1.

Introduction

The use of green fluorescent protein (GFP) as a marker to visualize specific proteins inside living cells has revolutionized cell biology.^{1, 2} The genetic sequence-encoding GFP has been optimized to facilitate both its expression in mammalian cells,^{3, 4} as well as its photophysics for best use in elucidating the distribution and intracellular trafficking of genetically tagged proteins.^{5, 6, 7, 8, 9} The advent of spectral variants of GFP has extended the utility of this technology by enabling the detection of protein-protein interactions in vivo. Specific proteins can now be tagged with florescent proteins (FPs) that fluoresce in the blue, green, yellow, orange, or red region.¹⁰ In conjunction with basic light microscopy, many molecular interactions can be implied based on simple colocalization studies.¹¹ This colocalization approach is however limited by the resolution of light microscopy (typically, $0.2\mu \mathrm{m}$ ). More sophisticated imaging technologies, including spectral imaging,¹² fluorescence lifetime imaging (FLIM),^{13, 14, 15} and imaging based on changes in fluorescence anisotropy,^{16, 17} are being applied to detect interactions between FP-tagged proteins at a resolution of $1–10\mathrm{nm}$ . One such approach is to monitor Förster resonance energy transfer (FRET) between a protein tagged with a donor-FP and a putative binding partner labeled with an acceptor FP. FRET efficiency can be measured by monitoring reciprocal change in the intensity of emission from donor and acceptor fluorophores. More advanced and precise measurement of FRET efficiencies can be achieved by monitoring the change of fluorescence lifetime of the donor fluorophore. A high transfer efficiency is often interpreted as an indication of close proximity $(<10\mathrm{nm})$ . Another approach, also based on FRET, is emerging as a method for dynamically monitoring the multimeric state of protein assemblies.^{17, 18, 19} In this method, fluorescence anisotropy is used to measure energy migration between proteins labeled with the same FP. High anisotropy is indicative of low FRET and vice versa. Furthermore, the amplitude of a drop in anisotropy encodes information on the number of fluorophores in a molecular complex that are participating in the energy migration.

The accurate interpretation of biological FRET experiments utilizing complex quantitative imaging modalities, such as anisotropy analysis or FLIM, relies on an in-depth knowledge of the photophysics of the fluorophores being used. Specifically, for FLIM, it is important to know the lifetime of a fluorophore in the absence of acceptors, and it is also essential to appreciate any other factors that might alter its fluorescence lifetime in the course of an experiment.^{20, 21, 22} For anisotropy analysis, it is important to know the fundamental anisotropy of a fluorophore (primarily a function of the orientation of its absorption and emission dipoles), how rigidly the fluorophore is attached to a protein of interest and it is also useful to understand any other factors that might alter the orientation of its emission relative to its excitation.^{20, 21, 22} Furthermore, these sophisticated types of imaging experiments are often difficult to interpret if other endogenous fluorophores are present. Under these circumstances, measured fluorescence lifetimes or anisotropies will be a function of both the fluorophores of interest, as well as the contaminating fluorophores. Unfortunately contaminating fluorophores, like flavins, nicotinamides, porphyrins, etc., are almost always present in living biological samples.

Two of the most popular spectral variant of GFP used for biological FRET studies are Cerulean,²³ a blue emitting FP, and Venus,²⁴ a yellow emitter. The crystal structure of both of these proteins have been reported.^{25, 26} Although the photophysics of GFP has been extensively studied,^{27, 28, 29, 30} this knowledge base may not universally apply to these spectral variants. Furthermore, it has long been suggested that in the FP family FRET might occur between an internal tryptophan and the FP fluorophore.^{31, 32, 33} If true, this could significantly alter fluorescent protein lifetime and anisotropy in a wavelength-dependent manner, and potentially complicate the interpretation of FRET between Cerulean and Venus. In this study, we characterize the lifetime and anisotropy of purified Cerulean and Venus and in so doing reveal: (i) Cerulean and to a lesser extent Venus fluorescence both decay as a double exponential, (ii) These fluorophores are rigidly fixed within their beta-barrel structure, and (iii) they both express efficient energy transfer between an internal tryptophan and their fluorophores upon UV excitation. Because the fluorophore within Cerulean and Venus are sequestered from the solvent/media due to their protein structure, we find they are resistant to dynamic quenching. We demonstrate how this can be exploited to remove the effects of endogenous fluorophores in these types of FRET experiments.

2.

Materials and Methods

Tris buffer (Trizma), Acrylamide (electrophoresis grade), flavin mononucleotide (FMN), and KI were from Sigma-Aldrich Inc., St. Louis, Missouri. 20X PBS, pH-7.5 was from AMRESCO®, Solon, Ohio. The water used for the analysis was deionized (from Millipore® deionizer unit). Both Cerulean and Venus were diluted in $0.1\mathrm{M}$ Tris, pH 8.5. In the iodide quenching studies with FMN PBS pH 7.5 was used as the buffer as fluorescence of FMN is sensitive to alkaline pH.

3.

Results and Discussion

The purity of Cerulean and Venus were accessed using Coomasie staining of the proteins and Western blot analysis (Fig. 1 ). The serial dilution of Cerulean and Venus (1.4 and $0.14\mathrm{mg}$ ) reveals a strong band at $27\mathrm{kDa}$ (FP, arrowheads in Fig 1) and some minor contaminants (double-headed arrows, Fig 1). Densitometry analysis of the Coomassie staining [Fig. 1a] revealed that Cerulean was 97% pure and Venus was 90% pure.

Fig. 1

Characterization of purified proteins. (a) Coomassie staining of a SDS PAGE gel shows $27\mathrm{kDa}$ (arrow heads) Cerulean (C) and Venus (V) protein bands. There seems to be minor contaminating bands at $60\mathrm{KDa}$ (double-headed arrow). (b) Western blot of Cerulean and Venus protein, where the Anti GFP antibody strongly cross reacts with the $27\mathrm{kDa}$ protein.

3.1.

Steady-State Fluorescence

Cerulean–Venus is an excellent donor acceptor pair, with selective donor excitation using the 405 or $442\text{-}\mathrm{nm}$ laser lines, and observation at $470\mathrm{nm}$ (no signal from acceptor at this wavelength). This significantly simplifies the FRET measurements and its interpretation. The excitation and emission spectra of both proteins are presented in Fig. 2 . There is a significant increase in the excitation intensity in the UV region, at $<300\mathrm{nm}$ . Emission spectra did not change with excitation (data not shown), suggesting either an electronic transition in the chromophore or that excitation energy absorbed by other moieties (such as aromatic amino acids) are efficiently transferred to the chromophore.

Fig. 2

Excitation and emission spectrum of (a) Cerulean and (b) Venus fluorescent proteins. Fluorescent protein excitation intensity in the UV $\left(280\mathrm{nm}\right)$ is thought to be due to FRET from aromatic amino acids to the fluorescent protein chromophore.

To study these possibilities further, we measured polarization spectra. Figure 3 shows steady-state excitation and emission polarization spectra of Cereluan, while Fig. 4 shows the same for Venus protein. Whereas the anisotropy values remain relatively constant within the emission spectra (b), the excitation anisotropies decrease significantly in shorter wavelengths (top panels). Such polarization dependences are common in (i) single fluorophore with different excited-state dipoles at higher electronic transition and (ii) a donor-acceptor system when excitation anisotropy is observed at acceptor emission. In fact, single-step energy transfer significantly affects anisotropy.

Fig. 3

(a) Excitation and (b) emission steady-state anisotropies of Cerulean. The high value of emission anisotropy, $>0.3$ , suggests limited flexibility of the fluorophore within the protein structure. The rapid decrease of the excitation anisotropy is consistent with FRET between aromatic amino acids and the fluorescent protein chromophore.

Fig. 4

(a) Excitation and (b) emission steady-state anisotropy of Venus. The high value of steady-state emission anisotropy (b) suggests limited flexibility of the chromophore within the barrel structure of the protein. The negative value of excitation anisotropy in the UV region is consistent with FRET occurring between aromatic amino acids and the fluorescent protein chromophore.

3.2.

Time-Resolved Fluorescence

Next, we measured fluorescence lifetimes of fluorescent proteins with UV excitation $\left(285\mathrm{nm}\right)$ and with direct long-wavelength excitation at 405 and $470\mathrm{nm}$ for Cerulean and Venus, respectively. The intensity decays of Cereluan and Venus fluorescent proteins with long-wavelength excitation, measured by time-correlated single-photon counting, are presented in Figs. 5 and 6 , respectively. One- and two-exponential fits to the experimental data are presented in Table 1 . For both proteins, the fluorescence intensity decays were best fitted using a two-exponent decay model, though single exponential fits with lifetimes of 3.18 and $3.03\mathrm{ns}$ for Cereluan and Venus may in many cases be acceptable. Lifetimes were significantly longer with UV excitation. Furthermore, strong negative lifetime components of $\sim 0.2\mathrm{ns}$ were observed (Table 1). This is a characteristic property of a process occurring in the excited state. A negative component in the intensity decay indicates that the fluorescent protein chromophore is receiving additional “pumping” after excitation. This is consistent with FRET occurring between the aromatic amino acids and the fluorescent protein chromophore.

Fig. 5

Time-domain intensity decay of Cerulean. The decay is well approximated with a two exponential decay model (see Table 1). The amplitude and fractional intensity weighted average lifetimes are 2.82 and $3.23\mathrm{ns}$ respectively.

Fig. 6

Time-domain intensity decay of the Venus. The decay is well approximated using a two exponential decay model (see Table 1). The amplitude and fractional intensity weighted average lifetimes are 2.87 and $3.03\mathrm{ns}$ respectively.

Table 1

Multiexponential analysis of fluorescence intensity decays of Cerulean and Venus.

Protein	Ex(nm)	Em(nm)	α1	T1 (ns)	α2	T2 (ns)	α3	T3 (ns)	X2
Cerulean	405	475	1.0	$3.18\pm 0.03$	—	—	—	—	4.7
			0.403	$1.52\pm 0.04$	0.597	$3.70\pm 0.02$	—	—	0.9
	285	475	1.0	$3.47\pm 0.09$	—	—	—	—	3.1
			0.936	$2.08\pm 0.03$	1.253	$4.02\pm 0.02$	$-1.189$	$0.16\pm 0.01$	1.0
Venus	470	525	1.0	$3.03\pm 0.01$	—	—	—	—	1.7
			0.160	$1.31\pm 0.07$	0.840	$3.17\pm 0.01$	—	—	0.9
	285	525	1.0	$3.35\pm 0.09$	—	—	—	—	1.2
			1.835	$3.31\pm 0.02$	$-0.835$	$0.2\pm 0.02$	—	—	1.0

This is visualized in Figs. 7 and 8 for Cerulean and Venus respectively. Although the data were fitted for the whole decay, we present the fit only for the initial $5\mathrm{ns}$ , in order to show the necessity of negative component to fit the decay. The short decay time (0.16 and $0.2\mathrm{ns}$ for Cerulean and Venus, respectively) associated with the negative decay component shows that the energy transfer process is very efficient $(>90\%)$ if we assume that the lifetime of unquenched tryptophan or tyrosine is $\sim 3\mathrm{ns}$ . This explains the absence of a significant tryptophan emission. The low steady-state excitation anisotropy value for Cerulean (Fig. 3) and negative value for Venus (Fig. 4) observed with excitation of $<350\mathrm{nm}$ is consistent with significantly different near-orthogonal orientations for donor (Trp) and acceptor (fluorescent protein chromophore) transition moments. However, characterization of Trp donor profile was not an option due to presence of mild spectroscopic impurities (as seen in SDS-Page and Western Blot).

Fig. 7

Initial intensity decay of Cerulean with the excitation in Trp excitation region $\left(285\mathrm{nm}\right)$ . The blue line and residuals show the fit with positive component only, while the green line and residuals show the fit with one negative component. A negative component is needed to fit the data better. (Color online only.)

Fig. 8

The initial intensity decay of Venus with the excitation in Trp excitation region $\left(285\mathrm{nm}\right)$ . The blue colored line and residuals show the fit with positive component only, while green line and residuals show the fit with one negative component. A negative component is needed to fit the data better.

Anisotropy decays of both proteins (Figs. 9 and 10 ) are similar, displaying high values of initial anisotropy and correlation times of $\sim 15\mathrm{ns}$ (Table 2 ). The fits to these decays can be modestly improved using a decay model with a second correlation time. However, these second correlation times are on the order of a few nanoseconds and are most likely associated with the nonspherical shape of the protein than with chromophore internal mobility. The very high steady-state anisotropies observed as well as the relatively slow anisotropy decays measured indicates that the fluorescent protein chromophores in Cerulean and Venus are rigidly oriented within the protein structure and have little freedom for rotation. This attribute should encourage polarization studies using fluorescent proteins, because the local fluorophore mobility is a bottleneck in many rotational diffusion investigations.

Fig. 9

Anisotropy decay of Cerulean. Two components are needed to fit the data well (See Table 2). The limiting anisotropy $\left(r_0\right)$ was 0.4, and the rotational correlation times ( ${\theta }_{\mathrm{a}}$ and ${\theta }_{\mathrm{b}}$ ) were 6.7 and $19.6\mathrm{ns}$ .

Fig. 10

Anisotropy decay of Venus. Two components are needed to fit the data well (See Table 2). The limiting anisotropy $\left(r_0\right)$ was 0.4, and the rotational correlation times ( ${\theta }_{\mathrm{a}}$ and ${\theta }_{\mathrm{b}}$ ) were 3.7 and $20.3\mathrm{ns}$ .

Table 2

Anisotropy decay analyses of Cerulean and Venus emission.

Protein	r1	θ1 (ns)	r2	θ2 (ns)	θ¯ (ns)	Xr2
Cerulean	$0.352\pm 0.002$	$14.5\pm 0.5$	—	—	14.5	1.6
	$0.088\pm 0.001$	$6.7\pm 0.4$	$0.266\pm 0.002$	$19.6\pm 0.6$	18.3	1.5
Venus	$0.374\pm 0.002$	$16.4\pm 0.6$	—	—	16.4	2.0
	$0.048\pm 0.001$	$3.7\pm 0.4$	$0.331\pm 0.002$	$21.4\pm 0.7$	20.3	1.8

3.3.

Fluorescence Quenching Study

GFP fluorescence is known to be resistant to many quenching reagents. This is thought to arise from the GFP $\beta$ -barrel protein structure that physically surrounds and sequester the GFP fluorophore from water soluble factors.^{25, 26} We next wanted to determine if the $\beta$ -barrel structure of Cerulean and Venus would also protect their fluorophores from quenching reagents. We also wanted to determine if the high-efficiency energy transfer observed with UV excitation was also insensitive to external quenchers. First, we noted that with long-wavelength excitations, even high concentrations of acrylamide or potassium iodide (KI) in the solution do not change the emissions of Cereluan or Venus proteins (data not shown). This indicates that their chromophores are not readily accessible to these quenchers.

Next, we used UV excitation to test if the tryptophan fluorophore is also protected by the $\beta$ -barrel structure. Figures 11 and 12 show the effect of acrylamide and/or KI presence on the UV-excited Cereluan and Venus proteins. There was little evidence for quenching of either Cerulean or Venus fluorescence. We conclude that the fluorescent protein chromophores as well as the tryptophan residue is effectively shielded from these quenchers. One can note that, in the presence of acrylamide, the emission spectrum of our purified Cerulean (and to a lesser extent Venus) was reduced at $400–450\mathrm{nm}$ (Fig. 11). Presumably, tryptophan in a small fraction of denatured fluorescent proteins can be quenched.

Fig. 11

Effect of acrylamide on (a) the fluorescence of Cerulean and (b) Venus.

Fig. 12

Effect of KI on the (a) fluorescence spectrum of Cerulean and (b) Venus.

We decided to test if the unique protective nature of the fluorescent protein $\beta$ -barrel structure can be exploited to suppress background fluorescence in fluorescent protein experiments. We prepared the mixture of Cereluan or Venus proteins with flavin mononucleotide (FMN). Flavins are an abundant chemical moiety found in cells. A major obstacle to live-cell florescence imaging is that flavin fluorescence often masks the fluorescence of less abundant exogenous fluorophores experimentally introduced into cells. To illustrate this, a solution of either Cerulean (Fig. 13 ) or Venus (Fig. 14 ) and the flavin FMN were prepared and their emission spectra recorded. As expected, the strong FMN emission masked the fluorescent protein spectra in these mixtures. Addition of 0.35 Molar KI selectively quenched FMN to reveal the previously masked fluorescence protein spectra. To illustrate this effect, we prepared and imaged a microscopy slide with immobilized Venus protein (Fig. 15, star-shaped spot). Next, a drop of FMN solution was applied to the slide, primarily over the immobilized Venus. This led to fluorescence dominated by FMN. The star-shaped Venus spot could not be distinguished. Finally, we progressively added KI solution [Fig. 15b]. As FMN was selectively quenched, the fluorescent Venus protein spot was again revealed.

Fig. 13

Effect of KI on the fluorescence mixture of Cerulean and FMN. (a) Normalized emission spectra of Cerulean and FMN. (b) Emission spectrum of mixture in absence (solid thick lines) and presence (dotted lines) of KI. Dashed line represent normalized spectrum of mixture in presence of KI.

Fig. 14

Effect of KI on the fluorescence mixture of Venus and FMN. (a) Normalized emission spectra of Venus and FMN. (b) Emission spectrum of mixture in absence (solid thick lines) and presence (dotted lines) of KI. Dashed line represent normalized spectrum of mixture in presence of KI.

Fig. 15

4.

Conclusions

In addition to excitation with visible light $(400–550\mathrm{nm})$ both, Cereluan and Venus fluorescent proteins can be efficiently excited in the UV region at wavelengths $\left(280\mathrm{nm}\right)$ , where tryptophan and tyrosine are excited. Surprisingly, appreciable emission from these aromatic amino acids was not observed in the emission spectra of fluorescent proteins when excited with UV light $\left(285\mathrm{nm}\right)$ . UV excitation is thought to be accompanied by a very efficient energy transfer from aromatic amino acids (presumably tryptophan and/or tyrosine) to the fluorescent protein chromophore because, under these conditions, a negative lifetime decay component (pumping) was observed in the fluorescence intensity decay. The lifetime associated with this lifetime decay component was $<200\mathrm{ps}$ . Assuming a nonquenched lifetime of tryptophan/tyrosine species of $\sim 3\mathrm{ns}$ , one can calculate a transfer efficiency of $>90\%$ . With such a high-energy-transfer efficiency, it is not surprising that the fluorescence spectra of tryptophan in fluorescent proteins cannot be measured accurately. Also, presence of trace amount of impurities in the sample (Fig. 1) contribute to the signal seen at the Trp emission region, making the interpretation even more complex. Furthermore, steady-state anisotropy measurements of the Cerulean or Venus fluorophore were only low with UV excitation, again indicating energy transfer between an aromatic amino acid and the fluorescent protein chromophore.

Time-resolved fluorescence anisotropy decay analysis also revealed that the emission dipole orientation of the Cereluan and Venus chromophores are rigidly fixed in the protein structure. This conclusion is based on our inability to detect any local, subnanosecond decay component in time-resolved anisotropy measurements. The presence of trace impurities in the sample makes the interpretation of fluorescence decay from the internal tryptophan more complex. This prevented us from determining if the orientation of its emission dipole is also rigidly fixed. Nonetheless, the negative steady-state anisotropy observed with UV excitation of Venus, which contains only one tryptophan residue, strongly indicates energy transfer from tryptophan residues to the chromophore region. The excitation anisotropy (while acceptor emission was observed) value dropped for Cerulean, but still remained positive. This can be due to presence of two tryptophan residues in Cerulean.

Both, tryptophan and the fluorescent protein chromophore are buried in the protein structure and not accessible to the external quenchers, acrylamide and/or KI. This is a unique advantage of fluorescent proteins over other probes. We have shown that quenchers can be used to selectively suppress nonspecific fluorescent background. In biological samples, such as cells or tissue, fluorescent background is a major obstacle in any fluorescence study, especially in those based on FRET. Our observation that the fluorescence of purified Venus is selectively spared while the fluorescence of FMN is almost completely quenched suggests that using KI solution in washing procedures will improve the ratio of fluorescent protein emission to background fluorescence. We understand that high concentrations of iodide will be toxic to the cell itself and thus cannot be used in live-cell imaging instantaneously. But this study proves the concept that using flavin quenchers can be a good strategy of improving signal-to-noise ratio in imaging experiments that use these fluorescent proteins. Use of other known quenchers of flavin, such as oxygen or reducing flavins in the cells, can be a potentially feasible future strategy. These simple experimental manipulations might be the key for allowing quantitative imaging of fluorescent proteins in cells with high background fluorescence (such as neurons) or for readily detecting the fluorescence emission of single fluorescent protein molecules by dramatically increasing the signal-to-noise ratio.