Coelenterazine | Nur77 Signal

Bioluminescence resonance energy transfer–based imaging of protein–protein interactions in living cells

Hiroyuki Kobayashi, Louis-Philippe Picard, Anne-Marie Schönegge and Michel Bouvier *

Abstract

Bioluminescence resonance energy transfer (BRET) is a transfer of energy between a luminescence donor and a fluorescence acceptor. Because BRET occurs when the distance between the donor and acceptor is <10 nm, and its efficiency is inversely proportional to the sixth power of distance, it has gained popularity as a proximity-based assay to monitor protein–protein interactions and conformational rearrangements in live cells. In such assays, one protein of interest is fused to a bioluminescent energy donor (luciferases from Renilla reniformis or Oplophorus gracilirostris), and the other protein is fused to a fluorescent energy acceptor (such as GFP or YFP). Because the BRET donor does not require an external light source, it does not lead to phototoxicity or autofluorescence. It therefore represents an interesting alternative to fluorescence-based imaging such as FRET. However, the low signal output of BRET energy donors has limited the spatiotemporal resolution of BRET imaging. Here, we describe how recent improvements in detection devices and BRET probes can be used to markedly improve the resolution of BRET imaging, thus widening the field of BRET imaging applications. The protocol described herein involves three main stages. First, cell preparation and transfection require 3 d, including cell culture time. Second, image acquisition takes 10–120 min per sample, after an initial 60 min for microscope setup. Finally, image analysis typically takes 1–2 h. The choices of energy donor, acceptor, luminescent substrates, cameras and microscope setup, as well as acquisition modes to be used for different applications, are also discussed.

Introduction

Protein trafficking and interactions with different partners are at the core of most physiological responses. Monitoring these processes in real time in living cells provides important information about the spatiotemporal regulation of multiple cell functions. In recent years, the use of fluorescence and luminescence tools has led to major breakthroughs in our understanding of cellular dynamics, by allowing monitoring of movement and interaction of proteins. Among the approaches used, RET has gained in popularity.
RET is a natural phenomenon occurring between two photoactive molecules1. It corresponds to the transfer of energy from a donor to an acceptor molecule through a non-radiative resonance process occurring through dipole–dipole coupling that happens only at a permissive distance and proper orientation. The transfer of energy results in the excitation of the acceptor, which then emits light at a specific wavelength. Two types of RET phenomenon have been mainly used to monitor biological processes: fluorescence RET (FRET)2,3 and BRET4,5, which use fluorescent and luminescent donors, respectively. For most RET pairs, efficient transfer can occur only if the distance between the donor and the acceptor is <10 nm, and the efficiency of transfer decreases as a function of the sixth power of the distance between them. The average size of a protein being ≅5 nm, changes in RET between donors and acceptors fused to proteins of interest reflect changes in the distance between the tagged proteins, which are consistent with the occurrence of macromolecular rearrangements. As a result, RET assays have been widely used to characterize protein–protein interactions and conformational changes within proteins or protein complexes.
BRET has been extensively used to investigate G protein–coupled receptor (GPCR) dynamics and signaling activity. For instance, BRET-based assays have been used to study receptor multimerization4,6,7, coupling to and activation of G protein8,9, trafficking10, engagement and activation of accessory proteins, such as β-arrestins11–15 and receptor activity–modifying proteins16, as well as post-translational modifications, such as ubiquitination17. More recently, BRET has also been used for ligand-binding assays using fluorophore-conjugated ligands and a receptor fused to an energy donor18,19. In most cases, the BRET signal was quantified by spectrometric measures using luminometers equipped with a monochromator or filters separating the donor and acceptor emissions. Classically, BRET is quantified by dividing the light signal of the acceptor by the luminescence emitted by the donor. Although useful, such spectrometric studies cannot provide information about the subcellular localization of the processes monitored. Although it has been difficult to image BRET signals with high spatial resolution because of the low light output of luciferase (the energy-luminescent donor), recent enzymatic improvement20 of Renilla luciferase (Rluc)-based BRET donors and the development of new luciferases such as NanoLuc (Nluc)21 have improved signal strength markedly, allowing the development of BRET imaging approaches, including imaging of protein–protein interactions in culture cells22–27 and even in living animals, using red-shifted BRET probes28,29 that overcome tissue absorption. In this protocol, we present procedures for BRET-based microscopic visualization of protein–protein interactions and trafficking that combine recent improvements in ultralow-light detectors, new generations of BRET probes and new approaches, such as enhanced bystander BRET (ebBRET)26,27. GPCR activation and trafficking are used as an example of the biological systems that can be studied using such high-resolution BRET microscopy imaging.
The protocol provides all the information needed to select the best BRET pairs and detection systems for different applications, as well as presenting the advantages and disadvantages of the different approaches. In addition, the step-by-step procedure allows investigators to easily perform BRET imaging experiments.
The major components of the BRET imaging microscopy system are illustrated in Fig. 1. The system used in the present protocol is composed of an inverted microscope connected to an electron-multiplying CCD (EMCCD) camera through a regular port equipped with a motorized filter wheel. The motorized system allows rapid switching between filter-on (GFP emission) or filter-off (corresponding to the entire light emitted, which is used as a measure of the donor emission). Because BRET is a luminescence-based assay, it does not require external illumination, but a light source for bright-field or epi-fluorescence microscopy is required for focusing purposes. Luminescence is produced by the direct application of the luciferase substrate to the cells.

Comparison with other methods

Unlike most other protein–protein interaction assays, such as co-immunoprecipitation, pull-down, protein ligation assays30 or cross-linking assays, RET-based assays can directly monitor protein–protein interactions in living cells. This noninvasive characteristic of the RET assay is particularly useful when the properties of the interacting proteins can be changed by their extraction or purification, or when the influence of the cellular environment is being studied. Another assay that allows the detection of protein–protein interaction in living cells is the protein complementation assay (PCA)31. PCA imaging is based on the reconstitution of a reporter protein (typically fluorescent proteins or luminescent enzymes) from its split fragments fused to the proteins of interest. Therefore, the PCA signal can be detected only when two PCA fragments reconstitute the functional reporter as a consequence of the interaction of the fusion proteins. Although this assay can provide robust signals, the background signal can be relatively high because of the propensity of some of the fragments to self-associate. In addition, the dynamics of the interactions can also be affected by the complementation process itself, because reconstitution of the reporter protein can stabilize the complex. Unlike RET assays, which allow the quantitative monitoring of each of the partners (by monitoring luminescence and fluorescence independently), with PCA no information can be directly obtained about the quantity or distribution of the noninteracting fragments. Despite these limitations, PCA remains a useful type of assay. In particular, modifications of the interacting fragments to reduce their affinity for one another have been used to considerably reduce the possible artifact linked to the self-association of the fragments32. Also, PCA using split firefly luciferase33, Rluc34, Nluc (NanoBit)35 or split fluorescent proteins36 has been combined with BRET to monitor the formation of up to four multiprotein complexes37–39.
Many RET-based microscopic imaging methods have been developed to monitor biological processes in the specific subcellular compartments where they occur. One of the major challenges of RET imaging is that of separating weak energy transfer signals from background. For that reason, FRET has been favored over BRET for imaging purposes because the level of signal resulting from the excitation (by light) of a fluorescent donor is greater than that from the bioluminescent donors used for BRET. However, the strong external illumination in FRET assays also increases the background autofluorescence signal, which can limit the signal resolution of FRET imaging. Off-peak excitation of the acceptor by the donor excitation light source is also a source of contamination that is dependent on the amount of acceptor, which is why acceptor photobleaching or fluorescence lifetime–FRET imaging is often used. The strong FRET illumination also causes phototoxicity and photobleaching, making long time-lapse measurement difficult. By contrast, BRET does not require an external light-mediated excitation of the donor because the energy emitted results from a bioluminescent reaction involving the oxidation of a substrate. It follows that no autofluorescence occurs, yielding a good signal-to-noise ratio (SNR). However, the low-light intensity characteristic of BRET assays requires a much more sensitive system for signal detection (see the ‘Limitations’ section below).

Limitations

Most of the BRET imaging limitations are related to the ability to collect sufficient light to obtain high-resolution images. This limitation depends on the brightness of the partners, the energy transfer efficiency of the sensor pair used and the kinetics of the phenomenon studied. It follows that BRET probes expressed at low levels are more difficult to image. Longer acquisition times can be used to mitigate this limitation to some extent. However, longer acquisition times limit the ability to image real-time dynamics because (i) the time required to generate a sufficiently high-quality image may be longer than the timescale of the phenomenon that is being investigated or (ii) the time required is longer than the lifetime of the luminescent signal. The low intensity of light may also make it more difficult to obtain valid quantification of the BRET changes observed. This is partly true when the signal of interest is close to the shot noise (the statistical random fluctuation of the photon counts).
To mitigate these limitations, selection of the brightest BRET partners and substrates, as well as the use of efficient optics and sensitive detectors is of primary importance. In the examples shown in this protocol, the minimum acquisition time to obtain images for both total luminescence and energy acceptor channels is 25 s. It follows that real-time kinetic analyses are limited to phenomena occurring on a timescale >25 s. Although this allows the quantitative assessment of phenomena such as receptor endocytosis40 or β-arrestin recruitment41, it does not permit the real-time analysis of G-protein activation42, for example.
One of the advantages of BRET imaging over spectrometric (plate reader) BRET measurement is the possibility of assessing the subcellular origin of the signal. However, the subcellular resolution remains moderate and does not allow easy distinction of discrete subcellular organelles. One of the problems leading to lack of resolution is the extent of time needed to detect sufficient signal. Indeed, cellular movements occurring during the acquisition period will lead to blurred images and could even have an impact on the accuracy of the BRET values calculated for a specific pixel. To limit this effect, we recommend never exceeding an acquisition time of 10 s (100 × 100-ms frames) for each wavelength + 5 s for processing, so that the measurements of the two wavelengths are performed within a time interval during which relatively small changes of shape or cell movement occur. For measurements that require longer acquisition times because of the low intensity of the signals, we recommend performing several rounds (up to ten) of 25-s measurements (10 s for each wavelength + 5 s for processing), and the integration of each of the images that are not substantially affected by cell movement or cell-shape changes. This allows increased accuracy of BRET measurements at a specific location.

Experimental design

Energy donors and acceptors

The most frequently used BRET donor is Rluc, an oxidase isolated from the bioluminescent sea pansy, R. reniformis. In the original BRET assays, the native Rluc was used4,5. However, Rluc mutants, such as Rluc8 (A55T, C124A, S130A, K136R, A143M, M185V, M253L and S287L)20 or RlucII (A55T, C124A and M185V)43,44 have improved enzymatic activity, providing greater brightness and making them donors of choice for BRET imaging. More recently, another luciferase from the sea shrimp, O. gracilirostris, known as Nluc21, has also emerged as a good choice for some BRET applications (see ‘Additional BRET donors’ section below).
The fluorophores used as BRET acceptors usually belong to the family of GFPs, originating either from Aequorea victoria or R. reniformis. The best fluorophores are selected on the basis of their excitation and emission spectra (depending on the donor/substrate used), as well as their quantum yield and Stokes’ shift45. The BRET signal can have different characteristics depending on the nature of the luciferases used as donors, the luciferase substrates and the acceptor selected. Different combinations and their advantages and limitations are described in the following sections.

BRET1 assays

Rluc catalyzes the oxidation of its native substrate, coelenterazine (CTZ), into coelenteramide, and the relaxation of coelenteramide to the ground state produces luminescence with an emission peak at ~480 nm46. GFP variants having an excitation wavelength that overlaps this emission spectrum, such as eYFP (excitation: 514 nm) or Venus-GFP (excitation: 515 nm)47, can be used as BRET acceptors for the Rluc-CTZ donor couple and are used in so-called BRET1 assays. A synthetic CTZ analog, CTZh (2-deoxy CTZ), is also often used as a substrate for BRET1 assays, showing results similar to those obtained with CTZ.
BRET1 was the first BRET assay format developed for protein–protein interaction analyses4,5. Because the wavelength profile of BRET1 is similar to that of the frequently used CFP and YFP FRET pairs, devices and constructs already available made it easy to perform BRET1 experiments. However, a major drawback of BRET1 is the poor signal separation between the donor and acceptor emission wavelengths, resulting in a suboptimal SNR. As shown in Fig. 2a, the donor and acceptor emission spectra for BRET1 pairs have a large overlap because the spectral width of Rluc-CTZh luminescence (≈85 nm) is much larger than the Stokes’ shift of eYFP (≈15 nm). It follows that the acceptor signal is not well resolved from the contaminating donor signal, making it difficult to quantify, especially when the efficiency of transfer is low.

BRET2 assays

BRET2 was developed with the objective of increasing the separation between the donor and acceptor emission spectra, so as to improve signal-to-background ratio14. BRET2 uses synthetic CTZ analogs, such as CTZ400A (also known as DeepBlueC or bisdeoxy-CTZ) or Me-O-e-CTZ (also known as Prolume Purple), which emits blue-shifted luminescence peaking at ~400 nm when oxidized by Rluc20. The emission of these substrates has a narrower spectral width than the substrates used for BRET1 (~50 versus 85 nm, Fig. 2b). BRET2 acceptors include A. victoria GFP mutants, such as GFP10 (ref. 48) or GFP2 (ref. 14), which have much larger Stokes’ shift than YFP (~90 versus 15 nm) and can be efficiently excited with blue-shifted luminescence. As a result, BRET2 provides greater signal separation than BRET1, resulting in a greater dynamic window (see Anticipated results). However, the luminescence signal of BRET2 tends to be weaker than that of BRET1, due to the low light output of blue-shifted substrates, and may be more difficult to monitor depending on the detection systems used. Figure 3 illustrates the difference in luminescence emitted by the substrates used for BRET1 and BRET2. The BRET1 substrate, CTZh, generates luminescence signals that are 3–17 times brighter than the three blue-shifted CTZ analogs tested (Fig. 3a). Among the blue-shifted substrates, Me-O-e-CTZ and Me-O-CTZ-O-Me (also known as Prolume Purple II) are brighter than CTZ400A, which was the original substrate described for BRET2 (ref. 14). Another limitation of BRET2 is that all blue-shifted substrates show a much faster signal decay (half-life ≈12 min) as compared to CTZh (half-life ≈22 min) (Fig. 3b). This difference in the half-life of luminescence can be easily appreciated in luminescence images taken from cells expressing RlucII-tagged β-arrestin2.
Indeed, the luminescence following CTZh addition is easily detectable for >30 min, whereas the image obtained with Me-O-e-CTZ as the substrate decayed rapidly and was barely detectable after 10 min (Fig. 3c).

Additional BRET acceptors

The GFP from R. reniformis (rGFP), can be used as a BRET acceptor that also allows a good separation between donor and acceptor emissions (Fig. 4). It also results in a larger BRET signal due to a better transfer efficiency between Rluc and rGFP49 than between Rluc and GFPs from other species. Indeed, the fact that the two proteins co-evolved in the same species resulted in an optimal dipole orientation for transfer in the Rluc-rGFP dimer50. The very efficient transfer can easily be seen when comparing BRET images obtained for the Rluc-rGFP to Rluc-GFP10 fusion constructs (Fig. 4a, two last rows). However, the fact that rGFP spontaneously interacts with Rluc precludes the use of this energy acceptor for studying protein–protein interactions. Yet this propensity of the Rluc and rGFP pair to interact with one another (albeit with low affinity) can be used to increase the signal originating from random collisions known as bystander BRET. This property has recently been used to monitor the localization of protein in specific subcellular domains26. For instance, such ebBRET can readily detect the translocation of Rluc-fused proteins to a specific subcellular domain or organelle harboring rGFP targeted to these sites with specific subcellular localization motifs. In such cases, Rluc spontaneously interacts with rGFP, enabling efficient BRET energy transfer only if Rluc and rGFP are in the same subcellular compartment (see ‘Anticipated results’).

Additional BRET donors

In addition to Renilla luciferase, luciferases from different animal species, firefly Photinus pyralis51, copepod Gaussia princeps52 and deep-sea shrimp O. gracilirostris18, have also been used in BRET applications. As reported by several groups24,25, the Oplophorus luciferase is especially interesting for BRET imaging because engineered luciferases based on the Oplophorus enzyme, such as eKAZ53 and Nluc21 show greater light output than Rluc, and their catalytic subunits (19 kDa) are smaller than that of Rluc (34 kDa)54. The enzymatic activity of Nluc21, similar to that of Rluc55,56, is more stable than firefly luciferase under different environmental conditions such as temperature, pH and salt concentration. These advantageous characteristics of Nluc led to the development of novel Nluc-based applications, such as ligand-binding assays using Nluc complementation57 and BRET with fluorophore-conjugated ligands58. Although Oplophorus luciferases can use CTZ analogs as substrates46,53, it should be noted that the emission wavelength and light intensity obtained for various analogs are different from those obtained with Rluc (Fig. 5). For instance, Rluc substrates such as MeO-e-CTZ and Me-O-CTZ-O-Me do not show luminescence with Nluc, whereas another Rluc substrate, CTZ400A, as well as the Nluc substrate furimazine, generated brighter signals with 5.2 and 3.4 times greater light output than CTZh when used with Nluc (Fig. 5a). Interestingly, CTZ400A, which emits light at 400 nm with Rluc (blue-shifted compared to CTZh), does not show such a blue-shifted spectrum with Nluc, and the peak emission is observed at 460 nm, similar to what is observed with other CTZ analogs, such as CTZh and furimazine (Fig. 5b). This means that CTZ400A represents the brightest and cheapest alternative for Nluc-based assays. However, the luminescence decay for furimazine is slower than that of CTZ400A when using Nluc, a property that could make it a better alternative for long real-time imaging. An advantage of Nluc for BRET imaging is the much slower decay of the luminescent signal, as compared to that of Rluc, for all substrates tested (compare Figs. 3b and 5c), allowing the possibility of longer acquisition times to monitor the kinetics of biological processes (Fig. 5d). This characteristic has been taken advantage of in recent studies monitoring ligand-binding kinetics using BRET-based assays with Nluc57,58, as well as for real-time BRET imaging of processes occurring over a timescale going from minutes to an hour25. However, Nluc cannot be used for ebBRET experiments, and thus Rluc still represents an advantageous alternative energy donor that is useful for some applications. It should also be noted that many biosensors using Rluc have already been developed and validated and can be readily used for BRET imaging. Novel Nluc-based sensors will certainly be developed but will require rigorous validation before they become available for imaging.

How to select a BRET donor and acceptor

The choice of the donor–acceptor pairs to use for BRET imaging largely depends on the specific process being imaged. Whether or not real-time imaging is sought, the timescale of imaging needed and the availability of already-validated biosensors are parameters that will influence the choice of BRET pairs.
For many applications, Rluc can be used as the donor in BRET1 (substrate: CTZh; acceptor: Venus) or BRET2 (substrate: CTZ400A or Me-O-e-CTZ; acceptor: GFP10 or GFP2) configurations. If the expression level of the protein fused to the energy donor is low, BRET1 using Venus as the acceptor would be preferable to GFP10 or GFP2 because it provides higher light output. However, if sufficient luminescence can be detected using blue-shifted substrates for Rluc (CTZ400A or Me-O-eCTZ), GFP2 or GFP10 could be better acceptor choices because BRET2 provides a greater dynamic window. In the case in which the luminescence signal generated by Rluc with any substrate is too low, Nluc is the preferred choice when using either A. victoria YFP variants25 or the newly characterized Discosoma coral variants, cyan-excitable orange fluorescent protein (CyOFP) and cyan-excitable RFP (CyRFP) with furimazine or CTZ400A as substrates. These fluorescent protein variants have excitation peaks similar to that of Venus (≈500 nm) but have a larger Stokes’ shift than Venus, yielding a greater separation between donor and acceptor emissions, thus opening possibilities of developing new generations of BRET imaging sensors59,60. Figure 5d illustrates the use of Nluc to image the dissociation of Gαq-Nluc from CyOFP-Gγ1 upon sustained stimulation with the angiotensin-II type-1 receptor (AT1R) agonist angiotensin II for 20 min. BRET-based ligand-binding assays using Nluc and the red-shifted fluorophore BODIPY 630/650 also took advantage of such larger Stokes’ shifts18. Although Nluc can clearly be advantageous for many applications, the existence of many validated sensors based on Rluc also makes that energy donor an appealing choice in many cases, as long as the expression levels are sufficient.
For real-time imaging applications, although Rluc-based BRET1, BRET 2 or Nluc-based BRET can all be used, the timescale of the process to image will determine the best choice. The imaging time being limited by the intensity and the half-life of the luminescence emission, Nluc-based BRET using either furimazine or CTZ400A would be the preferred energy donor because of its brightness and the long emission half-life (≈30 min), allowing imaging for a few hours. The next best choice would be BRET1, which provides ≈10 times less light than Nluc but three times more light than BRET2, with a half-life of ≈22 min, limiting the imaging to ≈1 h. Finally, BRET2 can also be used, but the lower light output and the shorter half-life of the signal (≈12 min) greatly limit the imaging time, which cannot extend to >≈20 min.
Although fluorescent proteins have been used more frequently as BRET energy acceptors, chemical fluorophores can also be used successfully for imaging19,61. Generally speaking, the spectrometric properties of chemical fluorophores are superior to those of fluorescent proteins, and good acceptors can be found for all BRET donors. However, the methods of conjugating the fluorophore to the protein of choice must be developed and optimized for each sensor.
To monitor protein translocation, ebBRET using Rluc as the donor, rGFP as the acceptor and MeO-e-CTZ as the substrate, is the superior choice. Both the excellent SNR and greater efficiency of transfer (due to the direct association of Rluc and rGFP, Fig. 4a) allow very robust monitoring of protein trafficking (see ‘Anticipated results’) that cannot be readily imaged using BRET1, BRET2 or Nluc-based BRET.

Distinguishing signal from noise

The above sections describe a number of novel donor–acceptor pairs with improved properties that allow their use for spatiotemporally resolved BRET imaging in various conditions. Still, one of the main limitations of BRET imaging remains the low level of the light output, which makes it difficult to distinguish it from noise. The main source of the noise for such low-level signals originates from the statistically random fluctuation of the photon counts that is known as ‘shot noise’. When considering only the shot noise, the SNR increases as the square root of the incoming photon number. It follows that to increase the SNR of BRET images by twofold, the amount of light emitted must be increased by fourfold. Accordingly, the BRET imaging experiment should be designed to maximize the signal output while constraining all other sources of noise to a minimum. The different sections of the protocol present technical and experimental procedures aimed at obtaining the best possible spatiotemporal resolution by limiting background noise, increasing the light output of the luciferase (see sections on donors and substrates above), improving the separation between donor and acceptor signals (see section on acceptors above), optimizing the light transmission of microscope optics and finally increasing the sensitivity of the detector. These latter aspects are discussed below.

Microscope setup

There are several bioluminescence imaging systems currently available on the market. These include the Olympus LV200 and the Atto Cellgraph. These systems are designed to minimize the contamination by external light during measurements, thus increasing the SNR. When combined with adequate objective lenses and sensitive cameras, these systems offer excellent performance. However, most wide-field microscope setups equipped with a highly sensitive camera should be amenable to performing BRET microscopy with limited modifications. Although the images collected for this report were obtained using a regular inverted microscope, more sophisticated microscopes equipped for photoactivated localization microscopy (PALM), stochastic optical reconstruction microscopy (STORM), total internal reflection fluorescence (TIRF) or calcium imaging should also be suitable for BRET microscopy.
To minimize the contaminating lights (background noise), the imaging equipment should be placed in a dark room, and the system should be further shielded from stray light by inserting the microscope into a dark box covering the microscope body. All openings of the dark box should be covered by dark seals hermetic to light. All the light from the equipment within the dark box should be turned off or masked, and the use of light-emitting devices in the dark room should be minimal. All illumination sources (pilot lamps and indicators) in the dark room should be masked, except for a maximally dimmed computer screen needed for monitoring the image during the acquisition. To verify the influence of external light, images obtained in the absence of sample with the camera shutter closed should be compared to those obtained when the shutter is open. When comparing mean signal intensity between these two acquisitions, differences of ≤1 photon per pixel per min should be targeted using a non-amplified camera or EMCCD in photon-counting mode.
The microscope should be equipped with objective lenses having high light-collecting capacity. The image brightness is known to be influenced by three objective lens parameters: transmission (TR), numerical aperture (NA) and magnification (M), and is determined by the following formula:
Therefore, they are the most important parameters to consider in selecting an objective lens for BRET imaging. Optimal lens magnification (M) is dependent on the surface area of the detector used. We usually combine a ×60 or ×100 lens with a 13 × 13-mm detector. It should be noted that sample brightness using a ×100 lens is ~30% of that obtained with a ×60 lens. The objective lens should be an oil immersion or water immersion lens having the highest available NA. Most high-quality objective lenses have high light transmission (TR) properties across the entire visible light spectrum. However, some lenses may have lower transmission in BRET2 assays, which emits violet light, which is at the lower edge of the visible spectrum. However, this limitation can be partly solved using lenses specifically designed for fluorescence measurement, because they tend to have higher transmission in the UV–violet range. The ×100 objective lens that we are using for BRET imaging (CFI Apochromat TIRF, Nikon) has a TR that drops from 82% at 510 nm to 48% at 400 nm. The relative transmission efficacy of the lenses can vary substantially between lens types and should be evaluated to select the best ones for the BRET pairs used as a function of the wavelength to be monitored.

Detector

Owing to the low signal, BRET imaging requires very sensitive cameras. The detection sensitivity of the camera is mainly determined by two parameters: (i) the quantum efficiency (QE) and (ii) the pixel size of the detector. In many scientific cameras, the peak QE at 500–600 nm is very high (80–90%), but drops rapidly to 10–60% in the UV–violet range, greatly compromising the detection sensitivity and thus the quality of the images when using a short-wavelength donor or acceptor. This is of particular concern for BRET2 assays, which require imaging of 400-nm violet signals originating from the luminescent donor. Therefore, the spectral response of the camera should be examined very carefully to select its QE characteristics as a function of the type of BRET imaging assay considered. The pixel size is the surface area of a single pixel on the sensor unit. Depending on the camera, it may range from 5 × 5 μm to 24 × 24 μm. The larger surface area has more chances to catch a photon, and improves the signal captured per pixel. Cameras having binning functions can bundle several pixels and treat the signal as if it were originating from a single pixel. Although larger pixels allow the detection of lower signals, making the camera more sensitive, it results in a lower-resolution image. The choice of a given pixel size is therefore always a compromise between sensitivity and resolution. We usually use a 170-μm2 (13 × 13 μm) or 680-μm2 (26 × 26 μm, 2 × 2 binning of 13 × 13 μm) pixel surface area using an EMCCD or CCD camera, respectively. Ideally, the size of the detector chip used should be ~13 × 13 mm (1,024 × 1,024 pixels of 13 × 13-μm pixels). This is based on the standard field of view (FOV) of the microscope (18 mm). Depending on the combination of microscope system and objective lens, larger detector chips might be used without overfilling the FOV.
The typical signal detected in BRET imaging is a couple of photons per pixel per second (1–10 e−/pixel/s) in our system. To limit the impact of the shot noise to a reasonably low level (<10 dB), we usually adjust exposure time so that the area of interest from the image corresponds to at least 100 photons captured per pixel. In addition to shot noise, another type of noise reducing image quality is known as ‘thermal noise’ or ‘dark current’ and corresponds to the random signal generated by thermal electrons produced by the camera itself. To reduce the thermal noise, the camera should be equipped with a cooling unit. In deep-cooled (typically below −60 °C) cameras, thermal noise is usually <0.01 e−/pixel/s, which is small enough for good BRET imaging. For the images presented in this report, we used either thermoelectric cooling (Pixis camera) or liquidnitrogen cooling (NuVu camera).
The final distinguishing factor between cameras is the ‘readout noise’, which originates from the signal processing of the camera circuit. EMCCD cameras are becoming increasingly popular for lowlight imaging because they have a lower readout noise relative to the signal. This is achieved by amplifying the photoelectron signals with high gain. However, the multiplication process of EMCCD cameras also amplifies other sources of noise62; it is therefore recommended to use an EMCCD camera in photon-counting mode63 in order to mitigate this negative effect of the high-gain amplification. Using a photon-counting strategy is also useful in minimizing the impact of cosmic rays on the image. Indeed, because the photon-counting image is generated as an integration of many (500–1,000) binary images with very short exposure, the high-energy signal from the cosmic ray is treated as a single-photon entry and therefore is diluted in the entire integrated image. Table 1 shows the comparison of the total noise level between CCDs and EMCCDs used in photon-counting mode or conventional mode. When the signal is strong enough, the noise level is similar between the two types of camera, independent of the detection mode because it reflects mainly the shot noise. By contrast, when the signal is very low, an EMCCD in photon-counting mode shows a much lower noise level than a CCD or an EMCCD in conventional mode. As a result, an EMCCD in photon-counting mode has a lower detection limit. This probably reflects the stochastic noise known as excess noise factor62. Photon counting is, indeed, known to be an effective way to reduce such noise in low-light condition63. Another advantage of EMCCD photon-counting mode over the conventional electron-multiplying (EM) mode results from the less detrimental influence of baseline signal drift that is sometimes observed as a function of imaging time, because the threshold used for photon counting is substantially higher than the baseline signal.
An EMCCD is also preferable to a CCD when short exposure times are required, as in the case of real-time imaging. This is well illustrated in Fig. 6a, where a weak luminescent signal could be detected for exposure times as short as 0.1–0.2 s using an EMCCD camera, but could not be detected with an exposure time <0.5–1.0 s with a CCD. Yet when the signal is of sufficient intensity, goodquality images can be obtained with both CCDs and EMCCDs. Indeed, Fig. 6b shows that the BRET between the G-protein subunits Gαq-RlucII and GFP10-Gγ1 expressed at the cell surface of HEK293 cells could be readily imaged with both cameras.

Anticipated results

Examples of BRET images that can be expected are provided in Figs. 4–7. A comparison of BRET image intensity obtained in three different BRET modes (BRET1, BRET2 and ebBRET) is presented in Fig. 4. The data were obtained with constructs genetically fusing the energy donor to the acceptor, providing excellent controls to test the imaging systems. Although ebBRET provides the brightest images, it cannot be used to monitor specific interactions between protein partners because it takes advantage of the self-assembly of Rluc and rGFP when they are present in the same compartment and could promote interactions between proteins that do not interact with one another normally. Thus, it is mainly suitable for monitoring translocation between compartments.
When comparing BRET1 and BRET2, although weaker luminescence is emitted in BRET2, the better separation of the acceptor and donor signals clearly increases the dynamic window of the BRET signal, allowing better imaging as compared to those produced with BRET1. This is illustrated in Fig. 4a, which shows images of total luminescence emitted (left panels), the light emitted by the acceptor (middle panels) and the calculated BRET signal (right panels) for cells expressing the donor alone (β-arrestin2-RlucII, rows 1 and 3) or a fusion between RlucII and Venus (BRET1; row 2) or GFP2 (BRET2; row 4) upon addition of CTZh (rows 1 and 2) or Me-O-e-CTZ (rows 3 and 4). As can be readily observed, a much greater background BRET signal is observed in BRET1 in the absence of acceptor (Fig. 4a, top row, right image and Fig. 4b, BRET1 donor only). This results from the overlap between the wide emission spectrum of Rluc and the emission of Venus, yielding a greater contamination of the donor emission signal in the acceptor channel (Fig. 4a, top row, middle image). A much lower background BRET is observed in BRET2. Because similar maximal BRET signals are observed for BRET1 and BRET2 (Fig. 4b, donor + acceptor), the lower background results in a much greater dynamic BRET window (maximal/background signals) for BRET2 (6.5-fold in BRET2 versus 1.7 in BRET1) (Fig. 4b). It follows that, although BRET1 generates brighter and longer-lasting signals that can be useful for imaging proteins expressed at low levels for extended periods of time, the dynamic window offered by BRET2 allows better separation between background and specific signals (Fig. 4c), making it a better choice for imaging when the process studied leads to small differences in BRET signals. Also shown in Fig. 4 is the fact that the BRET dynamic window observed for ebBRET is larger than those of both BRET2 and BRET1, owing to the high transfer efficiency observed for this donor–acceptor pair, making it the best choice to monitor protein translocation (see below).
A wide range of biological processes can be imaged by different BRET modalities. For example, we used BRET2 to monitor the activation of a heterotrimeric G protein (Gαqβ1γ1) by AT1R (Fig. 7a). The separation between Gα and Gβγ can be detected by measuring the decrease in BRET signal between the Gαq tagged with RlucII and the Gγ tagged with GFP10 following the activation of the receptor by angiotensin. This type of approach can be used to monitor the dynamic regulation of any protein–protein interaction. Both BRET increase and BRET decrease can be monitored, depending on the effect of a particular stimulus on the interaction. In Fig. 7b, we illustrate the use of ebBRET to image the subcellular redistribution of AT1R and the regulatory protein β-arrestin following activation with the AT1R agonist angiotensin-II. The translocation of β-arrestin to the plasma membrane upon receptor activation is visualized by monitoring the BRET between β-arrestin2-RlucII and rGFP tagged with a CAAX box from KRas (rGFP-CAAX)64 that targets it to the plasma membrane (Fig. 7b, top panels). The ensuing agonist-promoted endocytosis of the receptor that occurs can also be detected by imaging ebBRET between AT1R-RlucII and rGFP targeted to either the plasma membrane (rGFP-CAAX) or the early endosomes, using the early endosome localization motif, FYVE from endofin65 (Fig. 7b, middle and bottom panels, respectively). The use of rGFP selectively targeted to distinct subcellular organelles allows monitoring of the localization of an Rluc-tagged protein in these specific organelles. Here, examples for plasma membrane and endosome are provided, but similar experiments can be done for other compartments, such as endoplasmic reticulum, Golgi, nucleus, mitochondria.
The comparison between control and angiotensin-stimulated conditions for β-arrestin translocation (Fig. 7b, top panel) was done using the same cell population by taking images before and after treatment. This is possible because the recruitment of β-arrestin to the plasma membrane is relatively rapid following agonist treatment and thus can be monitored in real time before the luminescence signal decays to levels that compromise image quality. However, in the case of the receptor endocytosis (Fig. 7b, middle and bottom panels), the control and angiotensin conditions needed to be imaged in different cell populations because the luminescence signal decayed to levels incompatible with quality imaging before reliable endocytosis could be observed. It is important to note that statistically significant differences can be obtained between control and stimulated conditions, both when assessing the phenomenon in the same cell and when assessing it in different sets of cells (Supplementary Fig. 1).
Quantification of BRET signals in different subcellular compartments can also be achieved. This is illustrated in Fig. 8, which shows the dissociation of Gαq-Rluc from GFP10-Gγ1 upon stimulation of AT1R. For this purpose, bright cells from the FOV are manually segmented (Fig. 8a, cell mask), and the BRET is determined for each cell by dividing the light signal emitted in the acceptor channel by the total light detected. The individual BRET values for each cell are then averaged, and the values obtained under basal and receptor-stimulated conditions are compared. As shown in Fig. 8b, receptor stimulation leads to a statistically significant reduction in BRET. Next, the quantification is performed on a subcellular domain by dividing each cell area into peripheral and a central regions (Fig. 8a, central 50% and peripheral 50%) so as to isolate the signal largely originating from the plasma membrane. Figure 8c illustrates the distribution of pixel-by-pixel BRET levels for the population of cells in the FOV, expressed as histograms for both untreated and angiotensin-stimulated conditions. As can be seen, receptor activation leads to a reduction of the frequency of high BRET pixels. When considering only the signal coming from the periphery, receptor stimulation leads to a statistically significant reduction of the BRET signal (histogram in Fig. 8d), indicating that the receptor-promoted dissociation of Gαq-Rluc from GFP10-Gγ1 occurring at the cell surface can be detected and quantified. The raw image data, the cell masks and the MATLAB script used for this quantification are available as Supplementary Data 1.

References

1. Förster, T. Zwischenmolekulare energiewanderung und fluoreszenz. Ann. Phys. 437, 55–75 (1948).
2. Jares-Erijman, E. A. & Jovin, T. M. FRET imaging. Nat. Biotechnol. 21, 1387–1395 (2003).
3. Berney, C. & Danuser, G. FRET or no FRET: a quantitative comparison. Biophys. J. 84, 3992–4010 (2003).
4. Angers, S. et al. Detection of beta 2-adrenergic receptor dimerization in living cells using bioluminescenceresonance energy transfer (BRET). Proc. Natl. Acad. Sci. USA 97, 3684–3689 (2000).
5. 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).
6. Ayoub, M. A. et al. Monitoring of ligand-independent dimerization and ligand-induced conformationalchanges of melatonin receptors in living cells by bioluminescence resonance energy transfer. J. Biol. Chem. 277, 21522–21528 (2002).
7. Mercier, J.-F., Salahpour, A., Angers, S., Breit, A. & Bouvier, M. Quantitative assessment of beta 1- and beta2-adrenergic receptor homo- and heterodimerization by bioluminescence resonance energy transfer. J. Biol. Chem. 277, 44925–44931 (2002).
8. Galés, C. et al. Real-time monitoring of receptor and G-protein interactions in living cells. Nat. Methods 2, 177–184 (2005).
9. Galés, C. et al. Probing the activation-promoted structural rearrangements in preassembled receptor-Gprotein complexes. Nat. Struct. Mol. Biol. 13, 778–786 (2006).
10. Kobayashi, H., Ogawa, K., Yao, R., Lichtarge, O. & Bouvier, M. Functional rescue of beta-adrenoceptordimerization and trafficking by pharmacological chaperones. Traffic 10, 1019–1033 (2009).
11. Hamdan, F. F., Audet, M., Garneau, P., Pelletier, J. & Bouvier, M. High-throughput screening of G proteincoupled receptor antagonists using a bioluminescence resonance energy transfer 1-based beta-arrestin2 recruitment assay. J. Biomol. Screen. 10, 463–475 (2005).
12. Charest, P. G. & Bouvier, M. Palmitoylation of the V2 vasopressin receptor carboxyl tail enhances betaarrestin recruitment leading to efficient receptor endocytosis and ERK1/2 activation. J. Biol. Chem. 278, 41541–41551 (2003).
13. Terrillon, S., Barberis, C. & Bouvier, M. Heterodimerization of V1a and V2 vasopressin receptors determinesthe interaction with beta-arrestin and their trafficking patterns. Proc. Natl. Acad. Sci. USA 101, 1548–1553 (2004).
14. Bertrand, L. et al. The BRET2/arrestin assay in stable recombinant cells: a platform to screen for compoundsthat interact with G protein-coupled receptors (GPCRS). J. Recept. Signal Transduct. Res. 22, 533–541 (2002).
15. Azzi, M. et al. Beta-arrestin-mediated activation of MAPK by inverse agonists reveals distinct active conformations for G protein-coupled receptors. Proc. Natl. Acad. Sci. USA 100, 11406–11411 (2003).
16. Héroux, M., Breton, B., Hogue, M. & Bouvier, M. Assembly and signaling of CRLR and RAMP1 complexesassessed by BRET. Biochemistry 46, 7022–7033 (2007).
17. Perroy, J., Pontier, S., Charest, P. G., Aubry, M. & Bouvier, M. Real-time monitoring of ubiquitination inliving cells by BRET. Nat. Methods 1, 203–208 (2004).
18. Stoddart, L. A. et al. Application of BRET to monitor ligand binding to GPCRs. Nat. Methods 12, 661–663 (2015).
19. Stoddart, L. A., Kilpatrick, L. E. & Hill, S. J. NanoBRET approaches to study ligand binding to GPCRs andRTKs. Trends Pharmacol. Sci. 39, 136–147 (2018).
20. Loening, A. M., Fenn, T. D., Wu, A. M. & Gambhir, S. S. Consensus guided mutagenesis of Renilla luciferase yields enhanced stability and light output. Protein Eng. Des. Sel. 19, 391–400 (2006).
21. Hall, M. P. et al. Engineered luciferase reporter from a deep sea shrimp utilizing a novel imidazopyrazinonesubstrate. ACS Chem. Biol. 7, 1848–1857 (2012).
22. Xu, X. et al. Imaging protein interactions with bioluminescence resonance energy transfer (BRET) in plantand mammalian cells and tissues. Proc. Natl. Acad. Sci. USA 104, 10264–10269 (2007).
23. Coulon, V. et al. Subcellular imaging of dynamic protein interactions by bioluminescence resonance energytransfer. Biophys. J. 94, 1001–1009 (2008).
24. Kim, J. & Grailhe, R. Nanoluciferase signal brightness using furimazine substrates opens bioluminescenceresonance energy transfer to widefield microscopy. Cytometry A 89, 742–746 (2016).
25. Goyet, E., Bouquier, N., Ollendorff, V. & Perroy, J. Fast and high resolution single-cell BRET imaging.Sci. Rep. 6, 28231 (2016).
26. Namkung, Y. et al. Monitoring G protein-coupled receptor and β-arrestin trafficking in live cells using enhanced bystander BRET. Nat. Commun. 7, 12178 (2016).
27. Beautrait, A. et al. A new inhibitor of the β-arrestin/AP2 endocytic complex reveals interplay between GPCR internalization and signalling. Nat. Commun. 8, 15054 (2017).
28. De, A., Ray, P., Loening, A. M. & Gambhir, S. S. BRET3: a red-shifted bioluminescence resonance energytransfer (BRET)-based integrated platform for imaging protein-protein interactions from single live cells and living animals. FASEB J. 23, 2702–2709 (2009).
29. Yeh, H.-W. et al. Red-shifted luciferase-luciferin pairs for enhanced bioluminescence imaging. Nat. Methods 14, 971–974 (2017).
30. Fredriksson, S. et al. Protein detection using proximity-dependent DNA ligation assays. Nat. Biotechnol. 20, 473–477 (2002).
31. Kerppola, T. K. Visualization of molecular interactions by fluorescence complementation. Nat. Rev. Mol. Cell Biol. 7, 449–456 (2006).
32. Pelletier, J. N., Campbell-Valois, F. X. & Michnick, S. W. Oligomerization domain-directed reassembly ofactive dihydrofolate reductase from rationally designed fragments. Proc. Natl. Acad. Sci. USA 95, 12141–12146 (1998).
33. Ozawa, T., Kaihara, A., Sato, M., Tachihara, K. & Umezawa, Y. Split luciferase as an optical probe fordetecting protein–protein interactions in mammalian cells based on protein splicing. Anal. Chem. 73, 2516–2521 (2001).
34. Paulmurugant, R. & Gambhir, S. S. Monitoring protein–protein interactions using split synthetic Renilla luciferase protein-fragment-assisted complementation. Anal. Chem. 75, 1584–1589 (2003).
35. Dixon, A. S. et al. NanoLuc complementation reporter optimized for accurate measurement of proteininteractions in cells. ACS Chem. Biol. 11, 400–408 (2016).
36. Kerppola, T. K. Design and implementation of bimolecular fluorescence complementation (BiFC) assays for the visualization of protein interactions in living cells. Nat. Protoc. 1, 1278–1286 (2006).
37. Rebois, R. V. et al. Combining protein complementation assays with resonance energy transfer to detectmultipartner protein complexes in living cells. Methods 45, 214–218 (2008).
38. Héroux, M., Hogue, M., Lemieux, S. & Bouvier, M. Functional calcitonin gene-related peptide receptors areformed by the asymmetric assembly of a calcitonin receptor-like receptor homo-oligomer and a monomer of receptor activity-modifying protein-1. J. Biol. Chem. 282, 31610–31620 (2007).
39. Armando, S. et al. The chemokine CXC4 and CC2 receptors form homo- and heterooligomers that canengage their signaling G-protein effectors and βarrestin. FASEB J. 28, 4509–4523 (2014).
40. Fichter, K. M., Flajolet, M., Greengard, P. & Vu, T. Q. Kinetics of G-protein-coupled receptor endosomaltrafficking pathways revealed by single quantum dots. Proc. Natl. Acad. Sci. USA 107, 18658–18663 (2010).
41. Ahn, S., Shenoy, S. K., Wei, H. & Lefkowitz, R. J. Differential kinetic and spatial patterns of β-arrestin and G protein-mediated ERK activation by the angiotensin II receptor. J. Biol. Chem. 279, 35518–35525 (2004).
42. Lohse, M. J., Maiellaro, I. & Calebiro, D. Kinetics and mechanism of G protein-coupled receptor activation.Curr. Opin. Cell Biol. 27, 87–93 (2014).
43. Breton, B. et al. Multiplexing of multicolor bioluminescence resonance energy transfer. Biophys. J. 99, 4037–4046 (2010).
44. Leduc, M. et al. Functional selectivity of natural and synthetic prostaglandin EP4 receptor ligands. J. Pharmacol. Exp. Ther. 331, 297–307 (2009).
45. Rodriguez, E. A. et al. The growing and glowing toolbox of fluorescent and photoactive proteins. Trends Biochem. Sci. 42, 111–129 (2017).
46. Inouye, S. & Shimomura, O. The use of Renilla luciferase, Oplophorus luciferase, and apoaequorin as bioluminescent reporter protein in the presence of coelenterazine analogues as substrate. Biochem. Biophys. Res. Commun. 233, 349–353 (1997).
47. Nagai, T. et al. A variant of yellow fluorescent protein with fast and efficient maturation for cell-biological applications. Nat. Biotechnol. 20, 87–90 (2002).
48. Pfleger, K. D. G. & Eidne, K. A. Illuminating insights into protein-protein interactions using bioluminescence resonance energy transfer (BRET). Nat. Methods 3, 165–174 (2006).
49. Molinari, P., Casella, I. & Costa, T. Functional complementation of high-efficiency resonance energy transfer: a new tool for the study of protein binding interactions in living cells. Biochem. J. 409, 251–261 (2008).
50. Wampler, J. E., Hori, K., Lee, J. W. & Cormier, M. J. Structured bioluminescence. Two emitters duringboth the in vitro and the in vivo bioluminescence of the sea pansy, Renilla. Biochemistry 10, 2903–2909 (1971).
51. Yamakawa, Y., Ueda, H., Kitayama, A. & Nagamune, T. Rapid homogeneous immunoassay of peptides basedon bioluminescence resonance energy transfer from firefly luciferase. J. Biosci. Bioeng. 93, 537–542 (2002).
52. Li, F. et al. Buffer enhanced bioluminescence resonance energy transfer sensor based on Gaussia luciferase for in vitro detection of protease. Anal. Chim. Acta 724, 104–110 (2012).
53. Inouye, S., Sato, J., Sahara-Miura, Y., Yoshida, S. & Hosoya, T. Luminescence enhancement of the catalytic19kDa protein (KAZ) of Oplophorus luciferase by three amino acid substitutions. Biochem. Biophys. Res. Commun. 445, 157–162 (2014).
54. Inouye, S., Watanabe, K., Nakamura, H. & Shimomura, O. Secretional luciferase of the luminous shrimpOplophorus gracilirostris: cDNA cloning of a novel imidazopyrazinone luciferase. FEBS Lett. 481, 19–25 (2000).
55. Czupryna, J. & Tsourkas, A. Firefly luciferase and Rluc8 exhibit differential sensitivity to oxidative stress in apoptotic cells. PLoS ONE 6, e20073 (2011).
56. Shigehisa, M. et al. Stabilization of luciferase from Renilla reniformis using random mutations. Protein Eng. Des. Sel. 30, 7–13 (2017).
57. Hu, M.-J. et al. Development of a novel ligand binding assay for relaxin family peptide receptor 3 and 4 usingNanoLuc complementation. Amino Acids 50, 1111–1119 (2018).
58. Stoddart, L. A. et al. Development of novel fluorescent histamine H1-receptor antagonists to study ligandbinding kinetics in living cells. Sci. Rep. 8, 1572 (2018).
59. Chu, J. et al. A bright cyan-excitable orange fluorescent protein facilitates dual-emission microscopy and enhances bioluminescence imaging in vivo. Nat. Biotechnol. 34, 760–767 (2016).
60. Laviv, T. et al. Simultaneous dual-color fluorescence lifetime imaging with novel red-shifted fluorescent proteins. Nat. Methods 13, 989–992 (2016).
61. Machleidt, T. et al. NanoBRET—a novel BRET platform for the analysis of protein–protein interactions. ACS Chem. Biol. 10, 1797–1804 (2015).
62. Robbins, M. S. & Hadwen, B. J. The noise performance of electron multiplying charge-coupled devices. IEEE Trans. Electron Devices 50, 1227–1232 (2003).
63. Basden, A. G., Haniff, C. A. & Mackay, C. D. Photon counting strategies with low-light-level CCDs. Mon.Not. R. Astron. Soc. 345, 985–991 (2003).
64. Zacharias, D. A., Violin, J. D., Newton, A. C. & Tsien, R. Y. Partitioning of lipid-modified monomeric GFPs into membrane microdomains of live cells. Science 296, 913–916 (2002).
65. Schink, K. O., Raiborg, C. & Stenmark, H. Phosphatidylinositol 3-phosphate, a lipid that regulates membranedynamics, protein sorting and cell signalling. Bioessays 35, 900–912 (2013).