Determining the True Selectivity Profile of αv Integrin Ligands Using Radioligand Binding: Applying an Old Solution to a New Problem

Introduction

The alpha-v subset of the arginyl-glycinyl-aspartic acid (RGD) integrin family has been targeted over the past three to four decades for a number of diseases that have included fibrosis and cancer.^1–4 This subset of integrins, consisting of alpha-v beta-1 (αvβ1), alpha-v beta-3 (αvβ3), alpha-v beta-5 (αvβ5), alpha-v beta-6 (αvβ6), and alpha-v beta-8 (αvβ8), have been shown to interact with a range of endogenous ligands that play key roles in cell adhesion and activation of intracellular signaling pathways, enabling communication between the extra- and intracellular environments. ⁴ Over the past 10 years, there has been a significant body of literature implicating αvβ6 as a key player in idiopathic pulmonary fibrosis (IPF) and its role in activation of transforming growth factor β1 (TGFβ1).^5–7 As such, a drug discovery program to identify small-molecule αvβ6 inhibitors was initiated that identified early hits by cross-screening analogues from a focused screening set available within the GlaxoSmithKline (GSK) corporate collection.

Once a small-molecule drug target has been selected, the initial focus is upon generation of novel chemical entities that investigate the unexplored chemical space for the interaction with the biological target. The source of these new chemical templates can be generated from a number of different methods that can include high-throughput screening (HTS) of millions of small-molecule libraries or focused screening sets containing chemical starting points, redesigning existing templates from the scientific literature, and/or modeling from an endogenous ligand (if a structure for this is available). Regardless of the source of the small molecule, biological assays are required to test the activity as well as characterize and define the desired properties of a potential drug candidate. There exists a fine balance between the throughput of an individual or panel of assays and their limitations in terms of physiological relevance and interassay comparisons. In the case of αvβ6 as a potential therapeutic target for IPF, in addition to the desired properties for a small molecule (low molecular weight, CLogP and hydrogen bond donors/acceptors) to give it the best chance of progression as a drug, ⁸ there is the selectivity over the other RGD integrins, specifically the αv integrins due to the shared αv subunit. This is due to the potential liabilities associated with engagement of the non-αvβ6 αv integrins due to their involvement in angiogenesis/vascular permeability (αvβ3/5^9,10) and autoimmunity (αvβ8 ¹¹ ).

Historically, the determination of the ability of an integrin to compete with an endogenous ligand has been measured in either a cell adhesion assay^12,13 or competitive binding enzyme-linked immunosorbent assay (ELISA). ¹⁴ The IC₅₀ values generated via these techniques measure the ability of the integrin ligand to inhibit the binding of an endogenous peptide ligand to the integrin that is either recombinantly expressed in a cell line (cell adhesion) or a soluble integrin protein preparation (ELISA). Although efficient in being able to differentiate the rank order of IC₅₀ values of a set of ligands against a particular integrin and also being high throughput (allowing hundreds of compounds to be screened every week in a full concentration-response curve [CRC] format), one of the limitations of both these assay types is that a true affinity (K_I) value cannot be calculated. To measure a true affinity or K_I of a ligand, the Cheng-Prusoff equation ¹⁵ is required that allows the binding affinity (K_D) and the amount of competing ligand to be taken into consideration. This equation corrects for these parameters and allows a true affinity estimate for a ligand in the absence of a competing ligand. In the assay system mentioned above, the parameters required for the Cheng-Prussoff can either not be determined or are a crude estimate. In addition, these assay types do not allow a full characterization of the type of binding of the ligands in terms of being competitive and reversible (or not), and the kinetics of the ligand-integrin interaction cannot be ascertained.

As alternatives to the classical methods of determining integrin ligand affinity, a number of options include both new technologies (surface plasmon resonance [SPR], ¹⁶ fluorescence polarization [FP] binding ¹⁷ ) and more established methodology (radioligand binding ¹⁸ ). Again there are advantages and disadvantages with all of these: SPR allows the determination of k_on and k_off of unlabeled ligands, but analysis of small molecules is challenging due to the small changes in the resonance observed with low molecular weight ligands; FP binding assays are high throughput but suffer the same limitations as those described for the cell adhesion and ELISA formats; radioligand binding assays allow accurate binding characterization, including kinetics, but are low throughput, higher cost, and not as safe relative to nonradioactive alternatives. However, despite its limitations, the radioligand binding assay still remains one of the most informative methods for characterizing the interactions between a ligand and its protein/receptor binding partner. Interestingly, the method has only recently been used in characterization of the binding of ligands with the αvβ6 integrin^19,20 and has only been used sparingly in the integrin field.^21,22

Therefore, studies were completed to aid in the identification and profiling of high-affinity and αvβ6 selective small-molecule RGD mimetics by validation of the most appropriate test systems to investigate these characteristics within the constraints of throughput. In addition, to determine the dissociation profiles of small-molecule RGD mimetics from the αvβ6 integrin, the design and implementation of a novel radioligand integrin binding assay are described. This has allowed the opportunity to compare the kinetics of potential drug candidates at a much earlier stage of the optimization process and thus increase the chance of discovery and development of a drug with a long duration of action. This would only allow molecules with the most desirable profiles, in terms of affinity, selectivity, and kinetics, to be taken forward for further analysis and aid in the characterization and development of a novel αvβ6 small-molecule RGD mimetic inhibitor as a potential therapeutic agent for IPF.

Materials and Methods

Materials

All the small-molecule RGD mimetics used in this study were synthesized by the Fibrosis and Lung Injury Discovery Performance Unit Medicinal Chemistry group at GSK Medicines Research Centre (Stevenage, UK). Full-length LAPβ₁ (fLAP₁) was purchased from Sigma-Aldrich (Gillingham, UK). Truncated LAPβ₁ (tLAP₁—GRRGDLATIHG), truncated LAPβ₃ (tLAP₃— HGRGDLGALKK), and the αvβ6 selective peptide NAVPNLRGDLQVLAQKVART (A20FMDV2 derived from the foot-and-mouth disease virus ²³ ) were synthesized by Cambridge Research Biochemicals (Cleveland, UK) ( Fig. 1A ). Compound A (pan-αv small-molecule RGD mimetic) was radiolabeled with [³H] via catalytic reduction by RC TRITEC Ltd. (Teufen, Switzerland) and had a specific activity of 16.1 Ci/mmol ( Fig. 1A ). All other chemicals and reagents were purchased from Sigma-Aldrich unless otherwise stated. All ligands tested in cell assays and radioligand binding assays were completed in a final DMSO concentration of 0.1% and 1%, respectively, unless otherwise stated.

OPEN IN VIEWER

Figure 1.

The chemical structures of the arginyl-glycinyl-aspartic acid (RGD) mimetics used in this study (A) and the human αvβ6 integrin saturation (B), association, and dissociation binding (C) generated for [³H]compound A. For saturation studies, specific binding of [³H]compound A was measured by incubation of increasing concentrations of radioligand with the human αvβ6 integrin in the presence of either vehicle (1% DMSO) or 10 µM SC-68448. For association studies, specific binding of [³H]compound A was measured by incubation of radioligand (~ K_D and 5 × K_D [not shown]) with the human αvβ6 integrin over time in the presence of either vehicle (1% DMSO) or 10 µM SC-68448. For dissociation studies, soluble human αvβ6 integrin protein was preincubated for 1 h with a fixed concentration of radioligand (~ K_D) before dissociation was initiated by addition of 10 µM SC-68448. Plates were then filtered after a 24-h incubation for saturation or the time points indicated for association and dissociation studies, and the amount of radioligand bound to αvβ6 was measured by liquid scintillation spectroscopy. Specific binding was measured by subtracting the nonspecific binding (10 µM SC-68448) from the total radioligand binding in the presence of vehicle (1% DMSO) for saturation and association binding studies. Specific saturation binding data were fitted to a one affinity site model with Hill slope (see Data Analysis section). Association and dissociation were fitted to a one-phase association or dissociation model, respectively. Data shown are the mean ± SD of duplicate points and are representative of at least four individual experiments with similar results. Inset A: Scatchard transformation ²⁸ of specific binding with the straight-line fit constrained to the mean B_max and K_D values from nonlinear regression using the one affinity site model with Hill slope (where x = 0, y = B_max/K_D and y = 0, x = K_D).

Data Analysis

Analysis of all experiments was completed using Prism 5.0 (GraphPad Software, San Diego, CA). Specific binding data from association binding experiments were fitted to a one-phase association model (equation 1, where Binding is the level of binding observed in test well and k_ob is the observed rate constant in inverse units of time [min^–1]).

Binding = Binding t = 0 + ( Binding plateau − Binding t = 0 ) ( 1 − exp ( − k o b . min ) ) .

To calculate the association rate constant k_on (in units of inverse molar inverse time [M^–1 · min^–1]) from K_ob, equation 2 was used (where k_off is the dissociation rate constant in inverse units of time [min^–1]).

K o b = [ radioligand ] ⋅ k on + k off

Dissociation binding data for [³H]compound A (experiments individually fitted) and unlabeled integrin ligands (experiments globally fitted) were fitted to a one-phase dissociation model (equation 3, where k_off is the dissociation rate constant in inverse units of time [min^–1] and NSB _t=∞ is nonspecific binding determined at infinite time).

Binding = ( Binding t = 0 − Binding plateau ) . exp ( − k o f f ⋅ min ) + NSB t = ∞ .

From equation 3, k_off values were subsequently used to calculate dissociation t_1/2 values using the equation t_1/2 = 0.693/k_off. Specific binding data from saturation experiments were fitted to a one affinity site model with Hill slope (equation 4, where n_H = Hill slope coefficient) to determine K_D values. For visualization only, specific binding data from saturation experiments were also analyzed via Scatchard transformation. ²⁸

Binding = B max [ radioligand ] n H K D n H + [ radioligand ] n H .

All CRCs and competition binding displacement curves were fitted using nonlinear regression analysis (four-parameter logistic equation with variable slope ²⁹ ) with IC₅₀ values calculated from the fits. IC₅₀ values generated from competition binding curves against [³H]compound A (where [³H]compound A saturation binding curve n_H = 1) were converted to equilibrium dissociation constant (K_I) values using the Cheng-Prusoff equation ¹⁵ (equation 5, where L^* is the radioligand concentration).

K I = I C 50 1 + [ L * ] K D .

All statistical analyses were completed using Prism 5.0 (GraphPad Software), and differences of p < 0.05 were considered statistically significant. Statistical significance between two data sets was tested using a Student unpaired t test. One-way analysis of variance (ANOVA) was used for comparison of more than two data sets and, where significance was observed, an appropriate post-test completed. For determination of statistical difference between a single data set and a hypothetical value, a one-sample t test was carried out. Unless otherwise indicated, data shown graphically and in the text are either mean ± standard deviation (SD) or, where three or more data points/individual experiments have been completed, mean ± standard error of the mean (SEM).

Results

Characterization of [³H]compound A αvβ6 Integrin Binding

[³H]compound A saturation binding studies were carried out with the αvβ6 integrin to characterize the mode of binding and determine its affinity for the RGD binding site. Specific binding data from saturation experiments with [³H]compound A were fitted to a one affinity site model (with Hill slope [n_H]), and analysis resulted in the pK_D and n_H values shown in Table 1 . The binding to αvβ6 was saturable and high affinity ( Fig. 1B ) with the n_H not significantly different from unity (one-sample t test completed on log₁₀ n_H vs. 0, p > 0.05), suggesting the binding of [³H]compound A to αvβ6 followed the law of mass action at a single site. Saturation binding studies with [³H]compound A against αvβ1, αvβ3, αvβ5, and αvβ8 were also completed ( Table 1 ). Saturable binding of [³H]compound A with these integrins was also observed with n_H values not significantly different from unity (one-sample t test completed on log₁₀ n_H vs. 0, p > 0.05), again suggesting the binding of [³H]compound A to these integrins also followed the law of mass action at a single site. The affinity of [³H]compound A for αvβ3 and αvβ6 was comparable, as was its affinity for αvβ1, αvβ5, and αvβ8. [³H]compound A exhibited an ~9- to 20-fold higher affinity for αvβ3 and αvβ6 compared with αvβ1, αvβ5, and αvβ8 ( Table 1 ).

OPEN IN VIEWER

Table 1.

Binding Affinity of Compound A and [³H]compound A for the αv Integrins.

αv Integrin	[3H]compound A pKD	[3H]compound A nH	Compound A Radioligand pKI	Compound A Cell Adhesion pIC50
αvβ1	7.98 ± 0.12	1.59 (0.86, 2.32)	7.14 ± 0.02	6.40 ± 0.16
αvβ3	9.17 ± 0.15	1.09 (0.25, 1.92)	8.75 ± 0.04	7.20 ± 0.11
αvβ5	7.88 ± 0.15	1.00 (0.41, 1.58)	7.64 ± 0.03	7.10 ± 0.20
αvβ6	8.94 ± 0.07	1.00 (0.73, 1.26)	8.62 ± 0.03	6.80 ± 0.34
αvβ8	7.96 ± 0.13	1.26 (0.79, 1.72)	7.81 ± 0.04	7.40 ± 0.18

[³H]compound A specific saturation binding data were fitted to a one affinity site model with Hill slope to determine K_D (see Data Analysis section). Compound A competition binding data were fitted using nonlinear regression analysis (four-parameter logistic equation with variable slope ²⁹ ; see Data Analysis section). IC₅₀ values generated from competition binding curves were converted to K_I values using the Cheng-Prusoff equation ¹⁵ (see Data Analysis section). Data shown are mean values ± SEM for at least three individual determinations. pK_D, negative log₁₀ of K_D; pK_I, negative log₁₀ of K_I; pIC₅₀, negative log₁₀ of IC₅₀; n_H, Hill slope.

In kinetic binding studies, the association of [³H]compound A to the αvβ6 integrin was measured at ~K_D and 5 × K_D concentrations of radioligand. Global fitting of the association kinetic model to specific association binding data resulted in a k_on value of 9.7 ± 1.1 × 10⁷ M^–1 · min^–1 for [³H]compound A (mean ± SEM, n = 4). In addition to determining an accurate k_on value for [³H]compound A, the time it took to reach a plateau at these concentrations of radioligand was required to determine the optimum radioligand/integrin preincubation period in dissociation binding studies. For [³H]compound A, the plateau (i.e., when equilibrium or steady state) was achieved in 45 min when using a K_D concentration of radioligand ( Fig. 1C ). Therefore, the preincubation time used in dissociation studies was 1 h to ensure that equilibrium binding with the αvβ6 integrin had been achieved in the system at radioligand concentrations > K_D. To measure a dissociation t_1/2 value for [³H]compound A, dissociation was initiated via the addition of 10 µM SC-68448. A single-phase dissociation of [³H]compound A binding from the αvβ6 integrin was observed with complete dissociation achieved between 2 and 3 h ( Fig. 1C ). A t_1/2 value for [³H]compound A of 15.9 ± 5.0 min (mean ± SEM, n = 6) was observed.

Comparison of Cell Adhesion and Radioligand Binding Assays for Profiling Integrin Small-Molecule RGD Mimetics

To compare the rank order of activity of a range of novel small-molecule RGD mimetics between cell adhesion and radioligand binding assays for the αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 integrins, pIC₅₀ values generated in cell adhesion (as an empirical measure of competitor potency) were correlated with pK_I values (true affinity measure) ( Fig. 2 ). In addition, the average fold shift between pIC₅₀ and pK_I values within each integrin comparison was also calculated (fold shift = pK_I – pIC₅₀). The average fold shifts observed for αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 were 1.5 (0.7 to 2.2), 1.6 (1.3 to 1.8), 1.0 (0.5 to 1.8), 2.0 (1.4 to 2.8), and 0.5 (–0.3 to 1.2), respectively (mean with minimum and maximum values shown in parentheses, n = 16–47). The straight lines that corresponded to the average fold shifts for each integrin are shown as the solid lines on the individual correlations in Figure 2 . The largest average fold shift difference was shown with αvβ6 followed by αvβ3, αvβ1, αvβ5, and αvβ8. αvβ8 displayed the closest agreement between cell adhesion and radioligand binding, and this was the only integrin that had any values on or to the left of the x = y line (shown as dotted line Fig. 2E ). A similar trend was observed for αvβ6 when comparing [³H]compound A pK_D in saturation binding studies, or the pK_I generated using the unlabeled form in competition binding studies (compound A vs [³H]compound A) against the αvβ6 cell adhesion assay ( Table 1 ).

OPEN IN VIEWER

Figure 2.

Correlation of pIC₅₀ and pK_I values generated with arginyl-glycinyl-aspartic acid (RGD) ligands from cell adhesion or radioligand competition binding (against [³H]compound A) studies, respectively, with the αvβ1 (A), αvβ3 (B), αvβ6 (C), αvβ5 (D), or αvβ8 (E) integrin. Full concentration response curves in cell adhesion assays were generated by incubating RGD ligands at a range of concentrations with A549 cells (αvβ1) or K562 cells recombinantly expressing the αvβ3, αvβ5, αvβ6, or αvβ8 integrin in plates coated with endogenous integrin ligand. Plates were incubated for 30 min prior to removal of nonadherent cell populations, and the remaining adherent cells were then quantified by fluorescence using 2′,7′-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester (BCECF-AM). Full competition binding curves were generated by incubating small-molecule RGD ligands at a range of concentrations with the αvβ1, αvβ3, αvβ5, αvβ6, or αvβ8 soluble integrin protein and [³H]compound A. Plates were then filtered after a 2-h (αvβ5 and αvβ8) or 6-h (αvβ1, αvβ3, and αvβ6) incubation, and the amount of radioligand bound to integrin was measured by liquid scintillation spectroscopy. Cell adhesion concentration response curves and competition binding data were fitted using nonlinear regression analysis (four-parameter logistic equation with variable slope ²⁹ ) to generate IC₅₀ values. K_I values were calculated from IC₅₀ values (see Data Analysis section), and the pK_I values generated for each unlabeled integrin ligand against [³H]compound A were correlated for each integrin with the corresponding pIC₅₀ value generated from cell adhesion (dotted line shows the straight line x = y, solid line shows the average fold-shift of pIC₅₀ to pK_I for each integrin). pIC₅₀ and pK_I shown are the mean ± SD of at least two individual experiments carried out in singlicate.

Selectivity of Small-Molecule RGD Mimetics and Peptides for the αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 Integrins

To determine the selectivity prolife of a range of integrin ligands, competition binding curves were completed with [³H]compound A against αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 ( Fig. 3 ). The control small-molecule RGD mimetics SC-68448 and compound A were shown to be the least selective ligands ( Fig. 3 and Table 2 ). The ligand with the greatest selectivity was the tLAP₁ peptide for the αvβ6 integrin, with a minimum of ~3000-fold selectivity observed over αvβ1, αvβ3, αvβ5, and αvβ8. tLAP₁ pK_I values generated for αvβ1, αvβ3, αvβ5, and αvβ8 were comparable ( Table 2 ). Cilengitide was selective for the αvβ1, αvβ3, and αvβ5 integrins compared with αvβ6 and αvβ8 (up to 1 µM). Cilengitide demonstrated a comparable affinity for αvβ3 and αvβ5 with αvβ1 affinity ~9-fold and 19-fold less than αvβ5 and αvβ3, respectively. The novel small-molecule RGD mimetic compound B exhibited selectivity for αvβ6 over the other integrins tested by at least 93-fold.

OPEN IN VIEWER

Figure 3.

The selectivity profiles of arginyl-glycinyl-aspartic acid (RGD) ligands for the αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 integrins determined by competition binding studies. Full competition binding curves were generated by incubating unlabeled integrin ligand at a range of concentrations with the αvβ1, αvβ3, αvβ5, αvβ6, or αvβ8 soluble integrin protein and [³H]compound A. Plates were then filtered after a 2-h (αvβ5 and αvβ8) or 6-h (αvβ1, αvβ3, and αvβ6) incubation, and the amount of radioligand bound to integrin was measured by liquid scintillation spectroscopy. Total and nonspecific binding values were measured in the presence of vehicle (1% DMSO) and 10 µM SC-68448, respectively, and were used to calculate the percent inhibition of radioligand bound to each integrin tested. Competition binding data were fitted using nonlinear regression analysis (four-parameter logistic equation with variable slope ²⁹ ). Competition binding curves are shown for SC-68448 (A), tLAP₁ (B), cilengitide (C), compound A (D), and compound B (E) against αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8. Competition binding displacement curve data are the mean ± SEM of at least four individual experiments carried out in singlicate and were used to generate the pK_I values shown in Table 2 .

OPEN IN VIEWER

Table 2.

Selectivity Profiles of a Range of Integrin Small-Molecule RGD Mimetics and Peptides against the αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 Integrins Determined by Competition Binding Studies.

	pKI
αv Integrin	αvβ1	αvβ3	αvβ5	αvβ6	αvβ8
SC-68448	7.51 ± 0.09	8.89 ± 0.09	8.07 ± 0.08	8.56 ± 0.05	6.98 ± 0.04
Compound A	7.14 ± 0.02	8.75 ± 0.04	7.64 ± 0.03	8.62 ± 0.03	7.81 ± 0.04
tLAP1	5.40 ± 0.05	5.62 ± 0.09	4.77 ± 0.01	8.70 ± 0.07	5.07 ± 0.06
Cilengitide	7.10 ± 0.16	8.39 ± 0.24	8.04 ± 0.30	<6.35	<6.15
Compound B	8.61 ± 0.09	8.03 ± 0.05	8.48 ± 0.09	10.6 ± 0.04	8.45 ± 0.05

Competition binding data were fitted using nonlinear regression analysis (four-parameter logistic equation with variable slope ²⁹ ; see Data Analysis section). IC₅₀ values generated from competition binding curves against [³H]compound A were converted to K_I values using the Cheng-Prusoff equation ¹⁵ (see Data Analysis section). Data shown are mean values ± SEM for at least four individual determinations generated from the data shown in Figure 3 . pK_I, negative log₁₀ of K_I.

Dissociation Binding Kinetics of Unlabeled Ligands from the αvβ6 Integrin

To allow the ranking of unlabeled ligands in terms of rate of dissociation from the αvβ6 integrin, a novel methodology was developed where the association of [³H]compound A was used as a surrogate for the dissociation of an unlabeled ligand from a preformed unlabeled ligand/integrin complex. Using this method, the dissociation profiles observed for compound A (fast and full dissociation) and A20FMDV2 (slow and partial dissociation) closely matched those observed for the radiolabeled forms of these ligands ¹⁹ ( Fig. 1C ). SC-68448, compound A, and fLAP₁ all had similar fast dissociation profiles (t_1/2 values of 5.1, 9.3, and 14.2 min, respectively) while the profile of compound B was much slower (t_1/2 value of 121.3 min) in comparison ( Fig. 4 ).

OPEN IN VIEWER

Figure 4.

Dissociation binding kinetics of unlabeled small-molecule arginyl-glycinyl-aspartic acid (RGD) mimetics and peptides from human αvβ6 integrin. Dissociation of unlabeled ligands from the soluble αvβ6 integrin protein was initiated (following a 1-h preincubation with integrin) by addition of 50 × K_D (~50 nM) [³H]compound A. Plates were then filtered after the time indicated, and the amount of radioligand bound to αvβ6 was measured by liquid scintillation spectroscopy. Total and nonspecific binding values were measured at each time point in the presence of vehicle (1% DMSO) and 10 µM SC-68448, respectively, and were used to calculate the percent inhibition of radioligand bound to the αvβ6 integrin. Data shown are the mean ± SEM of at least four individual experiments carried out in quadruplicate.

Discussion

Compound A is a small-molecule compound that was discovered as part of a drug discovery structure-activity relationship (SAR) screening campaign to identify αvβ6 integrin RGD mimetics. Screening data on compound A completed in cell adhesion assays had shown it to be a pan-αv molecule that inhibited the binding of αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 with comparable IC₅₀ values (all comparisons made between IC₅₀ values at these integrins resulted in 1.0- to 2.5-fold differences). As a result of this, compound A was chosen to be radiolabeled with [³H] to allow an in-depth characterization of the binding profile of a small-molecule RGD mimetic with αvβ6 and other RGD-directed integrins. Data presented in this study have shown that [³H]compound A binds to the αvβ6 RGD binding site with high affinity, which has been confirmed by the lack of binding to αvβ6 in the presence of high concentrations of RGD peptides (tLAP₁ and A20FMDV2). The mode of binding of [³H]compound A with the αvβ6 RGD binding site has been shown to be a simple, competitive, and reversible interaction. Association with the αvβ6 RGD binding site is relatively fast in terms of an association rate constant for a ligand-protein/receptor interaction. Following a preincubation with integrin, to allow the radioligand-integrin complex to form, the radioligand can be fully competed off the integrin by a small-molecule RGD mimetic (SC-68448). In addition, it has been shown that [³H]compound A binds with high affinity to the αvβ3 integrin and with moderate affinity to αvβ1, αvβ5, and αvβ8. The interesting observation from the binding studies is that the selectivity profile for compound A is much different compared with that defined in the cell adhesion studies. In cell adhesion, compound A has a comparable potency between integrins, whereas in binding studies with [³H]compound A, there is a clear difference between αvβ3 and αvβ6 compared with αvβ5 and αvβ8. This highlights the pitfalls of using a panel of cell adhesion assays to define accurate selectivity profiles of unlabeled integrin ligands.

The use of radioligand binding assays to determine affinity with soluble protein preparations for the integrins αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 has allowed accurate K_I values to be calculated by incorporating the K_D and the concentration of radioligand ([³H]compound A) present in each competition binding assay, allowing the use of the Cheng-Prusoff equation. ¹⁵ The comparison of a radioligand binding assay with a cell adhesion assay format for αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 integrins has not only allowed a direct ligand-integrin affinity to be measured in a simple system redundant of the full cell architecture but, more important, has also allowed a direct comparison between integrins to estimate accurate selectivity profiles of ligands. The shifts observed to higher affinity measurements from cell adhesion to radioligand binding assay are unlikely to be solely due to the Cheng-Prusoff correction, as the radioligand format uses a soluble integrin receptor compared with a whole cell in cell adhesion studies. Consequently, the more complex cell system could result in a lower empirical measure of competitor potency as a result of increased nonspecific binding of unlabeled ligands to cellular proteins or loss of ligand due to cellular uptake mechanisms. As the background cell line (K562) used as the expression system for the integrins is kept constant (except for αvβ1), it could be hypothesized that these mechanisms would be comparable and affect each individual integrin cell adhesion assay to the same degree in terms of reducing the pIC₅₀ value observed. As the fold shifts between cell adhesion and radioligand binding are different between integrins, this suggests that the difference is likely to be due to the type of endogenous ligand used and/or its affinity for the individual integrins. If αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8 all had the same affinity for an endogenous ligand that could be used in all cell adhesion assays at the same concentration, comparisons between integrins from cell adhesion assays could be used to determine an accurate selectivity profile for a ligand. However, this is not the case in the panel of cell adhesion assays shown here where either the affinity of endogenous ligand varies (αvβ1, αvβ3, αvβ6, and αvβ8 with tLAP₁) or a different endogenous ligand is used (vitronectin is used for αvβ5 due to the lower affinity LAP₁ displays for this integrin). SC-68448 is a small-molecule RGD mimetic that was developed to interact with the αvβ3 integrin as a potential inhibitor of tumor angiogenesis and growth. ²⁷ The only data available for this molecule show it profiled against the αvβ3 and αIIbβ3 RGD integrins, where the chemistry efforts were focused on enhancing potency at αvβ3 and selectivity over αIIbβ3 alone. ²⁷ However, in this study, SC-68448 has been shown to have pan-activity across αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8, and as far as the authors are aware, this is the first time this has been demonstrated for this molecule. As such, it has been proven to be a useful tool for the investigation of the pharmacology of αv integrins in this study, as well as highlighting its potential for future use.

As part of a medicinal chemistry lead optimization campaign to identify novel, high-affinity, and selective αvβ6 small-molecule RGD mimetics, compound B was selected as being one of the early optimal chemical entities identified. In radioligand binding assays, compound B demonstrated a very high affinity (low pM) and selectivity for the αvβ6 integrin (minimum profile >93-fold selective over αvβ1, αvβ3, αvβ5, and αvβ8), as well as a slow dissociation profile. To generate these data, a number of different experimental designs were investigated to enable optimal profiling of the characteristics of compound B. From the correlation of cell adhesion pIC₅₀ and radioligand binding pK_I values, there appeared to be an upper limit of detection in both assays (~8.5 pIC₅₀ and ~10.5 pK_I) at which compound B was either close to or at in both formats. In enzymology, a tight binding limit can be observed when the concentration of an enzyme inhibitor required to cause inhibition is close to the concentration of the enzyme in the system. ³⁰ To investigate if a similar tight binding limit was being observed with the αvβ6 integrin in radioligand competition binding experiments, the concentration of integrin was lowered to observe if a shift in affinity occurred. To control this experiment, a low-affinity control (SC-68448) was included, and when the concentration of soluble αvβ6 protein was lowered from 0.3 to 0.075 nM (4-fold lower), no change in the affinity was observed. This would be expected for a low-affinity compound that was not at the tight binding limit. However, for compound B, a shift was observed that was shown to be significant.

When the law of mass action is applied to an interaction between a ligand and its protein/receptor binding partner, it has been shown that the ligand’s affinity (K_D or K_I) is a combination of the ligand’s kinetic constants (k_off and k_on). ¹⁸ At steady state, the K_D (or K_I in an unlabeled ligands case) is defined by the ratio of the k_off and k_on and generally for membrane-based protein targets is a measure of the likelihood of the ligand-receptor complex to dissociate. Therefore, a high-affinity ligand would be predicted to have a lower k_off value and a slower dissociation from the protein/receptor. To test this theory for the αvβ6 integrin and also endeavor to obtain an early estimate of unlabeled ligand dissociation kinetics, a novel radioligand binding assay format was developed. Unlabeled ligands were allowed to form a complex with the αvβ6 integrin for 1 h prior to being competed off the integrin with an excess (50 × K_D) of radioligand ([³H]compound A), using the association of the radioligand to measure the dissociation of the unlabeled ligand. Using this method, low-affinity ligands for αvβ6 were shown to dissociate quickly (SC-68448, fLAP₁, compound A) and a high-affinity ligand slowly (compound B). This assay format was validated by having compound A and A20FMDV2 characterized as controls. Identical dissociation profiles were observed via this method for compound A and A20FMDV2 compared with the classical methods used for the radiolabeled forms of these ligands. ¹⁹ The highest affinity ligand compound B was shown to have the slowest dissociation profile from the αvβ6 integrin, and full dissociation was not observed up to 48 h.

When endeavoring to demonstrate an accurate selectivity between integrins, it is imperative to ensure the same divalent metal cations are used between assays and that the cation used is physiologically relevant. The level of binding and affinity observed for a ligand at an αv integrin will depend upon the cation present,^20,31 and some nonphysiological cations like Mn²⁺ will activate the integrin into a high-affinity state, resulting in an overestimate of RGD-ligand affinity. There are examples where nonphysiological cations have been used when comparing selectivity between the αv integrins.^32,33 This is acceptable when looking to test a worst-case scenario in terms of lack of binding to an integrin. However, this can be misleading when affinity is demonstrated and then compared between integrins in the presence of high concentrations of nonphysiologically relevant cations (e.g., Mn²⁺). It is further exacerbated if the extent of the effect of the cation is potentially ligand and/or integrin specific.^19,20

In conclusion, as the radioligand binding assays used in this study use the same radioligand, whose affinity and concentration present in the system are precisely known, this technique offers the most accurate method for determining an affinity measurement for integrin ligands at a specific integrin, as well as providing the most accurate selectivity profile. Although the radioligand format is not amenable to the high throughput required for SAR screening, these radioligand binding studies have enabled the relative fold shifts between pIC₅₀ values generated in cell adhesion (as an empirical measure of competitor potency) with pK_I values (affinity measure) to be determined. Therefore, in SAR analysis using the high-throughput cell adhesion assays, these fold shifts can now be incorporated to predict an estimated affinity and also selectivity profile for larger sets of novel chemical templates. These have been used to drive lead optimization and deliver selective, high-affinity αvβ6 small-molecule RGD mimetics, with slow dissociation kinetics ²⁰ to enable duration of action, as potential therapeutics in IPF and other fibrotic diseases.