Engineering Redox-Sensitive Linkers for Genetically Encoded FRET-Based Biosensors

Introduction

The overall intracellular reduction/oxidation (redox) environment including organellar redox status exerts a profound influence on the normal cellular processes of DNA synthesis, enzyme activation, selective gene expression, cell cycle progression, proliferation, differentiation, and apoptosis (1–4). Thus, the ability to sense the redox environment in living cells and organisms has far reaching implications for understanding and manipulating basic cellular processes that underlie complex biomedical problems (5–8). Genetically encoded biosensors enable real-time and extended assessment of alterations in intracellular metabolism without cellular disruption (9–12).

A major advance for monitoring thiol redox potentials in live cells has resulted from the development of redox-sensitive forms of the green fluorescent protein (roGFPs) and the yellow fluorescent protein (rxYFP) (13–16). These redox biosensors have been characterized biochemically. Their spectroscopic and redox properties are quite different, and pitfalls of intracellular redox sensing using single spectroscopic probes have been comprehensively discussed recently (16).

Förster resonance energy transfer (FRET) is an optical technique whereby the excited state energy of a fluorescent donor molecule is transferred non-radiatively to a ground state acceptor molecule by means of long-range resonance coupling between the donor and acceptor transition dipoles (17). The distance (between 1 and 10 nm) and orientation between the donor and acceptor transition dipoles governs the efficiency of energy transfer. Because of this spatial sensitivity and signal specificity, FRET often provides an advantage over measurements with single fluorophores (18). Recently, the development of an extensive family of genetically encoded fluorescent proteins has led to several FRET metabolite biosensors (18–20). These sensors exploit FRET between two GFP variants (21).

Currently available redox biosensors employ single GFP variants, but do not employ FRET (15). In a FRET-based biosensor, a redox event induces a molecular conformational change in the redox sensor altering the distance between the FRET donor-acceptor pair, which in turn causes a detectable change in FRET efficiency as measured by changes in the emission spectra profiles (Fig. 1). The lack of linkers capable of both separating donor from acceptor and conferring redox sensitivity is a major drawback in engineering FRET-based redox biosensors. Current progress in the development of redox polypeptide switches makes this task achievable (22–26). In addition, the ability of helical linkers to control the distance and reduce the interference between GFP variants has been demonstrated recently (27–29).

Redox sensitivity in biological molecules is often mediated by thiol-containing cysteines, which can interact with vicinal thiols forming inter-or intramolecular disulfide bonds. For example, cysteine residues functionally involved in intracellular redox states via glutathione and thioredoxin pathways play a central role in mediating cellular responses to redox changes (9, 10). Formation and dissociation of disulfide bonds can significantly alter protein structure. Redox labile cysteine residues can form molecular switches by conferring functionality in one redox state (e.g., in the on-position) and attenuating the function in the opposite state (e.g., in the off-position).

In an attempt to build FRET-based redox biosensors, we present our first steps toward developing chimeric peptides serving as redox-sensitive linkers between donor and acceptor molecules. The linkers between the donor and acceptor molecules consist of α-helical structures in conjunction with different types of redox switches. The first switch represents the hexapeptide -WCGPCK- redox motif found in the thioredoxin family of oxidoreductases (30). The second and third switches consist of adjacent and dispersed cysteines embedded in an α-helical structure, respectively. The separation between donor and acceptor is varied by inserting helix-forming peptides of the general sequence A(EAAAK)_NA (N=2–8) (27, 28, 31).

In this report, we explored the use of redox-sensitive linkers as a tool for engineering FRET-based genetically encoded biosensors. The common FRET-compatible pair of GFP variants, enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP), was attached to various forms of designed redox linkers. The utility of the linkers was validated: (i) in vitro—through the constructs’ response to reduction with various concentrations of dithiothreitol (DTT) and ratios of oxidized and reduced forms of glutathione (GSSG/GSH); and (ii) in mammalian cell culture through oxidation and reduction with diamide and DTT. Finally, a convenient and robust method of measuring FRET efficiency is proposed for fusion fluorescent proteins. The high accuracy of the (ratio) _A method of analysis allowed simple and straightforward comparisons of linker designs.

Materials and Methods

CY-RL4, CY-RL5 and CY-RL6 constructs were subcloned from pECFP-C1 recombinants into NcoI/BamHI sites of pET19b (Novagen, San Diego, CA). LB medium used for culturing cells was supplemented with 50 μg/ml ampicillin. E. coli BL21 (DE3) chemically competent cells (Novagen) were transformed with the resulting recombinant plasmids and plated on LB agar plates. Single colonies were used to inoculate 5 ml of LB medium and grown overnight at 37 °C. Overnight cultures were diluted 1:200 and grown in 200 ml LB medium to an optical density (600 nm) of 0.5. After protein expression induction with 1 mM IPTG, the cells were grown for 3 hrs at 37 °C and harvested by centrifugation. The cells were suspended in 5 ml of lysis buffer consisting of 50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, pH 8.0, and lysed by sonication followed by centrifugation. Proteins were batch purified on Ni-NTA resins (Qiagen) following manufacturer’s protocol. Samples were desalted by repeated filtration through centrifugal concentrators Centricon-30 (Millipore, Billerica, MA) and diluted with buffer comprised of 50 mM NaH₂PO₄, 100 mM NaCl, pH 8.0. Protein content was determined by the BSA method (33).

Steady-state Fluorescence Spectroscopy.

All fluorescence measurements were carried out at 20 °C on a modified photon-counting fluorometer, ISS-PC (ISS, Champaign, IL). The magic angle condition (a 54.7° offset between the excitation and emission polarizer) was used to avoid polarization artifacts in the emission spectra. A 160 μl aliquot of diluted crude extract or 0.5 μM of purified construct was placed in a cylindrical microcell (3 mm diameter). To account for differences in the level of expression of FRET constructs in crude extracts, a direct correlation between levels of expression of GFP variants and their fluorescence intensity was used. For this purpose, samples were excited at 500 nm and prerecorded from 520 nm to 540 nm followed by dilution with a buffer consisting of 50 mM NaH₂PO₄, 100 mM NaCl, pH 8.0. Protein samples were pretreated with various concentrations of DTT or diamide for 5 min at room temperature. Subsequently, fluorescence emission spectra were acquired from 460 nm to 750 nm during excitation of ECFP at 440 nm and from 520 nm to 750 nm during excitation of EYFP with 500 nm light. The start of the emission wavelength scan was spaced by 20 nm from the excitation wavelength for all samples to avoid scattered light in the emission spectra.

All spectra were corrected for the wavelength dependent response of the emission monochromator and detector. All spectra were baselined by subtracting the mean value of the measured intensity from 730 nm to 750 nm, where the fluorescence intensity is zero. Fluctuations in lamp intensity were corrected automatically with a standard quantum counter. Routines in IgorPro (WaveMetrics, Lake Oswego, OR) software were written for data analysis and calculations using the fluorescence spectra.

(ratio)_A Method for Determination of FRET Efficiencies In Vitro.

Steady-state fluorescence emission spectra were measured for the determination of the FRET efficiency. The (ratio) _A method is a ratiometric method, which provides a sensitive, reproducible determination of FRET efficiencies (34). The (ratio) _A is the ratio of the acceptor fluorescence due to energy transfer (from excited donor (D) fluorophores to acceptor (A)) to the directly excited acceptor fluorescence. The more efficient the energy transfer, the greater the acceptor molecules’ emission intensity, and the larger the value of the (ratio) _A .

Three fluorescence emission spectra, as shown in Figure 3, are needed to measure the extent of energy transfer from donor to acceptor fluorophores by the (ratio) _A method: (1) the emission spectrum of donor and acceptor molecules during selective excitation of the donor molecules, F _DA ; (2) the emission spectrum of the acceptor molecules directly excited at the maximal absorption wavelength of the acceptor fluorophore, F _A ; and (3) the emission spectrum of the donor fluorophore alone, in the absence of acceptor fluorophores, under the same solvent conditions, F _D . The (ratio) _A is calculated from these spectra as follows (35):

$\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \[(\mathit{ratio})_{\mathit{A}}\ =\ \frac{[\mathit{F}_{\mathit{DA}}({\lambda}_{\mathit{exc}}^{\mathit{D}})\ {-}\ \mathit{F}_{\mathit{D}}({\lambda}_{\mathit{exc}}^{\mathit{D}})]}{\mathit{F}_{\mathit{A}}({\lambda}_{\mathit{exc}}^{\mathit{A}})}.\] \end{document}$

The excitation wavelengths λ _exc ^D and λ _exc ^a indicate the wavelengths for selective excitation of the donor or acceptor molecules. Doubly labeled F _DA (λ _exc ^D ) is composed of the fluorescence emission from both donor and acceptor molecules. The spectrum F _DA (λ _exc ^D ) is linearly decomposed with weighted components of previously recorded normalized F̂ _D (λ _exc ^D ) and F̂ _A (λ _exc ^A ) spectra. The weighted contribution of the donor, F _D (λ _exc ^D ), is then subtracted from F _DA (λ _exc ^D ) to give the contribution of F _A (λ _exc ^A )to the F _DA (λ _exc ^D ) spectrum (Fig. 3a). Thus, the numerator of Eq. (1) is the acceptor molecules’ fluorescence emission excited at λ _exc ^D The magnitude of the numerator is determined by the efficiency of energy transfer from donor to acceptor molecules, with a small contribution of the direct excitation of the acceptor at λ _exc ^D (some acceptor fluorescence that results from direct excitation of the acceptor is also present in F _DA (λ _exc ^D )). This contribution of directly excited acceptor (at λ _exc ^D ) is subtracted from (ratio) _A when calculating the FRET efficiency (see equation 2). The F̂ _A (λ _exc ^A )spectrum is measured directly on the D-A sample, so the denominator of Eq. (1) normalizes the FRET signal for sample-to-sample variation in concentration and fluorescence quantum yield of the acceptor molecules.

The FRET efficiency is then calculated according to Eq. (2),

$\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \[\mathit{E}\ =\ \frac{{\varepsilon}_{\mathit{A}}({\lambda}_{\mathit{exc}}^{\mathit{A}})}{{\varepsilon}_{\mathit{D}}({\lambda}_{\mathit{exc}}^{\mathit{D}})}\ \left[(\mathit{ratio})_{\mathit{A}}\ {-}\ \frac{{\varepsilon}_{\mathit{A}}({\lambda}_{\mathit{exc}}^{\mathit{A}})}{{\varepsilon}_{\mathit{A}}({\lambda}_{\mathit{exc}}^{\mathit{A}})}\right],\] \end{document}$

where ε _D and ε _A are the extinction coefficients of the donor and acceptor at the relevant excitation wavelengths. Note that the extinction coefficients of the donor and acceptor molecules are needed for the relevant excitation wavelengths to calculate the FRET efficiency. The extinction coefficient ratio ε _A (λ _exc ^D )/ε _A (λ _exc ^A ) of Eq. (2) removes the fluorescence component of F _DA (λ _exc ^D ) resulting from direct excitation of the acceptor molecules at the donor excitation wavelength. A second extinction coefficient ratio, ε _A (λ _exc ^A )/ε _D (λ _exc ^D ), accounts for the different efficiencies of light absorption of the donor and acceptor molecules.

The (ratio) _A values were calculated by measuring two fluorescence emission spectra per sample, as shown in Figure 3: the emission spectrum of ECFP and EYFP during excitation at 440 nm, where the absorbance of ECFP is dominant; and the emission spectrum of EYFP by excitation at 500 nm, where only EYFP absorbs. A third spectrum, the emission spectrum of pure ECFP (expressed without EYFP) excited at 440 nm, was recorded once for calculation of the (ratio) _A values for all samples. FRET efficiencies, E, were calculated from the (ratio) _A values as:

where the extinction coefficient ratios ε _A (500 nm)/ε _D (440 nm) = 1.8 and ε _A (440 nm)/ε _A (500 nm) = 0.076 have been substituted into Eq. (3). The extinction coefficients ratios were obtained using the extinction coefficients at the absorbance maxima reported in the literature: ε _D (433 nm) = 26,000 cm⁻¹M⁻¹ and ε _A (513 nm) = 84,000 cm⁻¹M⁻¹ (36, 37). Extinction coefficient spectra were calculated by multiplication of normalized fluorescence excitation spectra (provided by Clontech, Palo Alto, CA) for ECFP and EYFP by ε _D (433 nm) and ε _A (513 nm), respectively: ε _D (440 nm) = 25,000 cm⁻¹M⁻¹, ε _A (440 nm) = 3,300 cm⁻¹M⁻¹, and ε _A (500 nm) = 44,000 cm⁻¹M⁻¹.

Caution should be taken in interpreting the (ratio) _A if the absorption of the chromophore is sensitive to conditions such as ionic strength, pH, temperature, conformation changes, etc. For such cases, the extinction coefficients should be measured under all relevant experimental conditions. For our in vitro measurements (pH 8, constant temperature, and constant salt concentration), the absorption properties are expected to be stable.

GSH/GSSG Treatment.

Oxidized protein (≈0.5 μM) was incubated in buffer consisting of 50 mM NaH₂PO₄, 100 mM NaCl, pH 8.0, to which varying concentrations of oxidized (GSSG) and reduced (GSH) glutathione were added. Total concentration of glutathione was kept 10 mM. Treatment was performed for 10 min at 20 °C and spectra were collected as described above. Reduction potentials in millivolts for varying ratios of [GSH]²/[GSSG] were calculated from the Nernst equation (38).

Intracellular Analysis by Flow Cytometry.

CHO cells were cultured in 75 cm² tissue culture flasks (Corning) to confluence and then trypsinized, centrifuged, and resuspended in 1X Hank’s buffered saline solution (HBSS) to 1.0 × 10⁶ cells/ml. Cells were treated with either 2 mM diamide or 2 mM DTT for 5 min at room temperature before data collection. To test for linker reversibility, the cells oxidized with 2 mM diamide were reduced with 2 mM DTT. FRET measurements were taken after each treatment.

All flow cytometric data were collected using a BD LSR II (Becton Dickson, San Jose, CA) instrument. In each experiment at least 10,000 cells were analyzed. The solid state Coherent Sapphire blue 488 nm laser line at 20 mW and the Coherent VioFlame PLUS violet 405 nm laser line at 25 mW were used to excite EYFP and ECFP, respectively. EYFP signals were collected using a 530/30 bandpass filter in the primary laser pathway. The ECFP and FRET signals were collected using 450/50 and 525/50 bandpass filters, respectively, along with a 505 long-pass dichroic splitter filter inserted into the 405-nm violet laser pathway. All data were analyzed using FACS Express Version 3 (De Novo Software, Thornhill, Ontario, Canada).

Results

The efficacy of adjacent cysteine residues to function as a redox switch in a FRET-based sensor was studied in linker RL4 (Table 1). Two pairs of adjacent Cys were introduced into an α-helical linker comprised of five EAAAK repeats separating ECFP and EYFP in the construct CY-RL4. Alternatively, two linkers RL5 and RL6 were designed with four cysteines dispersed throughout the α-helical structure used for RL4. Additionally, a helix-stabilizing N-cap Ser-Pro-Glu was introduced to the N-terminal end of linker RL5 (Table 1) (40).

The Molecular Operating Environment (MOE) software was used to choose the spatial locations of the cysteines and to visualize the different connectivity patterns of disulfide bonds in order to estimate their effect on the helix-coil transition (41). The disulfide bonding patterns that were implemented into linkers RL5 and RL6, and which led to the greatest transition, were observed with bonds between cysteines 1–4 and 2–3, and to a lesser degree 1–3 and 2–4, respectively. The precise placements of the cysteines in the linker to obtain the desired bonding pattern were further determined with the aid of disulfide connectivity prediction tools (42, 43).

In contrast to crude cell lysates of CHO cells harboring expressed CY-RL1, CY-RL2, and CY-RL3 constructs, the linkers comprised of adjacent and dispersed cysteines were tested on purified FRET constructs expressed in E. coli. The typical profiles of emission spectra for these constructs and their sensitized emissions are shown in Figures 5a and 5b, respectively. The response of CY-RL4, CY-RL5, and CY-RL6 constructs to DTT reduction followed similar patterns, with the largest change for CY-RL5 (Fig. 5c). Meanwhile, oxidation of constructs with diamide did not significantly affect FRET as mentioned above for constructs in crude cell lysates (data not shown).

The response of the linkers to changes in the redox environment from treatment by GSH/GSSG was demonstrated for linkers RL5 and RL6 (Table 3). However, the DTT treatment elicited a greater response than did the GSH/GSSG pretreatment.