- Email: [email protected]
What is pharmacological ‘affinity’? Relevance to biased agonism and antagonism
Opinion
What is pharmacological ‘affinity’? Relevance to biased agonism and antagonism Terry Kenakin Department of Pharmacology, University of North Carolina School of Medicine, Chapel Hill, NC, USA
The differences between affinity measurements made in binding studies and those relevant to receptor function are described. There are theoretical and practical reasons for not utilizing binding data and, in terms of the quantification of signaling bias, it is unnecessary to do so. Finally, the allosteric control of ligand affinity through receptor–signaling protein interaction is discussed within the context of biased antagonism. In this regard, it is shown that both the bias and relative efficacy of a ligand are essential data for fully predicting biased effects in vivo. What we observe is not nature itself but nature exposed to our method of questioning. – Werner Heisenberg (1901–1976) Binding: ‘Langmuirian’ affinity Pharmacological affinity is a measure of the attraction a ligand has for a biological target. It can be quantified with an equilibrium dissociation constant, defined as the ratio of the rate that the ligand approaches the protein binding site (denoted k1 in temporal units such as s1M1) and the rate that the bound ligand diffuses away from the protein binding site (denoted k2 with units of s1). Thus the characteristic number used to quantify affinity is defined as k2/k1 and denoted KA. The root model for calculation of KA is the mass action equation, a relationship first derived by A.V. Hill [1] and made popular by the chemist Irving Langmuir 6 years later as the adsorption isotherm [2]. This equation (Equation 1) defines rA, the fraction of available binding sites bound by a concentration of ligand [A]: rA ¼
½ABmax ½A þ K A
[1]
where Bmax is the maximal number of binding sites and KA the equilibrium dissociation constant of the ligand–receptor complex. The application of this simple equation to the binding of ligands to proteins such as receptors has been of undeniable value to pharmacology, but the literal translation of the parameters obtained from experiments using this model should be made with caution. The originators of Corresponding author: Kenakin, T. ([email protected]). Keywords: biased signaling; biased agonism; receptor theory. 0165-6147/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tips.2014.06.003
434
Trends in Pharmacological Sciences, September 2014, Vol. 35, No. 9
this equation defined its use and limitations. For example, as pointed out by Colquhoun [3], Hill was applying the equation to determine only if ‘. . . [an] equation of this type can satisfy all the observations, than to base any direct physical meaning on . . . KA’. In the case of Langmuir’s application of the isotherm, he was primarily interested in the adsorption of gases to metal surfaces, a medium very different from receptor protein in that the binding surface is homogeneous and non-interactive with the ligand. The required homogeneity of the binding surface was highlighted by Langmuir himself who indicated that any heterogeneity (such as might be found for activated charcoal) could alter the interpretation of the parameters obtained; ‘. . .. but it is evident that [the] Equation, which appl[ies] to adsorption by plane surfaces, could not apply to adsorption by charcoal’ [2]. The available binding sites on seven transmembrane receptors (7TMRs) most often do not satisfy either of the prerequisites of uniformity and noninteraction (vide infra). Within the context of this discussion, KA values derived from Equation 1 will be defined as ‘Langmuirian binding’ affinities. The binding of a radioligand to a 7TMR can be measured, and Equation 1 applied to the resulting data to yield a measure of KA, but the question raised in this present discussion is – when it is appropriate for independent estimates through binding studies to be utilized as estimates of ‘functional affinity’? The term functional affinity will be used to denote a number that can describe pharmacological activities of a ligand; these include the level of receptor occupancy for determination of the amount of activating complex for cellular signaling and also the receptor occupancy that causes antagonism of more efficacious agonists in vivo. As a preface to this discussion, it is important to define the species that drugs bind to, in this case, 7TMRs. 7TMRs as ensembles of allosteric proteins 7TMRs are the major means by which chemical signals are transmitted from the extracellular space to the cell cytosol. Their main function is to change their shape (conformation) in response to interactions with extracellular ligands and intracellular signaling proteins. There is a great deal of evidence to suggest that proteins such as these do not stay in static conformations but instead exist in ensembles of different conformations that interchange with the available free energy of the system [4–9]. In this sense, the
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
binding surface for ligands is not homogeneous. This heterogeneity is further enhanced by the fact that ligand binding can selectively stabilize some conformations over others, and therefore the process of binding may change the relative proportions of these conformations [10]. If the ligand promotes a given conformation, which then reacts with a signaling protein or other cellular body, or even if the binding promotes stabilization of a different conformation, then the observed affinity of that ligand can be dependent upon these interactive processes. For a scheme where receptor R changes to another state R* upon binding of ligand A, the observed affinity of the ligand is not the molecular affinity for the R state but instead an affinity that depends upon the rate of the transformation to the R* state. g
K
A þ R @ AR @ AR ’
[2]
Thus, the receptor ‘isomerizes’ [11] (becomes another thermodynamic species) through the processes controlled by g and w (the conformational state of the 7TMR is not confined to AR) and the observed affinity will be an amalgam value for KA, very different from the static parameter defined by the adsorption isotherm [11]: KA ¼
K 1 þ g’
[3]
In addition, a very important feature of 7TMRs is their allosteric nature. Extracellular ligands such as hormones or neurotransmitters stabilize conformations of the receptor that then promote binding and activation of cellular signaling proteins. Such changes in receptor conformation constitute the ‘intrinsic efficacy’ of the ligand and form the basis of direct agonism. This function is allosteric in that the binding of a ligand to the receptor alters its affinity for signaling proteins as they interact at other loci on the receptor; this flow of energy forms an allosteric vector [12] that is bidirectional in that the binding of a signaling protein to a receptor will concomitantly alter the affinity of the receptor for a ligand in other regions of the protein. This reciprocation of affinity is concisely defined by the allosteric binding model [13,14]: αKA
A + BR
ARB αKB
KB
A + R +
B
KA
[4]
AR +
B
Therefore, the affinity of the receptor R for the ligand A is modified by the interaction of the receptor with allosteric ligand B by the factor a, and the affinity of the receptor for ligand B is modified to the same degree (namely a) by the binding of ligand A. Under these circumstances the observed affinity of the ligand A for the receptor will depend both on the nature and concentration of the co-binding allosteric ligand by the expression: K¼
K A ð1 þ ½B=K B Þ ð1 þ a½B=K B Þ
[5]
Because allostery is probe-dependent (i.e., every modulator will have a unique value of a for every co-binding
ligand), Equation 5 shows how the functional affinity K can vary with both the amount and type of allosteric co-binding ligand. This probe-dependent effect has been demonstrated in numerous experimental situations; for example, the allosteric inhibitor aplaviroc of the receptor CCR5 [chemokine (C-C motif) receptor 5] produces widely divergent effects on the affinity of chemokines from a greater than 30-fold decrease in the affinity of the receptor for CCL3 [chemokine (C-C motif) ligand 3] to nearly no effect on the affinity of the chemokine CCL5 [15]. Several allosteric ligands produce effects on ligand receptor affinity through interaction at extracellular sites but there is no a priori reason to suppose that these actions cannot be mediated via the cytosol. For example, a series of CCR4 and CCR5 antagonists have been shown to produce allosteric blockade of chemokine receptors by binding to sites inside the cell [16]. Similarly, signaling proteins such as G proteins or b-arrestin will function as allosteric modulators of the receptor, and it follows that the affinity of the receptor for different ligands (e.g., agonists) can differ by varying a-values as different signaling proteins associate with the receptor. This effect is dramatically illustrated by the difference in the X-ray crystallographic structure of the b2-adrenoceptor bound and not bound to a nanobody simulating a G protein [17,18]. Similarly, differences in the nature of the affinity changes with different signaling co-binding proteins are observed for ghrelin receptors in lipid discs – clear changes in the conformations of the receptors have been demonstrated following addition of Gq to nanodiscs [19]. Studies using exponential fluorescent lifetime decay analysis also show the creation of different conformations upon addition of b-arrestin [19]. k-Opioid receptor SCAM (substituted cysteine accessibility method) studies indicate changes in conformation in 7TM domains 6 and 7 with binding of Ga16 and/or Gai2 G protein subunits, and this is reflected in an 18-fold change in affinity for the ligand salvanoran [20]. In general, there is abundant evidence to show that receptors can complex with membrane proteins to yield multiple affinity states, a condition Black and Shankley referred to as ‘receptor distribution’ [21]. These types of data underscore the impact of membrane components on agonist affinity for receptors. Membrane components can alter antagonist affinity as well. For example, the membrane protein RAMP3 (receptor activity modifying protein [22]) produces dramatic effects on the blockade of human calcitonin receptor by the antagonist peptide AC66 [23]. The Schild regressions shown in Figure 1A,B show how cotransfection of RAMP3 produces a 10-fold decrease in the affinity of the antagonist AC66 in blocking the effects of amylin, but has no effect on the antagonism of human calcitonin; this illustrates both the allosteric control of antagonist affinity of a receptor for ligands by a membrane-bound species and also, the probedependence of that allosteric control. Functional affinity This discussion will define the term ‘functional’ affinity as being the KA value – defining the strength of binding of a ligand as it forms a complex with the receptor to both induce a cellular response and also to interfere with the 435
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Key: No RAMP
2.5
+ RAMP3
Log (DR-1)
Log (DR-1)
2.5
1.2
Agonist = hCalcitonin 3
2 1.5 1
Key: No RAMP + RAMP3
2 1.5 1
0.5
0.5
0
0 –10 –9.5
–8.5
–7.5
Log [AC66]
–9
–8
–7
Log [AC66]
Fraconal maximum response
Agonist = Amylin 3
TRENDS in Pharmacological Sciences
effects of another ligand as it attempts to co-bind to the receptor to induce a physiological response. In the latter case, several published pharmacological procedures are available to measure this value for different agonists, including orthosteric simple competitive antagonists (the Schild method [24]), non-competitive orthosteric antagonists (fitting to the non-competitive model [25]), partial agonists (Schild method, method of Stephenson [26] and method of Kaumann and Marano [27]), inverse agonists (Schild method), and allosteric antagonists (direct fitting to the allosteric model [28]). In terms of the production of response, the amount of ligand–receptor complex formed by an agonist is dictated by the KA; a useful model that can be used to relate ligand affinity and tissue response is the Black/Leff operational model [29]: Res ponse ¼
½AtEm ½Að1 þ tÞ þ K A
[6]
The efficacy of the ligand and the efficiency of stimulus transduction of the cell is denoted by the term t; Black and Leff stipulated that the KA is the reciprocal of the affinity of the ligand for the receptor that is used to determine the amount of drug–receptor complex transduced by the cell to process response [29]. From Equation 6 it can be shown that [30]: EC50 ¼
KA 1þt
[7]
Equation 7 demonstrates that as the response capability of the cell is reduced (i.e., reduction of receptor density) then t ! 0 and the EC50 ! KA. Therefore, the EC50 of agonists producing a low level of response closely approximates the KA. Thus, at least for agonism, the KA utilized in the Black/Leff model is the concentration of agonist around which agonist concentration–response curves collapse upon diminution of the receptor density (Figure 2). If the Black/Leff model is used for any procedures to quantify agonism, then this prerequisite must be met. It can be shown that the functional KA value (measured by low level activation EC50 measurements) can be differentially modified by various allosteric co-binding guests. For example, 436
0.8 0.6
x10–1
[Rt] = 1
x10–2
0.4
x10–3 0.2
KA 0 –9
–8
–7
–5
–6
–4
–3
Log [agonist] TRENDS in Pharmacological Sciences
Figure 2. The relationship between agonism and agonist affinity (denoted as the KA, equilibrium dissociation constant of the agonist–receptor complex) and receptor density as described by the Black/Leff operational model. The model dictates that, as receptor density is diminished, the concentration–response curves collapse upon the KA on the abscissal axis. If the Black/Leff model is used to describe agonism, then this prerequisite must be met.
Figure 3 shows the different EC50 values for the chemokine agonist CCL3L1 for activation of the CCR5 receptor when modified by two different allosteric modulators [31]. Thus, as the response capability of the system is reduced (i.e., conditions where EC50 ! KA), different co-binding ligands can produce different limiting EC50 values. This brings into question whether the low-level agonism EC50 values for a ligand can be considered the functional affinity of that ligand as it antagonizes other orthosteric ligands. Figure 4 shows that this can be shown to be the case for the a1-adrenoceptor in rat anococcygeus muscle; the pEC50 of 6.5 also corresponds to the pKB for oxymetazoline after alkylation of receptors and utilization of oxymetazoline as an antagonist with the Schild method [32]. This
Fracon CCR5 internalizaon
Figure 1. Control of antagonist binding affinity by membrane bound RAMP3 (receptor activity modifying protein 3). Schild regressions to the calcitonin antagonist AC66 show a 10-fold difference in the affinity of the antagonist in cells that do and do not co-express RAMP3 with the calcitonin receptor when amylin is the agonist. Allosteric probe dependence is shown by the fact that no difference is seen when calcitonin is used as the agonist. Data redrawn from [23].
1
1.2 1
CCL3L1 TAK779
α = 0.1
0.8 0.6
β=0
TAK779 0.4 Aplaviroc
CCL3L1
0.2 Aplaviroc
0 –11
–10
–9
Log [CCL3L1]
–8
α = 0.9
–7
β=0 TRENDS in Pharmacological Sciences
Figure 3. Differential effects of two allosteric antagonists (TAK779 and aplaviroc) on the concentration-response curve describing CCR5 [chemokine (C-C motif) receptor 5] internalization to the chemokine CCL3L1 [chemokine (C-C motif) ligand 3-like 1] in U373-MAGI-CCR5-E cells. The fact that the curves collapse around different KA values for the agonist demonstrates allosteric probe dependence and the difference in the effect of each allosteric antagonist on the affinity of CCL3L1. The effects are consistent with a 10-fold decrease in affinity for TAK779 and only a 1.1-fold decrease for aplaviroc. The allosteric effects on the receptor are defined as those changing the affinity of the agonist (denoted a) and those changing the efficacy of the agonist (denoted b) from the model of functional allosterism [55– 57]. Data redrawn from [31].
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
1.2
2
Norepinephrine
1
1.5
0.8
Log (DR-1)
Fraconal maximum response
Opinion
0.6 0.4
Oxymetazoline
1 0.5
0.2 0
0 –8
–7
–6
–5
–4
–7
Log [agonist]
–6
–5
Log [oxymetazoline] TRENDS in Pharmacological Sciences
Figure 4. Contractile effects of a-adrenoceptor agonists (norepinephrine and oxymetazoline) in rat anococcygeus muscle. The panel on the left shows that, in tissues alkylated with phenoxybenzamine (POB) to remove a portion of the aadrenoceptors (POB 0.1 mM, 10 min), oxymetazoline is a partial agonist with a pEC50 of 6.5. Removal of a greater portion of the receptor through controlled alkylation of receptor with phenoxybenzamine (0.3 mM, 10 min) produces a tissue that responds to norepinephrine but not to oxymetazoline. A Schild regression for oxymetazoline as an antagonist in these tissues yielded a pKB value of 6.5 (95% confidence limits 6.3–6.8). Data redrawn from [32].
emphasizes the link between ligand affinity and low response level EC50 values. Measuring agonist bias The term ‘bias’ will be used to denote the preferential ability of a ligand to cause the receptor to interact with a distinct signaling pathway in the cell as opposed to other pathways. It will be seen that the molecular determinants of bias are both the efficacy and affinity of the ligand for the receptor as it interacts with these pathways. The current model for biased signaling is based on the notion that different agonist molecules stabilize different receptor conformations interacting with multiple cytosolic guests [33] and that these go on to interact with signaling proteins to produce differential activation of signaling pathways [34]. The original model proposed ligand-based stabilization on theoretical grounds which has subsequently been verified experimentally with 19 F nuclear magnetic resonance (NMR) studies [35]. However, the original model did not adequately consider the guest-stabilization effect of signaling molecules as shown in X-ray crystallographic work [17,18] (Figure 5). In general, biased signaling should be considered to be a function of the Evoluon of the biased signaling model (A)
G proteinselecve
β-Arresnselecve
(B)
G proteinselecve
β-Arresnselecve
TRENDS in Pharmacological Sciences
Figure 5. Two schematic versions of the molecular model of biased signaling proposed in [34]. (A) Direct ligand stabilization of receptor states that then have a preference for signaling pathways as supported by 19F NMR data [35]. (B) Receptor states pre-stabilized by interaction with signaling molecules as suggested by X-ray crystallographic data [17,18].
ternary complex of ligand, receptor, and signaling molecule rather than being only a ligand–receptor effect [36]. This being the case, the bias of various molecules will necessarily vary with signaling pathway. The basis for the quantification of agonist bias is the calculation of the relative efficacy of a given agonist for stimulation of different cytosolic pathways. The currently known methods to calculate this are: (i) Transducer coefficients: DDlog(t/KA) values are obtained through application of the Black/Leff model for each pathway [37]. (ii) Relative activity (RA) values: DDlog(RA) values (RA = maximal response/EC50) are calculated for each pathway [38–40]. It should be noted that for concentration response curves of slope = 1, DDlog(RA) values are identical to the DDlog(t/KA) values [39]. (iii) blig values [41]: DDlog(t) values are calculated with the Black/Leff model from estimates of KA. Although the affinity used to obtain t values is not specified in the description of the method, if the Black/Leff method is used to obtain t, then by definition the functional affinity must be used (KA as shown in Figure 2). It is important to note that every currently known method of quantifying biased agonism utilizes the functional affinity of the agonist (KA) – either through application of the Black/Leff model (through t values [37,41]) or through the functional EC50 (through RA values [38–40]). This raises into question the ad hoc application of affinity measurements made in binding studies to the calculation of agonist bias. Binding estimates have been shown to differ from functional affinity measurements in many cases (for the reasons discussed above). For instance, whereas the binding affinity for 125I-human calcitonin in HEK 293 cells stably expressing human calcitonin receptors is 16 pM (pKd = 10.77 with 95% confidence limits of 10.63–10.91), the EC50 for calcium responses for human calcitonin is 426fold higher (EC50 = 7.2 nM, pEC50 = 8.14 + 0.2), a difference in the opposite direction from that expected to be produced by efficacy [42]. There is no a priori reason that an affinity measured in binding should accurately reflect an agonist response controlling functional affinity. For the quantification of biased signaling, a functional estimate of affinity must be applied either through application of an observed partial agonist EC50 or fitting of the Black/Leff operational model to the functional data. In the case of full agonists, no unique value of KA can be obtained, but the method yields ratios of t/KA for full agonists that are unique – and these can be applied to the calculation of DDlog(t/KA) for measurement of bias. Alternatively, the method of Barlow, Scott, and Stephenson [43] can be applied to obtain Dlog(t/KA) values directly when full and partial agonists are compared (Box 1). Considering that the functional affinity of a ligand can be estimated through a low-level agonist EC50 value in a functional assay (Figure 4), then the question can be asked – is there evidence to show variation in functional affinity with signaling pathways; in other words, does the affinity of an agonist for a receptor vary in response to the signaling pathway interacting with that receptor? Figure 6 shows that the EC50 for the 5-HT2A agonist MK212 in mouse 437
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Box 1. Dlog(t/KA) values through curve comparison. The method of comparing concentration response curves to two agonists devised by Barlow, Scott, and Stephenson [43] utilizes a double reciprocal relationship between equiactive concentrations of a full and partial agonist. Thus, considering two agonists A1 and A2, the following null relationship can be derived: 1 1 t A1 K A-2 t A1 -t A2 ¼ þ ½A1 ½A2 t A2 K A-1 t A2 K A-1
[I]
It can be seen that the slope of resulting linear regression of 1/[A1] upon 1/[A2] yields the ratio of the t/KA values for both agonists, the logarithm of which is Dlog(t/KA). Figure I shows chemokine CCL3L1 as A1 and CCL5 as A2; the linear plot indicates a value for Dlog(t/KA) for these agonists as 0.24. Data from [37].
(B)
100 90 80
50 40
1.8 1.4
CCL3L1
70 60
1/[A1] × 109
% Max. internalizaon
(A)
CCL5
30 20 10
1.0 0.8 0.6 0.4
0 -10
-9 -8 Log [Agonist]
-7
-6
Slope= ΔLog(τ/KA)= 0.24 2
4 6 1/[A2] × 108
8
TRENDS in Pharmacological Sciences
Figure I. A. Agonist dose-response curves for CCL3L1 and CCL5 for internalization of CCR5 receptors. Equiactive concentrations of agonist (denoted by broken lines) are compared in a double-reciprocal plot (B) to yield a straight line according to equation I in the Box. The slope of this line is the ratio of Log(t/KA) values for the agonists. Data redrawn from [37].
embryonic fibroblast cells is 100 nM for activation of ERK (extracellular signal-related kinase), but 2 mM for the production of inositol phosphate [44]. Owing to the fact that the maximal response for both of these effects are below the system maximal response, it is mathematically
Percentage maximal response
120 100
IP
80
ERK
60 40 20 0 –8
–6
–4
–2
–20
Log [MK212] TRENDS in Pharmacological Sciences
Figure 6. 5-HT2A (5-hydroxytryptamine 2A receptor)-mediated accumulation of inositol phosphate (IP) and extracellular signal-regulated kinase 1/2 (ERK1/2) phosphorylation in mouse embryonic fibroblasts as a function of concentrations of MK212 [6-chloro-2-(1-piperazinyl)pyrazine hydrochloride]. Broken lines show the expected curves when receptor density is diminished (collapse of the curve to the KA value as shown in Figure 2). Because the maximal responses to both pathways are submaximal, the EC50 values closely approximate to KA values, and these differ with the pathway activated. There is no single value for KA that can be used in the Black/Leff operational model that can fit both concentration–response curves. Data redrawn from [44].
438
impossible for a single estimate of the KA to fit Black/Leff operational model to these data; in other words, there must be different values of KA for MK212 interaction with the 5-HT2A receptor as it interacts with the two signaling pathways. Similar significant differences in EC50 values for partial agonists for a receptor activating different signaling pathways have been shown for m-opioid receptors in HEK293 cells [45] and U2OS OPRM cells [46], and for histamine H4 receptors in U2OS cells [47]. These data show that not only can ligands differ in the efficacies with which they interact with different signaling pathways, but also in their affinity for the receptor as it interacts with those pathways. Thus, in addition to biased agonism, this opens the possibility of therapeutically applicable biased antagonism (vide infra). In terms of agonist signaling bias, the independence of affinity and efficacy with different cellular activation can still yield a range of molecules with identical bias but differential capability to produce agonism. To describe biased agonists fully, the relative efficacy of the agonists for each pathway must be considered as well as the differential propensity to activate each pathway (bias). One way to do this is to plot some measure of the relative efficacy of the agonist for each pathway (as for example through the log of the relative maximal effects) as a function of its bias for each pathway. In this regard, bias is calculated through DDlog(t/KA) or DDlog(RA) values which consider both efficacy and affinity. Specifically, the efficacy of the agonists for various pathways will determine whether or not agonism for that pathway will occur, whereas the bias [i.e., DDlog(RA)] will determine the relative potency of the agonist for each
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Albuterol
0.5
Fenoterol 1.0
0.6
Ritodrine
1.5 1.3
Orciprenaline
Clenbuterol
LOG (relave maxima)
1.0
1.1
0.4
0.9 0.7 1.0
Epinephrine 0.5
0.8 0.8
Dobutamine
Cimeterol
0.3 0.1 1
–0.5
–0.1 0
0.5
1
1.5
–0.3
LOG (bias) –0.5 TRENDS in Pharmacological Sciences
Figure 7. Combinations of signaling bias (abscissae as DDlog(RA) values) and relative efficacy (ordinates as logarithms of ratios of maximal responses) for b1-adrenoceptor agonists causing the receptor to interact with two signaling pathways (unbroken lines, Gs protein activation; broken lines, b-arrestin interaction; RA, relative activity). Ligands can have similar bias (Ritrodine and Dobutamine) but very different signaling properties in terms of activation. The reference agonist for calculation of bias is epinephrine. The straight broken line reflects a total dependence of bias on efficacy alone with no contribution from affinity. Data from [48].
pathway response when agonism does occur. Some nuances between agonism and bias are illustrated with the series of b1-adrenoceptor agonists shown in Figure 7. This figure expresses the relative maximal responses of the agonists for G protein versus b-arrestin activation (ordinate values) as a function of the relative bias for each pathway compared to epinephrine calculated with DDlog(RA) values [48]. The relative scatter around the broken correlation line is noteworthy from the point of view that, if bias were solely a function of the relative efficacy of the agonists for each pathway, then a better correlation would be expected as log(relative maximum) and log(bias) would have the same origin (namely efficacy, t values). The fact that the data vary considerably from the linear broken regression line suggests that variable affinity plays a role in the bias. In addition, exemplified by the pattern shown in Figure 6, there are agonists for which it is mathematically impossible to fit both signaling curves with a single estimate of affinity (i.e., orciprenaline). Finally, it can be seen that agonists with nearly identical bias values can still have different signaling characteristics (i.e., potency vs maximal effects) as shown by the profiles for ritrodine and dobutamine. Such descriptions assist in developing molecules for different needs; in other words, whether the desired bias is to activate or block selectively a given signaling pathway. If selective activation of a pathway is the aim, then the efficacy of the molecule is the main determinant of activity, not bias. It should be noted that, although biased agonism has been suggested as the
most commonly cited reason for therapeutic improvement of a given receptor ligand, there is no reason that selective antagonism should not be more important in some cases. For example, the crucial property of the advanced angiotensin receptor biased ligand TRV000120 is its blockade of elevated endogenous angiotensin in heart failure rather than the added b-arrestin signaling [49]. The array of different combinations of relative potency and relative efficacy of a biased ligand underscores the variety of possible profiles anticipated in vivo with biased ligands. In many cases it would also be anticipated that differences in receptor densities of different organs will produce a mixture of profiles for the same ligand in vivo. Biased antagonism Biased antagonism is an established phenomenon for negative allosteric modulators. For example, the allosteric antagonist LP1805 [N,N-(2-methylnaphyl-benzyl)-2-aminoacetonitrile] converts normal signaling by NK2 receptors for the natural agonist neurokinin A (activation of Gaq and Gas) to a biased signal of enhanced Gaq but diminished Gas [50]. Similarly, the allosteric antagonist indole1 (Na-tosyltryptophan) changes normal prostaglandin D2 (PDG2) signaling of the chemoattractant receptor-homologous molecule expressed on TH2 cells (CRTH2) receptor (Gai and b-arrestin association with the receptor) to an exclusive Gai effect with no b-arrestin signal [51]. However, the situation is not as clear for orthosteric antagonists (or 439
Opinion biased agonists binding in low-sensitivity systems to induce antagonism of endogenous signaling). Theoretically, there is no reason to assume that the formation of different receptor–signaling protein complexes to yield variable affinities will not occur; this idea has been proposed in the literature. As stated by Baker and Hill: ‘. . . it is therefore possible that antagonists might differ in their affinity for different GPCR-signaling protein complexes (e.g., Gs- and barrestin-coupled forms of the b2-adrenoceptor) and show agonist- and signaling-pathway-dependent affinities..’ [52]. Variable orthosteric affinity for b-adrenoceptor antagonists has been demonstrated; for instance, the affinity of propranolol differs by a factor of 52 when used to block [3H]CGP12177 binding (pKB = 6.44) versus CGP12177 function (pK = 8.2) [53,54]. It should be noted that differential affinity due to signaling may not produce large effects (two to fivefold differences), but these variations may still obscure signaling bias significantly. In general, it can be questioned to what extent biased ligands will be encountered in pharmacology. From a theoretical point of view, if it is accepted that receptors exist in conformational ensembles and that ligands create new ensembles based on their differential affinity for the various conformational states; it is nearly inconceivable that two ligands would produce identical ensembles of a receptor protein – in other words, all ligands will therefore be biased to some extent. However, whether such a bias will have pharmacological consequences is dependent upon the nature of the various conformations that are stabilized. Assuming that only a select number of conformations couple to cellular proteins to cause an observed cellular effect, the conformational ensembles produced by many ligands may not result in an observed biased signaling profile. In addition, the likelihood of a given conformation emerging to cause an observed signaling bias depends upon the relative KA values for the various conformations in the ensemble. Another logical question raised by these ideas concerns to what extent biased affinity will be observed in the setting of antagonism? Equation 5 predicts that it will basically be operative only in cases where the ligand stabilizes a receptor conformation with preferential affinity for a co-binding ligand (i.e., the magnitude of a is >>1). Given this, it might be supposed that biased affinity will be operative only in cases where the ligand also signals toward the preferred pathway (i.e., t >1 in Equation 7) and this, in turn, would suggest that ligands with preferential affinity and low efficacy for a given pathway may not often be encountered. This latter question will not be resolved until careful quantitative studies of the trafficking of receptor antagonism to various signaling pathways are available. Another frontier in the therapeutic application of biased ligands will be the transference of in vitro estimates of bias (i.e., DDlog(t/KA) or DDlog(RA) values) in systems where distinct interactions of receptors and coupling proteins can be measured (i.e., within the allosteric vector [12]) versus where the stimulus escapes the allosteric vector and the cell type contributes to selective signaling. In this latter instance, the signaling bias can vary dramatically and become cell type-dependent; the availability of increasing 440
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
numbers of biased ligands will allow the assessment of this potential problem. Concluding remarks There are theoretical and practical reasons for not utilizing binding affinities as descriptors of functional effects. In addition, functional experiments can provide more accurate estimates of functional affinity which, in turn, are more relevant to functional receptor activity. Specifically, the functional KA value for partial agonists can be identified through fitting of the Black/Leff operational model, and t/KA ratios are identified for full agonists with the same model. Given this, binding experiments are an unnecessary, and possibly misleading, burden in assessing functional bias. Finally, the same ideas that describe signaling bias also predict biased antagonism, an effect thus far relatively unexplored in pharmacological drug development. It will be interesting to see to what extent this class of molecule can be exploited therapeutically should biased antagonism be determined to play a practical role in drug discovery.
References 1 Hill, A.V. (1910) The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J. Physiol. 40 (Suppl.), iv–vii 2 Langmuir, I. (1916) The constitution and fundamental properties of solids and liquids. Part I. Solids. J. Am. Chem. Soc. 38, 2221–2295 3 Colquhoun, D. (2006) The quantitative analysis of drug–receptor interactions: a short history. Trends Pharmacol. Sci. 2, 149–157 4 Fraunfelder, H. et al. (1988) Conformational substrates in proteins. Annu. Rev. Biophys. Biophys. Chem. 17, 451–479 5 Fraunfelder, H. et al. (1991) The energy landscapes and motions of proteins. Science 254, 1598–1603 6 Hilser, J. et al. (1998) The structural distribution of cooperative interactions in proteins: analysis of the native state ensemble. Proc. Natl. Acad. Sci. U.S.A. 95, 9903–9908 7 Hilser, V.J. et al. (2006) A statistical thermodynamic model of protein ensembles. Chem. Rev. 106, 1545–1558 8 Onaran, H.O. and Costa, T. (1997) Agonist efficacy and allosteric models of receptor action. Ann. N.Y. Acad. Sci. 812, 98–115 9 Onaran, H.O. et al. (2002) A look at receptor efficacy. From the signaling network of the cell to the intramolecular motion of the receptor. In The Pharmacology of Functional, Biochemical, and Recombinant Systems (Handbook of Experimental Pharmacology, Vol. 148) (Kenakin, T.P. and Angus, J.A., eds), pp. 217–280, Springer 10 Kenakin, T.P. and Onaran, O. (2002) The ligand paradox between affinity and efficacy: can you be there and not make a difference? Trends Pharmacol. Sci. 23, 275–280 11 Colquhoun, D. (1985) Imprecision in presentation of binding studies. Trends Pharmacol. Sci. 6, 197 12 Kenakin, T. and Miller, L.J. (2010) Seven transmembrane receptors as shapeshifting proteins: the impact of allosteric modulation and functional selectivity on new drug discovery. Pharmacol. Rev. 62, 265–304 13 Stockton, J.M. et al. (1983) Modification of the binding properties of muscarinic receptors by gallamine. Mol. Pharmacol. 23, 551–557 14 Ehlert, E.J. (1988) Estimation of the affinities of allosteric ligands using radioligand binding and pharmacological null methods. Mol. Pharmacol. 33, 187–194 15 Watson, C. et al. (2005) The CCR5 receptor-based mechanism of action of 873140, a potent allosteric non-competitive HIV entry-inhibitor. Mol. Pharmacol. 67, 1268–1282 16 Andrews, G. et al. (2008) An intracellular allosteric site for a specific class of antagonists of the CC chemokine G protein-coupled receptors CCR4 and CCR5. Mol. Pharmacol. 73, 855–867 17 Rasmussen, S.G.F. et al. (2011) Crystal structure of the b2 adrenergic receptor-Gs protein complex. Nature 477, 549–555
Opinion 18 Rasmussen et al. (2011) Structure of a nanobody-stabilized active state of the b2 adrenoceptor. Nature 469, 175–180 19 Mary, S. et al. (2012) Ligands and signaling proteins govern the conformational landscape explored by G protein-coupled receptor. Proc. Natl. Acad. Sci. U.S.A. 109, 8304–8309 20 Yan, F. et al. (2008) Ga-subunits differentially alter the conformation and agonist affinity of k-opioid receptors. Biochemistry 47, 1567–1578 21 Black, J.W. and Shankley, N.P. (1990) Interpretation of agonist affinity estimations: the question of distributed receptor states. Proc. R. Soc. Lond. B: Biol. Sci. 240, 503–518 22 Hay, D.L. et al. (2006) GPCR modulation by RAMPS. Pharmacol. Ther. 109, 173–197 23 Armour, S. et al. (1999) Pharmacological characterization of receptor activity modifying proteins (RAMPs) and the human calcitonin receptor. J. Pharmacol. Toxicol. Methods 42, 217–224 24 Arunlakshana, O. and Schild, H.O. (1959) Some quantitative uses of drug antagonists. Br. J. Pharmacol. 14, 48–58 25 Gaddum, J.H. et al. (1955) Quantitative studies of antagonists for 5hydroxytryptamine. Q. J. Exp. Physiol. 40, 49–74 26 Stephenson, R.P. (1956) A modification of receptor theory. Br. J. Pharmacol. 11, 379–393 27 Kaumann, A.J. and Marano, M.M. (1982) On equilibrium dissociation constants for complexes of drug receptor subtypes: Selective and nonselective interactions of partial agonists with two b-adrenoceptor subtypes mediating positive chronotropic effects of () isoprenaline in kitten atria. Naunyn Schmiedebeberg’s Arch. Pharmacol. 219, 216–221 28 Kenakin, T.P. (2012) Biased signaling and allosteric machines: new vistas and challenges for drug discovery. Br. J. Pharmacol. 165, 1659– 1669 29 Black, J.W. and Leff, P. (1983) Operational models of pharmacological agonist. Proc. R. Soc. Lond. [Biol.] 220, 141–162 30 Black, J.W. et al. (1985) An operational model of pharmacological agonism: The effect of E/[A] curve shape on agonist dissociation constant estimation. Br. J. Pharmacol. 84, 561–571 31 Muniz-Medina, V.M. et al. (2009) The relative activity of ‘function sparing’ HIV-1 entry inhibitors on viral entry and CCR5 internalization: Is allosteric functional selectivity a valuable therapeutic property? Mol. Pharmacol. 75, 490–501 32 Kenakin, T.P. (1984) The relative contribution of affinity and efficacy to agonist activity: organ selectivity of noradrenaline and oxymetazoline with reference to the classification of drug receptors. Br. J. Pharmacol. 81, 131–141 33 Kenakin, T.P. and Morgan, P.H. (1989) Theoretical effects of single and multiple transducer receptor coupling proteins on estimates of the relative potency of agonists. Mol. Pharmacol. 35, 214–222 34 Kenakin, T.P. (1995) Agonist–receptor efficacy II: agonist-trafficking of receptor signals. Trends Pharmacol. Sci. 16, 232–238 35 Liu, J.J. et al. (2012) Biased signaling pathways in b2-adrenergic receptor characterized by 19F-NMR. Science 335, 1106–1110 36 Onaran, H.O. and Costa, T. (2012) Where have all the active receptor states gone? Nat. Chem. Biol. 8, 674–677 37 Kenakin, T.P. et al. (2012) A simple method for quantifying functional selectivity and agonist bias. ACS Chem. Neurosci. 3, 193–203
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
38 Ehlert, F.J. (2005) Analysis of allosterism in functional assays. J. Pharmacol. Exp. Ther. 315, 740–754 39 Tran, J.A. et al. (2009) Estimation of the relative microscopic affinity constants of agonists for the active state of the receptor in functional studies on M2 and M3 muscarinic receptors. Mol. Pharmacol. 75, 381–396 40 Figueroa, K.W. et al. (2009) Selectivity of agonists for the active state of the M1 to M4 muscarinic receptor subtypes. J. Pharmacol. Exp. Ther. 328, 331–342 41 Rajagopal, S. et al. (2011) Quantifying ligand bias at seventransmembrane receptors. Mol. Pharmacol. 80, 367–377 42 Watson, C. et al. (2000) The use of stimulus-biased assay systems to detect agonist-specific receptor active states: Implications for the trafficking of receptor stimulus by agonists. Mol. Pharmacol. 58, 1230–1238 43 Barlow, R.B. et al. (1967) An attempt to study the effects of chemical structure on the affinity and efficacy of compounds related to acetylcholine. Br. J. Pharmacol. 21, 509–522 44 Strachan, R.T. et al. (2010) Genetic deletion of p90 ribosomal S6 kinase 2 alters patterns of 5-hydroxytryptamine 2A serotonin receptor functional selectivity. Mol. Pharmacol. 77, 327–338 45 McPherson, J. et al. (2010) m-Opioid receptors: correlation of agonist efficacy for signaling with ability to activate internalization. Mol. Pharmacol. 78, 756–766 46 Nickolls, S.A. et al. (2011) Understanding the effect of different assay formats on agonist parameters: a study using the m-opioid receptor. J. Biomol. Screen. 16, 706–716 47 Nijmeijer, S. et al. (2012) Analysis of multiple histamine H4 receptor compound classes uncovers Gai protein- and b-arrestin2 biased ligands. Mol. Pharmacol. 82, 1174–1182 48 Casella, I. et al. (2011) Divergent agonist selectivity in activating a1and a2-adrenoceptors for G-protein and arrestin coupling. Biochem. J. 438, 191–202 49 Violin, J.D. and Lefkowitz, R.J. (2007) b-Arrestin-biased ligands at seven transmembrane receptors. Trends Pharmacol. Sci. 28, 416–422 50 Maillet, E.L. et al. (2007) A novel, conformation-specific allosteric inhibitor of the tachykinin NK2 receptor (NK2R) with functionally selective properties. FASEB J. 21, 2124–2134 51 Mathiesen, J.M. et al. (2005) Identification of indole derivatives exclusively interfering with a G protein-independent signaling pathway of the prostaglandin D2 receptor CRTH2. Mol. Pharmacol. 68, 393–402 52 Baker, J.G. and Hill, S.J. (2007) Multiple GPCR conformations and signaling pathways: implications for antagonist affinity estimates. Trends Pharmacol. Sci. 28, 374–381 53 Baker, J.G. (2005) Site of action of b-ligands at the human b1adrenoceptor. J. Pharmacol. Exp. Ther. 313, 1163–1171 54 Baker, J.G. (2005) The selectivity of b-adrenoceptor antagonists at the human b1, b2 and b3 adrenoceptors. Br. J. Pharmacol. 144, 317–322 55 Kenakin, T.P. (2005) New concepts in drug discovery: collateral efficacy and permissive antagonism. Nat. Rev. Drug Discov. 4, 919–927 56 Ehlert, F.J. (2005) Analysis of allosterism in functional assays. J. Pharmacol. Exp Ther. 315, 740–754 57 Price, M.R. et al. (2005) Allosteric modulation of the cannabinoid CB1 receptor. Mol. Pharmacol. 68, 1484–1495
441
What is pharmacological ‘affinity’? Relevance to biased agonism and antagonism Terry Kenakin Department of Pharmacology, University of North Carolina School of Medicine, Chapel Hill, NC, USA
The differences between affinity measurements made in binding studies and those relevant to receptor function are described. There are theoretical and practical reasons for not utilizing binding data and, in terms of the quantification of signaling bias, it is unnecessary to do so. Finally, the allosteric control of ligand affinity through receptor–signaling protein interaction is discussed within the context of biased antagonism. In this regard, it is shown that both the bias and relative efficacy of a ligand are essential data for fully predicting biased effects in vivo. What we observe is not nature itself but nature exposed to our method of questioning. – Werner Heisenberg (1901–1976) Binding: ‘Langmuirian’ affinity Pharmacological affinity is a measure of the attraction a ligand has for a biological target. It can be quantified with an equilibrium dissociation constant, defined as the ratio of the rate that the ligand approaches the protein binding site (denoted k1 in temporal units such as s1M1) and the rate that the bound ligand diffuses away from the protein binding site (denoted k2 with units of s1). Thus the characteristic number used to quantify affinity is defined as k2/k1 and denoted KA. The root model for calculation of KA is the mass action equation, a relationship first derived by A.V. Hill [1] and made popular by the chemist Irving Langmuir 6 years later as the adsorption isotherm [2]. This equation (Equation 1) defines rA, the fraction of available binding sites bound by a concentration of ligand [A]: rA ¼
½ABmax ½A þ K A
[1]
where Bmax is the maximal number of binding sites and KA the equilibrium dissociation constant of the ligand–receptor complex. The application of this simple equation to the binding of ligands to proteins such as receptors has been of undeniable value to pharmacology, but the literal translation of the parameters obtained from experiments using this model should be made with caution. The originators of Corresponding author: Kenakin, T. ([email protected]). Keywords: biased signaling; biased agonism; receptor theory. 0165-6147/ ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tips.2014.06.003
434
Trends in Pharmacological Sciences, September 2014, Vol. 35, No. 9
this equation defined its use and limitations. For example, as pointed out by Colquhoun [3], Hill was applying the equation to determine only if ‘. . . [an] equation of this type can satisfy all the observations, than to base any direct physical meaning on . . . KA’. In the case of Langmuir’s application of the isotherm, he was primarily interested in the adsorption of gases to metal surfaces, a medium very different from receptor protein in that the binding surface is homogeneous and non-interactive with the ligand. The required homogeneity of the binding surface was highlighted by Langmuir himself who indicated that any heterogeneity (such as might be found for activated charcoal) could alter the interpretation of the parameters obtained; ‘. . .. but it is evident that [the] Equation, which appl[ies] to adsorption by plane surfaces, could not apply to adsorption by charcoal’ [2]. The available binding sites on seven transmembrane receptors (7TMRs) most often do not satisfy either of the prerequisites of uniformity and noninteraction (vide infra). Within the context of this discussion, KA values derived from Equation 1 will be defined as ‘Langmuirian binding’ affinities. The binding of a radioligand to a 7TMR can be measured, and Equation 1 applied to the resulting data to yield a measure of KA, but the question raised in this present discussion is – when it is appropriate for independent estimates through binding studies to be utilized as estimates of ‘functional affinity’? The term functional affinity will be used to denote a number that can describe pharmacological activities of a ligand; these include the level of receptor occupancy for determination of the amount of activating complex for cellular signaling and also the receptor occupancy that causes antagonism of more efficacious agonists in vivo. As a preface to this discussion, it is important to define the species that drugs bind to, in this case, 7TMRs. 7TMRs as ensembles of allosteric proteins 7TMRs are the major means by which chemical signals are transmitted from the extracellular space to the cell cytosol. Their main function is to change their shape (conformation) in response to interactions with extracellular ligands and intracellular signaling proteins. There is a great deal of evidence to suggest that proteins such as these do not stay in static conformations but instead exist in ensembles of different conformations that interchange with the available free energy of the system [4–9]. In this sense, the
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
binding surface for ligands is not homogeneous. This heterogeneity is further enhanced by the fact that ligand binding can selectively stabilize some conformations over others, and therefore the process of binding may change the relative proportions of these conformations [10]. If the ligand promotes a given conformation, which then reacts with a signaling protein or other cellular body, or even if the binding promotes stabilization of a different conformation, then the observed affinity of that ligand can be dependent upon these interactive processes. For a scheme where receptor R changes to another state R* upon binding of ligand A, the observed affinity of the ligand is not the molecular affinity for the R state but instead an affinity that depends upon the rate of the transformation to the R* state. g
K
A þ R @ AR @ AR ’
[2]
Thus, the receptor ‘isomerizes’ [11] (becomes another thermodynamic species) through the processes controlled by g and w (the conformational state of the 7TMR is not confined to AR) and the observed affinity will be an amalgam value for KA, very different from the static parameter defined by the adsorption isotherm [11]: KA ¼
K 1 þ g’
[3]
In addition, a very important feature of 7TMRs is their allosteric nature. Extracellular ligands such as hormones or neurotransmitters stabilize conformations of the receptor that then promote binding and activation of cellular signaling proteins. Such changes in receptor conformation constitute the ‘intrinsic efficacy’ of the ligand and form the basis of direct agonism. This function is allosteric in that the binding of a ligand to the receptor alters its affinity for signaling proteins as they interact at other loci on the receptor; this flow of energy forms an allosteric vector [12] that is bidirectional in that the binding of a signaling protein to a receptor will concomitantly alter the affinity of the receptor for a ligand in other regions of the protein. This reciprocation of affinity is concisely defined by the allosteric binding model [13,14]: αKA
A + BR
ARB αKB
KB
A + R +
B
KA
[4]
AR +
B
Therefore, the affinity of the receptor R for the ligand A is modified by the interaction of the receptor with allosteric ligand B by the factor a, and the affinity of the receptor for ligand B is modified to the same degree (namely a) by the binding of ligand A. Under these circumstances the observed affinity of the ligand A for the receptor will depend both on the nature and concentration of the co-binding allosteric ligand by the expression: K¼
K A ð1 þ ½B=K B Þ ð1 þ a½B=K B Þ
[5]
Because allostery is probe-dependent (i.e., every modulator will have a unique value of a for every co-binding
ligand), Equation 5 shows how the functional affinity K can vary with both the amount and type of allosteric co-binding ligand. This probe-dependent effect has been demonstrated in numerous experimental situations; for example, the allosteric inhibitor aplaviroc of the receptor CCR5 [chemokine (C-C motif) receptor 5] produces widely divergent effects on the affinity of chemokines from a greater than 30-fold decrease in the affinity of the receptor for CCL3 [chemokine (C-C motif) ligand 3] to nearly no effect on the affinity of the chemokine CCL5 [15]. Several allosteric ligands produce effects on ligand receptor affinity through interaction at extracellular sites but there is no a priori reason to suppose that these actions cannot be mediated via the cytosol. For example, a series of CCR4 and CCR5 antagonists have been shown to produce allosteric blockade of chemokine receptors by binding to sites inside the cell [16]. Similarly, signaling proteins such as G proteins or b-arrestin will function as allosteric modulators of the receptor, and it follows that the affinity of the receptor for different ligands (e.g., agonists) can differ by varying a-values as different signaling proteins associate with the receptor. This effect is dramatically illustrated by the difference in the X-ray crystallographic structure of the b2-adrenoceptor bound and not bound to a nanobody simulating a G protein [17,18]. Similarly, differences in the nature of the affinity changes with different signaling co-binding proteins are observed for ghrelin receptors in lipid discs – clear changes in the conformations of the receptors have been demonstrated following addition of Gq to nanodiscs [19]. Studies using exponential fluorescent lifetime decay analysis also show the creation of different conformations upon addition of b-arrestin [19]. k-Opioid receptor SCAM (substituted cysteine accessibility method) studies indicate changes in conformation in 7TM domains 6 and 7 with binding of Ga16 and/or Gai2 G protein subunits, and this is reflected in an 18-fold change in affinity for the ligand salvanoran [20]. In general, there is abundant evidence to show that receptors can complex with membrane proteins to yield multiple affinity states, a condition Black and Shankley referred to as ‘receptor distribution’ [21]. These types of data underscore the impact of membrane components on agonist affinity for receptors. Membrane components can alter antagonist affinity as well. For example, the membrane protein RAMP3 (receptor activity modifying protein [22]) produces dramatic effects on the blockade of human calcitonin receptor by the antagonist peptide AC66 [23]. The Schild regressions shown in Figure 1A,B show how cotransfection of RAMP3 produces a 10-fold decrease in the affinity of the antagonist AC66 in blocking the effects of amylin, but has no effect on the antagonism of human calcitonin; this illustrates both the allosteric control of antagonist affinity of a receptor for ligands by a membrane-bound species and also, the probedependence of that allosteric control. Functional affinity This discussion will define the term ‘functional’ affinity as being the KA value – defining the strength of binding of a ligand as it forms a complex with the receptor to both induce a cellular response and also to interfere with the 435
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Key: No RAMP
2.5
+ RAMP3
Log (DR-1)
Log (DR-1)
2.5
1.2
Agonist = hCalcitonin 3
2 1.5 1
Key: No RAMP + RAMP3
2 1.5 1
0.5
0.5
0
0 –10 –9.5
–8.5
–7.5
Log [AC66]
–9
–8
–7
Log [AC66]
Fraconal maximum response
Agonist = Amylin 3
TRENDS in Pharmacological Sciences
effects of another ligand as it attempts to co-bind to the receptor to induce a physiological response. In the latter case, several published pharmacological procedures are available to measure this value for different agonists, including orthosteric simple competitive antagonists (the Schild method [24]), non-competitive orthosteric antagonists (fitting to the non-competitive model [25]), partial agonists (Schild method, method of Stephenson [26] and method of Kaumann and Marano [27]), inverse agonists (Schild method), and allosteric antagonists (direct fitting to the allosteric model [28]). In terms of the production of response, the amount of ligand–receptor complex formed by an agonist is dictated by the KA; a useful model that can be used to relate ligand affinity and tissue response is the Black/Leff operational model [29]: Res ponse ¼
½AtEm ½Að1 þ tÞ þ K A
[6]
The efficacy of the ligand and the efficiency of stimulus transduction of the cell is denoted by the term t; Black and Leff stipulated that the KA is the reciprocal of the affinity of the ligand for the receptor that is used to determine the amount of drug–receptor complex transduced by the cell to process response [29]. From Equation 6 it can be shown that [30]: EC50 ¼
KA 1þt
[7]
Equation 7 demonstrates that as the response capability of the cell is reduced (i.e., reduction of receptor density) then t ! 0 and the EC50 ! KA. Therefore, the EC50 of agonists producing a low level of response closely approximates the KA. Thus, at least for agonism, the KA utilized in the Black/Leff model is the concentration of agonist around which agonist concentration–response curves collapse upon diminution of the receptor density (Figure 2). If the Black/Leff model is used for any procedures to quantify agonism, then this prerequisite must be met. It can be shown that the functional KA value (measured by low level activation EC50 measurements) can be differentially modified by various allosteric co-binding guests. For example, 436
0.8 0.6
x10–1
[Rt] = 1
x10–2
0.4
x10–3 0.2
KA 0 –9
–8
–7
–5
–6
–4
–3
Log [agonist] TRENDS in Pharmacological Sciences
Figure 2. The relationship between agonism and agonist affinity (denoted as the KA, equilibrium dissociation constant of the agonist–receptor complex) and receptor density as described by the Black/Leff operational model. The model dictates that, as receptor density is diminished, the concentration–response curves collapse upon the KA on the abscissal axis. If the Black/Leff model is used to describe agonism, then this prerequisite must be met.
Figure 3 shows the different EC50 values for the chemokine agonist CCL3L1 for activation of the CCR5 receptor when modified by two different allosteric modulators [31]. Thus, as the response capability of the system is reduced (i.e., conditions where EC50 ! KA), different co-binding ligands can produce different limiting EC50 values. This brings into question whether the low-level agonism EC50 values for a ligand can be considered the functional affinity of that ligand as it antagonizes other orthosteric ligands. Figure 4 shows that this can be shown to be the case for the a1-adrenoceptor in rat anococcygeus muscle; the pEC50 of 6.5 also corresponds to the pKB for oxymetazoline after alkylation of receptors and utilization of oxymetazoline as an antagonist with the Schild method [32]. This
Fracon CCR5 internalizaon
Figure 1. Control of antagonist binding affinity by membrane bound RAMP3 (receptor activity modifying protein 3). Schild regressions to the calcitonin antagonist AC66 show a 10-fold difference in the affinity of the antagonist in cells that do and do not co-express RAMP3 with the calcitonin receptor when amylin is the agonist. Allosteric probe dependence is shown by the fact that no difference is seen when calcitonin is used as the agonist. Data redrawn from [23].
1
1.2 1
CCL3L1 TAK779
α = 0.1
0.8 0.6
β=0
TAK779 0.4 Aplaviroc
CCL3L1
0.2 Aplaviroc
0 –11
–10
–9
Log [CCL3L1]
–8
α = 0.9
–7
β=0 TRENDS in Pharmacological Sciences
Figure 3. Differential effects of two allosteric antagonists (TAK779 and aplaviroc) on the concentration-response curve describing CCR5 [chemokine (C-C motif) receptor 5] internalization to the chemokine CCL3L1 [chemokine (C-C motif) ligand 3-like 1] in U373-MAGI-CCR5-E cells. The fact that the curves collapse around different KA values for the agonist demonstrates allosteric probe dependence and the difference in the effect of each allosteric antagonist on the affinity of CCL3L1. The effects are consistent with a 10-fold decrease in affinity for TAK779 and only a 1.1-fold decrease for aplaviroc. The allosteric effects on the receptor are defined as those changing the affinity of the agonist (denoted a) and those changing the efficacy of the agonist (denoted b) from the model of functional allosterism [55– 57]. Data redrawn from [31].
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
1.2
2
Norepinephrine
1
1.5
0.8
Log (DR-1)
Fraconal maximum response
Opinion
0.6 0.4
Oxymetazoline
1 0.5
0.2 0
0 –8
–7
–6
–5
–4
–7
Log [agonist]
–6
–5
Log [oxymetazoline] TRENDS in Pharmacological Sciences
Figure 4. Contractile effects of a-adrenoceptor agonists (norepinephrine and oxymetazoline) in rat anococcygeus muscle. The panel on the left shows that, in tissues alkylated with phenoxybenzamine (POB) to remove a portion of the aadrenoceptors (POB 0.1 mM, 10 min), oxymetazoline is a partial agonist with a pEC50 of 6.5. Removal of a greater portion of the receptor through controlled alkylation of receptor with phenoxybenzamine (0.3 mM, 10 min) produces a tissue that responds to norepinephrine but not to oxymetazoline. A Schild regression for oxymetazoline as an antagonist in these tissues yielded a pKB value of 6.5 (95% confidence limits 6.3–6.8). Data redrawn from [32].
emphasizes the link between ligand affinity and low response level EC50 values. Measuring agonist bias The term ‘bias’ will be used to denote the preferential ability of a ligand to cause the receptor to interact with a distinct signaling pathway in the cell as opposed to other pathways. It will be seen that the molecular determinants of bias are both the efficacy and affinity of the ligand for the receptor as it interacts with these pathways. The current model for biased signaling is based on the notion that different agonist molecules stabilize different receptor conformations interacting with multiple cytosolic guests [33] and that these go on to interact with signaling proteins to produce differential activation of signaling pathways [34]. The original model proposed ligand-based stabilization on theoretical grounds which has subsequently been verified experimentally with 19 F nuclear magnetic resonance (NMR) studies [35]. However, the original model did not adequately consider the guest-stabilization effect of signaling molecules as shown in X-ray crystallographic work [17,18] (Figure 5). In general, biased signaling should be considered to be a function of the Evoluon of the biased signaling model (A)
G proteinselecve
β-Arresnselecve
(B)
G proteinselecve
β-Arresnselecve
TRENDS in Pharmacological Sciences
Figure 5. Two schematic versions of the molecular model of biased signaling proposed in [34]. (A) Direct ligand stabilization of receptor states that then have a preference for signaling pathways as supported by 19F NMR data [35]. (B) Receptor states pre-stabilized by interaction with signaling molecules as suggested by X-ray crystallographic data [17,18].
ternary complex of ligand, receptor, and signaling molecule rather than being only a ligand–receptor effect [36]. This being the case, the bias of various molecules will necessarily vary with signaling pathway. The basis for the quantification of agonist bias is the calculation of the relative efficacy of a given agonist for stimulation of different cytosolic pathways. The currently known methods to calculate this are: (i) Transducer coefficients: DDlog(t/KA) values are obtained through application of the Black/Leff model for each pathway [37]. (ii) Relative activity (RA) values: DDlog(RA) values (RA = maximal response/EC50) are calculated for each pathway [38–40]. It should be noted that for concentration response curves of slope = 1, DDlog(RA) values are identical to the DDlog(t/KA) values [39]. (iii) blig values [41]: DDlog(t) values are calculated with the Black/Leff model from estimates of KA. Although the affinity used to obtain t values is not specified in the description of the method, if the Black/Leff method is used to obtain t, then by definition the functional affinity must be used (KA as shown in Figure 2). It is important to note that every currently known method of quantifying biased agonism utilizes the functional affinity of the agonist (KA) – either through application of the Black/Leff model (through t values [37,41]) or through the functional EC50 (through RA values [38–40]). This raises into question the ad hoc application of affinity measurements made in binding studies to the calculation of agonist bias. Binding estimates have been shown to differ from functional affinity measurements in many cases (for the reasons discussed above). For instance, whereas the binding affinity for 125I-human calcitonin in HEK 293 cells stably expressing human calcitonin receptors is 16 pM (pKd = 10.77 with 95% confidence limits of 10.63–10.91), the EC50 for calcium responses for human calcitonin is 426fold higher (EC50 = 7.2 nM, pEC50 = 8.14 + 0.2), a difference in the opposite direction from that expected to be produced by efficacy [42]. There is no a priori reason that an affinity measured in binding should accurately reflect an agonist response controlling functional affinity. For the quantification of biased signaling, a functional estimate of affinity must be applied either through application of an observed partial agonist EC50 or fitting of the Black/Leff operational model to the functional data. In the case of full agonists, no unique value of KA can be obtained, but the method yields ratios of t/KA for full agonists that are unique – and these can be applied to the calculation of DDlog(t/KA) for measurement of bias. Alternatively, the method of Barlow, Scott, and Stephenson [43] can be applied to obtain Dlog(t/KA) values directly when full and partial agonists are compared (Box 1). Considering that the functional affinity of a ligand can be estimated through a low-level agonist EC50 value in a functional assay (Figure 4), then the question can be asked – is there evidence to show variation in functional affinity with signaling pathways; in other words, does the affinity of an agonist for a receptor vary in response to the signaling pathway interacting with that receptor? Figure 6 shows that the EC50 for the 5-HT2A agonist MK212 in mouse 437
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Box 1. Dlog(t/KA) values through curve comparison. The method of comparing concentration response curves to two agonists devised by Barlow, Scott, and Stephenson [43] utilizes a double reciprocal relationship between equiactive concentrations of a full and partial agonist. Thus, considering two agonists A1 and A2, the following null relationship can be derived: 1 1 t A1 K A-2 t A1 -t A2 ¼ þ ½A1 ½A2 t A2 K A-1 t A2 K A-1
[I]
It can be seen that the slope of resulting linear regression of 1/[A1] upon 1/[A2] yields the ratio of the t/KA values for both agonists, the logarithm of which is Dlog(t/KA). Figure I shows chemokine CCL3L1 as A1 and CCL5 as A2; the linear plot indicates a value for Dlog(t/KA) for these agonists as 0.24. Data from [37].
(B)
100 90 80
50 40
1.8 1.4
CCL3L1
70 60
1/[A1] × 109
% Max. internalizaon
(A)
CCL5
30 20 10
1.0 0.8 0.6 0.4
0 -10
-9 -8 Log [Agonist]
-7
-6
Slope= ΔLog(τ/KA)= 0.24 2
4 6 1/[A2] × 108
8
TRENDS in Pharmacological Sciences
Figure I. A. Agonist dose-response curves for CCL3L1 and CCL5 for internalization of CCR5 receptors. Equiactive concentrations of agonist (denoted by broken lines) are compared in a double-reciprocal plot (B) to yield a straight line according to equation I in the Box. The slope of this line is the ratio of Log(t/KA) values for the agonists. Data redrawn from [37].
embryonic fibroblast cells is 100 nM for activation of ERK (extracellular signal-related kinase), but 2 mM for the production of inositol phosphate [44]. Owing to the fact that the maximal response for both of these effects are below the system maximal response, it is mathematically
Percentage maximal response
120 100
IP
80
ERK
60 40 20 0 –8
–6
–4
–2
–20
Log [MK212] TRENDS in Pharmacological Sciences
Figure 6. 5-HT2A (5-hydroxytryptamine 2A receptor)-mediated accumulation of inositol phosphate (IP) and extracellular signal-regulated kinase 1/2 (ERK1/2) phosphorylation in mouse embryonic fibroblasts as a function of concentrations of MK212 [6-chloro-2-(1-piperazinyl)pyrazine hydrochloride]. Broken lines show the expected curves when receptor density is diminished (collapse of the curve to the KA value as shown in Figure 2). Because the maximal responses to both pathways are submaximal, the EC50 values closely approximate to KA values, and these differ with the pathway activated. There is no single value for KA that can be used in the Black/Leff operational model that can fit both concentration–response curves. Data redrawn from [44].
438
impossible for a single estimate of the KA to fit Black/Leff operational model to these data; in other words, there must be different values of KA for MK212 interaction with the 5-HT2A receptor as it interacts with the two signaling pathways. Similar significant differences in EC50 values for partial agonists for a receptor activating different signaling pathways have been shown for m-opioid receptors in HEK293 cells [45] and U2OS OPRM cells [46], and for histamine H4 receptors in U2OS cells [47]. These data show that not only can ligands differ in the efficacies with which they interact with different signaling pathways, but also in their affinity for the receptor as it interacts with those pathways. Thus, in addition to biased agonism, this opens the possibility of therapeutically applicable biased antagonism (vide infra). In terms of agonist signaling bias, the independence of affinity and efficacy with different cellular activation can still yield a range of molecules with identical bias but differential capability to produce agonism. To describe biased agonists fully, the relative efficacy of the agonists for each pathway must be considered as well as the differential propensity to activate each pathway (bias). One way to do this is to plot some measure of the relative efficacy of the agonist for each pathway (as for example through the log of the relative maximal effects) as a function of its bias for each pathway. In this regard, bias is calculated through DDlog(t/KA) or DDlog(RA) values which consider both efficacy and affinity. Specifically, the efficacy of the agonists for various pathways will determine whether or not agonism for that pathway will occur, whereas the bias [i.e., DDlog(RA)] will determine the relative potency of the agonist for each
Opinion
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
Albuterol
0.5
Fenoterol 1.0
0.6
Ritodrine
1.5 1.3
Orciprenaline
Clenbuterol
LOG (relave maxima)
1.0
1.1
0.4
0.9 0.7 1.0
Epinephrine 0.5
0.8 0.8
Dobutamine
Cimeterol
0.3 0.1 1
–0.5
–0.1 0
0.5
1
1.5
–0.3
LOG (bias) –0.5 TRENDS in Pharmacological Sciences
Figure 7. Combinations of signaling bias (abscissae as DDlog(RA) values) and relative efficacy (ordinates as logarithms of ratios of maximal responses) for b1-adrenoceptor agonists causing the receptor to interact with two signaling pathways (unbroken lines, Gs protein activation; broken lines, b-arrestin interaction; RA, relative activity). Ligands can have similar bias (Ritrodine and Dobutamine) but very different signaling properties in terms of activation. The reference agonist for calculation of bias is epinephrine. The straight broken line reflects a total dependence of bias on efficacy alone with no contribution from affinity. Data from [48].
pathway response when agonism does occur. Some nuances between agonism and bias are illustrated with the series of b1-adrenoceptor agonists shown in Figure 7. This figure expresses the relative maximal responses of the agonists for G protein versus b-arrestin activation (ordinate values) as a function of the relative bias for each pathway compared to epinephrine calculated with DDlog(RA) values [48]. The relative scatter around the broken correlation line is noteworthy from the point of view that, if bias were solely a function of the relative efficacy of the agonists for each pathway, then a better correlation would be expected as log(relative maximum) and log(bias) would have the same origin (namely efficacy, t values). The fact that the data vary considerably from the linear broken regression line suggests that variable affinity plays a role in the bias. In addition, exemplified by the pattern shown in Figure 6, there are agonists for which it is mathematically impossible to fit both signaling curves with a single estimate of affinity (i.e., orciprenaline). Finally, it can be seen that agonists with nearly identical bias values can still have different signaling characteristics (i.e., potency vs maximal effects) as shown by the profiles for ritrodine and dobutamine. Such descriptions assist in developing molecules for different needs; in other words, whether the desired bias is to activate or block selectively a given signaling pathway. If selective activation of a pathway is the aim, then the efficacy of the molecule is the main determinant of activity, not bias. It should be noted that, although biased agonism has been suggested as the
most commonly cited reason for therapeutic improvement of a given receptor ligand, there is no reason that selective antagonism should not be more important in some cases. For example, the crucial property of the advanced angiotensin receptor biased ligand TRV000120 is its blockade of elevated endogenous angiotensin in heart failure rather than the added b-arrestin signaling [49]. The array of different combinations of relative potency and relative efficacy of a biased ligand underscores the variety of possible profiles anticipated in vivo with biased ligands. In many cases it would also be anticipated that differences in receptor densities of different organs will produce a mixture of profiles for the same ligand in vivo. Biased antagonism Biased antagonism is an established phenomenon for negative allosteric modulators. For example, the allosteric antagonist LP1805 [N,N-(2-methylnaphyl-benzyl)-2-aminoacetonitrile] converts normal signaling by NK2 receptors for the natural agonist neurokinin A (activation of Gaq and Gas) to a biased signal of enhanced Gaq but diminished Gas [50]. Similarly, the allosteric antagonist indole1 (Na-tosyltryptophan) changes normal prostaglandin D2 (PDG2) signaling of the chemoattractant receptor-homologous molecule expressed on TH2 cells (CRTH2) receptor (Gai and b-arrestin association with the receptor) to an exclusive Gai effect with no b-arrestin signal [51]. However, the situation is not as clear for orthosteric antagonists (or 439
Opinion biased agonists binding in low-sensitivity systems to induce antagonism of endogenous signaling). Theoretically, there is no reason to assume that the formation of different receptor–signaling protein complexes to yield variable affinities will not occur; this idea has been proposed in the literature. As stated by Baker and Hill: ‘. . . it is therefore possible that antagonists might differ in their affinity for different GPCR-signaling protein complexes (e.g., Gs- and barrestin-coupled forms of the b2-adrenoceptor) and show agonist- and signaling-pathway-dependent affinities..’ [52]. Variable orthosteric affinity for b-adrenoceptor antagonists has been demonstrated; for instance, the affinity of propranolol differs by a factor of 52 when used to block [3H]CGP12177 binding (pKB = 6.44) versus CGP12177 function (pK = 8.2) [53,54]. It should be noted that differential affinity due to signaling may not produce large effects (two to fivefold differences), but these variations may still obscure signaling bias significantly. In general, it can be questioned to what extent biased ligands will be encountered in pharmacology. From a theoretical point of view, if it is accepted that receptors exist in conformational ensembles and that ligands create new ensembles based on their differential affinity for the various conformational states; it is nearly inconceivable that two ligands would produce identical ensembles of a receptor protein – in other words, all ligands will therefore be biased to some extent. However, whether such a bias will have pharmacological consequences is dependent upon the nature of the various conformations that are stabilized. Assuming that only a select number of conformations couple to cellular proteins to cause an observed cellular effect, the conformational ensembles produced by many ligands may not result in an observed biased signaling profile. In addition, the likelihood of a given conformation emerging to cause an observed signaling bias depends upon the relative KA values for the various conformations in the ensemble. Another logical question raised by these ideas concerns to what extent biased affinity will be observed in the setting of antagonism? Equation 5 predicts that it will basically be operative only in cases where the ligand stabilizes a receptor conformation with preferential affinity for a co-binding ligand (i.e., the magnitude of a is >>1). Given this, it might be supposed that biased affinity will be operative only in cases where the ligand also signals toward the preferred pathway (i.e., t >1 in Equation 7) and this, in turn, would suggest that ligands with preferential affinity and low efficacy for a given pathway may not often be encountered. This latter question will not be resolved until careful quantitative studies of the trafficking of receptor antagonism to various signaling pathways are available. Another frontier in the therapeutic application of biased ligands will be the transference of in vitro estimates of bias (i.e., DDlog(t/KA) or DDlog(RA) values) in systems where distinct interactions of receptors and coupling proteins can be measured (i.e., within the allosteric vector [12]) versus where the stimulus escapes the allosteric vector and the cell type contributes to selective signaling. In this latter instance, the signaling bias can vary dramatically and become cell type-dependent; the availability of increasing 440
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
numbers of biased ligands will allow the assessment of this potential problem. Concluding remarks There are theoretical and practical reasons for not utilizing binding affinities as descriptors of functional effects. In addition, functional experiments can provide more accurate estimates of functional affinity which, in turn, are more relevant to functional receptor activity. Specifically, the functional KA value for partial agonists can be identified through fitting of the Black/Leff operational model, and t/KA ratios are identified for full agonists with the same model. Given this, binding experiments are an unnecessary, and possibly misleading, burden in assessing functional bias. Finally, the same ideas that describe signaling bias also predict biased antagonism, an effect thus far relatively unexplored in pharmacological drug development. It will be interesting to see to what extent this class of molecule can be exploited therapeutically should biased antagonism be determined to play a practical role in drug discovery.
References 1 Hill, A.V. (1910) The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J. Physiol. 40 (Suppl.), iv–vii 2 Langmuir, I. (1916) The constitution and fundamental properties of solids and liquids. Part I. Solids. J. Am. Chem. Soc. 38, 2221–2295 3 Colquhoun, D. (2006) The quantitative analysis of drug–receptor interactions: a short history. Trends Pharmacol. Sci. 2, 149–157 4 Fraunfelder, H. et al. (1988) Conformational substrates in proteins. Annu. Rev. Biophys. Biophys. Chem. 17, 451–479 5 Fraunfelder, H. et al. (1991) The energy landscapes and motions of proteins. Science 254, 1598–1603 6 Hilser, J. et al. (1998) The structural distribution of cooperative interactions in proteins: analysis of the native state ensemble. Proc. Natl. Acad. Sci. U.S.A. 95, 9903–9908 7 Hilser, V.J. et al. (2006) A statistical thermodynamic model of protein ensembles. Chem. Rev. 106, 1545–1558 8 Onaran, H.O. and Costa, T. (1997) Agonist efficacy and allosteric models of receptor action. Ann. N.Y. Acad. Sci. 812, 98–115 9 Onaran, H.O. et al. (2002) A look at receptor efficacy. From the signaling network of the cell to the intramolecular motion of the receptor. In The Pharmacology of Functional, Biochemical, and Recombinant Systems (Handbook of Experimental Pharmacology, Vol. 148) (Kenakin, T.P. and Angus, J.A., eds), pp. 217–280, Springer 10 Kenakin, T.P. and Onaran, O. (2002) The ligand paradox between affinity and efficacy: can you be there and not make a difference? Trends Pharmacol. Sci. 23, 275–280 11 Colquhoun, D. (1985) Imprecision in presentation of binding studies. Trends Pharmacol. Sci. 6, 197 12 Kenakin, T. and Miller, L.J. (2010) Seven transmembrane receptors as shapeshifting proteins: the impact of allosteric modulation and functional selectivity on new drug discovery. Pharmacol. Rev. 62, 265–304 13 Stockton, J.M. et al. (1983) Modification of the binding properties of muscarinic receptors by gallamine. Mol. Pharmacol. 23, 551–557 14 Ehlert, E.J. (1988) Estimation of the affinities of allosteric ligands using radioligand binding and pharmacological null methods. Mol. Pharmacol. 33, 187–194 15 Watson, C. et al. (2005) The CCR5 receptor-based mechanism of action of 873140, a potent allosteric non-competitive HIV entry-inhibitor. Mol. Pharmacol. 67, 1268–1282 16 Andrews, G. et al. (2008) An intracellular allosteric site for a specific class of antagonists of the CC chemokine G protein-coupled receptors CCR4 and CCR5. Mol. Pharmacol. 73, 855–867 17 Rasmussen, S.G.F. et al. (2011) Crystal structure of the b2 adrenergic receptor-Gs protein complex. Nature 477, 549–555
Opinion 18 Rasmussen et al. (2011) Structure of a nanobody-stabilized active state of the b2 adrenoceptor. Nature 469, 175–180 19 Mary, S. et al. (2012) Ligands and signaling proteins govern the conformational landscape explored by G protein-coupled receptor. Proc. Natl. Acad. Sci. U.S.A. 109, 8304–8309 20 Yan, F. et al. (2008) Ga-subunits differentially alter the conformation and agonist affinity of k-opioid receptors. Biochemistry 47, 1567–1578 21 Black, J.W. and Shankley, N.P. (1990) Interpretation of agonist affinity estimations: the question of distributed receptor states. Proc. R. Soc. Lond. B: Biol. Sci. 240, 503–518 22 Hay, D.L. et al. (2006) GPCR modulation by RAMPS. Pharmacol. Ther. 109, 173–197 23 Armour, S. et al. (1999) Pharmacological characterization of receptor activity modifying proteins (RAMPs) and the human calcitonin receptor. J. Pharmacol. Toxicol. Methods 42, 217–224 24 Arunlakshana, O. and Schild, H.O. (1959) Some quantitative uses of drug antagonists. Br. J. Pharmacol. 14, 48–58 25 Gaddum, J.H. et al. (1955) Quantitative studies of antagonists for 5hydroxytryptamine. Q. J. Exp. Physiol. 40, 49–74 26 Stephenson, R.P. (1956) A modification of receptor theory. Br. J. Pharmacol. 11, 379–393 27 Kaumann, A.J. and Marano, M.M. (1982) On equilibrium dissociation constants for complexes of drug receptor subtypes: Selective and nonselective interactions of partial agonists with two b-adrenoceptor subtypes mediating positive chronotropic effects of () isoprenaline in kitten atria. Naunyn Schmiedebeberg’s Arch. Pharmacol. 219, 216–221 28 Kenakin, T.P. (2012) Biased signaling and allosteric machines: new vistas and challenges for drug discovery. Br. J. Pharmacol. 165, 1659– 1669 29 Black, J.W. and Leff, P. (1983) Operational models of pharmacological agonist. Proc. R. Soc. Lond. [Biol.] 220, 141–162 30 Black, J.W. et al. (1985) An operational model of pharmacological agonism: The effect of E/[A] curve shape on agonist dissociation constant estimation. Br. J. Pharmacol. 84, 561–571 31 Muniz-Medina, V.M. et al. (2009) The relative activity of ‘function sparing’ HIV-1 entry inhibitors on viral entry and CCR5 internalization: Is allosteric functional selectivity a valuable therapeutic property? Mol. Pharmacol. 75, 490–501 32 Kenakin, T.P. (1984) The relative contribution of affinity and efficacy to agonist activity: organ selectivity of noradrenaline and oxymetazoline with reference to the classification of drug receptors. Br. J. Pharmacol. 81, 131–141 33 Kenakin, T.P. and Morgan, P.H. (1989) Theoretical effects of single and multiple transducer receptor coupling proteins on estimates of the relative potency of agonists. Mol. Pharmacol. 35, 214–222 34 Kenakin, T.P. (1995) Agonist–receptor efficacy II: agonist-trafficking of receptor signals. Trends Pharmacol. Sci. 16, 232–238 35 Liu, J.J. et al. (2012) Biased signaling pathways in b2-adrenergic receptor characterized by 19F-NMR. Science 335, 1106–1110 36 Onaran, H.O. and Costa, T. (2012) Where have all the active receptor states gone? Nat. Chem. Biol. 8, 674–677 37 Kenakin, T.P. et al. (2012) A simple method for quantifying functional selectivity and agonist bias. ACS Chem. Neurosci. 3, 193–203
Trends in Pharmacological Sciences September 2014, Vol. 35, No. 9
38 Ehlert, F.J. (2005) Analysis of allosterism in functional assays. J. Pharmacol. Exp. Ther. 315, 740–754 39 Tran, J.A. et al. (2009) Estimation of the relative microscopic affinity constants of agonists for the active state of the receptor in functional studies on M2 and M3 muscarinic receptors. Mol. Pharmacol. 75, 381–396 40 Figueroa, K.W. et al. (2009) Selectivity of agonists for the active state of the M1 to M4 muscarinic receptor subtypes. J. Pharmacol. Exp. Ther. 328, 331–342 41 Rajagopal, S. et al. (2011) Quantifying ligand bias at seventransmembrane receptors. Mol. Pharmacol. 80, 367–377 42 Watson, C. et al. (2000) The use of stimulus-biased assay systems to detect agonist-specific receptor active states: Implications for the trafficking of receptor stimulus by agonists. Mol. Pharmacol. 58, 1230–1238 43 Barlow, R.B. et al. (1967) An attempt to study the effects of chemical structure on the affinity and efficacy of compounds related to acetylcholine. Br. J. Pharmacol. 21, 509–522 44 Strachan, R.T. et al. (2010) Genetic deletion of p90 ribosomal S6 kinase 2 alters patterns of 5-hydroxytryptamine 2A serotonin receptor functional selectivity. Mol. Pharmacol. 77, 327–338 45 McPherson, J. et al. (2010) m-Opioid receptors: correlation of agonist efficacy for signaling with ability to activate internalization. Mol. Pharmacol. 78, 756–766 46 Nickolls, S.A. et al. (2011) Understanding the effect of different assay formats on agonist parameters: a study using the m-opioid receptor. J. Biomol. Screen. 16, 706–716 47 Nijmeijer, S. et al. (2012) Analysis of multiple histamine H4 receptor compound classes uncovers Gai protein- and b-arrestin2 biased ligands. Mol. Pharmacol. 82, 1174–1182 48 Casella, I. et al. (2011) Divergent agonist selectivity in activating a1and a2-adrenoceptors for G-protein and arrestin coupling. Biochem. J. 438, 191–202 49 Violin, J.D. and Lefkowitz, R.J. (2007) b-Arrestin-biased ligands at seven transmembrane receptors. Trends Pharmacol. Sci. 28, 416–422 50 Maillet, E.L. et al. (2007) A novel, conformation-specific allosteric inhibitor of the tachykinin NK2 receptor (NK2R) with functionally selective properties. FASEB J. 21, 2124–2134 51 Mathiesen, J.M. et al. (2005) Identification of indole derivatives exclusively interfering with a G protein-independent signaling pathway of the prostaglandin D2 receptor CRTH2. Mol. Pharmacol. 68, 393–402 52 Baker, J.G. and Hill, S.J. (2007) Multiple GPCR conformations and signaling pathways: implications for antagonist affinity estimates. Trends Pharmacol. Sci. 28, 374–381 53 Baker, J.G. (2005) Site of action of b-ligands at the human b1adrenoceptor. J. Pharmacol. Exp. Ther. 313, 1163–1171 54 Baker, J.G. (2005) The selectivity of b-adrenoceptor antagonists at the human b1, b2 and b3 adrenoceptors. Br. J. Pharmacol. 144, 317–322 55 Kenakin, T.P. (2005) New concepts in drug discovery: collateral efficacy and permissive antagonism. Nat. Rev. Drug Discov. 4, 919–927 56 Ehlert, F.J. (2005) Analysis of allosterism in functional assays. J. Pharmacol. Exp Ther. 315, 740–754 57 Price, M.R. et al. (2005) Allosteric modulation of the cannabinoid CB1 receptor. Mol. Pharmacol. 68, 1484–1495
441