FRET imaging reveals that functional neurokinin-1 receptors are monomeric and reside in membrane microdomains of live cells Bruno H. Meyer*†, Jean-Manuel Segura*, Karen L. Martinez‡, Ruud Hovius, Nathalie George, Kai Johnsson, and Horst Vogel§ Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Institut des Sciences et Inge´nierie Chimiques, CH-1015 Lausanne, Switzerland Edited by Axel T. Brunger, Stanford University, Stanford, CA, and approved December 13, 2005 (received for review September 2, 2005)
The lateral organization of a prototypical G protein-coupled receptor, the neurokinin-1 receptor (NK1R), was investigated in living cells by fluorescence resonance energy transfer (FRET) microscopy, taking advantage of the recently developed acyl carrier protein (ACP) labeling technique. The NK1R was expressed as fusion protein with ACP to which small fluorophores were then covalently bound. Our approach allowed the recording of FRET images of receptors on living cells with unprecedented high signal-to-noise ratios and a subsequent unequivocal quantification of the FRET data owing to (i) the free choice of optimal fluorophores, (ii) the labeling of NK1Rs exclusively on the cell surface, and (iii) the precise control of the donor–acceptor molar ratio. Our single-cell FRET measurements exclude the presence of constitutive or ligandinduced homodimers or oligomers of NK1Rs. The strong dependence of FRET on the receptor concentration further reveals that NK1Rs tend to concentrate in microdomains, which are found to constitute ⬇1% of the cell membrane and to be sensitive to cholesterol depletion. ACP labeling 兩 G protein-coupled receptor (GPCR) oligomerization
G
protein-coupled receptors (GPCRs) were for a long time presumed to be distributed in the plasma membrane exclusively in a monomeric form (1, 2), but recent reports have unveiled a more complex behavior; in particular, dimeric structures have been found for several GPCRs using biochemical and biophysical methods (3–9). Dimerization can occur between receptors of the same subtype (homodimerization) or of different subtypes (heterodimerization). Some GPCRs remain dimeric all of the time, whereas others cycle between monomeric and dimeric states in a ligand-regulated process (7). Although GPCR homodimerization seems to be important for receptor ontology and trafficking, heterodimerization might result in altered ligand selectivity and distinctive coupling to signal transduction pathways, providing an additional possibility for the fine tuning of cellular signaling. In addition to dimerization, the lateral distribution of GPCRs in cell membranes has been extensively debated recently. Several reports based on biochemical (10), plasmon-resonance spectroscopy (11), single-molecule microscopy (12), and fluorescence recovery after photobleaching experiments (13) propose that GPCRs are localized in microdomains, but a clear demonstration of the existence and nature of such microdomains in living cells remains elusive, in particular because the interpretation of biochemical data can be rather equivocal (14–17). Compartmentalization in form of microdomains was proposed to explain the efficiency of signal transduction at the low physiological surface concentrations of the signaling partners by their enrichment inside specialized signaling platforms (10, 18). Bioluminescence resonance energy transfer (BRET) and fluorescence resonance energy transfer (FRET) experiments have gained increasing interest to investigate these two central questions on GPCR signaling. (i) They can be performed directly in living cells (6). (ii) They offer the possibility to distinguish between 2138 –2143 兩 PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7
monomers, dimers, and higher-order oligomers from the dependence of the BRET or FRET signal on the donor–acceptor (DA) ratios and on the receptor expression levels in the membrane (6). (iii) They provide information on the lateral distribution of membrane proteins such as compartmentalization in microdomains (19, 20). Surprisingly few reports on GPCRs made full use of the possibility to obtain such quantitative information (21–23). Reliable quantitative FRET and BRET measurements on living cells are limited by the quality of the signals observed, which is directly related to the fluorophores used. Most studies use fluorescent proteins, respectively luciferase for BRET, and are therefore suffering from background signals arising from incompletely processed proteins inside the cell and from the high cell autofluorescence in this spectral region. This problem together with the difficulty to adjust DA ratios complicates quantitative investigations. In contrast, posttranslational labeling methods (24–27) are more versatile. In particular, we have shown previously the feasibility of the enzymatic transfer of fluorescent groups from labeled CoA to an acyl carrier protein (ACP) fused to the protein of interest (25) and demonstrated its advantages over labeling using fluorescent proteins: the possibility to choose optimal fluorophores, an exclusive labeling of the receptors translocated to the plasma membrane, and the control of the DA ratio (25, 28). In the present study, we have investigated the neurokinin-1 receptor (NK1R) as a prototypical example of a GPCR. The NK1R mediates processes such as nociception, neural inflammation, or smooth muscle contraction, and receives considerable attention as a drug target, for instance, for treatments of depression. Therefore, it is of importance to investigate the organization, i.e., dimerization and localization of the NK1R in the plasma membrane, and to correlate it with the receptor function. Although it was shown that NK1R can form heterodimers with the human -opioid receptor (29), it is unknown whether it is also able to form homodimers. In addition, detergent extraction studies showed that the NK1R resides in a detergent resistant membrane fraction, which raises the question of whether the receptors are localized in microdomains of the cell plasma membrane (30). Here, we used the advantages of the ACP labeling technique to address the questions of homodimerization and microdomain localization directly in living cells using FRET microscopy. The NK1R fused with ACP (ACP–NK1R) was labeled simultaneously with Cy3 (donor) and Cy5 (acceptor) at Conflict of interest statement: No conflicts declared. This paper was submitted directly (Track II) to the PNAS office. Abbreviations: NK1R, neurokinin-1 receptor; ACP, acyl carrier protein; DA, donor–acceptor; BRET, bioluminescence resonance energy transfer; GPCR, G protein-coupled receptor; D-PBS, Dulbecco PBS. *B.H.M. and J.-M.S. contributed equally to this work. †Present
address: Novartis Institutes for BioMedical Research, CH-4002 Basel, Switzerland.
‡Present
address: Nano-Science Center, University of Copenhagen, Universitetsparken 5, DK-2100 Copenhagen, Denmark.
§To
whom correspondence should be addressed. E-mail:
[email protected].
© 2006 by The National Academy of Sciences of the USA
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0507686103
different, well defined ratios. At expression levels of NK1R close to physiological conditions, no FRET signal was detected, whereas high FRET efficiencies with a strong dependence on DA ratio were observed at higher receptor concentrations. The dependence was not linear, excluding that the FRET signals were due to the presence of dimers. It was furthermore possible to rule out the presence of oligomers from the sharp dependence of the FRET signal on the expression level. Our FRET data can be explained to result from stochastic encounters of the donor and acceptor at an effective, local concentration ⬇80 times higher than expected from a homogenous distribution due to compartmentalization of the NK1R inside microdomains. Results Controlled Labeling of ACP–NK1R. Previous reports from our labo-
ratories have shown that the fusion protein ACP–NK1R expressed in HEK293 cells can be specifically labeled (25) and preserves its function to be activated by the agonist substance P, eliciting a calcium response at a similar effective concentration as the wildtype receptor (28). For FRET imaging, ACP-NK1Rs were doubly labeled with Cy3 (donor) and Cy5 (acceptor) by mixing the substrates at defined ratios before adding to the cells. Fig. 1 A and B show typical donor and acceptor fluorescence images of a living cell. Because only receptors correctly translocated to the plasma membrane are accessible for labeling, there is no background fluorescence inside the cells. Quantitative FRET experiments require exact knowledge of the DA ratios on the cells. To obtain the relation between the DA ratio on the cell surface and the corresponding DA ratio in solution used for labeling, HEK293 cells expressing very low amount (⬇25,000 receptors per cell) of ACP-NK1Rs were labeled using donor mole fractions in the labeling solution, xD,sol, ranging from 0.2–0.8. We did not observe FRET at such low expression levels so that the respective concentrations of donor and acceptor were directly proportional to the average fluorescence intensities in the corresponding channels. The observed donor mole fraction on the cell xD,cell varied linearly with xD,sol (Fig. 1C), which was used to predict the DA ratio on the cell surface from the concentrations of the fluorescent substrates used for labeling. NK1Rs Are Monomeric at Physiological Concentrations. Here, con-
trolled dual color labeling was used to investigate the degree of oligomerization of the NK1R by FRET imaging. We first studied HEK293 cells stably expressing ACP–NK1R at a low concentration (450 ⫾ 48 fmol兾mg protein; ⬇25,000 receptors per cell). At all donor mole fractions, xD, used for labeling, no FRET was detected (Fig. 2, black circles), strongly indicating that NK1Rs are monomeric at such low densities close to physiological concentrations Meyer et al.
Fig. 2. Dependence of the FRET efficiency on the donor mole fraction. HEK293 cells expressing ACP–NK1Rs were labeled with a mixture of Cy3 and Cy5 fluorophores at different donor mole fractions xD. The apparent FRET efficiency, Eapp,se, is plotted vs. xD for two populations of cells: Gray circles represent transiently expressing cells; black circles stem from stably expressing cells showing a 2.5 times lower expression. The expression level was similar for all investigated cells in a population. Each point is a mean (⫾SD) of 10 cells. Fitting the data with Eq. 8 plus an offset (solid curves) yields for the higher expressing cells offset ⫽ ⫺0.022 ⫾ 0.030 (i.e., zero within the experimental accuracy), E ⫽ 0.26 ⫾ 0.09, n ⫽ 3.8 ⫾ 1.3, and for the lower expressing cells with E fixed to 0.26 offset ⫽ 0.018 ⫾ 0.047, n ⫽ 1.09 ⫾ 0.28. The dotted curve represents a plot of Eapp,se calculated by using Eq. 12 in its range of validity, with ␣ ⫽ 81, R ⫽ 256 receptors per m2, and xD as an independent variable.
(31–33). To exclude that this observation was only due to the stable expression of the NK1R, FRET was measured on transiently expressing cells with similar expression levels leading to the same result. To know whether the FRET efficiency varied with the receptor expression, we then investigated HEK293 cells transiently expressing 2.5 times more ACP–NK1Rs than before (⬇63,000 receptors per cell). Strikingly, such a small concentration increase resulted in the apparition of a strong FRET signal (Fig. 2, gray circles). The dependence of the apparent FRET efficiency of sensitized acceptor emission, Eapp,se, on xD was analyzed to determine the degree of aggregation using an adapted model of Veatch and Stryer (34). A fit of Eapp,se as a function of xD using Eq. 8 yielded the true FRET efficiency, E ⫽ 0.26 ⫾ 0.09, and the aggregation number, n ⫽ 3.8 ⫾ 1.3. By contrast, a fit of the data at low expression levels (⬇25,000 receptors per cell) with E fixed at 0.26 using Eq. 8 yielded an aggregation number n ⫽ 1.09 ⫾ 0.28, confirming that NK1Rs are monomeric at physiological conditions. For dimers (n ⫽ 2), Eapp,se is expected to depend linearly on xD according to Eq. 8. Here, the aggregation number was significantly higher than 2, showing that NK1Rs also do not form dimers at higher expression levels but rather aggregates of roughly tetrameric size. From E and the Fo ¨rster distance for Cy3 and Cy5 of R0 ⫽ 50 Å (19), the average distance between donor and acceptor in a putative tetramer was calculated to be 59.5 ⫾ 4.6 Å in the same range as the estimated diameter of a GPCR of 40–50 Å (22). To interpret these findings accurately, it is essential to assess the potential influence of unlabeled receptors on the measured FRET. Unlabeled receptors might be present because of (i) incomplete labeling of the ACP–NK1R, (ii) nonconjugated CoA in the labeling solution, or (iii) presence of wild-type NK1R or other nonlabeled GPCRs forming dimers with the NK1R. Points i and ii can be excluded because saturation labeling experiments (data not shown and ref. 25) revealed that the labeling efficiency under the used conditions is ⬎95%, whereas analysis of the purity of conjugated CoA showed no trace of nonconjugated CoA. HEK293 cells do not express endogenously NK1Rs, but it cannot be excluded that other GPCRs present might form heterodimers with NK1R. The effect PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2139
BIOPHYSICS
Fig. 1. Controlled double-labeling of ACP–NK1R. (A and B) Fluorescence confocal micrographs show a HEK293 cell transiently expressing ACP–NK1Rs, which were labeled simultaneously with Cy3 and Cy5 at a Cy3 mole fraction of 0.67. The cell was imaged by using particular filter sets for donor (A) and acceptor (B). (Scale bar: 10 m.) (C) The donor mole fraction observed on the cells, xD,cell, as a function of the donor mole fraction used for the labeling, xD,sol. xD,sol is identical to xD,cell within experimental error as shown by a linear regression yielding an intercept of 0.035 ⫾ 0.073 and a slope of 0.94 ⫾ 0.12. Each point is the mean (⫾SD) of 10 cells.
Fig. 3. FRET dependence on cell surface receptor density. (A) HEK293 cells transiently expressing the ACP–NK1R were labeled at a donor mole fraction of xD ⫽ 0.67. The apparent FRET efficiency, Eapp,se (gray circles), is plotted as a function of the reduced acceptor concentration, cA (bottom), and the receptor surface density, R (top). Each data point represents an image containing one to three cells. The bold dotted curve at the bottom represents Eapp,se calculated with Eq. 12 assuming the absence of microdomains (␣ ⫽ 1; coefficients A1;2 and k1;2 taken for a Rc兾R0 ratio of 1.1; see Theory). The fine dotted horizontal line represents the maximal apparent efficiency determined from sensitized acceptor emission: Eapp,se,max ⫽ 0.626 ⫾ 0.042 (mean of the last 17 points ⫾ SD). The two vertical dashed lines represent the receptor surface densities for the data in Fig. 2. The solid curve is a fit over the first 47 points using Eq. 12 with only one free parameter ␣ ⫽ 81.1 ⫾ 0.3. The dotted curves represent Eapp,se calculated from Eq. 12 with ␣ ⫽ 81 ⫾ 50%. (B) Modulation of the apparent FRET efficiency by cholesterol extraction. FRET efficiency of transiently expressed ACP–NK1Rs labeled with xD ⫽ 0.6, before (0.34 ⫾ 0.048) and after (0.24 ⫾ 0.076) cholesterol extraction with 2% (wt兾vol) methyl--cyclodextrin for 45 min at room temperature (mean of 9 cells ⫾ SE; Student’s t test P ⱕ 0.005).
of unlabeled receptors can be estimated by fitting the data (Fig. 2, gray circles) using Eq. 10 and assuming varying mole fractions of unlabeled receptor x⬘U: Even a value of x⬘U of up to 35% did not noticeably change the resulting FRET efficiency. Furthermore, n was found to increase with x⬘U, excluding the possibility that the measured value of n ⬎ 2 would be due to the presence of unlabeled receptors. The possible coexistence of monomers and oligomers also does not account for n ⬎ 2, because it only results in the decrease of Eapp,se by a constant factor independent of xD (see FRET Efficiency Measured in the Presence of a Mixture of Oligomers and Monomers in Supporting Text, which is published as supporting information on the PNAS web site). Based on these results, we conclude that the NK1R is monomeric at physiological concentrations and does not form homodimers at higher expression levels. The existence of real tetramers at higher expression levels seems furthermore unlikely because (i) such a tetramer would exhibit an extremely low stability resulting in complete dissociation at physiological concentrations and (ii) no supporting structural indications for tetramers of GPCRs exist. Some GPCRs show an agonist-dependent dimerization (35). To learn whether this property is also found for the NK1R, we investigated the stable cell line expressing low levels of ACP–NK1R (⬇25,000 receptors per cell) labeled with a donor mole fraction of 0.6. Addition of 100 nM natural agonist substance P did not result in a significant increase of the apparent FRET efficiency (0.008 ⫾ 0.010 to 0.011 ⫾ 0.015; mean of 7 cells ⫾ SE.), indicating that the NK1Rs do not dimerize upon agonist binding. NK1Rs Reside in Membrane Microdomains. In a next step, we addressed the question of whether the observed FRET at the higher expression levels (⬇63,000 receptors per cell; 256 receptors per m2) might be due to a high frequency of stochastic encounters between donors and acceptors. To this aim, we transiently expressed ACP–NK1R in HEK293 cells and labeled the receptors using a constant donor mole fraction of 0.67. Transient transfection provided a heterogeneous population of cells covering a wide distribution of receptor expression levels, which in turn allowed us to measure FRET at various receptor surface densities. Fig. 3 shows Eapp,se both as a function of receptor surface density (R) and as a function of the concentration of acceptors per R20, 2140 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0507686103
cA ⫽ xARR20 (denoted also as reduced acceptor concentration). Eapp,se strongly depends on the receptor surface density and reaches saturation at Eapp,se,max ⫽ 0.626 ⫾ 0.042 at ⬇600 receptors per m2 (Fig. 3, fine dotted line). Eapp,se,max is expected to be reached when there is a maximal frequency of stochastic encounters, i.e., when the NK1Rs are closely packed (each receptor surrounded by six neighbors; n ⫽ 7). Under this condition the true efficiency can be estimated with Eq. 8 to be 0.34, which corresponds to a distance of closest approach Rc ⫽ 55.9 ⫾ 0.9 Å, in the same range as the estimated diameter of a GPCR of 40–50 Å (22). Virtually all NK1Rs are closely packed at this surface concentration because otherwise Eapp,se,max would be lower. The apparent FRET efficiency resulting from stochastic encounters between donors and acceptors at a reduced acceptor concentration cA can be accurately calculated by using an analytical expression derived by Wolber and Hudson (36) (Eq. 12), which only depends on the ratio of Rc兾R0, in our case 1.1 as calculated from the FRET efficiency of densely packed receptors. A plot of Eq. 12 (bold dotted curve in Fig. 3) shows that the surface density of NK1R is too low to allow sufficient stochastic encounters to account for the experimentally observed high apparent FRET efficiencies. A possible explanation for this apparent discrepancy would be that the effective density of NK1Rs at the cell surface is higher than the one estimated from the calibration. This explanation would be the case if the NK1R would not be homogeneously distributed but would reside in separated domains. NK1Rs expressed at an average surface concentration of 600 receptors per m2 would reach a local density of 40,000 receptors per m2 corresponding to a state of closest packing if they were concentrated in only ⬇1.5% of the cell membrane. A more precise estimate can be obtained by fitting the initial part of the data in Fig. 3 using Eq. 12, allowing only one free parameter, the ‘‘scaling’’ factor ␣, to account for the increased local concentration. The analytical expression in Eq. 12 is known to be valid only for a limited range between 0 and 0.5 of the local reduced acceptor concentration (36). The corresponding fit (bold curve in Fig. 3) yields a scaling factor ␣ ⫽ 81.1 ⫾ 0.3. Taking into account the scatter of the data, the variation of the scaling factor between cells can be estimated to be ⬇50% (dotted curves in Fig. 3). NK1Rs are therefore concentrated in microdomains representing only 1兾␣ ⫽ 0.8–2.5% of the total surface area of the plasma membrane. Meyer et al.
Cholesterol Depletion of the Plasma Membrane Decreases FRET.
Monastyrskaya et al. (30) showed that the NK1R is present in detergent-resistant membrane fractions from which it disappeared after cholesterol depletion using methyl--cyclodextrin (MBCD) treatment. To test whether the microdomains observed by FRET correspond to the detergent-resistant membrane fractions, the dependence of Eapp,se on cholesterol extraction was investigated on cells transiently expressing ACP–NK1R and showing high NK1R densities (Fig. 3B). After incubation with 2% MBCD at room temperature for 45 min (same conditions as in ref. 30), Eapp,se decreased significantly from 0.34 ⫾ 0.048 to 0.24 ⫾ 0.076 (mean of 9 cells ⫾ SE.), supporting a model where NK1Rs are localized in microdomains. Discussion The detailed insights into the organization of the NK1R in living cells presented here could only be gained by making use of the power of the ACP labeling for quantitative FRET measurements. First, ACP labeling enabled the recording of FRET images with excellent signal-to-noise ratios owing to the free choice of the fluorophores and the exclusive labeling of the cell surface receptors. Second, the ACP labeling allowed a precise control of the DA ratio, which was a requirement for assessing the oligomerization state of the NK1R by FRET. Presently, we cannot exclude that the N terminus of the NK1R might be involved in dimerization (6), and in turn that the fusion of ACP to the N terminus would abolish dimerization. However, because the ACP–NK1R was found to activate calcium signaling, homodimerization obviously is not required for functionality of the NK1R. The results concerning the dependence of FRET on the cell surface concentration of NK1R have to be considered in the context of physiological expression levels reported to be 550 fmol兾mg membrane protein in rat submaxillary gland membranes (31), 107 fmol兾mg membrane protein in guinea pig lung membranes (32), and 254 fmol兾mg membrane protein in rat brain membranes (33). In our experiments the lowest expression levels were 450 fmol兾mg total protein in cells stably expressing ACP– NK1R. Taking into account that the total protein content of a cell is higher than the membrane protein content, our lowest expression levels were still higher than those under physiological conditions. Yet, no FRET was detected, showing that NK1Rs do not dimerize or oligomerize at physiological expression levels. The strong dependence of FRET on the expression level of NK1R was not observed for some other GPCRs; stable BRET or FRET signals have been measured for receptor expression levels ranging from 1.4 to 26.3 pmol兾mg protein for 2-AR receptors (22), from 40 to 100 fmol兾mg membrane protein for CCR5 receptors (37), and from 200 to 1,000 fmol兾mg membrane protein for opioid Meyer et al.
receptors (38). Conversely, the FRET signal was also found to disappear at low expression levels for the SSTR5 receptor but could be recovered by addition of the agonist in contrast to what we observe on NK1R (35). These interesting differences in the dimerization behavior of GPCRs might reflect variations in the functionality yet to be unraveled. For instance, an attractive explanation for the absence of homodimers of the NK1R might be that this behavior is a means to favor the formation of heterodimers, corroborating previous reports on the association of NK1R with the -opioid receptor (29). It is worth noting that although the experiments exclude an essential role of NK1R dimerization for signal transduction, they do not address its importance for ontology and trafficking (7) because we do not probe intracellular receptors. The strong concentration dependence of the FRET signal is a clear signature that the proximity of the NK1R arises from stochastic encounters. It is remarkable that the amount of FRET can only be accounted for by postulating an 80 times higher concentration showing that the NK1R explores only a limited fraction of the plasma membrane consisting of microdomains with sizes below the optical resolution. The presence of NK1R in microdomains was already suggested by Holst et al. (39) to explain their observation of different pharmacological phenotypes of the NK1R that do not interchange in the cell membrane. Furthermore, NK1R was found to be present in the detergent-resistant membrane fraction upon detergent extraction in a cholesterol-dependent manner (30), which was interpreted as the NK1R being localized in liquid-ordered domains enriched in cholesterol and saturated lipids, i.e., mostly sphingolipids in mammalian cells. Our observation of a cholesteroldependent FRET signal supports the existence of such microdomains in a living cell, although interpretation of cholesteroldepletion experiments should always be considered with caution (14, 40). The maximal surface covered by the microdomains of ⬇1% of the plasma membrane is probably determined by multiple factors, in particular the amount of sphingolipids, cholesterol, and associated transmembrane proteins. Variations of these factors between individual cells are the likely origin of the scatter of the data observed in Fig. 3. It is remarkable that the FRET signal only occurred at concentrations above a critical surface density of 100 receptors per m2 indicating that the microdomains are very small with a diameter of ⬇10 nm. This small size might be the result of the stabilization of the microdomains by the NK1R, possibly owing to hydrophobic mismatching and receptor palmitoylation. As a result of their small size, the microdomains are populated with only one NK1R at physiological concentrations. The existence of the microdomains seems to be crucial for signal transduction because cholesterol depletion was shown to abolish signaling (30). A reason for this finding might be that the microdomains are able to recruit heterotrimeric G proteins as was observed in recent experiments (11, 18). This recruitment will lead to an effective increase by a factor of up to 80 of the respective concentrations of the various signaling partners inside the domains. Such enriched microdomains then would function as efficient signaling platforms. Materials and Methods Materials. Materials used included DMEM, FCS, and Dulbecco PBS (D-PBS) (all from Invitrogen), hygromycin B (Calbiochem), BSA (Fluka), methyl--cyclodextrin (Sigma), substance P (from K. Servis, University of Lausanne, Lausanne), bacitracin (Serva), 1,2-dioleoyl-sn-glycero-3-[phosphatidyl-rac-(1-glycerol)] (DOPG) (Avanti Polar Lipids), and 1,2-dipalmytoyl-sn-3-phosphatidylethanolamine-N–Cy5 (DPPE–Cy5) (from Silke Mark, Ludwig Institute, Epalinges, Switzerland). Synthesis of CoA–Cy3 and CoA–Cy5 and expression and purification of recombinant AcpS (PPTase from Escherichia coli) are described in ref. 25. Cell Culture and Transfection. Adherent HEK293 cells were grown
in DMEM supplemented with 2.5% FCS. The cultures were kept PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2141
BIOPHYSICS
Remarkably, at low surface densities the apparent FRET efficiency vanishes resulting in a deviation from the fitting curve and only starts to increase when the surface density exceeds a critical value of ⬇100 receptors per m2 of cell surface (i.e., ⬇8,000 receptors per m2 of microdomains). This behavior might be an indication that the microdomains are relatively small so that they become effectively populated with several receptors only at high surface densities of NK1R. In this case, considering that only one receptor is present per microdomain at the critical surface density, the area of a microdomain can be estimated to be ⬇125 nm2 (1 m2兾8,000), which corresponds to a microdomain diameter of ⬇10 nm. To make sure that the presence of microdomains can also explain the dependence of Eapp,se on xD (Fig. 2, gray circles) observed for cells having expression levels of 256 receptors per m2 (⬇63,000 receptors per cell), we calculated Eapp,se as a function of xD from Eq. 12 using the parameters obtained previously (Fig. 2, dotted curve). The dotted curve reproduces the data reasonably well in the range of xD between 0.5 and 1 where Eq. 12 is valid.
at 37°C in a humidified atmosphere with 5% CO2. For confocal microscopy, HEK293 cells were seeded (105 cells per ml) into 8-well plates (Nalge Nunc International) or transferred onto 0.17-mm thick, 25-mm diameter glass coverslips (Assistent, Berlin) deposited in 6-well plates (TPP, Trasadingen, Switzerland), containing DMEM兾FCS. At 16–20 h after splitting, cells were transfected by using the calcium phosphate method. Experiments were performed 24–55 h after transfection. Before optical imaging, the medium was replaced by D-PBS (with 0.1% wt兾vol BSA in the case of ligandbinding experiments). Stable HEK293 cell lines were produced from transiently transfected cells by selection with 200 g兾ml hygromycin B. Protein content was determined by using the Bradford protein assay (Pierce) with BSA as standard. ACP Labeling. Cells were first washed with D-PBS and then labeled in D-PBS, 10 mM MgCl2, 1 M AcpS, and 5 M CoA-substrate. For dual color labeling, the substrates CoA–Cy3 and CoA–Cy5 were first mixed at the desired ratio to a total concentration of 5 M and then added to the cell culture. The labeling was performed at 19°C for 40 min with an efficiency ⬎95%. Cells were then washed three times with D-PBS. FRET Microscopy. FRET efficiencies were determined by measuring
the sensitized acceptor emission. Laser-scanning confocal micrographs were recorded on a LSM 510 microscope (Zeiss) equipped with a 63⫻ water objective (1.2W Korr, Zeiss) excited at either 543 or 633 nm using helium–neon lasers. Detection of fluorescence signals of acceptor and donor兾FRET was achieved by appropriate filter sets. The settings for the acceptor were 633-nm excitation wavelength, dichroic mirror HFT 633, and LP 650 emission filter (acceptor channel). The donor and FRET images were recorded simultaneously with excitation at 543 nm and a dichroic mirror HFT 543, the emission beam was split with a NFT 635 dichroic mirror onto two detectors with a BP 560-615 filter for donor emission (donor channel), and a LP 650 filter for acceptor emission (FRET channel), respectively. The laser power for the excitation and gain of the detectors for the emission were adjusted for each experiment so that the signal covered the whole dynamic range of the instrument, yielding images with a good signal-to-noise ratio. We took care that the donor and acceptor signals were not saturated. Fluorescence intensities on the cell were calibrated by using measurements of fluorescence intensities of solutions of the donor or acceptor at known concentrations at the same experimental settings, allowing comparison of expression levels in various cells and calculation of observed donor mole fractions from xD,obs ⫽ [D]兾([D] ⫹ [A]), where [D] and [A] are donor and acceptor concentrations, respectively. Receptor surface densities were estimated by comparing acceptor signals on the cell with those measured under the same conditions on lipid vesicles containing DPPE–Cy5 at a known surface density (see Estimation of Receptor Surface Densities on Cells Using Artificial Lipid Vesicles in Supporting Text). After applying a constant threshold on the donor image to select the membrane comprising the labeled ACP–NK1, FRET ratios (FR) were calculated on a pixel-by-pixel basis by using (for a generalized derivation, see ref. 41, FRET Microscopy Using Sensitized Acceptor Emission in Supporting Text, and Fig. 4, which is published as supporting information on the PNAS web site) FR ⫽
Ff ⫺ S1Df , S4 Af
[1]
where Ff, Df, and Af are the experimental signals measured on cells using the FRET, donor, and acceptor filter set, respectively; S1 and S4 are cross-talk factors correcting for the donor emission detected in the FRET channel and the acceptor emission due to direct excitation, respectively, which were determined from measurements on donor or acceptor only, respectively (see FRET Micros2142 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0507686103
copy Using Sensitized Acceptor Emission in Supporting Text and Fig. 4). A histogram of FR values then was built and fitted with a Gaussian distribution to yield the mean FR. This procedure minimized possible artifacts introduced by the threshold at low FR values. Apparent FRET efficiencies for sensitized acceptor emission (Eapp,se) were calculated using Eapp,se ⫽ 共FR ⫺ 1兲关 A共 兲兾 D共 兲兴,
[2]
where A() and D() are the molar extinction coefficients of donor and acceptor, respectively, at the donor excitation wavelength (543 nm). For Cy3–Cy5, the ratio A()兾D() at 543 nm is 0.11. Image treatment and data analysis were performed in IGOR PRO 5 (WaveMetrics, Lake Oswego, OR). Unless differently stated, all indicated errors are 95% confidence intervals. Error bars represent standard deviations (SDs). Theory Relationship Between Donor Quenching and Sensitized Acceptor Emission. The true FRET efficiency, E, refers to the transfer of energy
between a donor and acceptor molecule in a single DA pair. In a large population of donors and acceptors, only a fraction of the molecules will contribute to FRET resulting in a measured apparent FRET efficiency, Eapp, deviating from E. Eapp depends on the method used to determine FRET, which is based on measuring either donor quenching, (Eapp,dq ⫽ 1 ⫺ (FDA兾FD), or sensitized acceptor emission, Eapp,se ⫽ (FAD兾FA ⫺1)[A()兾D()], where FDA is the donor fluorescence in presence of acceptor, FD is the donor fluorescence in absence of acceptor, FAD is the acceptor emission in presence of donor, and FA is the acceptor emission in absence of the donor. For a mixture of donor and acceptor molecules, where only a fraction of them performs FRET, Eapp,dq and Eapp,se are not necessarily identical. Most models described in the literature refer to FRET measurements using donor quenching, so that we derive in the following a simple equation allowing to relate Eapp,se and Eapp,dq. Energy conservation dictates that the excitations lost by energy transfer from the donor are gained by the acceptor. Excitations directly correspond to detected fluorescence by a proportionality factor comprising the fluorescence quantum yield and the detection efficiency. By using the fluorescence quantum yields, D and A, and the detection efficiencies, D and A, for donor and acceptor, respectively, one yields 1 1 共F ⫺ F DA兲 ⫽ 共F ⫺ F A兲. D D D A A AD
[3]
Fluorescence is proportional to number of molecules, extinction coefficients, fluorescence quantum yields, and detection efficiencies: FD ⬀ xDDDD and FA ⬀ xAAAA. FA x A A A A ⫽ . FD xD D D D
[4]
Combining Eqs. 3 and 4 yields
冉
A F AD ⫺1 D FA
冊 冉 ⫽
1⫺
冊
F DA x D , FD xA
Eapp,se ⫽ E app,dq
xD , xA
or
[5]
[6]
where xD and xA are the donor and acceptor mole fractions. FRET in Oligomers. Eapp varies in a characteristic manner with the
donor and acceptor mole fraction. A model was developed previously allowing to relate Eapp,dq to E and xD (42) Meyer et al.
[7]
where n is the number of units in an oligomer. By using Eqs. 6 and 7 and noting that xA ⫹ xD ⫽ 1, an expression for Eapp,se can be obtained Eapp,se ⫽ E
xD n⫺1 共1 ⫺ x D 兲. 1 ⫺ xD
[8]
Incomplete Labeling. Adair and Engelman (42) extended the model
for the case of incomplete labeling to Eapp,dq ⫽ E共1 ⫺ 共x⬘DU兲 n⫺1兲,
[9]
where x⬘DU is the sum of the mole fraction of donor labeled (x⬘D) and unlabeled (x⬘U) receptors. The prime denotes mole fraction on the cells distinct from mole fractions used for labeling (xD and xA) owing to the presence of unlabeled receptor. Combining Eqs. 6 and 9, Eapp,se for incomplete labeling is Eapp,se ⫽ E
xD 共1 ⫺ 共x⬘DU兲 n⫺1兲, xA
[10]
where x⬘DU ⫽ xD(1 ⫺ x⬘U) ⫹ x⬘U due to the fact that x⬘DU ⫽ x⬘D ⫹ x⬘U and x⬘D ⫽ xD(1 ⫺ x⬘U). FRET for Randomly Distributed Donors and Acceptors. For randomly distributed donors and acceptors in membranes, the energy transfer efficiency is a function of the Fo ¨rster distance (R0), the distance of closest approach of donor and acceptor (Rc), and the surface density of acceptors (19). Wolber and Hudson (36) calculated the dependence of the FRET efficiency on the donor and acceptor concentrations and showed that it could be described with an accuracy better than 1% by the analytical expression 1. Birnbaumer, L. (1990) FASEB J. 4, 3178–3188. 2. Chabre, M. & le Maire, M. (2005) Biochemistry 44, 9395–9403. 3. Angers, S., Salahpour, A. & Bouvier, M. (2002) Annu. Rev. Pharmacol. Toxicol. 42, 409–435. 4. Milligan, G. (2004) Mol. Pharmacol. 66, 1–7. 5. Park, P. S., Filipek, S., Wells, J. W. & Palczewski, K. (2004) Biochemistry 43, 15643–15656. 6. Pfleger, K. D. & Eidne, K. A. (2005) Biochem. J. 385, 625–637. 7. Terrillon, S. & Bouvier, M. (2004) EMBO Rep. 5, 30–34. 8. Stanasila, L., Perez, J. B., Vogel, H. & Cotecchia, S. (2003) J. Biol. Chem. 278, 40239–40251. 9. Fotiadis, D., Liang, Y., Filipek, S., Saperstein, D. A., Engel, A. & Palczewski, K. (2003) Nature 421, 127–128. 10. Ostrom, R. S. & Insel, P. A. (2004) Br. J. Pharmacol. 143, 235–245. 11. Alves, I. D., Salamon, Z., Hruby, V. J. & Tollin, G. (2005) Biochemistry 44, 9168–9178. 12. Lill, Y., Martinez, K. L., Lill, M. A., Meyer, B. H., Vogel, H. & Hecht, B. (2005) ChemPhysChem 6, 1633–1640. 13. Cezanne, L., Lecat, S., Lagane, B., Millot, C., Vollmer, J. Y., Matthes, H., Galzi, J. L. & Lopez, A. (2004) J. Biol. Chem. 279, 45057–45067. 14. Heerklotz, H. (2002) Biophys. J. 83, 2693–2701. 15. Simons, K. & Toomre, D. (2000) Nat. Rev. Mol. Cell. Biol. 1, 31–39. 16. Edidin, M. (1997) Curr. Opin. Struct. Biol. 7, 528–532. 17. Simons, K. & Vaz, W. L. (2004) Annu. Rev. Biophys. Biomol. Struct. 33, 269–295. 18. Perez, J.-B. (2005) Ph.D. thesis (Ecole Polytechnique Fe´de´rale de Lausanne, Lausanne, Switzerland). 19. Kenworthy, A. K. & Edidin, M. (1998) J. Cell Biol. 142, 69–84. 20. Zacharias, D. A., Violin, J. D., Newton, A. C. & Tsien, R. Y. (2002) Science 296, 913–916. 21. Ayoub, M. A., Couturier, C., Lucas-Meunier, E., Angers, S., Fossier, P., Bouvier, M. & Jockers, R. (2002) J. Biol. Chem. 277, 21522–21528. 22. Mercier, J. F., Salahpour, A., Angers, S., Breit, A. & Bouvier, M. (2002) J. Biol. Chem. 277, 44925–44931.
Meyer et al.
FDA ⫽ A1 e ⫺k1cA ⫹ A2 e ⫺k2cA , FD
[11]
where cA is the reduced acceptor concentration (acceptor surface concentration normalized by R20); A1,2 and k1,2 are constants that only depend on the Rc兾R0 ratio and were determined by Wolber and Hudson from a fit of their calculations for various Rc兾R0 ratios. For a Rc兾R0 ratio of 1.1, they obtained A1 ⫽ 0.6327, k1 ⫽ 1.3686, A2 ⫽ 0.3673, k2 ⫽ 0.4654 (36). Eq. 11 only accurately describes the calculations of Wolber and Hudson in the range of 0 ⱕ cA ⱕ 0.5. When receptors are localized in domains of the membrane, their local reduced acceptor concentration is increased from cA ⫽ R兾Aplasma membrane to cA ⫽ R兾Adomain, where R is the total number of receptors and Aplasma membrane and Adomains are the total area of the plasma membrane and domains, respectively. To account for this increase, we scaled the reduced acceptor concentration by a factor ␣ ⫽ Aplasma membrane兾Adomains to obtain the local reduced acceptor density. By using Eq. 6 we can relate the approximation of Wolber and Hudson to Eapp,se as Eapp,se ⫽ 共1 ⫺ 共A1 e ⫺k1cA␣ ⫹ A2 e ⫺k2cA␣兲兲
xD , 1 ⫺ xD
[12]
where cA ⫽ R(1 ⫺ xD)R20 with R being the receptor surface concentration. Eq. 12 is strictly accurate only for the case of one single continuous domain. When the domain is divided in several microdomains, a deviation occurs at low cA because xD and cA in the microdomains become discrete variables with the result of Eapp,se vanishing at low cA. We thank Drs. S. Mark and I. Lu ¨scher (Ludwig Institute, Epalinges, Switzerland) for the kind gift of Cy5-DPPE and Dr. J.-B. Perez for fruitful discussions on FRET and GPCR dimerization. This work was supported by Swiss National Science Foundation Grant NRP 47, Project 4047-057572, TopNano21 program Projects 5636.2 and 6266.1, and internal grants from Ecole Polytechnique Fe´de´rale de Lausanne. 23. Herrick-Davis, K., Grinde, E. & Mazurkiewicz, J. E. (2004) Biochemistry 43, 13963–13971. 24. Keppler, A., Gendreizig, S., Gronemeyer, T., Pick, H., Vogel, H. & Johnsson, K. (2003) Nat. Biotechnol. 21, 86–89. 25. George, N., Pick, H., Vogel, H., Johnsson, N. & Johnsson, K. (2004) J. Am. Chem. Soc. 126, 8896–8897. 26. Guignet, E. G., Hovius, R. & Vogel, H. (2004) Nat. Biotechnol. 22, 440–444. 27. Miller, L. W. & Cornish, V. W. (2005) Curr. Opin. Chem. Biol. 9, 56–61. 28. Meyer, B. H. (2005) Ph.D. thesis (Ecole Polytechnique Fe´de´rale de Lausanne, Lausanne, Switzerland). 29. Pfeiffer, M., Kirscht, S., Stumm, R., Koch, T., Wu, D., Laugsch, M., Schroder, H., Hollt, V. & Schulz, S. (2003) J. Biol. Chem. 278, 51630–51637. 30. Monastyrskaya, K., Hostettler, A., Buergi, S. & Draeger, A. (2005) J. Biol. Chem. 280, 7135–7146. 31. Luber-Narod, J., Boyd, N. D. & Leeman, S. E. (1990) Eur. J. Pharmacol. 188, 185–191. 32. Zhang, X. L., Mak, J. C. & Barnes, P. J. (1995) Peptides 16, 867–872. 33. Maruyama, M. (1986) Brain Res. 370, 186–190. 34. Veatch, W. & Stryer, L. (1977) J. Mol. Biol. 113, 89–102. 35. Rocheville, M., Lange, D. C., Kumar, U., Sasi, R., Patel, R. C. & Patel, Y. C. (2000) J. Biol. Chem. 275, 7862–7869. 36. Wolber, P. K. & Hudson, B. S. (1979) Biophys. J. 28, 197–210. 37. Issafras, H., Angers, S., Bulenger, S., Blanpain, C., Parmentier, M., LabbeJullie, C., Bouvier, M. & Marullo, S. (2002) J. Biol. Chem. 277, 34666 – 34673. 38. Gomes, I., Jordan, B. A., Gupta, A., Trapaidze, N., Nagy, V. & Devi, L. A. (2000) J. Neurosci. 20, RC110. 39. Holst, B., Hastrup, H., Raffetseder, U., Martini, L. & Schwartz, T. W. (2001) J. Biol. Chem. 276, 19793–19799. 40. Vrljic, M., Nishimura, S. Y., Moerner, W. E. & McConnell, H. M. (2005) Biophys. J. 88, 334–347. 41. Hoppe, A., Christensen, K. & Swanson, J. A. (2002) Biophys. J. 83, 3652–3664. 42. Adair, B. D. & Engelman, D. M. (1994) Biochemistry 33, 5539–5544.
PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2143
BIOPHYSICS
n⫺1 Eapp,dq ⫽ E共1 ⫺ x D 兲,
