Two Distinct Fluorescence States of the Ligand-Induced Green Fluorescent Protein UnaG

Abstract

UnaG is a recently discovered ligand-induced fluorescent protein that utilizes bound bilirubin (BR) as its fluorophore. The fluorescence of the UnaG-BR complex (holoUnaG) compares in quantum efficiency to that of enhanced green fluorescent protein, but it is superior in that the fluorophore formation is instantaneous and not dependent on oxygen; hence, much attention has been paid to UnaG as a new fluorescent probe. However, many important molecular properties of fluorescent probes remain unknown, such as the association/dissociation rates of BR, which determine the stability thereof, and the dispersibility of UnaG in aqueous solutions, which influence the functions of labeled proteins. In this study, we found, in the process of investigating the association rate, that the holoUnaG takes two distinct fluorescence states, which we named holoUnaG₁ and holoUnaG₂. The holoUnaG₁ initially forms after binding BR and then changes to the brighter holoUnaG₂ by a reversible intra-molecular reaction, thereby finally reaching an equilibrium between the two states. Spectroscopic analysis indicated that the intra-molecular reaction was associated not with a chemical change of BR but with a change in the environmental conditions surrounding BR. We also revealed that the molecular brightness ratio and equilibrium population ratio of the two states (holoUnaG₁/holoUnaG₂) were 1:3.9 and 6:4, respectively, using photon number counting analysis. From these results, we have suggested a novel schema, to our knowledge, for the formation of the UnaG and BR complex system and have determined the various rate constants associated therein. Additionally, using analytical ultracentrifugation, we established that UnaG in the apo-state (apoUnaG) and the holoUnaG are monomeric in aqueous solution. These findings provide not only key information for the practical use of UnaG as a fluorescent probe, but also the possibility for development of a brighter UnaG mutant by genetic engineering to constitutive holoUnaG₂.

Introduction

Fluorescent proteins (FPs) are indispensable tools for the life sciences. Fluorescent probes are used often to visualize the dynamics of labeled proteins and act as fluorescent sensors for various signals in living cells, such as Ca²⁺, pH, halide, and ATP concentration sensors (1, 2, 3, 4, 5, 6, 7). Beginning with the original discovery of the green fluorescent protein (GFP) from Aequorea Victoria, repeated improvements of fluorescent properties, like the expansion of the FP color palette and discoveries of GFP-like proteins from coral reefs, have made FPs even more powerful. Moreover, recent developments of ligand-induced fluorescent proteins (LIFPs) have provided further options for researchers using FPs. Near-infrared LIFPs, which utilize covalently bound biliverdin-related molecule (8, 9), surpassed the longer wavelength emission barrier of GFP-like fluorophores, leading to deep tissue/body live-cell imaging. Green LIFPs that utilize a non-covalently bound flavin-related molecule (10, 11) have been characterized by the features of instantaneous and oxygen-independent fluorophore formation, which the GFP family proteins do not have.

UnaG is a recently discovered green LIFP, which is a member of the fatty-acid-binding proteins (FABPs) and is the first FP derived from vertebrates (12). UnaG in the apo state (apoUnaG) specifically binds bilirubin (BR) inside the β-barrel and forms the UnaG-BR complex (holoUnaG). The holoUnaG generates fluorescence in high quantum efficiency (QE), ∼50%, which is equivalent to that of enhanced GFP, one of the brightest GFP mutants, although free BR in aqueous solution is almost dark. By utilizing the binding specificity to BR and the ligand-induced fluorescence, quick measurements of indirect BR concentrations in blood have been proposed as an application of UnaG (12). Moreover, UnaG has gained attention as a fluorescent probe because of its high fluorescence QE and the features of instantaneous and oxygen-independent fluorophore formation, as compared to GFP-family FPs in general, which require several hours for formation of fluorophores by post-transcriptional modification involving a molecular oxygen. Indeed, by employing these advantages, UnaG already has been applied to live-cell fluorescence imaging in a number of studies, as a fluorescent indicator for protein-protein interactions by bi-molecular fluorescence complementation (13), as a fluorescence sensor for hypoxic conditions in tissues (14), and as a fluorescent probe for BR distribution in cells (15).

Despite the great potential of UnaG as a fluorescent probe, important molecular properties for its use as an FP remain unknown. These include the association/dissociation rate constants of ligands that govern the fluorescence stability of LIFPs, and their dispersion in aqueous solution that would influence the functions of labeled proteins. As an example of the former, if the dissociation rate is not negligible, frequent exchange of the ligand could prevent photo-bleaching (16, 17) and realize bright and stable fluorescent imaging even under conditions of high-intensity excitation light (18). As an example of the latter, DsRed, the first red fluorescent protein discovered from coral reefs, forms a strong tetramer that disturbs some functions of labeled proteins in cells (19); therefore, monomeric red FPs were developed by genetically engineering DsRed (20).

In this study, while measuring the association/dissociation rates of UnaG and BR, we discovered two distinct fluorescence states in holoUnaG and the reversible transition between the two states. From these results, we have suggested a novel schema, to our knowledge, of the UnaG and BR complex system, and we have determined the various rate constants associated therein. Moreover, we also demonstrated that both apoUnaG and holoUnaG exist as monomers in aqueous solution even under conditions of 300 mM NaCl.

Materials and Methods

Chemicals

BR, dimethylsulfoxide (DMSO), and ampicillin sodium were purchased from Wako Pure Chemicals Industries (Osaka, Japan). Isopropyl-β-D-thiogalactopyranoside (IPTG) was purchased from Takara Bio (Shiga, Japan). Imidazole was purchased from Tokyo Chemical Industry (Tokyo, Japan). BR was dissolved in DMSO at 1 mM and stored at −20°C until use.

Preparation of proteins

A.K. and A.M. provided the plasmid encoding GST-UnaG (pGEX-UnaG). For large-scale purification to perform solution studies of UnaG, we made a 6×His-UnaG (a UnaG tagged by six histidine residues at the N-terminus) expression vector by transferring the UnaG coding region to the multiple cloning sites of pColdI (Takara Bio) using BamHI and EcoRI (pColdI-UnaG). The pColdI-UnaG was introduced into BL21 chemically competent Escherichia coli cells (Takara Bio). The cells were pre-cultivated in lysogeny broth with ampicillin (100 μg/mL) for 3 h at 37°C, and then grown for UnaG expression at 16°C for an additional 18 h with 0.5 mM IPTG. Cells were harvested by centrifugation and washed once by suspending in 20 mM potassium phosphate buffer adjusted to pH 7.5 (K-Pi buffer). Packed by centrifugation, cells were stored at −80°C until use. Packed cells were thawed and homogenized by sonication in K-Pi buffer, and the soluble fraction of the homogenate was applied to a HisTrap HP 1 mL column (GE Healthcare UK, Buckinghamshire, United Kingdom). The 6×His-UnaG bound to the column was washed with 10 mL of wash buffer (20 mM K-Pi, 500 mM NaCl, and 20 mM imidazole (pH 7.5)) and eluted with 2.5 mL of elution buffer (20 mM K-Pi, 500 mM NaCl, and 500 mM imidazole (pH 7.5)). Finally, the eluate was desalted with a PD-10 column (GE Healthcare) using K-Pi buffer containing sodium azide (0.02%, vol/vol), and the purified proteins were stored at 4°C until use. Although we used the purified 6×His-UnaG without digestion of the 6×His-tag, we refer to 6×His-UnaG as UnaG for simplification of the description in this study.

The plasmid for expression of UnaG N57A (the Arg57 residue was replaced with Ala) (12), which showed faint fluorescence, was constructed by using the KOD-Plus- Mutagenesis kit (TOYOBO, Osaka, Japan) according to the instruction manual. The correct generation of the mutation was confirmed by capillary DNA sequencing.

Analytical ultracentrifugation

Analytical ultracentrifugation sedimentation velocity analysis (AUC-SV) and sedimentation equilibrium analysis (AUC-SE) (21, 22) were performed to examine the sedimentation coefficient distribution (c(s)) and the molecular mass, respectively, of UnaG. This analysis was performed using Optima XL-I analytical ultracentrifuge (Beckman Coulter, Brea, CA) at 20°C. For AUC-SV, 400 μL of sample containing 15 μM UnaG (0.8 absorbance units at 280 nm) and 7.5 μM BR in K-Pi buffer was loaded into two channel cells. Absorbance profiles, as a function of radius, were measured at 280 and 500 nm and were collected every 6 min for 300 min (for a total of 50 times) at 55 K Rpm. Absorbance profiles were globally fitted by a continuous c(s) model using SEDFIT software (Peter Shuck, National Institutes of Health; http://www.analyticalultracentrifugation.com/). The c(s) range for the analysis was set at 0.5–15 S for examination in wide range and at 0.2–3 S for validation of peak shape, with a resolution of 200 in both ranges. For AUC-SE, 100 μL of a 1:1 mixture of UnaG and BR at 9.4, 5.6, and 3.8 μM (0.5, 0.3, and 0.2 absorbance units, respectively, at 280 nm) in K-Pi buffer was loaded into six channel cells. The absorbance profiles were collected at 25,000 Rpm and globally fitted using the ORIGIN software package supplied by Beckman Coulter. The density of the K-Pi buffer and the partial specific volume (Vp) of 6×His-tagged UnaG were calculated using the SEDNTERP software (http://sednterp.unh.edu/) as 1.002 g/mL and 0.721 cm³/g, respectively.

Spectroscopic studies

Measurements of excitation and fluorescence spectra and the fluorescence intensity (FI) time course of holoUnaG were performed using an RF-5300PC spectrophotometer (Shimadzu, Kyoto, Japan). For the measurements of the FI time course, the excitation and emission wavelength were set at 490 and 525 nm, respectively. Circular dichroism (CD) and absorbance spectra were recorded using a J-820 spectropolarimeter (JASCO, Tokyo, Japan). The formation of holoUnaG was triggered by the addition of a small volume of high-concentration BR dissolved in dimethylsulfoxide to 2 mL apoUnaG in a 10-mm-path quartz cuvette. To prevent the denaturation of UnaG, the final concentrations of DMSO were at most 0.2% (vol/vol) in all experiments. All data were collected at room temperature.

FI distribution analysis

FI distribution analysis (FIDA) (23, 24) was performed with an MF20 single-molecule fluorescence spectroscopy system (Olympus, Tokyo, Japan). We used the adjustment Dye Kit 488 nm (MF-D488PX-2) for control measurements. We measured the mixture of 100 nM UnaG and 10 nM BR in K-Pi buffer and the control sample (adjustment Dye Kit 488 nm, MF-D488PX-2) in glass-bottom 384-well dishes. Adding extra UnaG was necessary to reproduce the results, most likely due to non-specific adsorption of proteins to the dish surface; therefore, excess UnaG was considered as sacrificial protein for covering the dish surface area. All the samples were excited by a 488 nm Ar laser with a beam scanner at 3600 Rpm. Data collection time was set at 10 s for all experiments.

Results

FI increase in two phases after mixing of UnaG and BR

At first, we tried to measure the association rate of UnaG and BR by analyzing the kinetics of the FI time course spectrophotometrically after mixing UnaG and BR. However, in the process of the evaluation, we found that the FI time course did not agree with the conventional model, which assumes a simple fluorescent complex formation by binding BR. Fig. 1 A shows the FI time courses after adding BR to various concentrations of apoUnaG solution in a quartz cuvette. The FI time courses increased apparently in two phases, an immediate rapid phase followed by a slow phase. The rapid phase would correspond to the formation of holoUnaG by binding BR, because 1) the complex formation is the primary event after mixing of UnaG and BR, and 2) the amplitudes of each rapid phase (FI₀) were proportional to the UnaG concentration, which is equivalent to the amount of generated holoUnaG in the presence of excess BR (Fig. 1 B). In addition, the association of apoUnaG and BR was expected to proceed rapidly because of the high affinity of UnaG for BR (K_d = 98 pM) (12). On the other hand, the slow phase of the FI time course, normalized by each FI₀, seemed to show the same exponential curve (Fig. 1 C). The fitting analysis of each curve was achieved using Eq. 1 as the fitting function:

$$F{I}_{\text{nor}}\left(t\right)=A\left[1-\text{exp}\left(-k\times t\right)\right]+1,$$ (1)

with the single rate and amplitude (k = 4.34 × 10⁻³ ± 0.18 s⁻¹ and A = 0.78 ± 0.02; Fig. 1 D). The simple exponential kinetics of the slow phase, without dependence on the concentrations of holoUnaG and free BR, indicated that the slow phase was not caused by association between holoUnaG itself, holoUnaG and apoUnaG, and apoUnaG/holoUnaG and free BR.

Figure 1.

Two distinct rates in the FI time courses upon the mixing of UnaG and BR. (A) FI time courses after mixing UnaG and BR. BR was added to apoUnaG solution (1:1 BR/UnaG) at various concentrations (in ascending order, 0.25, 0.5, 1.0, 1.5, and 2.0 μM in final concentration) at 0 s. (B) Plot of the FI amplitude increases in the rapid phase (FI₀) versus the UnaG concentrations in the solution. (C) Normalized FI time courses in the slow phase displayed in (A) (black dots) with fitting curves achieved using Eq. 1 (gray lines). The time courses were normalized by the respective FI₀ and are presented with each 0.5 increment on the vertical axis for visibility. (D) Plot of the time constants and amplitudes of FI enhancement obtained from the fitting analysis in (C). The time constants (open circles, left axis) and the amplitudes (open squares, right axis) had no dependence on UnaG concentrations and were determined to be k = 4.34 × 10⁻³ ± 0.18 s⁻¹ and A = 0.78 ± 0.02 on average. Each curve presented in (A) and (C) is a typical curve selected from four independent experiments at each UnaG concentration. Data are averages from four independent experiments, and error bars in (B) and (D) indicate standard deviations.

The slow phase is also not caused by dissociation of holoUnaG

The kinetic analyses in Fig. 1 did not provide information about the possibility of involvement of dissociation of holoUnaG assemblies in the slow FI increase. Next, to investigate this possibility, we examined the sedimentation coefficient distribution (c(s)) of an equimolar mixture of apoUnaG and holoUnaG (7.5 μM each) using the AUC-SV in the same buffer as for the data presented in Fig. 1. Fig. 2 A shows the absorbance profiles measured at 280 nm, where both apoUnaG and holoUnaG were detected, with fitting curves by SEDFIT using c(s) values as parameters. The resulting c(s), with a single peak at ∼1.8 S (Fig. 2 B), indicated that both apoUnaG and holoUnaG have the same sedimentation coefficient of ∼1.8 S. Moreover, the absorbance profiles measured at 500 nm, where only holoUnaG was detected due to the absorbance of bound BR, provide almost the same c(s) value (Fig. S1), supporting the scenario of identical sedimentation coefficients for apoUnaG and holoUnaG. Additionally, almost the same c(s) value was obtained under conditions of high NaCl concentration (300 mM) (Fig. S2). The sedimentation coefficients of ∼1.8 S together with the primary structure of UnaG (140 amino acids, including the 6×His-tag and linker) suggest that both apoUnaG and holoUnaG exist as monomers in K-Pi buffer. These results show that the slow FI increase was also not caused by dissociation of holoUnaG assemblies.

Figure 2.

Analytical ultracentrifugation of apoUnaG and/or holoUnaG. (A) (Top) Sedimentation velocity absorbance profiles of an equimolar mixture of apoUnaG and holoUnaG (7.5 μM each) in K-Pi buffer (colored dots; higher color temperature indicates later time points) with fitting curves (black lines), using c(s) values as parameters with SEDFIT software. (Bottom) Residuals of the fittings in the top plot. The absorbance profiles were obtained at 55 K Rpm and 20°C, and were collected every 6 min for 300 min. Scans presented here are every 24 min for clarity. The fitting analysis was performed with a grid resolution of 200 for 0.5–15 S. (B) The c(s) as a result of the fitting analysis in graph (A). (Inset) The detailed c(s) obtained by limiting the range of c(s) to 1.0–2.5 S with a grid resolution of 200. Both c(s) curves showed a single peak without a shoulder. (C and D) (Top) Sedimentation equilibrium absorbance profiles for various concentrations of apoUnaG (C) and holoUnaG (D) (colored circles; blue, 9.4 μM; yellow, 5.6 μM; and red, 3.8 μM), and the fitting curves (black lines) that are analyzed globally by a nonlinear least-squares method. (Bottom) Residuals of the fittings in the top plots. The absorbance profiles were obtained at 25 K Rpm and 20°C.

We next confirmed the dispersion of monomers by directly determining the molecular masses of apoUnaG and holoUnaG using AUC-SE. We show the absorbance profiles of apoUnaG (Fig. 2 C) and holoUnaG (Fig. 2 D) in sedimentation equilibrium (after centrifugation for 16 h), with fitting curves given by global fitting analysis, and the residuals of the fittings (Fig. 2, C and D, lower plots). With the fitting analysis, the molecular masses of apoUnaG and holoUnaG were determined to be 20,928 and 21,299 Da, respectively, similar to the molecular masses calculated from the primary structures of apoUnaG and holoUnaG (including the 6×His-tag and linker), which are 18,651 Da and 19,236 Da, respectively. These results confirmed monomeric dispersion of both apoUnaG and holoUnaG in K-Pi buffer.

The kinetic analyses (Fig. 1) and the monomeric dispersion of both apoUnaG and holoUnaG (Fig. 2) indicated that the slow FI increase was associated with an intra-molecular reaction within the already formed holoUnaG molecule. These results suggested that holoUnaG exists in two states with different FIs, which we have named holoUnaG₁ and holoUnaG₂. The holoUnaG₁ molecule initially forms after binding BR and then changes to the brighter holoUnaG₂ molecule by this intra-molecular reaction.

The two distinct fluorescence states of holoUnaG

To clarify the presence of the two states of holoUnaG, we examined the distribution and molecular brightness ratio of holoUnaG₁ and holoUnaG₂ at the steady state of fluorescence, i.e., at the plateau of the slow phase, using FIDA (23). In Table 1, we have summarized the results of FIDA for the standard solution (Rhodamine-110) and the holoUnaG solution at steady state. The result from the Rhodamine-110 solution was well fitted by single-component analysis, whereas the holoUnaG sample showed good agreement with the two-component model showing a population ratio of 6:4 and molecular brightness ratio of 1:3.9 for holoUnaG₁/holoUnaG₂. This co-existence of holoUnaG₁ and holoUnaG₂ at steady state demonstrates the presence of the two states of holoUnaG and the reversibility of the intra-molecular reaction thereby, finally reaching equilibrium between two states.

Table 1.

Results of Fluorescence Intensity Distribution Analysis

Sample	q1 (kHz)	q2 (kHz)	C1 (%)	C2 (%)	χ2
Rhodamine-110	60.5 ± 0.4	–	100	–	1.0
holoUnaG	11.5 ± 2.1	44.9 ± 1.3	60.3 ± 1.6	39.7 ± 1.6	1.5

The results from Rhodamine-110 and holoUnaG were fitted by single- and two-component analysis, respectively. The variables q1 and q2 are the photon count rates of the components, i.e., the brightness of respective molecular species. C1 and C2 are the distributions of the respective molecular species. In all cases, the background photon count rate was fixed at 0.74 kHz, as measured with K-Pi buffer. Data were averaged from 10 independent experiments and are represented as the mean ± SD.

Details of the UnaG and BR complex system are shown in Fig. 3 to provide a more comprehensive understanding of its overall schema. The holoUnaG₁ forms through the binding of apoUnaG and BR before changing to the brighter holoUnaG₂ by a reversible intra-molecular reaction (k₁₂, k₂₁). The BR dissociation pathways for holoUnaG₁ and holoUnaG₂ (holoUnaGs) are represented by dashed arrows because the results obtained in this study indicated that 1) at least either k_off1 or k_off2 was not 0, and 2) the unique values of k_off1 and k_off2 could not be determined (as described below; see Fig. 6). If we had obtained the transient dynamics of the distribution of the two states during the slow phase, we could have clarified the reaction model. Note, however, that FIDA can be performed only at the steady state, since calibration requires at least ∼5 min.

Figure 3.

Proposed schema of the UnaG and BR complex system. The holoUnaG₁ was initially formed by the binding of apoUnaG and BR. Then, by a reversible intra-molecular reaction, holoUnaG₁ changed to holoUnaG₂ and brightness increased by a factor of 3.9. The apoUnaG, holoUnaG₁, and holoUnaG₂ finally reached a state of equilibrium. The BR dissociation pathways for holoUnaG₁ and holoUnaG₂ are indicated by dashed arrows, because 1) at least either k_off1 or k_off2 was not 0, and 2) the unique values of k_off1 and k_off2 could not be determined (for details, see main text).

Figure 6.

Dissociation rate constants of UnaG and BR. Shown is the FI time course upon addition of 80× excess N57A (weak-fluorescent mutant) apoUnaG to the WT holoUnaGs solution at the equilibrium state (open circles) and the fitting curve (gray line). The FI decay was well fitted with a single-exponential function with a time constant of 2.17 × 10⁻⁴ s⁻¹.

Spectroscopic studies to investigate the change in the BR fluorophore between the two states

To investigate the change in the BR fluorophore between holoUnaG₁ and holoUnaG₂, we measured the excitation, fluorescence, CD, and absorption spectra of the solution at the start and the plateau of the slow phase (at 0 and 10 min from the start of the slow phase), and examined each of the spectral changes between the two time points (Fig. 4). As for the excitation and fluorescence spectra, the intensity increased ∼1.7-fold (Fig. 4 A, inset) with the same spectral shapes, indicating that the intra-molecular reaction elevated the fluorescent efficiency of BR without change of the chemical structure. Likewise, no changes were observed in either the intensity or the spectral shape of the CD and absorption spectra (Fig. 4 B, upper and lower, respectively), indicating that the intra-molecular reaction did not alter the chemical structure or the conformation of BR. The absence of changes in the CD and absorption spectra also confirmed that the amount of bound BR remained unaltered during the slow phase. In addition, there were no changes in the CD spectra at the ultraviolet range, indicating that the secondary structure of UnaG was preserved during the intra-molecular reaction (Fig. S3).

Figure 4.

Spectroscopic studies to probe the change in bound BR. (A) The excitation and fluorescence spectra of 1 μM holoUnaG (both UnaG and BR concentrations were 1 μM) at 0 and 10 min from the start of the slow phase (black and red lines, respectively). Each spectrum was normalized by its own FI maximum. (Inset) The spectra normalized by the FI maxima at 0 min. (B) The CD (top) and absorbance (bottom) spectra of 1 μM holoUnaG are presented in colors corresponding to those in (A) (UnaG and BR concentrations were 1 and 5 μM, respectively; the condition was set for excess BR in solution to prevent BR dissociation for measurements over a long time period).

Analysis of the transition rates between holoUnaG₁ and holoUnaG₂

We then estimated the individual values of k₁₂ and k₂₁ by analyzing the normalized FI time course of the slow phase in Fig. 1 C with the function expressing the FI time course using only k₁₂ and k₂₁ as parameters. The function was obtained by solving differential equations for the schema (Fig. 3) under the condition that 1) the population of holoUnaG₁ was 100% at t = 0, as we could assume that no apoUnaG existed after adding to BR for a very large k_on value (as described below; see Fig. 5; Table 2), and 2) apoUnaG did not exist in the solution because of the high affinity of UnaG and BR (for derivation of the function, see the Supporting Discussion). We obtained

$${\text{FI}}_{\text{nor}}\left(t\right)=\left(\alpha -1\right)\times k_{12}/\left(k_{12}+k_{21}\right)\times \left\{1-exp\left[-\left(k_{12}+k_{21}\right)\times t\right]\right\}+1,$$ (2)

where α indicates the holoUnaG₂/holoUnaG₁ brightness ratio, now equal to 3.9, as obtained by the FIDA (Table 1). The time constant in the function (k₁₂ + k₂₁) corresponds to the apparent transition rate of the slow phase in Fig. 1 D (k = 4.34 × 10⁻³ s⁻¹), and the ratio 1/k₁₂:1/k₂₁ should correspond to the molecular distribution of 6:4 with the two states at equilibrium (Table 1). The k₁₂ and k₂₁ values satisfying the two restrictions are determined to be 1.72 × 10⁻³ and 2.62 × 10⁻³ s⁻¹, respectively (see Table 2, middle column).

Figure 5.

The association rate constant of UnaG and BR. (A) The FI time courses upon addition of various concentrations of BR (5, 10, and 20 nM in final concentration) to apoUnaG solutions (50 nM fixed concentration). The total concentrations of BR and UnaG in final concentration are denoted by [BR]₀ and [apoUnaG]₀, respectively. (Inset) The FI time courses focused on the very initial phase and the fitting lines are shown (in order, from the bottom, [BR]₀ = 5, 10, and 20 nM). The right axis shows the concentration of holoUnaG₁corresponding to the raw FI value. The method for converting the raw FI value to holoUnaG₁ concentration is described in the main text. (B) Plot of the initial velocities of holoUnaG₁ generation (V_init) versus [BR]₀ (open circles) and the linear regression line (dashed gray line). The association rate constant (k_on) was obtained from the slope of this plot (k_on = 7.05 × 10⁶ M⁻¹ s⁻¹).

Table 2.

Kinetic Rate Constants of the UnaG and BR Complex System with or without NaCl

Contents	0 mM NaCl	150 mM NaCl
kon	7.05 × 106 (M−1 s−1)	1.42 × 106 (M−1 s−1)
koff	2.17 × 10−4 (s−1)	5.13 × 10−4 (s−1)
k	4.34 × 10−3 (s−1)	6.69 × 10−3 (s−1)
k12a	1.72 × 10−3 (s−1)	2.66 × 10−3 (s−1)
k21a	2.62 × 10−3 (s−1)	4.03 × 10−3 (s−1)

The k₁₂ and k₂₁ rates were calculated from the apparent rate, k, using Eq. 2, the brightness ratio (1:3.9), and the population ratio (6:4) of holoUnaG₁ and holoUnaG₂.

^aValues determined by calculation.

The association rate constant of UnaG and BR

Next, we determined the association rate (k_on) of UnaG and BR by analyzing the initial velocity of the rapid phase on the FI increase. To slow the FI increase, BR and apoUnaG were mixed at low concentrations. We set the total concentration of BR ([BR]₀) and UnaG ([apoUnaG]₀) within 50 nM in the final concentration, and we measured the FI time courses upon mixing (Fig. 5 A). As described in Fig. 1, the FI time courses do not exhibit a simple exponential curve due to the FI enhancement in the slow phase. Then, we focused on the very initial part of the rapid phase, because the production of holoUnaG₂ and the resulting FI enhancement were negligible (Fig. 5 A, inset). To determine the absolute value of k_on, the right axis of the inset figure shows the concentration of holoUnaG₁. For this conversion, we used the relationship that the FI₀ for 1 nM holoUnaG₁ equals 5.60, which was measured in the same manner as for Fig. 1 C, with the spectrophotometer setup identical to that for Fig. 5 A. The plot of the initial velocities of holoUnaG₁ generation (V_init) versus [BR]₀ is shown in Fig. 5 B. Using the result of linear regression of this plot and the relationship of V_init = k_on × [apoUnaG]₀ × [BR]₀, the associated rate constant was determined to be k_on = 7.05 × 10⁶ M⁻¹ s⁻¹.

The dissociation rate constant of UnaG and BR

For the measurement of the apparent dissociation rate from holoUnaGs (k_off), we utilized the weak-fluorescent UnaG N57A mutant ((12); brightness <3% compared to wild-type (WT) UnaG) as an adsorbent of the free BR dissociated from holoUnaGs. First, we validated that the N57A mutant had a high affinity for BR, comparable to that of WT UnaG, from the fact that the FI of an equimolar mixture of WT holoUnaG and holo N57A mutant (1 μM each) without free BR at steady state decreased by half compared to the FI of 1 μM WT holoUnaG (Fig. S4). Fig. 6 shows that exponential FI decay was induced by the addition of a large excess of N57A (∼80× molar ratio) to WT holoUnaG solution at 0 min. Because the high concentration of N57A quickly depleted, the free BR dissociated from the WT holoUnaG in the solution. The time constant observed in the FI decay corresponded to the apparent dissociation rate of BR from the WT holoUnaG (k_off = 2.17 × 10⁻⁴ s⁻¹). The affinity calculated from k_off/k_on was K_d = 31 pM and was in good agreement with the reported affinity (K_d = 98 pM), supporting the validity of our measurement of the two rates.

As mentioned above, the unique values of k_off1 and k_off2 could not be determined, because it was impossible to distinguish between the individual BR dissociations from holoUnaG₁ and holoUnaG₂. However, the relationship k_off, k_off1, and k_off2 was obtained by solving the differential equations under the condition that the concentration of free BR is negligible (for derivation of Eq. 3, see the Supporting Discussion):

$$k_{\text{off}}=\left(-A+\sqrt{A^2-4B}\right)/2,$$ (3)

where A and B are defined as

$$A=k_{12}+k_{21}+k_{\text{off}1}+k_{\text{off}2}$$

$$B=k_{21}\times k_{\text{off}1}+k_{12}\times k_{\text{off}2}+k_{\text{off}1}\times k_{\text{off}2}.$$

The k_off1 and k_off2 are 0 or positive values that satisfy Eq. 3, i.e., k_off1 and k_off2 are restricted to a point on certain line in (k_off1, k_off2) space. As an example, the line expressing the possible set of (k_off1, k_off2) is shown in Fig. S5, using the parameters obtained in this study (Table 2, middle column: k_off = 2.17 × 10⁻⁴ s⁻¹, k₁₂ = 1.72 × 10⁻³ s⁻¹, and k₂₁ = 2.62 × 10⁻³ s⁻¹).

Effect of chloride ion concentration on the interaction of UnaG and BR

The effects of solution conditions on the interaction properties of UnaG and BR were assayed by measuring k, k_on, and k_off at high chloride ion concentration (K-Pi buffer containing 150 mM NaCl) in the same manner as for the data in Figs. 1, 5, and 6 (data not shown), and the results are summarized in the rightmost column of Table 2 to compare with the rates under the condition of 0 mM NaCl. We also measured the distribution and brightness ratios of holoUnaG₁ and holoUnaG₂ at 150 mM NaCl using FIDA. All rate constants were affected by the chloride ion, whereas the distribution and the brightness ratios were not. k_on was lowered by a factor of ∼0.20 by increasing NaCl to 150 mM, whereas k and k_off were increased by factors of ∼1.54 and 2.36, respectively. Moreover, we also calculated k₁₂, k₂₁ at 150 mM NaCl and summarized these values in the rightmost column of Table 2.

Discussion

Dispersibility of UnaG and other FPs

The dispersibility of FPs should be investigated to perform live-cell fluorescence imaging correctly, because the oligomer formation, in most cases, disturbs the function of labeled proteins, leading to abnormal cell behavior. As described above, the FPs derived from coral reefs tend to form strong oligomers, so the genetically engineered monomeric versions of those FPs are preferred by current researchers for fluorescence imaging (19, 20, 25). In this study, we demonstrated that both apoUnaG and holoUnaG exist as monomeric forms in aqueous solution even under conditions of high NaCl concentration, at least up to 7.5 μM. We also demonstrated that both apoUnaG and holoUnaG are monomers in aqueous solution in the presence of 1% w/v of PEG 6000, at least up to 6.5 μM (Fig. S6). These facts indicate that both apoUnaG and holoUnaG exhibit monomeric dispersion under various conditions. It is well known, however, that WT GFP forms a weak dimer at K_d = 0.11 mM (26), and the aggregation of the labeled proteins accompanied by the dimerization of the WT GFP could disrupt cellular structure and function (27). From this perspective, the dispersibility of UnaG in higher concentrations should be examined further.

The differences between the experimental and calculated molecular masses in this study (Fig. 2) arose from an error in the Vp, one of the physical constants of proteins used in the fitting analysis. This is because the Vp of UnaG (0.721 cm³/g) was estimated from the primary structure by SEDNTERP software, although the Vp values for proteins are affected by the conformation of the proteins. Indeed, substituting 0.708 cm³/g for the Vp can give the theoretical molecular mass of UnaG, which is within the range of Vp values generally.

Consideration of the transition between the two states in holoUnaG

At first, we assumed that the slow phase of the FI increase arose from additional BR binding to an unknown second ligand site in UnaG. This is because bovine serum albumin and human serum albumin, which are well-known BR binding proteins, both can bind two BRs with different affinities (28, 29). However, as shown in Fig. 1, the independence of the slow-phase kinetics from the concentration of free BR indicated that the slow phase was not caused by additional BR binding. In addition, as shown in Fig. 4, the unchanged CD spectra during the slow phase also demonstrated the absence of a second BR binding site, because the binding of free BR to UnaG and the resulting conformational fixation of BR inevitably caused the change in the CD spectrum (12, 30).

We next assumed a translocation or rearrangement of bound BR to another, deeper site with a change in brightness. We were aware that the liver-type FABP (FABP1), which has a high degree of homology and similar tertiary structure to UnaG, has a bound palmitic acid that causes translocation from the initial superficial site to a deeper site with a large conformational change (31). In the case of UnaG, however, the unchanged CD spectra during the slow phase (Fig. 4) indicated that the BR bound to holoUnaG did not alter its conformation, as seen in FABP1.

The kinetics of the slow phase (Fig. 1) and the co-existence of the two states at the steady state (Table 1) indicated that the slow phase was associated with a reversible intra-molecular reaction. It is well-known that some GFP-family proteins have the protonation/deprotonation states of their fluorophores accompanying the change of absorption spectra (32, 33, 34). In the case of holoUnaG, however, the unchanged shape of the absorption spectra during the slow phase (Fig. 4 B) indicated that bound BR does not exchange protons, because the protonation/deprotonation of fluorophores would be reflected in the shape of the absorption spectra between holoUnaG₁ and holoUnaG₂.

Proposed mechanisms of transition between the two states

The results of the spectroscopic analysis indicated that the reversible intra-molecular reaction was not associated with the chemical and conformational change of BR and UnaG (Fig. 4). These results implied that the intra-molecular reaction was associated with a change in the environmental conditions surrounding BR. The effects of chloride ion on k, k_on, and k_off (Table 2) indicate that the electrostatic force may play an important role in the transition between holoUnaG₁ and holoUnaG₂, as well as in the interaction of UnaG and BR. The effects on the k_on and k_off rates could be simply explained by the charge screening effect, where the electrostatic attraction and binding affinity between UnaG and BR are reduced by the surrounding ions in a solution, as also observed in the interaction of adipocyte-type FABP and the associated ligand (35). On the other hand, the increase in k rate at higher chloride ion concentration is highly suggestive of the involvement of electrostatics in the reversible intra-molecular reaction, because the increase in k indicates the reduction of the energy barrier separating holoUnaG₁ and holoUnaG₂, i.e., the activation energy of the intra-molecular reaction.

Based on the consideration described above, a bi-stable switch in the direction of a polar residue proximal to BR is one reasonable explanation for the reversible intra-molecular interaction and the resulting FI change. We speculated that a polar residue, associated not with BR but with another atom in holoUnaG₁, can be associated with BR in holoUnaG₂ through hydrogen bonds (H-bonds) or electrostatic interactions due to the bi-stable directional change, which is driven by thermal fluctuations over the energy barrier separating the two states. Indeed, the alternative direction of a polar residue proximal to the fluorophore and the resulting alternative interaction network between fluorophore and proximal residues were observed in enhanced GFP (36). In general, these changes are reflected in the fluorescence QE of fluorophores. They include changes in the electronic state of the fluorophore, stabilization/destabilization of the fluorophore by altering the network of H-bonds and other interactions (37), and the hydration state of the fluorophore (36, 38).

Arg57 is a potential candidate residue that may be associated with BR in this manner. As shown with the significant weak fluorescence of N57A holoUnaG (12), the fluorescence QE of holoUnaG was sensitive to the relationship between Arg57 and BR. Moreover, N57A did not exhibit the slow phase of FI increase after binding BR (Fig. S7). These results indicate that Arg57 plays a key role not only in the basal fluorescence efficiency of UnaG, but also in the transition between holoUnaG₁ and holoUnaG₂.

Significance of the findings of this study

The findings of this study provide key information for the practical use of UnaG. For example, when we quantify the amount of BR in vitro or in vivo from the fluorescence signal of holoUnaG, we must take into account the equilibration time (several minutes) for the two states. The QE of holoUnaG has been reported as 51% (12), although this was measured at equilibrium for the two states. The individual QEs of holoUnaG₁ and holoUnaG₂ correspond to 23 and 92%, respectively, according to their distribution at equilibrium (6:4) and the molecular brightness ratio (1:3.9) determined in this study. We suggest that the QE of holoUnaG₂ is the highest value for BR. The discovery of a higher QE for holoUnaG₂ opens the possibility for the development of a brighter UnaG with the construction of a constitutive holoUnaG₂ by genetic engineering.

Author Contributions

T.S., Y.S., and T.A. designed the research. A.K. and A.M. provided UnaG gene. Y.S. performed experiments. Y.S. and T.S. analyzed the data. T.S., Y.S., and T.A. wrote the manuscript. All authors discussed the results and commented on the manuscript.

Editor: Amy Palmer.