Green bioluminescence in Renilla species is generated by a ∼100% efficient RET (resonance energy transfer) process that is caused by the direct association of a blue-emitting luciferase [Rluc (Renilla luciferase)] and an RGFP (Renilla green fluorescent protein). Despite the high efficiency, such a system has never been evaluated as a potential reporter of protein–protein interactions. To address the question, we compared and analysed in mammalian cells the bioluminescence of Rluc and RGFP co-expressed as free native proteins, or as fused single-chain polypeptides and tethered partners of self-assembling coiled coils. Here, we show that: (i) no spontaneous interactions generating detectable BRET (bioluminescence RET) signals occur between the free native proteins; (ii) high-efficiency BRET similar to that observed in Renilla occurs in both fusion proteins and self-interacting chimaeras, but only if the N-terminal of RGFP is free; (iii) the high-efficiency BRET interaction is associated with a dramatic increase in light output when the luminescent reaction is triggered by low-quantum yield coelenterazine analogues. Here, we propose a new functional complementation assay based on the detection of the high-efficiency BRET signal that is generated when the reporters Rluc and RGFP are brought into close proximity by a pair of interacting proteins to which they are linked. To demonstrate its performance, we implemented the assay to measure the interaction between GPCRs (G-protein-coupled receptors) and β-arrestins. We show that complementation-induced BRET allows detection of the GPCR–β-arrestin interaction in a simple luminometric assay with high signal-to-noise ratio, good dynamic range and rapid response.
RET (resonance energy transfer)  between donor–acceptor pairs of chromophores linked to different proteins or protein domains has been extensively used to investigate nanoscale molecular interactions in protein biochemistry [2,3]. The progress in genetically encoded fluorescent protein, such as GFP (green fluorescent protein) and its spectral variants , has made FRET (fluorescence RET) a very attractive tool also in cell biology. Although the intrinsic structural limitations of GFP-like probes make rigorous measurements of distances impracticable, genetically encoded reporters allows us to investigate, at least in a qualitative fashion, the network of macromolecular interactions occurring in the living cell [5–7].
BRET (bioluminescence RET) is a recently introduced variation that exploits energy transfer occurring between a luciferase-bound donor and a compatible fluorescent protein acceptor [8–11]. BRET has two advantages over conventional FRET. First, it does not require incident light, which can be damaging to living cells under several conditions. Secondly, it may lead to assays with better signal-to-noise ratio, because endogenous luminescence is far lower than autofluorescence in mammalian cells. In practice, however, the benefits of the lower background are offset by the small efficiency of the RET process which is usually observed in such assays, even when donors and acceptors are fused in tandem within a single polypeptide chain .
Two types of BRET assays have been described so far. Both utilize the reaction of Renilla reniformis luciferase [Rluc (Renilla luciferase)] with CLZ (coelenterazine) as light emitter, and mutants of Aequorea victoria GFP as acceptors of energy transfer. One type uses the yellow-shifted mutant YFP (yellow fluorescent protein) (λmax=530 nm) , which provides a better spectral separation of the RET emission peak from that of native CLZ (λmax=475 nm). In the second type, the acceptor is a GFP mutant with a blue-shifted absorption spectrum . If used in conjunction with CLZ400A (didehydroxy CLZ; a blue-shifted and weakly luminescent didehydroxylated analogue), this assay can achieve a similar improvement of spectral resolution but at the expenses of dramatic reductions in quantum yield. In both cases, however, only a small fraction of photons are emitted via RET, even under conditions where covalent bonding ensure irreversible proximity between Rluc and the fluorescent acceptor [13,14].
Unlike artificially created probes, BRET can reach close to 100% efficiency  in some natural macromolecular systems [16–18]. For example, in the marine anthozoan Renilla, green luminescence is generated by a high-efficiency RET process, which effectively converts most of the blue photoemission of excited CLZ  into green light . This efficiency is the result of a specific protein–protein interaction between Rluc and an RGFP (Renilla GFP) . Presumably, the binding of the two proteins can optimize the spatial relationship between the chromophores, thus maximizing the transfer of energy via resonance. An additional important feature of this system, although still requiring clarification, is the apparent enhancement of quantum yield of the RET signal. Depending on the CLZ analogue used, the photon emission per mole of substrate is enhanced 5–100-fold in the presence of the GFP acceptor .
On the one hand, such features suggest that Rluc and RGFP would constitute an ideal reporter system for the investigation of protein–protein interactions. On the other, however, the tendency to spontaneous association might pose an impossible limitation, as it could allow interaction of proteins that would normally not do so. The importance of such a limitation is crucially dependent on the extent to which the ‘background’ affinity of the reporter system differs from that of the ‘reported’ proteins under study. Earlier measurements [21,23] described a significant affinity between Rluc and RGFP, with an apparent Kd located in the micromolar range. However, such determinations required conditions of virtually no ionic strength in the reaction, since the RET signal was readily disrupted at relatively low salt concentrations [21,23]. This suggests that the true affinity between Rluc and RGFP under physiological ionic strength might be far lower than the value estimated in those determinations.
In the present study, we have investigated whether the ‘natural’ BRET between the two Renilla proteins can be used as a reporter system. To evaluate the interaction between Rluc and RGFP under conditions that are as close as possible to the reaction occurring between genetically encoded tags, we engineered a series of specifically designed chimaeric proteins, and recorded the spectral properties of the luminescence both in cell extracts and intact cells.
Here, we show that no detectable spontaneous interaction occurs between Rluc and RGFP when they are co-expressed as individual proteins in cells. In contrast, both the highly efficient RET process and the characteristic enhancement in photon yield are readily observed when the two proteins are covalently linked into a single polypeptide chain, or are driven into proximity by a complementary pair of leucine-zipper peptides tethered to their sequences. We also show that the Renilla proteins can be used as an excellent reporter system for the study of the interaction between GPCRs (G-protein-coupled receptors) and β-arrestin in living cells.
Cell culture media, G418, hygromycin and FCS (fetal calf serum) were obtained from Invitrogen. All ligands and DEAE-dextran sulfate (500 kDa average mass) were from Sigma. c-Myc monoclonal antibody was from Santa Cruz Biotechnology. CLZ was from Prolume. CLZ400A (also known as Deep Blue®) was purchased from Biotium.
The RGFP cDNA was obtained from the plasmid pRrGFP (Prolume). All constructs were synthesized by PCR using standard procedures. To generate the fusion between the Renilla proteins the N-terminal methionine or the stop codon of RGFP were replaced respectively with a 21-mer (GDLGELSRILEQKLISEEDLL) or a 19-mer (EEQKLISEEDLGIPPARAT) linker peptide containing a c-Myc epitope (underlined). The obtained cDNAs were subcloned into pRluc-C1 or pRluc-N1 (PerkinElmer Life Sciences) to have Rluc–RGFP and RGFP–Rluc fusion proteins respectively.
The fusion between EYFP (enhanced YFP) (Clontech) and Rluc was obtained by linking the two cDNAs through a sequence encoding a 10-mer linker (GDLGELSRIL). Tagged versions of native Renilla proteins were obtained by extending the N-terminal of Rluc and the C-terminal of RGFP with a c-Myc epitope.
The CZ and NZ leucine zipper peptides tethered to c-Myc epitopes (boldface and underlined respectively in the following amino acid sequences) were generated by annealing two partially complementary synthetic oligonucleotides, carrying appropriate restriction sites at their ends. Subsequently, mutually primer synthesis was carried out.
The obtained fragments, encoding the CZ helix (MEEQKLISEEDLEQLEKKLQALEKKLAQLEWKNQALEKKLAQGGSGIPPARAT) and the NZ helix (EEQKLISEEDLGIQGGSGSGALKKELQANKKELAQLKWELQALKKELAQ), were tethered to the ends of Renilla proteins, which had been previously subcloned into pcDNA3 vector (Invitrogen) without methionine initiator or stop codon.
The sequence encoding the thrombin cleavage site and the hirudin domain (Leu38-Asp58) of the human PAR1 (proteinase-activated receptor 1) was PCR-amplified with only one substitution of glycine for Asp-39. This sequence was extended at the N- or C-terminus with a c-Myc epitope and the amplified fragments were inserted into the chimaeric constructs RGFP–Rluc or Rluc–RGFP, to generate the respective cleavable fusion proteins.
To construct the GPCR Rluc-tagged fusion proteins, we replaced the stop codon with a sequence coding for a 9-mer linker peptide (PGSPPARAT) in the human β2AR (β2-adrenergic receptor) and rat DOP (δ-opioid receptor), and a 15-mer peptide (KLAVPRARDPPARAT) in the rat V2R (vasopressin-2 receptor). The tagged receptors were then subcloned into the retroviral expression vector pQIXN (Clontech).
Rat β-arrestins 1 and 2, with their methionine initiator codon removed, were linked to Asp2 of RGFP through a peptide (EQKLISEEDLRT) and subcloned into pQIXH (Clontech).
Cell culture, transfection and extracts preparation
COS-7, HEK-293 cells (human embryonic kidney 293 cells) and GP2-293 cells were grown in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% (v/v) FCS, 100 units/ml penicillin G and 100 μg/ml streptomycin sulfate, in a humidified atmosphere of 5% CO2 at 37 °C.
Transient transfection of COS-7 cells was performed by the DEAE dextran/chloroquine procedure .
HEK-293 cell lines permanently co-expressing chimaeric GPCR–Rluc and RGFP–β-arrestins were generated using the pantropic retroviral expression system by Clontech. Briefly, recombinant retroviruses expressing receptor–Rluc or RGFP–β-arrestin fusion proteins were prepared by transfection of GP2-293 packaging cell with different retroviral vectors using FuGENE™, according to the manufacturer's instructions. Cells were allowed to increase the viral titre for 48–72 h before collecting the virus-containing supernatants. HEK-293 cells were infected with the RGFP–β-arrestins 1 or 2 retroviruses in the presence of 8 μg/ml polybrene for 48 h, and selected under hygromycin (100 μg/ml active drug). These polyclonal cells expressing RGFP–β-arrestin 1 or 2 were later super-infected with the different receptor–Rluc retroviruses, as described above, and selected under G418 (600 μg/ml active drug) and hygromycin. Receptor and β-arrestin expressions were confirmed by appropriate radioligand binding assay and Western-blot analysis respectively as described previously .
To prepare cytosol extracts for luminescence and fluorescence determinations, detached COS-7 cells, transiently transfected with vectors expressing the different fusion proteins, were homogenized in 50 mM Tris/HCl (pH 7.4) with a Dounce homogenizer and centrifuged for 10 min at 500 g, to pellet the nuclei. The resulting supernatants were centrifuged at 55000 g for 1 h at 4 °C, to separate the cytosol from the particulate fraction, and stored frozen at –80 °C. For measurements in living cell suspensions, cell monolayers were detached in Ca2+/Mg2+-free PBS containing 1 mM EDTA, centrifuged at 500 g for 5 min and resuspended in PBS supplemented with 0.1% D-glucose prior to the addition to the microcuvette. Protein determinations were performed as described by Peterson 
Measurements of luminescence and fluorescence
Bioluminescence and fluorescence measurements were made in a Fluorolog-3 spectrofluorimeter (Horiba Jobin Yvon), equipped with two identical emission monochromators and photon-counting detectors arranged in a T-format configuration. Emission spectra were corrected with company-provided correction files. Luminescence was recorded with the excitation lamp turned off. The Rluc substrates, CLZ and CLZ400A, were prepared as methanolic solutions (1 mM) and stored at –80 °C in light-shielded vials. Aliquots were dissolved in sample buffer before use and added to cell extracts at the final concentration of 5–10 μM.
Bioluminescence emission spectra were typically collected starting 2 s after the addition of substrate, using a combination of scanning steps and integration times such that a full spectrum could be completed in 10–25 s. When necessary, time-corrected spectra (i.e. corrected for the decay of Rluc emission during the acquisition time) were obtained by simultaneously recording light from both monochromators, one of which scanned the desired range of wavelengths, while the other was maintained at a fixed wavelength (either 475 or 510 nm) to record the rate of decay of light intensity, which was then used to apply the corrections to the spectrum.
To measure the ratios of light intensity at the emission maxima of donor and acceptor (from which BRET ratios were computed), the light emitted from the sample was read simultaneously from both monochromators, one set at 475 nm and the other at 510 nm, using a slit of 5 nm. Readings were started 5–8 s prior to substrate addition, to record dark current values, and were continued thereafter for 60–120 s using time increments and integration of 0.2 s. Small differences in light response between the two photomultipliers were corrected using factors computed by the simultaneous recording of the water Raman emission bands generated at similar wavelength maxima. All data points corresponding to the following 30 s starting 0.4 s after the peak value of light emission were used and averaged for calculation of BRET ratios.
Cytosol extracts (∼30 μg of total proteins) from COS-7 cells transiently transfected with plasmids encoding the cleavable fusion proteins were incubated in 50 μl of a buffer containing 1 mM Tris/HCl (pH 7.4) and 100 mM NaCl, in the presence and absence of thrombin (10 units/ml) for 30 min at 25 °C. The reactions were arrested by adding hirudin (15 units/ml) to each sample. For luminescence measurements, the samples were diluted in 50 μl of 1 mM Tris/HCl (pH 7.4). For Western-blot analysis, sample aliquots (15 μg of proteins) treated or not treated with thrombin were diluted in Laemmli sample buffer and separated by SDS/PAGE (10% gel). Immunoblotting was performed using monoclonal anti-c-Myc antibody (0.2 μg/ml), and the reactive bands were visualized by phosphatase staining with Promega reagents.
Fluorescence index of constructs
Since all our constructs were studied in cytosol extracts, we devised to measure an index related to the intrinsic fluorescence of the GFP domain, in order to compare different chimaeric proteins. For each cytosol prepared from cells transfected with a given construct, serial dilutions were used to build two titration curves relating respectively total fluorescence and c-Myc-immunoreactivity to the total protein concentration of the extract. Total fluorescence was measured in the spectrofluorimeter, adding each diluted aliquot into 1 mM Tris/HCl (pH 7.4) with excitation and emission monochromators set at 475 and 510 nm respectively. For the assessment of immunoreactivity, the same aliquots were diluted into Laemmli sample buffer, separated by SDS/PAGE (10% gel), and electrotransferred on to PVDF membranes (Millipore Immobilon). Immunoblotted bands were revealed using mouse anti-c-Myc antibody followed by fluorescein conjugate anti-mouse (Amersham Biosciences) as secondary antibody and were quantified by fluorescence scanning in a Typhoon 9200 (GE Healthcare).
From the slopes of such curves (obtained by linear regression), we calculated the FI [fluorescence intensity; in c.p.s. (counts/s)] and the IR [immunoreactivity; in RFU (relative fluorescence units)] per μg unit of extract. Their ratio FI/IR (c.p.s./RFU) is a number related to the intrinsic fluorescence of the expressed protein. It can be used to compare chimaeric proteins expressed in different cell extracts, under the assumption that the antibody affinity for the c-Myc epitope does not diverge significantly across the different constructs.
BRET assay of chimaeric receptors and β-arrestins
Luminescence emission spectra of cells expressing pairs of luminescent GPCR and fluorescent β-arrestins were measured as described above for intact cell determinations. Cell suspensions (1–3×105 cells) were incubated with or without the different ligands for 5 min in a 200 μl cuvette before adding CLZ (5 μM). For BRET determinations in attached monolayers, cells were plated on to either 96- or 24-well sterile white plastic plates (Packard View-plate) at a density of 1×105 and 5×105 cells per well respectively. To measure BRET ratios, the medium was replaced with PBS containing CLZ (5 μM) and, after 10 min, the plates were inserted in a plate luminometer (Victor Light, PerkinElmer) equipped with automatic injectors and two bandpass emission filters [blue: 470(20) nm and green: 510(20) nm; 3RD Millennium, Omega Optical, Brattleboro, VT, U.S.A.]. Following injection of agonist±antagonist, or PBS, the wells were incubated for a further 5 min and sequentially read through the green and blue filters (5 s integration). For recording the enhancement of light emission with the low quantum yield substrate CLZ400A, the medium was replaced with PBS containing CLZ400A (10 μM) with or without the receptor agonist, and the total luminescence of the wells was immediately read. For kinetics studies (as shown in Figure 9), cells expressing β2AR/β-arrestin 2 in 24-well plates were incubated for 5 min with CLZ400A. Next, the plates were inserted in the luminometer and the emission of total luminescence was recorded at 1 s intervals following a first injection (at 90 s) of either PBS or isoprenaline (100 nM final concentration) and a second injection (at 210 s) of PBS or alprenolol (10 μM).
Data analysis and calculations
Here, Y0 is the maximal emission of the peak and λ0 is the wavelength at the peak maximum, while b and Δλ are respectively an asymmetry factor and a parameter of peak spread, from which the peak width at half-maximal intensity can be computed as: W0.5=Δλ[sinh(b)/b].
A single Gaussian distribution was sufficient to describe the spectrum of Rluc alone, whereas a linear combination of three distributions was needed to accurately describe the spectra generated by the Rluc–RGFP interaction. Total light emission was computed either by numerical integration of the experimental data using the trapezoidal rule, or by summing the areas under the best-fitting distribution functions  calculated as:
BRET ratios were computed from the light intensities in c.p.s. recorded at the two wavelengths maxima (λ510 and λ475) corrected for spectral overlap. That is, BRET ratio=c.p.s.510(1–m)/c.p.s.475(1–n), where m is the ratio λ510/λ475 in the emission spectrum of Rluc/CLZ, and n is the ratio λ475/λ510 in the fluorescence emission spectrum of RGFP.
Absence of BRET signals in wild-type Renilla proteins
Efficient RET between Rluc and RGFP was previously described in vitro using purified proteins . To investigate whether energy transfer can occur in the extracts of cells expressing the genetically encoded proteins, we expressed either individually or jointly wild-type RGFP and Rluc in COS-7 cells.
Bioluminescence spectra triggered by the addition of 5 μM CLZ were recorded in extracts containing each individual protein, either co-expressed in the same cell or prepared as mixtures obtained by reconstituting individual extracts in various proportions. Although both RGFP and Rluc were well expressed, as indicated by the luminescence and fluorescence signals of the cell extracts, no detectable BRET emission could be detected. The bioluminescence spectrum of the RGFP/Rluc mix was always superimposable to that recorded for Rluc in the absence of RGFP (Figure 1A). Identical results were obtained either under very low (1 mM Tris/HCl) or normal (50 mM Tris/HCl and 100 mM NaCl) ionic strength reaction conditions.
RET between Rluc and RGFP can only be detected in fusion proteins
Thus RGFP and Rluc are not capable of forming a BRET emitting association complex under such conditions.
High- and low-efficiency BRET in fusion proteins
Next, we generated two dimeric fusion proteins in which the luciferase donor was alternatively linked to the N- or the C-terminal end of the fluorescent acceptor through spacer residues (see the subsection ‘cDNA constructs’ in the Experimental section). Bioluminescence spectra were recorded from COS-7 cell extracts expressing the fusion proteins.
In both cases emission spectra were bimodal, exhibiting, in addition to the 475 nm peak of Rluc, a second green-shifted component (λmax=510 nm), corresponding to the peak emission of RGFP (Figures 1B and 1C). This indicates that energy transfer from the CLZ excited state to the fluorescent chromophore readily occurs in such chimaeras. However, there was an impressive difference in transfer efficiency between the constructs. In the protein with RGFP in the N-terminal position, the BRET peak was predominant, as light was almost entirely emitted through the green pathway (Figure 1C). In contrast, when RGFP was downstream of Rluc, the broad emission spectrum of CLZ was the main component of the emitted light, and the BRET signal appeared as a small shoulder on the low energy side of the spectrum (Figure 1B).
Essential role of the N-terminus of RGFP
The difference in BRET efficiency between the two proteins suggests that the modifications due to the fusion (which in the construct with reduced BRET involve the GFP N-terminal and the Rluc C-terminal) might either impair the intrinsic fluorescence of GFP or interfere with the protein–protein interaction generating efficient energy transfer. Alternatively, the C-terminus modification in Rluc might adversely affect catalytic efficiency. To sort out such possibilities, we modified the chimaera with the best RET response (i.e. RGFP–Rluc) by extending either the N-terminus of the fluorescent protein domain or the C-terminus of the luciferase with ten histidine residues.
The fluorescence index (in c.p.s./RFU, see the Experimental section) of the N-terminus extended construct (4.4±0.6) was 60% compared with the C-terminus modified chimaera (7.4±1.1) and 40% compared with c-Myc tagged wild-type RGFP (10.7±1.9). This suggests that the two modifications have only minor effects on the intrinsic fluorescence of the fluorophor. We also measured excitation fluorescence spectra of modified constructs and wild-type GFP. They were superimposable, suggesting that there are no changes in the absorption properties of the chromophore.
In contrast, the luminescence spectra shown by the construct with the N-terminus extended GFP sequence displayed a marked reduction of the RET peak, similar to that observed for Rluc–RGFP, whereas the C-terminal extension of Rluc had no significant effect (Figure 2). Thus the results clearly indicate that a free N-terminus in RGFP is an essential requirement for the occurrence of the highly efficient energy transfer process.
Effect of the N-terminus modification of RGFP on BRET efficiency
Proteins tethered to self-assembling leucine zippers
We next evaluated if the two proteins could generate non-radiative energy transfer when brought into molecular proximity by the interaction of two tethered complementary partners. We used a pair of antiparallel ‘leucine zipper’ sequences, which are a useful self-assembling tool for the detection of very weak protein–protein interactions [29,30].
Four different fusion constructs were designed, such that the antiparallel peptide helices CZ and NZ were alternatively encoded on the C- and N- termini of the two proteins. This resulted in two complementary pairs of interacting proteins capable of associating with opposite orientations, only one of which involves GFP with a free α-amino group (see the scheme in Figure 3).
RET resulting from reassembling the interaction between Rluc and RGFP via leucine zippers
As observed for the fusion constructs described above, the bioluminescence spectra studied in reconstituted cell extracts show that only when the GFP N-terminus is free, high-efficiency BRET was observed (Figures 3A and 3B). This again underscores the fundamental role that the N-terminus of the fluorescent proteins plays for the occurrence of high-efficiency BRET.
To verify whether the interaction driven by the leucine zipper tags is also detectable in living cells, different ratios of cDNA mixes encoding GFP–NZ and CZ–Rluc fusion proteins were transfected in COS-7 cells and the luminescent spectra were compared with those of cells transfected using equivalent proportions of plasmids encoding the corresponding native proteins. BRET signals were clearly present in cells co-transfected with the leucine zipper-linked proteins, and the relative height of the 510 nm peak steadily rose as the fraction of acceptor-coding cDNA increased (Figure 4A). In contrast, no BRET was observed in cells expressing the native proteins, although the level of wild-type expression were roughly 10-fold greater than that of zipper chimaeras (Figure 4B) – as judged from the intensity of luminescence and fluorescence (results not shown). There is a clear correlation between the relative size of the RET peak (gauged as wavelength intensity ratios at 510 nm/475 nm) and the transfection ratio of the cDNAs coding the acceptor and donor. This is consistent with the notion that RET results from the leucine-zipper-driven self-association of Rluc and RGFP inside the cells (Figure 4C).
Comparison of luminescence spectra emitted by Renilla proteins intracellularly reassembled as leucine zipper chimaeras or native proteins
Cleavable constructs and reversibility of the high-efficiency BRET interaction
The results presented above clearly demonstrate that Rluc and RGFP cannot form a spontaneous complex in solution, but they do so when forced into close proximity via linked binding partners that have sufficient affinity. However, one important question is whether the complex, once formed, can dissociate when the interaction between the linked partners is interrupted.
To answer the question, we made two cleavable fusion proteins RGFP–Thr-Rluc and Rluc-Thr–RGFP. In both, we introduced a thrombin cleavage site within the linker region of the fused proteins and, to ensure efficient thrombin cleavage, we also inserted a flanking sequence corresponding to the ‘hirudin-like’ domain (Phe41-Asp58) of the protease-activated receptor-1 .
Western-blot analysis of cytosolic extracts containing the expressed proteins (shown for RGFP–Thr-Rluc in Figure 5A), prior to and following thrombin treatment, indicated that the enzyme can accomplish the complete cleavage of the chimaera under mild reaction conditions. The luminescence spectra of the cleavable proteins were identical with those of the corresponding non-cleavable constructs. In both cases, thrombin treatment converted the spectrum into the single broadband peak (λmax=475 nm) typical of Rluc emission in the absence of RGFP. Furthermore, the hydrolysis was specific, as no effect of thrombin was observed in the constructs lacking the cleavage site, nor was cleavage of the constructs detected in the presence of equimolar amounts of hirudin (results not shown).
Quantification of the RET efficiency using RGFP–Thr-Rluc and Rluc-Thr–RGFP cleavable fusion proteins
One advantage of the cleavable constructs is the possibility of comparing light output prior to and after thrombin treatment without altering reactant concentrations. This allows the quantification of the difference in RET efficiency between associated and dissociated proteins. Using this strategy, we measured the BRET ratios (e.g. the ratio of light emissions at peak wavelengths, 510/475, corrected for spectral cross-talk) in the RGFP–Thr-Rluc (17.1±0.1) and Rluc-Thr–RGFP (1.72±0.03) constructs prior to and following thrombin hydrolysis (Figure 5B). RET efficiency is thus 10-fold greater when the interaction between the proteins entails a free N-terminus in RGFP. In both chimaeras, RET was abolished by thrombin cleavage (Figure 5B). Therefore the results reassert the lack of spontaneous RET-producing interaction between free Rluc and RGFP in solution, but also indicate that even when the interaction is assisted by a covalent link, it can be readily reversed upon link removal.
Quantum yield enhancement of luminescence
An additional important feature emerged when we studied the luminescence spectra of the substrate CLZ400A in the cleavable construct. The emission spectrum of this analogue (λmax=396 nm) has marginal overlap with the absorption spectrum of RGFP. This makes it a poor donor of energy transfer in this system. Yet in the GFP–Thr-Luc construct, the spectrum of CLZ400A displayed a prominent narrow bandwidth peak at 510 nm, in addition to the typical 395 nm peak of the CLZ400A excited state, indicating energy transfer to the RGFP chromophore. Thrombin hydrolysis virtually abolished the green component, but also dramatically reduced light output through the blue pathway (Figure 6).
Quantum yield enhancement in CLZ400A emission
Spectral integration in control and thrombin-treated samples indicated that the enhancement in overall light output from free luciferase to RGFP-bound enzyme was roughly 44-fold. This gain of apparent quantum yield of the CLZ400A luminescence emitted from the GFP–Luc complex is of the same order of magnitude as that previously reported using purified proteins . Unlike those studies, however, we find that both the blue (395 nm) and the green (510 nm) emissions are enhanced as a result of the interaction of Rluc with RGFP.
The Renilla system as a reporter of GPCR–β-arrestin interactions
We speculated that this gain in luminescent emission can constitute a novel and alternative readout for BRET signal detection in protein–protein interaction studies. To verify this idea, we tested the performance of the Renilla proteins to act as reporters of the interaction between GPCR and arrestins .
A number of vectors encoding complementary sets of chimaeric proteins were prepared. To ensure the presence of a free α-amino group in its sequence, the acceptor RGFP was fused to the N-terminal ends of both β-arrestin 1 and 2. The Rluc donor was instead fused to the C-terminus of three GPCRs: the β2AR, the DOP and the V2R. Using retroviral vectors, six HEK-293 cell lines were prepared, each co-expressing one luminescent donor receptor with one fluorescent acceptor β-arrestin.
As shown in Figure 7, the luminescence spectra of CLZ400A recorded in intact cells co-expressing either β2AR–Rluc (Figure 7A) or DOP–Rluc (Figure 7B) with RGFP–β-arrestin 2 were sharply changed by the addition of the corresponding receptor agonists (isoprenaline and [D-Ala2, D-Leu5]enkephalin respectively). In both cases, agonist-induced receptor association to β-arrestin was revealed by a major increase of the 510 nm peak and an overall enhancement of total light emission over basal conditions. These changes were largely prevented in the presence of receptor antagonists (Figures 7A and 7B). We found no detectable agonist effects on luminescence in cells co-expressing DOP–Rluc or β2AR–Rluc with RGFP–β-arrestin 1, which is in agreement with results indicating that β2AR and DOP do not interact tightly with this protein. Conversely, AVP (arginine-vasopressin) induced strong luminescence enhancements when either RGFP–β-arrestin 2 or RGFP–β-arrestin 1 were present in cells expressing V2R–Rluc, a receptor known to interact with both β-arrestin types  (results not shown). A small, but detecTable 510 nm peak was present in the absence of agonists (see Figures 7A and 7B). This basal emission was absent from cells where β2AR and DOP were co-expressed with RGFP–β-arrestin 1 (results not shown). Thus it is likely that the signal reflects some extent of constitutive interaction between unbound receptor and β-arrestin, rather than background noise due to spontaneous binding of the reporters.
The Renilla BRET system as reporter of GPCRs–β-arrestin interactions
The light enhancement occurring when BRET is triggered via CLZ400A is large enough to allow for the detection of the interaction as a simple increase in total luminescence, without need of spectral resolution of the emitted light. This was tested as shown in Figure 8, where cells co-expressing V2R and β-arrestin 1 or 2 were plated in multiwell flasks, and the AVP-induced interaction of the V2R with the two β-arrestins was compared using two methods of detection. In one, we made conventional ratiometric determinations of the change in BRET ratios using cells incubated with CLZ. AVP produced a ∼3-fold enhancement of BRET ratio in both V2R–β-arrestin 1 and V2R–β-arrestin 2 cells, and both were prevented in the presence of the antagonist [d(CH2)31,D-Ile2,Ile4,Arg8]vasopressin (Figure 8A). In the second method, the enhancement of the total luminescence was recorded at consecutive time intervals following the addition of CLZ400A plus or minus the agonist AVP (Figure 8B). The ratios of light intensities in the presence and absence of agonist increased with time up to 10-fold, indicating that RET induced by the GPCR–β-arrestin association can also be revealed as an increase in total luminescence.
Ratiometric and luminometric detection of V2R–β-arrestin interaction
We further tested if this enhancement of quantum yield resulting from the binding of the two Renilla tags can also detect both the association and the dissociation of the arrestin–receptor complex in the cell. As shown in Figure 9, the addition of the agonist isoprenaline to cells co-expressing β2AR–Rluc and RGFP–β-arrestin 2 in the presence of CLZ400A induced a rapid increase in total luminescence, which was reversed upon further injection of the antagonist alprenolol. These results confirm and extend the results obtained with cleavable constructs, as they demonstrate that the reversibility of the RGFP–Rluc complex is maintained also when the Renilla reporters are linked to proteins that interact in living cells.
Kinetics of β-arrestin–receptor interaction monitored by changes of luminescence
The luciferase and the fluorescent protein of Renilla coelenterates represent the first natural bioluminescent system where radiationless energy transfer was investigated [16,20]. Although quantitative studies demonstrated that the process (modelled according to a Föster-type mechanism) can reach a surprising 100% efficiency in these proteins , no work has been done to assess if such molecules can make a useful gene reporter system for the detection of protein–protein interactions in cells.
In the present paper, we have reinvestigated the interaction between Rluc and RGFP with this question in mind. Unlike previous studies, which were based on the analysis of the purified proteins in diluted, low-salt solutions, here we adopted a protein engineering approach to probe their interaction. Using Rluc and RGFP sequences embedded into chimaeric vectors to force their intramolecular interaction, or proteins tethered to self-assembling α-helices to drive their intermolecular association, we tested which configuration was required for optimal energy transfer from the Rluc bioluminescent reaction centre to the GFP chromophore.
Taking advantage of thrombin cleavable constructs, we also measured differences in RET efficiency, and quantified the net gain of light output that occurs when the oxidation of the low-quantum-yield substrate CLZ400A is compared in the presence and absence of the acceptor RGFP.
We finally used the well-known interaction between membrane-bound GPCRs and β-arrestins [32–34] as a benchmark tool to test in living cells the performance of the Renilla proteins as a reporter system. Three clear results emerge from the present study.
First, there is no detectable background interaction between Rluc and RGFP when these molecules are expressed as free native proteins in cytosolic extracts or living cells. The inability to undergo spontaneous association is further confirmed by additional evidence. One is the total disruption of RET observed upon disconnection of the tethered proteins from the cleavable constructs. Another is the absence of ground reporter signals in the β-arrestin–receptor assay. In fact, the small levels of RET observed in the absence of agonist in cells expressing adrenergic or opioid receptors reflect a basal interaction between those GPCRs and β-arrestin 2, as it was not displayed when the same receptors were co-expressed with arrestin 1, a protein to which they bind far less strongly .
Thus we note that Rluc and RGFP do not generate more background noise than other pairs of reporters previously used for BRET assays, at least under the experimental conditions that are normally employed in such a type of assays. This finding, although positive with regard to the use as a reporter system, is puzzling when confronted with the dissociation constant of ∼1 μM estimated for the binding between Rluc and RGFP [21,23]. As we did not attempt to measure the intracellular concentrations of the expressed proteins, we must assume them to be equal to or below nanomolar, to account for the lack of any detectable spontaneous association. Alternatively, we may suspect that the binding affinity was overestimated in previous experiments.
This lack of spontaneous interaction may seem a paradox, if one considers that in living Renilla cells the two proteins readily interact, and the resulting RET is efficient enough to cause complete conversion of the emitted light spectrum. However, it was demonstrated in anthozoan cells that luciferase and GFP are sequestered into sub-micrometre-sized organelles called lumisomes . It is thus likely that high local concentration due to compartmentalization, and perhaps membrane docking mechanisms within the lumisome, are obligatory factors that promote the interaction between luciferase and GFP, despite the low intrinsic affinity between the proteins.
Molecular mass estimates obtained by gel chromatography also suggested that RGFP may exist in dimeric form [21,23]. It was proposed that the minimal required complex for high efficiency BRET emission might be the heterotrimer between one Rluc and two RGFP molecules . Our results obtained in RGFP–Rluc fusion proteins would suggest that a 1:1 complex is sufficient for an efficient interaction, although we cannot exclude a spontaneous association of the chimaeric constructs through their GFP domains. We found that BRET ratios of fusion proteins did not change even when they were diluted up to 20-fold (results not shown), which makes it difficult to believe that high-efficiency BRET depends on their self-association. Obviously, more detailed investigations are necessary to verify conclusively if dimeric GFP is an essential intermediate of the interaction.
Regardless of such unsolved questions on reaction strength and stoichiometry, our results clearly indicate that the tendency of Renilla proteins to form a complex is not an obstacle for the use of this system as BRET reporter of protein–protein interactions in transfected mammalian cells.
A second novel result in the present study is that the high-efficiency RET process generating the green luminescence of Renilla requires a free α-amino group in the molecule of RGFP. This finding explains previous observations showing that the reaction of purified RGFP with amino-group-modifying agents disrupted its RET acceptor activity . An effect of the free N-terminus in enhancing BRET efficiency was also reported for the A. victoria mutant GFP2, by comparing fusion constructs in which this protein was linked with either the N- or the C-terminus of Rluc .
The N-terminal modification of RGFP does not appear to produce important changes in the fluorescence of the fluorophor, or to prevent the transfer of energy between the optical centres of the two proteins. It does, however, cause a dramatic diminution of RET efficiency. The decrease is 10-fold, if deduced from the difference in BRET ratios, and affects either the intramolecular RET in fusion proteins or the intermolecular RET induced via the docking of antiparallel α-helices.
Our results suggest that the α-amino group in the N-terminus of GFP is crucial for the formation of the dimeric complex that allows RET to achieve maximum efficiency. This complex might optimize the physics of energy transfer to an extent that would never be possible by simply pulling into maximal proximity the molecular shells of the two proteins. This means that there are two different ways in which the Renilla proteins can be used as a reporter system.
One is when the RGFP N-terminal is blocked within the encoded fusion protein. The system in this case behaves and performs just like any other BRET reporter assay described so far [10,12]. There might be the advantage of a slightly improved BRET efficiency, due to the better spectral overlap between the emission of Rluc and the absorption of RGFP. However, that comes with the disadvantage of a poorer separation between the maxima of the BRET peak and the emission of CLZ.
The second is when GFP is encoded with a free N-terminus. The system, in this case, has the potential to deliver BRET emissions that display the same high efficiency as in the intact Renilla. Such an unprecedented efficiency should allow engineering assays with far superior sensitivity and accuracy than those described so far.
There is, however, an important caveat. Unlike conventional BRET assays, where the signal depends on the extent to which the reporters are brought into the range of distances in which energy transfer can occur, in the Renilla system BRET results from the complex that is formed when the reporters reach a threshold distance. Once that complex is made, it is not clear if the intermolecular distance and orientation of the optical centres can still be influenced by the molecular motion of the carrier proteins. For this reason, we consider this type of BRET assay much more akin to those based on functional complementation of split enzymes [37,38] or split fluorescent protein fragments [30,39,40]. Further work is necessary to investigate if this ‘complementation-induced’ BRET system can be useful for engineering intramolecular BRET biosensors (i.e. proteins carrying BRET labels tethered at distal ends of the molecule, so that the binding of signalling intermediates, like Ca2+ or cAMP, can induce conformational changes that are reported as changes of RET efficiency [41–45]). One clear application, however, is the utilization of this Renilla system for engineering single-chain optical sensors of proteolysis . As shown here, thrombin hydrolysis of the cleavable construct engineered with an N-terminal GFP caused a decrease in BRET ratio from 17.1 (±1.5) to 0.11 (±0.07). This means a >150-fold dynamic range in the response to thrombin hydrolysis. Conceivably, by inserting between the Rluc and the RGFP labels peptides containing specific cleavage sites for proteases of pharmacological interest (such as caspases or viral hydrolases), highly sensitive and specific probes can be prepared that may be particularly suited to use in high-throughput assay systems.
The third point emerging from the present study is that the quantum yield enhancement that occurs in the emission of the didehydroxy analogue CLZ400A can be effectively exploited as readout of the interaction between Rluc and RGFP. As deduced from the difference between intact and cleaved fusion proteins, we measured a more than 40-fold enhancement of light output when the luminescence emitted from CLZ400A/Rluc takes place in the presence of RGFP. This property allows measuring ‘complementation-induced’ BRET in a simple luminometric assay which does not require multiple filter readings. When adapted to the study of the GPCR–arrestin interactions in living cells, the system reports agonist-induced binding of receptor to arrestin as a severalfold enhancement of total light emission, a remarkable improvement of signal-to-noise ratio for the detection of a protein–protein interaction in an intact cellular system. Moreover, both the association and dissociation of the receptor–arrestin complex can be detected.
The mechanism underlying the phenomenon of quantum yield enhancement is still unclear. To explain the emission of low-quantum-yield analogues, such as CLZ400A, two singlet excited states were proposed. Both a neutral (λ=395 nm, bad RET donor) and a monoanionic (λ=475 nm, good RET donor) species would be produced during oxidation of CLZ400A. Since the mode of binding of this substrate to the enzyme pocket would cause fast quenching of the monoanionic state through solvent and/or protein interactions, there is no detectable emission at 475 nm. In the presence of the acceptor GFP, the energy transfer process would compete with the quenching and unveil the monoanion as 510 nm RET emission .
This theory was based on the observation that the addition of GFP to CLZ400A/Rluc induced the emergence of the 510 nm peak but did not change the 395 nm emission . In contrast, we found here that the interaction with GFP not only generates the green emission band, but also results in a dramatic enhancement of the bioluminescent emission at 395 nm, suggesting that GFP binding may change the overall catalytic properties of the enzymatic reaction between CLZ400A and Rluc, rather than merely ‘rescue’ one non-radiating intermediate of the reaction. A recent and interesting study shows that a single site mutation in the molecule of Rluc can produce enhancements of quantum yield that are identical with those induced by GFP binding to Rluc, and suggests that the mutated residue might be involved in the interaction of Rluc with the fluorescent protein . If the effect of GFP can be reproduced by a mutation in the Rluc sequence, it is likely that the mechanism of luminescence enhancement involves a conformational change that the binding of the fluorescent protein induces in the molecule of Rluc and cannot be solely explained by the dynamics of luminescence excited states.
In summary, our results demonstrate that the Renilla system provides several opportunities for engineering optical reporter assays that are useful for measuring macromolecular interactions in living cells. Besides its use as a conventional BRET assay, when both GFP and luciferase are fused in-frame with the C-terminal ends of the proteins under study, our results highlight a novel strategy for exploiting its properties. This relies on the natural capability of the Renilla proteins to form a high-efficiency RET-emitting complex when the proteins to which they are attached reach a permissive intermolecular distance.
This strategy, which we call ‘complementation-induced BRET’, requires that the energy acceptor must be positioned on the N-terminus of the protein of interest. Moreover, depending on the substrate that is used, complementation-induced BRET can be detected in two different ways: as ratiometric determination of RET efficiency, when using native CLZ, or as enhancement of total luminescence, when exploiting the large increase in quantum yield observed for CLZ400A.
An obvious limitation of complementation-induced BRET, as of any other assays based on functional complementation, is that it cannot provide a quantitative measure of the intermolecular distances between the interacting proteins. Quantification, at any rate, is a difficult task even in conventional FRET and BRET assays, particularly when they are based on genetically encoded labels , and the quantitative interpretation of the results is still a highly controversial matter [48,49]. Even if only capable of qualitative information, this new strategy should allow development of assays with unrivalled characteristics in terms of signal-to-noise ratio, accuracy and sensitivity.
We gratefully acknowledge Dr Antonio De Blasi (University of Rome ‘La Sapienza’, Rome, Italy) for the gift of cDNAs for β-arrestins and Dr Fabio Naro (University of Rome ‘La Sapienza’) for V2R. This work was supported in part by the FIRB (Il Fondo per gli Investimenti della Ricerca di Base) ‘Internationalization’ programme grant RBIN04CKYN_001.
fetal calf serum
resonance energy transfer
green fluorescent protein
- HEK-293 cells
human embryonic kidney 293 cells
relative fluorescence units
yellow fluorescent protein