ShadowR: a novel chromoprotein with reduced non-specific binding and improved expression in living cells

Here we developed an orange light-absorbing chromoprotein named ShadowR as a novel acceptor for performing fluorescence lifetime imaging microscopy-based Förster resonance energy transfer (FLIM-FRET) measurement in living cells. ShadowR was generated by replacing hydrophobic amino acids located at the surface of the chromoprotein Ultramarine with hydrophilic amino acids in order to reduce non-specific interactions with cytosolic proteins. Similar to Ultramarine, ShadowR shows high absorption capacity and no fluorescence. However, it exhibits reduced non-specific binding to cytosolic proteins and is highly expressed in HeLa cells. Using tandem constructs and a LOVTRAP system, we showed that ShadowR can be used as a FRET acceptor in combination with donor mRuby2 or mScarlet in HeLa cells. Thus, ShadowR is a useful, novel FLIM-FRET acceptor.


Results
To create a novel chromoprotein, we chose Ultramarine as a template 16 . Ultramarine is a monomeric chromoprotein with a relatively large extinction coefficient (64,000 M −1 cm −1 ) and low quantum yield (~0.001), but has many hydrophobic amino acids on its surface (Fig. 1a). We attempted to make Ultramarine more hydrophilic to reduce its non-specific hydrophobic interaction with cytosolic proteins. To do that, we first created a chimera by replacing the Ultramarine amino acid sequences (1-51, 118-157, and 231-236) with the corresponding regions of mCherry (Fig. S1). We chose this chimera because it is slightly more hydrophilic than Ultramarine (Fig. 1b). The rest of the sequence was not replaced because it surrounds the chromophore and is important for absorption and low quantum yield. Based on the crystal structure data of mCherry (Protein Data Bank ID: 2H5Q), we identified 105 amino acids whose side chains are directed outward from the chimera. Among them, 32 hydrophobic amino acids indicated by green arrowheads in Fig. S1 were selected to be replaced by more hydrophilic ones using single-amino acid saturation mutagenesis, or we replaced to either corresponding mCherry or Ultramarine amino acid (Fig. S1). When the outward directed amino acids are clustered, those are simultaneously subjected to   ShadowR are shown in blue and black letters, respectively. Extra amino acids (gray characters) were added to the N and C termini of Ultramarine to match the molecular size of ShadowR. Squares (gray, blue, red) indicate the amino acids (10-228 in Ultramarine/ShadowR) whose side chains are directed outward from the protein.
To identify amino acids, we utilized the crystal structure (Protein Data Bank ID: 2H5Q) of mCherry which shares 64% homology to Ultramarine. Blue and Red squares indicate that the amino acids in Ultramarine were replaced to more hydrophilic and hydrophobic amino acids, respectively. Gray squares indicate that the hydrophobicity of the amino acids is identical between Ultramarine and ShadowR. The chromophore tripeptide is highlighted with a magenta box. (b) Hydrophobicity index of respective proteins as calculated simply by summing the hydropathy index of amino acids 31 is indicated by the squares. (c) A homology model of ShadowR created by SWISS-MODEL 42 . The X-ray crystal structure of mCherry mutant (PDB ID 3NED) 43 was used as a template to represent ShadowR. Relatively hydrophobic acids (I, V saturation mutagenesis. The PCR products with saturated mutations were ligated into a bacterial expression vector and a genetic library was constructed. To screen the library for hydrophilic chromoproteins, we first identified vivid purple colonies under ambient light, confirming that the mutants have high absorption. We also confirmed that the colonies identified are not fluorescent under blue light illumination. Subsequently, we sequenced the identified colonies and picked mutants with more hydrophilic amino acids than the original sequence. When single-amino acid mutagenesis failed to produce purple colonies, the surrounding amino acids were simultaneously subjected to saturation mutagenesis or replaced to the corresponding Ultramarine or mCherry amino acids. We sequentially repeated this process for the 32 positions (Fig. S1), and finally identified a mutant that shows high absorption and non-fluorescence comparable to Ultramarine (Table 1), but with greater hydrophilicity calculated by using the reported hydropathy index of amino acids 31 (Fig. 1). As a result of these processes, 19 and 3 amino acids among 32 positions in the Ultramarine/mCherry chimera were replaced to more hydrophilic and hydrophobic ones, respectively (Figs 1c, S1). The three amino acids were replaced to hydrophobic ones, because the replacement of these amino acids to hydrophilic ones resulted in the loss of absorption. Eight amino acids were unchangeable because of the loss of absorption, and 2 amino acids were replaced to the different amino acids, but with the same hydropathy index 31 . The value of surface hydrophobicity of ShadowR is greatly reduced compared to that of Ultramarine or the chimera (Fig. 1b). Furthermore, electrostatic surface potential maps revealed that the electrostatic charge is increased at the surface of ShadowR compared with that Ultramarine (Fig. S2). We named this hydrophilic mutant ShadowR, where R stands for "red", since the absorption peak (585 nm) is similar to that of red fluorescent proteins. Using size-exclusion chromatography, we first confirmed that ShadowR is monomeric similar to Ultramarine (Fig. 2a). To detect the non-specifically bound cytosolic proteins of HEK293 cells to Ultramarine/ShadowR, we developed a new method called the non-specific binding assay (NSB assay). Ni + -nitrilotriacetate beads saturated with His-tagged Ultramarine or ShadowR were incubated with cell lysate (Fig. 2b), and non-specifically bound proteins were pulled down by centrifugation. Subsequent silver staining indicated lower levels of non-specific binding of ShadowR to cytosolic proteins than that of Ultramarine (Fig. 2c,d).
Spectral analysis of purified ShadowR confirmed that, similar to Ultramarine, ShadowR has an excitation peak at 585 nm, and molar extinction coefficient of 97,100 M −1 ·cm −1 ( Fig. 3a and Table 1). Quantum efficiency was not measurable because of the lack of fluorescence. Since the absorption spectrum of ShadowR significantly overlaps with the emission spectrum of mRuby2 (Fig. 3b) and mScarlet (Fig. 3c), mRuby2/ShadowR and mScarlet/ ShadowR may serve as FRET pairs.
To quantify the performance of mRuby2/ShadowR and mScarlet/ShadowR pairs as FRET pairs, we measured FRET efficiency and maturation efficiency using fusion proteins in HeLa cells and compared the results with those of ShadowR and Ultramarine proteins, as described previously (Fig. 4a) 18 . We expressed tandem constructs by lipofection, and measured the fluorescence lifetime of mRuby2 or mScarlet in the living HeLa cells by 2-photon FLIM-FRET ( Fig. 4a-f). We used 2-photon excitation for imaging because of its low phototoxicity compared with single-photon excitation 32 . Because of the low levels of expression of Ultramarine fusions (i.e., mRuby2/Ultramarine and mScarlet/Ultramarine) compared with those of ShadowR, we used a higher laser power for their imaging (See Fig. 4b legend). By analyzing the fluorescence lifetime decay curves, we measured the FRET efficiency and maturity of each acceptor separately, as described earlier 9,11,15 . While the FRET efficiencies of mRuby2 and mScarlet fusions with ShadowR were lower than those with Ultramarine (Fig. 4c,e), the maturity of ShadowR fusion proteins was comparable to that of Ultramarine fusion proteins (Fig. 4d,f). The slight decrease in FRET efficiency with ShadowR could be due to the difference in the relative orientation of the chromophore with regard to the donor and the acceptor, because of the amino acid sequence difference between Ultramarine and ShadowR. Next, we observed the chromophore maturation of ShadowR in E. coli (Fig. 4g) and found that the colonies expressing ShadowR exhibit more vivid purple color than those expressing Ultramarine, suggesting that ShadowR has better maturation in E. coli.
Next, we confirmed the expression of ShadowR in HeLa cells. Since ShadowR has no fluorescence, we fused A206K-mutated monomeric EGFP (mEGFP) 33 , mRuby2, or mScarlet with ShadowR to visualize ShadowR expression as fusion protein. The fluorescence level from individual cells was quantified by epifluorescence microscopy (Fig. 5). The cells expressing mEGFP/ShadowR showed higher fluorescence intensities compared with those expressing mEGFP/Ultramarine (Fig. 5a,b). Similar results were obtained with mRuby2 and mScarlet fusion proteins (Fig. 5a,c,d). Since there is spectral overlap between mScarlet/mRuby2 emission and ShadowR absorption (Fig. 3b,c), the increased brightness of mScarlet/mRuby2 could be due to the lower levels of complete  16,19 , respectively. Extinction coefficients were measured by the alkaline denaturation method (See Materials and Methods). Since EC and QE measurements could be operation sensitive, we carried out the side-by-side measurement of Ultramarine as a control. The differences in EC values of Ultramarine may be due to differences in experimental conditions or operational differences.
www.nature.com/scientificreports www.nature.com/scientificreports/ maturation of ShadowR compared with Ultramarine. However, this possibility is excluded since the maturation and FRET efficiency of ShadowR is comparable to those of Ultramarine ( Fig. 4c-f).
Since the tandem constructs with ShadowR exhibited brighter fluorescence (Fig. 5), we next performed western blotting and real time PCR to determine if the increased fluorescence observed with ShadowR was due to increased protein or mRNA expression (Fig. 6). Western blotting revealed increased expression of mEGFP/ShadowR and mRuby2/ ShadowR fusion proteins (Fig. 6a-c). Non-fused ShadowR also showed increased expression compared to Ultramarine (Fig. 6a,d). Slight band shift compared with Ultramarine fusion was observed for mEGFP/ShadowR and ShadowR. We rigorously checked their DNA sequences and confirmed that there is no unwanted insertion. Most likely, the band shift was due to the replacement of hydrophobic amino acids with charged hydrophilic amino acids or increased molecular weight (See Fig. 2a legend). Next, we quantified the mRNA levels of each construct, and found higher mRNA expression levels of ShadowR, mEGFP/ShadowR, and mRuby2/ShadowR than those of Ultramarine and its fusions (Fig. 6e). These results suggested that ShadowR exhibits enhanced mRNA and protein expression.
We further tested the performance of ShadowR using a genetically encoded optogenetic tool, the LOVTRAP system 34 . The LOVTRAP system consists of Zdk1 and LOV2 domains 35,36 , and their dissociation and association can be controlled by blue light. We fused mScarlet with LOV2 and ShadowR or Ultramarine with Zdk1, creating a LOVTRAP FRET construct (Fig. 7a). These pairs were expressed in HeLa cells and their blue light-dependent association and dissociation were imaged and quantified by 2pFLIM-FRET (Fig. 7b). We only tested mScarlet-LOV2, not mRuby2-LOV2, because mScarlet is much brighter in cells than mRuby2. In the absence of blue light, mScarlet-LOV2 bound to Ultramarine/ShadowR-Zdk1 in cells, but not Ultramarine/ShadowR alone, suggesting that Zdk1 binds to mScarlet-LOV2 (Fig. 7c). Next, HeLa cells expressing the LOVTRAP FRET construct were illuminated with blue light at 35 mW/cm 2 for 2 s. Immediately after illumination, the fluorescence lifetime of mScarlet in LOV2 increased by decreased FRET, and returned to basal levels in approximately 60 s, consistent with results of another study 34 . The binding fraction change (i.e., the fraction of mScarlet bound to Zdk1, see also Materials and Methods) of mScarlet-LOV2/ShadowR-Zdk1 was larger than that of mScarlet-LOV2/Ultramarine-Zdk1, suggesting that ShadowR is a superior chromoprotein in the LOVTRAP system. As a control experiment, we only expressed mScarlet-LOV2 and found that there was no binding fraction change after light illumination, suggesting that the change in binding fraction was due to the dissociation of Zdk1 (Fig. 7d-f). www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Here, we successfully developed a new chromoprotein, ShadowR, as a FLIM-FRET acceptor for pairing with mScarlet or mRuby2. Compared with the previously reported chromoprotein, Ultramarine 16 , ShadowR has superior property in terms of reduced non-specific binding to cellular proteins (Fig. 2b-d). The observed reduced non-specific binding of ShadowR is most likely due to the increased hydrophilic property compared with Ultramarine. Another feature of ShadowR is its increased protein expression in HeLa cells (Figs 5, 6), that facilitates imaging with lower laser power for reduced photodamage. It is not currently known if there is the b a  www.nature.com/scientificreports www.nature.com/scientificreports/ causal association between the increased expression and the surface hydrophilicity (Fig. 6a-d). Since the protein expression tend to correlate with mRNA expression 37,38 , the increased protein may be due to the increased levels of mRNA (Fig. 6e). A few possibilities should be considered regarding the mechanism underlying increased mRNA expression: 1) the expressed Ultramarine proteins inhibit the transcription machinery of mRNA (negative feedback) or 2) the transfected Ultramarine DNA adopts a structure that leads to inefficient transcription. However, the precise mechanism underlying increased mRNA levels is unknown and difficult to deduce from our experiments.
We previously reported the use of dark mCherry, mCherry I202Y , as a FRET acceptor 19 . The advantage of ShadowR over dark mCherry is its darkness. While the quantum efficiency of the dark mCherry is 0.02, the quantum efficiency of ShadowR is undetectable. This superior darkness may prevent artificial FRET signals due to fluorescence contamination 18  www.nature.com/scientificreports www.nature.com/scientificreports/ changes, compared with those of the Ultramarine version of constructs (Fig. 7). The reason of the enhanced FRET change could be due to reduced non-specific interactions of ShadowR with mScarlet or LOV2 compared with Ultramarine. In the dark state, mScarlet-LOV2 binds to ShadowR-Zdk1 via LOV2 and Zdk1. However, if additional non-specific interactions such as the binding of ShadowR to mScarlet or LOV2 exists and are weaker than those of Ultramarine due to the surface amino acid difference, the light-dependent separation of ShadowR-Zdk1 from mScarlet-LOV2 compared with that of Ultramarine fusions is facilitated. While ShadowR constructs exhibit the significant FRET signals, they show quite large cell-to-cell variability (Figs 4d, f, 7e, f). One of the future directions for improving ShadowR is to minimize this variability for more accurate measurement.
Taken together, we believe that ShadowR will be an additional useful tool for these studies, especially for FLIM-FRET.

Materials and Methods
Saturation mutagenesis. The synthesized gene encoding Ultramarine was purchased from FASMAC (Kanagawa, Japan). This gene in a customized pRSET vector (Invitrogen) was used as a template for constructing genetic libraries for ShadowR development. Sequential saturation mutagenesis to the targeted positions was performed by PCR amplification with degenerate primers in combination with overlapping PCR. Subsequently, the amplicons were subcloned into the customized pRSET vector. For making a library, the plasmid library was introduced into electro-or chemically competent cells, and the cells were grown for 18-20 h at 34 °C on LB agar plates supplemented with antibiotics.
Plasmid construction for mammalian expression. For all DNA construction described below, a modified pEGFP-C1 plasmid (Clontech), where a kanamycin resistance gene was replaced with an ampicillin resistance gene, was used as a backbone vector. The synthesized LOV2, Zdk1, and mScarlet genes were purchased from FASMAC (Kanagawa, Japan). The mRuby2 gene construct was a gift from Michael Lin (Addgene plasmid #40255). For Ultramarine and ShadowR, the respective genes were inserted into the vector by replacing EGFP. Extra sequences encoding amino acid sequences MVSKGEEDN and SDEMYK were fused to the N and C termini of Ultramarine, respectively, so it would match the molecular weight of ShadowR for reasonable comparison during experiments (Fig. 1a). To construct tandem protein plasmids, the Ultramarine (DNA sequence encoding amino acid residues 1-214) or ShadowR     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 in the following section.

Quantification of FRET efficiency and maturity.
To compare FRET efficiency between mScarlet/ mRuby2 and ShadowR/Ultramarine in HeLa cells, we fitted the fluorescence lifetime curve with a double exponential function convolved with an instrument response function, G(t), assuming that two fractions exist in the cells: (1) mature donor fluorescent protein (i.e., mScarlet or mRuby2) fused to an immature acceptor fluorescent protein (i.e., ShadowR or Ultramarine); (2) mature donor fused to a mature acceptor where FRET occurs and the fluorescence lifetime of the donor gets shorter:  (2), erfc is a complementary error function, t 0 is the time offset, σ G is the standard deviation of the IRF, and P free and P FRET are the populations of the free donor (i.e., donor fused to immature acceptor) and donor with FRET (i.e., donor fused to matured acceptor), respectively. τ free and τ FRET are the fluorescence lifetime of free donor and donor with FRET, respectively 9,11,41 . τ free can be independently measured (mScarlet (3.69 ns), mRuby2 (2.45 ns)). By fixing these values in Eq. (2), we obtained τ FRET values for mRuby2-Ultramarine (0.84 ns), mRu-by2-ShadowR (0.97 ns), mScarlet-Ultramarine (1.60 ns), and mScarlet-ShadowR (1.79 ns). The mean FRET efficiency (Y FRET ) between the donor and the mature acceptor was calculated as follows: FRET FRET free Using the obtained τ free and τ FRET values, we calculated the fraction of the mature acceptor or the binding fraction (donor undergoing FRET) in individual cells using the following formula as described elsewhere 9,11 : FRET free free m free FRET free FRET m

Data Availability
The data generated and analyzed during the current study are included in this published article and its supplementary information files. Datasets are available from the corresponding author on request.