure (PDB ID 5hcv43) of a close analog of 4 wherein the 1,3dihydrobenzimidazol-2-one was replaced by a 4H-1,4-benzoxazin-3-one as in 2, validating the accuracy of the modeling protocol. Visual inspection of the protein-ligand models (Figure 2) confirmed that the compounds exploit several key functional interactions of the steroid ligands. In all instances, the non-steroidal ligands contained functional groups that emulated the steroid Aring binding near the gatekeeper residues Gln776 and Arg817 in helices three and five respectively (Figure 2A, region 1). The interaction between these two helices is important for receptor activation across the oxosteroid receptor family, which suggest it may be targeted by compound design to modulate the degree of antagonism-agonism.44 In addition, this region is one of the few areas within the ligand binding pocket that readily accepts polar substituents, providing opportunity to control the physicochemical properties of the compounds.37 In the region between helices five and seven (Figure 2A, region 2), the C7 thioacetyl group of spironolactone rearranges the receptor conformation by modulating the position of Met852 with the C7 thioacetyl group.35 The structures of 2–4 all extended into the induced pocket with aromatic motifs overlapping the C7 substituent of spironolactone (Figure 2A, region 2). In a third region at the other end of the ligand binding pocket (Figure 2A, region 3), compounds 2–4 overlapped the D-ring of spironolactone and formed hydrogen bonds to the side chain of Asn770, which is fully conserved among the oxosteroid receptors and the extended hydrogen bonding network is important both for ligand recognition and for the positioning of helix 12 (not shown) in the active receptor conformation.36 Recently, Piotrowski et al. also reported a similar description of the three regions in the ligand binding pocket of MR.45 With the three key receptor regions identified, we generated 8


Page 8 of 72

Page 9 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


a pharmacophore model and searched the AstraZeneca compound collection for compounds with matching complementary structural features. Selection was done by substructure searches using motifs from known binders targeting the three key regions mentioned above. Subsequent iterative screening of the selected compounds provided a series of sulfonamide benzoxazinones as exemplified by rac-5 with pIC50 5.2 and an efficacy extrapolated to 100% in a GAL4 luciferase human MR reporter gene antagonist assay (hMRGAL4).46 Docking of (R)-5 into the induced fit model of MR in complex with 4 indicated that (R)-5 extended into all three regions of our pharmacophore model (Figure 2C). However, rac-5 exhibited poor kinetic solubility, pSol >6.0, and low metabolic stability in human liver microsomes (HLM) with CLint values of 120 µL/min/mg. Comparison of the docked pose of (R)-5 with tricycle 4 indicated that the ligandbinding pocket was suboptimally filled (Figure 2C). Fusing a phenyl ring to the piperidine of rac-5 to form a bicyclic tetrahydroquinoline moiety led to rac-6, which according to our modeling (of (S)-6) was predicted to fill the induced-fit pocket in the vicinity of Met852 (Figure 2A, region 2). Rac-6 was equipotent to rac-5 with an hMRGAL4 pIC50 5.3 (extrapolated efficacy 100%). Competitive binding of rac-6 with 3H-aldosterone in the LBD of human MR (hMRbind) was confirmed with a pKi 5.7. However, while rac-6 exhibited improved solubility compared to rac-5, its metabolic stability decreased as evidenced by HLM CLint increasing from 120 to >300 µL/min/mg (Figure 2C). A close inspection of the docked rac-5 and rac-6 models revealed that the sulfone linker between the ring systems lacked direct interactions with the receptor. Therefore, to improve physicochemical properties and improve binding affinity to MR, we replaced the sulfonyl group by a carbonyl resulting in resulting in rac-7 (R3a = CH3, Table 1), which maintained hMRbind pKi and hMRGAL4 potency compared to 8, despite being racemic, suggesting substitution (R3a ≠ H) was 9



still tolerated. Comparing the matched pair rac-6 versus rac-7, logD7.4 was reduced from 3.3 to 2.7 and consequently a significant improvement in LLE was calculated, here defined as hMRbind pKi - logD7.4, from 2.4 to 3.4. The reduction in lipophilicity also resulted in a reduction in HLM CLint from >300 to 128 µL/min/mg. Docking of (S)-7 isomer in both X-ray (PDB ID 2oax) and an induced-fit model placed the tetrahydroquinoline in the induced pocket between helices five and seven, but couldn’t distinguish between alternative poses of the tetrahydroquinoline moiety whereby the phenyl ring of the tetrahydroquinoline could be placed in regions 1 or 2 of the binding pocket (Figure 2A), vide infra. Given the plasticity observed for the flanking residues Met845 and Cys849, we decided to further explore hydrophobic interactions in this region.35 On preparation of rac-9 (R7 = F) the enantiomers were separated and tested to answer the question of stereochemical preference in binding to MR. Thus, (S)-9 showed an improvement in both binding and functional potency relative to rac-7 and was in these respects comparable with eplerenone (Table 1). Interestingly, enantiomer (S)-9 showed a 50-fold improved hMRbind Ki over (R)-9. Examining lipophilicity, logD7.4 was essentially unchanged from rac-7 to (S)-9 (Table 1) and as a result the LLE increased to 4.0 for (S)-9, which was a slight improvement considering rac-7 was compared with the isolated (S)-enantiomer of 9. However, metabolic stability of (S)-9 remained poor, HLM CLint 190 µL/min/mg, which highlighted the need to further improve the physicochemical properties and the metabolite pattern after incubation in HLM suggested the saturated system of the tetrahydroquinoline was subject to extensive oxidation. A crucial modification was therefore switch from the tetrahydroquinoline in 7–9 to the more polar 3,4-dihydro-2H-1,4-benzoxazine scaffold (Table 1, X = CH2  O), which would both remove the potentially labile benzylic methylene and reduce the lipophilicity. While hMRbind pKi was slightly reduced from 8 to 10 (6.2 10


Page 10 of 72

Page 11 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


to 6.0) and from (S)-9 to (S)-11 (6.9 to 6.8) the ∆logD7.4 was -0.4 and -0.3 respectively and the net increase in LLE was 0.2 units. Most importantly, the metabolic stability improved and HLM CLint was reduced from 27 to 4 µL/min/mg for 8 to 10 and from 190 to 8 µL/min/mg for (S)-9 to (S)-11 suggesting overall balanced properties for in vivo studies were achievable. As previously observed, the enantiomer (R)-11 exhibited much weaker MR binding affinity with a >63-fold drop in hMRbind Ki. We decreased the lipophilicity by introducing a nitrogen in the 5-position, by switching from 3,4-dihydro-2H-1,4-benzoxazine ((S)-11, Y = CH) to the more polar 3,4-dihydro-2H-pyrido[3,2b][1,4]oxazine ((S)-12, Y = N) , which resulted in an 8-fold lowered hMRbind Ki for (S)-12 relative to (S)-11, but this was compensated by retained good metabolic stability in HLM, an improved LLE by 0.4 units to 4.6 and a large improvement in solubility (pSol from 4.6 to 3.0). Taken together, this made (S)-12 a compound with overall well-balanced properties. However, due to synthetic challenges in making the functionalized 3,4-dihydro-2H-pyrido[3,2-b][1,4]oxazine scaffold and other electron-deficient heterocycles together with the reduced chemical stability of the exocyclic amide bond towards hydrolysis, this led us to not further pursue such replacements. Given the predicted suboptimal filling of the ligand binding pocket by compounds such as (S)9 and (S)-11 (R3a = CH3), larger R3a groups were investigated. An isobutyl group in compound (S)13 afforded an improvement in hMRbind for (S)-13 (hMRbind pKi 8.7 and an improvement of LLE to 4.9) over (S)-11 in line with an increase in logD7.4 to 3.8 from 2.6 although a drop-off in hMRGAL4, pIC50 7.2 (121%) was observed for (S)-13. However, the high lipophilicity of (S)-13 also led to reduced kinetic solubility (pSol 4.7) and a very poor HLM CLint >300 µL/min/mg. Learnings from the X-ray structure of (S)-13 in complex with human MR LBD

11



To better understand the binding mode of the series (Figure 2C) and to improve our understanding of the SAR, we determined the X-ray structure of MR LBD in complex with (S)-13 (PDB ID 6GEV, Figure 3).

Figure 3. X-ray structure (PDB ID 6GEV) of MR LBD (blue) in complex with (S)-13 (magenta sticks). Key residues are highlighted as orange sticks and putative hydrogen bonds are marked as dotted lines. Ser811 and Met852 were observed in multiple conformations as shown.

The X-ray structure showed (S)-13 extending into the same areas of the ligand binding domain as spironolactone confirming one of our two predicted binding modes for (S)-13 and previously (S)-7 (Supporting Information, Figure S14). At the center of helix three, the benzoxazinone scaffold of (S)-13 made a bidentate hydrogen-bonding with Asn770 and an additional hydrogen bond to Thr945. In contrast, spironolactone (PDB ID 2oax) was only positioned to make a weak hydrogen bond to the Nδ of Asn770. In addition, the benzoxazinone ether oxygen was within 12


Page 12 of 72

Page 13 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


hydrogen bonding distance of the Cys942 backbone nitrogen. This was a key interaction for MR binding affinity in this series demonstrated by 3,4-dihydro-1H-quinolin-2-one (S)-14 which exhibited a 460-fold loss of binding affinity relative to (S)-13. As (S)-14 will lack the putative interaction to the backbone nitrogen, Cys942 will engage in an intrahelical hydrogen bond to the carboxyl of Leu938 and therefore have a suboptimal angle for interaction with the (S)-14 ligand. Therefore, the relatively low affinity of (S)-14 was likely a consequence of a modulation of the ligand conformation and reduced complementary fit to the ligand binding pocket. Inspecting the electron density suggested that Ser811 adopted multiple conformations and that the benzoxazine oxygen of (S)-13 made a weak hydrogen bond interaction (distance 3.1 Å) to one of the two modeled rotamers. Looking at the structural state of the induced pocket in between helices five and seven, Met852 was also modelled in multiple conformations highlighting the flexibility of this region. The benzoxazine fluorine atom points directly at the Cα of Cys849 (not shown) but the distance was rather long (3.8 Å) and the gain in hMRbind when going from rac-7 to (S)-9 and from 10 to (S)-11 (Table 1) were more likely due to increased hydrophobic interactions and/or modulation of the electrostatics of the aromatic system. Finally, near the gatekeeper residues Gln776 and Arg817, the isobutyl extended into the same region as the steroid A-ring. It was clear that the close match between the ligand within the ligand binding pocket was facilitated by the (S)configuration. The (R)-enantiomer would likely come into a direct conflict with Trp806, Met807 and Ser810 above the steroid A and B rings in the aldosterone complex (not shown),36 which would explain the drop in hMRbind when comparing the enantiomeric pairs (S)-9 versus (R)-9 and (S)-11 versus (R)-11 (Table 1). This information gave us confidence to only prepare a single enantiomer where possible and develop stereoselective synthetic methods.

13



While the isobutyl substituent of (S)-13 had an extensive hydrophobic interaction surface to both helix three and helix five residues, it did not exploit the polar gatekeeper residues Gln776 and Arg817. Therefore, to improve physicochemical and DMPK properties further, we next focused our efforts on introducing polarity in the R3 side chain. Introducing polarity in the R3 sidechain The primary alcohol (S)-15 had comparable hMRbind pKi (6.7) and functional activity (hMRGAL4 pIC50 6.3 (110%)) relative to (S)-11 (hMRGAL4 pIC50 6.0 (118%)), the LLE improved from 4.2 to 4.9 for (S)-15. We hypothesized that the penalty in desolvation of the alcohol was compensated by establishment of a hydrogen bond with MR. The reduced lipophilicity of (S)-15 contributed to an improved kinetic solubility relative to (S)-11 (∆pSol -1.6) while the metabolic instability was slightly increased from 8 to 15 µL/mL/mg for (S)-15. In relation to (S)-11 the poor hMRbind selectivity over hGR remained unchanged although selectivity over hPR was retained. Next, we explored the extremes for interactions with the polar gate keeper region around Arg817 and/or Gln776 by introducing a basic and an acidic substituent in compounds (S*)-16 and (S)-17b, respectively. We exchanged F for a Cl as R7-substituent in (S)-17b to balance the expected low lipophilicity of the carboxylic acid derivative. Both compounds had poor functional activity, but the binding affinity to MR of the acid (hMRbind pKi 6.8) was remarkable in comparison to that of the amine (pKi 300 µL/min/mg. 14


Page 14 of 72

Page 15 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


In hPR, the nitrile group has been used as an efficient hydrogen bond acceptor to the gatekeeper residues (Gln725 and Arg766 in hPR).47, 48 A comparison of the residues lining the ligand binding pocket in MR relative GR, PR and AR showed that the polar gatekeeper residues are identical (Gln776 and Arg817, Figure 4) and a nitrile might therefore be able to pick up favorable interactions also in MR, if correctly positioned.

Figure 4. Pictorial view of (S)-13 and the closest shell of amino acids in the ligand binding domain of MR based on the X-ray structure of MR:(S)-13 (Figure 3). Residues that differ are highlighted for GR and PR (blue circles).

However, nitrile (S)-19c had a hMRbind pKi 6.3, which was lower than (S)-15 (pKi 6.7) and could not be easily rationalized by the decrease in lipophilicity, suggesting the nitrile was not correctly positioned. To optimize (S)-15 we explored ring contraction, but while modeling supported the 15



replacement of dihydrobenzoxazinone to a dihydroindole (20), both enantiomers lost binding affinity for human MR (only (R*)-20 reported in Table 1). This demonstrated the importance of a six-membered ring system and correct positioning of the axial R3a side chain (Figure 3). The said axial preference may be explained by allylic strain H-C-N-C=O, which exerts powerful control of the conformation in analogy with N-acylated piperidines.49, 50 The use of early estimates of dose to man (eD2M) predictions to assess compound quality We used LLE to drive compound design during lead generation and although it was a useful parameter it did not discriminate the quality difference between (S)-13 and (S)-15, which have similar LLEs, but differ markedly in their metabolic stability. We therefore complemented LLE by early estimates of daily dose to man (eD2M, Equation 1) going forward to assess the overall compound quality before designing the next generation of compounds and selection of promising compounds for evaluation in in vivo models (PK, PD and safety).51-53 This strategy was motivated by a desire both to keep the dose below 100 mg/day,54 which could have an impact on the maximal achievable clinical efficiency versus off-target related safety margins and a desire to keep cost of goods as low as possible. Equation 1. Calculation of human dose estimates. Equations were used to estimate dose from in vitro data. eD2M =

IC50 ⋅ X ⋅ CL ⋅ Ƭ F

wherein CL =

Qh ⋅ CLint Qh + CLint

eD2M = early dose to man. IC50 = potency in hMRGAL4 reporter gene antagonist assay. X = offset factor between in vivo and in vitro IC50 based on clinical data for eplerenone and spironolactone. CL = predicted human hepatic clearance. Ƭ = dosing interval. F = bioavailability. CLint = intrinsic clearance in human hepatocytes. Qh = hepatic blood flow. 16


Page 16 of 72

Page 17 of 72 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


To assess eD2M, we compiled published clinical data on exposures and urine albumin to creatinine ratios (UACR) for both eplerenone and spironolactone.55-57 Exposures of the active metabolites of spironolactone, canrenone and 7-alpha-thiomethylspironolactone, were assessed in addition to spironolactone exposure.58 The average unbound efficacious concentration (Cuss) was found to be approximately two-fold lower than the hMRGAL4 IC50 for both eplerenone and spironolactone, which was assumed to translate to other compounds and therefore applied to the eD2M predictions for all compounds (Cuss = hMRGAL4 IC50/2). Human clearance (CL, Equation 1) was predicted by the established well-stirred model59 based on human hepatocyte (hHep) CLint values rather than HLM CLint. Hepatocytes represent a more complete in vitro system regarding drug metabolizing enzymes compared to HLM, expressing non-P450 phase I metabolizing enzymes like aldehyde oxidase as well as phase II metabolizing enzymes like UDP-glucuronosyl transferase. The assumption was that the compounds were eliminated mainly via hepatic metabolism. For more advanced compounds, preclinical PK studies were performed and in vitro to in vivo correlation of observed rat clearance versus clearance predicted from rat hepatocytes (rHep) was confirmed (Chart 1, Supporting Information). Based on Equation 1, the eD2M for (S)13 and (S)-15 were >4700 mg/day and 840 mg/day respectively. It should be noted that the eD2M equation was based on average concentration at steady state and did not consider the half-life of the compound. Exposure and efficacy data in a rat PD model vide infra,33 demonstrated that unbound in vivo IC50 values correlated well with in vitro rMRGAL4 IC50 values, which supports that efficacious exposures and human dose can be assessed based on in vitro data for new compounds. The in vivo UACR IC50 was estimated using a PKPD model similar to Orena et al.60 Lead optimization 17



A breakthrough in compound design was achieved by introduction of the N-methyl amide in the R3 side chain yielding (S)-21. In comparison with the methyl ester (S)-18a_Me logD7.4 was reduced from 2.2 to 1.1. Since hMRbind affinity of (S)-21 (pKi 7.3) was comparable to that of (S)18a_Me (pKi 7.2), LLE improved from 5.0 to 6.2. On the other hand, hMRGAL4 pIC50 was reduced from 6.5 (117%) for (S)-18a_Me to 6.2 (106%) for (S)-21. (S)-21 had a 63-fold selectivity in hMRbind Ki over hGR, which was double that of (S)-17b and eplerenone. The metabolic stability in hHep was excellent (CLint

Identification of Mineralocorticoid Receptor Modulators with Low

Recommend Documents