Single-color, ratiometric biosensors for detecting signaling activities in live cells

Genetically encoded fluorescent biosensors have revolutionized the study of signal transduction by enabling the real-time tracking of signaling activities in live cells. Investigating the interaction between signaling networks has become increasingly important to understanding complex cellular phenomena, necessitating an update of the biosensor toolkit to allow monitoring and perturbing multiple activities simultaneously in the same cell. We therefore developed a new class of fluorescent biosensors based on homo-FRET, deemed FLuorescence Anisotropy REporters (FLAREs), which combine the multiplexing ability of single-color sensors with a quantitative, ratiometric readout. Using an array of color variants, we were able to demonstrate multiplexed imaging of three activity reporters simultaneously in the same cell. We further demonstrate the compatibility of FLAREs for use with optogenetic tools as well as intravital two-photon imaging.


Introduction
Genetically encoded biosensors have revolutionized the study of cell signaling by allowing the realtime monitoring of signaling activities, such as enzymatic activity or the release of second messengers, in live cells. They are therefore critical tools for uncovering the precise spatial and temporal regulation of signal transduction cascades. These biosensors can be divided into two broad classes: single-color and ratiometric. Single-color sensors, with an intensiometric activity readout, only occupy a single-color channel, allowing for more flexibility in multiplexed imaging experiments. However, they are sensitive to variations in probe concentration caused by changing expression levels or cell shape, as well as differences in imaging conditions, such as illumination intensity and focus. On the other hand, sensors with a ratiometric readout, such as those based on Fö rster Resonance Energy Transfer (FRET), cancel out many of these variations, enabling quantitative measurements of second messenger concentrations and better comparisons between experiments. However, the requirement for two distinct color channels limits their application in multiplexed imaging.
We therefore aimed to develop sensors that only occupy a single channel while still cancelling out the effects of varying imaging conditions and probe concentrations. Hence, rather than using the emission ratio between the FRET and donor channels, we instead used the loss of polarization of emitted light as a readout for FRET between two fluorescent proteins (FPs). Because this approach does not require the donor and acceptor to have distinct emission wavelengths, it can be used for either hetero-FRET (e.g. between different chromophores) or homo-FRET (e.g. between identical chromophores). Homo-FRET measurements have been useful for detecting protein clustering and protein oligomerization in live cells (Bader et al., 2011, Bader et al., 2009Gautier et al., 2001), but only recently has the possibility of using homo-FRET in biosensor designs been explored (Warren et al., 2015;Cameron et al., 2016). Here, we describe the development of a panel of single-color, genetically encodable biosensors based on homo-FRET for detecting kinase activity and second messenger dynamics. We call these sensors FLuorescence Anisotropy REporters, or FLAREs.

Results and discussion
To create our FLARE probes, we adapted existing FRET-based biosensors for homo-FRET measurements by replacing the traditional FRET pair with two FPs of the same color. The resulting biosensors include a molecular switch, which changes conformation in the presence of a particular biochemical activity, flanked by two spectrally similar FPs at the N-and C-termini. Changes in the conformation of the molecular switch, and thus the biochemical activities under study, are then monitored by observing the fluorescence anisotropy of the sensor using fluorescence polarization microscopy, with increased anisotropy corresponding to a lower-FRET state of the sensor, similar to the effect of increasing the intramolecular distance between the FRET pair ( Figure 1-figure supplement 1).
To develop a Protein Kinase A (PKA) activity FLARE, the molecular switch from A Kinase Activity Reporter 4 (AKAR4) (Zhang et al., 2001;Depry et al., 2011), composed of an FHA1 domain and PKA substrate (Figure 1a), was flanked between two FPs of the same color. The FHA1 domain binds to the PKA substrate when the latter is phosphorylated, altering the conformation of the molecular switch and leading to a change in FRET between the flanking homo-FRET pair. To test the effect of FP circular permutation on these biosensors, we developed two FLARE-AKAR variants based on the yellow FP mVenus: one in which the C-terminal FP was circularly permuted at position 172 (cp172Venus), consistent with the hetero-FRET AKAR4 sensor, and one without circular permutation. We expressed mVenus-cp172Venus FLARE-AKAR in HEK293T cells and captured a time-course using fluorescence polarization microscopy. Following PKA activation using a cocktail of 50 mM forskolin (Fsk), an adenylyl cyclase activator, and 100 mM 3-isobutyl-1-methylxanthine (IBMX), a general phosphodiesterase inhibitor, the anisotropy decreased from 0.29 ± 0.003 to 0.26 ± 0.003, a decrease of 0.028 ± 0.001 (N = 44, biological replicates, unpaired, two-tailed t-test, p<0.0001), with the kinetics of the decrease matching those observed with AKAR4 ( Figure 1b (Nagai et al., 2004;DiPilato and Zhang, 2009;Allen and Zhang, 2006). We observed a slight positive correlation between intensity and anisotropy change for Venus-cp172Venus FLARE AKAR; however, the expression level does not significantly impact the reporting ability of these sensors in general ( Figure 1-figure supplement 3a). The signal-to-noise ratio (SNR) of Venus-cp172Venus FLARE-AKAR was calculated to be 32 by dividing the magnitude of the anisotropy change upon maximal PKA stimulation by the standard deviation of the baseline before stimulation.
Subsequent control experiments confirmed that this change in anisotropy is caused by a change in the FRET state due to the conformational change of the sensor upon stimulation of PKA activity. PKA inhibition using 20 mM H-89 led to an immediate slope change and increase in anisotropy ( Figure 1b). On the other hand, a mutant version of the biosensor with a threonine-to-alanine (T-to-A) mutation at the phosphorylation site showed no change in anisotropy upon PKA stimulation with Fsk/IBMX or inhibition with H-89 ( Figure 1b, Figure 1-figure supplement 2b), suggesting that the observed changes in anisotropy were due to phosphorylation of the PKA substrate. Furthermore, isoproterenol, a b-adrenergic agonist, induced FLARE-AKAR responses in a dose-dependent manner (Figure 1-figure supplement 4). To further demonstrate that the change in anisotropy upon PKA stimulation was due to a change in FRET, we mutated the chromophore of the C-terminal cp172Venus in Venus-cp172Venus FLARE AKAR from GYG to GGG. We observed that the magnitude of the response to Fsk/IBMX decreased to approximately one-third of that of the wild-type sensor (Figure 1-figure supplement 5        In addition to the yellow sensors, we developed a panel of color variants of FLARE-AKAR, including EGFP-EGFP, mCherry-mCherry, mCerulean3-mCerulean3 (Markwardt et al., 2011), and mCeru-lean3-cp173Cerulean3 versions. All these variants exhibited a decrease in anisotropy in cells treated with Fsk/IBMX; however, the magnitude of the anisotropy decrease depended on the choice of FP, with the mVenus-cp172Venus variant having the largest response ( Figure 1c, Figure 1-figure supplement 2). As with the Venus variants, the Cerulean3-based FLARE-AKAR showed an increased dynamic range with a circularly permutated fluorescent protein at the C-terminal position. mCherry-mCherry FLARE-AKAR, being spectrally shifted from the AKAR4 heteroFRET sensor allowed for direct comparison of FLARE and heteroFRET sensors within the same cell. As shown in Figure 1figure supplement 6, changes in anisotropy in mCherry-mCherry FLARE-AKAR corresponded with the changes in normalized emission ratio in AKAR4, with similar kinetics.
We furthermore demonstrated the ability of FLARE-AKAR sensors to monitor kinase activity at particular subcellular compartments. By fusing Venus-cp172Venus FLARE-AKAR to targeting motifs from Lyn kinase and DAKAP1, we were able to detect PKA activity at the plasma membrane and outer mitochondrial membrane, respectively ( Figure 1-figure supplement 7). Moreover, we used untargeted FLARE sensors to detect differential PKA activity dynamics in different compartments; diffusable Venus-cp172Venus FLARE AKAR in HeLa showed that PKA activity has slower kinetics and a lower magnitude in the nucleus than the cytosol (Figure 1-figure supplement 8).
To demonstrate the generalizability of FLAREs, we developed a family of single-color kinase activity or activation reporters in various colors ( Figure 2). To construct a single-color Erk activity biosensor, we replaced the PKA sensor domain from Venus-cp172Venus FLARE-AKAR with the sensor domain from EKAR-EV, composed of a WW domain (PAABD), a flexible EV linker, and an Erk substrate peptide ( Figure 2a) (Harvey et al., 2008;Vandame et al., 2014). When expressed in HEK293T cells, Venus-cp172Venus FLARE-EKAR-EV exhibited a decrease in anisotropy of 0.02 ± 0.001 (N = 13) after treatment with 100 ng/mL epidermal growth factor (EGF) to activate the MAPK pathway ( , we likewise repeated this process for various color variants, and like the FLARE-AKARs, the sensors based on mVenus exhibited the largest responses. Furthermore, a myosin light chain kinase (MLCK) sensor was converted to a FLARE by exchanging the cyan FP for mVenus . Calmodulin In addition to biosensors for monitoring enzymes, we also developed FLAREs for monitoring second messenger dynamics ( Figure 3a). We developed a Ca 2+ FLARE by utilizing the sensor domain from the Cameleon family of biosensors, composed of CaM and the Ca 2+ /CaM-binding peptide M13 (Nagai et al., 2004;Miyawaki et al., 1999). When expressed in HEK293T cells, Venus-cp172Venus FLARE-Cameleon exhibited a decrease in anisotropy of 0.03 ± 0.002 (N = 10) upon addition of 1 mM ionomycin and 5 mM CaCl 2 , with the Venus-Venus, mCerulean3-mCerulean3, and   . The mVenus-based FLARE sensors tend to show larger dynamic ranges for a variety of FLARE sensors than mCerulean3 or mCherry variants, likely due to the superior extinction coefficient and quantum yield of mVenus, which make it a good FRET donor and acceptor. We further demonstrated the ability of Venus-cp172Venus FLARE Cameleon to detect submaximal responses by monitoring calcium transients in histamine-stimulated HeLa cells (Figure 3-figure supplement 1c). In order to determine the dissociation constant and Hill coefficients, we purified Venus-cp172Venus Cameleon and measured the fluorescence anisotropy in solutions of known free Ca 2+ concentration at different temperatures ( Figure 3-figure supplement 2). The resulting parameters are in good agreement with other Cameleon sensors (Nagai et al., 2004). Furthermore, we developed another calcium FLARE sensor, based on D1-ER (Palmer et al., 2004), with a sensitivity appropriate for calcium monitoring in the ER. The anisotropy decreases upon increasing calcium concentration and increases when calcium is depleted from the ER upon treatment with thapsigargin ( Figure 3figure supplement 3). In addition to Ca 2+ , we developed a FLARE to detect intracellular cAMP based on the ICUE family of sensors, in which a conformational change in a truncated form of the cAMP effector Epac leads to a decrease in FRET efficiency in the presence of cAMP (DiPilato and Zhang, 2009) (Figure 3a). When expressed in HEK293T cells, the fluorescence anisotropy of Venus-cp172Venus FLARE-ICUE increased by 0.02 ± 0.001 (N = 40) upon stimulation with Fsk/IBMX (Figure 3c, Figure 3-figure supplement 4).
The fact that FLAREs only occupy a single-color channel and are highly generalizable for different biosensors, as well as color variants, highlights their utility for multiplexed imaging applications. While multiplexing of hetero-FRET sensors is generally limited to two probes (Depry et al., 2013;Shcherbakova et al., 2012), we demonstrate that FLAREs can facilitate co-imaging of three biosensors simultaneously. We co-expressed mCherry-mCherry FLARE-AKAR, Venus-cpVenus FLARE-EKAR-EV, and mCerulean3-mCerulean3 FLARE-Cameleon in HEK293T cells and acquired a timecourse with sequential treatment using Fsk/IBMX, EGF, and thapsigargin. Clear and distinct decreases in anisotropy were observed in the red channel after Fsk/IBMX treatment, in the yellow channel after EGF stimulation, and in the cyan channel after thapsigargin treatment, corresponding to an increase in PKA activity, Erk activity, and intracellular Ca 2+ , respectively (N = 17) (Figure 4a, Figure 4-figure supplement 1).
We further aimed to show that FLARE sensors can be used to monitor multiple signaling activities simultaneously in different cellular contexts. For example, mCer3-mCer3 FLARE Cameleon, coimaged with other FLARE sensors in HeLa cells, can detect calcium responses to physiologically relevant stimulation conditions, such as histamine (Figure 4-figure supplement 2). Additionally, when mCer3-mCer3 FLARE AKAR was co-imaged with Venus-cp172Venus cameleon in HEK293T cells, activation of the b-adrenergic with isoproterenol led to a transient decrease in anisotropy in the cyan-channel (Figure 4-figure supplement 3). Furthermore, we used FLAREs to study the cAMP-Ca 2+ oscillatory circuit in pancreatic b-cells. MIN6 b-cells were transiently transfected with Venus-cp172Venus FLARE-ICUE and mCerulean3-mCerulean3 FLARE-Cameleon to simultaneously monitor cAMP and Ca 2+ dynamics, respectively (N = 19). We observed clear fluorescence anisotropy oscillations in both channels following stimulation with 20 mM tetraethylammonium chloride (TEA)       Summary of anisotropy changes after 50 mM forskolin and 100 mM IBMX with respect to baseline (upper right), calculated as the difference between the mean anisotropy from t = 5 min to t = 7.5 min and the anisotropy of the baseline before drug addition. Representative anisotropy pseudocolor image before (t = 0 min) and after (t = 7.5) stimulation with Fsk/IBMX (lower right). Dashed lines above and below time course reflect standard error of the mean. The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. FLARE second messenger biosensor panel.      Figure 4-figure supplement 4). Ca 2+ and cAMP exhibit synchronized oscillations, with Ca 2+ increases corresponding to cAMP decreases, consistent with previous findings (Landa et al., 2005;Ni et al., 2011). These data demonstrate that even the lower signal mCerulean3-based FLARE sensors can be used under sub-maximal and physiologically relevant stimulation conditions in multiplexed imaging experiments. In addition to multiplexed biosensor imaging, the fact that FLARE sensors occupy only a single color channel also permits simultaneously perturbing and monitoring biochemical activities using optogenetics and FLAREs, respectively. We coexpressed an mCherry-tagged, ER-targeted channelrhodopsin2 (hChR2) (Nagel et al., 2003;Lin et al., 2009;Markwardt et al., 2016), a light-gated calcium ion channel, with Venus-cp172 FLARE-Cameleon in REF52 cells, a rat embryonic fibroblast cell line. Illumination with blue light produced an immediate decrease in anisotropy, corresponding to an increase in intracellular Ca 2+ (Figure 4c, Figure 4-figure supplement 5). Control cells lacking hChR2 expression showed no change in anisotropy in the yellow channel.
In vivo, two-photon imaging of FLAREs was tested using a skeletal muscle preparation (Figure 4d,e). Exclusive excitation of mVenus (855 nm) or mCherry (1200 nm) was verified by imaging capillary tubes filled with recombinant proteins (Figure 4d). Plasmids encoding Venus-cp172Venus FLARE-Cameleon and mCherry FLARE-AKAR sensors were electroporated into the flexor digitorum brevis muscle of a live mouse (DiFranco et al., 2009;Kerr et al., 2015). Administration of an electrical current stimulated a rise in intracellular Ca 2+ concentration, as indicated by a decrease in FLARE-Cameleon anisotropy, independent of changes in FLARE-AKAR anisotropy (Figure 4e). Activation of AKAR was then induced by intraperitoneal injection of isoproterenol (0.5 mg/kg). Thus, FLAREs enable in vivo multiparametric biosensor measurements.
We have demonstrated that FLAREs are a highly generalizable, accessible platform for creating single-color sensors to detect biochemical activities in individual cells in real time. Their ratiometric readout allows for fluctuations in light intensity and probe concentration to be cancelled out, permitting quantitative measurements of intracellular concentrations. We showed that these sensors have an SNR of 3-32 (Table 1) and a dynamic range comparable with first-generation FRET sensors (Newman and Zhang, 2008;Zhou et al., 2015). Future development and optimization should further enhance their performance. We demonstrated that current FLAREs are already useful for multiplexed imaging applications. They can also be used in conjunction with optogenetic tools to enable all-optical interrogation of cellular regulation, and for intravital two-photon imaging to facilitate studies in tissues and living animals. FLAREs, by allowing researchers to monitor multiple activities within the same cell, as well as to both monitor and optogenetically perturb activities in the same cell, could be used to study the how the spatiotemporal regulation of biochemical activities in highly integrated pathways allows a small number of signals to produce diverse cellular behaviors. Continued on next page

Plasmid and construct construction
Cloning was performed using the pRSET-B vector using typical molecular cloning methods using polymerase chain reaction (PCR) with Phusion polymerase (New England Biolabs), restriction enzyme digestion, and ligation with T4 DNA ligase. To clone Venus-cp172Venus FLARE, AKAR4 was subcloned from a modified pCDNA3 to pRSET-B between the BamHI and EcoRI restriction enzyme sites. mVenus was then PCR amplified with primers encoding a BamHI site at the 5' end and an SphI site at the 3' end, and the resulting PCR product was digested with BamHI and SphI and ligated to pRSET-B AKAR4 digested with BamHI and SphI, with the mCerulean gene removed. Venus-cp172Venus FLARE AKAR was then sub-cloned back into a modified pCDNA3 vector using the BamHI and EcoRI sites. Other color variants were created by replacing the genes for the fluorescent proteins in other FLARE AKAR variants in pRSET-B, either between the BamHI and SphI sites for the N-terminal fluorescent protein, or SacI and EcoRI for C-terminal fluorescent proteins. Finalized constructs intended for mammalian expression were then sub-cloned into a modified pcDNA3 expression vector between the BamHI and EcoRI sites. FLARE variants of other sensors were created by amplifying the molecular switch from EKAR-EV, CKAR2, Cameleon and ICUE3 with primers encoding the SphI and SacI sites, digesting the PCR product with SphI and SacI enzymes, and ligating them to the relevant pRSET-B FLARE AKAR plasmid digested with SphI and SacI to remove the domains involved in the molecular switch for FLARE AKAR. The final constructs were then subcloned into a modified pCDNA3 expression vector between the BamHI and EcoRI sites. Targeted versions of the sensors were created either by PCR amplifying the sensor with primers containing the targeting sequence and ligating it to the pCDNA3 expression vector between BamHI and EcoRI, or by subcloning the construct into a plasmid already containing the targeting sequence. N-terminal targeting sequences were placed between HindIII and EcoRI, and C-terminal targeting sequences between EcoRI and XbaI. All cloning steps were performed using DH5a strain of E. coli. The threonine to alanine mutants for Venus-cp172Venus FLARE AKAR and FLARE EKAR were created by performing site-directed mutagenesis using a standard single-primer PCR-based protocol. The threonine to alanine mutant for Venus-cp172Venus FLARE CKAR, as well as the chromophoredead variant of Venus-cp172Venus FLARE AKAR, was created using Gibson assembly, amplifying the appropriate fragment with a primer containing the desired mutation.
The YFP MLCK FLARE sensor was created by replacing the CFP portion of an existing two color sensor Geguchadze et al., 2004) with a PCR amplified mVenus fragment flanked by XhoI and AgeI restriction sites. Moreover, the coding sequence for the CFP D1 ER sequence was manufactured by Genewiz. The sensor consists of two oxmCer3 proteins (https:// www.ncbi.nlm.nih.gov/pubmed/26158227) flanking the D1 ER calcium sensing domains (Palmer et al., 2004) and a C-terminal KDEL ER retention sequence.

Fluorescence polarization microscopy
Fluorescence anisotropy reporter co-imaging is described in more detail in Bio-protocol (Ross et al., 2019). Widefield images were collected using an Zeiss AxioObserver equipped for fluorescence polarization microscopy, using one of two setups. In the first setup, a wire grid polarizer (Meadowlark Optics) was placed in the excitation pathway between the LED illuminators and reflector turret containing filter cubes specific for CFP (Zeiss), YFP (Zeiss), and mCherry (Semrock). Images were generally collected using a 20 Â 0.75 NA objective lens. Polarizations parallel and perpendicular to the excitation polarizations were separated using Optical Insights Dual-View using their polarization splitting module. Both images were simultaneously collected in a single image collected by a watercooled Orca-R2 (Hamamatsu). In the second setup, a polarizer (Chroma) was placed in the excitation pathway between the xenon arclamp and the excitation filters. Images were collected using a 20 Â 0.45 NA objective lens. Polarizations were separated using an Opto-Split II LS image splitter, with two wire grid polarizers (Meadowlark) oriented parallel and perpendicular to the excitation polarizer. Images of both polarizations were collected using a Hamamatsu Flash 4.0 sCMOS camera. Two-photon imaging was performed using a Zeiss 7 MP with GaAsP non-descanned detectors housed at the University of Maryland School of Medicine confocal facility. Coherent Chameleon and OPO lasers were used for excitation. Fluorescence was filtered using an ET680 short pass filter for two-photon microscopy (Chroma) prior to separating polarizations with a one inch broadband polarizing beamsplitter cube (Thorlabs) mounted using a custom 3D printed cube. Images were collected using a 10Â, 0.3 NA Plan-apochromat objective lens. In vivo imaging was performed on C57Bl/6 mice under isoflurane anesthesia.

Image analysis
Image analysis was performed using Fiji (ImageJ) open-source image processing software. Polarization images were cropped and aligned using either the Zeiss Axiovision software or Fiji's built-in StackReg registration plugin. In Fiji, regions of interest (ROIs) were drawn around each cell, as well as one in the background. ROI intensities were background subtracted in each channel to estimate fluorescence emission intensity, and anisotropy was calculated as described by Lakowicz (2010). Anisotropies were calculated using the conventional equation : where g is the correction factor that accounts for differences in polarization transmission efficiencies within the instrument. The g-factor was calculated using an isotropic fluorescein solution as described by Piston and Rizzo (2008). Anisotropy was calculated by subtracting the anisotropy at each time point by the anisotropy at the time point right before drug addition. The magnitude of the anisotropy changes were calculated by taking the difference between the average anisotropy when the signal peaked or plateaued and the average anisotropy of the baseline time points before drug was added.