ShadowY: a dark yellow fluorescent protein for FLIM-based FRET measurement

Fluorescence lifetime imaging microscopy (FLIM)-based Förster resonance energy transfer (FRET) measurement (FLIM-FRET) is one of the powerful methods for imaging of intracellular protein activities such as protein–protein interactions and conformational changes. Here, using saturation mutagenesis, we developed a dark yellow fluorescent protein named ShadowY that can serve as an acceptor for FLIM-FRET. ShadowY is spectrally similar to the previously reported dark YFP but has a much smaller quantum yield, greater extinction coefficient, and superior folding property. When ShadowY was paired with mEGFP or a Clover mutant (CloverT153M/F223R) and applied to a single-molecule FRET sensor to monitor a light-dependent conformational change of the light-oxygen-voltage domain 2 (LOV2) in HeLa cells, we observed a large FRET signal change with low cell-to-cell variability, allowing for precise measurement of individual cell responses. In addition, an application of ShadowY to a separate-type Ras FRET sensor revealed an EGF-dependent large FRET signal increase. Thus, ShadowY in combination with mEGFP or CloverT153M/F223R is a promising FLIM-FRET acceptor.

SciEnTific RepoRts | 7: 6791 | DOI: 10.1038/s41598-017-07002 -4 To overcome this limitation, a very dark fluorescent protein named ShadowG was developed by directed evolution using sREACh 8 . Although this protein has superior darkness and minimized response variability, the FRET signal is relatively small, compared with that of sREACh 8 .
Here, we aimed to develop a dark yellow fluorescent protein that has high absorption and a low quantum yield as compared with sREACh to increase the FRET signal and prevent an artifact due to residual fluorescence. Saturation mutagenesis at the positions surrounding the chromophore of sREACh led to a new fluorescent protein named ShadowY that has a 7-fold lower quantum yield and a 1.2-fold greater extinction coefficient than those of sREACh. Furthermore, we confirmed that the pairing of ShadowY with mEGFP, Clover 13 , or its mutant (Clover T153M/F223R ) shows improved FRET signals with reduced cell-to-cell variability. Thus, mEGFP-ShadowY and Clover mutant-ShadowY are good FLIM-FRET pairs.

Results
To create a dark yellow fluorescent protein, we applied saturation mutagenesis to amino acid positions N144, N146, S147, and V148 surrounding the chromophore in the previously reported dark yellow fluorescent protein, sREACh ( Fig. 1) 12 . Because position W145 is crucial for reduction of quantum efficiency 7 , we avoided introducing a mutation at this position. We introduced the A206K monomeric mutation, while F223R was reversed (Fig. 1a); A206K increases the dissociation constant more than F223R does 14 . The PCR products with saturated mutations were ligated into a bacterial expression vector, and we thus constructed a genetic library. To screen the library for dark yellow fluorescent proteins that have a high extinction coefficient but a low quantum yield, we first identified vividly colored colonies under day light, confirming that the mutants have high absorption. Subsequently, the colonies were confirmed to have no fluorescence under blue light by a method similar to the one described elsewhere 8,15 . This screening identified a burnt orange colony under day light but no visible fluorescence under blue light. As a result, sequence analysis identified the following mutations: N144A, N146P, S147V, H148V, A206K, and R223F in sREACh (Fig. 1, Table 1). This mutant named sREACh#1 has a ~2-fold smaller quantum yield, while its extinction coefficient is comparable to that of sREACh (Table 1). Subsequently, with sREACh#1 as a template, saturation mutagenesis at positions Q204/S205/L207 located near the chromophore was carried out. After the screening, spectroscopic and sequence analysis revealed that one of the mutants named sREACh#2 has ~1.14-fold increased extinction coefficient and ~1.5-fold decreased quantum efficiency (Table 1). A similar procedure was also carried out at position K166/R168/H169, and we identified a mutant that shows improved darkness   1  11  21  31  41  51  VSKGEELFTG VVPILVELDG DVNGHKFSVS GEGEGDATYG KLTLKFICTT GKLPVPWPTL  VSKGEELFTG VVPILVELDG DVNGHKFSVS GEGEGDATYG KLTLKLICTT GKLPVPWPTL  VSKGEELFTG VVPILVELDG DVNGHKFSVS GEGEGDATYG KLTLKLICTT GKLPVPWPTL   61  71  81  91  101  111  VTTFGYGLQC FARYPDHMKQ HDFFKSAMPE GYVQERTIFF KDDGNYKTRA EVKFEGDTLV  VTTFGYGLMC FARYPDHMKQ HDFFKSAMPE GYVQERTIFF KDDGNYKTRA EVKFEGDTLV  VTTFGYGLMC FARYPDHMKQ HDFFKSAMPE GYVQERTIFF KDDGNYKTRA EVKFEGDTLV   121  131  141  151  161  171  NRIELKGIDF KEDGNILGHK LEYNYNSHNV YIMADKQKNG IKVNFKIRHN IEDGSVQLAD  NRIELKGIDF KEDGNILGHK LEYAWPVVNV YIMADKQKNG IKVNFSIYHN IEDGSVQLAD  NRIELKGIDF KEDGNILGHK LEYNWNSVNV YIMADKQKNG IKVNFKIRHN IEDGSVQLAD   181  191  201  211  221   and absorption relative to sREACh#2 (Table 1). We named this mutant ShadowY where Y stands for "yellow", and decided to pursue further analyses of this protein.
Spectral analysis of purified ShadowY confirmed that it has an excitation peak at 519 nm and an emission peak at 531 nm (Fig. 2a,b and Table 1), similar to those of sREACh (Table 1). Further analysis revealed that the molar extinction coefficient of ShadowY is 136,000 M −1 cm −1 : a 1.2-fold greater extinction coefficient than that of sREACh (Table 1). Quantum efficiency of ShadowY is 0.01, which is 7-fold smaller than that of sREACh (QE, 0.07; Table 1). Consistent with these results, two-photon excitation spectrum of ShadowY exhibited the low fluorescence compared with those of mEGFP and Clover (Fig. 2c), and the fluorescence lifetime of ShadowY (0.19 ns) is much shorter than that of sREACh (0.67 ns; Table 1, Fig. 2d).
We next characterized the folding and maturation kinetics of ShadowY by the urea-denaturation method as described previously 16 . The fluorescence of denatured ShadowY recovered in 73 sec: faster than recovery of sREACh (267 sec; Table 1, Fig. 2e), suggesting that ShadowY has superior folding properties. Next, chromophores (e,f) Fluorescence recovery of sREACh and ShadowY from a denatured (e) or reduced state (f). The respective fluorescent protein was excited at 517 nm with 5 nm bandwidth, and its fluorescence recovery was monitored at 531 nm with 5 nm bandwidth. Three independent experiments were averaged (the data are shown as mean ± SEM).
of the urea-denatured ShadowY were reduced with dithionite, and reoxidation time and recovery were monitored after dilution in urea-free buffer. Reoxidation time of ShadowY (140 min) is comparable to that of sREACh (130 min; Fig. 2f, Table 1).
Next, we tested the performance of ShadowY as an energy acceptor for 2-photon FLIM-FRET via comparison with sREACh in HeLa cells. We used 2-photon excitation for imaging because of the reduced phototoxicity compared with 1-photon excitation 17 . We chose mEGFP or Clover as an energy donor, because the emission spectra of these proteins significantly overlap with the excitation spectrum of ShadowY ( Fig. 3a,b). To quantify the performance of mEGFP-ShadowY and Clover-ShadowY pairs in comparison with mEGFP-sREACh and Clover-sREACh pairs, we fused these fluorescent proteins to the N and C termini of a light-sensitive LOV2-Jα helix domain from Phototropin 1 18,19 , respectively, creating mEGFP-LOV2-ShadowY, mEGFP-LOV2-sREACh, Clover-LOV2-ShadowY, and Clover-LOV2-sREACh as LOV2 FRET sensors (Fig. 4a), and monitored the blue-light-dependent structural change in HeLa cells by means of 2-photon FLIM-FRET (Fig. 4a,b). HeLa cells expressing the LOV2 FRET sensor were illuminated with blue light at 35 mW/cm 2 for 2 sec (Fig. 4b-d). Right after illumination, the fluorescence lifetime of mEGFP in LOV2 FRET sensors increased, i.e., FRET decreased, and returned in ~60 sec, consistent with another study 19 . The quantitative analysis indicated a significant increase in the fluorescence lifetime change of mEGFP-LOV2-ShadowY relative to mEGFP-LOV2-sREACh (Fig. 4e). Furthermore, we compared the cell-to-cell variability of FRET signals (Fig. 4g,h), and found that the variability of mEGFP-LOV2-ShadowY in both the basal state and after light illumination (before, 1.94 ± 0.05 ns; after, 2.13 ± 0.04 ns) is smaller than that of mEGFP-LOV2-sREACh (before, 1.96 ± 0.06 ns; after, 2.12 ± 0.05 ns). Taken together, these results suggest that ShadowY is superior FLIM-FRET acceptor.
To further characterize ShadowY, we measured FRET efficiency and maturation efficiency using tandem fluorescent proteins in HeLa cells as described previously 8 . We expressed tandem constructs (Fig. 5a), and the fluorescence lifetime of mEGFP, Clover, or Clover T153M/F223R was measured by 2-photon FLIM-FRET ( Fig. 5b-g). Because the fluorescence lifetime decay curves are convolution of both the FRET efficiency and maturity of an acceptor 5 , we measured these parameters separately, as described earlier 2,12 . Although FRET efficiencies of all the compared pairs showed comparable values (Fig. 5b,d,f), the maturity of ShadowY was found to be slightly better than that of sREACh in all pairs (Fig. 5c,e,g).
Next, mEGFP-ShadowY, Clover-ShadowY, and Clover T153M/F223R -ShadowY pairs were applied to a separate-type H-Ras FRET sensor 8,20 , and their FRET signals were compared (Fig. 6). We did not compare with sREACh because it has a bleed-thorough effect (Fig. S1). As a FRET donor, H-Ras was fused to mEGFP, Clover, or Clover T153M/F223R , and as an acceptor, the Ras-binding domain of Raf1 was fused to ShadowY (Fig. 6a). The donor  and acceptor were fused via the P2A sequence to ensure equal expression of these molecules 21 and to minimize the response variability due to the imbalanced expression of the donor and acceptor. We transfected HeLa cells with these FRET sensors and compared their response signals as a binding fraction change (Fig. 6b-g). After stimulation with epidermal growth factor (EGF), H-Ras was rapidly activated (within a few minutes; Fig. 6b,c). When the FRET response signals of Ras sensors were compared, all the three FRET sensors showed a similar signal change (Fig. 6d-g), with the values of 21.34 ± 1.05 (mEGFP-ShadowY), 18.39 ± 0.77 (Clover-ShadowY), 20.62 ± 0.86 (Clover T153M/F223R -ShadowY), respectively (Fig. 6d).

Discussion
Here, we successfully developed a new dark yellow fluorescent protein, ShadowY, as a FLIM-FRET acceptor for pairing with mEGFP or the Clover mutant. ShadowY has superior properties in terms of absorption and folding kinetics relative to sREACh (Fig. 2, Table 1). These factors most likely contribute to the increased FRET signals and the reduced cell-to-cell variability, compared with those of sREACh (Figs 4 and 5). Furthermore, although sREACh is difficult to apply to a separate-type FRET sensor because of bleed-through fluorescence contamination 8 , ShadowY does not have this problem because of the superior darkness relative to sREACh (Fig. S1). An application of ShadowY to an LOV2 and H-Ras FRET sensors yielded a large FRET change (Figs 4c,d and 6c), which is larger than that of the previously reported mCherry or ShadowG version of sensors 8 .
In the past decade, several types of dark fluorescent proteins have been identified and applied to FRET imaging, photoacoustic imaging, and structural analysis 7,8,12,[22][23][24][25] . We believe that ShadowY will be an additional useful tool for these studies, especially for FLIM-FRET.

Materials and Methods
Saturation mutagenesis. The sREACh gene in a customized pRSET vector (Invitrogen) served as an initial template for construction of genetic libraries. First, a XhoI restriction site was silently introduced at the positions corresponding to amino acid residues L141 and E142 in sREACh (Fig. 1a). Saturated mutagenesis was performed by PCR amplification of sREACh (the fragment corresponding to amino acid positions 141-238) with degenerate primers. These primers are as follows: For sREACh#1, FW 5′-gagactcgagtacNNBtggNNBNNBNNBNNBgtctatatcatggccga-3′, RV 5′-gagaggatcccttgtacagctcgtccat-3′, and XhoI and BsrGI are used for subcloning into the custom pRSET vector; for sREACh#2, FW 5′-gagagggcccgtgctgctgcccgacaaccactacctgagctacNNBNNBaagNNBagcaaagaccccaacg-3′, RV 5′-gagaggatcccttgtacagctcgtccat-3′, and ApaI and BsrGI sites were used for subcloning; for ShadowY, FW 5′-gagactcgagtacgcttggcccgtggtgaatgtctatatcatggccgacaagcagaagaacggcatcaaggtgaacttcNNKatcNN-KNNKaacatcgaggacggca-3′, RV 5′-gagaggatcccttgtacagctcgtccat-3′, and XhoI and BsrGI sites were used for subcloning. The plasmid was then introduced into electrocompetent cells, and the cells were grown for 18-20 h at 34 °C on LB agar plates supplemented with antibiotics.
Plasmid construction. In all DNA construction procedures described below, a modified pEGFP-C1 plasmid (Clontech) served as a backbone vector. For construction of the EGFP or Clover with the CAAX motif of K-Ras (corresponding to amino acid residues 173-188), EGFP or Clover fused to CAAX via a linker encoding the peptide SGLRSRAQASNSAV was inserted into the vector by replacing EGFP. To create cytosolic mCherry, sREACh, or ShadowY as shown in Fig. S1, the EGFP in the vector was replaced by the respective genes. To create the tandem fluorescent protein constructs (shown in Fig. 5), the respective combination of fluorescent proteins was fused with a linker encoding the peptide SGLRSG in the vector.
For construction of the LOV2 FRET sensor, a donor fluorescent protein was fused to the N terminus of the LOV2 domain (DNA sequence corresponding to amino acid residues 404-546 in Phototropin 1) via a linker encoding the peptide ASM. The acceptor fluorescent protein was fused to the C terminus of the LOV2 domain via the linker peptide KLGNS.
For construction of the H-Ras FRET sensors, we fused an acceptor fluorescent protein to the C terminus of the Ras-binding domain of Raf1 (amino acid residues 50-131 with two mutations: K65E and K108A) 20 via the linker peptide GSG. Subsequently, H-Ras fused to a donor fluorescent protein via the linker peptide SGLRSRG was fused to the C terminus of the acceptor protein via the P2A sequence 21 so that the Ras-binding domain and H-Ras parts were translated into different polypeptides within the cell.

Fluorescent properties of the fluorescent proteins.
His-tagged fluorescent proteins were overexpressed in Escherichia coli DH5α cells using a modified pRSET vector (Invitrogen) and purified on a Ni + -nitrilotriacetate column (HiTrap, GE Healthcare). Excitation and emission spectra of the fluorescent proteins diluted in PBS were recorded on a spectrofluorometer (RF-6000; Shimadzu). Matured-protein concentrations were calculated from the extinction coefficient of the chromophore after denaturation in 0.1 N NaOH (40,000 M −1 cm −1 at 446 nm) 26 . The extinction coefficients of fluorescent proteins were determined by dividing the peak optical density by the molar concentration of matured proteins. Quantum efficiency of the proteins was determined by a comparison with that of Clover (0.76) as described elsewhere 13 .
Two-photon excitation spectra were measured under the two-photon fluorescence microscope (FVMPE-RS; Olympus). An Insight Ti:Sapphire laser (Spectra-Physics) with the power of 3.4-4.5 mW at the respective wavelength under the objective lens was used to excite the purified fluorescent proteins. Raw fluorescence intensity values were corrected by dividing them by squared laser power used for each wavelength.

Refolding and reoxidation.
To measure the refolding time of ShadowY after denaturation, the proteins were dissolved in denaturation buffer (8 M urea, 1 mM dithiothreitol) and heated at 95 °C for 5 min as described previously 16 . The refolding was initiated by diluting the denatured protein with a 100-fold volume of renaturation buffer (5 mM KCl, 2 mM MgCl 2 , 50 mM Tris-HCl pH 7.5, 1 mM dithiothreitol) at room temperature. For the reoxidation experiment, 5 mM dithionite was added into the denaturation buffer to reduce the chromophore. The respective fluorescent protein was excited at 517 nm with 5 nm bandwidth, and its fluorescence recovery was monitored at 531 nm with 5 nm bandwidth in a spectrofluorometer (RF-6000; Shimadzu). construction of a fluorescence lifetime image, the mean fluorescence lifetime in each pixel was translated into a color-coded image 2, 28 . Analysis of the lifetime change and binding-fraction change was conducted as described elsewhere 2,29,30 . In Fig. 4, blue LED (244-87-470-50E-40; CoolLED) with a band pass filter (FF01-469/35-25; Chroma) was used for illumination to induce the structural change of LOV2 FRET sensors.

Analysis of the fluorescence lifetime image.
To generate fluorescence lifetime images, we acquired the mean fluorescence lifetime in each pixel by calculating the mean photon arrival time <t>as where t o is obtained by fitting the whole image with single exponential or double exponential functions convolved with an instrument response function as described previously 2,30 . After that, the mean fluorescence lifetime in each pixel was converted to the corresponding color. FRET efficiency and the fraction of the donor fluorescent protein undergoing FRET were calculated as in other studies 2,5,8,30 . along with the corresponding binding fraction (averaged over 8 to 10 min) after EGF stimulation (red). The data are also presented as mean ± SD on the right.