Oligomerization of yeast α-factor receptor detected by fluorescent energy transfer between ligands

Abstract

G-protein-coupled receptors (GPCRs) comprise a large superfamily of transmembrane receptors responsible for transducing responses to the binding of a wide variety of hormones, neurotransmitters, ions, and other small molecules. There is extensive evidence that GPCRs exist as homo-and hetero-oligomeric complexes; however, in many cases, the role of oligomerization and the extent to which it occurs at low physiological levels of receptor expression in cells remain unclear. We report here the use of flow cytometry to detect receptor-receptor interactions based on fluorescence resonance energy transfer between fluorescently labeled cell-impermeant ligands bound to yeast α-mating pheromone receptors that are members of the GPCR superfamily. A novel, to our knowledge, procedure was used to analyze energy transfer as a function of receptor occupancy by donor and acceptor ligands. Measurements of loss of donor fluorescence due to energy transfer in cells expressing high levels of receptors were used to calibrate measurements of enhanced acceptor emission due to energy transfer in cells expressing low levels of receptors. The procedure allows determination of energy transfer efficiencies over a 50-fold range of expression of full-length receptors at the surface of living cells without the need to create fluorescent or bioluminescent fusion proteins. Energy transfer efficiencies for fluorescently labeled derivatives of the receptor agonist α-factor do not depend on receptor expression level and are unaffected by C-terminal truncation of receptors. Fluorescently labeled derivatives of α-factor that act as receptor antagonists exhibit higher transfer efficiencies than those for labeled agonists. Although the approach cannot determine the number of receptors per oligomer, these results demonstrate that ligand-bound, native α-factor receptors exist as stable oligomers in the cell membranes of intact yeast cells at normal physiological expression levels and that the extent of oligomer formation is not dependent on the concentration of receptors in the membrane.

Significance

The role of oligomerization of receptors in the important family of G-protein-coupled receptors remains poorly understood. We describe an approach for detecting oligomerization of the yeast α-mating pheromone receptor, Ste2p, a member of the G-protein-coupled receptor family, in the plasma membranes of intact cells. The approach uses flow cytometry to detect fluorescence resonance energy transfer between differentially labeled ligands bound to Ste2p. It was used to show that the extent of oligomerization of native Ste2p is insensitive to receptor concentration over a 50-fold range of expression levels and that truncation of the C-terminal cytoplasmic tails of receptors does not affect oligomerization.

Introduction

G-protein-coupled receptors (GPCRs), the largest class of transmembrane receptors, transduce an extracellular stimulus, such as the binding of a ligand, to an intracellular response, typically the exchange of GTP for GDP bound to a cytoplasmic heterotrimeric G-protein. GPCRs are involved in diverse signaling pathways, including hormonal regulation, neurotransmission, and sensory perception. Although there is no obvious requirement that the basic pathway of GPCR signaling should involve interactions between receptors, there is extensive evidence that the receptors exist as homo- and hetero-oligomers (1, 2, 3, 4, 5). The functional effects of such oligomerization are not generally well understood, except in the case of certain class C family GPCRs that appear to exist as obligate dimers. For one class C receptor, the GABA_B receptor, both correct intracellular targeting and signal transduction appear to require hetero-oligomerization (5, 6, 7, 8). For other classes of GPCRs, monomeric receptors appear to be capable of activating G-protein-coupled responses (9,10), despite evidence that they exist as oligomers in cells, and evidence for positive and negative (11, 12, 13, 14, 15, 16) cooperative effects on ligand binding and signaling responses.

Techniques used to assay receptor oligomerization are often difficult to interpret. Cross-linking may trap complexes that normally exist only transiently. Coimmunoprecipitation may be inefficient or lead to artifactual association or dissociation in detergent micelles. Measurements of energy transfer between fluorescent or bioluminescent proteins covalently fused to receptors are subject to uncertainties stemming from perturbing effects of attaching the fused proteins and from the large size of attached fluorescent proteins. Covalent labeling of receptors with small fluorophores raises problems with labeling specificity or requires use of purified receptors. Bimolecular fluorescence complementation is an irreversible process that may not report the equilibrium distribution of protein oligomers (17). An additional problem with most of these approaches is that they do not discriminate between receptors localized in different intracellular and cell surface locations.

Variations in levels of expression of receptors can have major effects on measurements of receptor oligomerization and there is concern that much of the evidence for receptor oligomerization has been obtained in cells expressing high levels of receptors. There are several ways in which high-level expression could lead to overestimation of receptor oligomerization. Some GPCRs undergo moderate-affinity transient interactions in cell membranes such that mass action could promote receptor oligomerization at high receptor concentrations (18, 19, 20, 21, 22, 23). Overexpression of receptors to greater-than-normal levels in two-dimensional membrane compartments could also drive nonphysiological oligomerization.

Another way that receptor abundance can affect measurements of receptor oligomerization is through random collisions or close encounters that may occur between noninteracting receptors (24, 25, 26). It has long been known that proximity among noninteracting randomly distributed fluorophores present at high concentrations in membranes can lead to energy transfer (27). This problem may be exacerbated by the tendency of some receptors to become sequestered and concentrated in membrane subdomains (28,29). Corrections for energy transfer resulting from random collisions provide the basis for a series of still-controversial protocols for conducting bioluminescence resonance energy transfer (BRET) measurements of receptor oligomerization, in which expression levels and ratios of donors and acceptors are varied to distinguish true oligomerization from proximity effects (24,30, 31, 32, 33, 34, 35).

The α-factor receptor Ste2p, mediating the response to the α-mating pheromone in yeast, is a GPCR that has served as a model for understanding many aspects of G-protein signaling. Although the receptor shares little sequence similarity with mammalian receptors, it activates a heterotrimeric G-protein that exhibits considerable sequence identity with mammalian G-proteins and has been shown to be functionally interchangeable with mammalian GPCRs (36, 37, 38). Evidence for oligomerization of Ste2p receptors in yeast membranes has been obtained from energy transfer between fluorescent and bioluminescent proteins fused to Ste2p (39, 40, 41, 42), from intermolecular cross-linking (43, 44, 45, 46), from bimolecular fluorescence complementation (43), and from coimmunoprecipitation of differentially tagged receptors (47). Solubilized α-factor-bound Ste2p purified from heterologous expression systems also forms dimers (46), exemplified by the recent cryo-electron microscopy structure of dimeric Ste2p reconstituted with a G-protein that had also been expressed as subunits in heterologous systems (48).

In addition to physical evidence for receptor-receptor association, several types of functional interactions between different coexpressed STE2 alleles have been identified. These include dominant negative effects of mutant receptors (39,49, 50, 51), dominant effects of normal receptors over constitutively active mutant receptors (50,52,53), and co-internalization of internalization-incompetent receptor alleles with coexpressed internalization-competent alleles (42,47,54). However, ligand binding to one subunit of an oligomer cannot compensate for a binding defect in the other subunit; thus, a single receptor unit appears to mediate both ligand binding and G-protein activation (55). Furthermore, it has not been possible to detect conformational changes in Ste2p resulting from ligand binding by a separate receptor with which it is co-oligomerized (54).

Previous studies of the oligomeric state of Ste2p in intact yeast cells have all been conducted using yeast strains that overexpress C-terminally truncated receptors fused to fluorescent or bioluminescent proteins, raising the question of whether oligomerization could be a result of overexpression or truncation or fusion. Furthermore, BRET studies of Ste2p fusion proteins in isolated yeast membranes uncovered very different efficiencies of BRET in comparing full-length versus C-terminally truncated receptors (39). An additional problem with previous energy transfer studies of Ste2p oligomerization is that none of them have distinguished between receptors localized at the cell surface and the sizeable fraction of Ste2p that is often located in intracellular compartments (41,54,56,57).

To address the question of how receptor expression levels, C-terminal truncation, and fusion to fluorescent proteins might affect measurements of receptor oligomerization, we have developed procedures for monitoring interactions among full-length unfused yeast α-factor receptors based on detection of fluorescence resonance energy transfer (FRET) between two small organic fluorophores conjugated to α-factor, the normal receptor agonist. Because the fluorescent forms of α-factor are cell impermeant, the approach allows quantitation of energy transfer between receptor-bound ligands that reside specifically at the surface of living cells. FRET efficiencies were determined using flow-cytometry-based measurements of fluorescence emissions as a function of varying ratios of receptor-bound ligands containing donor and acceptor fluorophores. These methods are used to compare interactions between the fluorescent agonists bound to full-length and C-terminally truncated alleles of Ste2p expressed at various levels in cells. The results reveal that the efficiency of energy transfer remains constant as expression levels vary over a 50-fold range extending at its lower end to approximately the normal endogenous expression level. Our approach should be applicable to diverse receptors for which fluorescently tagged ligands are available, allowing detection of oligomerization of unmodified receptors in native cells.

Materials and methods

Strains and plasmids

The host strain for FRET measurements of plasmid-encoded forms of Ste2p was A232 (MATa cry1^R Ade2-1 His4-580 Lys2_oc Trp1_am Tyr1_oc SUP4-3^ts Leu2 Ura3 bar1-1ste2-Δ (49)). Strain A3365 contains a C-terminally truncated STE2 gene on multicopy plasmid pMD1422 (50). Strain A448 contains full-length STE2 on multicopy plasmid pMD240 (49). Stain A453 contains full-length STE2 encoded on CEN plasmid pMD149 (49). Strain A1244 contains a C-terminally truncated allele of STE2 integrated into the genome. It was created by transforming an empty multicopy URA3 vector in strain A530 (50) containing a STE2 gene encoding the C-terminally truncated protein into the chromosome. Yeast transformations were accomplished using the one-step procedure adapted by Chen et al. (58) or by a lithium acetate-polyethylene glycol method (59) performed in the absence of carrier DNA. Saturation biding of [K⁷(BTR),Nle¹²]α-factor ([K⁷(BODIPY-TR),Nle¹²]α-factor) was conducted using strain A3277, containing STE2 on multicopy plasmid pMD1230 (60) transformed into strain A575 (MATa ste2-Δ far1-Δ bar1⁻ cry1^R Ade2-1 His4-580 Lys2_oc Trp1_am Tyr1_oc SUP4-3^ts Leu2 Ura3 FUS1ρp[FUS1-lacZ TRP1]) (52). Strain A454 expressing no Ste2p because of deletion of the chromosomal STE2 gene (60) was used as a negative control.

Fluorescent ligands

[K⁷(NBD),Nle¹²]α-factor and [d-Tyr³K⁷(NBD),Nle¹²]α-factor have been described previously (61, 62, 63). BTR (BODIPY-TR) derivatives were synthesized as described previously for [K⁷(NBD),Nle¹²]α-factor (62), but using the N-hydroxysuccinimide ester of BTR (Invitrogen, Carlsbad, CA) (methyl {p-[4,4-difluoro-5-(2-thienyl)-3a,4a-diaza-4-bora-s-indacen-3-yl]phenoxy}acetate) instead of the 7-nitrobenz-2-oxa-1,3-diazol-4-yl (NBD) fluoride. All fluorescent ligands were more than 98% homogeneous as judged by reversed-phase high pressure liquid chromatography and exhibited the calculated molecular mass.

Preparation of fluorescent ligand solutions

Stock solutions of labeled ligands were prepared as previously described (60). Briefly, each ligand was dissolved in 150 μL of 100% methanol, and then 150 μL of water was added to bring the mixture up to 300 μL final volume. The resulting absorbance at each respective dye’s maximal absorbance (470 nm for NBD; 588 nm for BTR; 625 nm for Bodipy-630/650) was divided by the extinction coefficient (23,000 M⁻¹ cm⁻¹ for NBD; 68,000 M⁻¹ cm⁻¹ for BTR; 101,000 M⁻¹ cm⁻¹ for Bodipy-630/650) to give the stock ligand concentration. Ligands being added simultaneously were premixed before incubation with cells. All ligands were diluted to the appropriate concentrations in a manner to ensure the same volume of ligand was added to every sample analyzed during a single experiment. All ligands were maintained in black microcentrifuge tubes (Argos Technologies, Vernon Hills, IL) because of the light sensitivity of the fluorescent dyes.

Fluorimeter measurements

Bulk fluorescence measurement of fluorescent ligands bound to yeast cells at a density of 5 OD₆₀₀ (optical density at 600 nm) was performed using a Photon Technologies Quantamaster spectrofluorometer (Birmingham, NJ). Each strain was also monitored for autofluorescence by incubating with 0 nM ligand with the same added volume of 50% methanol/50% water. The spectra of cells alone from the same yeast culture were subtracted from each of the spectra of cells incubated from ligand. In each case, the subtracted spectra were adjusted for different baselines by normalizing to the average of the emissions from 690 to 700 nm, at which emissions from bound ligands are expected to be minimal.

Ligand energy transfer using flow cytometry

Flow cytometric measurements of the fluorescence of ligands bound to yeast cells were conducted essentially as described previously (60,64). Approximately 1.5 × 10⁶ yeast cells, grown overnight to an OD₆₀₀ of 1.0, were mixed with fluorescent ligand stocks prepared in a 50% methanol solution in bovineserumalbumin-coated black microcentrifuge tubes and ice-cold 20 mM acetate buffer (pH 4.6) sufficient to bring the final volume to 400 μL. Samples were maintained on ice to minimize endocytosis (60). Fluorescence emission was detected using Becton Dickinson LSRII cytometers (East Rutherford, NJ). Side and forward scatter measurements were collected for each cell, and fluorescence was measured in three channels that we refer to as the NBD, BTR, and transfer channels. The NBD channel was excited at 488 nm with detection using an emission filter centered at 525 nm with a bandwidth of 50 nM. The BTR channel was excited at 633 nm with detection using a filter centered at 660 nm with a 20 nm bandwidth. The transfer channels was excited at 488 nm with detection using an emission filter centered at 630 nm with a 30 nm bandwidth. Fluorescence in each of the three channels was also monitored for the same strains incubated with each ligand alone. The concentration of methanol was maintained constant in all samples. After gating out ∼10% of events that were outliers in dot plots of forward versus side scatter, the mean fluorescence emission value for 10,000 cells was recorded in each of the three channels.

Equilibrium binding assays

Assays used to determine B_max- and K_D-values for each fluorescently labeled α-factor were performed using flow cytometry to measure cell-associated fluorescence essentially as previously described (60,64,65). Individual least-squares fits to data from replicate binding experiments were performed based on single-site binding isotherms with a nonspecific component. Because of the low total concentrations of receptors in cellular samples, no correction was applied for ligand depletion (60). Errors are displayed as standard errors of the mean for the replicate data sets.

Results

General considerations for measurement of FRET between receptor-bound ligands

In addition to the parameters that determine FRET efficiency in general, measurements of energy transfer efficiency between differentially labeled ligands are subject to the following factors:

• 1)The number of receptors per oligomer. This number has not been well established for Ste2p. Most of the previous studies do not provide discrimination between dimers and higher-order oligomers, although there have been reports of Ste2p adopting both dimeric (46,48) and higher-order (46,57) states. Although the data presented below can be fitted by assuming that Ste2p is either dimeric or in higher-order oligomeric states, in the absence of additional constraints, receptors were assumed to exist as dimers. Implications of this assumption with respect to the available structure of dimeric Ste2p (48) are discussed further below.
• 2)Varying receptor occupancy by different ligands. The occupancy of each labeled ligand bound to receptors, a critical parameter in calculating FRET efficiencies, can be calculated based on free ligand concentrations and the dissociation constant for each ligand, according to the formalism of the Gaddum equation (66). This equation describes the fractional occupancy, f_A, of binding sites by ligand A, in the presence of a second ligand, ligand B, as

$${\text{f}}_{\text{A}}=\frac{\left[\text{A}\right]}{\left[\text{A}\right]+{\text{K}}_{\text{A}}\left(1+\left[\text{B}\right]/{\text{K}}_{\text{B}}\right)},$$ (1)

• where [A] and [B] are the unbound concentrations of the two ligands and K_A and K_B are their dissociation constants.

• 3)Possible cooperativity between receptors in oligomers. Despite the evidence for cooperativity among some co-oligomerized GPCRs (11, 12, 13, 14, 15, 16) and for functional interactions among Ste2p receptors (39,49, 50, 51), the binding of α-factor to Ste2p does not exhibit any obvious cooperative effects (64,67). Nonetheless, to avoid the need to consider cooperativity, energy transfer measurements were performed at high overall ligand concentrations so that the relative overall concentrations of the donor and acceptor ligands are directly reflected in the occupancies of receptors by each ligand, obviating the need to consider possible cooperative interactions that could occur at lower overall receptor occupancies.
• 4)Background emissions from cellular autofluorescence, light scattering, nonspecifically bound fluorescent ligand, and unbound ligand in solution. We describe below procedures for correcting for these sources of fluorescence.

The binding of several fluorescent α-factor analogs to yeast cells has previously been characterized. The introduction of certain modifications of α-factor at position K⁷, in particular, has minimal effects on binding affinity or signaling efficacy (60, 61, 62, 63). When bound to Ste2p, the α-factor derivative [K⁷(NBD),Nle¹²]α-factor provides a particularly robust fluorescent signal, apparently because of sequestration of the NBD fluorophore in a binding site environment that is more hydrophobic than the environment experienced by the fluorophore in aqueous solution. To allow the detection of energy transfer between labeled ligands, we synthesized an α-factor analog with excitation and emission maxima that are red-shifted compared with NBD. This analog, [K⁷(BTR),Nle¹²]α-factor, contains a BODIPY-TR fluorophore at position K⁷ (Fig. 1). In saturation binding experiments assayed by flow cytometry as described previously (60,64,65), [K⁷(BTR),Nle¹²]α-factor bound to cells with an affinity (K_D = 8 ± 2 nM, n = 6; Fig. S1) that is similar to that of [K⁷(NBD),Nle¹²]α-factor (60), although associated with higher nonspecific binding than that seen for [K⁷(NBD),Nle¹²]α-factor (Fig. S1). This increase in the ratio of nonspecific/specific binding most likely results from a reduced emission of specifically bound BTR-labeled α-factor compared to the strongly enhanced emission of specifically bound NBD-labeled ligand (60, 61, 62, 63). We also investigated the properties of an α-factor derivative labeled at K⁷ with the fluorophore BODIPY-630/650. However, this was found to bind to Ste2p with much lower affinity than [K⁷(NBD),Nle¹²]α-factor or [K⁷(BTR),Nle¹²]α-factor and was not pursued further.

Figure 1.

Schematic chemical structures of the fluorescent α-factor analogs (a) [K⁷(NBD),Nle¹²]α-factor and (b) [K⁷(BTR),Nle¹²]α-factor. Nle was substituted as a stable replacement for methionine at position 12 of each peptide (61,62).

FRET between ligands measured by fluorimetry of cell suspensions

Energy transfer between [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor could be detected in cell suspensions using a spectrofluorometer to detect enhanced acceptor emission. To maximize the fluorescent signal, measurements were performed using cells expressing C-terminally truncated α-factor receptors from a multicopy plasmid. Removal of the C-terminal cytoplasmic tail of the receptor increases the number of receptors at the cell surface, apparently by inhibiting endocytosis. As shown in Fig. S2, a major contribution to the fluorescence emission came from cellular autofluorescence and light scattering. To minimize the background, all emission spectra were first normalized to those of cells alone (from the same culture) based on the emission in the range 690–700 nm where emission from the bound ligands is minimal. The emission spectrum of cells alone was then subtracted from the spectra of ligand-bound samples. As shown in Fig. S3, for cells that do not express receptors, this correction procedure resulted in relatively flat baselines over much of the emission range, though with substantial deviations in the range 500–550 nm. When excited at 470 nm, the corrected spectra of cells expressing C-terminally truncated receptors that had been incubated with [K⁷(NBD),Nle¹²]α-factor, exhibited an emission peak at 510 nm from bound NBD, whereas cells incubated with [K⁷(BTR),Nle¹²]α-factor alone, as expected, exhibited little emission at any wavelength (Fig. 2 a). However, when excited at 588 nm, cells incubated with [K⁷(BTR),Nle¹²]α-factor alone exhibited strong emission at 625 nm characteristic of the emission of BTR, apparently because of enhancement of emission from [K⁷(BTR),Nle¹²]α-factor upon binding to cells (Fig. 2 b).

Figure 2.

Spectrofluorometer measurements of emission of [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor bound to yeast cells of the strain (A3365) expressing C-terminally truncated Ste2p receptors from a multicopy plasmid and strain A454 containing a deletion of the STE2 gene. [K⁷(NBD),Nle¹²] α-factor was present, where indicated, at a concentration of 20 nM. [K⁷(BTR),Nle¹²] α-factor was present, where indicated, at a concentration of 80 nM. Traces were normalized to have the same average difference over the range 691–700 nm compared to cells in the absence of ligand. (a) Samples excited at 470 nm. (b) Samples excited at 588 nm. Correction was applied as described in Materials and methods. Note that emissions from the samples excited at 588 nM were multiplied by a factor of 0.2 compared to samples excited at 470 nm.

470 nm excitation of receptor-expressing cells incubated with both 20 nM [K⁷(NBD),Nle¹²]α-factor and 80 nM [K⁷(BTR),Nle¹²]α-factor resulted in a significant emission peak at 625 nm (characteristic of the emission of BTR) that is apparent in both the corrected and uncorrected spectra shown in Figs. 2 a and S2 a, respectively. This must be the result of energy transfer, as cell-bound donor [K⁷(NBD),Nle¹²]α-factor exhibits little emission at 625 nm and cell-bound acceptor [K⁷(BTR),Nle¹²]α-factor is only minimally excited at 470 nm. This result is consistent with previous demonstrations of energy transfer among overexpressed C-terminally truncated Ste2p receptors on intact cells (39, 40, 41, 42,57,68).

Receptor-expressing cells incubated with both donor and acceptor ligands also exhibited decreased emission at the donor emission maximum of 510 nm compared with cells incubated with 20 nM [K⁷(NBD),Nle¹²]α-factor alone. Most of this decrease was due to competition, as occupancy by donor is expected to decrease from 67 to 15% in comparing cells incubated with 20 nM donor alone to cells incubated with 20 nM donor plus 80 nM acceptor. The contribution of energy transfer to reduced donor emission is likely to be small, as only a low percentage of receptors (12%) would be expected to exist as components of dimers in which one component is bound to donor and the other component is bound to acceptor. Attempting to maximize the decrease in donor emission due to energy transfer leads to a conundrum: increasing the donor/acceptor ratio to increase receptor occupancy by donor will decrease the fraction of donor-bound receptor dimers that also contain a bound acceptor and decrease the proportional contribution of energy transfer to overall donor emission. On the other hand, decreasing the donor/acceptor ratio to increase the proportion of donor-bound dimers that also contain acceptors will decrease overall donor emission compared to the substantial background.

When excited at 588 nm, cells incubated with both donor and acceptor exhibited decreased emission at 625 nM compared with cells incubated with acceptor alone (Fig. 2 b). Because the presence or absence of donor should not affect directly excited acceptor, this reduction in acceptor emission is due to competition from the donor [K⁷(NBD),Nle¹²]α-factor that reduces acceptor binding from the expected 91% in the presence of 80 nM acceptor alone to 77% when 20 nM donor is also present.

In principle, it could be possible to use a (ratio)_A approach (69,70) for analysis of energy transfer based on the increase in acceptor emission detected by fluorimetry of cell suspensions. This is an analytic approach based on the comparison of the emission from a single sample excited, separately, at wavelengths corresponding to the characteristic excitation maxima of the donor and that of the acceptor. However, we encountered two significant barriers to applying this approach to fluorimetry-based measurements of cell suspensions: 1) the determination of actual energy transfer efficiencies depends directly on the ratio of the extinction coefficients of the donor and acceptor at their respective characteristic excitation wavelengths. In this case, the relevant extinction coefficients would have to be those for the receptor-bound fluorophores, as the physical environments and optical properties of the fluorophores bound to Ste2p appear to be very different from those of the same compounds in solution (60, 61, 62, 63). The extremely low concentrations of receptors in cell suspensions with high levels of light scattering preclude direct measurement of the absorbances of receptor-bound ligands. 2) Even for samples with the highest level of Ste2p expression, high backgrounds from cellular autofluorescence and light scattering of cell suspensions, coupled with the difficulty of comparing differently treated suspensions prevented reliable quantitation of fluorescence emission from the suspensions. Thus, although measurements of cell suspensions in the spectrofluorometer provided clear evidence for energy transfer between [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor bound to Ste2p, reliable quantitation of energy transfer efficiencies could not be achieved from these experiments and would be much more difficult for low endogenous levels of receptor expression.

FRET between ligands detected by flow cytometry

Energy transfer between ligands could also be readily detected by flow cytometry using different available fluorescent channels, as shown in Figs. 3 and 4 and Table 1. Compared with measurements in a spectrofluorometer, flow cytometry avoids contributions due to light scattering, nearly eliminates the contribution of unbound fluorescent ligands in solution, and can detect differences in fluorescent properties of different populations of cells in a sample. Initial experiments used cells that overexpress C-terminally truncated Ste2p from a multicopy plasmid. As shown in Fig. 3, all cells showed a basal level of autofluorescence in the absence of any labeled ligand; thus, the mean value of the autofluorescence in different channels was subtracted from all samples, based on comparisons of cells from the same culture with and without ligands. Compared to cells without ligand, cells overexpressing truncated Ste2p with bound [K⁷(NBD),Nle¹²]α-factor alone exhibited substantial fluorescence emission at 525 nm when excited at 488 nm (NBD channel), very low levels of emission at 660 nM when excited at 633 nm (BTR channel), and a low level of increased emission at 630 nm when excited at 488 nm (transfer channel) apparently due to direct NBD emission at the longer wavelengths. Binding of [K⁷(BTR),Nle¹²]α-factor by itself to these cells resulted in no detectable emission in the NBD channel, substantial emission in the BTR channel, and some emission in the transfer channel, presumably due to limited direct excitation of the BTR fluorophore at 488 nm. A population of cells exhibiting fluorescence levels equivalent to autofluorescence is visible as a broad shoulder on the left side of the histogram peaks in each channel for samples incubated with labeled ligands. This is attributable to cells that spontaneously lose the Ste2p-encoding plasmid, despite growth on selective media (71).

Figure 3.

Flow cytometry histograms of [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor bound to cells overexpressing C-terminally truncated α-factor receptors (strain A3655) detected in (a) the NBD channel, (b) the BTR channel, and (c) the transfer channel. For the sample containing both donor and acceptor ligands, the receptor occupancies of the two ligands indicated in the legend were determined as described in the text based on measurements in the BTR channel relative to control samples of the same cells expressing 100% donor or 100% acceptor. To see this figure in color, go online.

Figure 4.

Two-dimensional plots of fluorescence emission of [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor bound to cells overexpressing C-terminally truncated α-factor receptors (strain A3655). (a) Dot plot of emission in the transfer channel versus the NBD channel. (b) Dot plot of emission in the transfer channel versus the BTR channel. Samples are the same as shown in Fig. 3. Each dot in the plots represents the data from one cell.

Table 1.

Measurement of energy transfer efficiencies by flow cytometry

	Relative Ste2p surface expressiona	Energy transfer efficiencyb	Nonspecific [K7(BTR),Nle12]α-factor binding contributionc
Ste2p-Δ305–431, multicopy (strain A3365)	1, n = 12	0.59 ± 0.05, n = 15	0.22 ± 0.02, n = 6
Full-length Ste2p, multicopy (strain A448)	0.13 ± 0.03, n = 12	0.52 ± 0.04, n = 12	0.21 ± 0.02, n = 6
Ste2p-Δ305–431, chromosomal (strain A1244)	0.12 ± 0.01, n = 6	0.60 ± 0.04, n = 3	0.31 ± 0.03, n = 3
Full-length Ste2p, CEN plasmid) (strain A453)	0.021 ± 0.002, n = 12	0.61 ± 0.05, n = 9	0.52 ± 0.03, n = 3
Ste2p-Δ305–431, multicopy (strain A3365) (d-Tyr3 ligands)	1d	0.83 ± 0.07, n = 3	0.14 ± 0.06, n = 3e

^aRelative surface expression determined based on emission from donor-only samples under conditions in which the donor fraction is 1. Errors are presented as standard error of the mean.

^bEnergy transfer efficiency for strains expressing C-terminally truncated receptors expressed from a multicopy plasmid is determined from loss of donor fluorescence. Transfer efficiencies for all other strains are based on enhanced emission in the transfer channel. Errors are presented as standard error of the mean.

^cNonspecific binding contribution is the fraction of total emission [K⁷(BTR),Nle¹²]α-factor-only samples at a concentration of 300 nM derived from fitting of saturation binding for each strain, detected in the transfer channel. Errors are presented as standard error of the mean.

^dBased on emission from normal [K⁷(NBD),Nle¹²]α-factor.

^eFor d-Tyr³ samples, a nonspecific correction equal to 14% of the donor-only emission (Df = 1) in the NBD channel was applied for the donor ligand [d-Tyr³K⁷(NBD),Nle¹²]α-factor.

When both [K⁷(NBD),Nle¹²] and [K⁷(BTR),Nle¹²]α-factor were incubated with the cells overexpressing truncated receptors, the fluorescence in the NBD and BTR channels decreased (Fig. 3, a and b), whereas fluorescence in the transfer channel increased (Fig. 3 c). The decreased emission in the NBD and BTR channels is due to competition for binding sites by the opposite ligand and, in the case of the NBD channel, potentially due to energy transfer from NBD to BTR. The levels of emission in the transfer channel are more than fivefold greater than the emission seen for cells incubated with [K⁷(BTR),Nle¹²]α-factor alone and more than 10-fold greater than the level for [K⁷(NBD),Nle¹²]α-factor alone, despite lower amounts of each ligand bound when both are present. This provides strong evidence for energy transfer, again consistent with previous whole-cell studies of overexpressed C-terminally truncated receptors fused to fluorescent or bioluminescent proteins (39, 40, 41, 42,57,68).

The same trends are apparent when the data are expressed as dot plots showing cell populations in Fig. 4. Cells with the highest levels of emission in the transfer channel exhibit the highest levels of emission in the NBD and BTR channels, indicating that these are the cells in the population that express the highest number of receptors. The dot plots also show that the increases in emission due to energy transfer affect the entire population of cells, shifting to higher positions on the y axes of Fig. 4, a and b and providing greater signals in this channel than can be observed for the signals from either individual ligand. As noted for the histogram displays, cells with both ligands bound exhibit higher fluorescence in the transfer channel than cells with either singly bound ligand.

The (ratio)_A method cannot be used to determine the energy transfer efficiency from flow cytometry because of the unknown differences in excitation intensity and emission sensitivity in the different fluorescent channels and the lack of knowledge of the relative extinction coefficients of the donor and acceptor. An additional complication in this system, compared with the covalently labeled double-stranded DNAs used in the (ratio)_A method of Clegg, is the variation in the extent of labeling with the two fluorophores, including the possibility that receptor dimers may bind two donor or two acceptor ligands.

For cells with high fluorescence emission due to expression of high levels of C-terminally truncated receptors encoded on multicopy plasmids, we used a curve fitting approach to measure the efficiency of energy transfer from the loss of donor emission at various ratios of donor to acceptor ligand. To avoid considering any possible effects of cooperativity of ligand binding on receptor occupancy and to minimize the contributions of unoccupied receptors, flow cytometric fluorescence measurements were conducted at high total concentrations of the two ligands, varying the fractional occupancy of each by varying the donor/acceptor ratio. Thus, we used maximal total ligand concentrations ∼15 and 30 times greater than the K_D-values of [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor, respectively, which should maintain receptors more than 93% occupied by one or the other ligand. The low remaining concentration of unoccupied receptors may lead to a slight systematic underestimation of energy transfer efficiencies, as a small fraction of ligands will not participate in energy transfer because they are bound to receptor dimers in which one ligand binding site is empty. However, because the same ligand concentrations were used to generate various donor fractions in all experiments described below, such low levels of incomplete occupancy should not affect relative measures of energy transfer efficiency.

Emission in the NBD channel for the flow cytometer can be expressed as

$${\text{F}}_{\text{D}}={\text{A}}_0+{\text{F}}_{\text{DD}}+{\text{F}}_{\text{DA}}+{\text{F}}_{\text{A}}+{\text{F}}_{\text{NsDD}}+{\text{F}}_{\text{NsDA}},$$ (2)

where F_D is the total emission in the donor channel, A₀ is the cellular autofluorescence in the NBD channel, and F_DD is the emission of donor ligand in the NBD channel for ligands that are specifically bound to dimers that contain two donor ligands. F_DA is the emission in the NBD channel of donor ligand bound in dimers that contain one donor and one acceptor. In the absence of energy transfer, F_DA would be the emission from donor bound to dimers that contain an acceptor ligand in the second site. In the presence of energy transfer with efficiency E, the remaining donor emission is proportional to (1 − E). F_A is the small emission in the donor channel from acceptor ligands bound to receptors, and F_NsDD and F_NsDA are the emissions in the donor channels due to nonspecifically bound donor and acceptor ligands, respectively (which may also include a small contribution from unbound ligand that is found within the illumination volume of the flow cytometer).

Of these terms, the autofluorescence, A₀, can be readily determined from flow cytometry of cells that are not incubated with any ligand and thus can be subtracted from every sample. F_DD for a particular sample can be extrapolated from a sample of cells incubated with donor alone, corrected for a nonspecific binding contribution and the donor fraction (Df) of particular samples. As can be seen in Figs. 2, 3, and 4, the term F_A, the acceptor emission in the donor channel is a small contribution that can be extrapolated from a sample containing solely acceptor ligand (Df = 0), corrected for the fraction of receptors occupied by acceptor ligand. Thus, for these cells, after correcting for autofluorescence and tabulating in terms of bound donor ligands, F_D can be written as

$${\text{F}}_{\text{D}}=\text{ k}1\left[{\text{R}}_{\text{T}}\right]\left({\left(\text{Df}\right)}^2+\left(1-\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)\right)+\text{ k}2\left[{\text{R}}_{\text{T}}\right]\left(1-\text{Df}\right)+{\text{N}}_{\text{sDD}}\left[[{\text{K}}^7\left(\text{NBD}\right),{\text{Nle}}^{12}]\alpha \text{-}\text{factor}\right]+{\text{N}}_{\text{sDA}}\left[\left[{\text{K}}^7\left(\text{BTR}\right),{\text{Nle}}^{12}\right]\alpha \text{-}\text{factor}\right].$$ (3)

Here, R_T is the total number of receptors per cell. Df is the fractional occupancy of receptors by donor so that the number of receptors per cell that are in dimers containing two donors is R_T(Df)² and the number in dimers containing one donor and one acceptor is R_T(Df)(1 − Df). k1 is a parameter that relates the emission in the donor channel to the number of bound donors per cell. k2 relates the emission in the donor channel to the number of bound acceptor ligands per cell. These parameters can be measured from conditions in which Df = 1 or Df = 0 so that only donor or only acceptor is bound to cells. N_sDD and N_sDA are proportionality constants that relate the amount of nonspecifically bound donor and acceptor ligands to the free concentrations of the ligands. (Note that these, and additional expressions describing emission from bound ligands, are expressed in terms of individual receptors bound to ligands rather than in terms of receptor dimers.)

When total ligand concentrations are maintained with different ratios of donor/acceptor, the concentrations of the donor and acceptor ligands are proportional to Df or (1 − Df), respectively, and with proportionality constants k3 replacing N_sD and k4 replacing N_sA. These contributions of nonspecific binding can be estimated from saturation binding experiments so that

$${\text{F}}_{\text{D}}=\text{ k}1\left[{\text{R}}_{\text{T}}\right]\left({\left(\text{Df}\right)}^2+\left(1-\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)\right)+\text{ k}2\left[{\text{R}}_{\text{T}}\right]\left(1-\text{Df}\right)+\text{ k}3\left(\text{Df}\right)+\text{ k}4\left(1-\text{Df}\right).$$ (4)

For any set of conditions, the constants k2 and k4 can be combined into k5 = k2(R_T) + k4 so that

$${\text{F}}_{\text{D}}=\text{ k}1\left[{\text{R}}_{\text{T}}\right]\left({\left(\text{Df}\right)}^2+\left(1-\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)\right)+\text{ k}5\left(1-\text{Df}\right)+\text{ k}3\left(\text{Df}\right).$$ (5)

Thus, when Df = 1, F_{D =} k1[R_T] + k3.

When DF = 0, F_D = k5.

For cells overexpressing high levels of truncated receptors, the nonspecific binding component for [K⁷(NBD),Nle¹²]α-factor is less than 2% at the levels of ligand used in these experiments (60). Thus, for these cells, we disregard k3.

Because of the critical dependence of F_D on the fractional occupancies of donor and acceptor ligands, it was important to have accurate estimations of these occupancies, which can be affected by inaccuracies in determining the exact concentrations of ligand stocks, adsorption of the low concentrations of ligands to tube surfaces, and uncertainties in determining the exact binding affinities of the two ligands. Thus, rather than relying on estimation of bound ligands based on the concentrations of added stocks, we directly measured the acceptor fractions and used these to calculate donor fractions (because the two fractions should sum to a value of ∼1 at the supersaturating concentrations of ligand used in these experiments). Fractional acceptor occupancy was determined based on levels of directly excited [K⁷(BTR),Nle¹²]α-factor emission for each sample containing different concentrations of the two ligands, which should be proportional to acceptor occupancy plus a contribution from nonspecific binding of acceptor, both of which are proportional to the acceptor ligand concentration. As can be seen in Figs. 2, 3, and 4, emission from [K⁷(NBD),Nle¹²]α-factor in the BTR channels is negligible. Thus, F_A, the emission in the acceptor channels is

$${\text{F}}_{\text{A}}=\text{ k}6\left[{\text{R}}_{\text{T}}\right]\left(1-\text{Df}\right)+{\text{N}}_{\text{sAA}}\left[\left[{\text{K}}^7\left(\text{BTR}\right),{\text{Nle}}^{12}\right]\alpha \text{-}\text{factor}\right].$$ (6)

The first term of this expression is the emission in the BTR channel from specifically bound [K⁷(BTR),Nle¹²]α-factor in which k6 relates the detected emission in this channel to the concentration of bound [K⁷(BTR),Nle¹²]α-factor. The second term of the expression is the contribution to the emission resulting from nonspecifically bound [K⁷(BTR),Nle¹²]α-factor. Because both of these terms are proportional to (1 − Df) under these experimental conditions in which only donor fraction is varied, the first two terms of this expression can be folded into one constant k7, which can be readily determined from samples containing 100% acceptor and no donor. Thus, F_A can be expressed as

$${\text{F}}_{\text{A}}=\text{ k}7\left(1-\text{Df}\right).$$ (7)

A typical plot of the donor fraction based on the above expression versus the calculated donor fraction is shown in Fig. S4. The most reproducible results were obtained when donor fractions were determined individually for each sample of ligand-treated yeast culture being assayed, consistent with the possibility that any given culture of cells contains a variety of sizes and shapes that could affect flow cytometric measurements and could have different variations in ligand concentration due to adsorption. Attempts to reduce variation by using the slope of a regression analysis of cell fluorescence against forward scattering did not provide any detectable reduction in variation (data not shown).

We used least-squares fitting of F_D as a function of a series of measured donor fractions for that culture to determine values of E for each culture based on loss of donor emission. A typical set of fits, along with expected curves for different values of E is shown in Fig. 5. We obtained a mean value of 59 ± 5% for E (n = 15; five different experiments each using three independent yeast transformants). This indicates that the spacing between fluorophores is less than the Förster radius for receptor-bound fluorescent donor-acceptor pair; however, uncertainties in underlying optical parameters prevent us from reliably determining the applicable Förster radius.

Figure 5.

Plots of fluorescent emission in the NBD channel as a function of measured donor occupancy. Black, red, and blue solid circles show measured emission for three independent isolates of strain A3365, expressing C-terminally truncated Ste2p from a multicopy plasmid, measured in parallel. Solid, black, red, and blue lines show least-squares fits to data for the three isolates, yielding energy efficiency (E) values of 0.40, 0.58, and 0.66, respectively. The black dotted line indicates the predicted trace for E = 0. The dashed black line shows the predicted trace for E = 1. To see this figure in color, go online.

Dependence of FRET on levels of receptor expression and presence of the receptor C-terminal tail

Previous studies of FRET using fluorescent proteins fused to α-factor receptors were only capable of detecting interactions between C-terminally truncated receptors, which are present at increased abundances at the cell surface compared with normal receptors (42). Subsequently, BRET between luciferase and green fluorescent protein fused to α-factor receptors in isolated cell membrane fractions provided the sensitivity required to detect co-oligomerization of full-length receptors (39). In these experiments, energy transfer efficiencies for interactions between full-length receptors were found to be much lower than for truncated receptors. Although this could be due to spacing and/or flexibility of the fluorescent and luminescent proteins fused at the end of the 130 amino acid tail, compared with fusions placing these reporters directly at the membrane surface (39), these results raised the possibility that receptor truncation could affect oligomerization. They also left open the question of the state of oligomerization of receptor in intact cells.

We used FRET between receptor-bound ligands to provide direct comparisons of the oligomerization states of truncated and full-length receptors expressed at different levels at the surfaces of intact yeast cells. However, we found that measurements based on the loss of donor emission, such as those described above, provide insufficient signal/noise ratio for reliable determinations of energy transfer efficiency for cells expressing full-length receptors at lower expression levels. The low signal/noise ratio stems, at least in part, from the fact that such measurements rely on detection of modest decreases in donor emissions compared to the large and variable signal observed in the absence of transfer. In contrast, detection of FRET based on enhanced acceptor emission provides more reliable quantitation, as the enhanced emission is compared with a low background seen in the absence of transfer (see Figs. 2, 3, and 4). However, determination of energy transfer efficiencies based on enhanced emission in the transfer channel requires either use of the (ratio)_A approach, which, as described above, is not applicable for flow cytometry measurements, or detailed knowledge of optical properties of the receptor-bound fluorophores, which are difficult to measure directly. We circumvented this problem by using energy transfer efficiencies from loss of donor emission in cells expressing high levels of receptors to calibrate the missing parameters relating emission in the transfer channel to energy transfer efficiency. This allows the determination of energy transfer efficiencies in cells expressing low levels of receptors. For analyzing such measurements, we express emission in the transfer channel as F_T, based on the formalism described by Clegg (69,70):

$${\text{F}}_{\text{T}}={\text{F}}_{\text{DD}}+{\text{F}}_{\text{DA}}+{\text{F}}_{\text{A}}+{\text{Ns}}_{\text{TD}}+{\text{Ns}}_{\text{TA}},$$ (8)

where F_DD is bleedthrough of direct donor emission into the transfer channel from dimers containing donors only and from the residual nontransferred emission of dimers containing donor and acceptor, F_DA is the emission in the transfer channel from energy transfer occurring in dimers containing both specifically bound donor and acceptor, F_A is emission in the transfer channel due to direct excitation of acceptor, and N_sTD and N_sTA are the emissions of nonspecifically bound donor and acceptor ligands in the transfer channel. This can be expressed as

$${\text{F}}_{\text{T}}=\text{ k}8\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{D}}\left(\left(\text{Df}\right)\left(1-\text{Df}\right)\left(1-\text{E}\right)+{\left(\text{Df}\right)}^2\right)+\text{ k}9\left(\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{A}}\left(\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)+\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{A}}{\phi }^{\text{A}}\left(1-\text{Df}\right)\right)+{\text{Ns}}_{\text{TD}}\left[\left[{\text{K}}^7\left(\text{NBD}\right),{\text{Nle}}^{12}\right]\alpha \text{-}\text{factor}\right]+{\text{Ns}}_{\text{TA}}\left[\left[{\text{K}}^7\left(\text{BTR}\right),{\text{Nle}}^{12}\right]\alpha \text{-}\text{factor}\right],$$ (9)

where k8 is a proportionality constant relating the detected signal in the transfer channel to fluorescence emission in the transfer channel due to bleedthrough of donor emission from donor and the number of receptors per cell. k9 is a proportionality constant relating the detected signal in the transfer channel to fluorescence emission from acceptor (due to energy transfer and direct excitation at 488 nm) and the number of receptors per cell. ε^D and ε^A are the extinction coefficients of the donor and acceptor, respectively, at the excitation wavelength of the transfer channel (488 nm). φ^D and φ^A are the quantum yields of the donor and acceptor, respectively, and Ns_TA and Ns_TD are proportionality constants relating the emissions in the transfer channel from nonspecifically bound donor and acceptor, respectively, to the total concentrations of each ligand.

When Df = 0, F_T = F_TDf0, [[K⁷(NBD),Nle¹²]α-factor] = 0, and [[K⁷(BTR),Nle¹²]α-factor] = 300 nM, thus

$${\text{F}}_{\text{T}}={\text{F}}_{\text{TDf}0}=\text{ k}9\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{A}}{\phi }^{\text{A}}+{\text{Ns}}_{\text{TA}}\left(300\text{nM}\right).$$ (10)

Define k10 = Ns_TA(300 nM), which can be measured from saturation binding experiments for each strain, so that

$$\text{k}9=\left({\text{F}}_{\text{TDf}0}–\text{ k}10\right)/\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{A}}{\phi }^{\text{A}}.$$ (11)

When Df = 1, F_T = F_TDf1, [[K⁷(BTR),Nle¹²]α-factor] = 0, and [[K⁷(NBD),Nle¹²]α-factor] = 150 nM, so that

$${\text{F}}_{\text{T}}={\text{F}}_{\text{TDf}1}=\text{k}8\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{D}}+{\text{Ns}}_{\text{TD}}\left(150\text{nM}\right).$$ (12)

Define k11 = Ns_TD(150nM) so that

$$\text{k}8=\left({\text{F}}_{\text{TDf}1}-\text{ k}11\right)/\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{D}}.$$ (13)

Thus,

$$\begin{array}{l}{\text{F}}_{\text{T}}=\left(\left({\text{F}}_{\text{TDf}1}-\text{k}11\right)/\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{D}}\right)\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{D}}\left(\left(\text{Df}\right)\left(1-\text{Df}\right)\left(1-\text{E}\right)+{\left(\text{Df}\right)}^2\right)+\\ (\left({\text{F}}_{\text{TDf}0}–\text{ k}10\right)/\left(\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{A}}{\phi }^{\text{A}}\right)(\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{D}}{\phi }^{\text{A}}\left(\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)+\left[{\text{R}}_{\text{T}}\right]{\epsilon }^{\text{A}}{\phi }^{\text{A}}\left(1-\text{Df}\right))+\text{ k}11\left(\text{Df}\right)+\text{ k}10\left(1-\text{Df}\right),\end{array}$$ (14)

$$\begin{array}{l}{\text{F}}_{\text{T}}=\left({\text{F}}_{\text{TDf}1}-\text{k}11\right)\left(\left(\text{Df}\right)\left(1-\text{Df}\right)\left(1-\text{E}\right)+{\left(\text{Df}\right)}^2\right)+\left({\text{F}}_{\text{TDf}0}–\text{ k}10\right)\left({\epsilon }^{\text{D}}/{\epsilon }^{\text{A}}\right)\left(\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)+\\ \left({\text{F}}_{\text{TDf}0}–\text{ k}10\right)\left(1-\text{Df}\right)+\text{ k}11\left(\text{Df}\right)+\text{ k}10\left(1-\text{Df}\right),\end{array}$$ (15)

and

$${\text{F}}_{\text{T}}=\left({\text{F}}_{\text{TDf}1}-\text{k}11\right)\left(\left(\text{Df}\right)\left(1-\text{Df}\right)\left(1-\text{E}\right)+{\left(\text{Df}\right)}^2\right)+\left({\text{F}}_{\text{TDf}0}–\text{ k}10\right)\left({\epsilon }^{\text{D}}/{\epsilon }^{\text{A}}\right)\left(\text{E}\right)\left(\text{Df}\right)\left(1-\text{Df}\right)+{\text{F}}_{\text{TDf}0}\left(1-\text{Df}\right)+\text{ k}11\left(\text{Df}\right).$$ (16)

Because of the low nonspecific binding component of the donor [K⁷(NBD),Nle¹²]α-factor (64) and low emission of [K⁷(NBD),Nle¹²]α-factor in the transfer channel, k11, the nonspecific contribution of the donor [K⁷(NBD),Nle¹²]α-factor to emission in the transfer channel is negligible at all but the lowest expression levels of receptor in cells where Ste2p is expressed from a CEN plasmid. For this particular strain, nonspecific binding was measured from saturation binding experiments as described previously (60,64).

Although we do not know a priori the ratio of the extinction coefficients, ε^D/ε^A, for fluorophores bound to receptors, the above determination of the energy transfer from loss of donor emission for cells expressing high levels of receptors was used to derive a ε^D/ε^A ratio of 28.4 ± 1.0 (n = 15) by fitting emission in the transfer channel as a function of donor fraction for the high-expressing strain A3365. Such a high ratio reflects the fact that the exciting wavelength (488 nm) is near the absorption maximum for NBD but well below the maximum for BTR. It also most likely reflects an enhancement of the absorption of the environmentally sensitive NBD fluorophore when [K⁷(NBD),Nle¹²]α-factor is bound to receptors, as a lower ratio was detected for NBD- and BTR-labeled α-factor antagonists that display fluorescent properties quite different from those of the labeled agonist (see below). A representative fit to fluorescence measurements in the transfer channel, along with predicted traces for different values of the ratio of extinction coefficients, is shown in Fig. 6.

Figure 6.

Plots of fluorescent emission in the NBD channel as a function of measured donor occupancy. Black, red, and blue solid symbols show measured emission for three independent isolates of strain A3365, expressing C-terminally truncated Ste2p from a multicopy plasmid, measured in parallel. Solid, black, red, and blue lines show least-squares fits to data for the three isolates, yielding ε^D/ε^A ratios of 26.5, 29.8, and 27.2, respectively. The black dotted line indicates the predicted trace for a ratio of 25% greater than the overall mean of 28. The dashed black line shows the predicted trace for a ratio of 25% less than the overall mean (for replicate shown in red). The predicted traces are shown for parameters derived from replicate 3 (black circles). To see this figure in color, go online.

The above formalism allows calculation of energy transfer efficiencies based solely on measurements in three separate fluorescent channels made at different donor/acceptor ratios accompanied by determinations of levels of nonspecific binding determined in separate saturation binding experiments. The measurements do not require any additional correction for levels of receptor expression. This approach was used to determine energy transfer efficiencies between ligands at receptor expression levels that varied over a 50-fold range, from nearly the endogenous level provided by STE2 expressed from a centromere-based plasmid to the high levels of surface expression of Ste2p resulting from expression of C-terminally truncated receptors from multicopy plasmids (Table 1). Representative sets of fits to measurements made on cells expressing intermediate levels of receptors (as full-length Ste2p expressed from a multicopy plasmid) are shown in Fig. 7. Fits to measurements made on cells expressing low levels of full-length Ste2p from a CEN plasmid are shown in Fig. 8. In each case, predicted fits for different energy transfer efficiencies are shown for reference. Although the reliability of such measurements is greater at high expression levels, no statistically significant difference in the energy transfer efficiency between the two labeled α-factor derivatives was detected for strains with any of the different levels of surface expression (Table 1; compare Figs. 7 and 8). Furthermore, no difference in energy transfer efficiency was detected in comparing C-terminally truncated to full-length receptors.

Figure 7.

Plots of fluorescent emission in the transfer channel as a function of measured donor occupancy. Black, red, and blue solid symbols show measured emission for three independent isolates of strain A448, expressing full-length Ste2p from a multicopy plasmid, measured in parallel. Solid, black, red, and blue lines show least-squares fits to data for the three isolates, yielding energy transfer efficiencies of 0.43, 0.87, and 0.59, respectively. The black dotted line indicates the predicted trace for E = 0. The long dashed black line indicates the predicted trance for E = 0.35. The short dashed black line indicates the predicted trace for E = 0.75. The predicted traces are shown for parameters derived from replicate 3 (black circles). To see this figure in color, go online.

Figure 8.

Plots of fluorescent emission in the transfer channel as a function of measured donor occupancy. Black, red, and blue solid symbols show measured emission for three independent isolates of strain A453, expressing full-length Ste2p from a CEN plasmid, measured in parallel. Solid, black, red, and blue lines show least-squares fits to data for the three isolates, yielding energy transfer efficiencies of 0.55, 0.65, and 0.52, respectively. The black dotted line indicates the predicted trace for E of 0. The short dashed black line indicates the predicted trance for E = 0.4. The long dashed black line indicates the predicted trace for E = 0.7. The predicted traces are shown for parameters derived from replicate 3 (black circles). To see this figure in color, go online.

Energy transfer between fluorescent antagonists

The peptide d-Tyr³-α-factor binds to Ste2p with similar affinity as normal α-factor but fails to elicit a detectable pheromone response (63). Versions of this peptide labeled on the K⁷ position with either NBD or BTR also bind to Ste2p with affinities similar to labeled forms of normal α-factor (28 ± 4 nM, n = 3 for the NBD-labeled antagonist; 9 ± 5 nM, n = 3 for the BTR-labeled antagonist; Fig. S5). However, the fluorescence emission from the NBD-labeled form of d-Tyr³-α-factor is not as blue shifted as that of [K⁷(NBD),Nle¹²]α-factor and is ∼10-fold less intense than that of receptor-bound [K⁷(NBD),Nle¹²]α-factor, indicating that the fluorophore of the antagonist resides in a more polar environment than that of the agonist (63). This reduced fluorescence of specifically bound [d-Tyr³K⁷(NBD),Nle¹²]α-factor resulted in the need for greater corrections for nonspecific binding compared with NBD-labeled normal α-factor.

Based on loss of donor emission, measurements using varying ratios of NBD- and BTR-labeled d-Tyr³ ligands in cells expressing high levels of C-terminally truncated Ste2p yielded a higher energy transfer efficiency between donor and acceptor ligands (0.83) than was observed for the labeled versions of normal α-factor (Table 1). This level of energy transfer, when combined with measurements of fluorescence emission in the transfer channel as a function of fractional occupancy of the antagonist donor, leads to an estimated ratio of extinction coefficients ε^D/ε^A of 4.5 ± 0.5, which is six times lower than that determined for labeled agonists. This is consistent with the observed decrease in fluorescence emission from receptor-bound NBD-labeled antagonist compared with the NBD-labeled agonist.

The observed increased efficiency of transfer compared with comparable labeled agonists suggests that the K⁷ positions of the receptor-bound antagonists are closer together than is the case for agonists. A different binding mode for antagonists is consistent with the reduced fluorescence emission of the NBD-labeled antagonist, the reduced spectral shift of the NBD-labeled antagonist, and the different spectrum of receptor mutations reported to affect the fluorescence emission of receptor-bound antagonist (63). However, there may also be contributions to the altered energy transfer that arise from different fluorescent properties or molecular mobilities of the fluorescently tagged antagonists, compared with the corresponding labeled agonists.

Discussion

We present here an approach for determining energy transfer efficiencies between differentially labeled fluorescent ligands bound to low abundance proteins in living cells. The approach allows determination of absolute energy transfer efficiencies despite the inability to directly measure the values of extinction coefficients and quantum yields for receptor-bound forms of environmentally sensitive fluorescent ligands. It is based on the use of flow cytometry, which allows simultaneous measurement of fluorescence emission for individual cells in three separate channels corresponding to directly excited donor emission, directly excited acceptor emission, and an energy transfer channel. Compared with fluorescence detection of cell suspensions using a spectrofluorometer, flow cytometric measurements are relatively unaffected by light scattering or by the presence of excess unbound ligand. Compared with fluorescence-microscopy-based approaches for measuring FRET, the use of flow cytometry allows the collection of unbiased quantitative data from a very large number of cells in suspension without concern for complications arising from photobleaching. Expressions relating fluorescence in the different optical channels to loss of donor emission and enhancement of acceptor emission at different donor/acceptor ratios served as the basis for least-squares fitting of energy transfer efficiencies to fluorescence measurements. Energy transfer efficiency was initially measured based on loss of donor fluorescence (which does not require knowledge of extinction coefficients or quantum yields) for cells that express high levels of receptor. The ratio of the extinction coefficients of receptor-bound donor and acceptor determined at the high receptor expression levels was then used to extend the measurements to low expression levels based on quantitation of enhanced acceptor emissions.

Two aspects of the analytical approach described here that appeared to be important for obtaining reliable results were 1) the use of high total concentrations of donor plus acceptor ligands. This effectively removed the need to consider populations of unoccupied receptors in the analysis, allowed extrapolation of the fraction of donor-occupied receptors from measurements of the acceptor-occupied fraction, and removed possible complications due to cooperativity of ligand binding by receptors; and 2) experimental determination of the proportions of bound donor and acceptor based on measurements of fluorescence in the acceptor channel. This minimized inaccuracies in determining ligand concentrations in different samples and stemming from uncertainties in determining binding affinities of different ligands.

Despite the large number of studies of receptor oligomerization using fusions of receptors to fluorescent proteins, and the wide application of fluorescent ligands for detecting and characterizing receptors (72,73), energy transfer between fluorescent ligands bound to receptors on intact cells has previously only rarely been applied for monitoring receptor oligomerization. 1) Early studies of epidermal growth factor (EGF) receptor dimerization looked for energy transfer between differentially labeled forms of EGF bound to receptors on intact mammalian cells. FRET was detected by fluorimetry of bulk solutions, by an indirect approach based on reduced photobleaching of donor in the presence of acceptor, and by fluorescence lifetime measurements (74). (We have attempted to detect ligand FRET via fluorescence lifetime measurements of cells in suspension and by fluorescence lifetime imaging microscopy in whole cells but find the number of detectable lifetime components to be too high to allow reliable determinations.) 2) A confocal microscopy study detected FRET between differentially labeled ligands for the luteinizing hormone receptor based on reduced photobleaching of donor in the presence of acceptor (75). 3) A confocal microscopy study of homo- and heterodimerization of somatostatin receptors using differentially labeled ligands (76) detected FRET based on intensity measurements in two emission channels upon illumination at a single wavelength—a technology known to be subject to considerable uncertainties in situations where the stoichiometries and concentrations of donor and acceptor can vary (77,78). 4) Time-resolved FRET studies of cell suspensions measured in a plate reader were used to enhance the signal/noise ratio of energy transfer between differentially labeled ligands for vasopressin, oxytocin, and dopamine receptors (68,79).

In view of the diversity of fluorescent ligands currently available for GPCRs and other receptors, we expect that the approaches that we describe here will find application to the study of additional receptor-ligand systems, particularly because they can be used for native cells that have not been engineered to overexpress fusion proteins. However, application to each new system would require development of labeled ligands and individual optimization. The use of flow cytometry also raises the possibility of using fluorescence-activated cell sorting for screening for variant forms of receptors that exhibit decreased or enhanced oligomerization.

In contrast to previous studies of Ste2p in intact cells that detected oligomerization of overexpressed, truncated forms of the receptor fused to fluorescent proteins that were present in multiple subcellular compartments, we have been able to draw the following major conclusions:

• 1)Concentration independence of FRET: Energy transfer between ligands bound to α-factor receptors occurs with approximately equal efficiency over an experimentally determined ∼50-fold range of expression levels (Table 1) extending from the high levels of C-terminally truncated receptors used in previous studies of Ste2p oligomerization to low, near-endogenous levels. Thus, previous reports of receptor oligomerization are unlikely to have resulted from high receptor densities in the membrane. This is a critical observation because high receptor densities have been proposed to drive receptor oligomerization via mass action (20, 21, 22) and to give rise to energy transfer because of proximity effects among noninteracting donors and acceptors (24, 25, 26,31, 32, 33, 34)}. The measurements of receptor expression levels that we conducted are relative and not absolute. However, an estimate of the absolute receptor density at the surface of cells expressing Ste2p from the normal endogenous locus can be derived from the consensus determination of 5000 receptors per cell (https://www.yeastgenome.org/) distributed over a yeast cell surface area of ∼250 μ² (80). This yields a density of 20 receptors per μ², indicating that the spacing between randomly distributed receptors should be on the order of thousands of ångstroms—enough to preclude random associations of diffusing receptor as a basis for energy transfer.
• 2)Oligomerization at the cell surface: Fluorescent pheromone analogs are membrane impermeant and are, in fact, detected only at the cell surface by fluorescence microscopy when endocytosis is inhibited (60). Thus, detection of energy transfer between receptor-bound [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor demonstrates oligomerization of receptors located at the cell surface. In contrast, substantial populations of the cellular complement of Ste2p fused to fluorescent proteins are found in intracellular compartments (42,54,56), raising the possibility that previous measurements of energy transfer between such receptor fusions in cells, using fluorimetry or BRET of cell suspensions (39, 40, 41, 42), may include contributions from receptors in intracellular locations.
• 3)Analysis of full-length receptors: Previous analyses of Ste2p oligomerization in intact cells have made use of receptors containing truncations of the C-terminal tail (39, 40, 41, 42,57). However, BRET analysis of isolated membranes containing full-length or C-terminally truncated Ste2p fused to luciferase and green fluorescent protein gave very different BRET efficiencies (39). We report here that energy transfer between receptor-bound ligands occurs with similar efficiency regardless of whether the receptors are full length or C-terminally truncated. Thus, the previously observed differences in energy transfer efficiency in comparing full-length and truncated receptors are apparently due to different locations or flexibilities of the fused reporter proteins.
• 4)Avoiding the use of fusion proteins: Unlike previous measurements of Ste2p oligomerization in intact cells (39, 40, 41, 42,57), the approach described here does not require fusion of receptors to large fluorescent or luminescent proteins that could affect tendencies of receptors to self-associate. Instead, oligomerization is detected using ligands derivatized with small fluorescent probes. However, an intrinsic drawback of using fluorescent ligands is that the approach is incapable of detecting ligand-dependent changes in oligomerization state.

The measured energy transfer efficiency of 59% that we have determined for agonist ligands implies that the separation between the fluorophores of the receptor-bound ligands is less than the Förster radius, R_0, for these fluorophores. These studies, along with previous characterization of these ligands (60, 61, 62, 63), indicate that receptor binding results in significant changes in the physical environments of the fluorophores that strongly affect their fluorescent properties. Lack of knowledge of the extinction coefficients, quantum yields, local refractive indices, and orientations of excitation and emission dipoles for the receptor-bound fluorophores makes it difficult to reliably determine the R₀ for the receptor-bound states of the donor and acceptor ligands. Other FRET pairs in which NBD is used as a donor with acceptor fluorophores excited in the same range as BTR have been reported to have R₀ values of 40–70 Å (81, 82, 83). Compared with these FRET pairs, the extended spectral overlap between NBD emission and BTR excitation and the apparently high quantum yield of receptor-bound [K⁷(NBD),Nle¹²]α-factor suggest that the applicable R₀ for the receptor-bound [K⁷(NBD),Nle¹²]α-factor and [K⁷(BTR),Nle¹²]α-factor can be expected to be at the higher end of this range (81).

In the structure of dimeric Ste2p bound to α-factor, the distance between the nitrogen atoms of the ε-amino groups of K⁷ of bound ligands is ∼19.8 Å (Fig. 9 a) (48). Based on the r⁶ dependence of energy transfer efficiency (81), the distance between donor and acceptor yielding an efficiency of 59% will be 94% of the R₀ distance. Even if we use the lower limit of estimated R₀ values for FRET pairs involving NBD, an energy transfer efficiency of 59% corresponds to an estimated distance between fluorophores of 37 Å, considerably greater than the distance between lysine ε-amino groups in the Ste2p structure. Part of this discrepancy can be explained by finite linker lengths of the attachments of the fluorophores to α-factor. The NBD-fluoride used for labeling (61) results in a very short linker, but the reagent used for BTR labeling contains a seven-atom aminohexanoyl spacer between the fluorophore and the amine-reactive N-hydroxysuccinimide group. However, even the interposition of this ∼10 Å linker does not seem adequate to explain the large distance between donor and acceptor fluorophores indicated by the measured energy transfer efficiency.

Figure 9.

(a) Structure of the dimer of Ste2p bound to α-factor. The arrow shows the 19.8 Å distance measured between the two ε-amino groups of the K⁷ residues of the bound molecules of α-factor. (b) View of the side chain of α-factor projecting into the cavity formed by extracellular ends of the first and second transmembrane helices of the receptor. The figure was prepared from the coordinates PDB: 7AD3 (48) using the PyMOL Molecular Graphics System (Schrödinger, LLC).

As shown in Fig. 9 b, in the structure of α-factor bound to Ste2p, the lysine side chain of residue K⁷ of α-factor is nestled into a cavity formed primarily from the extracellular ends of the first and second transmembrane helices of the receptor (48). The available space in this cavity is clearly too small to accommodate either the NBD or the BTR fluorophores attached to the lysine ε-amine. Thus, binding of the fluorophore-conjugated ligand must be accompanied by translation or rotation of residue K⁷ to allow its side chain to point toward unoccupied space outside the receptor helical bundle. This is likely to be a significant factor in increasing the apparent distance between receptor-bound fluorophores. Such a rearrangement of the lysine side chain is consistent with the observation that the environment of NBD attached to position K⁷ of α-factor is strongly dependent on the exact number of methylene groups connecting the fluorophore to the peptide backbone (61,62). However, modification of position K⁷ does not significantly disturb the rest of the interactions of α-factor with the receptor, as K⁷ is one of the few positions that can be modified with minimal effects on binding or signaling (48,60,63,84, 85, 86).

The approach described here cannot reliably report on the number of monomers per oligomer. Expressions for analyzing energy transfer in trimeric, tetrameric, or larger oligomers as a function of donor/acceptor ratios can be readily derived by analogy to the approach that we have described here for dimers. Such expressions predict dependencies of fluorescence emission on donor/acceptor ratios that are similar to those for dimers and thus can be readily fitted to flow cytometry data that we have presented (results not shown), precluding the use of these data to directly determine oligomer size. However, fits of the data to higher-order oligomeric models require the pairwise energy transfer efficiencies between individual donors and acceptors to be much lower than is the case if receptors are dimers because of the increased number of possibilities for energy transfer in a larger oligomer. Thus, if Ste2p exists as higher-order oligomers, the pairwise energy transfer between receptor-bound ligands would necessarily be lower than what we have calculated by assuming that receptors exist as dimers. This, in turn, would require the ligands bound to receptors in higher-order oligomers to be farther apart than we calculate from the analysis based on dimers, making our experimental results even more difficult to reconcile with the proximity of ligand binding sites in the available structure of ligand-bound dimeric receptors. This provides a tentative indication that Ste2p oligomers in the membrane of intact cells may not be larger than dimers.

Measurement of FRET between bound ligands, obviously, cannot address questions of whether oligomerization state might depend on ligand binding. Although analyses of Ste2p oligomerization in yeast cells and membranes have failed to find any ligand dependence (42,47) and the cryo-electron microscopy structure of ligand-bound Ste2p is dimeric (48), there have been reports of formation of higher-order oligomers in the presence of α-factor (46).

Forms of the α-factor antagonist [d-Tyr³]α-factor labeled with NBD and BTR at position K⁷ exhibited higher levels of energy transfer than those observed for similarly labeled normal α-factor. This suggests that the fluorophores of the receptor-bound antagonist ligands are closer together than is the case for labeled forms of normal α-factor. However, some of this agonist-antagonist difference may be due to altered physical environments of the fluorophores in the two types of ligands that affect the relative R₀-values and the extinction coefficients, as suggested by very different ratios of donor:acceptor extinction coefficients for the two ligands. Taken together, these findings are consistent with previous observations of the different binding modes and fluorescence properties of Ste2p-bound fluorescent α-factor versus antagonists (63).

Author contributions

S.M.C. performed most of the experiments and helped to develop the approach. R.S. performed some experiments and helped to develop the analysis. F.N. provided the labeled peptides and participated in the writing of the manuscript. M.E.D. conceived of the overall approach, performed some experiments, analyzed the data, and wrote most of the manuscript. All listed authors contributed to editing the manuscript.

Editor: Vasanthi Jayaraman.