Section5:Practical Use of Fluorescence Polarization in Competitive Receptor Binding Assays

From Assay Guidance Wiki

A fluorophore whose absorption vector is aligned with polarized excitation light is selectively excited. If the fluorophore tumbles rapidly relative to its fluorescent lifetime then it will be randomly orientated prior to light emission and therefore will show a low polarization value (situation A above). However, if this fluorophore’s rotation is slowed down so that it tumbles slowly with respect to the fluorescent lifetime (e.g. by binding to a large receptor as shown in B above) it will not rotate much before light emission and will show a high polarization value. The dependence of polarization on fluorescent life-time is shown below.

[The graph above contains simulated data using the Perrin equation (Cantor and Schimmel, 1980) and taking the limiting polarization as 0.5 using T = 293 K and assuming a spherical protein in water with the fluorescence probe rigidly attached]

Typical fluorophores include fluorescein- or BODIPY-labels that have fluorescence lifetimes allowing FP measurements to be made between a small labeled-ligand (<1500 Da) and a large receptor (e.g. > 10,000 Da).

The increase in polarization can be measured with several microplate readers where the fluorescence is measured using polarized excitation and emission filters. Two measurements are performed on every well. Data is obtained for the fluorescence perpendicular to the excitation plane (the “P-channel”) and fluorescence that is parallel to the excitation plane (the “S-channel”). For screening applications, the millipolarization units (mP) are often calculated using:

The proper use of S and P channel data requires two corrections. First, accurate calculation of polarization using fluorescent readers requires calculation of the instrument “G-factor”. This factor corrects for any bias toward the P channel. For microplate readers, a 1 nM fluorescein solution is typically used and the G-factor that yields a value of 27 mP is entered (27 mP is the known value for a 1 nM fluorescein solution at R.T). Secondly, the S and P values should have the background fluorescence subtracted (determined using assay buffer without labeled-ligand in the well).

Fluorescence Polarization and Receptor Binding

Receptor-binding FP assays use a small molecule labeled ligand (so called tracer) and a large unlabeled receptor. An example is a fluorescently labeled-steroidal ligand binding to a nuclear receptor-ligand binding domain (kits of this type are sold by Invitrogen/PanVera). This type of assay typically yields a minimum signal of approximately 50 mP for the unbound tracer and a maximum signal of approximately 300 mP when the tracer is fully bound to the receptor.

Validate Activity of Fluorescent Tracer

The receptor binding activity of a fluorescent-labeled tracer can be determined in a competition assay using a radiolabeled ligand and traditional methods of receptor binding (filtration, SPA, charcoal precipitation, etc.). As shown in the figure below, some loss of receptor binding activity may occur following fluorescent tagging. It is important to identify lower binding activity prior to further experiments with the fluorescent tracer. Functional receptor assays, such as cAMP measurement, calcium mobilization or GTPγS binding, can also be performed to determine if there has been a loss in biological activity as a result of the labeling process.

Choosing Tracer and Receptor Concentrations

The Kd of the tracer and the amount of tracer bound under the chosen assay conditions will be required for analysis of competitive binding parameters. Typically, the Kd can be estimated using radioligand-binding techniques (SPA, filtration) discussed in previous sections, provided there is not significant deviation in the potency of the tracer and the unlabeled molecule (see figure above). It may be useful to perform a tracer calibration curve by varying the amount of tracer and ensuring that the polarization signal is constant over a reasonable concentration range, inclusive of the estimated Kd. By definition, the polarization signal is independent of the intensity of the tracer. This also identifies the variability at the tracer concentration to be used. The polarization signal as a function of tracer concentration is shown for a representative tracer in the figure below. Note that as the signal nears the limits of sensitivity for the detector, the variation increases.

The amount of bound tracer can be measured in an experiment where the tracer is held at a constant concentration near its Kd and the receptor concentration is then varied. An example of this type of experiment using the glucocorticoid receptor (GR) included in the FP kit available from Invitrogen/PanVera is shown below. Here the ligand-binding domain of GR is varied using a constant Kd concentration of a labeled-steroidal ligand (Fluormone™, Invitrogen/Panvera kits; Data provided by Pharmacopeia).

In these types of FP experiments no correction for nonspecific binding (NSB) is performed as was shown in earlier sections for radioligand-binding experiments. This is because the tracer (what is the radioactive ligand concentration in traditional assays) is held constant at a concentration usually near the Kd and the protein receptor concentration is then varied over several orders of magnitude. However, this assumption should be checked by observing the polarization of the ligand in the absence of receptor. (Caution: it is possible to observe increasing FP signals when membrane receptors are used due to light scattering. In those cases, a correction may need to be made by measuring the signal in the presence and absence of the fluorescent tracer). If binding to non-specific buffer components or microtiter plates surfaces is observed then this tracer should be avoided. An analytical treatment of FP competitive-binding data has recently been presented by Roehrl et al. (2004) that allows one to quantify the effect of non-specific binding on FP titration curves.

Examination of the curve above allows one to choose a receptor concentration that yields an acceptable assay window (typically a ΔmP of between 150 mP and 300 mP).

Pharmacological Profile

Sensitivity to known competitors should be checked at this stage to ensure that the developed FP assay is adequate for the intended purpose. An example pharmacological profile using fluorescence polarization is shown below.

Ligand Depletion

The FP assay format is homogenous in nature and therefore lends itself to simple “mix and read” protocols. However, to obtain an acceptable signal, the assay must be set-up with a large fraction of the tracer bound to the receptor (typically >80 %). The high amount of bound tracer requires a specific set of equations to be used when interpreting FP derived competition binding results.

In these cases, where a large amount of bound tracer exist, the Cheng-Prusoff equation as mentioned in the discussion of heterologous competition-receptor binding (see p. 27) will always lead to an overestimation of the Ki from the IC50. This is because the Cheng-Prusoff equation is strictly given as:

{Eq. 1}

In the case of FP displacement-binding, the free ligand term [Lf] can not be substituted for the total ligand concentration [L] because there is little free ligand available. This differs from the typical saturation-binding experiments mentioned in previous sections.

Three equations have been presented in the literature to provide a solution to this situation for simple competitive-binding. Munson and Rodbard (Munson and Rodbard, 1988) provide a correction that takes into account the amount of bound tracer. This takes the form of:

{Eq. 2}

Where yo is the bound/free ratio of tracer and Lo is the total tracer concentration.

Huang provides an alternative form of this correction in terms of the fraction of bound tracer (Huang, 2003). Rearrangement of Equation 15 given in Huang to solve for Ki yields:

{Eq. 3}

Where Fo is the fraction of tracer bound and Lo is the total tracer concentration. Huang’s result is redundant with the earlier Munson and Rodbard equation except for expressing the equation in terms of the fraction of tracer bound. Therefore, Eq. 2 and Eq 3 yield the same correction (see below).

The final correction often used in this situation is the one derived by Kenakin (1993). Here the equation is expressed in terms of total receptor concentration (Ro), the total tracer concentration (Lo as above) and the bound tracer concentration (Lb).

{Eq. 4}

These equations should be used instead of Cheng-Prusoff when > 10% of the tracer is bound to the receptor in the assay.

Application of Ligand Depletion Equations Once a suitable choice of receptor and tracer concentrations have been made and the resulting assay has been shown to be useful for competitive binding analysis, one can calculate the amount of bound tracer under the assay conditions taking the lower and upper asymptotes as values for free and bound tracer respectively.

Some example competition-binding data (Fluormone™ kit, Invitrogen/Panvera) are shown in Table I to illustrate the differences between using the Cheng-Prusoff equation without correction for the amount of bound tracer or each of the above equations which correct for tracer depletion. For these competition-binding experimental results the conditions were:

Equilibrium dissociation constant, Kd = 0.6 nM (Fluormone™ ligand), determined using
Bound Tracer Concentration, Lb = 0.9 nM, determined from receptor concentration experiment at constant tracer (Lo), by reading the mP signal and determining the % of maximum
Total Tracer Concentration, Lo = 1 nM, concentration set near the Kd value
Total Receptor Concentration, Ro = 4 nM (GR ligand-binding domain), determined from receptor concentration experiment at constant tracer – yields statistically valid assay with robust signal

These concentrations yield the following terms required for Equations 2-4:

Bound/Free ratio of Tracer, yo = Lb/(Lo – Lb) = 0.9/(1-0.9) = 9
Fraction of Tracer Bound, Fo = Lb/Lo = 0.9/1 = 0.9

Table I. Comparison of Ki values determined from ligand depletion correction formulas. Values are in nM. IC50 shown is the measured IC50 under the assay conditions described in the text. All other values are calculated values. Data provided by Pharmacopeia.

		(1)	Ki, nM
Ligand	IC₅₀	Cheng-Prusoff	Munson-Rodbard	Huang	Kenakin
Cortisone	8.0	3.0	0.24	0.24	0.6
Dexamethasone	3.6	1.3	-0.16	-0.16	0.3
Estradiol	815	306	74	74	63
Testosterone	229	86	20	20	18
Compound 1	6.4	2.4	0.09	0.09	0.5
Compound 2	1000	375	91	91	77

A graphical representation of the data is shown below.

Application of Cheng-Prusoff under these conditions can lead to more than 10-fold overestimations of Ki. In many cases all three equations yield similar corrections and as mentioned above Munson & Rodbard and Huang yield identical values. However, one issue with the Munson & Rodbard and Huang type corrections is that certain combinations of IC50, Kd and bound tracer yield impractical negative values of Ki. This has been discussed in the literature as a breakdown in additional assumptions buried within these equations such as competitive inhibition with a single binding site. For this reason, the Kenakin equation is commonly chosen for performing this correction. Additionally, curve fitting to the equations given in Roehrl et al. (2004) can be used to examine if complete inhibition is achieved as well as the KD of the competitor compound.

Detection of Fluorescent Interference From Compounds in FP Screens

All FP experiments start with measuring polarized prompt fluorescence from the assay well. This makes these experiments susceptible to fluorescence interference by compounds present in the well. However, a helpful method to address this issue has been presented by Turconi et al. (2001). This paper calculates the total fluorescence intensity from a well (given by S + 2P; see references in above paper) and the observed anisotropy[1] from the each well to flag false positive wells due to fluorescence interference.

An example is provided below to illustrate the use of this method. Plots of the total fluorescence intensity (normalized to the control well values, e.g. the total fluorescence intensity of the assay in the absence of compounds) versus the anisotropy are shown below.

Three cases are illustrated in the figure above. In case A, the compounds in the wells are not active or fluorescent. Therefore the measured Fluormone tracer is bound to GR ligand-binding domain (GR-LBD) and the anisotropy values are clustered around the 0% inhibition value. Furthermore, there is no change in fluorescence intensity in the compound-containing wells relative to the control wells. In case B, the compounds in the well are active in the assay but not fluorescent. Therefore, the tracer is being displaced from the GR-LBD and the anisotropy values distribute from high to low inhibition values. Again, there is no change in the total fluorescence intensity. In case C, the compounds appear active as they show a decrease in anisotropy values suggesting that the tracer has been displaced from the GR-LBD. However there is a correlation between decreasing anisotropy and increasing fluorescence intensity in the wells with the lowest anisotropy values showing more than a 35-fold increase in the fluorescent intensity relative to control values. This suggests that the measured FP is due to the compounds themselves rather than the tracer.

In typical FP-receptor binding experiments the tracer is kept at a low nM concentration while the compounds that are being screened are typically in the μM range. If these compounds are fluorescent at the detection wavelengths then their fluorescence can easily overcome that of the tracer. As compounds in screening campaigns are typically of low molecular weight (<500 Da) they will exhibit low anisotropy values. Compounds in case C were of this type and subsequent secondary assays showed them to be inactive. A final case not shown above is where the compounds are both fluorescent and active. Turconi et al. present an equation that can be used to fit the fluorescent intensity data to the case where anisotropy changes without displacement of the ligand (see Equation 4 and discussion therein of Turconi et al.). The solid line in case C above shows an example of this fit. One can then evaluate outliers from this curve fit in terms of potential active but fluorescent compounds.

It is also possible to observe changes in polarization that are due to fluorescent compounds present as aggregates. In this case, the fluorescence intensity will increase along with the polarization as long as the aggregation does not quench the fluorescence. Additionally, light scattering from particulates or compound participates can lead to apparently high polarization values. For receptor binding experiments as described above this superfluous increase in polarization may mask any decrease in polarization due to an active compound and thus result in a false negative. Careful examination of the fluorescence intensity versus polarization plots should identify these artifacts.

[1] Anistropy is derived by measuring the S and P channels as described above, however the fluorescence is expressed with the denominator representing the total fluorescence intensity from the sample. The equation for calculating anisotropy is given by:

Anisotropy and polarization are related by the equations given below where P is the polarization and a is the anisotropy:

In general, anisotropy is more useful analyzing complex systems or mixtures as the equations are simpler to express in terms of anisotropy. (Cantor and Schimmel, 1980). Arguably, screening data should be presented in terms of anisotropy rather than polarization but this convention has not been adopted as yet.