Corticotropin-releasing factor (CRF) receptor antagonists are under preclinical and clinical investigation for stress-related disorders. In this study the impact of receptor-ligand binding kinetics on CRF1 receptor antagonist pharmacology was investigated by measuring the association rate constant (k1), dissociation rate constant (k−1), and kinetically derived affinity at 37°C. Three aspects of antagonist pharmacology were reevaluated: comparative binding activity of advanced compounds, in vivo efficacy, and structure-activity relationships. Twelve lead compounds, with little previously noted difference of affinity, varied substantially in their kinetic binding activity with a 510-fold range of kinetically derived affinity (k−1/k1), 170-fold range of k−1, and 13-fold range of k1. The k−1 values indicated previous affinity measurements were not close to equilibrium, resulting in compression of the measured affinity range. Dissociation was exceptionally slow for three ligands (k−1 t1/2 of 1.6–7.2 h at 37°C). Differences of binding behavior were consistent with in vivo pharmacodynamics (suppression of adrenocorticotropin in adrenalectomized rats). Ligand concentration-effect relationships correlated with their kinetically derived affinity. Two ligands that dissociated slowly (53 and 130 min) produced prolonged suppression, whereas only transient suppression was observed with a more rapidly dissociating ligand (16 min). Investigating the structure-activity relationship indicated exceptionally low values of k1, approaching 100,000-fold less than the diffusion-limited rate. Retrospective interpretation of medicinal chemistry indicates optimizing specific elements of chemical structure overcame kinetic barriers in the association pathway, for example, constraint of the pendant aromatic orthogonal to the ligand core. Collectively, these findings demonstrate receptor binding kinetics provide new dimensions for understanding and potentially advancing the pharmacology of CRF1 receptor antagonists.
Physiological responses to stressful stimuli are mediated by corticotropin-releasing factor (CRF), a 41-amino acid peptide that acts on the pituitary to regulate the hypothalamic-pituitary axis and the central nervous system to modulate behavioral responses to stress (Bale and Vale, 2004). CRF activates the CRF1 receptor, a class B G protein-coupled receptor (GPCR) (Bale and Vale, 2004; Grigoriadis, 2005). Pathophysiological conditions can arise from dysregulation of the stress axis, including depression, post-traumatic stress disorder, and relapse in substance abuse. Consequently, CRF has received considerable attention in drug discovery for psychiatric disease. The search for tractable new mechanisms for depression treatment, beyond monoamine modulators, is stimulated by the fact that existing drugs do not work effectively in approximately 30% of patients (Grigoriadis, 2005).
Antagonism of the CRF1 receptor has been proposed for rebalancing a dysfunctional stress axis (Holsboer, 2000; Grigoriadis, 2005). Successful development of drug-like small-molecule antagonists of the CRF1 receptor (Fig. 1) required application of advanced pharmacology technology. Low-affinity lead compounds were identified in one of the early successes of high-throughput screening (Hodge et al., 1999). Screening was required because CRF is too large and complex to be used as a chemical starting point for small-molecule medicinal chemistry. The compounds are allosteric modulators of the CRF1 receptor (Hoare et al., 2003); allosteric inhibition solved the molecular weight problem, enabling a small molecule to inhibit binding of a peptide 10 times its size. CRF1 receptor antagonists were some of the first allosteric GPCR ligands to be tested clinically. In the first report of behavioral effects in humans, 5-[7-(dipropylamino)-2,5-dimethylpyrazolo[1,5-a]pyrimidin-3-yl]-N,N,4-trimethylpyridin-2-amine (NBI 30775; Fig. 1) significantly reduced Hamilton depression and anxiety scores in a small group of patients with exceptionally high baseline scores (Zobel et al., 2000). In larger trials, 3,6-dimethyl-4-(pentan-3-yloxy)-2-(2,4,6-trimethylphenoxy)pyridine (CP-316,311; Fig. 1) failed to demonstrate efficacy in the treatment of major depression (Binneman et al., 2008), and pexacerfont [N-[(2R)-butan-2-yl]-8-(6-methoxy-2-methylpyridin-3-yl)-2,7-dimethylpyrazolo[1,5-a][1,3,5]triazin-4-amine] (Fig. 1) did not demonstrate efficacy compared with placebo for the treatment of generalized anxiety disorder (Coric et al., 2010).
Ideally, a compound with the highest possible efficacy for antagonizing the CRF1 receptor would provide the best tool for testing the utility of this mechanism for treating psychiatric disorders. Designing receptor-ligand interactions to optimize therapeutic outcome might now be possible because of an explosion in our knowledge of and ability to measure receptor-ligand interactions and the subsequent modulation of receptor and cellular activity (see Kenakin, 2007 and accompanying reviews). A simple strategy for maximizing antagonist efficacy is to maximize receptor residence time, by slowing antagonist dissociation from the receptor (Copeland et al., 2006; Vauquelin and Van Liefde, 2006; Brinkerhoff et al., 2008; Tummino and Copeland, 2008). This strategy requires an understanding of the kinetics of receptor-ligand interaction, which has been evolving since the 1960s (Paton, 1966; Rocha, 1969). With respect to GPCRs, antagonists of the angiotensin II AT1 receptor exemplify the potential of slow dissociation to increase antagonist effect (Verheijen et al., 2004). The slowly dissociating antagonist candesartan produces a greater maximal antihypertensive effect than more rapidly dissociating antagonists such as losartan (Hansson, 2001). Candesartan also produces a longer duration of effect (Lacourcière and Asmar, 1999). Ligand binding kinetics on other GPCRs are being extensively investigated, including the H1 histamine receptor, for which binding kinetics have been interpreted in the context of receptor structure and changes of Gibbs free energy (Wittmann et al., 2011); the M3 muscarinic acetylcholine receptor, for which antagonist binding has been reevaluated kinetically (Dowling and Charlton, 2006) and agonist dissociation rate correlated with intrinsic activity (Sykes et al., 2009); and the μ-opioid receptor, for which slow buprenorphine dissociation contributes to the pharmacodynamics of the drug in humans (Yassen et al., 2006). Small-molecule binding kinetics are now being measured on the CRF1 receptor. In a recent article, residence time was shown to be the primary determinant of insurmountable antagonism in vitro and CRF1 receptor binding in vivo (Ramsey et al., 2011). Pharmacological screening methods have been described for optimizing the dissociation rate constant (Miller et al., 2011; Ramsey et al., 2011).
In this study, we first measured the association and dissociation rate constants of small-molecule binding to the CRF1 receptor. We discovered that the kinetics of binding differed substantially between lead molecules and association and dissociation were remarkably slow. We then investigated the extent to which the binding kinetics rationalize pharmacodynamic differences and structure-activity relationships.
Materials and Methods
The following compounds were synthesized by using published methods (see the following reviews for original references: Gilligan et al., 2000b; Kehne and De Lombaert, 2002; Gross et al., 2005; Li et al., 2005; Zorrilla and Koob, 2010): 5-chloro-4-N-(cyclopropylmethyl)-2-methyl-4-N-propyl-6-N-(2,4,6-trichlorophenyl)pyrimidine-4,6-diamine (NBI 27914), N-butyl-N-ethyl-2,5,6-trimethyl-7-(2,4,6-trimethylphenyl)-7H-pyrrolo[2,3-d]pyrimidin-4-amine (antalarmin), CP-316,311, 8-(2,4-dichlorophenyl)-N,N-bis(methoxymethyl)-2,7-dimethylpyrazolo[1,5-a][1,3,5]triazin-4-amine (DMP696), 3-(4-methoxy-2-methylphenyl)-2,5-dimethyl-N-(pentan-3-yl)pyrazolo[1,5-a]pyrimidin-7-amine (DMP904), NBI 30775, 3-(2,4-dichlorophenyl)-9-(heptan-4-yl)-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4(12),5,7-tetraene (NBI 34041), (10S)-9-(cyclopropylmethyl)-3-(2,4-dichlorophenyl)-10-ethyl-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4(12),5,7-tetraene (NBI 35965), 5-(4-chloro-2-methoxyphenyl)-N-[(1S)-1-(4-fluorophenyl)pentyl]-1-methyl-N-propyl-1H-1,2,4-triazol-3-amine (NBI 46200), 10-(2-chloro-4-methoxyphenyl)-11-methyl-N-(pentan-3-yl)-1,8,12-triazatricyclo[7.3.0.03,7]dodeca-2,7,9,11-tetraen-2-amine (ONO-2333Ms), pexacerfont, and 4-(2-chloro-4-methoxy-5-methylphenyl)-N-[(1S)-2-cyclopropyl-1-(3-fluoro-4-methylphenyl)ethyl]-5-methyl-N-(prop-2-yn-1-yl)-1,3-thiazol-2-amine (SSR125543A).
Analogs of NBI 35965 and NBI 34041 were as described in Gross et al. (2005): 3-(2,4-dichlorophenyl)-6-methyl-9-(nonan-5-yl)-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4(12),5,7-tetraene (NBI 34416), 3-(2,4-dichlorophenyl)-6-methyl-9-(pentan-3-yl)-1,2,5,9-tetraazatricyclo [6.3.1.04,12]dodeca-2,4(12),5,7-tetraene (NBI 34417), 9-cyclohexyl-3-(2,4-dichlorophenyl)-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4(12),5,7-tetraene (NBI 34802), (10S)-3-(2-chloro-4-fluorophenyl)-9-(cyclopropylmethyl)-10-ethyl-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4,6,8(12)-tetraene (NBI 37606), (10S)-9-(cyclopropylmethyl)-10-ethyl-3-(4-fluorophenyl)-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4,6,8(12)-tetraene (NBI 37608), 3-(2,4-dichlorophenyl)-9-(heptan-4-yl)-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4,6,8(12)-tetraene (NBI 49721), and 3-(2,4-dichlorophenyl)-6-methyl-9-(propan-2-yl)-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4,6,8(12)-tetraene (NBI 78194).
[3H]NBI 35965 was prepared as described previously (Gross et al., 2005). [3H]NBI 30775 was prepared by using the same method, using the 6-bromo pyrazolopyrimidine intermediate. 125I-[Tyr0]sauvagine was from PerkinElmer Life and Analytical Sciences (Waltham, MA). Dulbecco's phosphate-buffered saline (DPBS) and cell culture supplies were from Invitrogen (Carlsbad, CA). Fetal bovine serum was from Thermo Fisher Scientific, Waltham, MA. All other reagents were from Sigma-Aldrich (St. Louis, MO). Low-binding 96-well plates were from Corning (Palo Alto, CA). Unifilter GF/C plates and Microscint 20 scintillation fluid were from PerkinElmer Life and Analytical Sciences.
Radioligand Binding Assays Conditions.
Binding experiments were performed in low-binding 96-well plates in assay buffer, comprised of DPBS (1.5 mM KH2PO4, 8.1 mM Na2HPO4, 2.7 mM KCl, and 138 mM NaCl), supplemented with 10 mM MgCl2 and 2 mM ethylene glycol-bis[β-aminoethyl]-N,N,N′,N′-tetraacetic acid, set to pH 7.4 with NaOH. All binding assays using [3H]NBI 35965 or [3H]NBI 30775 included 10 μM GTPγS to uncouple receptor from G protein and protease inhibitors [1:1000 dilution of protease inhibitor cocktail P8340 from Sigma-Aldrich, final assay concentrations of 0.1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 850 nM aprotonin, 422 μM bestatin hydrochloride, 1.4 μM N-(trans-epoxysuccinyl)-l-leucine 4-guanidinobutylamide, 1.9 μM leupeptin hemisulfate, and 1.53 μM pepstatin A]. Nonspecific binding in every experiment was defined by the addition of 4 μM NBI 34041. All compound concentrations listed within represent final concentrations in the assay. Specific binding for each concentration of radioligand at each time point was determined by subtracting nonspecific binding from total binding. The cell membranes used were isolated from human embryonic kidney 293 Flp-In cells (Invitrogen) stably expressing the CRF1 receptor at 62 pmol/mg membrane protein (Hoare et al., 2004). Membranes were isolated by high-pressure nitrogen cavitation and differential centrifugation as described previously (Hoare et al., 2003). The assay mixture (total volume of 200 μl) was incubated for the times and temperatures indicated on a temperature-controlled reaction block designed to fit the 16 round-bottom 96-well plates used (J-KEM, St. Louis, MO; see http://www.jkem.com/crbhof.html). At the end of the incubation period bound and free radioligand were separated by rapid vacuum filtration onto Unifilter GF/C filter plates. GF/C filter plates were pretreated with 0.5% polyethylenimine in distilled water for 30 min and prerinsed with 200 μl per well 1% bovine serum albumin in DPBS immediately before harvesting of the assay plate by using a cell harvester (UniFilter-96 Filtermate; PerkinElmer Life and Analytical Sciences). Filters were then washed two times with 400 μl of DPBS. Filter plates were dried, 50 μl of Microscint 20 was added, and the plate was monitored for radioactivity by using a TopCount NXT (PerkinElmer Life and Analytical Sciences) at 30% efficiency. The total amount of radioligand added to the assay was measured by using a 1600TR liquid scintillation counter (PerkinElmer Life and Analytical Sciences) at 47% efficiency.
Radioligand Dissociation Assays.
Radioligand dissociation experiments were performed by preincubating cell membranes with radioligand for 2 h. The target concentration of radioligand, temperature, amount of membrane protein, and dissociation time period used were: 10 nM [3H]NBI 35965 at 22°C, 4 μg of membrane protein, and 5 h; 10 nM [3H]NBI 30775 at 22°C, 9.5 μg of membrane protein, and 5 h; 3 nM [3H]NBI 35965 at 37°C, 5 μg of membrane protein, and 7 h; and 3 nM [3H]NBI 30775 at 37°C, 5 μg of membrane protein, and 7 h. Dissociation was initiated by the addition of 4 μM NBI 34041. Bound radioligand was harvested at 12 time points by using the method described above. Nonspecific binding was measured by including 4 μM NBI 34041 in the preincubation phase of the experiment. In each experiment, one duplicate set of wells did not receive unlabeled compound in the dissociation phase of the experiment to measure the stability of total radioligand binding over time. Total radioligand binding remained stable over the time period measured (5 h at 22°C and 7 h 37°C; data not shown).
Radioligand Association Assays.
Radioligand association experiments were initiated by the addition of cell membranes to wells containing radioligand. For [3H]NBI 35965 at 22°C six concentrations were tested ranging from 450 pM to 20 nM (n = 3) using 4 μg of membrane protein, and at 37°C a single concentration of approximately 3 nM (n = 2) was tested. For [3H]NBI 30775 at 22°C six concentrations were tested ranging from 1 to 20 nM (n = 3) using 8 μg of membrane protein, and at 37°C a single concentration of approximately 3 nM (n = 2) was tested. For all assays at 37°C, before addition of membrane, plates and membrane solution were warmed to 37° by using the heat block and a water bath, respectively. Bound radioligand was harvested as described above at 12 time points between 2 min and 2.5 h. The association experiments to determine the receptor kinetic rate constants of the unlabeled ligands were performed in the same manner using approximately 3 nM [3H]NBI 30775, either in the absence or the presence of a range of concentrations of unlabeled ligand. For lead compounds (Table 1), four or five concentrations of unlabeled ligand and 15 time points were used (ranging from 60 s to 4 h). For kinetic SAR (see Figs. 6⇓–8), two concentrations of unlabeled ligand and 16 time points were used (ranging from 20 s to 4 h).
Competition Binding Assays.
Unlabeled ligands were competed against 125I-sauvagine, in the absence of GTPγS, or [3H]NBI 35965, in the presence of 10 μM GTPγS, at 22°C for 2 h. Twelve concentrations of unlabeled ligand were tested, ranging from 10 μM to 31.6 pM by 3.16-fold serial dilution. The concentration of radioligand varied from 63 to 110 pM for 125I-sauvagine and 1.3 to 1.8 nM for [3H]NBI 35965. The amount of membrane protein in the assay was 2 μg per well for both radioligands.
Measurement of Adrenocorticotropin in Adrenalectomized Rats.
All animal studies were approved by the Institutional Animal Care and Use Committee at Neurocrine Biosciences. Rats were received at 175 to 200 g from Charles River Breeding Laboratories (Portage, MI) and housed in a 12-h light cycle for 1 week before adrenalectomy. Rats were anesthetized with isoflurane, and the whole adrenal capsule was plucked out of the intraperitoneal cavity. All external incisions were closed by using wound clips. The ability of adrenalectomized animals to maintain normal electrolyte levels was compromised, fat metabolism was altered, and glucose storage was impaired. Water containing 0.9% NaCl + 1.0% sucrose and regular rat chow were provided ad libitum. Their diet was also supplemented daily with two pellets of sweetened condensed milk chow (Research Diets, Inc., New Brunswick, NJ) and a 3-ml subcutaneous injection of lactated Ringer's solution. Adrenalectomy was verified by plasma corticosterone measurements. Seven days after surgery, blood was drawn from the tail vein, and corticosterone radioimmunoassay (MP Biomedicals, Solon, OH) was performed on the serum. Only rats with a corticosterone level below 10 ng/ml were used in the study. Seven days after adrenalectomy, rats were implanted with femoral vein catheters. Approximately 4 days later, rats were prepared for blood sampling by attaching their catheters to PE50 tubing and a syringe and acclimated to individual opaque sampling cages for 2 h. These cages allowed sampling to occur without disturbance to the rat. After a baseline blood sample (0.3 ml), rats received intravenous injection of vehicle (10% cremaphore) or test compound in vehicle. Blood samples were taken at the time points indicated in Fig. 5. Blood volumes were replaced with 5 U/ml heparinized saline. Blood samples were stored on ice with EDTA and plasma separated by centrifugation at 4°C and then stored at −80°C for subsequent measurement of adrenocorticotropin by radioimmunoassay (MP Biomedicals). The plasma compound concentration in these blood samples was measured as follows. Proteins in plasma samples were precipitated with 200 μl of acetonitrile, after adding 25 to 50 μl of internal standard. The organic layer was then isolated by centrifugation and dried. Following reconstitution with 30:70:0.1% acetonitrile/water/formic acid, the material was introduced into a SCIEX API-3000 liquid chromatography/tandem mass spectrometry system for analysis (ESPI+) (Danaher Corp., Washington, DC).
Data and Statistical Analysis.
All data were analyzed by using Prism 4.0 (GraphPad Software, Inc., San Diego, CA). Radioligand dissociation data (specific binding) were fit to monophasic and biphasic association equations, with the plateau of the specific binding set to zero, and the best fit was determined by using a partial F test. A monophasic model (eq. 1) fit best in all cases (p > 0.05): k−1(L) is the radioligand dissociation rate constant, t is time in minutes, and Yt = 0 is specific binding at the initiation of the dissociation phase of the assay. Radioligand association binding data (specific binding) were fit to monophasic and biphasic association equations, and the best fit was determined by using a partial F test. A monophasic model (eq. 2) fit best for all experiments (p > 0.05): where t is time in minutes, kobs is the observed association rate constant, Y is specific binding at time t, and Ymax is specific binding at infinite t. To determine the radioligand association rate constant tested at 22°C, where multiple concentrations of radioligand were used, kobs was plotted against the radioligand concentration, and the data were analyzed by linear regression by using eq. 3: where k1(L) is the radioligand association rate constant and [L] is the radioligand concentration. The fitted value of the slope yields k1(L). The Y intercept (k−1(L)) was fixed at the mean k−1(L) value measured from radioligand dissociation experiments (Table 1). To determine the radioligand association rate constant at 37°C where a single concentration of radioligand was tested, eq. 4 was used, with k−1(L) set to its directly measured value from the dissociation experiments: The association and dissociation rate constants of unlabeled ligands were measured by using the method of Motulsky and Mahan (1984) in which association of a radioligand is measured in the absence and presence of the unlabeled test ligand. The analysis assumes the model: where R is the receptor, L is the radioligand, k1(L) is the radioligand association rate constant, k−1(L) is the radioligand dissociation rate constant, I is the unlabeled ligand, k1(I) is the unlabeled ligand association rate constant and k−1(I) is the unlabeled ligand dissociation rate constant. Kinetic constants for the unlabeled ligands were determined by measuring the time course of association of [3H]NBI 30775 in the absence of unlabeled ligand and the presence of a range of unlabeled ligand concentrations. Specific radioligand binding (RL) was globally fit to a two-component exponential curve by nonlinear regression using the kinetics of competitive binding equation provided in Prism 4.0: where X is time (minutes), Y is specific radioligand binding (cpm), and Bmax is the total amount of receptor in the assay (cpm). Radioligand association data for the control (no unlabeled ligand) and for the presence of multiple concentrations of unlabeled ligand were fitted globally to this equation (Dowling and Charlton, 2006). The globally fitted parameters were Bmax, k1(L), k1(I), and k−1(I). The fixed parameters were k−1(L), [L], and [I]. The association rate constant of [3H]NBI 30775 (k1(L)) was fit for each experiment as an internal control, and k1(L) determined by the global fit closely matched the k1(L) value determined by direct measurement of [3H]NBI 30775 association alone (Table 1). In early experiments on the lead compounds (Table 2) four or five concentrations of compound were tested in each experiment and used in the global fitting. Subsequently, it was determined that fitted values with an equivalent S.E. within the fit could be obtained from using two concentrations of unlabeled ligand in the experiment. Consequently, most data for the kinetic SAR (see Figs. 6⇓–8) were obtained from assays using two concentrations of unlabeled ligand. The global r2 values from the fits ranged from 0.90 to 0.98.
Throughout this article, the association rate constant from either labeled or unlabeled ligands is denoted as k1 and the dissociation rate constant for labeled or unlabeled ligands is denoted as k−1. The half-life of drug dissociation from the receptor (t1/2), also equal to the median residence time, was calculated from the dissociation rate constant, k−1, by using the following equation: The kinetically derived affinity (kinetic Kd for radioligands and kinetic Ki for unlabeled ligands) was determined by using eq. 7:
Statistical Comparison of Rate Constant Data.
We compared the different groups of the structure-activity relationship data (see Figs. 6⇓–8) by using one-way analysis of variance, followed by the Newman-Keuls multiple comparison test to compare pairs of compounds. The results are given in the legends to Figs. 6 to 8. The logarithm of the k1 value was used in the test because it was assumed to be normally distributed. The linear value of k−1 has been shown to be normally distributed (Christopoulos, 1998). The log value but not the linear value of Kd (k−1/k1) is normally distributed, so it was assumed that the log value of the denominator, k1, was normally distributed.
Receptor Binding Kinetics of Radiolabeled CRF1 Receptor Antagonists.
Numerous nonpeptide antagonists have been developed that bind with high affinity to the CRF1 receptor, but the kinetics of their interaction with the receptor have not been systematically evaluated. We measured the kinetic parameters of these ligands' interaction with the CRF1 receptor, specifically the association rate constant k1, dissociation rate constant k−1, and kinetically derived affinity (Kd or Ki, equal to k−1/k1). We first examined the tritiated form of NBI 35965 ([3H]NBI 35965) and [3H]NBI 30775 (Fig. 1). The kinetics of binding were consistent with a single-site binding interaction with the CRF1 receptor; time-course data were fit best by monoexponential rate equations (Fig. 2), and the dependence of the observed association rate on radioligand concentration was linear up to 45 nM [3H]NBI 35965 and 20 nM [3H]NBI 30775 (data not shown). At 22°C, a temperature commonly used in CRF1 receptor binding assays, radioligand dissociation was markedly slow (Fig. 2A; Table 1). For [3H]NBI 35965 k−1 was 0.0023 min−1 (median residence time or k−1 t1/2, of 5 h). [3H]NBI 30775 dissociated too slowly to enable accurate measurement of k−1 because receptor occupancy had decreased by only 10% after 5 h (Fig. 2A). A similarly slow rate of dissociation of NBI 30775 at room temperature has been reported (Ramsey et al., 2011), and slow antagonist dissociation at room temperature has been observed for numerous GPCRs (Anthes et al., 2002; Fierens et al., 2002; Dowling and Charlton, 2006; Tummino and Copeland, 2008). In an attempt to better define the dissociation rate constant, the assay temperature was increased to 37°C. This increase markedly accelerated radioligand dissociation; the residence time of [3H]NBI 35965 decreased from 5 h to 17 min (Fig. 2A; Table 1). Dissociation of [3H]NBI 30775 was accelerated to a slow but measurable rate (k−1 t1/2 of 3.2 h; Fig. 2A; Table 1). A similar magnitude of k−1 increase was also observed for [3H]candesartan on the AT1 receptor (Fierens et al., 2002). Raising the incubation temperature to 37°C also accelerated radioligand association, 5.5-fold to 22 × 106 M−1min−1 for [3H]NBI 35965 and 2.8-fold to 31 × 106 M −1min−1 for [3H]NBI 30775 (Fig. 2B; Table 1). These rates of ligand association and dissociation at 37°C were unusually slow compared with antagonists of other GPCRs. The k1 values of 22 and 31 × 106 M−1min−1 are much less than the diffusion-limited rate constant [approximately 10,000 × 106 M−1min−1 (Copeland et al., 2006)]. These k1 values are also less than values reported for several antagonists of class A GPCRs, in the range of 1000 × 106 M−1min−1 (see references cited in Tummino and Copeland, 2008). In addition, dissociation of [3H]NBI 30775 from the CRF1 receptor was exceptionally slow, even at 37°C (t1/2 of 3.2 h).
Binding Kinetics of High-Affinity Unlabeled CRF1 Receptor Antagonists.
Slow antagonist dissociation can contribute to maximizing the duration and possibly the extent of in vivo efficacy of antagonist compounds (Copeland et al., 2006). Therefore, we determined whether other CRF1 receptor antagonists beyond NBI 30775 dissociated slowly. In addition, slow binding kinetics can distort measurements of binding affinity (Arányi and Quiroga, 1980; Motulsky and Mahan, 1984) owing to lack of equilibrium in the assay. Equilibrium is approximated in competition binding assays by a time interval at least 3-fold the residence time of the slowest-dissociating ligand in the assay (Motulsky and Mahan, 1984). For the CRF1 receptor, competition binding assays by us and other groups have typically been performed for 1 to 2 h at 22°C. Under these conditions NBI 35965 and NBI 30775 binding is not even close to equilibrium; an incubation time of at least 15 h would be required for NBI 35965 to reach equilibrium with the CRF1 receptor at 22°C (calculated from data in Table 1). Consequently, we re-evaluated the binding affinity of lead compounds by using the affinity derived from kinetic measurements (k−1/k1), a method for determining binding affinity that avoids distortion of affinity measurements resulting from lack of equilibration.
We used the method of Motulsky and Mahan (1984) to measure the binding kinetic constants of the antagonists, which determines the association and dissociation rate constants of an unlabeled compound from its effect on the association rate of a radioligand. Figure 3 shows representative data for four ligands with varying k−1 values. For unlabeled NBI 35965 and NBI 30775 (Fig. 3, B and C), the association and dissociation rate constants and kinetically derived affinity closely matched those of the radiolabeled versions of the compounds at both 22 and 37°C (Table 1), validating the method. To characterize binding kinetics of other compounds (Fig. 1) we chose the physiologic assay temperature of 37°C, instead of 22°C, to accelerate dissociation into a measurable range for slowly dissociating compounds. [3H]NBI 30775 was used as the radioligand because it provided the highest total binding/nonspecific binding ratio of the radioligands tested, and it dissociates slowly, enabling more accurate measurement of the k−1 value of slowly dissociating ligands (Motulsky and Mahan, 1984).
We evaluated the comparative kinetic pharmacology of 12 lead CRF1 receptor antagonists, selected based on progression to advanced preclinical or early clinical testing (Fig. 1). Compounds tested were CP-316,311, NBI 30775, NBI 35965, pexacerfont, antalarmin, DMP696, DMP904, NBI 27914, NBI 34041, NBI 46200, ONO-2333Ms, and SSR125543A.
Under standard in vitro assay conditions, competition against a radiolabeled peptide agonist at 22°C, all of these ligands have been described as high-affinity ligands, with little noted difference between their affinity for the CRF1 receptor (Li et al., 2005) (Fig. 4A). Measuring their binding kinetics revealed considerable differences of receptor-binding activity (Fig. 4; Table 2). The dissociation rate constant varied by 170-fold, from a k−1 t1/2 of 2.6 min for NBI 27914 to 7.2 h for SSR125543A (Fig. 4B; Table 2). Three ligands bound with a long residence time (k−1 t1/2 of >1 h; DMP904, NBI 30775, and SSR125543A; Figs. 3, C and D and 4B; Table 2). A long CRF1 receptor residence time has been reported for NBI 30775 and DMP-904 at room temperature (Ramsey et al., 2011). The association rate constant varied from 2.6 × 106 M−1min−1 for pexacerfont to 33 × 106 M−1min−1 for SSR125543A (Table 2), values much less than the diffusion-limited rate constant [approximately 10,000 × 106 M−1min−1 (Copeland et al., 2006)]. Remarkably, the kinetic Ki (k−1/k1) varied by 510-fold, from 49 pM for SSR125543A to 25 nM for NBI 27914 (Fig. 4A). This range was greater than the range of apparent affinity determined by using the previously used (traditional) binding assay conditions (competition against 125I-sauvagine at 22°C; 25-fold range from 0.3 nM for SSR125543A to 7.4 nM for pexacerfont; Fig. 4A; Table 3). We are currently investigating the difference of affinity further by using a direct measurement of ligand binding, frontal affinity chromatography coupled to mass spectrometry (Slon-Usakiewicz et al., 2005). In these experiments, membranes were immobilized by using artificial phospholipids coupled to a solid support. Ligands were continuously infused and affinity was measured as a function of the elution time, the ligand detected directly by using mass spectrometry. In our first experiments, the affinity of DMP696 was 23 nM, and for NBI 30775 it was 0.60 nM.
We investigated the reason for the discrepancy between the affinity range from kinetic measurements and traditional assay conditions. This difference was not caused by the different nature of the radioligands used (nonpeptide antagonist versus peptide agonist), because the apparent ligand affinity in competition against the nonpeptide antagonist [3H]NBI 35965 at 22°C was not significantly different from the apparent affinity measured by using 125I-sauvagine (Table 3). Next, we investigated the extent to which lack of equilibration at 22°C could affect the measurement of apparent affinity in a competition assay, using simulated data. We simulated inhibition of [3H]NBI 35965 binding by unlabeled ligands for 2 h at 22°C. Kinetics of [3H]NBI 35965 binding were from Table 1 and the dose was set at 1.5 nM. The unlabeled ligand k−1 was varied, and the k1 value was set at 10 × 106 M−1min−1. Under these conditions, as the residence time of the unlabeled ligand approached and then exceeded the 2-h incubation time, the apparent affinity approached a limit of approximately 0.5 nM such that compounds with markedly different k−1 and Ki values could no longer be distinguished on the basis of their apparent affinity. This affinity limit matched the apparent boundary of affinity measured experimentally (Fig. 4). We conclude that the lack of equilibration underlies the compressed affinity range determined from traditional binding assay conditions.
In Vivo Pharmacodynamics of CRF1 Receptor Antagonists with Varying Kinetic Binding Activity.
As described above, measuring binding kinetics revealed previously unappreciated differences of receptor binding pharmacology and, for some ligands, exceptionally slow dissociation of the receptor-ligand complex (Table 2; Fig. 4). We next investigated the extent to which this newly revealed in vitro binding activity translated to the in vivo pharmacology of the ligands. Quantitative pharmacodynamics of CRF1 receptor antagonism was assessed by measuring plasma adrenocorticotropin levels in adrenalectomized rats (Fig. 5). This model was used because it allows detection of sustained in vivo efficacy (Rivier et al., 1999) by using a quantitative biomarker (adrenocorticotropin).
Sustained, elevated adrenocorticotropin was observed in adrenalectomized rats out to 6 h after vehicle administration (Fig. 5A). The tonic adrenocorticotropin level was more than 1000 pg/ml (see legend to Fig. 5) compared with an average of 160 pg/ml in intact animals in our laboratory. This response was blocked by 100 μg of astressin, a nonselective peptide CRF1/CRF2 receptor antagonist, but not by 100 μg of asressin2B, a CRF2-selective peptide antagonist (data not shown), which is consistent with activation of the CRF1 receptor underlying the elevated adrenocorticotropin level and in agreement with previous studies (Rivier et al., 1999). The effects of nonpeptide antagonists were then evaluated on this prolonged CRF1 receptor-mediated response. We compared the PK/PD relationship of three ligands, NBI 30775, NBI 34041, and NBI 35965. The apparent affinity of the three ligands was similar when measured with original assay conditions (apparent Ki of 2.6, 1.4, and 1.1 nM for NBI 30775, NBI 34041, and NBI 35965, respectively; Table 3; Fig. 4A). The ligands were well differentiated by their dissociation rate constant and kinetically derived affinity (residence times of 2.2 h, 53 min, and 16 min for NBI 30775, NBI 34041, and NBI 35965, respectively, with corresponding Ki values of 0.36, 2.3, and 1.7 nM; Table 2; Fig. 4).
At the highest dose (10 mg/kg), all three ligands reduced adrenocorticotropin acutely (1 h postinjection; Fig. 5, A–C). After a longer duration a clear difference emerged between NBI 35965 and the other two ligands. The adrenocorticotropin level returned to the vehicle level by 2 h for NBI 35965 (Fig. 5C), whereas the response was sustained for 4 to 6 h for NBI 30775 and NBI 34041 (Fig. 5, A and B). Comparing the time course of compound clearance indicated that the difference between NBI 30775 and NBI 35965 could not be accounted for by pharmacokinetics; the pharmacokinetic profile of the compounds was very similar (Fig. 5D). The more sustained adrenocorticotropin suppression produced by NBI 30775 compared with NBI 35965 can be explained by prolonged occupancy of the CRF1 receptor (k−1 t1/2 of 2.2 h and compared with 16 min; Table 2; Fig. 4B). At the lower dose of 1 mg/kg, NBI 30775 and NBI 34041 sustainably reduced adrenocorticotropin by a slightly smaller increment, whereas NBI 35965 had no appreciable effect (Fig. 5, A–C). The PK/PD relationship was investigated in more detail by comparing the concentration-effect data of the ligands. The level of adrenocorticotropin reduction at the time of peak effect (1 h) was plotted against the plasma concentration of ligand at this time point (Fig. 5E). The rank order of potency was NBI 30775 > NBI 34041 = NBI 35965 (Fig. 5E), the same as that for binding affinity (0.36 nM for NBI 30775, 1.7 nM for NBI 34041, and 2.3 nM for NBI 35965; Table 2). Collectively, the comparative in vivo pharmacodynamics can be explained by the kinetics of antagonist binding to the CRF1 receptor; long sustained duration of action of NBI 30775, compared with NBI 35965, can be explained by the longer receptor residence time; and the higher in vivo potency of NBI 30775 was consistent with the higher receptor binding affinity of the ligand.
Kinetic Structure-Activity Relationships of CRF1 Receptor Antagonists.
We next used kinetic measurements to investigate the structure-activity relationships that ultimately resulted in the development of lead compounds from initial, lower-affinity chemical starting points (Arvanitis et al., 1999; Hodge et al., 1999). SAR studies have identified a number of key structural features of the ligands that result in high-affinity interaction with the CRF1 receptor (reviewed in Gilligan et al., 2000b; Kehne and De Lombaert, 2002) (diagrammatic representation in Fig. 9, adapted from the models developed in these studies). We assessed the contribution of the association and dissociation rate constants to the affinity-enhancing role of these elements of chemical structure. A graphical representation of these binding constants is used in Figs. 6, 7, and 8 to provide a visual tool for evaluating ligand SAR (see legend to Fig. 6 for details.) These features are evident from the common structural groups of the compounds in Fig. 1. A heterocyclic core bears a potential hydrogen-bond acceptor nitrogen atom. Attached to the core is a benzyl group or aromatic heterocycle (the lower or pendant aromatic group), separated from the core nitrogen by a one- or two-atom spacer. The aromatic group is oriented orthogonal to the plain of the core in the bioactive conformation (Hodge et al., 1999). On the opposite side of the core, N-, C-, or O-linked groups, often aliphatic, modify receptor binding affinity.
The optimal, orthogonal relationship between the lower aromatic group and the core is maintained by flanking substituents on these rings that enforce this twisted bioactive conformation (Hodge et al., 1999; Gilligan et al., 2000b; Kehne and De Lombaert, 2002). Within the tricyclic series we examined, these substituents are the ortho-position substituent on the lower ring (R1; Fig. 6) and the 4-position substituent on the core pyridine (R2; Fig. 7) (Gross et al., 2005). We first investigated the effect of the R1 substituent on the kinetics of binding, using NBI 37608 (R1 = H) and NBI 37606 (R1 = Cl) (Fig. 6). In the absence of a 2-position substituent, the association rate constant was very low (k1 of 0.048 × 106 M−1min−1 for NBI 37608; Fig. 6), indicating a substantial kinetic barrier to ligand association. Substitution at the ortho position with Cl reduced this barrier, increasing the association rate by 33-fold (to 1.6 × 106 M−1min−1; NBI 37606; Fig. 6). These kinetic data support the model developed previously in Fig. 9, that the binding pocket in the receptor for the lower aromatic ring is constrained relative to the binding site for the core. The pocket is too rigid to accommodate the freely rotating lower aromatic of NBI 37608, so successful collision is limited by the low probability that the group is rotated to the optimal orthogonal position by chance when the ligand encounters the receptor (Fig. 9), manifested as a low k1 value (Fig. 6). Constraining the lower aromatic in the orthogonal position (NBI 37606) optimizes the presentation of this group to the postulated rigid binding pocket on the receptor (Fig. 9), resulting in a greater frequency of successful collisions with the receptor, manifested as an increase of k1. The ortho-substituent did not significantly affect ligand dissociation (Fig. 6), suggesting it does not contribute to stabilizing the receptor-ligand complex once the ligand has associated with the receptor. In combination, the rate constants indicate the affinity-enhancing effect of the ortho-substituent was driven by an increase of k1.
We next examined the R2 substituent flanking the core pyridine (Fig. 7). This small substituent on the ligand core increases CRF1 receptor affinity in a variety of ligand structures (Arvanitis et al., 1999; Gilligan et al., 2000b; Kehne and De Lombaert, 2002). We investigated the kinetics of the affinity enhancement by using nonsubstituted NBI 49721 (R2 = H) and NBI 34041 (R2 = CH3) (Fig. 7). The k1 value of NBI 49721 was exceptionally low (0.12 × 106 M−1min−1; Fig. 7), implying a poorly configured structure for successful collision with the receptor. Substitution of H of NBI 49721 with the methyl group of NBI 34041 increased k1 69-fold, a similar magnitude to the effect of ortho-substitution of the lower aromatic (33-fold; Fig. 6). This increased association is consistent with the core methyl substituent stabilizing the orthogonal relationship of the lower aromatic group with respect to the core, increasing the probability of successful collision of this group with a rigid binding pocket on the receptor (Fig. 9). The core methyl substituent also significantly slowed dissociation (decreased k−1; Fig. 7), implying the methyl substituent stabilizes the receptor-ligand complex. This finding is consistent with the methyl group interacting with a hydrophobic pocket on the receptor (Fig. 9), likely small given the size constraint at this position in the ligand pharmacophore (Arvanitis et al., 1999; Hodge et al., 1999; Gilligan et al., 2000b; Kehne and De Lombaert, 2002). In combination, the rate constants indicate the affinity-enhancing effect of the core-flanking R2 substituent resulted from increasing k1 and decreasing k−1.
In early SAR studies ligand affinity was increased by the presence of branched alkyl or heteroalkyl chains on the opposite side of the core from the lower aromatic group. The effect on affinity of these upper aliphatic groups depended on their size, branching pattern and, where a chiral center was present, the stereochemical configuration (Gilligan et al., 2000a; Gross et al., 2005). We evaluated the kinetics underlying the affinity contribution of the upper aliphatic groups at the R3 position of the NBI 34041 scaffold (Fig. 8). A small, 3-carbon aliphatic group in this position (NBI 78194 of Fig. 8) bound weakly to the receptor (Ki of 200 nM), owing to a low value of k1 (0.12 × 106 M−1min−1) and rapid dissociation (k−1 t1/2 of 1 min). Increasing the chain length to five methylene units (NBI 34417 of Fig. 8) increased affinity 16-fold. Underlying this increase was an 18-fold increase of k1 and a 4-fold slowing of dissociation (Fig. 8). The former implies the addition of the methylene units reduces a kinetic barrier in the association pathway. The latter implies the longer alkyl chain increases the stability of the receptor-ligand complex, possibly by anchoring this region of the ligand within a hydrophobic binding pocket (Gross et al., 2005) (Fig. 9). Extension to seven methylene units (NBI 34041) slowed dissociation 3.0-fold, suggesting a slightly stronger stabilizing interaction, without appreciably affecting association (Fig. 8). Extending the chain to nine methylene units (NBI 34416 of Fig. 8) reduced affinity by decreasing k1 18-fold and accelerating dissociation 3.8-fold (Fig. 8), indicating a size constraint on the binding kinetics. Constraining the aliphatic group in a cyclohexyl ring (NBI 34802 of Fig. 8) decreased k1 and increased k−1, suggesting flexibility of the aliphatic region is necessary for optimal association with and dissociation from the receptor. It is noteworthy that the binding kinetics of the 3-propyl, 5-nonyl, and cyclohexyl analogs were similar (Fig. 8), suggesting optimal size and flexibility are required for successful collision of the upper aliphatic group with the receptor.
Overall, the kinetic SAR indicates that the improvement of affinity in the tricyclic series has resulted from an increase of k1 together with a decrease of k−1. In all three regions of the receptor we investigated, optimizing the groups increased k1. The association rate constant of even the most optimized compounds was low (highest k1 of 20 × 106 M−1min−1 for NBI 35965; Table 2) relative to the diffusion-limited rate constant [approximately 10,000 × 106 M−1min−1 (Copeland et al., 2006)]. The kinetic data reveal that the development of high-affinity ligands has involved increasing k1 approximately 100-fold. Given the exceptionally low k1 values of the starting points (100,000-fold less than diffusion), the remaining shortfall of k1 in high-affinity compounds is made up for by the unusually long residence times (up to 53 min for NBI 34041; Table 2).
The CRF system is the principal regulator of stress responses (Bale and Vale, 2004). Drug discovery has yielded CRF1 receptor antagonists as potential treatments for depression and other disorders of the stress axis (Holsboer, 2000; Grigoriadis, 2005). Testing the utility of this mechanism will be aided by designing antagonists with maximal efficacy for blocking the actions of CRF at the CRF1 receptor. Prolonging receptor residence time is one approach for achieving strong in vivo and clinical efficacy (Copeland et al., 2006; Vauquelin and Van Liefde, 2006; Brinkerhoff et al., 2008). This approach requires an understanding of the kinetics of receptor-ligand interaction. In this study we reevaluated the pharmacology of previously identified CRF1 antagonists in the context of their receptor binding kinetics.
Measuring binding kinetics of 12 lead compounds redefined their binding pharmacology (summarized in Fig. 4). No appreciable differences of affinity between the compounds have been noted previously but the kinetic affinity (k−1/k1) ranged 510-fold, from 49 pM for SSR125543A to 25 nM for NBI 27914 (Table 1). Long ligand residence times likely compressed the observed affinity range in previously used assays, which used relatively short incubations at room temperature. The wide range of affinity resulted in large part from the range of residence time, from 7.2 h for SSR125543A to 2.6 min for NBI 27914 (Fig. 4; Table 1). Three compounds bound with an unusually long residence time of greater than 1 h at 37°C (SSR125543A, NBI 30775, and DMP-904). Although slow ligand dissociation at room temperature has been reported for numerous GPCRs (Anthes et al., 2002; Fierens et al., 2002; Dowling and Charlton, 2006; Tummino and Copeland, 2008), including the CRF1 receptor (Miller et al., 2011; Ramsey et al., 2011), few studies reported k−1 at the physiologic temperature. In the most studied case for GPCRs, at 37°C candesartan dissociates from the AT1 receptor with a k−1 t1/2 of 58 min (Fierens et al., 2002).
The redefinition of lead compound binding activity modifies the interpretation of in vivo efficacy. For example, in a recent study (Ramsey et al., 2011), the in vivo EC50 for occupancy of the CRF1 receptor was correlated with the kinetically defined affinity. In the present study, newly revealed differences of binding activity explained differential compound pharmacodynamics for suppressing adrenocorticotropin in adrenalectomized rats. The rank order of efficacy (NBI 30775 > NBI 34041 = NBI 35965; Fig. 5E) was the same as the rank order of kinetically defined binding affinity (Fig. 4A; Table 2). Prolonged suppression of adrenocorticotropin was observed with NBI 30775 but not NBI 35965 (Fig. 5, A and C), despite their nearly identical pharmacokinetic profiles (Fig. 5D), consistent with the sustained antagonism of NBI 30775 resulting from its long residence time (Fig. 4B; Table 2) (Vauquelin and Van Liefde, 2006). It is worth noting that adrenocorticotropin provides a biomarker that can be translated from preclinical to clinical pharmacology.
Conceivably, the kinetic redefinition of CRF1 receptor-antagonist interaction might affect the interpretation of ligand efficacy in humans. Comparison of clinical efficacy between compounds will need to consider differences of accurately determined binding affinity. The kinetically derived affinity measurement unmasked differences between compounds tested clinically to date; NBI 30775 binds with 33-fold higher affinity than CP-316,311 and 53-fold higher affinity than pexacerfont (Fig. 4A; Table 2). With respect to the dissociation rate constant, in the study of Ramsey et al. (2011) k−1 was used in a PK/PD simulation to suggest appreciable occupancy (>50%) was not achieved in the human behavioral and endocrine trial of NBI 30775 (Zobel et al., 2000; Künzel et al., 2003; Ramsey et al., 2011). NBI 30775 did not significantly affect plasma adrenocorticotropin in these patients (Künzel et al., 2003). Comparison with other receptor systems suggests the residence time could affect human pharmacodynamics. At the AT1 receptor, the slowly dissociating antagonist candesartan produces a greater maximal antihypertensive effect than the more rapidly dissociating antagonist losartan (Hansson, 2001) and elicits a prolonged effect that endures sufficiently to tolerate a missed dose in humans (Lacourcière and Asmar, 1999). The prolonged pharmacodynamics of the μ-opioid receptor partial agonist buprenorphine, used for managing opiate withdrawal, might result in part from slow dissociation from the receptor (Yassen et al., 2006). A long residence time has been posited to explain the prolonged efficacy of the M3 antagonist tiotropium (Dowling and Charlton, 2006), which is used to treat chronic obstructive pulmonary disease. The long-acting antihistamine deslorataidine (the active metabolite of loratadine) dissociates slowly from the H1 receptor (Anthes et al., 2002). It is tempting to speculate that a long residence time of a CRF1 receptor antagonist might aid the detection of a measurable and/or prolonged effect in humans on endocrine biomarkers (e.g., adrenocorticotropin) or on scores of psychiatric dysfunction.
Application of binding kinetic information will aid medicinal chemistry strategies for the future optimization of CRF1 receptor antagonists. For most receptor-ligand interactions, the association rate is diffusion-limited, and ligand SAR is driven by changes in the dissociation rate (Tummino and Copeland, 2008). By contrast, for CRF1 receptor antagonists the association rate is limited by ligand interaction with the receptor. This conclusion is based on the k1 value being at least 300-fold lower than the rate of diffusion, k1 being highly dependent on the chemical structure of the ligand (Table 2; Figs. 6⇑–8), and the k1 SAR being consistent with previously described models of small-molecule interaction with the CRF1 receptor (Fig. 9; Gilligan et al., 2000b; Kehne and De Lombaert, 2002). More refined models, such as those being developed for the H1 receptor (Wittmann et al., 2011), will require structural information on the CRF1 receptor or a homologous class B GPCR. Collectively, the retrospective kinetic SAR indicates affinity-improving changes overcame kinetic barriers in the ligand association pathway (for example, restricting rotational freedom of the lower aromatic ring) and stabilized the receptor-ligand interaction (optimizing upper aliphatic chain length).
Unusual binding kinetic behavior can be a manifestation of a more complex binding mechanism than a simple one-step binding and dissociation process (mechanism A in Tummino and Copeland, 2008). In a two-step binding model (mechanism B), ligand associates with and dissociates from the receptor in an initial complex (defined by the rate constants k1 and k2), which then undergoes a transition to form a final complex (formation defined by k3, deformation by k4). A common manifestation of this model in kinetic data is a hyperbolic dependence of the observed association rate on the ligand concentration (Strickland et al., 1975; Tummino and Copeland, 2008). Here, the dependence was linear up to a radioligand concentration of 20 nM for [3H]NBI 30775 and 45 nM for [3H]NBI 359655, albeit at room temperature. According to simulations we have performed, the two-step model can apply to these data when the initial interaction is transient (large values of k1 and k2) and when k3 is >30-fold k4. Under these conditions, the final complex predominates, the observed dissociation rate constant is almost equal to k4, and k1, k2, and k3 define the observed association rate constant. Using these inferences, the two-step model can provide a simple mechanistic explanation for the kinetic SAR; ligand associates rapidly with the receptor (diffusion limited) to form an initial state, from which it also dissociates rapidly. This state then transitions into the final complex, with the structure of the ligand defining k3 and k4, and hence the observed association and dissociation rate constants. In a third model, mechanism C (Tummino and Copeland, 2008), the receptor transitions between two states, one that binds ligand, one that does not. For the CRF1 receptor as for all class B GPCRs, the large N-terminal domain could act as a gate that needs to open to allow access of small molecules to the membrane-spanning domain of the receptor, a process consistent with mechanism C. Deletion of the N-terminal domain (Hoare et al., 2004) did not appreciably affect k1 and k−1 of [3H]NBI 30775 (data not shown), suggesting this domain is not responsible for the unusual kinetics of small-molecule binding to the CRF1 receptor.
In summary, investigating CRF1 receptor binding kinetics redefined the pharmacology of CRF1 receptor antagonists. By overcoming the artificial compression of affinity measurement caused by the lack of equilibration, substantial differences of the binding behavior between lead compounds was revealed. These differences translated to differential pharmacodynamics in vivo. Very slow ligand dissociation was observed that could maximize compound efficacy and prolong the duration of efficacy. The molecular basis of ligand SAR in the evolution of these compounds involved spatial constraint of the ligand and potentially the binding pocket in the receptor and kinetic barriers in the association pathway that were lowered by optimizing ligand structure. Applying this new knowledge of CRF1 receptor-ligand interactions will aid future development of therapeutic ligands targeting this receptor.
Participated in research design: Fleck, Hoare, Bradbury, and Grigoriadis.
Conducted experiments: Fleck, Hoare, and Pick.
Contributed new reagents or analytic tools: Fleck, Pick, and Grigoriadis.
Performed data analysis: Fleck, Hoare, and Pick.
Wrote or contributed to the writing of the manuscript: Fleck, Hoare, and Grigoriadis.
We thank Jacek Slon-Usakiewicz for frontal affinity chromatography coupled to mass spectrometry data and John P. Williams for valuable guidance and discussion.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- corticotropin-releasing factor
- G protein-coupled receptor
- Dulbecco's phosphate-buffered saline
- guanosine 5′-O-(3-thio)triphosphate
- structure-activity relationship
- NBI 27914
- NBI 30775
- NBI 34041
- 3-(2,4-dichlorophenyl)-9-(heptan-4-yl)-6-methyl-1,2,5,9-tetraazatricyclo[6.3.1.04,12]dodeca-2,4(12), 5,7-tetraene
- NBI 34416
- NBI 34417
- NBI 34802
- NBI 35965
- NBI 37606
- NBI 37608
- NBI 46200
- NBI 49721
- NBI 78194
- Received October 4, 2011.
- Accepted January 23, 2012.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics