The Development of Diphenyleneiodonium Analogs as GPR3 Agonists

Abstract

G protein-coupled receptor 3 (GPR3) is an orphan receptor potentially involved in many important physiological processes such as drug abuse, neuropathic pain, and anxiety and depression related disorders. Pharmacological studies of GPR3 have been limited due to the restricted number of known agonists and inverse agonists for this constitutively active receptor. In this medicinal chemistry study, we report the discovery of GPR3 agonists based off the diphenyleneiodonium (DPI) scaffold. The most potent full agonist was the 3-trifluoromethoxy analog (32) with an EC₅₀ of 260 nM and 90% efficacy compared to DPI. Investigation of a homology model of GPR3 from multiple sequence alignment resulted in the finding of a binding site rich in potential π-π and π-cation interactions stabilizing DPI-scaffold agonists. MMGBSA free energy analysis showed a good correlation with trends in observed EC₅₀s. DPI analogs retained the same high receptor selectivity for GPR3 over GPR6 and GPR12 as observed with DPI. Collectively, the DPI analog series shows that order of magnitude improvements in potency with the scaffold were attainable; however, attempts to replace the iodonium ion to make the scaffold more druggable failed.

G protein-coupled receptors (GPCRs) are the largest family of integral membrane proteins, and compounds targeting GPCRs represent approximately 35% of approved drugs¹. Over the past three decades, a number of GPCRs have been identified based on their protein structure without known endogenous ligands giving rise to a large group of orphan GPCRs. G protein-coupled receptor 3 (GPR3) is an orphan receptor that appears to be constitutively active resulting in activation of adenylyl cyclase through Gαs. Whether cAMP accumulation is the result of true constitutive activity or the presence of a chronic, ubiquitous ligand is not known². GPR3 belongs to a family of GPCRs, including GPR6 and GPR12, which share greater than 50% identity³. The subfamily is part of the MECA cluster of GPCRs being closely related to melanocortin, endothelial differentiation gene, cannabinoid, and adenosine receptors^{3b, 3c}.

GPR3 is highly expressed in the mammalian brain including the medial habenula, cerebral cortex, striatum, hippocampus, and cerebellum^{3b, 3c, 4}. Lower levels are also expressed in the testis, ovaries, and eye^3c. GPR3 constitutively activates Gαs leading to increased levels of cAMP in mammalian cell lines and neurons^{4a, 5}. GPR3 regulation of cAMP levels plays a role in promoting neurite outgrowth and survival^5–6. Additionally, cAMP signaling through GPR3 during postnatal development inhibits the proliferation of cerebellar granule cells⁷. GPR3 can also signal, independent of G-proteins, through β-arrestin 2⁸. GPR3 activation of β-arrestin along with regulation of β-amyloid precursor protein leads to stimulation of amyloid-β (Aβ) production⁹. GPR3 is also expressed in oocytes of multiple mammalian species where it is known to play an important role in maintaining meiotic arrest in prophase I¹⁰. The blocking of GPR3 constitutive activity leads to decreased cAMP levels and subsequent maturation of the oocyte through resumption of meiosis¹¹.

Work with GPR3 knockout mice has identified GPR3 as a novel drug target for several CNS mediated diseases including drug abuse, neuropathic pain, anxiety and depression related disorders, and Alzheimer’s Disease. GPR3 knockout mice self-administer more cocaine, demonstrate increased locomotor activity to acute doses of cocaine, establish conditioned place preference at lower doses, and demonstrate a decrease in days to reach self-administration acquisition². Nociceptive responses after sciatic nerve injury are higher in GPR3 knockout mice for thermal (both hot and cold) stimuli, but not mechanical stimuli¹². In addition, GPR3 knockout mice have reduced morphine anti-nociception in a tail immersion test¹². GPR3 knockout mice also display an increase in anxiety-like behaviors including decreased exploratory behavior in an open-field test, decreased exploration of novel objects, and aversion to open arms in an elevated plus-maze^4a. In addition, they display increased aggressive behavior in resident-intruder tests, and increased despair-like behaviors in forced swim and tail suspension tests^4a. However, a second study did not observe an altered level of anxiety in the open-field and the elevated plus maze in GPR3 knockout mice compared to wild-type¹³. Recent work has demonstrated that the genetic deletion of GPR3 reduces amyloid pathology in four different AD transgenic mouse models, and knockout of GPR3 alleviated cognitive deficits in APP/PS1 mice¹³. Interestingly, GPR3 regulation of Aβ appears to be exclusively through a β-arrestin pathway, not through cAMP stimulation^{9b, 14}. Finally, GPR3 appears to play a role in the function of brown adipose tissue with GPR3 knockout resulting in late onset obesity in mice¹⁵. In addition, GPR3 activation of thermogenic adipocytes counteracts metabolic disease, such that decreased body weight, increased calorie-burning, and increased glucose tolerance is observed in mice overexpressing GPR3 in brown and beige adipocytes¹⁶.

Despite the discovery, cloning and expression of GPR3 in multiple cell lines, as well as development of a GPR3 knockout mouse, the receptor remains an orphan. Previous work indicated that sphingosine 1-phosphate (S1P) can enhance cAMP accumulation in an HEK cell line transiently transfected with GPR3, but only by ~20% at 1 μM¹⁷. Subsequent studies demonstrate that S1P does not activate GPR3 mediated internalization in the β-arrestin PathHunter assay¹⁸, or cAMP accumulation in CHO or HEK cells^{4a, 19}. Due to these conflicting results, GPR3 remains classified as an orphan GPCR. Interestingly, a recent publication suggests that the N-terminal tail of GPR3 can act as a regulator of receptor constitutive activity. N-terminal truncations of GPR3 demonstrated that amino acids 18–27 were critical for receptor intrinsic activity and that a peptide of amino acids 18–27 could activate cAMP accumulation through GPR3 but not CB1¹⁶.

Two publications have described the discover of small molecule ligands for GPR3. A group developed a series of GPR3 inverse agonists, including the selective compound AF64394²⁰. In a high-throughput screening campaign, Ye et al. (2014) discovered several weak agonists, including diphenyleneiodonium (DPI) chloride. DPI was able to increase cAMP in a concentration dependent manner and demonstrated selectivity for GPR3 over GPR6 and GPR12¹⁹. In addition, DPI was able to increase calcium mobilization in a HEK Gα16 cell line, increase GPR3 internalization in U2OS cells, and cause translocation of β-arrestin 2 to the membrane¹⁹. However, while it is the most potent GPR3 agonist reported to date, DPI has weak potency for GPR3 activation, with a reported EC₅₀ for cAMP accumulation in the low micromolar range.

Herein we report a medicinal chemistry study of DPI as a potential scaffold for the discovery of GPR3 agonists. The design strategy had two primary goals. The first goal was to determine if it was possible to replace the iodonium ion with another atom or combination of atoms, the latter of which might be required to replace the size of the iodonium ion. The second goal was to conduct a structure activity relationship (SAR) study around the DPI scaffold to see if we could discover a more potent compound for GPR3 activity. A set of 20 constrained biphenyl analogs was assessed with either a sulfur, oxygen, carbon, or nitrogen atom in place of the iodonium ion. In addition, 20 DPI analogs which vary by phenyl ring substituent were synthesized to determine if phenyl ring substitution has an effect on compound potency and efficacy. In addition, a GPR3 homology model was developed and docking studies were conducted to determine potential interactions between GPR3 and the DPI scaffold.

Compound Synthesis

Compounds 2, 4, 6, 10, 12, 13, 14, 15, 16, 17, and 21 in Table 1 were synthesized following the synthetic methodology described in the Supplemental Material. Compounds 3, 5, 7, 8, 9, 11, 18, 19 and 20 were commercially available.

Table 1.

GPR3 Agonist Activity of Non-iodonium DPI Analogs

Compound #	Structure	EC50 (μM)	% Emax
2		382 ±149	19 ± 3*
3		67 ± 14	50 ± 5*
4		86 ± 42	40 ± 3*
5		29 ± 2	26 ± 3*
6		60 ± 13	72 ± 4*
7		45 ± 27	33 ± 2*
8		7.6 ± 0.6	44 ± 6
9		40 ± 4	69 ± 8*
10		114	39 ± 3*
11		> 100 μM	ND
12		21 ± 9	24 ± 12*
13		ND	4 ± 3*
14		> 100 μM	21 ±11*
15		> 100 μM	16 ± 7*
16		158 ± 69	24 ± 10*
17		66 ± 49	44 ± 11*
18		81 ± 11	63 ± 1*
19		> 100 μM	18 ± 2*
20		5.1 ± 0.2	13 ± 1*
21		105 ± 24	60 ± 4*

ND = not determined.

^*Precent Emax reported is for the highest concentration tested, 100 μM.

All of the DPI analogs in Table 2 were synthesized following the synthetic route as described for 3-methyldiphenyleneiodonium triflate (25), shown in Scheme 1. 2-Bromoaniline (43) was coupled to 4-methylbenzeneboronic acid (44) using typical Suzuki coupling conditions to form biphenyl intermediate 45. A Sandmeyer reaction was used to convert the aniline to iodide 46, followed by oxidative cyclization using meta-chloroperoxybenzoic acid to form the desired DPI analog 25. The use of 4-substituted boronic acids ensured only one compound since it is a symmetrical oxidative cyclization. The same is true when using anilines substituted para to the bromide.

Table 2.

GPR3 Agonist Activity of Iodonium DPI Analogs

Compound #	Structure	EC50 (μM)	% Emax
1		5.17 ± 0.75
22		3.86 ± 1.47	110 ± 5
23		ND	8 ± 3*
24		1.89 ± 0.46	104 ± 11
25		6.12 ± 2.13	105 ± 3
26		1.53 ± 0.23	80 ± 6
27		0.921 ± 0.181	72 ± 2
28		1.31 ± 0.22	73 ± 4
29		0.937 ± 0.138	70 ± 5
30		1.34 ± 0.17	99 ± 1
31		369 ± 71	25 ± 3*
32		0.260 ± 0.043	90 ± 3
33		1.30 ± 0.62	44 ± 4
34		ND	40 ± 19*
35		3.02 ± 0.62	66 ± 6
36		0.923 ± 0.172	91 ± 5
37		0.816 ± 0.343	43 ± 6
38		19.3 ± 6.3	36 ± 3*
39		0.277 ± 0.078	60 ± 3
40		0.455 ± 0.057	87 ± 2
41		0.137 ± 0.010	70 ± 6
42		0.230 ± 0.042	56 ± 6

ND = not determined.

^*Precent Emax reported is for the highest concentration tested, 100 μM.

Scheme 1.

a. 1,2-dimethoxyethane, (PPh₃)₂PdCl₂, K₂CO₃, 80°C; b. THF, 4.0M HCl; NaNO₂, H₂O; Kl, H₂O; c. DCM, mCPBA, 0°C; CF₃SO₃H.

Structure-Activity Relationship Study for DPI scaffold

All of the compounds in Tables 1 and 2 were tested for their ability to stimulate cAMP accumulation through hGPR3 stably expressed in HEK293 cells. Table 1 shows the results of an attempt to replace the iodonium ion of the DPI scaffold with a carbon, nitrogen, sulfur, or oxygen bridge, or a combination of those atoms. None of the iodonium ion replacement compounds were full agonists when tested up to 100 μM. The most efficacious compound was 5,6-dihydro-5,6-dimethylphenanthridine (6) at 72% compared to DPI, but it had decreased potency with an EC₅₀ of 60 μM. Sixteen of the twenty compounds were ≤ 50% efficacious compared to DPI. The most potent two atom bridge was benzo[c]cinnoline with an EC₅₀ of 29 μM but it was only 26% efficacious. A nitrogen bridge was the most effective replacement. Carbazole (8) had an EC₅₀ value of 7.6 μM but was only 44% efficacious. Methylcarbazole (9) was the second most efficacious compound at 69%, but it also had decreased potency with an EC₅₀ of 40 μM. None of the carbon bridged compounds were active. Of the three sulfur bridged compounds, dibenzothiophene (18) was the most active, but was a weak partial agonist with an EC₅₀ of 80 μM and 63% efficacy compared to DPI. An oxygen bridge seemed to make the scaffold even weaker, as seen with 3-fluorodibenzofuran which had an EC₅₀ > 100 μM and was only 60% efficacious. Ring substituted analogs of 8 were also made analogous to active compounds 27 and 29 in Table 2 but showed no change in activity from 8 (data not shown).

Table 2 shows the results of the 20 DPI analogs which vary by phenyl ring substitution, in addition to commercially available DPI as its hydrochloride salt (1) and synthesized DPI as its triflate salt (22). As expected, the synthetic sample of DPI (22) and the commercially available compound (1) had roughly the same potency, 3.86 ± 1.47 μM and 5.17± 0.75 μM respectively. Nine of the twenty analogs had EC₅₀ values less than 1 μM, and four had eC₅₀ values less than 300 nM. While substitution improved potency, it tended to weaken efficacy compared to DPI. Only seven of the compounds maintained efficacy above 80% and were considered full agonists. Six of the compounds showed efficacy between 60% and 80% and were considered to be partial agonists. Ring expansion to the 10H-Dibenz[b,e]iodinium (23) rendered the compound inactive. Phenyl ring substitution at the 3-position improved potency in every case except the 3-methyl substituted compound (25) which was slightly weaker with an EC₅₀ of 6.12 μM and the 3-methyl sulfone substituted compound (31) which was inactive. Phenyl ring substitution at the 5-position as seen in analog 34 also rendered the compound inactive.. Addition of a 4-fluoro group to the 3-methyl compound, 3-fluoro compound, and 3-trifluoromethyl compound to form 35, 36, and 37 respectively, slightly increased potency but had mixed effects on efficacy.

The most potent full agonist was the 3-trifluoromethoxy analog (32) with an EC₅₀ of 260 ± 43 nM and 90 ± 3% efficacy compared to DPI. Attempts to improve its full agonist activity failed. The 3-trifluoroethoxy analog (33) had a potency that was nearly an order of magnitude less than 32 and greatly reduced efficacy (44% vs 90%). The 3,4-difluoromethylenedioxy analog 38 was inactive. Substituents on the other phenyl ring had mixed results on potency but reduced efficacy in all 4 cases resulting in partial agonists. The most potent partial agonist was the 3-fluoro-7-trifluoromethoxy analog (41) which was twice as potent with an EC₅₀ of 137 ± 10 nM but was slightly less efficacious at 70 ± 6 % compared to DPI (Figure 1). These analogs represent the most potent GPR3 full and partial agonists discovered to date.

Figure 1.

Agonist activity of DPI and compounds 32 and 41 in the HEK hGPR3 cAMP accumulation assay. Data are expressed as a percentage of the maximal DPI response. cAMP data are the mean ± SEM of three to seven independent experiments conducted in duplicate.

GPR3 Model Development and Docking Studies

GPR3 homology models were built using four different templates, cannabinoid receptor 1 (CB1), sphingosine-1-phosphate receptor 1 (S1PR), β2 adrenergic receptor (ADBR2), and adenosine alpha 2A receptor (ADORA2A), using a common MUSCLE algorithm²¹ multiple sequence alignment (Supplemental Figure 1). As shown in Supplemental Table 1, all four templates had modest sequence identity to the GPR3 target: CB1(24%), S1PR (26%), ADBR2 (21%) and ADORA2A (23%). All aspects of the MUSCLE multiple sequence alignment gave rise to expected X.50 Ballesteros-Weinstein conserved residues (N58^1.50, D86^2.50, R134^3.50, W161^4.50, A201^5.50, P262^6.50, P294^7.50) with the exception of A201^5.50. This residue has only 78% conservation in helix 5 compared to the other conserved residues which have greater than 88% conservation in class A receptors²². The alignments shown in Supplemental Figure 1 illustrate correspondence of template and key conserved residues essential to a common hydrophobic core (V/I/L 6.40/F6.44) and (W6.48/Y7.53), in addition to activation microswitches: TM3-TM6 ionic locks, aromatic-toggle, 3–7 lock, and tyrosine toggle linked to the proline nPxxy in TM7 switch residues, identified from previous study of class A GPCRs²³. The alignments confirmed the conservation of all of these motifs in the GPR3 sequence. The full-length constructed model of GPR3 based on the crystal structure of agonist bound CB1 (5XR8)²⁴ yielded the highest Qualitative Model Energy Analysis (QMEAN)^{21, 25} score as shown in Supplemental Table 2. Annealing and molecular dynamics (MD)-equilibration did not alter this relative QMEAN assessment^{25a, 26}

The CB1 template GPR3 homology model was advanced to preliminary Schrodinger GLIDE-XP docking and induced fit surveys of selected DPI scaffold agonists (1, 18, 28, 32, 41). The goal of the initial docking and induced fit was to examine the nature of the DPI agonist scaffold interaction in order to rationalize the sensitivity of iodonium ion replacement in the scaffold, and the sensitivity to small substituent variations off the DPI-scaffold. Figure 2 (left) shows the top GLIDE XP generated docking poses for DPI (1) and analogs (32, 41 and 18) with a range of agonist potencies spanning 2 orders of magnitude (EC₅₀s = 5.2 μM, 260 nM, 137 nM, and 80 μM respectively). Figure 2 (right) shows selected LIGPLOT representations for 1 and 32. These poses emphasize how the cationic ligands are situated in a box of surrounding aromatic residues. The predicted binding site in GPR3 for these DPI-scaffold ligands is constrained by W260 (W6.48, a conserved hydrophobic core residue found in class A GPCRs), Y278 which is conserved between GPR3, GPR6 and GPR12 (Supplemental Table 1), F120 poised below the DPI-analogues, Y280 in a second coordination sphere but not in direct contact with the cationic DPI-scaffold ligands, and Y188 which is present in extracellular loop-2 (ECL2) between helices 3 and 4. Abundant π-stabilizing contacts between GPR3 and the ligands are possible, including a potentially novel π-cation interaction providing stabilization for the bound iodonium containing ligand.

Figure 2.

Left: GLIDE XP poses of 1 (DPI), 32, 41 and 18 docked into equilibrated GPR3 model showing similar disposition of the residues in the orthosteric binding site. Right: LIGPLOT depictions of 1 and 32 binding site interactions illustrating prevalence of DPI-scaffold potential interaction with Y188 in the ECL2.

Both Schrodinger Cavity Finder and complimentary VINA-blind docking (with a large box encompassing the entire model) identified the preferred binding site, set to be in the transmembrane intrahelical region near the extracellular domain. Supplemental Figure 2 shows that the VINA-blind docking and Schrodinger-cavity prediction were in agreement and that the MMGBSA predicted binding free energies paralleled the DPI-scaffold EC₅₀ potency assessments. Additionally, examination of both the GLIDE-XP induced fit and VINA+MMGBSA predicted binding pose positioning resulted in ligand substituent variations directly contacting the putative W260/F120 aromatic toggle switch at the bottom of the blind docking identified preferred binding site^{23a, 23c, 25, 27}.

Two groups have developed GPR3 homology models and examined the nature of inverse agonist (AF64394)²⁸ and agonist (DPI) binding^27c. Bharathi and Roy employed an iterative threading refinement server approach (i-Tasser)²⁹ to produce a full length GPR3 model followed by docking and dynamics study of AF64394 binding, but it appears the structure (Figure 2) may be incorrect²⁸, having a protonated center in the [1,2,4]triazolo[1,5-a]pyrimidine based scaffold. Capaldi and co-workers built their homology model using the agonist form of the adenosine receptor template (5G53 employing the GOmodo web server³⁰ followed by HADDOCK docking. Similar to our model these groups predict an orthosteric binding pocket with key π-π interactions for both the inverse agonist and agonist bound forms including aromatic π-π stacking interactions with F97, F101, W43 and Y280 for AF64394²⁸ and potential interactions of DPI with W260, F120 and F263^27c.

DPI analog specificity for GPR3

In the original Ye et al. paper, DPI demonstrated specificity in the cAMP accumulation assay for GPR3 over GPR6 and GPR12¹⁹. This may be explained by unique π-stabilizing contacts between GPR3 and this cationic ligand that are not present in GpR6 or GPR12. GPR3, GPR6 and GPR12 have a high global sequence identity (56–58%: Supplemental Table 1) that translates to similar binding sites. However, there are several substantive differences between amino acids defining the predicted binding sites of GPR3, GPR6 and GPR12 (Supplemental Table 3). One potential key difference is the presence of a tyrosine in the ECL2 for GPR3 (Y188) compared to an arginine in the same location for GPR6 (R220) and GPR12 (R192). A C177-C184 salt bridge was suggested to be present in the ECL2 of GPR3 from three (CB1, S1PR1, and B2AD) of the four template-based constructions. Our model predicts a potential DPI-GPR3 Y188 interface through what appears to be a π-cation interaction. Combined with the lack of a tyrosine residue in GPR6 or GPR12, this interaction may partially explain the selectivity of DPI analogs for GPR3 over GPR6 or GPR12. Previous studies on a GPR6 homology model suggest GPR6 R220 points down into the binding pocket similar to what we propose for GPR3 Y188 and acts as a key hydrogen bond donor of ligand interactions³¹. In addition, this type of loop residue insertion into orthosteric binding sites is present in CB1-antagonist structures³² and the μ-opiate receptor agonist structure (5C1M) where a portion of the N-terminus inserts into the orthosteric binding site stabilizing the BU72-agonist³³.

We tested the ability of both DPI (1) and the 3-trifluoromethoxy analog (32) to stimulate cAMP accumulation in HEK cells stably expressing hGPR6 or hGPR12. Neither DPI (1) nor 32 were able to increase the cAMP response over basal levels in either the HEK GPR6 cell line or the HEK GPR12 cell line, indicating that this analog of DPI maintains the receptor specificity between GPR3 and GPR6 or GPR12 (Figure 3). Both DPI (1) and 32 were tested for activity in a HEK parental cell line. Neither compound increased cAMP in the HEK parental cell line up to 10 μM (Figure 3). Also as demonstrated in Figure 3, due to the constitutively active nature of these receptors, over expression of GPR3, GPR6 or GPR12 in HEK cells causes an increase in basal cAMP levels compared to a parental cell line, similar to results from other studies^{3b, 17, 34}.

Figure 3.

Agonist activity of DPI and compound 32 in the cAMP accumulation assay in HEK cells over-expressing either hGPR3, hGPR6, hGPR12 or HEK293 cells alone. DPI and 32 only increase cAMP accumulation in cells overexpressing GPR3, but not GPR6, GPR12, or in HEK293 cells alone. Data are the mean ± SEM of three independent experiments conducted in duplicate, **** p<0.0001.

Based on the predicted π-cation interaction between GPR3 Y188 and the iodonium ion of DPI we introduced a point mutation into GPR3 substituting the tyrosine for an arginine at position 188, to mimic the amino acid found in the same position for GPR6 and GPR12. We also mutated the comparable arginine at 220 in GPR6 to a tyrosine (GPR6 R220Y). Interestingly, both DPI (1) and 32 were able to increase cAMP accumulation in the GPR3 Y188R mutant similar to the wildtype receptor (Figure 4). In addition, substitution of the tyrosine into the GPR6 receptor at position 220 in ECL2 did not confer activity for DPI (1) or 32. These data suggest that the potential π-cation interaction between the iodonium ion and Y188 is not necessary for DPI activation of GPR3, and Y188 does not, on its own, confer the selectivity of DPI and its analogs for GPR3 over GPR6 and GPR12.

Figure 4.

Both DPI and compound 32 retained activity in the GPR3 Y188R mutant receptor in the cAMP accumulation assay with similar potency and efficacy compared to wildtype GPR3. The agonists also displayed no activity in the GPR6 R220Y mutant similar to GPR6 wt, and cells transfected with control plasmid (pcDNA3.1). Data are the mean ± SEM of three independent experiments conducted in duplicate.

MMGBSA scoring for docked configurations of compound 32 at GPR3 vs GPR6 (−31.3 vs. −26.4 kcal/mol, respectively) revealed overall significantly worse free-energies for GPR6 binding compared to GPR3 so in general, the models align with the observed selectivity. A more detailed examination of the potential contribution of each amino acid through residue decomposition analysis of the amino acids within 4 Å of the docked agonist (32), suggested that Y188 does not contribute overwhelmingly to the MMGBSA-free-energy score. Other residues such as W260 and T264 also contribute to the stabilization of the bound charged ligand (Supplemental Table 4). Further computational and mutational study of the receptor is necessary to determine the structural basis for DPI analog selectivity of GPR3 over the other family members.

Mulliken Charge and QM(DFT) Energetic Analysis for a Model System Mimicking the Binding Site Interactions

The increase in activity of DPI analogs with highly electronegative phenyl ring substituents, like 32, and the presence of potential π-cation interactions with GPR3, suggested that the charge density of the iodonium ion might be driving potency. Mulliken charges were calculated to determine if the SAR of the DPI analogs was driven in part by the π-cation and π-stacking interactions observed in the model. Supplemental Figure 3 illustrates that frontier molecular orbitals in the energetic vicinity of the HOMO (highest occupied molecular orbital) /LUMO (lowest unoccupied molecular orbitals) are found on the cationic ligand (LUMO) and surrounding aromatic residues (HOMO-3 through HOMO). This suggests the π-cation stabilization may be important in terms of electron density overlap and in a charge-transfer, π-cation stabilization motif, with transfer from the highest occupied orbital regime to the unoccupied orbital. The nature of the HOMO-1 thru LUMO+2 orbitals were examined for selected ligands (1, 18, 28, 32, 41) and surrounding residues. As shown in Figure 5, Mulliken and potential derived charges were examined as a function of SAR substitutions. Variations of positions X- and Z- altered the charge density in the DPI scaffold to a modest degree as shown by monotonic changes in the iodonium atom. The substituent effects modified the E(complex) QM-derived interaction energy in a manner that correlated with the observed EC₅₀ values for each of the ligands (R²=0.94) confirming that halogen charge density may be a factor in potency.

Figure 5.

A simplified model calculation demonstrating π-cation interactions as a partial determinant for the trend in potency of the DPI scaffold as correlated with potential binding-affinity. A) Table of QM derived Mulliken charges and their QM derived interaction energies compared to agonist potency. B) Stick figure of DPI scaffold and a phenyl ring. C) Linear relationship between QM interaction energies and functionally determined EC₅₀ of select ligands.

GPR3 is involved in the regulation of several physiological processes, with GPR3 agonists having potential as therapeutic targets for drug abuse, neuropathic pain, and anxiety and depression related disorders. There is a limited number of available GPR3 agonists and inverse agonists for pharmacologic evaluation of this constitutively active, orphan receptor. In this study, we have conducted a SAR investigation of the DPI scaffold focused on two goals, replacement the iodonium ion and effects of phenyl ring substitution. So far, iodonium replacement has not resulted in compounds with increased activity. The most active compound was a nitrogen replacement, carbazole (8), which had an EC₅₀ of 7.6 μM roughly similar to DPI but was only 44% efficacious. All of the other analogs were much less potent and most were less efficacious. Of the DPI phenyl substituent analogs, it was discovered that a 3-trifluoromethoxy group (32) increased potency an order of magnitude over DPI without losing significant efficacy (EC₅₀ 260 nM, E_max 90%). Several analogs of 32 were synthesized, leading to the discovery of the 3-fluoro-7-trifluoromethoxy analog (41) which had increased potency, but was slightly less efficacious. Analog 41 is the most potent GPR3 agonist reported to date. A GPR3 homology model was developed and docking studies indicated pronounced π-π stacking and π-cation interactions as major stabilizing facets for DPI scaffold binding. In addition, MMGBSA free energy analysis showed a good correlation with trends in observed EC₅₀s. Compound 32 retained high receptor selectivity for GPR3 over GPR6 and GPR12, other closely related orphan GPCRs. Our homology model suggested a possible π-cation interaction between tyrosine 188 and the iodonium ion as a potential mechanism for DPI selectivity between GPR3 and the other two family members. However, DPI and 32 retained activity at GPR3 Y188R mutant receptors, suggesting other receptor-ligand interactions may be critical for receptor activation and selectivity. Overall, GPR3 represents a novel CNS target and further development of potent and selective agonists will be critical to probe the role of GPR3 in important neurophysiological and neuropathological conditions.