mBeRFP: a versatile fluorescent tool to enhance multichannel live imaging and its applications

The field of cell and developmental biology relies more and more on the analysis of protein dynamics in cells, tissues or organisms using live imaging. In addition to major advances in fluorescence microscopy, the discovery of the green fluorescent protein (GFP) in the jellyfish Aequorea victoria (Shimomura et al., 1962) and its first use as a fluorescent marker in non-jellyfish organisms (Chalfie et al., 1994) were a crucial breakthrough for the field of non-invasive protein imaging. Since the discovery of GFP, many other fluorescent proteins (FPs) have been identified, developed and enhanced, providing a powerful toolkit for visualization of dynamic processes in vivo (Chudakov et al., 2010). However, in classical fluorescence imaging, i.e. without a spectral detector and linear unmixing, the number of FPs that can be tracked in a living tissue is often limited by the intrinsic spectral properties of FPs, as well as by the instrumentation available. Indeed, to detect specifically two FPs simultaneously, their excitation and emission peak should be separated by at least 60 nm (Thorn, 2017), so usually only the pairs cyan fluorescent protein (CFP)/yellow fluorescent protein (YFP), or the commonly used GFP/red fluorescent protein (RFP) are imaged together. In some cases, with an adapted set of filters, a third color can be added with little crosstalk, such as RFP with the couple CFP/YFP (Boulina et al., 2013), or blue fluorescent protein (BFP) (mTagBFP2; Subach et al., 2011) with GFP/RFP. However, in this latter case, near-UV excitation is required, which is highly phototoxic for long-term live-imaging experiments (Gorgidze et al., 1998; Icha et al., 2017). Thus, combining more than two colors for live imaging with low phototoxicity, sufficient brightness and photostability of FPs appears challenging, especially in model organisms in which the environment differs greatly from assessments performed in vitro to characterize the properties of these FPs (Heppert et al., 2016; Wiedenmann et al., 2009).

During the last two decades, an interest in monomeric red FPs characterized by a large Stokes shift (LSS RFP, i.e. a difference between the emission peak and the excitation peak larger than 150 nm) appeared. These proteins are particularly interesting because they can be excited by blue light, allowing the simultaneous excitation of the LSS RFP with CFP or GFP. Among these LSS RFP, mKeima (Kogure et al., 2006), LSSmKate1, LSSmKate2 (Piatkevich et al., 2010a,b) and mBeRFP (Yang et al., 2013) were generated. However, their full potential in living organisms has not yet been tested.

Here, we provide a detailed characterization of mBeRFP for in vivo live-imaging applications in Drosophila. First, we cloned and expressed in flies mBeRFP, and showed that it can be easily combined with green and red FPs without crosstalk. We demonstrate that mBeRFP can be combined with enhanced GFP (eGFP) in live imaging, offering the interesting possibility of imaging the two fluorophores using a single light source. We also discuss the potential uses of this protein for quantitative and biophysical applications.

To test the potential of fluorescent extended Stokes shift proteins in Drosophila, we constructed a set of plasmids expressing either cytoplasmic or nuclear mBeRFP under the control of a UAS promoter (Fig. 1A), as well as a cytoplasmic LSSmKate2, and generated the corresponding transgenic flies. We expressed these constructs using different Gal4 promoters (Fig. 1B,C; Fig. S1A-C) and showed that, at the same laser power, the fluorescence intensity of the LSSmKate2 was not as high as that of mBeRFP (Fig. S1C,D), so we focused on mBeRFP. Then, to ensure that the expression of mBeRFP did not induce developmental defects or cytotoxicity, we followed tissue elongation in pupal leg disk expressing or not expressing cytoplasmic mBeRFP. This revealed that morphogenesis was not affected by the expression of mBeRFP, as shown by the normal elongation of the leg and the correct formation of the folds (Fig. 1D; Movie 1).

We then tested the possibility of using mBeRFP as a tag to follow endogenous protein. As a proof of concept, we generated an α-catenin-mBeRFP fusion protein and used α-catenin-TagRFP as a control. Although mBeRFP was less fluorescent than TagRFP when imaged using a similar laser power on fixed samples (Fig. 1G), no significant difference in terms of location and morphogenesis was observed between these two fusion proteins (Fig. 1E,F). These data suggest that mBeRFP could be used in vivo in Drosophila. In addition, we developed ‘promoter-free’ versions of mBeRFP (Fig. S1E).

To investigate the possibility of using mBeRFP for live imaging, we determine whether (1) it can be detected at low laser power to ensure tissue viability, (2) it does not affect the dynamics of protein of interest and (3) it is not too sensitive to photobleaching. To check these characteristics, we used the α-catenin-mBeRFP fusion protein. We first followed the adherens junction dynamics in the apterous domain during the leg development (Fig. 2A; Movie 2). At large scale, tissue dynamics appeared unperturbed all along the movie. At a cellular scale, we were able to follow the extrusion of apoptotic cells from the apical surface of the leg disk (Fig. 2A′), with a duration similar to our previous observations (Schott et al., 2017). This showed that mBeRFP could be efficiently detected and used to follow dynamic processes in living samples.

Then, we assessed the impact of the mBeRFP fusion on protein dynamics by performing fluorescence recovery after photobleaching (FRAP) experiments on α-catenin fused to mBeRFP, TagRFP or GFP (Fig. 2B-E). This technique allows the kinetics of diffusion of a protein of interest to be evaluated by monitoring the recovery of fluorescent signal following photobleaching. FRAP on α-catenin-mBeRFP junctions indicated that the mobile fraction and the half time recovery were not significantly different from those of α-catenin-TagRFP or α-catenin-GFP, suggesting that the protein dynamics is mainly unaffected by the fusion of mBeRFP. Of note, the mobile fraction of α-catenin-TagRFP was significantly lower than that of α-catenin-GFP.

Next, to address the question of photobleaching, we imaged wing disks expressing α-catenin-GFP and α-catenin-mBeRFP and measured the fluorescence intensity at each time frame, when excited with a 458 nm laser (Fig. 2F,G). In this experiment, the fluorescence of both α-catenin-GFP and α-catenin-mBeRFP was stable over time. The same experiment was carried out with wings expressing both E-cadherin-GFP and nuclear mBeRFP with the same result (Fig. S2; Movie 3), providing reassuring evidence that this fluorescent protein can be used for live imaging.

To ensure that mBeRFP can be used in combination with classical FPs, we analyzed the extent of crosstalk in fixed tissues under optimal conditions for fluorescence detection. To test the combination with green-emitting FPs, we used eGFP. Theoretically, eGFP and mBeRFP can be coupled without crosstalk using adapted emission windows (Fig. 3A). In practice, we fixed emission windows from 490 nm to 550 nm for eGFP and from 600 nm to 700 nm for mBeRFP (Fig. 3A). Although the maximal excitation wavelength differs between these two proteins, the higher brightness of eGFP compared with mBeRFP (Thorn, 2017) offers the interesting possibility of exciting the two proteins with the same wavelength. Indeed, at 458 nm wavelength we could excite mBeRFP to more than 90% and eGFP to 60% (Fig. 3A). Then, we imaged at the same time α-catenin-mBeRFP and MyoII-GFP using either a classical confocal mode or a spectral mode to completely avoid potential spectral fluorescence overlap and crosstalk (Fig. 3B) (Zimmermann et al., 2003). The crosstalk measurement revealed a total absence of crosstalk between eGFP and mBeRFP with these settings (Fig. 3C), confirming the possible combination of these two proteins. To test the potential impact of the presence of mBeRFP on GFP fluorescence, we measured the fluorescence of the MyoII-GFP inside or outside the expression domain of α-catenin-mBeRFP. We observed no difference between the two domains (Fig. 3D,E), suggesting that the presence of mBeRFP does not affect GFP fluorescence. To test further the specificity and precision in terms of spatial distribution of the two fusion proteins eGFP and mBeRFP using a single excitation wavelength, we used random illumination microscopy. This newly developed super-resolution technique offers the possibility of achieving unprecedented spatial resolution (around 100 nm) in living tissues (Mangeat et al., 2021). We concomitantly imaged MyoII-GFP and α-catenin-mBeRFP with a single illumination source at 445 nm and were able to resolve distinctly the two cortical myosin belts of neighboring cells separated by the adherens junctions (Fig. S3A,B).

We also tested the possibility of combining mBeRFP with red-emitting FPs. To do so, we compared the emission rate of mBeRFP and TagRFP when excited at 561 nm (Fig. 3A). We imaged leg disks expressing either α-catenin-TagRFP or α-catenin-mBeRFP and compared the fluorescence intensity outside and inside the expression domain (Fig. 3F). As expected, the fluorescence intensity of α-catenin-TagRFP was high inside the expression domain while the fluorescence intensity of α-catenin-mBeRFP was very close to the background intensity. We noted, however, a significant residual collection of fluorescence that could be the consequence of the existence of a second excitation peak of the mBeRFP protein at 580 nm, as characterized from purified protein (Yang et al., 2013). To extend this analysis, we measured the crosstalk between mBeRFP and TRITC dye after fixation of α-catenin-mBeRFP-expressing leg and staining for actin with phalloidin-TRITC. Similar to the eGFP/mBeRFP pair, we imaged the mBeRFP/TRITC pair using either spectral or classical mode when the sample was illuminated with a 561 nm light source. The analysis of the spectral contamination revealed an absence of crosstalk between mBeRFP and TRITC for the settings used in this assay (Fig. S3C,D). Thus, although mBeRFP can be weakly excited at 561 nm, its emission between 570 and 700 nm is close to the background when the laser power is low, suggesting that this fluorescent protein can be easily coupled with red-emitting FPs without disrupting either the observation or the quantification.

In summary, these data showed that mBeRFP can be combined with green or red-emitting FPs or dyes without crosstalk. In addition, this tool could even allow the detection of five distinct channels using four light sources and only three tracks in classical confocal mode (Fig. 3G).

In the following experiments, we chose to focus on the acto-myosin cytoskeleton (MyoII-GFP) together with the adherens junctions (α-catenin-mBeRFP), two very dynamic markers, as an example of simultaneous evaluation of tissue mechanics and cell shape changes, a central interest in the field of mechanobiology. Our aim was to show that this new tool could be of interest for long-term, live-imaging experiments and could be combined easily with other markers. We followed cell division in larval wing disk and confirmed that mBeRFP can be easily combined with GFP-fused protein for live imaging (Fig. 4A; Movie 4). Moreover, using the mBeRFP/GFP pair with a single excitation source rather than the GFP/RFP pair using two different lasers sequentially could (1) greatly reduce phototoxicity because the samples are illuminated at a single wavelength (instead of two different wavelengths) and (2) reduce twice the acquisition time, a crucial parameter for studying protein dynamics.

For FRAP experiments, we decided to bleach these proteins at the level of adherens junctions and followed the fluorescence recovery over time. As expected, we observed a higher stability of the adherens junction protein α-catenin compared with myosin (Fig. 4B; Movie 5) and found approximately the same rate of mobile fraction described in other systems (67% compared with 68% and 25% compared with 22%, respectively for MyoII and α-catenin; Kasza et al., 2014; Yonemura et al., 2010). For laser ablation, in the example shown in Fig. 4C and Movie 6, we were able to follow both the vertex displacement, based on the α-catenin-mBeRFP signal, and the fluorescence intensity of myosin at the same time. This showed that after ablation there is a relaxation of the junction, as expected, followed by a step in which the vertices come together concomitantly with a strong accumulation of myosin at the level of the cut (Fig. 4C′,C″), suggesting the establishment of a myosin-dependent repair mechanism, similarly to what has been observed during wound repair (Antunes et al., 2013; Zulueta-Coarasa and Fernandez-Gonzalez, 2018).

To summarize, we have shown that mBeRFP can be easily used with low photobleaching and combined with classical FPs, such as GFP and RFP, for multiple applications including long-term live imaging and opto-manipulations. Moreover, the fluorescence of this protein is not disrupted by the fixation protocol (Fig. 3G), allowing imaging of an additional channel in fixed tissue, and making this FP a versatile tool for imaging in model organisms. Additionally, the combination of mBeRFP with GFP allows the study of two distinct proteins of interest using the same illumination source. A single illumination source rather than two could be potentially less deleterious for tissues in terms of phototoxicity, even if this last property depends on the wavelength used as well as several other parameters (Icha et al., 2017).

If we consider excitation and emission spectra, mBeRFP could also be coupled with CFP, allowing both proteins to be excited together using a single light source, as previously shown (Yang et al., 2013), and YFP family proteins. Thus, it would potentially be possible to combine the most popular FRET pair – CFP/YFP (Bajar et al., 2016) – with an additional marker coupled to mBeRFP and image them simultaneously using a single illumination source at 458 nm.

Furthermore, considering how close the mBeRFP excitation spectrum is to those of CFP and GFP, we hypothesized that mBeRFP and GFP could be simultaneously observed using multiphoton microscopy, as similarly shown for CFP and GFP (Sahai et al., 2005), although it should offer better detection of each FP. Similarly, it has previously been shown that LSSmKate1, a mBeRFP sister protein, can be combined and simultaneously excited at an optimal wavelength (870 nm) with CFP and GFP for multicolor imaging using two-photon microscopy (Piatkevich et al., 2010a).

Overall, considering the possibility of easily combining mBeRFP with other FPs, its relatively high brightness and stability in living tissues, the wide number of potential applications for live imaging as well as the possibility of using it on fixed tissues, we strongly believe that this fluorescent protein and the versatile constructs we generated will constitute important tools for cell, developmental and mechano-biologists in their current research.

LSSmKate2 and mBeRFP were individually amplified by PCR with specific primers from pLSSmKate2-N1 (Addgene plasmid #31867) and LK1-MpEF1+mBeRFP+Nos-T35S-T (gift from F. Federici, Pontificia Universidad Católica de Chile, Santiago, Chile), and then cloned into the KpnI site of pUASt using In-Fusion technology (Clontech). The forward and reverse primers used to amplify LSSmKate2 were, respectively, GGCCGCGGCTCGAGGGTACCATGAGCGAGCTGATTAAGGAG and AAAGATCCTCTAGAGCTAATTAAGCTTGTGCCCCAGT. The forward primer used to amplify mBeRFP was GGCCGCGGCTCGAGGGTACCATGGTGTCTAAGGGCGA. The reverse primers used were AAAGATCCTCTAGAGTTAATTAAGTTTGTGCCCCAGT or AAAGATCCTCTAGAGTACCTTGCGCTTTTTCTTGGGAGCTCCCTCATTAAGTTTGTGCCCCAGT to generate, respectively, the cytoplasmic or nuclear version of mBeRFP.

To generate these constructs, the same strategy described above was used. Briefly, the mBeRFP coding sequence was cloned into the KpnI site of phsp70, which was derived from the pUASt by removing UAS sequence.

This construct was carried out by amplifying α-catenin coding sequence from a pUAS-Dαcatenin-TagRFP (kindly provided by K. Sugimura, Kyoto University, iCeMS, Japan) then cloning in-frame with mBeRFP in the pUASt-mBeRFP plasmid digested with KpnI, using In-Fusion technology. Forward and reverse primers used to amplify α-catenin coding sequence were, respectively, GGCCGCGGCTCGAGGATGTTAAAACCTGATAAAATGGGCA and CCCTTAGACACCATGGCAACAGCGTCAGCAGGACT.

Transgenic lines carrying UAS-LSSmKate2, UAS-mBeRFP, UAS-NLSmBeRFP and UAS-α-catenin-mBeRFP insertion at attP40 or attP2 landing sites were generated in this study by standard procedures using PhiC31/attB-mediated integration. Injections were performed by the CBI Drosophila facility (Toulouse, France).

ap^md544-Gal4 (BDSC_3041), pdm2-Gal4 (BDSC_49828), E-cad-GFP (BDSC_60584), UAS-α-catenin-GFP (BDSC_58787) were obtained from Bloomington Drosophila Stock Center (BDSC). sqh-eGFP^KI[29B] (named MyoII-GFP in this article) were previously described (Ambrosini et al., 2019). Dll^EM212-Gal4 and UAS-α-catenin-TagRFP were, respectively, gifts from G. Morata (Centro de Biología Molecular Severo Ochoa, Madrid, Spain) and K. Sugimura (Kyoto University, iCeMS, Japan).

Imaginal leg and wing disks were dissected 2 h after pupae formation or at third instar larval stage in 1× PBS. Tissues were fixed in 4% paraformaldehyde for 20 min, then washed in PBS and mounted in Vectashield containing (or not) either DAPI or phalloidin rhodamine (Vector Laboratories). For Fig. 3G, samples were washed in 0.3% Triton X-100, 1% bovine serum albumin in PBS (BBT) after fixation and incubated overnight at 4°C with rat anti-E-cadherin antibody (Developmental Studies Hybridoma Bank, DCAD2) diluted at 1:50 in BBT. After washes in BBT, tissues were incubated with 1:200 anti-rat IgG 647 for 2 h at room temperature with 1:500 phalloidin TRITC (Fisher Scientific). Then, samples were washed in 0.3% Triton X-100 in PBS and mounted on slides in Vectashield with DAPI (Vector Laboratories).

For live imaging, FRAP and laser ablation experiments (Figs 1D, 2A-E and 4B-C″) in leg tissue, imaginal leg disks were dissected from prepupae in Schneider medium supplemented with 15% fetal calf serum, 0.5% penicillin-streptomycin and 2 μg/ml 20-hydroxyecdysone (Sigma-Aldrich, H5142) and mounted on slides.

For live imaging of cell division and photobleaching experiments in wing tissue (Figs 2F,G and 4A; Fig. S2), imaginal wing disks were dissected from early third-instar larvae in Schneider's insect medium supplemented with 15% fetal calf serum, 0.5% penicillin-streptomycin and mounted on slides.

Images were acquired with a Zeiss LSM880 confocal microscope (Carl Zeiss) equipped with a 30 mW 405 nm diode laser, a 34.4 mW 458/488/514 nm argon multiline laser, a 13 mW 561 nm DPSS laser and a 3 mW 633 nm HeNe laser on a Zeiss Axio Observer microscope. The microscope was also fitted with Plan-Apochromat 40×/NA 1.3 Oil DIC UV-IR M27, 63× C-Apochromat NA 1.2 Water Corr (Fig. 2C,D, FRAP and laser ablation) and 40× C-Apochromat NA 1.2 Water Corr (Fig. 2A,B, live imaging and photobleaching assay) objectives. z-stacks were acquired using either the laser-scanning confocal mode or the spectral mode and the adapted illumination source. A linear unmixing process was performed following spectral acquisition to separate each fluorophore based on their spectra.

Fig. 3G was generated using three different tracks. MyoII-GFP and α-catenin-mBeRFP were imaged simultaneously (track 1) using a 458 nm laser with two emission windows (490-510 nm and 642-695 nm, respectively, for GFP and mBeRFP); nuclei (DAPI) and actin (Phalloidin-TRITC) were imaged in a second track using 405 nm and 561 nm lasers, and E-cadherin (Alexa 647) with a 633 nm laser in a third track.

Random illumination microscopy (RIM) was then performed using a home-made system that has been previously described (Mangeat et al., 2021). Images were acquired every 4 ms using an inverted microscope (TEi Nikon) equipped with a 100× magnification, 1.49 N.A. objective (CFI SR APO 100XH ON 1.49 Nikon) and Abbelight two sCMOS camera (ORCA-fusion, Hamamatsu) system band pass filters (Semrock): FF01-514/30-25 for GFP, FF01-630/92-25 for mBeRFP, Fast diode lasers (Oxxius) with wavelength centered at 445 nm (LBX 445 100 CSB OE) were used for the excitation of both GFP and mBeRFP. A spatial light phase binary modulator (QXGA fourth dimensions) was conjugated to the image plane to create speckle random illumination. Image reconstruction was then performed as previously detailed (Mangeat et al., 2021) with the upgraded version of AlgoRIM V1.2: https://github.com/teamRIM/tutoRIM.

Photobleaching was performed on the LSM880 confocal microscope. GFP and/or mBeRFP were excited simultaneously using the 458 nm laser source and photons were collected by GaAsP detectors. Images were acquired from imaginal wing disks every 6 min over 240 min (41 time frames).

FRAP was performed on the LSM880 confocal microscope described above. For FRAP experiments, GFP and/or mBeRFP were excited simultaneously using the 458 nm laser source, while TagRFP was excited with 561 nm, and photons were collected by GaAsP detectors. Photobleaching was set through the bleaching module provided with ZEN software. Briefly, the region of interest (ROI) was illuminated five times using a bleach dwell time of 0.60 μs/pixel with 100% 488 nm or 561 nm laser power. Images were acquired every 900 ms, ten times before bleaching and 100 times post-bleaching.

Laser-ablation experiments were carried out using a pulsed DPSS laser (532 nm, pulse length 1.5 ns, repetition rate up to 1 kHz, 3.5 μJ/pulse) steered by a galvanometer-based laser scanning device (DPSS-532 and UGA-42, from Rapp OptoElectronic, Hamburg, Germany) and mounted on an LSM880 confocal microscope (Carl Zeiss). The microscope was fitted with a 63× C-Apochromat NA 1.2 Water Corr objective (Carl Zeiss). Photo-ablation of the apical junction was carried out in the focal plane by illuminating at 95-98% laser power for 500 ms. Time-lapse images were acquired using a 458 nm light source, every 370 ms, from 5 s before to at least 200 s after ablation, with a pixel size of 0.13 μm/pixel, and photons were collected in two emission windows (from 490 nm to 550 nm for eGFP and from 600 nm to 700 nm for mBeRFP).

To assess the crosstalk, photobleaching or fluorescence intensity, the mean of fluorescence intensity was measured in the ROI using ImageJ (Figs 1G and 2B-G; Figs S1D, S2, S3D).

The crosstalk in classical laser-scanning mode between GFP and mBeRFP was calculated as the difference between the rate of eGFP emitted in the mBeRFP emission window when imaged at 458 nm with the spectral mode and the corresponding rate in the classical mode. The crosstalk in classical laser-scanning mode between mBeRFP and TRITC was calculated as the difference between the rate of mBeRFP emitted in the TRITC emission window when imaged at 561 nm with the spectral mode and the corresponding rate in the classical mode.

Vertex displacement was evaluated by kymograph using a home-made macro in ImageJ.

The normality and the variance of data sets was determined using Prism 8 (GraphPad). A two-tailed unpaired Student's t-test was used to assess the significance in Figs 1G, 2D,E and 3F, and Fig. S1D. A two-tailed paired Student's t-test was used in 3E.

We would like to thank Bruno Monier for his constructive comments on the manuscript and Thomas Mangeat, from the LITC platform, for his help on RIM experiments.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200495.reviewer-comments.pdf.