Electronic Sculpting of Ligand-GPCR Subtype Selectivity: The Case of Angiotensin II

GPCR subtypes possess distinct functional and pharmacological profiles, and thus development of subtype-selective ligands has immense therapeutic potential. This is especially the case for the angiotensin receptor subtypes AT1R and AT2R, where a functional negative control has been described and AT2R activation highlighted as an important cancer drug target. We describe a strategy to fine-tune ligand selectivity for the AT2R/AT1R subtypes through electronic control of ligand aromatic-prolyl interactions. Through this strategy an AT2R high affinity (Ki = 3 nM) agonist analogue that exerted 18,000-fold higher selectivity for AT2R versus AT1R was obtained. We show that this compound is a negative regulator of AT1R signaling since it is able to inhibit MCF-7 breast carcinoma cellular proliferation in the low nanomolar range.

0-10 nM (8 data points in triplicate). Non-specific binding was determined in presence of 6 μM cold AII. Competition assays were performed using a concentration of [ 125 I]-AII of 1 nM and various concentrations of unlabeled ligands, as indicated in the Figure 3. Samples were incubated for 2 hours at 4°C. Receptor-bound and free radioligand were separated by filtration through Whatman GF/B filters, pre-soaked with 0.3% polyethylamine. The filters were washed with 5 ml of ice-cold binding buffer and transferred to scintillation tubes. Radioactivity was counted on a Beckman LS6000 liquid scintillation counter and data were analyzed by non-linear regression using Prism software (GraphPad). K i values were calculated according to the Cheng and Prusoff equation with a K D for [ 125 I]-AII of 1.8 nM (AT1R) and 2.3 nM (AT2R). Experiments on whole cells were performed in both AT 1 R-and AT 2 R-HEK-293 stable cell lines, as previously described in detail 4 with minor modifications. Novel peptides and references compounds, CGP42112 and candesartan, were examined for their ability to inhibit specific binding of [ 125 I]-Ang II, and nonspecific binding was determined using 10μM cold Ang II. Following non-linear regression, IC 50 values, representing the concentration at which each ligand displaced 50% binding of [ 125 I]-Ang II in either AT 1 R-or AT 2 R-HEK-293 cell lines, were estimated.Log ratios of IC 50 values for each ligand at AT 1 R:AT 2 R were determined as a measure of AT 2 R selectivity.
Peptide synthesis and sample preparation. The synthesis of the following peptides: 6 -P 7 -F 8 ) was achieved using Fmoc/tBu methodology. 2-Chlorotrityl chloride resin and N α -Fmoc (9-fluorenylmethyloxycarbonyl) amino acids were used for the synthesis. Peptide purity was assessed by analytical HPLC, mass spectrometry (FABMS, ESIMS) and amino acid analysis. The samples were prepared for NMR spectroscopy by dissolving the peptide in 0.01 M KPi buffer (pH = 4), containing 0.02 M KCl. 2,2-dimethyl-2-sila-pentane sulfonate (DSS) was added to a concentration of 1 mM as an internal chemical shift reference. Peptide concentration was commonly 5 mM in 90% 1 H 2 O / 10% 2 H 2 O. Trace amounts of NaN 3 were added as a preservative.
MCF7 breast carcinoma cell proliferation assay. MCF7 breast carcinoma cells in log phase were treated with drug vehicle alone (which was not toxic to cells) or increasing concentrations of [Y] 6 -AII. Cells were counted daily and the effects of the analogue on proliferation expressed as cell numbers relative to controls which received only the drug vehicle ( Figure S8). [Y] 6 -AII efficiently inhibited proliferation at concentrations of 10 -8 M and showed evidence of anti-proliferative effects at 10 -9 M. The analogue showed an IC50 of ~ 5 x 10-8 M in MCF cells.
NMR spectroscopy. NMR spectra (2D 1 H-1 H TOCSY and 2D 1 H-1 H NOESY) were acquired at 500 MHz using a Bruker Avance 500 spectrometer. For water suppression excitation sculpting with gradients was used. The mixing time for TOCSY spectra was 80 ms. Mixing times for NOESY experiments were set to 100, 200, 350 and 400 ms to determine NOE build -up rates. A mixing time of 350 ms provided sufficient crosspeak intensity without introducing spin-diffusion effects in the 2D -NOESY. Inter-proton distances for were derived by measuring cross-peak intensities in the NOESY spectra. Structure calculations were performed with the program CYANA. The mole fraction of the peptide molecules in the cis isomeric form (Xcis) was obtained by measuring the areas of well-resolved peaks corresponding to the same proton resonance in the cis and trans forms in 1D spectra. To investigate solvent protection values for amide protons, the amide proton temperature coefficients (Δδ/ΔT) were measured in a range of temperatures from 283K to 308K. A plot of Δδ/ΔT vs. the chemical shift deviation (CSD) of the measured amide proton resonances at 283 K was constructed with appropriate random coil chemical shift correction 7 . The dashed line (Δδ/ΔT = −7.8 (CSD) −4.4) represents the cut off of Δδ/ΔT between exposed and sequestered NHs of proteins. Gradients above the dashed line indicate exposed NHs, whereas those below indicate sequestered NHs. Diffusion ordered spectroscopy (DOSY) spectra were recorded using the Bruker microprogram stebpgp1s19 at 292 K. The eddy current delay (Te) was set to 5 ms. The diffusion time was adjusted to 100 ms .The duration of the pulse field gradient, δ g , was optimized in order to obtain 5% residual signal with the maximum gradient strength with the resulting δ value of 3.6 ms. The pulse gradient was increased from 2 to 95% of the maximum gradient strength using a linear ramp 16k data points in the F2 dimension (20 ppm) and 16 data points in the F1 dimension were collected. Final data sizes were 16k×128.
Modeling studies. Generation of the homology model. A Blast search for the sequence of the human type-1 angiotensin II receptor (P30556) and human type-2 angiotensin II receptor (P50052) revealed as the closest template the structure of the human chemokine receptor CXCR1 (PDB ID 2LNL). However, as this structure is an Apo-form with a completely closed binding pocket and therefore not suitable for our purpose, we used instead the closely-related CXCR4 chemokine Receptor in complex with a cyclic peptide (PDB ID 3OE6).Importantly, the compounds of interest, AII and the Y6 analogue, are agonists which likely bind to the active AT1R and AT2R. In order to represent well the active AT1R and AT2R, we used as a second template the closest-related active GPCR, which was found to be the beta2 adrenergic receptor (PDB ID 3SN6). Homology modeling was carried out using the MOE package. 8 After sequence alignment (default settings) and manual corrections which ensured alignment of highly conserved residues as well as integral helices, 10 models were generated using the Amber12:EHT force field. 9 The best models for the AT1R and AR2R were chosen for subsequent docking studies.
Docking studies. In order to obtain a reasonable starting structure, we used the GOLD software (Version 5.0) 10 and docked AII and the Y6-analouge into the AT1R and the AT2R applying a constraint between the alpha-carboxylate group of the Phe8 and the residues K199(AT1R)/K215(AT2R). In order to dock the peptide into the target, the binding site was defined as a large region of 30 Å around the NZ atom of Lys199. As scoring function, we applied the GoldScore. All other parameters were set to default.
Conformational space analysis. Subsequently, the conformational space of the obtained complexes was extensively explored using LowModeMD Search for the ligand (completely free) and receptor residues within 12 Å of the ligand (fixed backbone atoms and completely free loop regions), as implemented in the MOE package. The LowModeMD Search method 11 generates conformations using a short ~1 ps run of molecular dynamics which is followed by an energy minimization. Obtained conformations are stored in a database and ranked according to their potential energy. The LowModeMD Search was carried out using the Amber12:EHT force field 9 and Born solvation.

Final model selection.
Complexes with low energy obtained during the conformational space analysis were validated using available mutagenesis studies (see below). Final selected models were subjected to a comprehensive energetic minimization (Amber12:EHT force field 9 and Born solvation) using the MOE package 8 .
Calculation of the binding pocket volume. The volume was determined for the AT1R and AT2R in complex with [Y] 6 analogue using the VMD plugin VOLSURF (default settings, http://www2.fc.up.pt/PortoBioComp/Software/Volarea/Home.html). The ligand was removed prior to volume calculations that were carried out using a box spanning the complete transmembrane binding site (yellow box, Figure S9).