Agonist bias and agonist‐dependent antagonism at corticotrophin releasing factor receptors

Abstract The corticotropin‐releasing factor (CRF) receptors represent potential drug targets for the treatment of anxiety, stress, and other disorders. However, it is not known if endogenous CRF receptor agonists display biased signaling, how effective CRF receptor antagonists are at blocking different agonists and signaling pathways or how receptor activity‐modifying proteins (RAMPs) effect these processes. This study aimed to address this by investigating agonist and antagonist action at CRF1 and CRF2 receptors. We used CRF1 and CRF2 receptor transfected Cos7 cells to assess the ability of CRF and urocortin (UCN) peptides to activate cAMP, inositol monophosphate (IP1), and extracellular signal‐regulated kinase 1/2 signaling and determined the ability of antagonists to block agonist‐stimulated cAMP and IP1 accumulation. The ability of RAMPs to interact with CRF receptors was also examined. At the CRF1 receptor, CRF and UCN1 activated signaling in the same manner. However, at the CRF2 receptor, UCN1 and UCN2 displayed similar signaling profiles, whereas CRF and UCN3 displayed bias away from IP1 accumulation over cAMP. The antagonist potency was dependent on the receptor, agonist, and signaling pathway. CRF1 and CRF2 receptors had no effect on RAMP1 or RAMP2 surface expression. The presence of biased agonism and agonist‐dependent antagonism at the CRF receptors offers new avenues for developing drugs tailored to activate a specific signaling pathway or block a specific agonist. Our findings suggest that the already complex CRF receptor pharmacology may be underappreciated and requires further investigation.


| INTRODUC TI ON
The neuropeptide, corticotropin releasing factor (CRF) is a member of the secretin peptide family. 1,2 CRF is expressed throughout the central nervous system and in peripheral tissues. 3 CRF is closely related to the three urocortin (UCN) peptides, which share a similar structure, and have overlapping receptors and functions. 4 The expression of the three UCN peptides are less well-characterized than CRF, however, they are expressed at both overlapping and discrete sites within the CNS. 5,6 CRF is best characterized for its roles in stress and anxiety. 5 In the pituitary, CRF stimulates the release of adrenocorticotropic hormone (ACTH). ACTH subsequently elevates circulating glucocorticoid steroids, which allow an organism to respond to stressful situations. 5,7 Elevated CRF concentrations have also been observed in a number of psychiatric disorders, indicating that drugs targeting this system could have utility in treating stress and anxiety. 7 The UCN peptides have also been implicated in stress responses; potentially regulating the recovery from stressful stimuli and ACTH release from the pituitary. [8][9][10] The UCN peptides have additional functions centrally, including modulating social behaviors, and peripherally, such as the modulation of vasodilation, cardiac output, and the control of metabolism. [11][12][13][14][15] This suggests that drugs targeting the UCN axis could have added utility in treating other behavioral disorders, cardiovascular disease, obesity, and diabetes.
The CRF peptide family can bind two G protein-coupled receptors (GPCRs); CRF 1 and CRF 2 . The CRF 1 receptor binds CRF and UCN1 with high affinity, whereas the CRF 2 receptor can bind CRF, UCN1, UCN2, and UCN3 with high affinity. 16,17 The CRF receptors have already been exploited in drug discovery resulting in small molecule CRF 1 antagonists which have been explored in clinical trials to treat anxiety, depression and addiction. 18 Several CRF 1 receptor antagonists are reportedly safe and clinical trials are ongoing, whereas other CRF receptor antagonists have been discontinued due to a lack of efficacy. 18 The underlying basis for these differences is not well understood but it is possible that CRF receptor pharmacology is more complicated than presently appreciated. For example, both the CRF 1 and CRF 2 receptors have functional splice variants and have been reported to form heterodimers with a receptor activity-modifying protein (RAMP). [19][20][21] However, there are conflicting reports for RAMP interactions. 22 RAMPs can form heterodimers with several GPCRs altering cell surface expression, receptor trafficking, pharmacology, and/or signaling properties. 23 The activation or specific inhibition of a discrete signaling pathway can be associated with a biological event, driven by biased ligands. Such molecules can activate the same receptor, giving rise to a different pattern of signaling and potentially different biological outcomes. 1,24 This has led to considerable interest in exploring biased signaling for both endogenous ligands and in drug discovery, where the goal is to activate or block a specific pathway. 25,26 Exploiting the therapeutic potential of the UCN peptides also relies on a greater understanding of how these peptides trigger CRF receptor signaling.
In order to enable further exploitation of CRF receptor-mediated signaling pathways in drug discovery, we profiled several signaling pathways in response to the endogenous peptides CRF and the three UCN peptides. Moreover we investigated how effectively CRF receptor antagonists block CRF-or UCN1-stimulated receptor signaling.

| Plasmid constructs
Plasmids containing hCRF 1α and hCRF 2α (GenBank accession numbers AY457172 and AY449734) were obtained from the cDNA Resource Centre (Bloomsburg University) and are referred to as CRF 1 and CRF 2 in this manuscript. Hemagglutinin (HA) epitope-tagged human Calcitonin (CT) receptor (CT (a) splice variant) and myc-tagged human RAMP1 were used as described previously. 27 HA-tagged human calcitonin receptor-like receptor (CLR), N-terminal FLAG-tagged human RAMP3 and N-terminal myc-tagged human RAMP3 were a gift from Professor David Poyner (Aston University) and Professor Patrick Sexton (Monash University), respectively. N-terminal FLAG-tagged human RAMP2 was used as described previously. 28 All receptors and RAMPs were cloned in pcDNA3.0 or pcDNA3.1 plasmid vectors. All plasmid inserts were sequenced at the Centre for Genomics, Proteomics and Metabolomics (University of Auckland) and the sequences were verified prior to use.

| cAMP measurement in Cos7 cells
cAMP accumulation in Cos7 cells transfected with hCRF 1α or hCRF 2α receptors was measured using the LANCE cAMP detection assay kit (PerkinElmer Life and Analytical Sciences) as described previously. 30 Briefly, culture media was removed and replaced with DMEM containing 0.1% bovine albumin serum (BSA) and 1 mmol/L 3-isobutyl-1-methylxanthine (IBMX) for 30 minutes at 37°C. Cells were incubated for an additional 15 minutes at 37°C with media or agonist concentrations of 1 pmol/L to 1 µmol/L for agonist only assays or 10 pmol/L to 10 µmol/L concentrations for antagonist experiments. Antagonists were added simultaneously or immediately prior to the addition of agonists. To stop cell stimulation and extract cAMP, the contents of the wells were replaced with 50 µL absolute ethanol. Samples were left at −20°C for a minimum of 10 minutes.
The ethanol was evaporated and replaced with 50 µL of cAMP detection buffer. Samples were shaken for 15 minutes before 5 µL cell lysate was transferred to a white 384-well optiplate and processed for cAMP quantification as described previously. 31 Samples were read using an Envision plate reader (PerkinElmer). cAMP concentrations were determined from a standard curve generated in duplicate.

| Measurement of extracellular signal-regulated kinase 1/2 phosphorylation in Cos7 cells
The AlphaLISA Surefire Ultra extracellular signal-regulated kinase 1/2 (ERK1/2) phosphorylation (pERK1/2) assay kit (PerkinElmer Life and Analytical Sciences) was used to measure pERK1/2 in cell lysates after agonist stimulation. For these assays 20% FBS was used as a positive control. Following peptide stimulation at 37°C for 0-30 minutes with 100 nmol/L agonist (time courses) or 7 minutes with media alone or 1 pmol/L to 1 µmol/L agonist, media was removed and 20 µL of lysis buffer added to each well. Plates were shaken at room temperature for 10 minutes. Each sample was then transferred to a 384-well optiplate and pERK1/2 measurement performed as per the manufacturer's instructions. Samples were read using an Envision plate reader (PerkinElmer).

| IP 1 measurement in Cos7 and HEK-293S cells
The IP-one Gq assay kit (Cisbio) was used to quantify accumulated myo-inositol-1-phosphate (IP 1 ), a by-product of IP 3 produced after receptor-mediated Gαq activation in Cos7 and HEK-293S cells.
Briefly, culture media was removed and replaced with DMEM containing 0.1% BSA for 30 minutes at 37°C. Cells were then incubated with media containing 50 mmol/L lithium chloride (LiCl) for an additional 0-120 minutes (time courses) or 90 minutes at 37°C in the presence or absence of an antagonist. Agonist concentrations of 1 pmol/L to 10 µmol/L were used for agonist only assays or 10 pmol/L to 10 µmol/L concentrations for antagonist experiments.
After cell stimulation, 14 µL of stimulation buffer was added to each well. Three microlitres of each detection antibody were added in turn to the plate and incubated at room temperature for 1 hour. Fifteen microliters of sample was then transferred to a 384-well optiplate and read by an Envision plate reader (PerkinElmer). IP 1 concentrations were calculated from a standard curve generated in duplicate.

| Measurement of RAMP cell surface expression by ELISA
Cos7, HEK-293T or HEK-293S cells were plated into 96-well plates, transfected and assayed for cell-surface expression as described previously. 32 Briefly, transfected cells were fixed using 4% paraformaldehyde in PBS for 20 minutes, then washed twice with PBS. One hundred microliters of PBS was added to each well and the plates stored at 4°C until further analyzed. The PBS was aspirated, and the cells incubated at room temperature with 0.6% hydrogen peroxide in PBS for 20 minutes. Cells were washed twice in PBS and blocked with 10% goat serum/PBS for 1 hour at room temperature.

| Experimental design and data analysis
In all experiments the position of peptides and antagonists (pharmacology) or transfected receptors (ELISAs) on assay plates was randomized during each independent experiment, which are independent biological replicates. In all cases, duplicate, triplicate, or quadruplicate technical replicates were conducted for each independent experiment. For Concentration-response experiments were then conducted with the same experimental design for cAMP, ERK1/2, and IP 1 pathways. For signaling assays, the relevant control peptide (CRF or UCN1) was included on each assay plate in each independent experiment. The requirement for multiple concentrations of agonist/antagonist to be made-up by a single operator for individual assays resulted in blinding not being feasible. All group sizes were designed to be equal at n = 5 independent experiments. However, when F tests performed on individual experiments indicated that a single curve could fit to both agonist and antagonist curves or no agonist concentration-response curve could be fitted to the data, neither pK B nor pEC 50 values could be determined, respectively. Therefore, no statistical comparisons were performed and experiments were curtailed at n = 3-4 individual experiments. For antagonism of UCN1-mediated IP 1 accumulation by CP-376,395 at the CRF 2 receptor, one additional experiment was performed. All data were plotted and analyzed using GraphPad Prism 6.0 or 7.0 (GraphPad Software Inc). Data points are the mean ± standard error of the mean (SEM) from n separate experiments, combined.

| Agonist assays
For agonist signaling assays data were fitted with a four-parameter logistic equation. F tests were performed to determine if the Hill slope was significantly from one (GraphPad Prism). When the Hill slope was not significantly different from one the curves were constrained to one and pEC 50 values obtained. When the Hill slope was significantly different from one, this parameter was unconstrained.
To combine the data, maximal responses (E max ) were determined and the data expressed as a percentage of the E max obtained for matched UCN1 on the same assay plate. For pERK1/2 time course assays, the data were normalized to the response from 20% FBS conducted in parallel. Data normalization was necessary due to variation introduced by transient receptor transfection.

| Signaling bias
Agonist signaling bias was calculated as published previously. 33 UCN1 acted as a full agonist at all signaling pathways examined and was used as the reference ligand. Briefly, using equations for the Operational Model for Bias, GraphPad Prism 7.0 was used to determine Log(τ/K A ). The Hill slope was determined using the procedure outlined for agonist assays. Thus, the Hill slope was constrained to one for cAMP and IP 1 assays and unconstrained for pERK1/2, the E max parameter was shared or set to 100 if the fit was ambiguous, to reflect the reference ligand response. ΔLog(τ/K A ) was then calculated by subtracting the reference ligand Log(τ/K A ) from each test ligand Log(τ/K a ). ΔΔLog(τ/K A ) ratios were determined by comparing the ΔLog(τ/K A ) for each signaling pathway to the reference signaling pathway (cAMP). The bias factor for each ligand was defined as the inverse log of the ΔΔLog(τ/K A ) for a given ligand.

| Antagonist assays
For antagonist assays, pK B antagonist potency values were calculated using pEC 50 values from concentration response curves of agonist alone, or agonist in the presence of one or three different antagonist concentrations. Initially, F tests were performed to determine if both the agonist alone and agonist in the presence of antagonist data sets could be fitted using a single curve. When a single curve did not fit all data sets, pK B values were calculated. When the E max in the presence of antagonist was not significantly different from the agonist alone curve (F test), the data were analyzed using global Schild analysis for competitive antagonists (Graphpad Prism). F tests were then performed to determine if the Schild slope was significantly from one.
When the Schild slope was not significantly different from one, this parameter was constrained to one and antagonist pK B values were obtained. When the E max in the presence of antagonist was significantly different from the agonist alone curve (F test), the method of Gaddum for an insurmountable or non-competitive antagonist was used to determine antagonist potency. 34 To generate curves, data points were receptor, CP 376-395 data sets were further analyzed by fitting the operational model of allosterism to the combined data sets. 35 No intrinsic activity of CP-376,395 was observed, therefore τ B was constrained to 0. The E max was constrained to the maximum % value in the control (agonist alone) curve and E min was set to 0%. When the Hill slope of the control curve was equal to one, n was set to one. All other parameters were shared between all data sets. The β value was constrained to 0 when initial fits reported an ambiguous value which was near 0. The CRF 2 data sets used a single antagonist concentration and therefore could not be fitted to the operational model of allosterism.

| ELISA assays
To compare the cell surface expression of RAMP1 and 2 between receptors, the data were normalized to the maximum surface expression generated by CLR and RAMP1 or 2 because CLR gives reproducibly high surface expression of both RAMP1 and RAMP2. 32,36 Data normalization was necessary due to variation introduced by transient receptor transfection. For FLAG-RAMP3, normalization was not performed.

| Statistical analysis
The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology. 37 All data were plotted and analyzed using GraphPad Prism 6.0 or 7.0 (GraphPad Software Inc). pEC 50 and pK B values were averaged from separate biological replicates (individual experiments) to generate mean values. For signaling data, pEC 50 and pK B which are log values and assumed to be normally distributed, significant differences were determined using parametric tests. When two values were compared, an un-paired two-tailed Student's t test was used. When more than two values were compared, a one-way ANOVA with post hoc Dunnett's test was used. For cell surface expression of RAMP1 and RAMP2 (ELISAs), the mean normalized surface expression from individual experiments were combined. Significant differences were determined using one-way ANOVA with post hoc Dunnett's test. In all cases statistical significance was defined as P < .05.

| CRF receptors in transfected Cos7 cells exhibit biased signaling
The pharmacology of the CRF 1 and CRF 2 receptors was characterized in transiently transfected cells by determining the ability of their endogenous ligands to activate different intracellular signaling pathways. We selected three signaling molecules for interrogation; cAMP, IP 3 (via IP 1 ) and ERK1/2. cAMP and IP 1 are important signaling molecules for Gαs and Gαq signaling, respectively, and pERK1/2 is reportedly important for downstream effects of CRF. 38,39 We first established that no endogenously expressed CRF-responsive receptor was functional in Cos7 cells ( Figure S1).
Experiments were then performed to determine the optimal time point for pERK1/2 and IP 1 accumulation to conduct concentrationresponse experiments at CRF receptors. The data suggested that in response to CRF and UCN1, 7 minutes for pERK1/2 and 90 minutes for IP 1 accumulation were the optimal time points for both receptors ( Figure S2). Interestingly, 100 nmol/L CRF failed to stimulate IP 1 accumulation during the time course for CRF 2 activation ( Figure S2D), suggesting that its ability to signal via this pathway is less potent than at the CRF 1 receptor, when compared to UCN1.
Concentration-response experiments at the CRF 1 receptor revealed that CRF and UCN1, but not UCN2 and UCN3 exhibited a F I G U R E 1 Intracellular signaling of CRF, UCN1, UCN2, and UCN3 in Cos7 cells expressing CRF 1 or CRF 2 receptors. A, Stimulation of cAMP accumulation by peptides at CRF 1 receptors. B, Stimulation of ERK1/2 phosphorylation by peptides at CRF 1 receptors. C, Stimulation of IP 1 accumulation by peptides at CRF 1 receptors. D, Stimulation of cAMP accumulation by peptides at CRF 2 receptors. E, Stimulation of ERK1/2 phosphorylation by peptides at CRF 2 receptors. F, Stimulation of IP 1 accumulation by peptides at CRF 2 receptors. Data points are the mean ± SEM of the combined data from five independent experiments, performed in triplicate. CRF, corticotropin releasing factor; ERK1/2, extracellular signal-regulated kinase 1/2; IP 1 , inositol monophosphate concentration-dependent increase in cAMP, pERK1/2, and IP 1 accumulation (Figure 1A-C; Table 1). For both cAMP and IP 1 accumulation, the maximal responses to CRF and UCN1 were similar and the Hill slope was not significantly different from 1. For pERK1/2 the maximal responses were similar, and the Hill slope was 0.43 and 0.52 for CRF and UCN1, respectively. The relative potency between CRF and UCN1 was similar at all pathways, although CRF was significantly more potent at cAMP signaling and UCN1 was significantly more potent when IP 1 signaling was measured. However, when comparing potency between pathways, CRF and UCN1 were approximately 100-fold more potent at cAMP signaling than either IP 1 or pERK1/2 (Table 1). To quantify whether UCN1 or CRF displayed a preference for a specific signaling pathway, biased signaling was assessed using the ΔΔlog(τ/K A ) method 33 (Table 1). This suggested that CRF was approximately 11-fold biased for cAMP over IP 1 accumulation relative to UCN1 at the CRF 1 receptor.
In comparison to the CRF 1 receptor, the CRF 2 receptor displayed greater variation in ligand responses between pathways. CRF, UCN1, UCN2, and UCN3 all produced concentration-dependent increases in cAMP, pERK1/2, and IP 1 accumulation at this receptor (Figure 1D-F; Table 1). The Hill slope was not significantly different from one when CRF, UCN1, UCN2, or UCN3 stimulated cAMP and IP 1 accumulation . However, when pERK1/2 was measured, the Hill slope was 0.28, 0.51, 0.23, and 0.30 for CRF, UCN1, UCN2, and UCN3, respectively. Interestingly, the peptides were not equally active. UCN1 and UCN3 activated cAMP signaling more potently than UCN2 and CRF ( Figure 1D; Table 1). In contrast, there was no significant difference in the ability of CRF, UCN1, UCN2, and UCN3 to produce pERK1/2, although UCN3 trended toward being the most potent ( Figure 1E; Table 1). For IP 1 accumulation, UCN1 was approximately 10-, 25-, and 67-fold more potent than UCN3, UCN2, and CRF, respectively ( Figure 1F; Table 1). CRF also displayed a lower E max than the UCN peptides for the accumulation of IP 1 , suggesting that CRF is a partial agonist at this receptor via this pathway. These differences in signaling profiles were supported by analysis of biased signaling (Table 1), whereby relative to cAMP and pERK1/2, CRF, and UCN3 displayed lower potencies for the activation of IP 1 signaling. This suggests that CRF and UCN3 are biased agonists relative to UCN1 with a preference for stimulating cAMP over IP 1 accumulation (approximately 47-and 27-fold, respectively).

| Characterization of antagonist pharmacology at CRF receptors
Overall, the signaling behavior and identification of biased signaling for cAMP over IP 1 accumulation by CRF and UCN3 relative to UCN1 at the CRF 2 receptor indicated that the activation of these receptors is more complex than is currently appreciated. To further understand CRF receptor signaling behavior, the ability of antagonists to block CRF, and in some experiments, UCN1-mediated cAMP and IP 1 accumulation at CRF receptors were investigated. Three antagonists were selected; α-helical CRF  , astressin 2B , and CP-376,395. 40  The majority of prior antagonist characterization has been conducted using competitive binding and IC 50 format assays. Although these types of assays give a snapshot of antagonist activity, they do not have the depth of Schild analysis, which can highlight additional molecule behavior, such as partial agonism and insurmountable antagonism.
Thus, where possible, we elected to undertake Schild-style analysis.

| α-helical CRF (9-41) weakly discriminates between CRF receptors in transfected Cos7 cells
α-Helical CRF  , has been reported as a competitive antagonist of CRF and UCN1 at both CRF 1 and CRF 2 receptors. 43 CRF-stimulated cAMP accumulation was antagonized by α-helical CRF  at the CRF 1 receptor (Figure 2A). Interestingly, α-helical CRF  also weakly stimulated cAMP accumulation with an E max of 14.8% indicating that it can act as a weak partial agonist of this receptor ( Figure S3A). Similar partial agonism by α-helical CRF  at the CRF 1 receptor has previously been reported. 44 Despite the elevation in basal cAMP with α-helical CRF  , global Schild analysis fitted the data well. The Schild slope was not significantly different from one and was therefore constrained to one. α-helical CRF  antagonized CRF at the CRF 1 receptor with a pK B of 6.77 (Table 2).
α-Helical CRF  was approximately 10-fold more potent at antagonizing CRF-induced cAMP accumulation at the CRF 2 receptor, compared to the CRF 1 receptor ( Figure 2B; Table 2). No partial agonism was observed for α-helical CRF  at the CRF 2 receptor ( Figure S3B). The Schild slope was not significantly different to one and global Schild analysis indicated that α-helical CRF  antagonized CRF at the CRF 2 receptor with a pK B of 7.73 (Table 2).

| Astressin 2B exhibits probe-dependent antagonism at CRF receptors
Astressin 2B , is a highly modified truncated peptide, which is reported to be a selective antagonist of the CRF 2 receptor. 41 Interestingly, CRF-mediated cAMP signaling was antagonized by astressin 2B at the F I G U R E 2 Antagonism of CRF-mediated cAMP signaling by α-helical CRF  in Cos7 cells expressing CRF 1 or CRF 2 receptors. A, Antagonism of cAMP accumulation by α-helical CRF  at CRF 1 receptors. B, Antagonism of cAMP accumulation by α-helical CRF   Note: Antagonist potency values (pK B ) were determined using global Schild analysis for cAMP signaling or the Gaddum method for insurmountable antagonism for IP 1 accumulation. Data were analyzed by a student's t test.
Data are mean ± SEM of the combined data from n independent experiments. NC; no curve could be fitted to the data.
Astressin 2B was at least 1000-fold more potent at antagonizing either CRF or UCN1-induced cAMP signaling at the CRF 2 receptor, compared to the CRF 1 receptor (Figure 3; Table 2). However, astressin 2B also acted as a weak partial agonist of the CRF 2 receptor, stimulating cAMP accumulation with an E max of 11.0%; this was not F I G U R E 3 Antagonism of CRF or UCN1-mediated cAMP signaling by astressin 2B in Cos7 cells expressing CRF 1 or CRF 2 receptors. A, Antagonism of CRF-mediated cAMP accumulation by astressin 2B at CRF 1 receptors. B, Antagonism of UCN1-mediated cAMP accumulation by astressin 2B at CRF 1 receptors. C, Antagonism of CRF-mediated cAMP accumulation by astressin 2B at CRF 2 receptors. D, Antagonism of UCN1-mediated cAMP accumulation by astressin 2B at CRF 2 receptors. Data points are the mean ± SEM of the combined data from 5 (A, C and D) or 3 (B) independent experiments, performed in triplicate. CRF, corticotropin releasing factor the case at the CRF 1 receptor ( Figure S4A,B). Global Schild analysis indicated that astressin 2B was approximately 10-fold more potent at antagonizing CRF-mediated (pK B of 9.52) than UCN1-mediated (pK B of 8.44) cAMP signaling ( Figure 3C,D; Table 2). These findings suggest that astressin 2B behaves as an agonist-or probe-dependent antagonist at the CRF receptors, favoring the antagonism of CRF over UCN1-mediated cAMP signaling.

| CP-376,395 exhibits probe-dependent antagonism at the CRF 1 receptor
CP-376,395 is a small molecule antagonist reported to be selective for the CRF 1 receptor. 42 In contrast to α-helical CRF  and astressin 2B , which are larger peptide antagonists, CP-376,395 displayed no evidence of partial agonism. CP-376,395 effectively antagonized both CRF and UCN1-mediated cAMP accumulation at the CRF 1 receptor ( Figure 4A,B; Table 2). Global Schild analysis indicated that CP-376,395 was approximately 50-fold more potent at antagonizing CRF-mediated (pK B of 6.99) than UCN1-mediated (pK B of 5.34) cAMP accumulation (Table 2). Interestingly, this finding suggests that CP-376,395 behaves as an agonist-or probe-dependent antagonist, favoring the antagonism of CRF over UCN1-mediated cAMP accumulation. CP-376,395 was used to stabilize a CRF 1 receptor crystal structure. 45 The structure suggests that CP-376,395 binds at an allosteric site and may thus act as an allosteric modulator.  Table 2). F tests conducted on individual data sets suggested a reduction in E max , indicative of a non-competitive antagonist. To confirm that this was not a non-specific effect on this pathway, the ability of 100 µmol/L CP-376,395 to antagonize IP 1 accumulation at the calcitonin receptor was tested ( Figure S5). CP-376,395 had no effect on IP 1 accumulation at the calcitonin receptor, suggesting that the effects of CP-376,395 on E max were CRF 1 receptor-dependent. The reduction in E max indicated that global Schild analysis was not an appropriate method to analyze antagonism, therefore, the method of Gaddum was used to determine antagonist potency for a non-competitive or insurmountable antagonist. 34 This suggested that CP-376,395 was approximately 8-fold more potent at antagonizing CRF-mediated (pK B of 7.55) than UCN1-mediated (pK B of 6.67) IP 1 accumulation at the CRF 1 receptor ( To confirm the specificity of CP-376,395 for the CRF 1 receptor, antagonist activity was compared at the CRF 2 receptor. One hundred micromolar CP-376,395 had no effect on either CRF or UCN1mediated cAMP accumulation at the CRF 2 receptor (Figure 5A,B; Table 2). This suggests that CP-376,395 had a pK B of <4 at the  accumulation with a pK B of 3.40 (Table 2).

| CRF receptors do not increase RAMP expression at the cell surface
Previous research has suggested that both the CRF 1α and CRF 1β receptor splice variants can interact with RAMP2, as determined by enhancement of RAMP2 cell surface expression. 19,21 RAMP2 was also reported to increase Gαq-coupling of the CRF 1β receptor variant. 19 We hypothesized that interactions between the CRF receptors and RAMPs could further alter signaling and antagonist behavior. To address this question, we first sought to confirm that the CRF 1 receptor can affect RAMP2 surface expression, and to compare this to the CRF 2 receptor. Two robust RAMP partners-CLR and CTR were used as positive controls, and additional class B GPCRs (glucagon, PAC 1 , and VPAC 1 receptors) were also examined in parallel. In Cos7 cells, the cell surface expression of myc-tagged RAMP1 was significantly increased in the presence of the CLR, CTR, and VPAC 1 receptors. However, no change in RAMP1 surface expression was observed with CRF 1 and CRF 2 receptors ( Figure 6A).
Similar results were observed in HEK-293S cells ( Figure 6B). Only

| D ISCUSS I ON AND CON CLUS I ON S
The CRF receptors have been the target of intensive efforts to develop new drugs, with the major clinical focus on the treatment of anxiety, depression, and drug-dependence. 18 Although no specific CRF receptor targeted therapy has been approved by regulatory au- Data were analyzed by one-way ANOVA followed by a post hoc Dunnett's test. *P < .05. Data points are the mean ± SEM of the combined data from five independent experiments, performed in quadruplicate. CLR, calcitonin receptor-like receptor; CRF, corticotropin releasing factor; RAMP, receptor activity-modifying protein the ability of species-matched CRF receptors and agonists to activate cAMP, pERK1/2, and IP 1 accumulation and calculated bias for these agonists. 33 Overall, the CRF receptors displayed similar agonist pharmacology to that described in the literature. 49 However, we observed some subtle differences between signaling pathways with magnitudes similar to those reported at related receptors. Specifically, CRF displayed biased signaling towards accumulation of cAMP over IP 1 relative to UCN1 at the CRF 1 re- modulation. The peptide antagonists, α-helical CRF  and astressin 2B , displayed similar antagonist potency to previous reports. 44,58,59 However, both α-helical CRF  and astressin 2B displayed weak partial agonism at the CRF 1 and the CRF 2 receptors, respectively. This phenomenon may have resulted in the antagonist determination being inaccurate and potentially over-estimated. The appearance of partial agonism is perhaps unsurprising as several antagonists derived by truncating the endogenous peptide for class B GPCRs display this property and partial agonism has previously been reported for α-helical CRF  in a receptor dependent manner. 29,31,44,60 However, in other studies partial agonism is either not observed or reported. 58,61 This inconsistency may reflect the difficulty in detecting a weak agonist response, differences in receptor expression or batch-dependent variation in the antagonist prepara- This was followed-up by a second study that showed both CRF 1α and CRF 1β , but not CRF 2β , interact with RAMP2 and was supported by prior in vivo data from RAMP2 −/− mouse models, which display a weaker plasma ACTH response to CRF. 19,21 Given the close evolutionary relationship between all class B GPCRs and specifically between the CRF receptors, we hypothesized that RAMPs may also interact with CRF 2α receptors. Surprisingly, in the current study, neither CRF 1α nor CRF 2α increased RAMP1 or RAMP2 cell surface expression in the three cell lines tested. However, the current study is somewhat in agreement with a recent report, where the CRF 1 receptor only weakly interacted with RAMP2 and the CRF 2 receptor did not interact with either RAMP1 or RAMP2. 22 Experiments using RAMP3 were halted as neither construct was functional in our assays. We confirmed that the sequences of the FLAG-RAMP3 were the same as has been reported previously. 19,67 There is no clear explanation for this difference. In contrast to the previous study, the presence of RAMP2 did not alter Gαq coupled CRF 1 receptor signaling. The reasons for the discrepancy in the effect of CFR 1α on RAMP2 between studies is not clear. However, this may relate to differences between the expression level of RAMP2 or CFR 1α between studies and the capacity to detect weak or uncommon interactions. We also tested in parallel three other class B GPCRs that had previously been shown to interact with RAMPs; the glucagon, VPAC 1 and PAC 1 receptors. 19,22,68 The results from these other receptors tested was also mixed. The PAC 1 receptor, which is reported to interact with all three RAMPs, 22 only translocated RAMP2 in Cos7, but not HEK-293S or HEK-293T cells in the current study, although the effect in Cos7 cells was very small. In one study, the VPAC 1 receptor was reported to translocate all three RAMPs to the cell surface, 69 however in a second study VPAC 1 only interacted with RAMP2 or RAMP3. 22 In the current study, we observed an increase in RAMP1 surface expression in all three cell lines, but only saw increased surface expression with RAMP2 in Cos7 cells. This suggests that for VAPC 1 receptors, translocation of RAMP1 to the cell surface is more robust than for RAMP2. Inconsistencies between studies have also previously been reported for the glucagon receptor. 22,[67][68][69] In contrast to initial studies where surface expression was increased, we observed no significant changes in RAMP2 cell surface expression when co-expressed with the glucagon receptor. 69 Similarly, two distinct studies reported that co-expression of RAMP2 with the glucagon receptor had the opposite effects on cAMP production. 67,68 These differences between studies may simply reflect the difficulties associated with investigating non-obligate heterodimers and that more sensitive methods may be required to detect subtle interactions between receptors and RAMPs. However, differences may also relate to the precise cellular to activate a specific signaling pathway or block a specific agonist through a CRF receptor. Based on our findings we propose that the already complex pharmacology associated with the CRF receptors may be underappreciated and requires further investigation.

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guide topha rmaco logy.
org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, 73 and are permanently archived in the Concise Guide to PHARMACOLOGY 2017. 40