Potent Anti-Inflammatory, Arylpyrazole-Based Glucocorticoid Receptor Agonists That Do Not Impair Insulin Secretion

Abstract

Glucocorticoids (GCs) are widely used in medicine for their role in the treatment of autoimmune-mediated conditions, certain cancers, and organ transplantation. The transcriptional activities GCs elicit include transrepression, postulated to be responsible for the anti-inflammatory activity, and transactivation, proposed to underlie the undesirable side effects associated with long-term use. A GC analogue that could elicit only transrepression and beneficial transactivation properties would be of great medicinal value and is highly sought after. In this study, a series of 1-(4-substituted phenyl)pyrazole-based GC analogues were synthesized, biologically screened, and evaluated for SARs leading to the desired activity. Activity observed in compounds bearing an electron deficient arylpyrazole moiety showed promise toward a dissociated steroid, displaying transrepression while having limited transactivation activity. In addition, compounds 11aa and 11ab were found to have anti-inflammatory efficacy comparable to that of dexamethasone at 10 nM, with minimal transactivation activity and no reduction of insulin secretion in cultured rat 832/13 beta cells.

Glucocorticoids (GCs) are steroid hormones synthesized endogenously in the adrenal glands, which operate by acting as ligands for the glucocorticoid receptor (GR). The GR is an intracellular receptor located in the cellular cytoplasm and bound to chaperone proteins in the nonligand bound state. The GR is expressed by nearly all cell types, including pancreatic β-cells, and is the product of a single gene, NR3C1, that is post-translationally modified to produce functionally distinct subtypes of this transcriptional regulatory protein.¹

GCs are a staple of modern medicine due to their anti-inflammatory and immunomodulatory effects. GC-based therapies have been successfully applied to treating a wide range of inflammatory conditions, autoimmune disorders, and cancers—in addition to aiding in organ and tissue transplantation.²⁻⁴ Despite the obvious beneficial properties of GCs, they pose a long list of serious side effects including Cushing’s disease, muscle atrophy, metabolic syndrome, osteoporosis, and diabetes.⁵⁻⁷ In recent years, reducing autoimmune-mediated inflammation in patients at risk for developing steroid induced diabetes mellitus (SIDM) has been of interest; however, current GC therapies are known to be diabetogenic by transcription-mediated mechanisms that modify the secretion of insulin from pancreatic β-cells, regulate β-cell proliferation, alter body composition, and promote insulin resistance.⁸⁻¹¹ The reduction in insulin sensitivity places extra stress on β-cells to produce increasing quantities of insulin, which increases the risk of β-cell failure.^12,13 The need for improved GC agonist compounds is due to side effects of current compounds upon chronic use, such as immunosuppressive agents for organ transplants. This long-term usage dramatically increases the risk for diabetes.^14,15

Diabetes is the seventh leading cause of death in the United States, affecting an estimated 34.2 million people or 10.5% of the U.S. population in 2018.^16,17 This disease also serves as a major contributor to heart disease and stroke and is the leading cause of kidney failure, nontraumatic lower limb amputations, and associated with cases of blindness in adults. The incidence of steroid induced diabetes mellitus (SIDM) among patients receiving traditional postoperative treatments for solid organ transplantation is relatively high in the case of lung (60%),¹⁸ heart (29%),¹⁹ liver (24%),²⁰ and kidney (17%)²¹ transplants. It should be noted that SIDM would likely be of great concern in pancreatic organ or islet transplantation, as these procedures manage postoperative care similarly.^22,23 These situations, as well as any condition requiring prolonged GC use, could greatly benefit from a synthetic GC analogue with a “dissociated” profile (i.e., transrepression and desired transactivation activity with little or no unwanted effects). While it is difficult to summarize the cascade of intricate biological mechanisms initiated by GCs into a simplified model, the assumptions that transrepression elicits the desired anti-inflammatory effects while transactivation is the primary mode responsible for the undesired side effects still provides a framework to quickly assess the efficacy of GC analogues.²⁴⁻²⁷

Upon ligand-initiated GR activation, the GR–GC complex is translocated to the cellular nucleus where it can facilitate changes in gene transcription. Dimerization of the ligand activated GR binds to glucocorticoid-responsive elements (GREs) to regulate the expression of target genes. GR action promotes activation of genes which encode proteins that regulate inflammatory signals but can also directly repress the transcription of genes induced by pro-inflammatory cytokines, such as the cytokine IL-1β.^28,29 The latter occurs through a monomeric GR-GC complex with transcription factors such as NF-κB or AP-1.

Current synthetic GCs suppress inflammation induced by IL-1β but also reduce insulin sensitivity and modulate insulin secretion in β-cells.³⁰⁻³⁴ With the recently renewed interest in development of glucocorticoid receptor agonists,³⁵ this study provides novel insights into nonsteroidal arylpyrazole-based glucocorticoid receptor agonists (APGRAs) by investigating known³⁶ and new molecules with more in-depth biochemical assays. These compounds were subjected to in vitro assays evaluating both transactivation and transrepression, and candidates displaying the most potent and/or selective anti-inflammatory properties were further evaluated to determine how insulin secretion was affected in cultured β-cells. Additionally, computational docking experiments were performed to explore whether the molecules bound in the normal steroidal site or the deacylcortivazol (DAC) expanded GC binding pocket, to assess which region of the expanded pocket the molecules accessed, and to determine structure–activity relationships (SARs) between the APGRA analogues and the GR.

The total synthesis of all APGRA analogues tested in this study were prepared by a previously described synthesis published by Merck scientists,³⁶ with modifications made to the published method (Scheme 1). A change of reagent is employed in the conversion of secondary alcohol 11 to ketone 12, using a standard PCC oxidation in place of the costly oxidation with TPAP using NMO as a sacrificial oxidant. The molecules investigated by Ali and co-workers, focused on alterations to the R² moiety while leaving the 4-fluorophenyl R¹ group as a constant.³⁶ Modifying the identity of R¹ was achieved through the addition of the appropriate phenylhydrazine reagent to β-ketoaldehyde 6, thereby installing the desired R¹-arylpyrazole moiety. Additionally, to further enhance the scope of functionalities examined in our biological evaluation, a previously uninvestigated cyanomethylene unit was included as an R² analogue in the series.

Scheme 1. Reagents and Conditions: (i) AcOH, H₂O, 75 °C, 1 h; (ii) l-proline, DMF, rt, 120 h; (iii) p-TsOH, 4 Å mol sieves, ethylene glycol, rt, 30 min; (iv) ethyl formate, NaH, toluene, 0 °C, 3 h; (v) NaOAc, AcOH, rt, 3 h; (vi) 6 N HCl, THF, reflux, 3.5 h; (vii) Ph₃PCH₂OCH₃Cl, KHMDS, THF, rt, 24 h; (viii) 4 N HCl, rt, 36 h; (ix) THF, −78 °C, 1 h; (x) PCC, 4 Å mol sieves, DCM, rt, 6 h; (xi) MeLi, THF, −78 °C, 1 h.

Briefly, the total synthesis of the steroid core begins with a Robinson annulation of compounds 1 and 2 to form intermediate trione 3, which upon asymmetric intramolecular cyclization gives Wieland-Miescher ketone 4. Selective ethylene glycol protection of compound 4 leaves the α,β-unsaturated ketone available for enolate formation and subsequent addition to ethyl formate, giving β-ketoaldehyde 6. Addition of the appropriate phenylhydrazine to 6 completes the core structure of the APGRAs as compounds 7a, 7b, and 7c. The acid-catalyzed deprotection of acetal 7 provided synthon 8, which was formylated via Wittig reaction to afford aldehyde 10, which provided an intermediate that could be diversified to provide the analogues presented in this study. Separation of starting material from the product of the one-pot Wittig/hydrolysis sequence from ketone 8 to aldehyde 10 was futile, and extra care was taken to obtain aldehyde 10 in high purity. This was done by isolating intermediate enol-ether 9, which after purification was subjected to hydrolysis conditions to afford compound 10 in sufficient purity to proceed with the next reaction.

Insertion of the R² group into the GC scaffold was accomplished through generation of the appropriate alkyl lithium or Grignard reagent followed by the addition of aldehyde 10, resulting in a series of 2° alcohols (11) that were subjected to biological testing. Intermediate ketone 12 was obtained through a PCC oxidation of 11 and after purification was treated with MeLi to afford a series of 3° alcohols (13), which were also assayed. Ali and co-workers found this reaction sequence led to stereoinduction in all subsequent transformations yielding a new chiral center at the formation of 4.³⁶ On the basis of these observations, the conformation shown for 11 and 13 is predicted to the be the major product with a ∼24% diastereomeric excess (de) (Scheme 1).

It is well established that mouse, rat, and human beta cells secrete large amounts of insulin in response to glucose and reduce insulin secretion in response to GC exposure.^29,37−39 Since there is excellent agreement across species, a model rat isolated cell line, 832/13, was used in this study as it has been shown to recapitulate the key responses of human cell lines well.⁴⁰ A culture of 832/13 cells and measurements of insulin secretion using an insulin ELISA kit were used as previously described.⁴⁰ Six-point dose–response curves were used to generate EC₅₀ and saturating concentrations of each steroid compound. For GSIS, saturating concentrations of each commercial steroid or N-arylpyrazole based steroid were used as indicated in the figure legends.

The 832/13 cells were grown in 24-well plates to 50% confluence and then transiently transfected with 25 ng of indicated plasmid per well using TransFectin Lipid Reagent (Bio-Rad) according to the manufacturer’s instructions. Cell lysis, luciferase assays, and normalization to total protein content were carried out as described previously.³¹ Briefly, transrepression is measured using a previously reported CCL2-promoter luciferase plasmid construct that is activated with a known pro-inflammatory signal, IL-1β, followed by GR agonism.^23,41 Ligand-activated transactivation is measured using a previously reported 3XGRE-promoter luciferase plasmid construct in which increases in luciferase activity are measured in response to GR agonists.^30,42

Ligand–receptor interactions were modeled between the molecules subjected to in vitro testing and the extended pocket of the GR ligand-binding domain (LBD; Figure 3). The crystal structure of the GR LBD bound to DAC, a potent APGRA, and the fourth LXXLL motif of steroid receptor coactivator 1 (SRC1–4; PDB ID: 3BQD) was used as the model for molecular docking calculations. The human GR in complex with budesonide (PDB ID: 5NFP) was used in molecular docking studies but was not chosen as the preferred model as the results corroborated those of the quantum chemical experiments.³⁵ The 3BQD crystal structure was chosen as the best template as consistent trends in binding were observed, which validated previous work that assumed the enlarged ligand binding pocket caused by the phenylpyrazole moiety would also house the smaller APGRA analogues.⁴³ The GR LBD was prepared for docking studies by removing the bound ligand and excess water molecules. Ligands were optimized in Chem3D (PerkinElmer) prior to docking, which was performed between a rigid 3BQD GR and flexible ligand using AutoDock Vina (The Scripps Research Institute).⁴⁴ The resulting docking conformations were viewed with PyMol (Schrödinger).⁴⁵ Ligand-residue interactions were further analyzed with LigPlot+ (European Molecular Biology Laboratory).⁴⁶ Relative binding affinities (AutoDock Vina) can be found in the Supporting Information (SI section 1).

Figure 3.

Molecular docking experiment between APGRA 11ab (structure shown in panel A) and GR LBD (3BQD). (B) 2D representation of ligand–receptor interactions where a red semicircle represents a hydrophobic interaction, and a cyan dotted line represents a hydrogen bond. (C) 3D representation of ligand–receptor interactions with cyan dotted lines displaying hydrogen bonding.

The nature of the GC–GR interactions was further examined by means of density functional theory (DFT). The receptor pocket was modeled by including all main group atoms within 6 Å of the cocrystallized ligand (budesonide) with solvent molecules removed. All valences were capped with appropriate atoms based on geometry of the included fragments. Constrained geometry optimizations were performed on docking output orientations by relaxing all hydrogen atoms and the full glucocorticoid molecule within the pocket with atoms of the pocket frozen during convergence. The BP86 density functional was employed along with Grimme’s D3 dispersion correction with the Becke-Johnson damping function (abbreviated as -D3(BJ)) and the def2-SVP basis set.⁴⁷⁻⁵¹ The resolution of identity (RI) approximation was used for rapid integral evaluation.⁵²

Single-point calculations were performed on the BP86-D3(BJ)/def2-SVP optimized geometries at the B3LYP- D3(BJ)/def2-TZVPP level for higher accuracy.⁵³⁻⁵⁵ All DFT computations were performed with the TURBOMOLE 7.2 quantum chemical software package⁵⁶ (SI section 1).

Previous work by Ali and co-workers reported the synthesis of APGRAs containing only F substitution on the aryl moiety in which compounds 11a(a–f), 13aa, 13ac, and 13ae were originally reported as compounds 26, 10, 11, 18, 23, 29, 28, 21, and 25, respectively.³⁶ Another limitation of this work was that the bioassays used did not directly measure the ability of the compounds to impact transactivation and transrepression at the transcriptional level and instead relied on measures that could have impacted either pre-, during, or post-transcription. Therefore, this work aimed to investigate this promising class of anti-inflammatory compounds by expanding the library of APGRAs and using bioassays that directly measured the impact of these molecules on transcriptional outcomes. Although the difficulty in translating in vitro activity to in vivo success has been documented recently, this further validates the need to thoroughly investigate GR mechanisms with comprehensive biological assays to find avenues that better predict in vivo activity.⁵⁷⁻⁵⁹ Directly studying the ability of the ligand activated GR to dimerize and bind to the GRE allows for a more encompassing view on transactivation potential compared to traditional reporter gene assays such as tyrosine amino transferase (TAT) or glutamine synthase (GS) assay.

The ability of the APGRAs to both reduce IL-1β induced inflammation (CCL2 assay, transrepression) and enhance transcription of a synthetic GRE-containing promoter luciferase-reporter gene (3xGRE assay, transactivation) was tested and benchmarked against both dexamethasone (Dex) and hydrocortisone (HC). The pIC₅₀ for transrepression by Dex was 8.2, and this compound also showed the highest activity in this assay. Transactivation occurred at nearly the same concentration (pEC₅₀ = 8.7) with Dex as well. Hydrocortisone showed 87.8% of the transrepression of Dex, and the pIC₅₀ for this compound was slightly lower as well (7.6). Interestingly, HC displayed a ∼100 fold concentration difference between transrepression and transactivation activities, with the latter having a pEC₅₀ of 5.6; although this compound’s transactivation potential was still 96.8% of Dex.

Two moieties on the APGRA scaffold were chosen to be manipulated for the purpose of identifying a ligand that displayed favorable ligand properties (Scheme 1). Three different sets of APGRA derivatives were synthesized that varied in the substituent at R¹ leading to a series of fluorophenyl-, tolyl-, and anisolyl-pyrazoles. Molecules in the fluorophenyl- series have been reported, and these molecules showed desirable properties.^36,60 The tolyl- and anisolyl-APGRA’s have not been previously reported. The fluorophenyl- analogues showed a much-improved transrepressive profile over the tolyl- and anisolyl- series, with the majority of the compounds having a pIC₅₀ > 8, and are therefore more likely to have useful anti-inflammatory activity. However, the tolyl- and anisolyl- analogues elicited lower transcription of the GRE-promoter luciferase gene fusion, indicating that they could be less likely to promote side effects (assuming that transactivation is the most likely cause of side effects).

Generally, compounds in series 13 bearing side chains with 3° alcohols displayed better anti-inflammatory activity than those in series 11 with 2° alcohols in the side chain, and the maximum activity for the 2° alcohols also often occurred at 1 to 2 orders of magnitude higher concentrations. Unfortunately, the more active compounds also tended to elicit more activation of the GRE as well (Table 1). Therefore, the slightly higher anti-inflammatory properties of the series 13 compounds must be weighed against benefits from the series 11 compounds that were more successful in differentiating transrepression from transactivation. If one assumes that the bulkiest substituent on the side chain will be preferentially positioned away from the bridgehead methyl between the two rings, then the changes in activity may be more impacted by an apparent change in the stereochemistry of the alcohol that would be displayed on the top face for the 2° and the bottom face of the 3° alcohols, respectively, than by the addition of the methyl substituent (see the structures in Scheme 1).

Table 1. Biological Analysis of Transrepression and Transactivation Profiles.

			CCL2a,b				3xGREa,c
compd	R1	R2	pIC50	Emax	% Dexd	% HCd	pEC50	Emax	% Dexd	% HCd
HC			7.6	75.0	87.8	100.0	5.6	10.5	96.8	100.0
Dex			8.2	85.5	100.0	113.8	8.7	10.9	100.0	103.3
11aa	F	Me	8.2	74.8	87.5	99.6	7.2	1.8	16.5	17.0
13aa	F	Me	7.4	67.6	79.1	90.1	7.2	9.3	85.2	88.0
11ab	F	Ph	7.4	67.3	78.7	89.6	7.0	2.8	26.1	27.0
13ab	F	Ph	8.1	80.8	94.6	107.7	8.4	6.6	60.5	62.5
11ac	F	4-F-Ph	8.0	63.6	74.5	84.8	7.1	4.9	44.8	46.2
13ac	F	4-F-Ph	8.3	73.4	85.9	97.8	7.8	9.0	82.5	85.2
11ad	F	2,3-F-4-Ani	7.8	58.1	68.0	77.4	7.2	13.3	122.8	126.8
13ad	F	2,3-F-4-Ani	8.7	74.7	87.4	99.5	8.2	10.6	97.3	100.5
11ae	F	3-Thioph	8.3	76.8	89.9	102.3	6.9	6.5	60.3	62.2
13ae	F	3-Thioph	8.2	84.7	99.1	112.8	8.4	8.4	77.7	80.3
11af	F	t-Bu	9.6	35.4	41.5	47.2	6.1	1.4	12.8	13.3
11ag	F	Cyano Me	5.8	51.0	59.7	68.0	5.6	1.2	11.3	11.6
11ba	CH3	Me	9.7	25.6	30.0	34.1	5.8	1.3	11.7	12.1
13ba	CH3	Me	7.2	53.5	62.6	71.3	9.4	0.8	7.7	8.0
11bb	CH3	Ph	6.4	39.4	46.1	52.4	5.0	1.9	17.9	18.4
11bb	CH3	Ph	7.5	42.6	49.9	56.8	5.9	2.2	19.8	20.5
11bc	CH3	4-F-Ph	6.6	33.1	38.7	44.1	4.5	1.0	9.5	9.8
13bc	CH3	4-F-Ph	6.2	42.0	49.2	56.0	6.5	0.9	8.1	8.3
11bd	CH3	2,3-F-4-Ani	6.8	16.9	19.7	22.5	4.0	1.5	13.9	14.4
13bd	CH3	2,3-F-4-Ani	8.1	45.7	53.5	61.0	5.7	2.0	18.2	18.8
11be	CH3	3-Thioph	7.4	36.7	42.9	48.9	5.0	1.1	10.2	10.5
13be	CH3	3-Thioph	7.0	56.3	65.9	75.0	5.8	1.1	10.4	10.7
11ca	OMe	Me	9.8	12.7	14.8	16.9	6.6	1.2	10.9	11.2
13ca	OMe	Me	5.6	53.7	62.8	71.5	7.1	1.0	9.6	9.9
11cb	OMe	Ph	5.9	31.3	36.6	41.7	5.0	1.2	10.8	12.0
13cb	OMe	Ph	6.3	62.4	73.0	83.1	5.3	1.2	10.7	11.0
11cc	OMe	4-F-Ph	7.1	16.7	19.5	22.3	4.8	1.3	11.6	14.8
13cc	OMe	4-F-Ph	6.6	48.4	56.7	64.5	6.9	0.9	8.5	8.8
11cd	OMe	2,3-F-4-Ani	8.5	39.1	45.7	54.0	8.1	1.6	14.3	15.3
13cd	OMe	2,3-F-4-Ani	5.9	31.1	36.4	41.4	5.0	2.3	21.5	22.2
11ce	OMe	3-Thioph	6.3	34.4	40.3	45.9	5.0	1.6	14.9	9.3
13ce	OMe	3-Thioph	5.8	56.4	66.0	75.2	5.0	1.3	11.8	12.1

^aValues are the mean of at least two experiments done in duplicate.

^bCCL2-promoter-luciferase-reporter activity assay.

^c3xGRE-promoter-luciferase-reporter activity assay.

^dEfficacy represented as percentage of maximal response of dexamethasone or hydrocortisone.

For each of the R¹ analogues, five different groups at the R² position on the APGRA scaffold were also explored (Scheme 1). Two further R²’s (cyanomethylene- and t-butyl-) were only prepared in the fluororophenyl- series, as these analogues lacked the ability to favor one transcription mode over the other and inclusion of these functional groups led to unexpected synthetic difficulties. The suite of molecules showed a broad variation in the ability to dissociate their activity profile and differentiate anti-inflammatory activity from transactivation potential (Table 1). In most cases, the dominant factors driving the activities of the APGRAs seemed to be provided by the arylpyrazole moiety; however, the inclusion of an aryl group at R² was able to increase selectively the transrepression activity in several molecules from the tolyl- and anisolyl- series to ∼66–75% that of Dex, although higher concentrations were needed (see 13be, 13cb, and 13ce in Table 1). It is noteworthy that none of these compounds elicited transactivation activity above background levels. These results indicate that more exploration may provide further analogues of interest.

The most desirable compounds for this study have a profile with ≥75% of the transrepression and ≤25% of the transactivation activity of Dex, respectively, at similar concentrations. While more of these compounds are likely to be useful in further studies aimed toward understanding the mechanisms of action, only molecules 11aa and 11ab that contained F at R¹ and also bore a side chain with a 2° alcohol with methyl or phenyl at R² displayed these properties (Figure 1 and Table 1). Therefore, these two compounds along with the corresponding molecules 13aa and 13ab containing 3° alcohols were further tested for their ability to limit insulin secretion (Figure 1c). All but 13ab were able to perform better than Dex in the GSIS bioassays, with the 2° alcohols 11aa and 11ab and 3° alcohols 13aa showing no reduction in insulin secretion at both 3 mM and 15 mM glucose concentrations. The fact that 13aa was observed to elicit 85% of the transactivation potential of Dex in the 3xGRE assay is noteworthy, as this compound did not display GSIS inhibition as expected from the commonly accepted transrepression/transactivation model. Together, these results suggest that the design of anti-inflammatory compounds with decreased side effects is achievable and that our current understanding of the role of the GRE in physiological regulation is still incomplete and warrants further mechanistic study.

Figure 1.

Bioassay data showing GR agonist activity and impact on glucose-stimulated insulin secretion. (A) CCL2-promoter-luciferase-reporter activity assay; x-axis, log concentration (molar); y-axis, %-maximal IL-1β response (relative promoter activity). (B) 3xGRE-promoter-luciferase-reporter activity assay; x-axis, log concentration (molar); y-axis, fold over control. (C) 832/13 cells were exposed to 100 nM of the indicated GR agonist compounds for 18 h, followed by measurements of glucose-stimulated insulin secretion.

Further exploration into the binding of APGRAs to the GR was accomplished through computational experiments. Both molecular docking studies, performed using AutoDock Vina (The Scripps Research Institute),⁶¹ and DFT calculations were used in this endeavor. A crystal structure of budesonide bound to the GR has been reported (PDB ID: 5NFP),³⁵ and the coordinates from this structure were used to construct the truncated standard binding pocket for the experiments reported herein. Further, the binding of DAC to the GR elicits an expansion of the standard binding pocket to accommodate the extra bulk of the arylpyrazole moiety appended to the intact steroidal core, and the crystal structure for the protein–ligand complex (PDB ID: 3BQD) was used to model an expanded binding pocket (Figure 2).⁶²

Figure 2.

Expanded ligand binding pocket of APGRA, deacylcortivizol, bound to GR (left) compared to ligand binding pocket of steroidal GC, budesonide, bound to GR (right). Residues R611 and Q570 are depicted to exhibit the conformational changes that lead to an expanded ligand binding pocket.

Previous studies modeling the interaction of APGRAs with structures similar to those reported herein have assumed that the pyrazole ring of the compounds would situate in the binding pockets at the location normally occupied by the A ring of the steroid ligands. This assumption always places the aryl moiety of the APGRAs either within the expanded binding pocket or in the region of the natural binding pocket that would expand to accommodate the added steric bulk.^43,62 However, the APGRAs are smaller in size than typical steroids and could potentially reside fully in either pocket if the orientation of the molecule is not fully analogous to that of the steroidal ligands. Therefore, these efforts commenced with a comprehensive exploration of the interactions of these compounds with the standard binding site.

Computational docking of the APGRAs in the standard pocket led to random results in which multiple ligand orientations were possible, and many of these differed from the orientation that budesonide adopted in the original crystal structure (SI section 1). To overcome the limitations of a rigid receptor model, DFT calculations were also performed on a truncated 5NFP receptor pocket with the synthesized compounds. It was hypothesized that flexibility of the residues surrounding the ligands may lead to a discovery of key residue interactions; however, results from these experiments also led to disordered protein–ligand interactions (SI section 1). For these experiments, a subset of molecules were modeled in the least sterically accessible conformation with respect to the orientation of the side chain, and the repositioning of the alcohol did not lead to more reliable binding results (data not shown). Therefore, it is unlikely that these compounds reside within the standard steroid binging pocket.

When performing molecular docking studies with the expanded GR pocket, the ligands all showed an analogous orientation, and clear trends in interactions with the receptor began to emerge. The orientation for the top calculated binding affinities mimicked that of DAC in the original crystal structure, and key residue interactions that have been reported in previous studies were observed with a majority of the synthesized compounds.⁶² Note, DFT calculations with the expanded binding pocket were not performed as the MM level of theory was deemed sufficient for this study.

Compared to the GR binding pocket, computational binding of the APGRAs to the expanded binding pocket resulted in reproducible orientations that showed defined interaction between the ligand and protein (Figure 3). Hydrogen bonding between residue Q570 and the nitrogen atom in the arylpyrazole moiety was observed, and the aryl moiety was tightly enclosed by nearby residues M604, A607, L608, R611, and F623. The Dex-bound structure shows similar hydrogen-bonding between the C3-ketone and residues Q570 and R611, with further contacts between residue N564 and the C11-hydroxyl group. Further, additional hydrophobic and π interactions between the protein and the arylpyrazole moiety play a role in increasing binding affinity.⁶³ Overall, the fluorophenyl- series was able to reside deeper within the expanded binding pocket, while the tolyl- and anisolyl- derivatives tended to be shifted out of the expanded binding pocket and further into the canonical binding pocket. This aligned the molecules further from helix-5 and closer to helix-10 in the GR LBD and weakened the critical hydrophobic and π interactions. This shift disrupts H-bonding to Q570 for the tolyl- and anisolyl- analogues. Molecules 13be, 13cb, and 13ce gain a π interaction between the aromatic side chain of the molecule and Y735, and this new contact may be responsible for the recovery of transactivation activity for these compounds (SI Section 1).

A re-evaluation of previously reported and an expansion of structures for a set of arylpyrazole GC analogues was undertaken to determine whether transrepression and transactivation activity could be differentiated in order to decrease the side effects, such as inhibition of insulin secretion, of these potentially therapeutic entities. This work showed that members of the previously synthesized fluorophenyl series of analogues (11aa, 13aa, 11ab, and 13ab) had anti-inflammatory activity approaching that of Dex and equal to that of HC, while 11aa and 11ab also displayed suppressed transactivation activity when compared to those drugs. All of these molecules, except 13ab, did not inhibit insulin secretion in the presence of glucose stimulation. These observations indicate that therapeutic potential of the 2° alcohols, 11aa and 11ab, has been overlooked previously. Further, the lack of insulin inhibitory side effects for 13aa, despite the noted transactivating properties, highlights a knowledge gap concerning the nuanced regulatory mechanisms employed by the GRE. Similar knowledge gaps have been reported, promoting the need for further investigation of mechanisms effecting glucose metabolism.⁵⁷

Members of the tolyl- and anisolyl- series of APGRA analogues were not as potent as the fluorophenyl- series; however, several of these compounds (13be, 13cb, 13ce) had anti-inflammatory potentials that were ≥66% and ≥75% that of Dex and HC, respectively. These molecules also displayed no transactivation above background, which makes them entities of interest for further studies. The computational studies showed conclusively that the arylpyrazoles with a small substituent at the para- position of the aryl moiety fit tightly within the expanded binding pocket that was previously observed in crystals of the DAC-GR complex. These studies also suggest that adjusting the placement of the ligands between the expanded and canonical steroid binding pocket will be useful for tuning the activity of future anti-inflammatory compounds that target the GRE.

GR signaling remains a complex area of study, and this work highlights that efforts to re-evaluate and further examine the phenomena that arise from the interplay of GRE transactivation and transrepression are still needed. Further studies into the SAR for therapeutically useful GC analogues, as well as in vivo investigations of the regulatory mechanisms for ligand-GR complexes, are underway with 11aa, 11ab, and 13aa, since these compounds displayed a greater ability than DEX to separate transrepression and transactivation activity and have shown the potential to avoid inhibition of insulin secretion and potentially the debilitating drug induced diabetes that plagues current long-term GC active compounds.

Reagents were purchased and used without further purification, except for methyl vinyl ketone (1), which was distilled immediately prior to use. Dry ethereal solvents were obtained by distillation from a potassium-benzophenone ketyl still; all other dry solvents were obtained via distillation over CaH₂. Thin-layer chromatography (TLC) was carried out using Sorbent Technologies silica G w/UV254 TLC plates. All purified compounds of interest were visualized as single spots using short-wave UV light; more permanent staining of TLC spots was achieved using vanillin, p-anisaldehyde, or ceric ammonium molybdate (Hannesian’s) TLC staining solutions. ¹H NMR and ¹³C NMR spectra were recorded on either a Varian Mercury Vx 300 MHz or Varian Inova 500 MHz instrument (Palo Alto, CA) in solutions of CDCl₃. High resolution mass spectrometry (HRMS) was performed using either a JEOL DART source operating in positive ion mode coupled to a JEOL AccuTOF JMS-T100LC mass spectrometer (Akishima, Tokyo, Japan), an Applied Biosystems QStar Elite mass spectrometer (Foster City, CA) equipped with a Sciex Turboionspray ESI source (Framingham, MA) operating in positive ionization mode, or an Exactive Plus Orbitrap mass spectrometer with an ESI source operating in positive ionization mode. Optical rotation was measured using a PerkinElmer 241 polarimeter (Waltham, MA) equipped with a sodium source lamp. IR spectra were collected using a Thermo Scientific Nicolet iS5 FTIR spectrometer (Waltham, MA) as a dry film on single NaCl plates. LCMS analysis was conducted on a Thermo Scientific Dionex UltiMate 3000 UHPLC system coupled to an Exactive Plus Orbitrap mass spectrometer (Waltham, MA) with an ESI source operating in positive ionization mode. All annotated characterization data, including NMR spectra of final compounds, and bioassay data can be found in the Supporting Information.

All air/moisture sensitive reactions were done in an argon atmosphere under anhydrous conditions, unless otherwise stated. All solvents and reagents were used without further purification unless otherwise stated. All heated reactions were performed using oil baths. Analytical thin-layer chromatography was performed using silica gel (normal phase) with a mobile phase consisting of EtOAc/hexanes unless otherwise stated. ¹H NMR chemical shifts were recorded in parts per million (ppm) Abbreviations used for ¹H NMR are s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, dd = doublet of doublets, and td = triplet of doublets.

Compound 6 was synthesized as previously reported.^36,64 The general procedure used to generate compound 10a was used to obtain both 10b and 10c, the tolyl- and anisolyl- pyrazole derivatives, which are novel arylpyrazole compounds. Detailed synthetic procedures, chromatographic methods, and characterization data of novel compounds can be found in the Supporting Information (SI Section 1).

Glossary

Abbreviations

APGRA

aryl pyrazole glucocorticoid receptor agonist

AP-1

activator protein 1 (transcription factor)

DART

direct analysis in real time

ESI

electrospray ionization

GC

glucocorticoid

GR

glucocorticoid receptor

GRE

glucocorticoid response element

hGR

human glucocorticoid receptor

IFN-γ

type II interferon

IL-1β

interleukin-1β

NR3C1

nuclear receptor subfamily 3, group C, member 1

NF-κB

B-cell nuclear factor kappa-light-chain-enhancer

R_f

retention factor

SAR

structure–activity relationship

TNF-α

tumor necrosis factor alpha

T1DM

type-1 diabetes mellitus

UHPLC

ultrahigh performance liquid chromatography

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00379.

• SI section 1: Synthetic procedures, characterization, raw and processed biological data, and molecular docking data (PDF)

• SI section 2: Smiles table with relative binding affinities (CSV)

Author Present Address

^∥ Department of Chemistry, Princeton University, Princeton, NJ 08544, USA

Author Present Address

^⊥ Department of Chemistry, University of California, Davis, CA 95616, USA

Author Present Address

^# Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830, USA

Author Present Address

^∇ Department of Biochemistry and Chemistry, Brigham Young University, Provo, UT 84602, USA

Author Contributions

^§ These authors contributed equally to this work.

NIH AI138136 (to J.J.C.)

The authors declare no competing financial interest.