Journal home
Advance online publication
Current issue
Press releases
Guide to authors
Online submissionOnline submission
For referees
Free online issue
Contact the journal
For Advertisers
About this site
For librarians
Application notes
NPG Resources
Nature Biotechnology
Nature Protocols
Nature Genetics
Nature Chemical Biology
Nature Cell Biology
Nature Neuroscience
Nature Reviews Genetics
Nature Reviews Molecular Cell Biology
Nature Reviews Drug Discovery
Nature Conferences
NPG Subject areas
Clinical Medicine
Drug Discovery
Earth Sciences
Evolution & Ecology
Materials Science
Medical Research
Molecular Cell Biology
Browse all publications
Nature Methods - 3, 1001 - 1006 (2006)
Published online: 5 November 2006; | doi:10.1038/nmeth978

A rigorous experimental framework for detecting protein oligomerization using bioluminescence resonance energy transfer

John R James1, Marta I Oliveira2, 3, Alexandre M Carmo2, 3, Andrea Iaboni1 & Simon J Davis1

1 Nuffield Department of Clinical Medicine and Medical Research Council, Human Immunology Unit, Weatherall Institute of Molecular Medicine, The University of Oxford, Oxford Radcliffe Hospital, Oxford, OX3 9DU, UK.

2 Group of Cell Activation and Gene Expression, Instituto de Biologia Molecular e Celular, Universidade do Porto, Rua do Campo Alegre, 823, 4150-180, Porto, Portugal.

3 Instituto de Ciências Biomédicas de Abel Salazar, Universidade do Porto, Largo do Prof. Abel Salazar, 2, 4099-003, Porto, Portugal.

Correspondence should be addressed to Simon J Davis

Bioluminescence resonance energy transfer (BRET), which relies on nonradiative energy transfer between luciferase-coupled donors and GFP-coupled acceptors, is emerging as a useful tool for analyzing the quaternary structures of cell-surface molecules. Conventional BRET analyses are generally done at maximal expression levels and single acceptor/donor ratios. We show that under these conditions substantial energy transfer arises from random interactions within the membrane. The dependence of BRET efficiency on acceptor/donor ratio at fixed surface density, or expression level at a defined acceptor/donor ratio, can nevertheless be used to correctly distinguish between well-characterized monomeric and oligomeric proteins, including a very weak dimer. The pitfalls associated with the nonrigorous treatment of BRET data are illustrated for the case of G protein–coupled receptors (GPCRs) proposed to form homophilic and/or mixed oligomers on the basis of previous, conventional BRET experiments.

Please visit methagora to view and post comments on this article
BRET relies on nonradiative energy transfer between donor and acceptor fluorophores1. The bioluminescent protein, luciferase, oxidizes its substrate, coelenterazine, releasing photons. An appropriate fluorophore in close-enough physical proximity, typically <100 Å, can be excited to a higher energetic state, before emitting photons with longer wavelengths. An optimized version of this technology, the BRET2 assay, uses DeepBlueC, an analog of the natural substrate with maximal emission at 410 nm, and a UV-GFP variant (GFP2) that is excited at this wavelength and emits at 515 nm, giving a spectral separation of >100 nm. The key advantage of BRET over other forms of resonance energy transfer technology, such as fluorescence resonance energy transfer (FRET), is the very high signal/noise ratio gained from luminescence detection, which is unaffected by photobleaching or other optical effects, and allows protein interactions to be detected at physiological expression levels. Proteins of interest are expressed as 'BRET pairs' after being genetically fused to either Renilla luciferase (Luc) or 'GFP2' (GFP).

BRET efficiency (BRETeff), that is, the ratio of GFP emission to that of Luc emission, is dependent on the inverse sixth-power of the mean separation of donor and acceptor. For donors and acceptors forming constitutive oligomeric structures, BRETeff will therefore usually be high. A potential complication of all RET experiments is that background signals may arise from random interactions if donor and acceptor levels are sufficiently high. BRETeff in this case will generally be lower than that for oligomers as a result of the larger average separation of donors and acceptors. In principle, however, BRETeff maxima for bona fide oligomeric interactions may nevertheless be comparable to those arising from random interactions if the subunits are well separated and/or interact weakly, complicating discrimination between the two types of interactions.

Theoretical considerations2, 3, 4, 5 predict that, in addition to differences in BRETeff maxima that may or may not be obvious, the dependence of BRETeff on fluorophore concentrations will differ systematically for proteins interacting randomly versus those that form oligomers, in two types of experiments. For the first, or 'type-1', experiment, wherein the combined number of donors and acceptors is held constant, BRETeff for random interactions will be independent of acceptor/donor ratios above a certain threshold2, 3. This is because, at acceptor/donor ratios high enough to prevent competition between donors for acceptors, each donor will 'experience' the same acceptor environment (Fig. 1a; a fuller explanation of the theoretical predictions is available in Supplementary Note online). In contrast, BRETeff for oligomers is predicted4 to be highly dependent on the relative donor concentration because donors self-associate, reducing the overall efficiency of energy transfer (Fig. 1b,c). In this case, the acceptor/donor ratio (f) should affect BRETeff in a hyperbolic fashion that is dependent on the stoichiometry of the oligomer (n); reducing the relative concentration of donors will increase BRETeff to saturation (equation (1) is derived in Supplementary Methods online).

Figure 1. Effects of acceptor/donor ratio and surface density on BRETeff.
Figure 1 thumbnail

(a) BRETeff for random interactions is unaffected by donor concentration because as acceptor/donor ratio increases and total surface density remains essentially constant (n.c., no change), each donor continues to 'experience' a similar separation from the nearest acceptor. This independence breaks down if there are more donors than available acceptors, that is, when the acceptor concentration ceases to be effectively constant. (b,c) Molecules that can oligomerize are sensitive to changes in the acceptor/donor ratio because donors and acceptors compete to form multimeric complexes. Oligomerization reduces the effective surface density, reducing BRETeff from random interactions. (d) Increasing surface density at a given acceptor/donor ratio forces monomeric acceptor and donor molecules into closer proximity, increasing the likelihood of BRET owing to random interactions. At low surface densities, BRETeff will tend to zero. (e) Noncovalent oligomers are biased toward oligomer formation at higher surface densities but will also exhibit density-dependent BRET owing to random interactions. BRETeff should not decrease to zero at low densities, however, owing to the effect of affinity on complex formation. (f) Covalent or other constitutive oligomers are expected to produce BRETeff that is independent of surface density because folding and complex formation are generally density independent. However, at high surface densities the random interactions of oligomers will contribute to the overall signal. Fluorescing and nonfluorescing acceptor molecules are shown as yellow and open circles, respectively, and donors shown as blue circles. The BRET-permissible area surrounding donors is shown approximately to scale.

Full FigureFull Figure and legend (48K)

For the second, or 'type-2', experiment, wherein expression level is varied and acceptor/donor ratio kept constant, BRETeff for random or very weak interactions will be pseudo-linearly dependent on expression and may fall to zero at very low expression levels (Fig. 1d,e). Conversely, for oligomers BRETeff is expected to be expression level–independent because, generally, expression will itself be dependent on the correct assembly of the oligomer, making this type of experiment especially useful for identifying obligate oligomers (Fig. 1f). We test these predictions for a series of monomeric and oligomeric proteins of known stoichiometry and then consider the case of class-A GPCRs proposed to form oligomeric structures in previous BRET experiments6.

Conventional BRET experiments
The monomeric, noninteracting type-I membrane proteins CD2 and CD86 mediate cell adhesion and signaling in the immune system7, 8, 9 (Fig. 2a). When we expressed CD2 and CD86 separately as BRET pairs (by cotransfecting genes encoding CD2Luc and CD2GFP or CD86Luc and CD86GFP), or as a single BRET pair (by cotransfecting CD2Luc and CD86GFP), in 'conventional' BRET experiments in HEK-293T cells, that is, involving maximal expression at a single ratio of donors and acceptors, BRETeff was approximately one-fifth that for a control protein consisting of a soluble, fused form of Luc and GFP (sGFP-Luc; Fig. 2b). Fluorescence-activated cell sorting (FACS) analysis and confocal microscopy confirmed that energy transfer resulted almost exclusively from interactions at the cell surface (Fig. 2c,d) rather than intracellular aggregation or other artifacts resulting from overexpression. This indicated that, for type-I membrane proteins, substantial energy transfer results solely from random, nonspecific protein interactions. BRETeff for a type-I membrane glycoprotein that dimerizes in crystal lattices and in solution, CD80 (refs. 10,11), was almost twice that of the monomers (P < 0.001), but considerably less than that of sGFP-Luc (Fig. 2b).

Figure 2. Monomeric proteins expressed at the cell surface give substantial resonance energy transfer in conventional BRET experiments.
Figure 2 thumbnail

(a) Schematic of the architecture of proteins used in this study, approximately to scale. Scale bar, 10 nm. (b) BRETeff for the indicated proteins expressed as BRET pairs, normalized against the value obtained for sGFP-Luc. "BP" denotes single proteins coexpressed as both Luc and GFP fusions. Error bars represent mean plusminus s.d., n = 8. (c) FACS analysis of CD86GFP-transfected HEK 293T cells after 24 h, stained with anti–CD86-PE (Serotec). (d) Confocal microscopy–based analysis of CD86GFP (left) and beta2ARGFP (right) expression, demonstrating GFP localization to the cell surface.

Full FigureFull Figure and legend (33K)
Type-1 experiments
We determined whether these BRETeff differences complied with the theoretical predictions for the two classes of interactions by examining the dependence of BRETeff on acceptor/donor ratio (the type-1 experiment). Twenty-four hours after transfection with equivalent amounts of total DNA comprised of various ratios of the GFP- or Luc-tagged fusion protein–expressing vectors, 293T cells expressed physiological levels of each of the cell surface proteins (that is, 50,000–500,000 copies/cell12; Fig. 3a). Comparison of light emission in the 370–450 nm and 500–530 nm ranges, measured for cells expressing the GFP- and Luc-tagged fusion proteins, with that obtained for cells expressing sGFP-Luc, allowed BRETeff to be plotted against the ratio of GFP and Luc concentrations ([GFP]/[Luc]). This was important as it permitted stoichiometric information to be derived, and rendered the method instrument-independent, facilitating interexperimental comparisons. Others have shown that fluorescence levels are largely unaffected by the precise structural context of the N- or C-terminal fusions13, which we confirmed by showing that, for all constructs tested, equivalent levels of cell-surface expression gave exactly the same GFP fluorescence (Fig. 3a). We also failed to see any correlation between cytoplasmic tail length and BRETeff for cytoplasmic domains less than approx150 residues (data not shown). Others have reported construct-specific variation in fluorophore intensity14 but this may have been due to differential protein expression levels not directly assayed in their experiments.

Figure 3. Type-1 experiments: varying the acceptor/donor ratio distinguishes between BRET arising from random versus oligomeric interactions.
Figure 3 thumbnail

(a) FACS analysis using PE-conjugated antibodies to the specified antigens and QuantiBRITE beads, demonstrating that GFP expression correlates with equivalent cell-surface staining. (b) Solid lines show fits of data obtained for the indicated BRET pairs to equation (1), that is, for dimers, and the dotted lines show the fit (for [GFP]/[Luc] > 2) to a constant value, as predicted for random interactions. Only the model that gave the best fit is shown. (c) Residual BRETeff values after nonlinear least-squares fitting to the indicated model, plotted as a moving average of the data. A good fit reduces residuals to zero. (d) Cells expressing CD80 or CD86 as a BRET pair were incubated with phosphate-buffered saline (– CTLA-4Fc) or 50 mug/ml CTLA-4Fc (+ CTLA-4Fc) before assaying for BRET.

Full FigureFull Figure and legend (47K)
BRETeff values obtained for CD2 and for CD86 expressed as BRET pairs, and for the two proteins coexpressed as a single BRET pair, exhibited similar maxima (approx0.2) and were independent of the [GFP]/[Luc] ratio beyond a value of 2 (Fig. 3b): fits of the data to a constant beyond this [GFP]/[Luc] ratio were better than hyperbolic fits (equation 1), fully in accord with the theoretical predictions (Fig. 3c; fits to all the data, along with the goodness of each fit represented numerically as the root mean square of the residuals, are available in Supplementary Fig. 1 online and Supplementary Table 1 online, respectively). Other noninteracting proteins expressed as BRET pairs, that is, ALCAM and PD-L1, gave equivalent data (Fig. 3b). Notably, all the monomers gave the same BRETeff maxima (approx0.2), implying that this is the threshold above which all oligomeric type-I membrane proteins may be identifiable. Conversely, the covalent type-I homodimers CD28 and CTLA-4 (Fig. 2a) yielded data that gave very good fits to equation 1 (n = 2), with BRETeff approaching the unitary value assigned to sGFP-Luc (Fig. 3b). As in the conventional experiment (Fig. 2b), BRETeff for CD80 expressed as a BRET pair (Fig. 3b) was higher than that for the monomers and gave a better fit to equation 1 than did BRETeff for CD86 (Fig. 3c). The asymptote is substantially smaller than that for CD28 or CTLA-4, however, consistent with the coexistence of CD80 dimers and monomers in dynamic equilibrium. The observation that BRETeff is enhanced by coincubation with soluble, bivalent ligand (CTLA-4Fc; Fig. 3d), which would be expected to stabilize CD80 dimers11, supported this interpretation. A chimeric protein consisting of the CD80 extracellular domain and CD86 transmembrane and cytoplasmic domains yielded similar data, implying that, at the cell surface, CD80 dimerizes via its extracellular domain, presumably in the manner observed in crystals of soluble CD80 (refs. 10,11 and data not shown). The affinity of soluble CD80 self-association is very low (50 muM)10, indicating that type-1 experiments readily identify very weak homophilic interactions. The weak self-association of CD80 has recently been confirmed in FRET-based analyses15.

Type-2 experiments
It is not always feasible to vary the [GFP]/[Luc] ratio while keeping the combined number of donors and acceptors constant, in which case the expression level can be varied and the acceptor/donor ratio kept constant5 (type-2 experiment). We transfected cells with constant ratios of donor- and acceptor-encoding vectors and measured BRETeff at successive time points, that is, at increasing expression levels. BRETeff levels for CD86 and CD2 exhibited a strict dependence on surface density, whereas for the covalent dimer, CTLA-4, it was largely expression-level independent (Fig. 4). The slight increase in BRETeff observed in the case of CTLA-4 can be attributed to random interactions between homodimers and is further evidence that such interactions contribute substantially to BRETeff, even though it is often assumed that transfer between oligomers can be neglected. For CD80, BRETeff was expression-level dependent, but did not fall to zero at low expression levels, as expected for nonconstitutive homodimers (Fig. 1e). The evidence that CD80 exhibited behavior distinct from that of randomly interacting proteins, although statistically significant (P < 0.001), was less compelling for this type of analysis than for the type-1 experiment (Fig. 3b,c).

Figure 4. Type-2 experiments: BRET signals arising from random interactions of molecules are linearly related to surface density at a given acceptor/donor ratio.
Figure 4 thumbnail

The expression level was calculated from Luc emission and fitted to linear equations for all molecules. Note that for the oligomers, CTLA-4 and CD80, the intercept of the ordinate is nonzero.

Full FigureFull Figure and legend (19K)
Re-examination of class-A GPCR dimerization
Without exception, the authors of more than forty previous BRET-based studies of membrane protein oligomerization concluded that the proteins they were studying form dimers or higher-order structures6. The majority of studies addressed the quaternary structures of GPCRs, the largest family of cell-surface signaling proteins encoded by the mammalian genome. As the studies for the most part used conventional BRET experiments, that is assays in which donor- and acceptor-encoding plasmids were used at a single ratio16, 17, we examined whether a more rigorous analysis of BRET data supported these conclusions.

We chose the class-A GPCR most studied in BRET experiments, human beta2-adrenergic receptor (beta2AR). We used a second class-A GPCR, mouse cannabinoid receptor 2 (mCannR2) and a heterodimerizing class-C GPCR, the neuronal gamma-amino-n-butyric acid type beta receptor (GABAbetaR) for control purposes (Fig. 2a). In experiments in which we systematically varied the [GFP]/[Luc] ratio but kept the overall expression levels constant, BRETeff for beta2AR and mCannR2 expressed as BRET pairs were of the same low maximum values reported elsewhere (approx0.1; refs. 14,16), and exhibited [GFP]/[Luc] ratio independence beyond a value of approx2 (Fig. 5a). When beta2ARLuc and mCannR2GFP were coexpressed as a BRET pair, these two functionally unrelated proteins yielded the same low BRETeff values exhibiting [GFP]/[Luc] ratio independence (Fig. 5a).

Figure 5. Two native class-A GPCRs are monomeric at the cell surface.
Figure 5 thumbnail

(a) Both beta2AR and mCannR2, expressed separately as BRET pairs or together as a BRET pair, and beta2AR and CD2 expressed as a BRET pair, give similar low BRETeff values exhibiting [GFP]/[Luc] ratio independence beyond a [GFP]/[Luc] ratio of 2, with lines representing the best fit to the data, as described in the Figure 3 legend. (b) Variation of the acceptor/donor ratio for beta2AR shows no evidence for oligomerization at three different levels of surface expression, with the highest level being approx2 times 106 molecules/cell. (c) BRETeff values for a heterodimerizing class-C GPCR, GABAbetaR, consisting of GABAbetaR1 and GABAbetaR2 subunits, exhibit the hyperbolic relationship with acceptor/donor ratio expected for a dimer, as do BRETeff values obtained for a fusion protein consisting of beta2AR and the heterodimerizing coiled-coil domains of the GABAbetaR complex (beta2ARcoil1 and beta2ARcoil2). beta2ARcoil2Delta is a truncated version of beta2ARcoil2 with only half the GABAbetaR2 coiled-coil. (d) The decrease in interaction when beta2AR is expressed as a BRET pair with GABAbetaR2 is due to the homodimerization of GABAbetaR2 at the cell surface, which decreases by half the effective density of molecules present.

Full FigureFull Figure and legend (41K)
The BRETeff maxima observed in these experiments were substantially lower than those obtained for CD2 or CD86. The likely explanation for this is that the larger hydrodynamic diameter of GPCR proteins (>50 Å versus 30 Å)18 increases the distance of closest approach, reducing the maximum BRETeff obtainable via random, or any other, interaction. For type-I membrane proteins the limiting factor is very likely to be the dimensions of the Luc and GFP fluorophores. Supporting this possibility, coexpression of beta2ARLuc with CD2GFP as a BRET pair gave a BRETeff maximum slightly higher than that obtained for beta2AR expressed as a BRET pair (Fig. 5a). We tested the possibility that we had not expressed enough GPCR, that is, that we were sampling a point in the equilibrium dominated by monomers, by increasing expression. At GPCR levels as high as 2 times 106 molecules/cell, that is, double that used by others14, 17, the same low BRETeff values and [GFP]/[Luc] ratio independence seen at lower expression levels were obtained (Fig. 5b). In the type-2 experiment, in which the [GFP]/[Luc] ratio was held constant and overall expression level varied, BRETeff for beta2AR expressed as a BRET pair exhibited the same strict dependence on expression level, with BRETeff falling to zero at low surface densities (P < 0.01), as did BRETeff values obtained under the same conditions for CD2 and CD86 expressed as BRET pairs (Fig. 4).

Dimerization of a beta2AR/GABAbetaR chimera
We also sought confirmation that class-A GPCR dimerization, had it occurred, would have been detectable using the new approaches, by generating beta2AR and GABAbetaR chimeras. Heterodimerization of the GABAbetaR1 and GABAbetaR2 subunits is driven, at least in part, by the interactions of cytoplasmic sequences present in each subunit that together form 'coiled-coil' domains19, 20 (Fig. 2a). Coexpression of GABAbetaR1Luc and GABAbetaR2GFP as a BRET pair yielded BRETeff data exhibiting the hyperbolic relationship expected for dimers in addition to a BRETeff maximum more than double those obtained for beta2AR or mCannR2 (Fig. 5c). Having established that GABAbetaR heterodimerization was readily detectable, we genetically fused the coiled-coil domains of GABAbetaR1 (coil1) and GABAbetaR2 (coil2), along with Luc and GFP, to the seven transmembrane domain–containing region of beta2AR, giving beta2ARcoil1Luc and beta2ARcoil2GFP. BRETeff values obtained by coexpression of beta2ARcoil1Luc and beta2ARcoil2GFP as a BRET pair were almost indistinguishable from those obtained for coexpression of GABAbetaR1Luc and GABAbetaR2GFP (Fig. 5c). A second pair of constructs coexpressed as a BRET pair, beta2ARcoil2DeltaLuc and beta2ARcoil2DeltaGFP, for which the coiled-coil domain of GABAbetaR2 was truncated by 50% and therefore likely to interact more weakly, gave BRETeff values intermediate between those obtained for the native forms of beta2AR and GABAbetaR (Fig. 5c), consistent with BRETeff being sensitive to the affinity of oligomerization.

GABAbetaR2 is a poor specificity control for GPCR interactions
In previous studies of GPCR interactions, coexpression with an 'irrelevant' GPCR21, 22, 23, 24 has been used as a control for nonspecific interactions. Notably, the same irrelevant GPCR, the GABAbetaR2 subunit of GABAbetaR, was used on each occasion. We replicated these experiments by coexpressing beta2ARLuc and GABAbetaR2GFP as a BRET pair, and found substantial (50%) decreases in BRETeff compared to beta2AR expressed alone as a BRET pair (Fig. 5d). Given that our irrelevant control for GPCR coexpression experiments, mCannR2, gave data indistinguishable from those obtained for beta2AR alone, the question arose as to whether GABAbetaR2 constituted an appropriate control for homodimerization. Expression of the GABAbetaR2 subunit alone as a BRET pair gave BRETeff values that were substantially higher than those obtained for beta2AR and exhibited a hyperbolic relationship with [GFP]/[Luc] ratio (Fig. 5d). This indicated that the GABAbetaR2 subunit homodimerized as readily as it formed heterodimers with the GABAbetaR1 subunit, as described elsewhere25 (Fig. 5c). The significance of this is that GABAbetaR2 homodimerization will halve the effective number of molecules available for energy transfer via random interactions with the test protein, reducing BRETeff. Coexpression of CD2Luc and CTLA-4GFP as a BRET pair reproduced this effect (data not shown). These considerations suggest that GABAbetaR2 is an inappropriate control for identifying GPCR dimers.

The use of BRET to discriminate between monomeric and oligomeric cell-surface proteins is not necessarily straightforward. This is because the interpretation of BRET data, as in the case of other resonance energy transfer experiments, is complicated by the dependence of energy transfer on several factors, including the distance of nearest approach (that is the cross-sectional area of donor and acceptor chimeras), expression level, subunit affinity and stoichiometry. An additional complication for analyses of cell-surface proteins is that random interactions contribute a larger fraction of the total energy transfer than seems previously to have been appreciated. Unless BRETeff is very high, the strategy implemented in conventional BRET experiments, that is, single estimates of energy transfer derived at maximum expression levels, is of limited use.

The prediction that the BRET 'signature' for randomly interacting proteins in type-1 experiments consists of the independence of BRETeff and acceptor/donor ratio, whereas, for type-2 experiments, it comprises the strict expression-level dependence of BRETeff, greatly assists in the interpretation of the data. The independence of BRETeff and acceptor/donor ratio for randomly interacting proteins in type-1 experiments will only be apparent when the [GFP]/[Luc] ratio becomes sufficiently high that the donors are no longer competing for acceptors at the cell surface and acceptor concentration is essentially constant. The threshold for this effect appears to be an acceptor/donor ratio of approx2, as also determined empirically for FRET data5. For CD80, the case for dimerization was strengthened by control experiments with class-matched, bona fide monomers. The pitfalls associated with choosing inappropriate 'specificity' controls, however, was illustrated for beta2AR and GABAbetaR2 coexpression. In assigning a particular stoichiometry to a given protein, consideration should be given to (i) the BRETeff maximum, (ii) the relationship of BRETeff with acceptor/donor ratio and overall expression level in type-1 and type-2 experiments, respectively, (iii) comparisons with suitable controls and, ideally, (iv) the effects of forced changes in native stoichiometry induced by, for example, mutagenesis.

According to these criteria, class-A GPCRs exemplified monomeric behavior, which challenges the notion6 that these GPCRs are innately predisposed to forming homo- or hetero-oligomers. When this does occur, as in the case of the class-C GABAbetaR heterodimer, GPCR dimerization could be readily detected by varying acceptor/donor ratio at constant expression. The possibility that random interactions of class-A GPCRs were responsible for the energy transfer observed previously has been addressed quantitatively only once, for beta2AR, and ruled out14. On that occasion, as in each of the more quantitative BRET analyses published, rather than increasing acceptor/donor ratio by decreasing the amount of donor, the acceptor concentration was varied and donor level kept constant. In such experiments the acceptor/donor ratio and total surface density vary simultaneously, making it very difficult to observe the type-1 or type-2 signatures of randomly interacting proteins. We were also unable to reproduce the type-2 observation14 that BRETeff for beta2AR is independent of expression level at a constant acceptor/donor ratio, finding instead that it was entirely expression-level dependent. A second group17 also obtained expression level–dependent BRETeff data for beta2AR and three other class-A GPCRs, but overlooked the possibility that this constituted clear-cut evidence for energy transfer resulting from random interactions only.

Our results indicate that nonradiative energy transfer between monomers can reach a level previously assigned only to the formation of oligomeric structures. Given that proteins occupy >25% of the cell surface by area26, this degree of nonspecific energy transfer should probably not have been surprising. As there is no detectable energy transfer between GFP tethered to the membrane and luciferase expressed in the cytoplasm (J.R.J. and S.J.D.; unpublished data), it seems that the cell surface constitutes a microenvironment greatly favoring the cis interactions of proteins, such as those required for 'signalosome' formation27. Our results also imply that, over time, in addition to the assembly of nascent oligomeric complexes, resonance energy transfer will be sensitive to local changes in protein density. This offers new opportunities for following the reorganization of both interacting and noninteracting cell-surface molecules in the course of, for example, receptor triggering28 and immunological synapse formation29. Rigorous quantitative analysis of the data will be required to tease out these effects.

BRET assay.
We collected cells from wells 24 h post-transfection using phosphate-buffered saline, pelleted them at 600g for 3 min and resuspended them at approx1.5 times 106 cells/ml in minimal essential medium. For each transfection, we added 10 muM DeepBlueC (final concentration; PerkinElmer) to 100 mul of cells in a 96-well OptiPlate (PerkinElmer) and collected light emission in the 410 plusminus 40 nm (BRET-A) and 515 plusminus 15 nm (BRET-B) wavelength ranges three times for each range, integrated over 1 s on a Fusion microplate analyzer (PerkinElmer). To determine GFP and Luc expression, we dispensed 100 mul of cells in a separate well, excited them at 420 nm and measured emission at 515 plusminus 15 nm three times over 1 s, to obtain the total fluorescence units (FU). We then incubated the cells in the same well with 10 muM coelenterazine-h for 2 min before reading total emission three times integrated over 1 s, to obtain the total luminescence units (LU). We calculated BRET values, after background subtraction, as BRET-B/BRET-A, corrected for luciferase expression alone (typically 7% of BRET-A luminescence). As the concentration of tagged molecules is proportional to the signal detected (Fig. 3a), the acceptor/donor ratio can be calculated as [GFP]/[Luc] = (kGFU)/(kLLU) = K(FU/LU), where kG and kL are constants specific to GFP and Luc, respectively, and K = kG / kL. The FU/LU value calculated for the sGFP-Luc expression in each experiment gives 1/K because the acceptor/donor ratio is fixed at one. We used this derived value of K to convert the measured acceptor/donor ratio (FU/LU) to one of concentrations ([GFP]/[Luc]), before plotting against the BRETeff at that value.

For the type-2 BRET analysis, we transfected the cells with a particular acceptor/donor ratio (12:1), and then collected samples at regular intervals so that the expression level could be varied systematically, before being assessed as above for BRET. We determined the level of protein expression using total Luc expression as this gave the most reliable measure of low-level expression.

Statistical analysis.
We used the two-tailed Student's t-test, where appropriate, to assess the difference between data sets. We used the nonlinear least-squares fitting function of Origin 7.5 (Originlab) to fit functions to data and estimate parameters, as described in Supplementary Methods.

Additional methods.
Description of DNA constructs, cell culture and transfection, confocal microscopy and FACS analysis are available in Supplementary Methods.

Note: Supplementary information is available on the Nature Methods website.

Author contributions
J.R.J., M.I.O. and A.I. executed the experiments; J.R.J., A.M.C. and S.J.D. formulated the experiments and wrote the manuscript.

Received 31 July 2006; Accepted 17 October 2006; Published online: 5 November 2006.

  1. Xu, Y., Piston, D.W. & Johnson, C.H. A bioluminescence resonance energy transfer (BRET) system: application to interacting circadian clock proteins. Proc. Natl. Acad. Sci. USA 96, 151–156 (1999). | Article | PubMed | ChemPort |
  2. Wolber, P.K. & Hudson, B.S. An analytic solution to the Forster energy transfer problem in two dimensions. Biophys. J. 28, 197–210 (1979). | PubMed | ISI | ChemPort |
  3. Fung, B.K. & Stryer, L. Surface density determination in membranes by fluorescence energy transfer. Biochemistry 17, 5241–5248 (1978). | Article | PubMed | ISI | ChemPort |
  4. Veatch, W. & Stryer, L. The dimeric nature of the gramicidin A transmembrane channel: conductance and fluorescence energy transfer studies of hybrid channels. J. Mol. Biol. 113, 89–102 (1977). | Article | PubMed | ISI | ChemPort |
  5. Kenworthy, A.K. & Edidin, M. Distribution of a glycosylphosphatidylinositol-anchored protein at the apical surface of MDCK cells examined at a resolution of <100 A using imaging fluorescence resonance energy transfer. J. Cell Biol. 142, 69–84 (1998). | Article | PubMed | ISI | ChemPort |
  6. Pfleger, K.D. & Eidne, K.A. Monitoring the formation of dynamic G-protein-coupled receptor-protein complexes in living cells. Biochem. J. 385, 625–637 (2005). | PubMed | ISI | ChemPort |
  7. Jones, E.Y., Davis, S.J., Williams, A.F., Harlos, K. & Stuart, D.I. Crystal structure at 2.8 A resolution of a soluble form of the cell adhesion molecule CD2. Nature 360, 232–239 (1992). | Article | PubMed | ISI | ChemPort |
  8. Bodian, D.L., Jones, E.Y., Harlos, K., Stuart, D.I. & Davis, S.J. Crystal structure of the extracellular region of the human cell adhesion molecule CD2 at 2.5 A resolution. Structure 2, 755–766 (1994). | Article | PubMed | ISI | ChemPort |
  9. Collins, A.V. et al. The interaction properties of costimulatory molecules revisited. Immunity 17, 201–210 (2002). | Article | PubMed | ISI | ChemPort |
  10. Ikemizu, S. et al. Structure and dimerization of a soluble form of B7–1. Immunity 12, 51–60 (2000). | Article | PubMed | ISI | ChemPort |
  11. Stamper, C.C. et al. Crystal structure of the B7–1/CTLA-4 complex that inhibits human immune responses. Nature 410, 608–611 (2001). | Article | PubMed | ISI | ChemPort |
  12. Barclay, A.N. et al. The Leucocyte Antigen Factsbook (Academic Press, San Diego, 1997).
  13. Shaner, N.C. et al. Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein. Nat. Biotechnol. 22, 1567–1572 (2004). | Article | PubMed | ISI | ChemPort |
  14. Mercier, J.F., Salahpour, A., Angers, S., Breit, A. & Bouvier, M. Quantitative assessment of beta 1- and beta 2-adrenergic receptor homo- and heterodimerization by bioluminescence resonance energy transfer. J. Biol. Chem. 277, 44925–44931 (2002). | Article | PubMed | ISI | ChemPort |
  15. Bhatia, S., Edidin, M., Almo, S.C. & Nathenson, S.G. Different cell surface oligomeric states of B7–1 and B7–2: implications for signaling. Proc. Natl. Acad. Sci. USA 102, 15569–15574 (2005). | Article | PubMed | ChemPort |
  16. Angers, S. et al. Detection of beta 2-adrenergic receptor dimerization in living cells using bioluminescence resonance energy transfer (BRET). Proc. Natl. Acad. Sci. USA 97, 3684–3689 (2000). | Article | PubMed | ChemPort |
  17. Ramsay, D., Kellett, E., McVey, M., Rees, S. & Milligan, G. Homo- and hetero-oligomeric interactions between G-protein-coupled receptors in living cells monitored by two variants of bioluminescence resonance energy transfer (BRET): hetero-oligomers between receptor subtypes form more efficiently than between less closely related sequences. Biochem. J. 365, 429–440 (2002). | Article | PubMed | ISI | ChemPort |
  18. Chen, Y., Couvineau, A., Laburthe, M. & Amiranoff, B. Solubilization and molecular characterization of active galanin receptors from rat brain. Biochemistry 31, 2415–2422 (1992). | Article | PubMed | ChemPort |
  19. White, J.H. et al. Heterodimerization is required for the formation of a functional GABA(B) receptor. Nature 396, 679–682 (1998). | Article | PubMed | ISI | ChemPort |
  20. Jones, K.A. et al. GABA(B) receptors function as a heteromeric assembly of the subunits GABA(B)R1 and GABA(B)R2. Nature 396, 674–679 (1998). | Article | PubMed | ISI | ChemPort |
  21. Canals, M. et al. Homodimerization of adenosine A2A receptors: qualitative and quantitative assessment by fluorescence and bioluminescence energy transfer. J. Neurochem. 88, 726–734 (2004). | Article | PubMed | ChemPort |
  22. Canals, M. et al. Adenosine A2A-dopamine D2 receptor-receptor heteromerization: qualitative and quantitative assessment by fluorescence and bioluminescence energy transfer. J. Biol. Chem. 278, 46741–46749 (2003). | Article | PubMed | ISI | ChemPort |
  23. Percherancier, Y. et al. Bioluminescence resonance energy transfer reveals ligand-induced conformational changes in CXCR4 homo- and heterodimers. J. Biol. Chem. 280, 9895–9903 (2005). | Article | PubMed | ChemPort |
  24. Breit, A., Lagace, M. & Bouvier, M. Hetero-oligomerization between beta2- and beta3-adrenergic receptors generates a beta-adrenergic signaling unit with distinct functional properties. J. Biol. Chem. 279, 28756–28765 (2004). | Article | PubMed | ISI | ChemPort |
  25. Uezono, Y. et al. Coupling of GABAB receptor GABAB2 subunit to G proteins: evidence from Xenopus oocyte and baby hamster kidney cell expression system. Am. J. Physiol. Cell Physiol. 290, C200–C207 (2006). | PubMed | ChemPort |
  26. Saxton, M.J. Lateral diffusion in an archipelago. Distance dependence of the diffusion coefficient. Biophys. J. 56, 615–622 (1989). | PubMed | ChemPort |
  27. Jun, J.E. & Goodnow, C.C. Scaffolding of antigen receptors for immunogenic versus tolerogenic signaling. Nat. Immunol. 4, 1057–1064 (2003). | Article | PubMed | ISI | ChemPort |
  28. Davis, S.J. & van der Merwe, P.A. The structure and ligand interactions of CD2: implications for T-cell function. Immunol. Today 17, 177–187 (1996). | Article | PubMed | ISI | ChemPort |
  29. Davis, D.M. & Dustin, M.L. What is the importance of the immunological synapse? Trends Immunol. 25, 323–327 (2004). | Article | PubMed | ChemPort |
We thank A. Wise (GlaxoSmithKline) for the gift of the GABAbetaR2 template, and E. Evans (Nuffield Dept. Clinical Medicine, Oxford University) and J. McIlhinney (MRC Anatomical Neuropharmacology Unit, Oxford University) for helpful discussion. This work was supported by the Wellcome Trust, the Rhodes Trust and the Programa Operacional Ciência e Inovação 2010, cofunded by the European Regional Development Fund.

Competing interests statement:  The authors declare that they have no competing financial interests.


Natureproducts is an online service detailing information about specific products used in this article, you can view the product descriptions, request information and compare with other similar products. The products used are listed in alphabetical order.

A-Z product listingbiocompare
96-well OptiPlate (PerkinElmer)
anti–CD86-PE (Serotec)
DeepBlueC (PerkinElmer)
Fusion microplate analyzer (PerkinElmer)
Origin 7.5 (Originlab)
See more natureproducts
Previous | Next
Table of contents
Download PDFDownload PDF
Send to a friendSend to a friend
rights and permissionsRights and permissions
CrossRef lists 115 articles citing this articleCrossRef lists 115 articles citing this article
Save this linkSave this link
More articles like this
Figures & Tables
Supplementary info
See also: News and Views by Lohse
Export citation
Export references



Search buyers guide:


Nature Methods
ISSN: 1548-7091
EISSN: 1548-7105
Journal home | Current issue | Archive | Press releases |
Nature Publishing Group, publisher of Nature, and other science journals and reference works©2006 Nature Publishing Group | Privacy policy