Anda di halaman 1dari 33

Practical Aspects of Radioligand Binding

The technique of radioligand binding has revolutionized the ability to characterize both receptors and the ligands (substrates) that interact with them; in this context, the term receptor is used to define any protein of biological interest that interacts with a ligand (or substrate) that can be radiolabeled. Radioligand binding can be used to characterize receptors in their natural environment (wild type) as well as those transfected into cell lines. It can be used to study receptor dynamics and localization, to identify novel chemical structures that interact with receptors, and to define ligand activity and selectivity. The technique thus serves as a valuable adjunct to other pharmacological and molecular biological approaches. This unit reviews the major approaches to developing a binding assay. A number of other excellent articles dealing with the practical and theoretical aspects of ligand binding assays are also available (Motulsky and Neubig, 1997; Limbird, 1996; Williams et al., 1995; Kenakin, 1993; UNIT 1.1). For many routine binding assays, essentially all that is required is a suitable radioligand and a crude homogenate of a tissue known to contain the receptor. The homogenate and the radioligand are mixed, and at an appropriate time (empirically determined by trial and error) the unbound radioligand (L*, or free) is rapidly separated from the ligand bound to the receptor (L*R, or bound), usually by rapid filtration. Tissue sources in a radioligand binding assay can include tissue slices (UNIT 8.1), subcellular fractions of a tissue, or intact cellular preparations that include native, immortalized, or transfected cells (UNIT 6.3). Historically, the existence of receptors, or specific ligand binding sites, was inferred from pharmacological data (UNIT 1.1). The biochemical demonstration that these low-abundance proteins actually existed required the invention and development of radioligand binding assays. Until the advent of molecular cloning techniques (UNITS 6.1 & 6.3), relatively little was known about the molecular nature of receptors, except for what could be gleaned using radioligand binding. Binding assays have now become routine, and computer programs for analysis of binding data are commonplace (Motulsky and Neubig, 1997). In contrast to other biochemical assays, radioligand binding assays can yield large amounts of data in a very short time (on the order of days) using minimal amounts of tissue and ligand. This advantage has led to the technique of high-throughput screening, a major milestone in the drug discovery process that allows hundreds of thousands of compounds to be evaluated for in vitro activity using a microtiter plate format (Williams and Gordon, 1996). Compounds can currently be run through a battery of more than 80 in vitro radioligand binding assays to assess their selectivity to a variety of receptors, enzymes, and signal transduction targets. These data provide a valuable starting point in assessing functional and in vivo activity, and may provide information on potential side effect liabilities in a candidate molecule. Radioligand binding is also used in combination with autoradiography (UNIT 8.1) to visualize receptors in situ: thin microtome-generated sections of tissue are labeled and juxtaposed to X-ray film to produce photographic images of receptor density. Ligands labeled with short-lived isotopes (e.g., 11F, 13C, or 99Tc) can be used in vivo in both animals and humans, using positron emission tomography (PET) to visualize ligand bound to receptor in living tissue. This technique is used to measure receptor dynamics in various disease states, especially in the central nervous system (CNS); to assess the access of drugs to tissues; and to measure receptor occupancy in real time, which may make it possible to titrate drug efficacy in the clinical setting and avoid side effects. The following series of criteria must be met in order to validate the binding assay (Cuatrecasas and Hollenberg, 1976). 1. Binding should be saturable, indicating a finite number of receptor sites. However, in some instances, nonspecific binding can appear to be saturable (see Binding Specificity, discussion of nonspecific binding behavior). 2. The binding affinity, defined as the dissociation constant (Kd), should be consistent with values determined for physiological receptors (e.g., 100 pM to 10 nM). 3. Binding should be reversible, consistent with a physiological mechanism for terminating the effect of a ligand at the receptor. 4. The tissue and subcellular distribution of the specific binding should be consistent with what is known about the proposed physiological effects of the endogenous ligand, and with

UNIT 1.3

Receptor Binding Contributed by Michael McKinney


Current Protocols in Pharmacology (1998) 1.3.1-1.3.33 Copyright 1998 by John Wiley & Sons, Inc.

1.3.1

what is known about the localization of the receptor. 5. The pharmacology of binding for both agonists and antagonists should be consistent with the pharmacology of the natural ligand in functional and whole-animal test paradigms. By extrapolation, there should also be negative pharmacological data (e.g., ligands that are known not to interact with the targeted receptor should not affect radioligand binding). 6. A simultaneous correlation of binding with biological concentration-response curves in identical tissue preparations should be generated. 7. Activity in a binding assay should be predictive of activity in a relevant animal model of receptor function. In general, items 1 through 5 are part of the process of characterizing a binding assay, while items 6 and 7 address additional properties of compounds examined in the assay. Radioligand binding assays only measure the affinity and density of a ligand binding site. The efficacy and the pharmacodynamic and pharmacokinetic properties of the ligand are not measured, and must be assessed by functional and in vivo analysis of ligand properties. For G proteincoupled receptors, it is possible to assess whether a ligand is an agonist or an antagonist (UNITS 1.1 & 1.2) by conducting a GTP shift experiment. This is not always a robust measure, and it is more useful to assess functional activity using a reporter system (UNIT 6.2) or an intact tissue system (see Chapter 4). The techniques of molecular cloning have provided a considerable amount of additional data in characterizing receptor subtypes, their pharmacological and molecular properties, and their anatomical distribution (UNIT 6.1).

forward rate = k +1 [ L*] [ R ]


Equation 1.3.2

The constant k+1 has units of (time1 concentration1). Generally, this reaction is reversible, with the L*R complex dissociating to reform L* and R. The rate of the reverse reaction is dependent on the amount of L*R and the magnitude of the reverse rate constant (k1). The constant k1 is expressed in units of time1.
reverse rate = k 1 [ L * R ]
Equation 1.3.3

At equilibrium, the forward and reverse reactions are equal in rate. Therefore, the amounts of L*, R, and L*R remain constant. Like a bimolecular chemical reaction, the ratio of the rate constants in a radioligand binding reaction is equal to the thermodynamic equilibrium binding constant (Kd).
Kd = k 1 k +1 = [ L*] [ R ] [L * R]

Equation 1.3.4

FUNDAMENTALS OF RADIOLIGAND BINDING ASSAYS


In its simplest form, the binding of a radioligand to a receptor is analogous to a bimolecular reaction according to the Law of Mass Action. The radioligand (L*) combines with the receptor (R) to form a complex (L*R).
L* + R

k1

k+1

L* R

Equation 1.3.1

Practical Aspects of Radioligand Binding

The rate of the forward reaction (left to right) is determined by the concentrations of L* and R, and by the forward rate constant (k+1), as follows.

The Kd is expressed in molar units of concentration (e.g., nanomolar or picomolar). The binding affinity of a receptor for a ligand is a molecular consequence of its structure, and the Kd is used to identify and classify receptors based on this affinity. Therefore, the determination of Kd is a primary goal in developing a binding assay once the optimal conditions for specific binding (see Binding Specificity) have been established. In the assay, the species measured is the bound ligand (i.e., the L*R complex). The receptor, which is embedded in the plasma membrane, is readily isolated from the aqueous reaction mixture by filtration. By quantifying the radioactivity trapped on the filter, the amount of radioligand bound to the tissue during the incubation is quantified. In equilibrium binding assays, the unbound and bound ligand are separated from each other after the forward and reverse binding reactions achieve equivalent velocities. In kinetic binding assays, the reaction is interrupted at various times during the formation or dissociation of the L*R complex. The Kd value can be determined with either type of assay. If the binding is bimolecular, the Kd will be similar using the two different approaches. The kinetic binding assay also al-

1.3.2
Current Protocols in Pharmacology

30 total binding 25 B max specific binding 20


Bound [ 3 H]NMS (fmol/ tube)

15

10

nonspecific binding

0 0 0.5 1.0 1.5

Free [ 3 H]NMS (nM)

Figure 1.3.1 Saturation binding to muscarinic receptors on N1E-115 mouse neuroblastoma cells. Six concentrations of [3H]N-methylscopolamine ([3H]NMS), with or without 10 M unlabeled NMS, were incubated with 300,000 intact cells/tube for 45 min at 15C before rapid filtration was performed to separate bound from free. The total binding is the sum of the specific and nonspecific binding. Nonspecific binding is defined as the amount of binding found in the tube containing both the radioligand and unlabeled NMS.

lows for the determination of the association and dissociation rate constants (k+1 and k1, respectively). In an equilibrium binding assay, the radioligand is incubated with a receptor source until sufficient time has elapsed for the forward and reverse reactions to attain equal velocities before separating free from bound ligand. With a saturation binding experiment, assays are performed using a series of radioligand concentrations, ranging up to a concentration at which virtually all of the receptors are occupied with ligand. An example of this is shown in Figure 1.3.1, where the binding of [3H]N-methylscopolamine ([3H]NMS), a muscarinic cholinergic receptor antagonist, to muscarinic receptors in N1E-115 mouse neuroblastoma cells is shown. The concentration of [3H]NMS is plotted on the abscissa, and the amount of radioligand bound to the filters at each concentration is plotted on the ordinate. The figure shows total, specific, and nonspecific binding, where total binding is the sum of specific and nonspecific binding (see Binding Specificity).

While a certain amount of nonspecific binding is always present, the more useful radioligands and assays display minimal nonspecific binding (<20% of the total), enhancing the precision of the assay by increasing the signalto-noise ratio. Because there are a finite number of physiologically relevant receptors in the tissue, specific binding becomes maximal (i.e., approaches Bmax) as the concentration of radioligand increases (Fig. 1.3.1). However, nonspecific binding is not saturated with radioligand and therefore continues to increase as a function of radioligand. The concentration of radioligand at which the amount of specific binding is one half the Bmax approximates the equilibrium binding dissociation constant (Kd). However, obtaining the Kd and Bmax from a plot such as that shown in Figure 1.3.1 is inappropriate because it is necessary to locate the top of the curve (Bmax) precisely. To use a graphical method, the data are usually transformed to yield a straight line (see Graphical Methods). However, it is preferable to fit the specific binding data with a

Receptor Binding

1.3.3
Current Protocols in Pharmacology

slope =
B/F

1 Kd

Bmax

Figure 1.3.2 The Scatchard (Rosenthal) plot. The specific bound (B) is plotted on the x axis, and the ratio of specific bound to free (B/F) is plotted on the y axis. The x intercept is the maximal amount of specific binding (Bmax). The slope of the plot is the negative of the inverse of the equilibrium binding dissociation constant (Kd).

Practical Aspects of Radioligand Binding

mathematical model in a computer program to determine these parameters (see Use of Computer Modeling Techniques). The traditional graphical method is the Scatchard (or Rosenthal) plot (Fig. 1.3.2). However, this plot of the concentration of bound/free ligand (B/F) versus the concentration of bound ligand (B) is invalid as both the ordinate and abscissa contain a common term that influences the slope of the plot independently of the data. For this reason, all binding data should routinely be evaluated using a nonlinear regression analysis program (e.g., GraphPad, LIGAND, or EBDA) that allows the computer to fit alternative models to the data. Refined analysis of a saturation isotherm reveals whether the binding represents a simple bimolecular association or is more complex (e.g., if multiple subtypes of receptors have different binding properties or if there is cooperativity between receptors). An alternative to the saturation experiment is the competition (or displacement) binding assay, where increasing concentrations of unlabeled ligands ([D]) compete for the receptor with a fixed concentration of radioligand. The competition between the muscarinic receptor antagonist [3H]quinuclidinyl benzilate ([3H]QNB) and the unlabeled agonist carbamylcholine (carbachol) illustrates this point (Fig. 1.3.3). The displacement assay is more economical than a saturation binding assay

because a single, low concentration of radioligand is used, conserving the expensive radiolabeled compound. When it is known that competition between the two ligands is for a single class of binding site (nominally, the Hill slope = 1; see Graphical Methods), the displacement curve is analyzed by computer to obtain the IC50 (the concentration of unlabeled competitor necessary to displace 50% of the specifically bound radioligand), and this value is mathematically corrected using the Cheng-Prusoff equation (Cheng and Prusoff, 1973; see Graphical Methods, discussion on semilogarithmic plots) to obtain the Ki, the equilibrium dissociation binding constant for the unlabeled species. Because the displacement of [3H]QNB by carbachol in N1E-115 neuroblastoma cells occurs over a concentration range spanning six orders of magnitude, it appears that the interaction between R and L* is quite complex (Fig. 1.3.3). The curve passing through the data is the computer-assisted fit of a receptor model, which suggests that there are three separate binding sites for this radioligand in these cells. Kinetic radioligand binding assays are used to determine the association and dissociation rate constants (k+1 and k1, respectively). To obtain the k+1 value, known amounts of the ligand and receptor are incubated together and the amount of binding is determined as a function of time. To determine the k1, the dissociation of ligand from the receptor is monitored

1.3.4
Current Protocols in Pharmacology

60 50
[3H]QNB bound (fmol/million cells)

40 30 20 10 0

log[carbachol] (M)

Figure 1.3.3 Competition between a radiolabeled antagonist and an unlabeled agonist for muscarinic receptors on N1E-115 cells. [3H]Quinuclidinyl benzilate ([3H]QNB; 0.2 nM) was incubated with 200,000 intact cells/tube and various concentrations of carbachol for 75 min at 15C, and the suspensions were rapidly filtered to terminate the reactions. The level of nonspecific binding was determined using 1 M atropine.

over time. For examples of kinetic experiments and their analysis, see Determination of Rate Constants. When the L*R binding reaction is simple bimolecular, the semilogarithmic plots of transformed binding data are linear, and the slopes of such plots are used to determine the k+1 and k1 values. When the ratio of the two rate constants, determined by kinetic experiments, is found to approximate a value of Kd that has been determined independently in an equilibrium binding assay, there is evidence that the ligand-receptor reaction is simple bimolecular. In cases where the receptor-ligand reaction is more complex, it is necessary to use a computer with an appropriate model to determine rate and equilibrium constants (see Use of Computer Modeling Techniques).

GUIDELINES FOR ESTABLISHING A RADIOLIGAND BINDING ASSAY Radioligand Selection


The selection of the radioligand depends on ligand stability, specific activity, and pharmacological selectivity. It is always preferable, when possible, to use an antagonist radioligand that is able to recognize and bind to the receptor regardless of whether the receptor is coupled to membrane-associated signal transduction ele-

ments (e.g., G proteins) or is in a desensitized state. Antagonists bind to receptors with much higher affinity than do agonists. Additionally, agonists induce conformational changes in receptor-effector complexes that can cause ligand-receptor complexes to exist in multiple states with different binding potencies. Another advantage of antagonist radioligands is that they do not activate the receptor, which, in the case of binding with metabolically active cells, could result in the desensitization of the site. Agonists tend to label only a portion of the actual receptor present. Almost all receptors exist at concentrations of less than 10 pmol/mg protein, and usually considerably less. For these reasons, most binding assays are designed for labeled antagonists. In many cases, the radioligand is commercially available from either NEN Life Sciences or Amersham. For many receptors there are several radioligands that may be used. The radioligand is usually labeled with tritium (3H), although 32P, 33P, or 35S can also be used (Filer et al., 1989). With peptides and proteins, tyrosine residues can be labeled with 125I or 123I. The effect of incorporation of a radioactive moiety must be evaluated to determine that the pharmacological properties of the ligand are not significantly altered. This is a particular

Receptor Binding

1.3.5
Current Protocols in Pharmacology

concern with iodine radionuclides because of their size. The specific activity of the isotope, measured either in bequerels (Bq, where 1 Bq = 1 disintegration/sec) or in Curies/mmol (Ci/mmol, where 1 Ci = 3.7 1010 Bq), must be sufficiently high to allow detection of lowabundance binding sites. In general, compounds with a specific activity of <20 Ci/mmol do not make good radioligands. Producing 14Clabeled ligands of sufficiently high specific activity is expensive and difficult, so they are rarely used. After labeling, the radioligand must be purified. It is prudent to routinely verify the purity of the starting material, whether it is purchased or synthesized in the lab (Filer et al., 1989). Stability The level of purity of the radioligand must be assayed periodically. In some cases the radioligand requires special storage procedures (e.g., the addition of antioxidants or protease inhibitors, in the case of a peptide radioligand) to slow or prevent its degradation. These procedures are similar to those used to prevent the degradation of receptors in the tissue preparation (see Tissue Preparation, discussion on receptor stability). Affinity and pharmacological selectivity The development of a binding assay and its subsequent validation requires the demonstration of high-affinity binding of the radioligand to the receptor. A high-affinity radioligand is desirable because it allows separation of bound from free ligand by filtration. A high-affinity ligand-receptor complex will not dissociate significantly during filtration. For further details on the advantages of filtration over centrifugation as a separation method, see Separation of the Receptor-Ligand Complex (L*R) from Unbound (Free) Ligand. In selecting a radioligand, the primary consideration is its pharmacological selectivity for the receptor of interest, e.g., whether the ligand binds only to the targeted receptor. The majority of commercially available ligands have sufficient selectivity (i.e., they are >100 times more specific for their target receptor than for other sites). The receptor selectivity of the radioligand at the concentration(s) used must be taken into consideration, since the specificity of a compound can vary with concentration. At higher radioligand concentrations, interactions may occur with other, nontargeted receptors. If it is not possible to obtain a selective ligand, it may be possible to block binding of the ligand

to nontargeted receptors with another unlabeled selective compound, thus reducing the inherent complexity of the data analysis. Optically active radioligands If a ligand is racemic containing a chiral center, use of the stereochemically active form of the radioligand is preferable. With few exceptions, most receptors will differentiate between the optical isomers of compounds. That is, the receptor will typically bind one of the isomers with higher affinity than the other.

Tissue Preparation
Tissue disruption and washing Generally, the tissue must be disrupted to increase access of the radioligand to the receptor population. A tissue homogenizer (e.g., Polytron) or a sonicator is usually used to disrupt tissue for use in a binding assay. Typically, the tissue is kept cold or frozen until it is disrupted in the ice-cold homogenizing vessel and buffer. Allowing the temperature to rise, or running the homogenizer/sonicator for too long, can result in significant loss of receptor binding from denaturation or from activation of proteolysis. Tissue disruption is not necessary when performing in situ receptor autoradiography (where thin tissue sections, prepared using a microtome, are used to retain anatomical relationships), or when performing binding assays using intact cellular preparations (e.g., to address regulation of binding sites; UNIT 8.1). Because of diffusion barriers, release of intermediary agents, uptake of radioligand, and other factors, the relative affinity of compounds for binding sites in intact cellular preparations may differ from those found using tissue homogenates, receptors removed from their membrane environment, or those transfected into novel cell systems (Kenakin, 1993). Endogenous ligands for the targeted receptor can interfere with radioligand binding. This is prevented by using a series of washing steps and incubations (with or without degradative enzymes, depending on the ligand in question) to remove the endogenous ligand. Extensive washing (repeated pelleting via centrifugation followed by resuspension in fresh buffer) is required to remove endogenous -aminobutyric acid (GABA). For monoamines, incubation at 37C is required to allow endogenous enzymes to degrade the endogenous ligand (e.g., norepinephrine, dopamine). For adenosine receptors, incubation with the catabolic

Practical Aspects of Radioligand Binding

1.3.6
Current Protocols in Pharmacology

enzyme adenosine deaminase (ADA) is required to remove endogenous adenosine. Similarly, binding to G proteincoupled receptors (GPCRs) can be affected by GTP, which is present in relatively high concentrations in the cell. The nucleotide should be removed from the homogenate by washing to avoid modulation of the binding state of GPCRs. Buffer selection In most cases, a homogenization/assay buffer is selected that yields the highest signal-tonoise ratio for specific and nonspecific binding. As a consequence, almost all binding assays are performed under nonphysiological conditions, in a medium having an ionic strength or pH unlike anything that would exist in vivo. When it is preferable to measure binding under physiological conditions, a solution such as Krebs Ringer or Hanks balanced salt solution (HBSS) may be used. Some receptor-radioligand combinations require special ions. For example, opioid receptor binding is modulated by sodium (UNIT 1.4), GABAA receptor binding is modulated by chloride (UNIT 1.7), and the N-methyl-D-aspartate (NMDA) subtype of the glutamate receptor is modulated by magnesium. Receptor stability Many receptors are stable when frozen in situ. Tissue blocks may be kept at 80C or in liquid nitrogen for long periods of time (months) before use in a binding assay. Binding behavior may change as a consequence of freeze-thaw cycles or long-term storage (>3 months), and the stability of binding must be evaluated as part of the characterization of the binding assay. When tissues are homogenized, the receptors are relatively stable as long as they remain in their membrane phospholipid environment. However, tissue homogenization can liberate high levels of proteases, sometimes requiring the addition of protease inhibitors (e.g., phenylmethylsulfonyl fluoride, an inhibitor of trypsin-like proteases) or calcium-chelating salts (e.g., EDTA or EGTA) to the buffer system in order to preserve the receptor. Likewise, peptide radioligands are susceptible to degradation by proteases in tissue homogenates. When adding protease inhibitors, it is important to ensure that the inhibitors do not themselves interact with the ligand. Proteolysis can also be reduced, but not eliminated, by conducting the assay at lower temperatures. However, reducing the temperature slows reactions; at low temperatures, several hours may

be required for the system to achieve equilibrium. The receptor or ligand may also be susceptible to destruction in an oxidative or reductive environment. For example, it may be necessary to retain the methionine residues in a reduced state by including a sulfhydryl reagent, such as 2-mercaptoethanol or dithiothreitol, in the buffer. Catecholamine ligands are readily oxidized by dissolved oxygen, requiring the addition of ascorbic acid or other antioxidants to enhance their stability.

Binding Assay Conditions


Protein (receptor) concentration The equations used to analyze binding data are based on two assumptions: that the receptor concentration is low and that the free radioligand concentration (F) at the end of the assay is essentially equivalent to the concentration of ligand present at the beginning of the assay. If the receptor concentration is increased such that F changes significantly (by virtue of a significant proportion of the ligand being bound), the receptor binding site affinity will be underestimated. As a general rule, the receptor concentration should be <10% of the radioligand Kd. For example, if the Kd for the radioligand is 1 nM, the assay volume is 1 ml, and the Bmax is 100 fmol/mg tissue, the tissue concentration in the assay should not exceed 1 mg/tube. If it is important to identify the true Kd (that determined at an infinitely low receptor concentration), several saturation binding assays should be performed over a range of tissue concentrations (typically over the range of 100 to 1000 mg tissue/tube). The apparent Kd determined at each tissue concentration is plotted versus the tissue concentration, and the results extrapolated to zero tissue. Example 1: Calculation of the amount of expected binding in a radioligand binding assay The expected amount of ligand binding is dependent in part upon the receptor concentration in the assay. The majority of receptors are present in tissues at a concentration of 10 to 1500 fmol/mg protein. Radioligands are typically labeled to a specific activity of 20 to 60 Ci/mmol, although iodinated radioligands have a theoretical specific activity of 2200 Ci/mmol. In this example, an assay is conducted to detect a receptor present at 500 fmol/mg protein. For each assay, 0.1 mg protein is included in a total assay volume of 1 ml. The specific

Receptor Binding

1.3.7
Current Protocols in Pharmacology

activity of the radioligand is 50 Ci/mmol, the scintillation counter efficiency is 50% (1 cpm/2 dpm), and 2.2 1012 dpm correspond to 1 Ci. If one ligand molecule binds to each receptor, the amount of radioactivity associated with 1015 mole (1 fmol) of receptor molecules is:
50 Ci 1 mmol 2.2 1012 dpm 1 Ci 1 mmol 1012 fmol

1 cpm 2 dpm

= 55 cpm / fmol

Equation 1.3.5

If the binding sites are fully occupied with the radioligand, the amount of specific binding per assay tube is:
500 fmol 1 mg protein 0.1 mg protein tube 55 cpm 1 fmol

= 2750 cpm / tube


Equation 1.3.6

Practical Aspects of Radioligand Binding

Temperature Most receptor proteins, and many proteins involved in signal transduction, are embedded in a lipid bilayer. Conformational changes are dependent in part on the constitution of the lipids in the cell membrane, from both a bulk perspective and a microscopic level. Receptor conformation may also vary as a function of the intrinsic ligand efficacy. The binding of a ligand to a receptor involves the displacement of water molecules from the binding site and the formation of noncovalent bonds between the ligand and receptor. Receptor- and ligand-specific differences exist in the chemical nature of these bonds and the amount of water displaced. Importantly, all these processes are affected by temperature. From a practical viewpoint, it is important to control the incubation temperature to reduce variability in the assay. From a theoretical perspective, the dependence of equilibrium binding on temperature provides insight into the molecular nature of ligand-receptor interactions. Differences in the enthalpy or entropy of binding among receptors in different tissues have been used to support the concept of distinct opioid receptor subtypes (Wild et al., 1994). The thermodynamics of equilibrium binding is also used to differentiate between modes of ligand-receptor interaction. For example, the effect of temperature on -adreno-

ceptor binding is more pronounced with agonists (for which binding is largely enthalpic in nature) than with antagonists (for which binding mainly involves a decrease in entropy; Weiland et al., 1979). Agonist binding to adenosine receptors is entropy-driven, which is interpreted to result from removal of water from the binding pocket (Borea et al., 1996). Agonist binding to the GABAA receptor is also entropydriven. In this case, conformational changes in this multimeric ion channel receptor to produce an active conformation have been used to explain differences between agonists and antagonists (Maksay, 1994). For muscarinic receptors, the hydrogen bonding potential of ligand residues, rather than agonist efficacy, appears to be the main determinant of binding thermodynamics (Waelbroeck et al., 1993). Binding assays can be performed at a variety of temperatures, depending on the requirements of the experiment. Room temperature (25C) is convenient, but temperature variation must be controlled to obtain reproducible results. Membranes undergo a transformation from a liquid crystalline phase to a liquid phase that is determined in part by the content of unsaturated fatty acids and in part by temperature. Typically, the transition temperature is 21 to 22C. Although Arrhenius (or vant Hoff) plots for equilibrium receptor binding to adrenergic (Contreras et al., 1986) or muscarinic (Waelbroeck et al., 1993) receptors do not display a pronounced inflection at this temperature, it is conceivable that some receptor processes are affected by the lipid phase of the membrane. Because variations in room temperature can occur over the time period of the assay, they can skew the results of a given assay, or produce artifactual variability among assays performed over a period of time. Thus, assays performed at room temperature should be maintained in a thermostatically controlled water bath or incubator to stabilize the temperature. In some instances, the assay is performed at physiological temperature (37C), so the results may be compared more readily to in vivo ligand-receptor behavior. In other cases, binding assays may be performed well below room temperature (e.g., on ice) to prevent protein conformational changes such as those associated with receptor desensitization, or to slow the action of proteases. In some instances, the choice of assay temperature and duration may be determined empirically to provide the best signal-to-noise ratio for the particular receptor and ligand combination.

1.3.8
Current Protocols in Pharmacology

Separation of the Receptor-Ligand Complex (L*R) from Unbound (Free) Ligand


Since most of the radioligand remains unbound at the end of a binding reaction, an efficient method for its removal, without the loss or dissociation of the receptor-ligand complex, is necessary to accurately quantify the amount of bound ligand. Separation is usually performed at low temperatures in order to reduce the ligand dissociation rate. Bennett and Yamamura (1986) have characterized various separation methods by considering the typical range of dissociation rates for radioligands. Removal of excess, unbound radioactivity by various washing procedures allows the determination of bound radioactivity directly by conventional spectrometry or by use of scintillation proximity assays (SPA, Amersham). The radioactivity thus determined is considered the total binding. However, in a typical experiment, the radioligand is bound by, adsorbed onto, or sequestered into many sites or compartments (nonspecific binding), in addition to being bound to the receptor target (specific binding). The chemical nature of the radioligand, the method of tissue homogenization, and tissuespecific biochemical processes all influence the degree of nonspecific radioligand binding. Filtration Filtration is the most efficient and convenient method of separating free from bound radioligand, because it requires less handling and manipulation of samples. For instance, using a microtiter-plate format, nearly 300 separate reactions can be terminated and washed in less than 5 min. To terminate the reaction and wash 180 tubes using high-speed centrifugation can take more than 3 hr. Filtration is also preferable to centrifugation because the nonspecific binding is usually lower, as a result of the more thorough washing of the tissue homogenate. After incubation of the ligand with the tissue preparation, the contents of the assay tubes (or microtiter plates) are aspirated onto filters where the tissue particles and bound ligand are trapped, while the unbound ligand passes into the effluent. The filters are then washed repeatedly with cold buffer using a commercially available manifold and house vacuum. While glass fiber filters can be used when assaying solubilized receptors, microfiltration is also possible. Filtration can be used for radioligands with Kd values in the range of 10 to 30 nM and below, since the L*R complex does not disso-

ciate significantly (<10%) during the 15 sec required to rinse the filters four or five times. Centrifugation When radioligand affinity approaches a Kd of >100 nM, rapid centrifugation using either a microcentrifuge or a full-sized centrifuge can provide a reliable estimate of binding. However, it is more difficult to extensively wash the L*R complex to reduce the amount of unbound ligand, so that the level of nonspecific binding is usually much higher than with filtration. Centrifugation binding experiments are typically performed in 1.5-ml plastic microcentrifuge tubes or 15-ml polypropylene tubes. L*R complexes are then pelleted using either a refrigerated high-speed microcentrifuge or a refrigerated full-sized centrifuge. In some instances, the homogenate is layered on top of an oil phase and the experiment terminated by centrifuging the L*R complex through the oil; the unbound radioactivity remains in the aqueous phase above the oil. Microcentrifugation can be completed with a 60sec burst. With a full-sized centrifuge, 5- to 10-min centrifugations are required to effectively pellet the L*R complex, and subsequent careful washing of the pellet (using a syringe and ice-cold buffer) removes excess unbound radioactivity. When using a low-affinity radioligand, the choice between the two centrifugation methods is decided on the basis of the evaluation of ligand binding characteristics. The microcentrifuge oil method can often result in radioligand dissociation as the L*R complex moves through the oil. Other methods More specialized separation methods include the use of column chromatography, selective adsorption with activated charcoal, selective precipitation with salts or an antibody, and dialysis. However, these methods are rarely used today due to their inconvenience.

Binding Specificity
Minimizing nonspecific binding In a typical binding assay, the radioligand will become bound or sequestered into nonreceptor sites. Some nonspecific binding sites are in the tissue preparation, others may be on the filters or centrifuge tubes or pellets used to separate bound from unbound radioactivity. For some ligands, nonspecific binding can be reduced by presoaking filters in 0.1% polyethyleneimine (PEI) at room temperature for

Receptor Binding

1.3.9
Current Protocols in Pharmacology Supplement 8

30 min. Also, certain ligands, especially peptides and proteins, adhere to assay tubes and significantly reduce the concentration of free radioligand, resulting in erroneous data. In some cases, it is possible to prevent this by preabsorbing an excess of unlabeled ligand to the tubes. The best radioligands yield nonspecific binding well below 10% of the total binding when assayed at a radioligand concentration equivalent to its Kd value. However, when the choice of radioligand is limited, it may be necessary to perform experiments with high levels of nonspecific binding. Nonspecific binding behavior and quantitation In some instances, nonspecific binding may appear to be saturable (e.g., when the concentration of unlabeled drug is low, or when the drug interacts with a receptor that is not completely blocked). However, nonspecific binding sites are sometimes described as nonsaturable in the range of ligand concentrations that are used to specifically saturate the targeted receptor (generally <100 nM). The amount of nonspecific binding is then linearly related to ligand concentration. That is, while specific binding describes a hyperbola with an asymptote at the maximal equilibrium concentration of receptor-ligand complex, nonspecific binding appears as a straight line (Fig. 1.3.1). Nonspecific binding must be quantitated at every concentration of receptor and radioligand by including assay tubes that contain, in addition to receptor and radioligand, a 100- to 1000fold excess of an unlabeled ligand known to bind specifically to the targeted receptor. If the unlabeled ligand is present at a concentration sufficient to occupy more than 99% of the receptor, it will prevent the radioligand from binding to the receptor. The remaining bound radioactivity that is measured represents the nonspecific binding. The nonspecific binding at each radioligand concentration is subtracted from the total binding to obtain the specific binding, which is defined as that ligand bound to the pharmacologically relevant receptor (Fig. 1.3.1). Ideally, the unlabeled ligand used to define nonspecific binding should be structurally dissimilar from the radioligand in order to avoid a phenomenon known as isotope dilution. This refers to the situation where an excess of an unlabeled ligand identical in structure to the radioligand provides an apparent reduction in total binding that in reality is only a reduction in specific activity of the radiolabeled species.

Practical Aspects of Radioligand Binding

If assays are conducted without taking into account nonspecific binding, erroneous data can be generated. For many binding assays, specific binding is 70% to 95% of the total binding, providing a good signal-to-noise ratio and reproducibility for the assay. In developing new radioligand binding assays where neither the radioligand nor the assay conditions are optimal, specific binding may be 50% or less, making the assay less reproducible and precise. As the conditions for binding assays are developed empirically and are usually nonphysiological, the process of developing an assay is iterative and dynamic (see Developing a New Binding Assay). When a radioligand is known to bind to more than one receptor and only one of these receptors is targeted in the binding assay, it is sometimes possible to use an unlabeled (cold) compound that is selective for the receptor of interest. When this unlabeled competitor is present at the proper concentration, it displaces the specific binding to the receptor of interest, but not the radioactivity bound either to the other receptors in the preparation or to nonreceptor sites. When measuring nonspecific binding it is advisable to use the lowest possible concentration of unlabeled ligand, because the displacement of radioligand from nonspecific sites or from nontargeted receptors can occur with excessively high concentrations. With saturation binding, the level of nonspecific binding can be assessed mathematically by fitting the data to a model containing terms for both the receptor-dependent binding and the nonspecific binding. The disadvantage of this method is the difficulty in assessing whether nonspecific binding is linear, as it should be. Allosterism, or cooperativity, adds another layer of complexity to binding studies. This occurs when the ligand binds to more than one site on a receptor, or when receptors in close association with one another are influenced by the binding of a ligand to another receptor site. Data analysis of allosterism or cooperativity is different from that which applies to simple competitive interactions. The effect of positive and negative cooperativity (nH > 1 and nH < 1, respectively) on equilibrium binding is shown in Figure 1.3.4 (see Graphical Methods, discussion on Scatchard plots and Figure 1.3.8). Receptors with even a minor difference in composition, such as those which occur in species orthologs, may display marked differences in binding behavior. A classical example of this phenomenon is the case of the human 5-HT1Db and the rat 5-HT1B serotonin receptors. These

1.3.10
Supplement 8 Current Protocols in Pharmacology

1.0

0.8

n H = 1.5 n H = 1.0

Fraction bound

0.6

n H = 0.5

0.4

0.2

0 0 1 2 3 [L]/Kd 4 5

Figure 1.3.4 Effects of allosterism or cooperativity on receptor binding. The x axis is the free concentration of the ligand, expressed in relationship to the Kd value. The y axis is the fractional specific binding to the receptor, where 1.0 is equivalent to full saturation of the binding sites (Bmax). The nH is the Hill coefficient, a number that can be used to express the degree of cooperativity in the binding reaction. When |nH| > 1, the receptors interact with positive cooperativity; when |nH| < 1, the receptors interact with negative cooperativity.

receptors are 93% identical in amino acid sequence, but display different pharmacological profiles. Site-directed mutagenesis of threonine-355 in the human sequence to the asparagine found at this location in the rat sequence yields a mutant human receptor that has the same pharmacological profile as the native rat receptor (Metcalf et al., 1992). Studies with human and rat neurotensin receptors provide another example of this phenomenon (Pang et al., 1996). A -napthalene derivative of neurotensin (8-13) binds more avidly to the rat than to the human ortholog. Computer modeling suggested that the tyrosine in the human receptor sequence was responsible for the difference in binding. Site-directed mutagenesis established that the difference in binding potency was actually due to a tyrosine at residue 339 in the human receptor sequence. Thus, when this residue was mutated to the phenylalanine present in the rat sequence, the binding potency

converted to that displayed with the wild type rat receptor (Pang et al., 1996). Because receptors are composed of L-amino acids, they bind ligands in a stereoselective manner. If the ligand has a chiral center and exists in optical isoforms, the receptor may recognize the enantiomers or diastereomers with differing energies of binding. A recently developed glycine antagonist for the metabotropic glutamate receptor has 16 optical forms, each with distinct binding properties (Pellicciari et al., 1996). However, stereoselectivity is not an absolutely reliable criterion. For instance, the binding of the R- and S-enantiomers of the nicotinic analgesic epibatidine to the 42 form of the nicotinic cholinergic receptor, labeled by [3H]cytisine, does not exhibit stereoselectivity (Sullivan et al., 1994). In a time dependence study, the on- and off-rates for a radioligand should be compatible with the known rate of action of the ligand in
Receptor Binding

1.3.11
Current Protocols in Pharmacology

functional assays. For an agonist ligand that has been characterized extensively in classical dose-response studies, this rate may be known. In this instance, the association rate constant for binding should approximate the value obtained in the functional assay under similar conditions (e.g., buffer composition, temperature). When agonists bind to metabolically active cells, the binding characteristics may be modulated in several ways. The ligand may be accumulated or metabolized by the cells, or may cause desensitization of the receptors. These processes alter the apparent pharmacological characteristics of the L*R interaction (Motulsky et al., 1985).

Developing a New Binding Assay


As noted above, many of the characteristics of a binding assay are highly empirical. Thus, the buffer used, the tissue preparation, the assay temperature, and the duration of the reaction are all derived on a trial-and-error basis. The initial binding experiment described below is designed to test each of these variables. When developing a new binding assay, the ligand must be rigorously checked for purity and stability. It is often prudent to dilute the new radioligand in alcohol and store aliquots at 20C to avoid excessive radiolysis (Filer et al., 1989). The initial binding experiment should use a thoroughly washed tissue homogenate known to contain the receptor target. The tissue may be preincubated alone or with appropriate enzymes or detergents to remove endogenous ligands. Tissue homogenates may be made from fresh or previously frozen tissue. In some cases, use of frozen tissue can improve the signal-to-noise ratio. To test a variety of assay conditions, the homogenate is resuspended in a series of different buffers (e.g., 50 mM TrisCl, pH 6.0, 6.5, 7.0, 7.5, and 8.0; HEPES buffer; Krebs Ringer, pH 7.4) and incubated with ligand for different periods of time (e.g., 10, 30, 45, 60, 90 and 120 min) at a variety of controlled temperatures (e.g., 4, 22, and 37C). Conditions for both total and nonspecific binding should be determined in triplicate, and the reaction should be terminated by filtration over PEI-soaked glass fiber filters. At this point, binding will have been assessed as a function of temperature, pH, incubation time, and the need for ions (as assessed by any improvement in binding in Krebs Ringer). In the initial binding assay for the benzodiazepine receptor, a physiological buffer was used. It was subsequently shown that 50 mM TrisCl, pH 7.5, yielded the same data,

Practical Aspects of Radioligand Binding

thereby reducing the time and cost of buffer preparation. If the signal-to-noise ratio is poor (<40%), additional trials should be performed to determine more optimal conditions. There are instances where a radioligand has identical pharmacology to its unlabeled species, but does not have similar binding characteristics. Thus, for unknown reasons, the radioligand may not provide any specific binding at a receptor for which optimal binding conditions have been established for the unlabeled ligand. If no specific radioligand binding is obtained, or if there is minimal overall binding, a centrifugation assay should be attempted to check for low-affinity binding to the receptor. This can be done using a microcentrifuge and conducting the assay in a small aqueous volume on top of an oil layer. The incubation is then terminated by a rapid, cold centrifugation, which pellets the L*R complex, leaving the unbound radioactivity in the aqueous layer. The tip of the polypropylene microcentrifuge tube can then be cut with a blade and the protein in the tip solubilized in scintillation fluid containing a detergent. An alternative is to use a high-speed (e.g., 48,000 g) centrifuge, where pelleting of the L*R complex requires a 10-min centrifugation and washing is carefully performed using ice-cold buffer and a syringe so as not to agitate the pellet and reduce specific binding. If no binding is observed at this time, the purity, stability, and structure of the radioligand should be assessed. Whether the incubation is terminated by filtration or centrifugation, some time should be spent optimizing the specific binding, perhaps by screening other unlabeled ligands. The Kd and Bmax values for the new ligand should be determined by both saturation and kinetic analysis (see Analysis of Binding Data) and the subcellular and regional distribution of specific binding correlated with the reported functional pharmacology of the ligand. The pharmacology of the radioligand should be determined using a set of unlabeled ligands that are known to interact with the target receptor. Ideally, some of these should exist as isomers to permit assessment of the stereospecificity of the binding. Finally, a series of ligands thought to be inactive at the target receptor should be assessed at a concentration up to 100 M to provide additional information regarding the pharmacological properties and substrate selectivity of the binding site. Radioligand binding assay development is an iterative process. This is exemplified by the development of 4-phosphonomethyl-2-piper-

1.3.12
Current Protocols in Pharmacology

idinecarboxylic acid (CGS19755), a specific ligand for inhibiting the NMDA receptor (Murphy et al., 1987, 1988). In 1985, the only selective ligands for the NMDA receptor were 2amino-5-phosphonopentanoic acid (AP-5) and 2-amino-7-phosphonoheptanoic acid (AP-7), both of which have receptor affinities in the micromolar range. A rigid analog of AP-7, 3(2-carboxypiperazine-4-yl) propyl-1-phosphonic acid (CPP), was found to be more potent and a high-speed centrifugation binding assay using [3H]CPP was developed (Murphy et al., 1987). The assay was time-consuming and laborious, but provided a Kd measurement for CPP of 201 nM. The availability of this assay resulted in the identification of the even higheraffinity ligand, CGS19755 (Kd = 9 nM; Murphy

et al., 1988), which reached Phase III clinical trials for the treatment of stroke. Because CGS19755 had a high affinity for the receptor, it was the first radioligand that could be used in a high-throughput filtration assay for this site.

ANALYSIS OF BINDING DATA


The mathematical models used for analyzing binding data are based on considerations of molecular interactions in solution phase. The models do not take into consideration the fact that, in reality, the receptor is almost always in a lipid membrane phase. In practice, however, if the concentrations of the receptor and ligand are kept low, the solution phase equations are sufficiently accurate (Kenakin, 1993; UNIT 1.2). Of the various models proposed for receptors,

Saturation binding

A 1.0
Fraction bound

1.0

0.5

0.5

0 1 2 3 [L]/Kd 4 5

0 2 1 0 1 2 log([L]/Kd)

Competition binding

C 1.0
Fraction bound

D 1.0

0.5

0.5

0 1 2 3 4 5 [D]/IC50

0 2 1 0 1 2 log([D]/IC50)

Figure 1.3.5 Appearance of binding data using different methods of plotting. Saturation binding (A and B) and competition binding data are shown (C and D). The lefthand panels (A and C) show specific binding (expressed as fraction bound) plotted versus ligand concentration (expressed as a concentration ratio with respect to the Kd) on the x axis, on arithmetic plots (both axes plot untransformed values). The righthand panels (B and D) are semilogarithmic plots, which show specific binding plotted on the y axis (untransformed) versus the logarithm of the ligand concentration.

Receptor Binding

1.3.13
Current Protocols in Pharmacology

the occupancy model is most often invoked when describing radioligand binding (Fig. 1.3.5, panels A and C; UNIT 1.2).

d[ L ] = [ L ] 2.303 d (log[ L ])
Equation 1.3.11

Graphical Methods
Semilogarithmic plots The response of a tissue to a ligand, L (or L*, if the ligand under discussion is radioactive), is usually displayed as a function of the logarithm of the ligand concentration, [L]. The shape of the resulting curve described over a large range of drug concentrations is sigmoidal. The center of the curve, or the inflection point, is where 50% of the total response is observed. When the binding interaction is simple bimolecular, the slope of the fractional binding curve at the region of inflection in a semilogarithmic plot (Fig. 1.3.5B,D) is 0.576, if the amount of ligand bound is low and the total ligand concentration can be considered equal to the free ligand concentration. The value of 0.576 is derived as shown in Equations 1.3.7 to 1.3.13. The concentration of bound ligand, B, is a saturable function of the concentration of free ligand (F = [L]), the Kd, and the Bmax.
B= [ L ] Bmax [L] + Kd

Substituting Equation 1.3.11 into Equation 1.3.8 gives


dB [ L ] 2.303 d (log[ L ]) = 1 4 Kd

Equation 1.3.12

Again, the substitution [L] = Kd (at half-saturation) can be made, giving the slope for the binding curve at the inflection point.
dB d (log[ L ]) = K d 2.303 4 Kd = 2.303 4 = 0.576

Equation 1.3.13

Equation 1.3.7

Substituting [L] = Kd (at the point of half-saturation) into Equation 1.3.7 and differentiating gives
dB d[ L ] = Bmax 4 Kd

Equation 1.3.8

If the binding is expressed fractionally, then Bmax = 1. In a semilogarithmic plot as shown in Figure 1.3.5, the x axis is dependent on log[L], not on [L]. The slope of a semilogarithmic plot must then be the derivative, dB/d(log[L]). Equations 1.3.9 through 1.3.11 provide an expression for d(log[L]) in terms of d[L].
ln[ L ] = 2.303 log[ L ]
Equation 1.3.9

d (ln[ L ]) = d[ L ] / [ L ] = 2.303 d (log[ L ])


Equation 1.3.10 Practical Aspects of Radioligand Binding

This slope should not be confused with the Hill slope (see below), which is an exponent in an equation and which will have an absolute value of 1.0 for all of the data in Figure 1.3.5. The sigmoidal shape of the specific binding curve is a consequence of the transformation of the abscissa, because in a simple arithmetic plot the curve is hyperbolic with a maximum value that approaches a constant ordinate value (Fig. 1.3.5A,C). Theoretically, the ligand concentration must be increased to infinity to determine the ordinate value of the asymptotes of hyperbolic or sigmoidal curves. The asymptote must be estimated by extrapolation. A more accurate way to establish this value is with a Scatchard plot. However, the most accurate and most mathematically appropriate way of determining the parameters of saturation and competition curves is a direct fit to the equilibrium binding equation by nonlinear regression analysis on a computer (UNIT 1.2). In a competition between a radioligand and an unlabeled ligand (D), the position of the binding curve in a semilogarithmic plot is a function not only of the affinity of the unlabeled ligand, but also of the affinity and concentration of the radioligand (Fig. 1.3.6; see Example 2, below). With a high-affinity (low Kd) radioligand, or at higher ligand concentrations, the IC50 for the unlabeled compound will be higher and the curve will be shifted rightward (as with curves 2 and 3 in Fig. 1.3.6), suggesting a lower affinity of the unlabeled ligand. In analyzing such competition curves, the Cheng-Prusoff correction (Cheng and Prusoff, 1973) is used

1.3.14
Current Protocols in Pharmacology

1.0

0.8

Fraction bound

0.6 curve 1 0.4 2 3

0.2

0 7 6 5 log[D] (M) 4 3 2

Figure 1.3.6 Effect of increasing the radioligand concentration in a competition binding assay. This is a semilogarithmic plot of specific binding (expressed on the y axis as fraction of maximal binding) in a hypothetical experiment. The drug D is an unlabeled competitor and the logarithm of its concentration is plotted on the x axis. Curves 1, 2, and 3 represent the appearance of the data when progressively larger concentrations of the radioligand are used. For further explanation, see Example 2: The Cheng and Prusoff Correction.

to convert the IC50 to Ki, which is an absolute value. Thus,


Ki = IC 50 1 + [L] / Kd

Equation 1.3.14

where Ki and IC50 refer to the unlabeled ligand, [L] is the concentration of the radioligand used in the assay, and Kd is the equilibrium binding dissociation constant for the radioligand. The following postulates and assumptions should be kept in mind when using this correction. First, equilibrium is assumed, because equations for steady state yield invalid data when the binding system is not at equilibrium (Ehlert et al., 1981). If the correct kinetic equations are used in such nonequilibrium situations, it is possible to indirectly derive the correct binding constants for unlabeled compounds (Schreiber et al., 1985). Second, the reaction is presumed to be simple bimolecular. If the reaction is not and if the Hill slope is significantly different from unity, the resulting value is not the Ki but the K50 (a

pseudo-Ki), a number with little meaning (i.e., it equates to Ki only under conditions of competitive inhibition). Binding is not of a simple competitive nature when (1) the receptors bind ligand cooperatively, (2) the receptors are desensitized or internalized during the assay, or (3) the receptors exist in multiple independent forms with different binding affinities for the ligands. Because of the dependence on simple bimolecular binding, the Cheng-Prusoff correction is usually inappropriate for use when an antagonist is used to inhibit the response to an agonist in a functional assay (a null method). Because the Hill slopes for agonist dose-response curves in functional assays are frequently of nonunit value, simple bimolecular models cannot be used. It is thus inappropriate to use the Cheng-Prusoff correction to calculate the Ki value for an antagonist in a functional assay unless it can be shown that the Hill slope for the agonist dose-response curve is unity (Lazareno and Birdsall, 1993). Third, the concentration of receptors in the binding assay must be much lower than the

Receptor Binding

1.3.15
Current Protocols in Pharmacology

binding constants for the ligands since higher receptor concentrations require the use of a more complex correction to obtain the Ki value (Jacobs et al., 1975; Linden, 1982). Example 2: The Cheng-Prusoff correction In Figure 1.3.6, bound ligand from an assay experiment is shown as a function of log[D], where [D] is the concentration of a competing unlabeled ligand. The concentrations of radioligand ([L*]) for curves 1, 2, and 3 are 1 nM, 10 nM, and 30 nM, respectively. Note that the absolute levels of specifically bound radioligand are normalized to 100% for all three curves. The equilibrium dissociation constant for the radioligand (Kd) under these conditions is 0.5 nM, and the shapes of the displacement curves are consistent with a simple competitive interaction (Hill slope = 1). By computer analysis, the IC50 values for the unlabeled compound were found to be 3 M, 21 M, and 61 M for curves 1, 2, and 3, respectively. The following calculations demonstrate that the Ki for the unlabeled compound is 1 M in all three experiments. Curve 1 Radioligand concentration = 1 nM Radioligand Kd = 0.5 nM Cheng-Prusoff correction = 1 + (1 nM/0.5 nM) = 3 Ki of the competitor = IC50/3 = 3 M/3 = 1 M Curve 2 Radioligand concentration = 10 nM Radioligand Kd = 0.5 nM Cheng-Prusoff correction = 1 + (10 nM/0.5 nM) = 21 Ki of the competitor = IC50/21 = 21 M/21 = 1 M Curve 3 Radioligand concentration = 30 nM Radioligand Kd = 0.5 nM Cheng-Prusoff correction = 1 + (30 nM/0.5 nM) = 61 Ki of the competitor = IC50/61 = 61 M/61 = 1 M The Scatchard (Rosenthal) plot This method was originally described by Scatchard (1949), and was refined by Rosenthal (1967), for calculating the binding properties of ligands bound to proteins in solution. The Scatchard transformation is based on the occu-

pancy theory of receptor models, as previously shown in Equation 1.3.1.


L+R

k1

k+1

LR

Equation 1.3.15

At equilibrium, the forward and reverse reactions are equal in velocity, and the Kd is given as [L] [R] Kd = [ LR ]
Equation 1.3.16

Bmax is the total number of receptors.


Bmax = [ R ] + [ LR ]
Equation 1.3.17

By solving for [R] (Equation 1.3.18), substituting into Equation 1.3.16, and rearranging, [LR] is determined as a function of the concentration of free ligand ([L]), Kd, and Bmax (Equation 1.3.19). This is equivalent to Equation 1.3.7.
[ R ] = Bmax [ LR ]
Equation 1.3.18

[ LR ] =

Bmax [ L ] [L] + Kd

Equation 1.3.19

As LR is the bound ligand and L is the free ligand, Equation 1.3.19 can be rewritten:
bound = Bmax free free K d

Equation 1.3.20

This can be rearranged to the form used in the Scatchard plot.


bound free = Bmax Kd bound Kd

Equation 1.3.21

Practical Aspects of Radioligand Binding

This transformation is extensively used in analysis of receptor binding in membrane preparations. Thus, following an equilibrium saturation experiment, the ratio of bound/free can be plotted versus bound (Fig. 1.3.7), with

1.3.16
Current Protocols in Pharmacology

the x intercept representing the maximum number of binding sites (the Bmax) and the slope being used to determine the Kd (slope = 1/Kd). Example 3 (below) describes the data analysis used to produce this graph. Until computers were used to facilitate the fitting of data with models using nonlinear regression methods, the Scatchard plot was the standard way to determine the Bmax and Kd for receptor binding sites. It has been extensively and justifiably criticized for its deficiencies, especially when compared to nonlinear curve fitting methods (Rodbard et al., 1980; Feldman, 1983; Burgisser, 1984). Most notably, the transformation of the binding data produced by dividing the bound ligand by the free ligand results in distortion, such that the upper extremity of the Scatchard plot (region of low free ligand concentration) is overemphasized in a linear regression (Rodbard et al., 1980). This results in a skewed determination of Bmax and Kd. Other assay factors such as nonequilibrium

conditions, radioligand accumulation or degradation in the tissue, or receptor isomerization, are also magnified by a Scatchard transformation (Ketelslegers et al., 1984; Beck and Goren, 1983). Despite these reservations, the Scatchard plot has considerable heuristic value, particularly for demonstrating whether the binding data are simple or complex in nature. Scatchard plots that are convex (i.e., the middle of the plot is curved upward relative to the extremities) indicate positive cooperativity, while concave plots indicate negative cooperativity or binding site heterogeneity (Fig. 1.3.8). Cooperativity refers to interactions between receptors upon binding of a ligand at multiple sites (usually one ligand on each receptor). With positive cooperativity, ligand binding to one of the interacting receptors sequentially facilitates binding at the other(s), while in negative cooperativity binding is progressively less avid (Monod et al., 1965; Koshland et al., 1966).

200
Bound/ free (fmol/mg/ nM)

R = 0.979

slope = 4.309 100


Kd = 0.232 nM B max = 61 fmol/ mg

0 0 10 20 30 40 50 60

Bound (fmol/ mg)

Figure 1.3.7 Scatchard (Rosenthal) plot of the equilibrium binding of [3H]NMS to muscarinic receptors on N1E-115 neuroblastoma cells. The data shown in Figure 1.3.1 and Table 1.3.1 were used to construct this plot, and a full discussion is given in the text. The linear regression correlation coefficient for this plot (R) is 0.979, the Kd value (the negative inverse of the slope) is 0.232 nM, and the Bmax (the x intercept) is 61 fmol/mg. The data in Figure 1.3.7 were normalized to mg protein, whereas the data in Figure 1.3.1 are expressed in fmol/tube specific binding.

Receptor Binding

1.3.17
Current Protocols in Pharmacology

100

n H = 1.0

n H = 0.5

Bound/ free (% B / [L])

50 n H = 1.5

0 0 20 40 60 80 100

Bound (% B max)

Figure 1.3.8 Effect of allosterism or cooperativity on the appearance of binding when shown on a Scatchard (Rosenthal) plot. When multiple sites interact in negative cooperativity (nH < 1), the plot of binding is concave. When multiple sites interact with positive cooperativity (nH > 1), the Scatchard transform is convex. The normal appearance of the Scatchard plot for simple bimolecular interactions (no cooperativity; nH = 1), is the straight-line plot. Note that binding of the ligand to multiple independent sites which exhibit different binding affinities will produce a concave Scatchard plot (apparent negative cooperativity).

Allosterism is a variant of cooperativity where a single receptor molecule or complex contains distinct but interacting binding sites for two or more ligands. The seminal example of allosterism is that of the GABAA/benzodiazepine ligand-gated receptor (Rabow et al., 1995). Example 3: Scatchard analysis This example is based on the experiment shown in Figure 1.3.1. The untransformed binding data are shown in Table 1.3.1. Six concentrations (0.06 nM to 1.4 nM) of [3H]NMS were incubated with intact N1E-115 mouse neuroblastoma cells (0.5 mg tissue in a 2-ml assay) at 37C until equilibrium was achieved (60 min). Total binding of the radioligand was measured in triplicate at each concentration, and the amount of nonspecific binding was determined in duplicate at each concentration by adding 1 mM atropine to some tubes to displace the radioligand. The specific activity of [3H]NMS was 53.5 Ci/mmol and the efficiency of the scintillation counter was 43%.

Practical Aspects of Radioligand Binding

The concentrations of the radioligand used in the assay were determined by measuring the radioactivity in two aliquots of each dilution of stock. First, it is necessary to calculate the factor needed to convert cpm to fmoles of ligand. Since the specific activity of the radioligand was 53.4 Ci/mmol and there are 2.2 1012 dpm/Ci, it follows that there are 53.4 2.2 1012 dpm/mmol ligand (or 117.5 dpm/fmol ligand). With a 43% counting efficiency (1 dpm = 0.43 cpm), there are 50.5 cpm/fmol ligand. As an example, the results of converting total binding from cpm to fmol using this factor are shown in Table 1.3.1. For each of the six points in this assay, the free concentration of [3H]NMS is calculated by subtracting the total radioactivity bound to the filter from the total amount of ligand added to the tube. For example, at the lowest ligand concentration, 123 8.3 = 114.7 fmol. Thus, the lowest free concentration of [3H]NMS is 114.7 fmol/2 ml, or 0.0574 nM.

1.3.18
Current Protocols in Pharmacology

Table 1.3.1

Sample Raw Data for Scatchard Analysis

Total [3H]NMS concentration 0.061 nM (123 fmol/tube) 0.12 nM (240 fmol/tube) 0.233 nM (466 fmol/tube) 0.47 nM (930 fmol/tube) 0.87 nM (1747 fmol/tube) 1.41 nM (2816 fmol/tube)

Total bindinga (cpm) 455, 395, 413 (421) 593, 643, 634 (623) 985, 942, 983 (970) 1250, 1249, 1332 (1277) 1442, 1438, 1495 (1458) 1382, 1785, 1769 (1645)

Nonspecific bindinga (cpm) 98, 101 (100) 127, 121 (124) 133, 144 (138) 183, 182 (182) 266, 255 (261) 433, 442 (438)

Average total binding (fmol) 8.3 12.3 19.2 25.3 28.9 32.6

aReplicate cpm values shown, with average cpm indicated in parentheses.

Table 1.3.2 Sample Transformed Data for Scatchard Analysis

Specifically Free [3H]NMS bound [3H]NMS (nM) (fmol/mg) 0.0574 0.114 0.223 0.452 0.859 1.392 12.7 19.8 33 43.4 47.4 47.8

Bound/free (fmol/mgnM) 221.5 173.5 147.8 95.9 55.2 34.3

In this calculation, nonspecific binding is considered to be removed from the reaction. Note that, depending on its physical or biological nature, nonspecific binding may be readily reversible, in which case nonspecifically bound ligand can be considered available for specific binding at the target receptor. In such a case, the amount of free radioligand at each point in the saturation curve can be calculated as the total ligand added minus the specific binding. The amount of specifically bound [3H]NMS at each point is determined by subtracting the nonspecific binding from the total binding. At the lowest concentration of [3H]NMS used, nonspecific binding is 100 cpm, which converts to 2.0 fmol. Specific binding is thus 8.3 2.0 = 6.3 fmol. This value is generally normalized to the tissue concentration: 6.3 fmol/0.5 mg tissue = 12.6 fmol/mg tissue. Although not shown here, an alternative procedure for determining nonspecific binding takes advantage of its linearity (see Binding Specificity). In this case, the nonspecifically bound radioligand is plotted versus the total [3H]NMS added, and linear regression is used to derive more accurate estimates of the nonspecific binding at each point. Ideally, non-spe-

cific binding would be evaluated by both methods, although the latter is more involved. In Table 1.3.2, transformations of free and bound radioligand concentrations suitable for Scatchard analysis are shown. For the Scatchard plot, bound/free is plotted against the specifically bound [3H]NMS and the data are examined for linearity (Fig. 1.3.7). In this case, the data appear to be linear since the data points can be fitted with a straight line. The slope of this line is 1/Kd in units of nM1, with the x intercept representing Bmax, in fmol/mg tissue. For this particular plot, the slope is 4.309 nM1, which yields a Kd value of 0.232 nM. The Bmax is 61 fmol/mg tissue. Compare the Scatchard plot (Fig. 1.3.7) with the plot of the untransformed data that has not been normalized to the tissue concentration (Fig. 1.3.1). The Hill plot This analysis was originally intended to demonstrate the number of binding sites for oxygen on hemoglobin (a protein that consists of four subunits) and the positively cooperative nature of this binding process (Hill, 1910). The amount of bound ligand (B) is expressed as a
Receptor Binding

1.3.19
Current Protocols in Pharmacology

saturable function of the concentration of the free ligand (F):


B Fn B = max n Kd + F
Equation 1.3.22

where n indicates the number of binding sites and Kd is an amalgamation of dissociation constants for the individual subunits, which can be derived mathematically. A rearrangement of this equation yields an expression of binding as the logarithm of the ratio of occupied sites divided by unoccupied sites. In this equation the nomenclature has reverted to Kd, which can be viewed as the empirical or observed Kd.
log

Bmax

= n log F log K d B

Equation 1.3.23

A plot of this term versus the logarithm of the concentration of the free ligand is thus a logit-log plot, except for the use of the logarithm to the base 10 for the Hill plot and the use of the natural logarithm (ln) in the logit-log plot. An example of a Hill plot is shown in Figure 1.3.9 (see Example 4 below for details

on the construction of this particular plot). The n in this equation is referred to as the Hill slope, and it is often abbreviated as nH. Hill plots can also be constructed for competition curves, in which case nH has a negative sign. It is not necessary to know the Bmax to construct a Hill plot, because the data can be expressed as percent bound (%B) with the value on the y axis becoming the logarithm of %B/(100 %B). For example, in a competition experiment, data for a given concentration of radioligand are plotted as the logarithm of %B/(100 %B) versus the logarithm of the concentration of the unlabeled ligand (log[D]). The displacement of a radioligand by an unlabeled ligand yields a Hill plot with a negative slope. Thus, if the competition between the radioligand and the unlabeled compound is simple competitive at a single binding site in the preparation, then nH = 1. In this case, logF equals the logIC50 of the displacing agent at y = 0. If the equilibrium binding constant and the concentration for the radioligand are known, these values can be used in the Cheng-Prusoff equation (see Equation 1.3.14) to obtain the Ki of the unlabeled ligand. Typically, the Hill plot is used to determine the complexity of binding. Simple bimolecular interactions yield a Hill slope of unity. However, when the absolute value of the Hill slope is statistically different from unity it can mean

0.6 0.4

slope = 0.89

(Bmax B)

0.2 0 0.2 0.4 0.6


x intercept = 9.6521 (0.228 nM)

log

11

10 logF (M)

Practical Aspects of Radioligand Binding

Figure 1.3.9 Hill plot for the saturation binding of [3H]NMS to muscarinic receptors on N1E-115 cells. The data used to make Figures 1.3.1 and 1.3.7 and Table 1.3.1 were used for this Hill plot. This is a log-log plot, in which the logarithm of the ratio of bound to unbound is plotted on the y axis, and the logarithm of the free concentration of the radioligand is plotted on the x axis. The Hill slope is 0.89, which in this case is not significantly different from unity (P > 0.05).

1.3.20
Current Protocols in Pharmacology

that the simple model is not appropriate. A value >1 may be indicative of positive cooperativity, although the value of the Hill slope does not necessarily quantify the number of distinct ligand binding sites, as Hill originally intended. With neurotransmitter receptors, the binding isotherms of 3H-labeled antagonists commonly yield Hill slopes of unity, while agonist binding produces Hill plots with slopes significantly <1. As an example, Figure 1.3.10 shows a Hill plot of the experiment shown in Figure 1.3.3. In this case, the Hill slope is 0.25. This unusually low value is a consequence of the binding of carbachol to at least three distinct sites in the N1E-115 cells, each with a distinct binding affinity (0.26 nM, 0.1 M, and 25.7 M). These cells express both the m1 and m4 muscarinic receptor subtypes (Yasuda et al., 1993), and the binding of an agonist induces two different affinity states for each subtype, giving rise to very complex competition curves (Fig.1.3.3). If the absolute value of the Hill slope is significantly different from unity, the Cheng-

Prusoff equation is not valid to convert an IC50 to a Kd value. Rather, the appropriate mathematical model and a robust fitting method should be used for analysis of such binding data. For example, the low Hill slope values determined for agonist binding to receptors that couple with GTP-binding proteins indicate the presence of multiple binding states for a single molecular class of receptor protein, suggesting a ternary complex. Alternatively, the data might suggest the existence of multiple, noninteracting receptor subtypes, in which case a model containing independent binding capacities and ligand binding constants should be used. Certain experimental manipulations may simplify the analysis, including the use of preliminary analyses to quantify the number of each receptor subtype present, or the use of a GTP analog to convert the receptors to a single agonist binding state. Example 4: Construction of a Hill plot The data used in the Scatchard analysis (Table 1.3.2) can be transformed to the data shown

1.0 0.8 0.6

(100 %B )

0.4 0.2 0 0.2 0.4 0.6 0.8 1.0 n H = 0.25

log

%B

10

log[carbachol] (M)

Figure 1.3.10 Hill plot for the competition between [3H]NMS and carbachol for muscarinic receptors on N1E-115 cells. The binding data shown in Figure 1.3.3 were used to construct this plot. The ratio of bound/unbound in this case was expressed using percentages of maximal specific binding (%B), rather than the Bmax (as was done in Figure 1.3.9). The absolute value of the Hill slope in this case was 0.25, which was significantly different from unity (P > 0.001). The apparent negative cooperativity results from the binding of carbachol to three different sites on N1E-115 cells with different affinities.

Receptor Binding

1.3.21
Current Protocols in Pharmacology

Table 1.3.3

Sample Transformed Data for Constructing a Hill Plot

B (fmol/mg) 12.7 19.8 33 43.4 47.4 47.8

F (nM) 0.0574 0.114 0.223 0.452 0.859 1.392

logF 10.24 9.94 9.65 9.34 9.07 8.86

B/(Bmax B) 0.2629 0.4806 1.1786 2.4659 3.4853 3.6212

log[B/(Bmax B)] 0.5802 0.3182 0.0714 0.392 0.5422 0.5589

in Table 1.3.3. While the first three columns are self-explanatory, the Bmax determined in the Scatchard plot was used in the calculations for the final two columns. The Hill plot is constructed by plotting the last column on the ordinate versus the logF column on the abscissa (Fig. 1.3.9). As shown, the Hill plot of a radioligand binding site saturation curve has a positive slope that passes through the value of y = 0 (when the binding sites are 50% occupied) at the point at which logF = logKd. The absolute value of the Hill slope is unity if the binding is simple bimolecular (nH = 1), and logF at y = 0 provides a true measure of logKd for the radioligand. Indeed, it is probably a more accurate method than the Scatchard plot for deriving Kd, as two plots are used in the derivation. In this example, nH = 0.89 and the x intercept = 9.6521, corresponding to a Kd of 0.228 nM, which approximates the value determined in the Scatchard analysis. Alternatively, instead of using the Scatchard Bmax, the maximal observed binding can be used, as for the example shown in Figure 1.3.10. The linear regression correlation coefficient (R) for this Hill plot is 0.979. Using the slope and correlation coefficient, standard statistical procedures can be used to determine whether the slope is significantly different from unity. Computer programs such as PHARM/PCS (Tallarida and Murray, 1986) are also available for this purpose. In the present example, the PHARM/PCS program indicates that the slope is not significantly different from unity (P < 0.05), which can be interpreted to mean that the binding is probably simple bimolecular. This is also consistent with the interpretation of the data as developed in the Scatchard plot.

dissociation of the ligand from the receptor is usually a unimolecular process governed only by the off-rate, and because it is not necessary to know the number of receptors in the assay, as is the case for determination of the association rate constant. A plot of the logarithm of the concentration of ligand bound at time t minus that bound at t0 versus time invariably yields a straight line for the dissociation of a ligand from a single receptor. However, if the receptor isomerizes, it distributes into multiple forms with unique dissociation rate constants, yielding a complex curve. This is indistinguishable from the situation when the ligand binds to multiple receptors. Additional analysis is required to establish that multiple binding states of a single receptor type are present (Jarv et al., 1979). To determine the dissociation rate constant, the receptor and radioligand are incubated until binding equilibrium is reached. The complete binding reaction (previously shown in Equation 1.3.1) is k+1 L* + R L* R k1

Equation 1.3.24

The forward reaction is eliminated either by rapidly diluting the assay mixture to an approximation of infinite volume or by adding excess unlabeled ligand.

L * + R L * R
k 1

Equation 1.3.25

The rate of this reaction is modified from Equation 1.3.3.

Determination of Rate Constants


Dissociation rate constant For most binding studies it is easier to determine the dissociation rate constant than the association rate constant. This is because the

d[ L * R ] / dt = k 1 [ L * R ]
Equation 1.3.26

Practical Aspects of Radioligand Binding

Following separation of the bound and free radioligand, the only ligand measured is that bound to receptor, which decreases over time.

1.3.22
Current Protocols in Pharmacology

The dissociation rate constant is then determined from the slope of a plot of the logarithm of the ratio of bound ligand versus time. This is shown in Equation 1.3.29, which is derived from Equation 1.3.26 in the following manner. Equation 1.3.26 is rearranged,
d[ L * R ] / [ L * R ] = k 1 dt
Equation 1.3.27

lected time points, the differential dissociation of [3H]NMS can be used to reduce radioligand binding to a particular receptor subtype to insignificance, simplifying the analysis of its binding to the remaining receptor populations. Example 5: Determination of the dissociation rate constant The binding of [3H]NMS to muscarinic receptors in intact N1E-115 mouse neuroblastoma cells at 37C was studied in a 2-ml assay using an isotonic physiological phosphate buffer at pH 7.4. The assay was performed in two stages. In the first stage, [3H]NMS binding was allowed to reach equilibrium by incubating 300,000 cells/tube with 0.56 nM radioligand for 45 min at 37C. In the second stage, a high concentration (10 M) of unlabeled NMS was added at various times to the preequilibrated reactions to permit analysis of the dissociation of [3H]NMS over a period ranging from 30 sec to 45 min. A 24-well filtration manifold was used to separate bound from free ligand. Since the dissociation experiment involved a total of 48 tubes, two filter strips were used. A schedule of incubation times, NMS addition, and filtration was determined so that all the reaction mixtures for one filter (24 tubes for 12 time points in duplicate) were filtered at once, followed 10 min later by filtration through the second filter (24 tubes for 12 time points in duplicate). To allow all of the tubes for the 12 time points to be filtered at the same time, the schedule was reversed so that the unlabeled NMS was first added to the tubes for the 45 min dissociation of [3H]NMS and last to the tubes for 30 sec dissociation. The amount of binding at t = 0 was determined to be 51.8 fmol/million cells. Nonspecific binding was assessed in separate sets of tubes and subtracted from the total binding to obtain the specifically bound [3H]NMS. With this protocol, the data shown in Table 1.3.4 were obtained. The untransformed data in the center column, which is the specifically bound [3H]NMS measured at each nonzero time point, were plotted versus time (left column) as shown in Figure 1.3.11A. In Figure 1.3.11B, the transformed data in the right column were plotted versus time. These transformed data are nonlinear, as would be expected if [3H]NMS dissociates from more than a single class of binding site. A preliminary estimate for a single value for k1 could be made by fitting a straight line through the transformed data. In this case, k1 would be 0.08 min1. However, the data are more properly analyzed using a computer pro-

and then integrated to give


ln[ L * R ] = k 1 t + c
Equation 1.3.28

The term c is a constant of integration equivalent to ln[L*R0], the natural logarithm of the amount of bound ligand at t = 0, when the dissociation experiment begins. The complete equation is
ln([ L * R t ] / [ L * R 0 ] = k 1 t
Equation 1.3.29

where [L*Rt] is the amount of specifically bound radioligand at a given time t, [L*R0] is the amount of specifically bound radioligand at t = 0 (i.e., at equilibrium), and k1 is the dissociation rate constant (in units of time1). Thus, a plot of ln([L*Rt] / [L*R0]) versus time yields a straight line with a slope of k1. When there are multiple states or there are multiple binding sites that differ in their dissociation rate constants, this plot is nonlinear and the data are best analyzed by direct fitting to the appropriate equation (see Use of Computer Modeling Techniques), rather than by determining the slopes from the plot. A special binding assay method that makes use of the differential dissociation rates of radioligands from a heterogeneous receptor population is exemplified by the elegant work of Christophe and associates (Waelbroeck et al., 1990). Muscarinic receptors are encoded by five genes, and four of the resultant receptor subtypes are expressed at significant levels in brain tissue. Most antagonists, including NMS, bind with the same, or nearly the same, affinity to these receptor subtypes. However, NMS dissociates from muscarinic receptors with significantly differing off-rates. In a competition binding assay, the unlabeled ligand will compete more effectively with a subtype that has a faster dissociation rate. Thus, at carefully se-

Receptor Binding

1.3.23
Current Protocols in Pharmacology

Table 1.3.4 Sample Data for the Determination of the Dissociation Rate Constant

t (min) 45 30 20 15 10 7 5 3 2 1 0.5

Bt (fmol/million cells) 3.4 6.1 10.7 14.5 21.3 27.1 32.5 39.7 42.3 48.0 49.1

ln(Bt/B0) 2.718 2.137 1.58 1.273 0.889 0.648 0.467 0.267 0.203 0.077 0.055

gram with an appropriate binding model. Other experiments with N1E-115 cells have shown that they express two subtypes of muscarinic receptors, and it is known from the literature that [3H]NMS dissociates from each subtype with a different rate. Therefore, the data in Figure 1.3.11 were fitted using a computer model of two classes of binding sites. At the major site (79% of the total sites), k1 = 0.114 min1, while at the less abundant site (21% of the total), k1 = 0.027 min1. Association rate constant In the simple bimolecular reaction between L* and R, the rate of formation of L*R (forward reaction from Equation 1.3.24) is second order with respect to the concentration of reactants. The forward rate is modified from Equation 1.3.2.
d[ L * R ] / dt = k +1 [ L*] [ R ]
Equation 1.3.30

The slope of the line tangent to the plot of d[L*R] versus t, at the point at which the plot passes through the origin, can, in principle, be used to calculate the value of k+1, if [RT] is known (Rodbard, 1973). However, because of the inaccuracy of determining this tangent on a manual plot, this procedure is rarely used. As the binding reaction proceeds, a significant amount of L*R is formed and the reverse reaction begins to contribute (negatively) to the amount of L*R that can be measured at any given time t.
d[ L * R ] / dt = ( k +1 [ L*] [ R ]) ( k 1 [ L *R])

Equation 1.3.31

Typically, k1 is determined first (see Dissociation Rate Constant, above), and k+1 is calculated from the data obtained by conducting the association experiment at several different concentrations of [L*]. The complete equation describing the association of ligand to receptor is
ln

Practical Aspects of Radioligand Binding

When the receptor and radioligand are initially mixed, this equation describes the formation of L*R when [R] = [RT], the total concentration of receptors present. [RT] is determined independently by equilibrium binding and by Scatchard analysis; for details, see the Scatchard (Rosenthal) plot (above). If the concentration of receptors is much lower than the concentration of radioligand ([RT] << [L]), the amount of free ligand is not appreciably reduced by the amount of ligand bound to the receptor, and therefore [L] is essentially a constant during the assay. Under these conditions, the equation is first-order with respect to receptor concentration at the outset of the reaction.

= ( k1 + k+1[ L*]) t [ L * R eq ] [ L * R t ]
[ L * R eq ]
Equation 1.3.32

where [L*Req] is the concentration of ligandreceptor complexes at equilibrium and [L*Rt] is the concentration of ligand-receptor complexes at any time t. A plot of the lefthand portion of Equation 1.3.32 versus time yields a straight line with a slope of (k1 + k+1[L*]). Note that [L*] is assumed to be a constant for each association curve. An independent determination of k1 is

1.3.24
Current Protocols in Pharmacology

unnecessary when directly fitting association data to this equation with a computer and an iterative technique. Example 6: Determination of the association rate constant In an experiment performed in parallel with that shown in Example 5 above, the binding of 0.56 nM [3H]NMS to muscarinic receptors in N1E-115 neuroblastoma cells was studied at 37C using 300,000 cells per 2-ml assay. A

schedule of reaction initiation and termination times was established so that all of the incubations were terminated at the same time using filtration. With this protocol, the data shown in Table 1.3.5 were obtained. Since equilibrium was attained after 10 min, the amount of binding at 60 min was taken as Beq, the amount of binding at equilibrium, and the amount of binding at each time t (Bt) was used to calculate the values in the righthand column. The untransformed data are plotted in

A
60
Bound [ 3 H]NMS (fmol/ million cells) In(Bt /B0)

50 40 30 20 10

3 0 5 10 15 20 25 30 35 40 45

t (min)

Figure 1.3.11 Dissociation of [3H]NMS from muscarinic receptors on N1E-115 cells. The radioligand (0.56 nM) was incubated with intact N1E-115 cells (300,000 cells/tube) for 45 min at 37C, at which time equilibrium is reached. The dissociation of radioligand from the receptors was followed for various periods of time after the addition of 10 M NMS, and the reactions terminated by rapid filtration. Nonspecific binding was also determined at these various times (not shown) by adding 10 M NMS to some tubes for the duration of the experiment; the nonspecific binding did not vary with time, and it was subtracted from the total binding to obtain the specific binding, which is plotted. (A) An arithmetic plot of the time course of dissociation. (B) Plot of the natural logarithm (ln) of the amount bound at any time (t), expressed as a fraction of the amount bound at t = 0 (i.e., B0). Note the inflection point in the lower plot is at about t =20 min.

Receptor Binding

1.3.25
Current Protocols in Pharmacology

Table 1.3.5 Sample Data for the Determination of the Association Rate Constant

t (min) 0.5 1 3 5 7 10 15 20 30 45 60

Bt (fmol/million cells) ln[Beq/(Beq Bt)] 15.5 25.4 43.4 51.6 54.9 57.1 57.5 61.5 59.0 57.6 56.1 0.31 0.58 1.38 2.20 2.92 4.11

Practical Aspects of Radioligand Binding

Figure 1.3.12A. The transformed data (right column) are plotted in Figure 1.3.12B. The latter is very nearly linear; however, close examination reveals nonlinearity at times earlier than 1 min. If this early portion is ignored and a line fitted through these data, a slope of 0.43 min1 is obtained. If this slope is set equal to (k1 + k+1[L*]), and k1 from the major class of sites (0.114 min1; see Example 5) is inserted, the value for k+1 is determined to be 0.56 nM1 min1. Again using the k1 for the major site, the ratio of k1/k+1 is 0.114/0.56 = 0.204 nM. This value approximates the equilibrium binding constant (Kd) determined by Scatchard or Hill analysis if it is assumed that only this class of receptor is present in N1E-115 cells. Since the Kd determined by kinetic analysis (on- and off-rates) is very similar to the values determined with equilibrium binding (0.232 nM, 0.228 nM), it appears there is only one class of binding sites. Although there is small percentage of another class of muscarinic receptor binding sites for [3H]NMS cells, their presence is not revealed with a Scatchard plot composed of only six points. To obtain a curvilinear Scatchard plot that could reliably reveal the second group of binding sites (19% of the total), many more data points are required from the binding assay (typically >20). Direct fitting of the association data shown in Figure 1.3.12 using iterative nonlinear fitting yields a k+1 value of 0.699 nM1 min1. A more accurate determination of binding parameters is possible if the data are directly fitted with a mathematical model rather than being transformed and analyzed on a manual plot, even though, in this instance, the k+1 value determined

from the plot (0.56 nM1 min1) is similar to the computer-derived value (0.699 nM1 min 1).

Use of Computer Modeling Techniques


There are many published and commercially available computer programs for the analysis of binding data. Most spreadsheet or graphical programs (e.g., Lotus) incorporate iterative fitting techniques for data. In many instances, the user can formulate the mathematical model. The choice of model and the method of fitting to the data should be defensible on both biological and statistical grounds. The simplest hypothesis should be selected unless there is information to support a more complex model (Kenakin, 1993). For example, the addition of a third binding site to a two-site model may be inappropriate if only two sites are known to actually exist in the tissue. The addition of parameters to a receptor-ligand binding model always allows a closer fit of the model to the data to produce a lower sum-of-squared residuals; however, the question then becomes whether the increase in parameters has significantly reduced the variance. To permit statistical discrimination between alternate models, the data must be of sufficient quality and quantity to permit confident assessment of model parameters. This means the variance should be as low as experimentally possible and there should be a sufficient number of data points. A measure of variance is calculated by determining the differences between the data and the fitted curve at each point, and then squaring these differences and summing them. Variance can be minimized by performing replicates of

1.3.26
Current Protocols in Pharmacology

the binding assay and, up to a point, the statistical power may be strengthened by increasing the number of data points in the assay. Receptor models Generally, the receptor models used in computer fitting of equilibrium binding data contain equations that describe the binding to one site or to multiple, noninteracting sites. The derivation of these models and their interpretation with regard to their appearance in the

Scatchard plot can be traced back at least to Feldman (1972). With the simplest model of multiple, independent sites (two binding sites), the Scatchard plot is concave and the tangents to the extremities of the curve describe two major populations of receptors, high-affinity and low-affinity. Because these tangents are difficult to draw accurately, the Scatchard plot should not be used. Examples of the more commonly encountered equations are shown below with a descrip-

A
60 50
Bound [ 3 H]NMS (fmol/ million cells)

40 30 20 10

10 t (min)

15

60

B
5

( Beq Bt )
In

Beq

6 t (min)

10

Figure 1.3.12 Association of [3H]NMS to muscarinic receptors on N1E-115 cells. The radioligand concentration was 0.56 nM, 300,000 cells/tube were used, and the temperature was 37C. Panel (A) is a plot of the untransformed specific binding measured at various times (t) after starting the incubations. The plot in panel (B) is a logarithmic transform of the ratio of the amount bound at equilibrium (Beq) to that remaining unbound (Beq Bt) at any time (t).

Receptor Binding

1.3.27
Current Protocols in Pharmacology Supplement 10

tion of their use. While in all such cases there is only one independent variable ([L*] or [D]) and one dependent variable ([L*R]), two or more parameters must be determined. These parameters are constants in the equations to be solved by the computer using the data sets (pairs of [L*], [L*R] or [D], [L*R]) to iteratively refine parameter estimates until the model equation best fits the data. The parameters are thus adjusted so that the differences between the values of [L*R] actually determined in the experiment and the corresponding values calculated from the model are minimized. Normally, when a computer is used to determine binding parameters certain constraints are applied such as allowing only nonnegative values. A common case encountered in binding experiments is the presence of two independent binding sites (usually two different receptor molecules). In a saturation equilibrium binding assay with a radioligand (L*) that distinguishes two sites, the amount of specifically bound radioligand [L*R] is
[L * R] = B1 [ L*] K1 + [ L*] + B2 [ L*] K 2 + [ L*]

bind an unlabeled competitor (D) with differing affinities, the following equation is used:
[L * R] = B1 [ L*] ([ L*] + K d ) (1 + [ D] / K1 ) + B2 [ L*] ([ L*] + K d ) (1 + [ D] / K 2 )
Equation 1.3.34

Equation 1.3.33

Practical Aspects of Radioligand Binding

where B1 and B2 are the capacities of the two sites (i.e., the Bmax values for each receptor), and K1 and K2 are the equilibrium binding dissociation constants for the radioligand at the respective sites. With this equation, the independent variable is [L*], the dependent variable is [L*R], and there are four parameters to be determined (B1, B2, K1, and K2). If the radioligand recognizes both sites with the same affinity, then the equation simplifies to one term on the righthand side, and only two parameters must be determined. Conversely, if more than two independent binding sites are present, additional analogous terms may be added to the two righthand terms shown in Equation 1.3.33. Moreover, a term describing nonspecific binding can be added ([L*] KNS), in which the amount of nonspecific binding is constrained to be linearly dependent on [L*]. With iterative computer-based minimization techniques, the addition of this term is a more accurate way to obtain an estimate of the level of nonspecific binding. When a radioligand (L*) is competitively displaced from two independent sites that bind the radioligand with the same Kd and that also

where Kd refers to the dissociation constant for the radioligand, usually determined in a separate saturation experiment, and K1 and K2 are the two binding constants for the unlabeled drug D at the two independent binding sites that have respective capacities of B1 and B2. In this case, the independent variable is [D] and the dependent variable is [L*R]. [L*] is fixed at a known value and this value is inserted into the model before iterative fitting to the data. The Kd for the radioligand is determined in independent experiments by analysis of saturation curves (either with the computer or by Scatchard analysis). The six parameters on the righthand side of Equation 1.3.34 are thus reduced to four, which can be estimated iteratively by the computer. Constraints are applied during the calculation of the four parameters so that the computer does not attempt to fit the data with negative parameter values. An alternate form of Equation 1.3.34 uses affinity binding constants (the inverse of the dissociation constant):
[L * R] = B1 (1 + [ L*] K A ) 1 + [ L*] K A + [ D] K1 + B2 (1 + [ L*] K A ) 1 + [ L*] K A + [ D] K 2
Equation 1.3.35

in which KA, K1, and K2 are the affinity constants (expressed in liter/mol) for the radioligand and displacing agent, respectively. Multiplicity of binding sites can occur in a single receptor population if the receptor changes conformation and consequently binds the ligand with a different affinity. Formally, the two binding sites are not independent because they interconvert. An example of this is the ternary complex model (DeLean et al., 1980; Wreggett and DeLean, 1984; UNIT 1.2), which describes a mechanism by which multiple binding sites result from the interaction of the agonist-receptor complex with GTP-bind-

1.3.28
Supplement 10 Current Protocols in Pharmacology

ing proteins (Tolkovsky and Levitzki, 1981). Agonist binding to GTP-coupled receptors in situ is typically complex, with concave Scatchard plots and Hill slopes <1.0. The use of the complete ternary complex model requires the determination of several additional parameters describing the interactions of receptor with agonist, receptor with GTP-binding protein, and GTP-binding protein with GTP/GDP. In most binding experiments only some of these parameters can be determined, while others are estimated or determined independently. Statistically, the binding data are fit equally well by models containing two independent receptors with different affinities as with models that describe the agonist-induced interconversion of a single receptor into multiple binding states (Abramson et al., 1987). In such cases, use of the goodness-of-fit after minimization does not aid in the selection of the most appropriate binding model. For this, additional biochemical information is required. Iterative fitting methods The equations that describe equilibrium or time-dependent radioligand binding to receptors are nonlinear in one or more of their parameters. This means linear regression cannot be used to determine binding parameters without some kind of data transformation, such as with the Scatchard plot or the Lineweaver-Burke plot. Since data transformation can distort the estimates of parameters, a better approach is to fit the nonlinear model directly to the data. This is accomplished by computer iteration, with the operator supplying the model and initial estimates of the parameters, and the computer using an algorithm to progressively alter the parameter values to reduce variance to a minimum (a level usually selected by the operator). A number of computer programs use the Gauss-Newton procedure modified by the Marquardt-Levenberg method of approaching the minimal sum-of-squared residuals (Marquardt, 1963). Examples include ALLFIT, which is used with dose-response experiments (DeLean et al., 1978), and LIGAND, which is popular for binding analysis (Munson and Rodbard, 1980). When using an iterative fitting routine, the mathematical model is defined or selected and the operator supplies initial estimates of the parameter values. Using numerical approximations to the derivatives of the binding equation and matrices with elements determined by the number of data points and parameters, the computer calculates new parameter values which, when inserted into the model, reduce the vari-

ance between the model and the data. With each such iteration the parameter values are refined in a step-wise manner to eventually reach the minimum variance (lowest sum-of-squared residuals), which is defined as the point at which successive iterations do not improve the fit (reduce the minimum) by more than a certain predetermined fraction (e.g., 0.01). This result of iterative fitting is referred to as convergence. The number of iterations required to converge on a solution depends on the fitting algorithm, the quality of the data, the complexity of the model, and the accuracy of the initial parameter estimates. In some instances, the computer will seemingly iterate forever, as if it cannot converge. In such cases it is likely the data quality is poor or that some data are missing. In this regard, the data at the extremities of the binding curve are particularly necessary for convergence. With actual binding data, false (local) minima may be determined by these iterative methods, and the computer may provide inaccurate parameter estimates. It is important to be aware of the possibility of encountering a local minimum when evaluating a fit provided by the computer. The better the data (i.e., the closer the points lie to the fitted curve) and the more data points collected, the more powerful is the method in finding the true (global) minimum sum-of-squared residuals and in providing good estimates of the parameters. To assess whether the solution is at the lowest minimum sum of squares, run the program on the data several times altering the initial values for the parameters. Even when the computer converges on a solution in a reasonable number of iterations, rerun the problem by inputting another set of estimates for the parameters to determine whether the same, or nearly the same, solution is given. If so, it can be assumed that the computer is producing a reliable set of parameter estimates. An alternative to minimization methods using differential equations is the simplex method of Nelder and Mead (1965). A comparison of the methods in their application to enzyme kinetics is provided by Lam and Cross (1979). An example of the use of the simplex minimization is Karlsson and Neil (1989). Statistical analysis To mathematically define the goodness-offit of the parameter estimates in a single experiment, computer programs sometimes provide a standard deviation or standard error for the estimate of each parameter. Because the fitting method is nonlinear, this parameter error is

Receptor Binding

1.3.29
Current Protocols in Pharmacology

Practical Aspects of Radioligand Binding

itself an estimate (using a linear approximation) and cannot be used in reporting the reliability of parameter estimates. To obtain reliable values, repeated experiments must be run and standard statistical tests must be performed on sets of parameter estimates obtained in the independent experiments. For a saturation experiment based on a simple bimolecular equilibrium model in which two parameters (Bmax and Kd) are to be determined, ten data points (sets comprised of values of [L*] and [L*R]) determined in triplicate are usually sufficient for calculation of good estimates of the two parameters. In this case, the degrees of freedom (the number of data points minus the number of parameters) is eight. If the experiment is repeated several times it is usually possible to obtain standard errors of 10% or less for the parameter estimates. When the complexity of the model is increased as with two binding sites, more parameters must be estimated (four in this case). With good data quality, ten data points will provide reliable estimates of these four parameters, although the estimates will be at the outer edge of the accuracy of the methodology (with six degrees of freedom). With replicate assays, the estimates of the parameters will vary severalfold. In an individual assay, the confidence intervals for the parameters (i.e., estimates of the possible range of parameter values, provided by the iteration routine) may reach an order of magnitude or more. The simple way to resolve this problem is to increase the number of data points. For a two-site model, 15 to 30 (or more) data points in each independent experiment are typically necessary. If a more complex model is biologically reasonable entailing additional parameters, and when experiments produce better data (lower variance and more data sets), the more complex mathematical model may be considered. In the analysis of the competition curve shown in Figure 1.3.3, it was necessary to obtain 36 data points in quadruplicate to attain statistical reliability in the fitting of a three-site model. For a more complex model to be justified, there should be a sufficient improvement in fit gained by adding parameters (i.e., there should be a sufficient reduction in the variance). In general, for a typical binding assay with 10 to 15 data points the sum of squares should be reduced several-fold. To statistically validate the choice of the more complex model, the F-ratio test is the most appropriate (Kenakin, 1993). The data are fit to each of the two alternate models using the iterative computer

program, ensuring that the true minimum is achieved with each fit and that the parameter estimates are reasonable. The sums-of-squared residuals are then obtained for each model. The formula for calculating the F-ratio (DeLean et al., 1978) is given in Equation 1.3.36,

F=

(SSA SSB ) / (dfA dfB ) SSB / dfB


Equation 1.3.36

where SSA and SSB refer to the sum-of-squared residuals after fitting to models A and B, and dfA and dfB are the respective degrees of freedom (the difference between the number of data points and the number of parameters). As an example, a comparison between two models used to account for an experiment comprised of ten data points is shown here. Model A has one binding site and two parameters (Kd and Bmax). The sum-of-squared residuals (SSA) = 57, and there are 8 degrees of freedom (dfA = 10 data points 2 parameters). Model B has two binding sites and four parameters (K1, B1, K2, and B2); SSB = 21 and dfB = 10 data points 4 parameters = 6. Using Equation 1.3.36, the F-ratio is calculated as follows:
F= (57 21) / (8 6) 21 / 6
Equation 1.3.37

= 5.143

A table providing the F-distribution for the desired level of probability (typically, P = 5% is used) may be consulted in a textbook on statistics. Columns correspond to the difference in degrees of freedom (dfA dfB, in the numerator of Equation 1.3.36), while rows correspond to the number of degrees of freedom for the less complex model (dfB, in the denominator of Equation 1.3.36). For the above example, the F-ratio for significance at the 5% level is 10.92, shown at the intersection of the column headed by the numeral 2 and the row headed by the numeral 6. Since the calculated value of 5.143 is less than 10.92, the two-site model is not justified at this level of probability. The improvement in fit gained by adding two parameters is not statistically justified because the variance was not sufficiently reduced. To attain significance, the sum of squares would need to be reduced approximately another two-fold, or the number of data points would need to be doubled. Studies using the two-independent-bindingsite model have shown that with typical data

1.3.30
Current Protocols in Pharmacology

the iterative fitting techniques can detect a minor binding site with abundance as low as 10%, if the ligand binds to the two sites with a 100-fold or so difference in affinity (DeLean et al., 1982). When the sites are more equal in abundance, the difference in affinity can be as low as 10-fold without the two sites becoming indistinguishable (McKinney et al., 1985). The Students t test is typically used to assess whether there are significant differences in parameter values obtained under differing conditions, with P < 0.05 taken as the minimal level of significance. The t distribution is an approximation of the normal (Gaussian) distribution, and tests of population differences associated with the normal distribution assume that the data follows the normal curve. As Gaddum (1945) pointed out, it is essential to verify whether the data actually follow a normal distribution before using one of these tests. It is also necessary to assume that the variance is constant over the range of the data being analyzed. For data such as those used for dose-response curves and for receptor binding, where the independent variable is the compound or radioligand concentration, the EC50 or Kd values are not normally distributed, although their logarithmic transforms are (Fleming et al., 1972). Thus, the correct way to determine whether differences between such data sets are statistically significant is to compute the logarithms of the Kd values or EC50 values, obtain the means and standard errors of the logarithmic values of the data, and determine the level of significance from t tables. To present the mean in arithmetic form, the antilogarithm of the mean is used. The standard error of the mean in arithmetic form is calculated as the product of the arithmetic mean and the standard error of the averaged logarithmic transforms (DeLean et al., 1982). Logarithmic transforms are not required for the statistical analysis of binding capacities (Bmax).

such that either the radioligand or the receptor preparation are unstable over time. Include total and nonspecific binding assay tubes at regular intervals (e.g., every 36 tubes) to correct for the drift.

LITERATURE CITED
Abramson, S.N., McGonigle, P., and Molinoff, P.B. 1987. Evaluation of models for analysis of radioligand binding data. Mol. Pharmacol. 31:103111. Beck, J.S. and Goren, H.J. 1983. Simulation of association curves and Scatchard plots of binding reactions where ligand and receptor are degraded or internalized. J. Recept. Res. 3:561577. Bennett, J.P. Jr. and Yamamura, H. 1986. Neurotransmitter, hormone, or drug receptor binding methods. In Neurotransmitter Receptor Binding, 2nd ed. (H.I. Yamamura, S.J. Enna, and M.J. Kuhar, eds.) pp.61-90. Raven Press, New York. Borea, P.A., Dalpiaz, A., Varani, K., Gessi, S., and Gilli, G. 1996. Binding thermodynamics at A1 and A2A adenosine receptors. Life Sci. 59:13731388. Burgisser, E. 1984. Radioligand-receptor binding studies: Whats wrong with the Scatchard analysis? Trends Pharmcol. Sci. 5:142-145. Cheng, Y.-C. and Prusoff, W.H. 1973. Relationship between the inhibition constant (KI) and the concentration of inhibitor which causes 50 per cent inhibition (I50) of an enzymatic reaction. Biochem. Pharmacol. 22:3099-3103. Contreras, M.L., Wolfe, B.B., and Molinoff, P.B. 1986. Thermodynamic properties of agonist interactions with the beta adrenergic receptor-coupled adenylate cyclase system. I. High- and lowaffinity states of agonist binding to membranebound beta adrenergic receptors. J. Pharmacol. Exp. Ther. 237:154-164. Cuatrecasas, P. and Hollenberg, M.D. 1976. Membrane receptors and hormone action. Adv. Protein Chem. 30:251-451. DeLean, A., Munson, P.J., and Rodbard, D. 1978. Simultaneous analysis of families of sigmoidal curves: Application to bioassay, radioligand assay, and physiological dose-response curves. Am. J. Physiol. 235:E97-E102. DeLean, A., Stadel, J.M., and Lefkowitz, R.J. 1980. A ternary complex model explains the agonistspecific binding properties of the adenylate cyclase-coupled -adrenergic receptor. J. Biol. Chem. 255:7108-7117. DeLean, A., Hancock, A.A., and Lefkowitz, R.J. 1982. Validation and statistical analysis of a computer modeling method for quantitative analysis of radioligand binding data for mixtures of pharmacological receptor subtypes. Mol. Pharmacol. 21:5-16. Ehlert, F.J., Roeske, W.R., and Yamamura, H.I. 1981. Mathematical analysis of the kinetics of competitive inhibition in neurotransmitter recep-

TROUBLESHOOTING
Individual binding assays exhibit their own idiosyncrasies. However, there are some generic guidelines that are applicable to all binding assays. Loss of specific binding. Assess radioligand purity and possibility of degradation. Make fresh buffers and a new tissue preparation. Ligand pharmacology changes. Assess radioligand purity and possibility of degradation. Make fresh buffers and a new tissue preparation. Total and specific binding decrease over the course of the assay. The assay conditions are

Receptor Binding

1.3.31
Current Protocols in Pharmacology

tor binding assays. Mol. Pharmacol. 19:367371. Feldman, H.A. 1972. Mathematical theory of complex ligand-binding systems at equilibrium: Some methods for parameter fitting. Anal. Biochem. 48:317-338. Feldman, H.A. 1983. Statistical limits in Scatchard analysis. J. Biol. Chem. 258:12865-12867. Filer, C., Hurl, S., and Wan, Y.-P. 1989. Radioligands: Synthesis and handling. In Receptor Binding. (M. Williams, P.B.M.W.M. Timmermans, and R.A. Glennon, eds.) pp. 105-135. Marcel Dekker, New York. Fleming, W.W., Westfall, D.P., De La Lande, I.S., and Jellett, L.B. 1972. Log-normal distribution of equieffective doses of norepinephrine and acetylcholine in several tissues. J. Pharmacol. Exp. Ther. 181:339-345. Gaddum, J.H. 1945. Lognormal distributions. Nature 156:463-466. Hill, A.W. 1910. The possible effects of the aggregation of the molecules of hemoglobin on its dissociation curves. J. Physiol. (Lond.) 40:iv-vii. Jacobs, S., Chang, K.-J., and Cuatrecasas, P. 1975. Estimation of hormone receptor affinity by competitive displacement of labeled ligand: Effect of concentration of receptor and of labeled ligand. Biochem. Biophys. Res. Commun. 66:687-692. Jarv, J., Hedlund, B., and Bartfai, T. 1979. Isomerization of the muscarinic receptor-antagonist complex. J. Biol. Chem. 254:5595-5598. Karlsson, M.O. and Neil, A. 1989. Estimation of ligand binding parameters by simultaneous fitting of association and dissociation data: A Monte Carlo simulation study. Mol. Pharmacol. 35:59-66. Kenakin, T. 1993. Radioligand binding experiments. In Pharmacologic Analysis of Drug-Receptor Interaction, 2nd ed., pp. 385-410. Raven Press, New York. Ketelslegers, J.-M., Pirens, G., Maghuin-Rogister, G., Hennen, G., and Frere, J.-M. 1984. The choice of erroneous models of hormone-receptor interactions: A consequence of illegitimate utilization of Scatchard graphs. Biochem. Pharmacol. 33:707-710. Koshland, D.E., Nemthy, G., and Filmer, D. 1966. Comparison of experimental binding data and theoretical models in proteins containing subunits. Biochemistry 5:365-385. Lam, C.F. and Cross, A.P. 1979. Comparative study of parameter estimation procedures in enzymic kinetics. Comput. Biol. Med. 9:145-153. Lazareno, S. and Birdsall, N.J. 1993. Estimation of competitive antagonist affinity from functional inhibition curves using the Gaddum, Schild and Cheng-Prusoff equations. Br. J. Pharmacol. 109:1110-1119. Practical Aspects of Radioligand Binding Limbird, L. 1996. Cell Surface Receptors: A Short Course in Theory and Methods, 2nd ed. Kluwer Academic Publishers, Boston.

Linden, J. 1982. Calculating the dissociation constant of an unlabeled compound from the concentration required to displace radiolabel binding by 50%. J. Cyclic Nucleotide Res. 8:163-172. Maksay, G. 1994. Thermodynamics of gamma-aminobutyric acid type A receptor binding differentiate agonists from antagonists. Mol. Pharmacol. 46:386-390. Marquardt, D.W. 1963. An algorithm for leastsquares estimation of nonlinear parameters. J. Soc. Indust. Appl. Math. 11:431-441. McKinney, M., Stenstrom, S., and Richelson, E. 1985. Muscarinic responses and binding in a murine neuroblastoma clone (N1E-115). Mediation of separate responses by high affinity and low affinity agonist-receptor conformations. Mol. Pharmacol. 27:223-235. Metcalf, M.A., McGuffin, R.W., and Hamblin, M.W. 1992. Conversion of the human 5-HT1Db serotonin receptor to the rat 5-HT1B ligand-binding phenotype by Thr355Asn site directed mutagenesis. Biochem. Pharmacol. 44:1917-1920. Monod, J., Wyman, J., and Changeux, J.-P. 1965. On the nature of allosteric transitions: A plausible model. J. Mol. Biol. 12:88-118. Motulsky, H. and Neubig, R. 1997. Analyzing radioligand binding data. In Current Protocols in Neuroscience (J.N., Crawley, C.R., Gerfen, R., McKay, M.A., Rogawski, D.R., Sibley, and P., Skolnick, eds.) pp. 7.5.1-7.5.55. John Wiley & Sons, New York. Motulsky, H.J., Mahan, L.C., and Insel, P.A. 1985. Radioligand, agonists and membrane receptors on intact cells: Data analysis in a bind. Trends Pharmacol. Sci. 6:317-319. Munson P.J. and Rodbard, D. 1980. LIGAND: A versatile computerized approach for characterization of ligand-binding systems. Anal. Biochem. 107:220-239. Murphy, D.E., Schnieder, J., Boehm, C., Lehmann, J., and Williams, M. 1987. Binding of [3H]3-(2carboxypiperazin-4-yl)propyl-1-phosphonic acid to rat brain membranes: A selective, highaffinity ligand for N-methyl-D-aspartate receptors. J. Pharmacol. Exp. Ther. 240:778-784. Murphy, D.E., Hutchison, A.J., Hurt, S.D., Williams, M., and Sills, M.A. 1988. Characterization of the binding of [3H]CGS 19755: A novel N-methyl-D-aspartate antagonist with nanomolar affinity in rat brain. Br. J. Pharmacol. 95:932938. Nelder, J.A. and Mead, R. 1965. A simplex method for function minimization. Computer J. 7:308313. Pang, Y.-P., Cusack, B., Groshan, K., and Richelson, E. 1996. Proposed ligand binding site of the transmembrane receptor for neurotensin (8-13). J. Biol. Chem. 271:16060-16068. Pellicciari, R., Marinozzi, M., Natalini, B., Costantino, G., Luneia, R., Giorgi, G., Moroni, F., and Thomsen, C. 1996. Synthesis and pharmacological characterization of all sixteen stereoisomers of 2-(2-carboxy-3-phenylcyclopropyl)glycine.

1.3.32
Current Protocols in Pharmacology

Focus on (2S,1S,2S,3R)-2-(2-carboxy-3phenylcyclopropyl)glycine, a novel and selective group II metabotropic glutamate receptor antagonist. J. Med. Chem. 39:2259-2269. Rabow, L.E., Russek, S.J., and Farb, D.H. 1995. From ion currents to genomic analysis: Recent advances in GABA-A receptor research. Synapse 21:189-274. Rodbard, D. 1973. Mathematics of hormone-receptor interaction. I. Basic principles. Adv. Exp. Med. Biol. 36:289-329. Rodbard, D., Munson, P.J., and Thakur, A.K. 1980. Quantitative characterization of hormone receptors. Cancer 46:2907-2918. Rosenthal, H.E. 1967. Graphic method for the determination and presentation of binding parameters in a complex system. Anal. Biochem. 20:525-532. Scatchard, G. 1949. The attractions of proteins for small molecules and ions. Ann. N.Y. Acad. Sci. 51:660-672. Schreiber, G., Henis, Y.I., and Sokolovsky, M. 1985. Rate constants of agonist binding to muscarinic receptors in rat brain medulla. Evaluation by competition kinetics. J. Biol. Chem. 260:87958802. Sullivan, J.P., Decker, M.W., Brioni, J., DonnellyRoberts, D., Anderson, D.A., Bannon, A.W., Kang, C.-H., Adams, P., Piattoni-Kaplan, M., Buckley, M.J., Gopalakrishnan, M., Williams, M., and Arneric, S.P. 1994. (+)-Epibatidine elicits a diversity of in vitro and in vivo effects mediated by nicotinic acetylcholine receptors. J. Pharmacol. Exp. Ther. 271:624-631. Tallarida, R.J. and Murray, R.B. 1986. Manual of Pharmacologic Calculations with Computer Programs, 2nd ed. Springer-Verlag, New York. Tolkovsky, A.M. and Levitzki, A. 1981. Theories and predictions of models describing sequential interactions between the receptor, the GTP regulatory unit, and the catalytic unit of hormone dependent adenylate cyclases. J. Cyclic Nucleotide Res. 6:139-150. Waelbroeck, M., Tastenoy, M., Camus, J., and Christophe, J. 1990. Binding of selective antagonists to four muscarinic receptors (m1 to m4) in rat forebrain. Mol. Pharmacol. 38:267-273. Waelbroeck, M., Camus, J., Tastenoy, M., Lambrecht, G., Mutschler, E., Kropfgans, M., Sperlich, J., Wiesenberger, F., Tacke, R., and Christophe, J. 1993. Thermodynamics of antagonist binding to rat muscarinic M2 receptors: Antimuscarinics of the pridinol, sila-pridinol, diphenidol and sila-diphenidol type. Br. J. Pharmacol. 109:360-370. Weiland, G.A., Minneman, K.P., and Molinoff, P.B. 1979. Fundamental difference between the molecular interactions of agonists and antagonists with the -adrenergic receptor. Nature 281:114-117.

Wild, K.D., Porreca, F., Yamamura, H.I., and Raffa, R.B. 1994. Differentiation of receptor subtypes by thermodynamic analysis: Application to opioid receptors. Proc. Natl. Acad. Sci. U.S.A. 91:12018-12021. Williams, M. and Gordon, E. M. 1996. Drug discovery: An overview. In A Textbook of Drug Design and Development, 2nd ed. (P. KrogsgaardLarsen, T. Liljefors, and U. Madsen, eds.) pp. 1-34. Harwood Academic Publishers, Chur, Switzerland. Williams, M., Deecher, D.C., and Sullivan, J.P. 1995. Drug receptors. In Burgers Medicinal Chemistry and Drug Discovery, 5th ed. (M.E. Wolff, ed.) pp. 349-397. John Wiley & Sons, New York. Wreggett, K.A. and DeLean, A. 1984. The ternary complex model. Its properties and application to ligand interactions with the D2-dopamine receptor of the anterior pituitary gland. Mol. Pharmacol. 26:214-227. Yasuda, R.P., Ciesia, W., Flores, L.R., Wall, S.J., Li, M., Satkus, S.A., Weisstein, J.S., Spagnola, B.V., and Wolfe, B.B. 1993. Development of antisera selective for m4 and m5 muscarinic cholinergic receptors: Distribution of m4 and m5 muscarinic receptors in rat brain. Mol. Pharmacol. 43:149157.

KEY REFERENCES
Kenakin, 1993. See above. A complete treatise for the advanced student. McGonigle, P. and Molinoff, P.B. 1994. Receptors and signal transduction: Classification and quantitation. In Basic Neurochemistry: Molecular, Cellular, and Medical Aspects, 5th ed. (G.J. Siegel, ed.) pp. 210-230. Raven Press, New York. An introductory treatment within the context of neuropharmacology. Weiland, G.A. and Molinoff, P.B. 1981. Quantitative analysis of drug-receptor interactions: I. Determination of kinetic and equilibrium properties. Life Sci. 29:313-330. Useful for explaining common plotting methods. Yamamura, H.I., Enna, S.J., and Kuhar, M.J. 1985. Neurotransmitter Receptor Binding, 2nd ed. Raven Press, New York. This volume has been a standard in the field for many years and is especially useful for beginners.

Contributed by Michael McKinney Mayo Clinic Jacksonville Jacksonville, Florida

Receptor Binding

1.3.33
Current Protocols in Pharmacology

Anda mungkin juga menyukai