Direct comparison of fluorescence- and bioluminescence-based resonance energy transfer methods for real-time monitoring of thrombin-catalysed proteolytic cleavage

Biosensors and Bioelectronics 24 (2009) 1164–1170

Contents lists available at ScienceDirect

Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios

Direct comparison of fluorescence- and bioluminescence-based resonance energy transfer methods for real-time monitoring of thrombin-catalysed proteolytic cleavage H. Dacres ∗ , M.M. Dumancic, I. Horne, S.C. Trowell Food Futures Flagship, CSIRO Entomology, GPO Box 1700, Canberra, ACT 2601, Australia

a r t i c l e

i n f o

Article history: Received 11 April 2008 Received in revised form 30 June 2008 Accepted 1 July 2008 Available online 22 July 2008 Keywords: FRET BRET Thrombin Renilla luciferase Fluorescent proteins

a b s t r a c t In this study, a representative FRET system (CFP donor and YFP acceptor) is compared with the BRET2 system (Renilla luciferase donor, green fluorescent protein2 (GFP2 ) acceptor and coelenterazine 400a substrate). Cleavage of a thrombin-protease-sensitive peptide sequence inserted between the donor and acceptor proteins was detected by the RET signal. Complete cleavage by thrombin changed the BRET2 signal by a factor of 28.9 ± 0.2 (R.S.D. (relative standard deviation), n = 3) and the FRET signal by a factor of 3.2 ± 0.1 (R.S.D., n = 3). The BRET2 technique was 50 times more sensitive than the FRET technique for monitoring thrombin concentrations. Detection limits (blank signal + 3 b , where  b = the standard deviation (S.D.) of the blank signal) were calculated to be 3.05 and 0.22 nM thrombin for FRET and BRET2 , respectively. This direct comparison suggests that the BRET2 technique is more suitable than FRET for use in proximity assays such as protease cleavage assays or protein–protein interaction assays. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Resonance energy transfer (RET) offers the convenience of real-time ratiometric quantification of inter- and intra-molecular protein interactions (Li et al., 2006). RET is the non-radiative transfer of energy from an excited state donor to a ground state acceptor. The signal strength and efficiency of RET depend upon the extent of the overlap between the donor emission and acceptor absorption spectra, the quantum yield of the donor and the molar absorptivity of the acceptor as well as the distance between, and relative orientations of the donor and acceptor dipoles (Forster, 1959). In the last decade there have been an increasing number of publications describing the use of RET-based probes in proximity assays for monitoring protein–protein interactions. RET-based assays have been described for monitoring GPCR (G-protein-coupled receptors) oligomerization (Overton and Blumer, 2002; Kroeger et al., 2002), GPCR interactions with G-proteins (Gales et al., 2005) or with ␤-arrestin (Coulon et al., 2008; Molinari et al., 2008). RET-based proximity assays also include complementation assays (Eglen, 2007) and protease assays (Mitra et al., 1996). RET-based assays may monitor either the increase of a signal due to enhancement of RET or the reduction in the RET signal due to dissociation of a

∗ Corresponding author. E-mail address: [email protected] (H. Dacres). 0956-5663/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2008.07.021

complex or a proteolytic cleavage. A variety of RET techniques have been employed in proximity assays with the main ones being FRET and BRET techniques. Although first characterised in 1959 (Forster, 1959) it is only in the last decade that RET techniques have been widely adopted as a research tool in the biosciences. The uptake of RET techniques has been facilitated by the development of a range of RET donors and acceptors, the increasing ease with which they can be genetically encoded and the increased sophistication and availability of instruments that can perform RET measurements. FRET is the most commonly used RET system and both donors and acceptors are fluorescent. The most widely used is CFP as the donor and YFP as the acceptor (Piston and Kremers, 2007). BRET, which replaces the donor fluorophore of FRET with a luciferase, requires the addition of a substrate to initiate bioluminescent emission and hence, energy transfer. Acceptor emission must be due to energy transfer rather than excitation at the acceptor excitation wavelength. Two common implementations of BRET comprise Renilla luciferase (RLuc) with either coelenterazine h (BRET1 ) (em = ∼475 nm) or coelenterazine 400a (Clz400a) substrate (BRET2 ) (em = 395 nm) as the donor system coupled to either of the GFP mutants, YFP (BRET1 ) (em = ∼530 nm) or green fluorescent protein2 (GFP2 ) (BRET2 ) (em = ∼510 nm). The BRET2 system offers superior spectral separation between the donor and acceptor emission peaks of ∼115 nm compared to ∼55 nm for the BRET1 system (Pfleger and Eidne, 2006). However, although arguments have been made to justify the

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

choice between BRET or FRET (Boute et al., 2002), there has been no direct experimental comparison of the two techniques in terms of their sensitivity, kinetics or other properties to provide a basis to choose between them in any particular situation. In this study a commonly used FRET (CFP-YFP) and a BRET2 (GFP2 -RLuc) system were directly compared by placing a thrombinspecific cleavage sequence (Chang, 1985) between the donor and acceptor proteins allowing the thrombin cleavage of this substrate to be detected via the RET signal (Supplementary data, Fig. 1(a)). Thrombin, an extracellular protein, is a well-understood protease in the regulation of blood coagulation (Coughlin, 1999) that selectively cleaves the Arg-Gly bonds of fibrinogen to form fibrin. The results from this study will enable the applicability of FRET and BRET2 techniques for proximity assays to be assessed in quantitative terms. 2. Materials and methods 2.1. Materials All primers were purchased from Geneworks (Supplementary data, Table 1). The vectors pRSET-CFP (Invitrogen) and pcDNA6.2/NYFP-DEST (Invitrogen) were used as a source of CFP and YFP for FRET analysis (Supplementary data, Table 2). The vector GFP2 -MCS-RLuc (PerkinElmer) was used as the source for both GFP2 and RLuc for BRET2 analysis (Supplementary data, Table 2).

1165

buffer (50 mM sodium phosphate buffer, 300 mM NaCl, pH 7.0). The cells were lysed by French press (−18,000 psi) and the soluble protein fractions were isolated by centrifugation at 9300 × g (4 ◦ C) for 15 min. Proteins were purified using cobalt affinity chromatography according to the supplied instructions (BD Talon (BD Biosciences, Clontech)). Following elution of the purified protein with 150 mM imidazole, the buffer was exchanged for cleavage buffer (50 mM Tris (pH 8.0), 100 mM NaCl, and 1 mM EDTA) by dialysis. Aliquots of 200 ␮L protein were snap frozen on dry ice and stored at −80 ◦ C. Protein concentrations were determined by absorbance at 280 nm and calculated according to the method of Gill and Von Hippel (1989). The calculated protein concentration was confirmed using the Bradford Assay (BioRad) using BSA as the standard. 2.4. Instrumentation 2.4.1. Spectral measurements All spectral scans were recorded with a plate-reading SpectraMax M2 spectrofluorometer (Molecular Devices). The reactions were carried out in 96-well plates (PerkinElmer). Fluorescence spectral scans were recorded from 450 to 600 nm with an excitation wavelength of 430 nm using a 455-nm emission cutoff filter. Spectral scans of BRET2 constructs were recorded using the luminescence scan mode scanning between 380 and 600 nm.

2.2. Construction of RET proteins GFP2 , RLuc and YFP (Constructs 1–3, Supplementary data, Table 2) were amplified by polymerase chain reaction (PCR) using the primers shown in Supplementary data (Table 1) and cloned into pGEM®-T Easy vector (Promega). This resulted in the respective insertion of BamHI and BsrGI restriction sites directly upstream and downstream from the amplified gene. The amplicons were inserted into the BamHI and BsrGI sites of the pRSET vector, replacing CFP, to give pRSET-GFP2 , pRSET-RLuc and pRSET-YFP. DNA sequencing confirmed correct amplicon sequence. GFP2 , RLuc, CFP and YFP were amplified by PCR using primers (Supplementary data, Table 1) for construction of fusion pairs 5–8 (Supplementary data, Table 2). EcoRI and PstI or PstI and BsrGI restriction sites were introduced upstream and downstream, respectively, from the four different amplified genes (Supplementary data, Fig. 1(b)). The fusion pairs 5–8 (Supplementary data, Table 2) were then inserted into the EcoRI and BsrGI sites of the pRSET vector DNA and confirmed by sequencing. The thrombin recognition sequence (LQGSLVPRGSLQ) (Zhang, 2004) was synthesized by annealing oligomers C1 and C2 (Supplementary data, Table 1) followed by digestion with PstI. The sequence GSLVPRGS was then inserted between the fusion pairs (5–8, Supplementary data, Table 2) by the PstI site situated between the donor-acceptor constituents of the fusion pair. The fusion protein contains a 6× His-tag at its N-terminus. All constructs were confirmed by sequencing. 2.3. Expression and purification of RET-based proteins Proteins were expressed in E. coli strain BL21 DE3 (Novagen). An overnight culture was grown from a single colony in LB containing 100 ␮g/mL ampicillin and 2% glucose at 37 ◦ C, 200 rpm. Expression was induced by inoculating 500 mL LB containing 100 ␮g/mL ampicillin to an A600 of 0.1 and induced at 37 ◦ C (200 rpm) for 3.5 h followed by overnight incubation at 22 ◦ C (200 rpm). Cells were harvested 24 h after inoculation. For protein purification, cells were harvested by centrifugation at 4335 × g (4 ◦ C) for 15 min and resuspended in equilibration

2.4.2. Simultaneous dual emission detection Simultaneous dual emission RET measurements were carried out with a POLARstar OPTIMA microplate reader (BMG LabTech). Simultaneous emission measurements used either the BRET2 emission filter set comprising of RLuc/Clz400a emission filter (410 nm band-pass 80 nm) and the GFP2 emission filter (515 nm band-pass 30 nm) or the FRET filter set consisting of a CFP excitation filter (450 nm band-pass 10 nm) and the respective CFP (500 nm band-pass 10 nm) and YFP (530 nm band-pass 10 nm) emission filters. 2.5. Thrombin assays Purified fusion protein (1 ␮M) was incubated with varying amounts of thrombin protease (Amersham Biosciences) in cleavage buffer (final volume of 100 ␮L) at 30 ◦ C for up to 90 min. RET analysis was carried out in 96-well plates with incubation of specified fusion protein aliquots with thrombin for 90 min at 30 ◦ C followed by recording the RET signal. For BRET2 measurements Clz400a substrate (5 ␮M) was added following the 90-min period and a 0.50-s integration time used. Real-time analysis was also performed using the POLARstar OPTIMA plate reader and a whole plate assay was employed for FRET analysis or a well-by-well assay for BRET2 analysis (20 s integration time). 2.5.1. Kinetic analysis Km (Michaelis constant) and Vmax (maximum velocity) were obtained by non-linear regression analysis to the Michaelis– Menten equation using GraphPad Prism for the Mac version 5.0a. 2.6. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) analysis Proteins (2.5 ␮g) were diluted in 1× sample loading buffer (Invitrogen) for SDS–gel electrophoresis (NuPAGE system: 12% Bis–Tris gel with MOPS running buffer (Invitrogen)). Bands were visualised following staining with Fast stain TM (Fisher).

1166

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

2.7. Hirudin inhibition of thrombin protease Recombinant hirudin from yeast (Sigma) was incubated with thrombin at room temperature for 10 min prior to implementation of thrombin assays. 2.8. RET ratio determinations FRET ratio was calculated as emission ratio measured at 510 nm/530 nm (Zhang, 2004). BRET2 ratio was calculated as emission ratio measured at 515 nm/410 nm (Pfleger and Eidne, 2006). 3. Results and discussion 3.1. Orientation of the FRET and BRET pairs Two proteins were generated for each RET pair, in which the donor molecule (RLuc or CFP) was inserted either at the N- or the C-terminus of the fluorescent acceptor (GFP2 or YFP). This resulted in four fusion proteins referred to as RLuc-RG-GFP2 , GFP2 -RGRLuc, CFP-RG-YFP and YFP-RG-CFP, where RG refers to the Arg-Gly cleavage site in the LQGSLVPRGS linker sequence. The measured FRET ratio of CFP-RG-YFP was 0.27 ± 0.01 (S.D., n = 3) compared to 1.18 ± 0.01 (S.D., n = 3) for YFP-RG-CFP. Previous studies using these FRET probes have all placed CFP at the N-terminus of the fusion protein most likely due to the higher FRET ratio in this orientation (Pollock and Heim, 1999; Mahajan et al., 1999; Tyas et al., 2000; Zhang, 2004; Felber et al., 2004; Vinkerborg et al., 2007). The measured BRET2 ratios for RLuc-RG-GFP2 and GFP2 -RGRLuc were 1.00 ± 0.02 (S.D., n = 3) and 6.21 ± 0.18 (S.D., n = 3), respectively. This effect was observed previously using BRET probes with enhancements of 10- and 3-fold when the interaction between the proteins allowed a free N-terminus in GFP2 (Ohmiya, 2005; Molinari et al., 2008). These results confirm that the CFP-RG-YFP and GFP2 -RG-RLuc fusion construct signal enhancements of 4.39 and 6.19, respectively, compared to the constructs YFP-RG-CFP and RLuc-RG-GFP2 signals are consistent with the efficiency of energy transfer being sensitive to orientation and distance. All further studies were carried out using CFP-RG-YFP and GFP2 -RG-RLuc fusion proteins.

the BRET2 fusion protein in crude extracts from COS-7 cells produced a >150-fold change in the ratio, five times greater than that observed in this study, when monitored using integrated spectral scans (Molinari et al., 2008). Molinari et al. assumed protein concentrations to be in the nanomolar range and as the proteins were not purified there was little evidence that the spectral change was due to only thrombin cleavage of the BRET2 protein. It is possible that other crude extract components may have interfered with the assay. The use of a dual emission detection plate reader with bandpass filters is more convenient than integrating spectral scans for monitoring events in real time. 3.2.2. Specificity The recognition site for thrombin is the Arg-Gly linkage in the peptide sequence LQGSLVPRGS and for Caspase-3 is DEVD (Mahajan et al., 1999). Caspase-3 was therefore used to confirm the specificity of the thrombin RET assay reaction. Replacing thrombin with 54 nM Caspase-3 resulted in no measurable change in FRET or BRET2 ratio (Fig. 1). Pre-incubation of thrombin with two anti-thrombin units of hirudin, which is known to stably inhibit thrombin by binding to its catalytic and its anion binding exosite I (Zhang, 2004), also abolished any changes in RET ratio. Specificity was further confirmed by SDS–PAGE following incubation of RET constructs with thrombin (Fig. 2). Thrombin treatment of the FRET fusion resulted in qualitative conversion of the single 59.7 kDa (CFP-RG-YFP) polypeptide into components of 32.4 and 27.3 kDa corresponding to CFP with a His-tag and YFP without. Similarly, thrombin treatment of the 68.8 kDa BRET2 fusion protein converted it into components of 36.4 and 32.4 kDa corresponding to untagged RLuc and His-tagged GFP2 . The thrombin cleaved His-tagged GFP2 and CFP are 0.9 kDa higher than the native His-tagged GFP2 and CFP (31.5 kDa) because thrombin cleavage leaves the sequence LQGSLVPR attached to the C-terminus of these His-tagged proteins. In all cases, the migration of bands relative to the pre-stained protein ladder implied molecular weights 4 kDa higher than their true (calculated) values, a discrepancy we attribute to idiosyncrasies of the gel buffer system. For both RET fusion proteins, pre-incubation with hirudin prevented the conversion of the RET fusion protein into its two components (Fig. 2) confirming that the cleavage observed in the absence of hirudin was due to the specific interaction of thrombin with the fusion proteins.

3.2. Effect of full cleavage of the Arg-Gly linkage 3.3. Spectral analysis of RET probes 3.2.1. RET assays When the CFP-RG-YFP and GFP2 -RG-RLuc fusions were completely cleaved using 54 nM thrombin for 90 min the RET ratios changed by factors of 3.2 ± 0.1 (R.S.D., n = 3) and 28.8 ± 0.2 (R.S.D., n = 3), respectively (Fig. 1). Notably these final RET ratios were not significantly different (P < 0.05) from those obtained by mixing 1 ␮M each of the donor and acceptor pairs for both RET systems suggesting full cleavage occurs under these conditions (Fig. 1, control). Therefore, under identical assay conditions the change in recorded RET ratios was nine times higher for BRET2 than for FRET. A major disadvantage associated with FRET is due to the need for a light source to energise the donor fluorophore (Piston and Kremers, 2007). This causes a problem referred to as “cross-talk” due to the direct excitation of the acceptor at the donor excitation wavelength. Differences between the FRET and BRET2 ratios recorded here, following thrombin cleavage, demonstrate the practical implications of this problem for FRET measurements. Previous studies monitoring complete cleavage of FRET substrates with the proteases thrombin (Zhang, 2004) and trypsin (Pollock and Heim, 1999) demonstrated 4.6- and 4.2-fold changes in the FRET ratio, respectively. Complete thrombin hydrolysis of

The full emission spectra of the purified RET proteins, before and after treatment with 54 nM thrombin confirmed results obtained using band-pass filters (Fig. 3). An enhanced fluorescence emission was recorded at  = 530 nm upon excitation at 430 nm prior to thrombin treatment. The FRET ratio changed by a factor of 4.2 (530 nm/480 nm) following thrombin treatment. This is consistent with previously reported 4.6- and 4.2-fold changes in the FRET ratio calculated using single wavelength measurements (Zhang, 2004; Pollock and Heim, 1999). The changes in FRET ratios calculated from single wavelength measurements are slightly higher than those obtained here using band-pass filters (excitation filter: 450 nm band-pass 10 nm and emission filters: 500 nm band-pass 10 nm and 530 nm band-pass 10 nm) due to subtle differences in the excitation and emission wavelengths. For spectral scans, CFP emission intensity is measured at 480 nm with an excitation wavelength of 430 and a 455-nm cut-off filter. This eliminates interference from Rayleigh scattering (Cox et al., 2001), which has the same wavelength as the incident light and would tend to mask the CFP emission peak. When using band-pass filters to overcome this problem, an excitation wavelength of 450 nm

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

1167

Fig. 1. Change in RET ratio following thrombin cleavage (normalized FRET or BRET2 ratio (mean ± R.S.D., n = 3)) of 1 ␮M of fusion proteins: (a) CFP-RG-YFP and (b) GFP2 RG-RLuc following treatment (90 min, 30 ◦ C) with 54 nM thrombin, 54 nM Caspase-3 or 2 units thrombin (54 nM) following pre-treatment (10 min, room temperature) with 2 units hirudin. Controls consist of 1 ␮M of donor and acceptor molecules.

is used with a CFP emission 10 nm band-pass filter centred at 500 nm, red-shifting the measurement wavelength to a suboptimal level. The bioluminescent spectrum before thrombin treatment was bimodal with a peak at 420 nm representing RLuc emission and a second peak at 510 nm representing GFP2 emission (Fig. 3). This indicates energy transfer from the excited state of Clz400a to GFP2 . Upon cleavage with thrombin the green component of the spectrum disappeared. The BRET2 ratio changed by a factor of 26.2 (510 nm/420 nm) which is slightly lower than the ratio change obtained with the microplate reader due to the Renilla emission filter (410 nm band-pass 80 nm) consisting of a magnifying lens to increase the intensity of the bioluminescent signal (Fig. 1). There was also a marked decrease (2.9-fold) in the light output at 420 nm upon thrombin cleavage although this decrease is not observed after normalization (Fig. 3). The decrease in quantum yield of the Clz400a substrate upon thrombin cleavage of GFP2 -RLuc has been reported previously (Molinari et al., 2008) as has the enhancing effect of Renilla GFP (RGFP) (∼3-fold) when fused to its own luciferase (Waud et al., 2001). Molinari et al. (2008) suggest that this

quantum yield enhancement could be due to the GFP2 fusion changing the catalytic properties of RLuc by inducing a conformational change. 3.4. Kinetics analysis of RET probes Optimization of the FRET assay sensitivity requires the substrate concentration to be optimized. Increasing CFP-RG-YFP concentration from 0.5 to 8 ␮M changed the FRET ratio by a factor of 1.6 (Supplementary data, Fig. 2) consistent with the observations of Felber et al. (2004). This represents an enhancement in the efficiency of energy transfer as the concentration of substrate increases. Although monomeric CFP and YFP do not dimerize until their concentrations reach 100 ␮M (Felber et al., 2004; Mitra et al., 1996) head to tail dimerization of CFP-RG-YFP may be energetically more favourable. A 1-␮M concentration was chosen for the fusion protein because it gave a relatively high FRET ratio whilst allowing for direct comparisons with the data of Zhang (2004). There are few, if any, precedents for using FRET-based measurements to determine enzyme kinetics. Increasing CFP-RG-YFP

Fig. 2. SDS–PAGE analysis of purified His-tagged RET proteins. 2.5 ␮g protein loaded per lane. From left to right: (a) molecular markers (kDa), CFP, YFP, CFP-RG-YFP, CFP-RG-YFP following incubation with 54 nM for 90 min at 30 ◦ C; CFP-RG-YFP same conditions as previous lane except that thrombin was pre-incubated with 2 units hirudin for 10 min at room temp; (b) molecular markers (kDa), RLuc, GFP2 , GFP2 -RG-RLuc, GFP2 -RG-RLuc following incubation with 54 nM for 90 min at 30 ◦ C; GFP2 -RG-RLuc, same conditions as previous lane except that thrombin was pre-incubated with 2 units hirudin for 10 min at room temperature.

1168

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

Fig. 4. Initial velocity of CFP-RG-YFP cleavage as a function of its concentration (mean ± R.S.D., n = 3) in the presence of 27 nM thrombin.

Fig. 3. Normalized emission spectra of 1 ␮M of specified RET constructs before and after treatment with thrombin (54 nM, 90 min, 30 ◦ C): (a) CFP-RG-YFP (ex = 430 nm, normalized by the intensity at 480 nm) and (b) GFP2 -RG-RLuc (5 ␮M Clz400a substrate, normalized by the intensity at 420 nm); N = normalization wavelength.

concentration caused a decrease in the magnitude of the response to 27 nM thrombin (Supplementary data, Fig. 3). At higher substrate concentrations the relative size of the change in the FRET signal is reduced relative to the elevated starting signal. The “inner filter” effect may also contribute to this observation (Palmier and Van Doren, 2007). At higher fluorophore concentrations some of the emitted fluorescence can be re-absorbed by the fluorophore itself resulting in deviation from the Beer–Lambert law producing a nonlinear response (Parker and Rees, 1962). The fluorescence intensity deviated from linearity above 2 ␮M substrate concentration providing evidence that the inner filter effect may influence fluorescence intensity at higher concentrations (Supplementary data, Fig. 4). An alternative approach to monitoring enzyme kinetics is to monitor changes in the donor or acceptor fluorescence intensity (excitation 430 nm/emission 530 nm). Using this approach, the rate of change in the YFP signal in response to 27 nM thrombin increased with increasing CFP-RG-YFP concentration (Fig. 4). The maximum velocity (Vmax ) was determined to be 4.6 ± 0.8 pmoles CFP-RG-YFP cleaved per minute (mean ± S.D., n = 3) and the Michaelis constant (Km ) was 2.0 ± 0.6 ␮M (mean ± S.D., n = 3). Medintz et al. (2006) determined the Vmax to be 0.2 ± 0.01 ␮M min−1 and Km to be 0.46 ± 0.12 ␮M using a Quantum dot (QD) donor and quencher dye QXL-520 as acceptor with a thrombin cleavage site between the two. Medintz et al. (2006) measured the QD emission intensity, a similar approach to the one taken here. The number of substrate molecules cleaved by each thrombin site was calculated to be 2.11 min−1 (kcat ). To enable direct comparison with the calculated kinetic rate constant for the hydrolysis of the Arg-Gly peptide bond in fibrinogen we converted kcat into the same units as those used by Marsh et al. (1983). The kcat value of 767 pM (NIH unit s)−1 calculated for the FRET substrate was very similar to the value

of 730 pM (NIH unit s)−1 determined for the intact fibrinogen A␣ chain. This is consistent with the observation (Sheraga, 2004) that as the length of the peptide containing the thrombin cleavage sequence is increased the rate of substrate hydrolysis approaches that of intact fibrinogen. It was not possible using direct approaches to determine the kinetic parameters for the BRET-monitored thrombin assay. In this case, increasing the concentration of the GFP2 -RG-RLuc substrate changes the ratio of luciferase to its Clz400a substrate. At constant Clz440a concentration, the time-dependent decay of the bioluminescence signal was accelerated by increasing the GFP2 -RG-RLuc concentration making the thrombin-catalysed cleavage difficult to ascertain. Clearly the simplicity of the FRET system makes it more suitable than BRET for determining enzyme kinetic parameters that require variation of the substrate concentration. Using a constant substrate concentration of 1 ␮M, thrombin concentrations were varied to enable comparison of reaction rates between the FRET and BRET2 (Supplementary data, Fig. 5). At the highest thrombin concentrations (108 nM) the FRET signal reached saturation due to all of the substrate being cleaved by thrombin after approximately 3600 s. A similar trend was observed with the BRET2 kinetic response, with substrate cleavage appearing to be complete after 500 s for higher thrombin concentrations (54 and 108 nM). Once the maximum BRET2 signal had been reached a decrease in the BRET2 signal followed. This was due to the short biochemical half-life of Clz400a (∼300 s) with RLuc and GFP2 signals decreasing to such low levels that they could not be detected above the background signal. It may, in future, be possible to prolong the BRET2 signal using Clz400a derivatives with increased

Fig. 5. Initial rate of change of normalized RET ratio vs thrombin concentration for BRET2 (1) and FRET (2) measurements (mean ± R.S.D., n = 3).

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

1169

Fig. 6. (a) Calibration of thrombin concentration with 1 ␮M CFP-RG-YFP (FRET measurements) and GFP2 -RG-RLuc (BRET2 measurements) and the linear calibration plots for (b) FRET and (c) BRET2 measurements (mean ± R.S.D., n = 3).

half-lives (>1 h), which carry ether protecting groups at the carbonyl of the imidazopyrazinone ring (Levi et al., 2007), once they become commercially available. Comparisons of initial rate changes in the normalized RET ratio (RET s−1 ) suggest that the BRET2 fusion protein is cleaved 10 times faster than the FRET fusion protein with calculated rate constants of 1 × 10−4 BRET2 units s−1 nM−1 thrombin and 1 × 10−5 FRET units s−1 nM−1 thrombin, respectively (Fig. 5). This is the first direct comparison of FRET and BRET2 reaction rates. The difference in reaction rates was not predicted as the enzyme reaction is the same; however, it may be due to improved accessibility of the thrombin cleavage site when RLuc replaces a fluorescent protein as a donor molecule. RLuc and RGFP are known to spontaneously associate in nature (Ward and Cormier, 1979). Possibly the association drives a conformational change that enhances the accessibility of the thrombin cleavage site. 3.5. Sensitivity and limits of detection Comparison of the sensitivities of the two assays was achieved by incubating 1 ␮M of the FRET or BRET2 probes with increasing concentrations of thrombin (Fig. 6). This is the first report of the use of BRET2 for quantitative determination of thrombin. The RET signal was measured following a fixed period of time (90 min) and divided by the starting signal. The RET ratio changes linearly with increasing concentrations of thrombin up to 21.6 nM for FRET measurements (Fig. 6a and b) and 5.4 nM for BRET2 measurements (Fig. 6a and c). Calibrations were linear in these regions with R2 values exceeding 0.98. Comparison of the gradients of the calibrations revealed that the BRET2 method is ∼50 times more sensitive to changing thrombin concentrations than FRET. A FRETbased immunoassay, which labelled antibody variable regions with blue and green fluorescent proteins could determine 70–1000 nM of antigen hen egg lysozyme (Arai et al., 2000). Replacement of the FRET components with RLuc and YFP resulted in a 10-fold improvement in the sensitivity with a measurable concentration

range of 7–700 nM (Arai et al., 2001). Although Arai and coworkers used a different FRET and BRET pair from those used in this study, the observed differences in sensitivity are of a similar magnitude. The detection limits (blank signal + 3 × standard deviation of the blank (Miller and Miller, 2000)) for thrombin calculated from the curves (Fig. 6b and c) were 3.05 and 0.22 nM for FRET and BRET2 , respectively. This is comparable with a minimum detectable thrombin concentration of 1 nM thrombin reported by Zhang (2004) using FRET measurements. The BRET2 system approaches the level of the most sensitive thrombin assay reported, which is a FRET-based assay. Grant et al. (2004) covalently attached a fluorescein donor and a coumarin acceptor to the thrombin cleavage site using Fmoc (9-fluorenylmethoxycarbonyl) chemistry. The detection limit was 54 pM, i.e. 4-fold more sensitive than the BRET2 assay reported here. Because the BRET2 donor and acceptor can be genetically encoded they may be more broadly applicable, particularly to in situ and in vivo studies than chemically synthesized donors and acceptors. Thrombin concentrations associated with plasma clots have been measured as 0.459 nM using both chromogenic and clotting methods (Meddahi et al., 2004), i.e. within the range detectable by the BRET2 assay demonstrating the latter’s potential for directly measuring physiologically relevant concentrations of thrombin. 4. Conclusions and future directions FRET and BRET2 probes were directly and quantitatively compared in monitoring a thrombin cleavage assay. Upon complete cleavage by thrombin the BRET2 signal changed by a factor of 28.9 ± 0.2 (R.S.D., n = 3) and the FRET signal by a factor of 3.2 ± 0.1 (R.S.D., n = 3). These responses were specific for thrombin hydrolysis of the RET constructs. Comparison of kinetic data demonstrated that the BRET2 construct was cleaved at a rate 10 times faster than the FRET construct. Comparison of calibration graphs showed that the BRET2 calibration was 50 times more sensitive for thrombin quantification compared to FRET. Detection limits were calculated

1170

H. Dacres et al. / Biosensors and Bioelectronics 24 (2009) 1164–1170

to be 3.05 and 0.22 nM of thrombin for FRET and BRET2 measurements, respectively. The BRET2 detection limit is 13.9 times lower for BRET2 because the FRET uses a substrate rather than a light source to initiate energy transfer, therefore, increasing the signalto-noise ratio. Improvements in both RET techniques can be made by using improved variants of donor and acceptor materials. The use of CyPet and YPet (Nguyen and Daugherty, 2005) or mCerulean and mVenus (Piston and Kremers, 2007) in place of standard CFP and YFP fluorophores improves donor quantum yields and acceptor extinction coefficients, and therefore, energy transfer efficiency. The BRET2 assay can be improved by using either alternative donor luciferases or improved bioluminescent substrates (Levi et al., 2007). For instance, the use of a RLuc variant (RLuc8) increased the quantum yield of Clz400a substrate by a factor of 32 compared to native RLuc (De et al., 2007). The combination of higher quantum yield and prolonged light signal offers the possibility of applying BRET2 to real-time monitoring of protein–protein interactions. BRET analysis has recently been applied to real-time single cell imaging of protein–protein interaction (De et al., 2007; Coulon et al., 2008). In conclusion, these results are a direct assessment of the most favoured FRET technique (CFP and YFP pair) compared to the commonly used BRET2 technique. It is demonstrated that BRET2 is more sensitive, has faster reaction kinetics and can detect lower concentrations of analyte under exactly the same reaction conditions. Although other factors, such as equipment availability or the convenience of the assay may affect the choice of method, our findings suggest that BRET2 is more suitable for use in proximity assays. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.bios.2008.07.021. References Arai, R., Ueda, H., Tsumoto, K., Mahoney, W.C., Kumagai, I., Nagamune, T., 2000. Protein Eng. 13 (5), 369–376.

Arai, R., Nakagawa, H., Tsumoto, K., Mahoney, W., Kumagai, I., Ueda, H., Nagamune, T., 2001. Anal. Biochem. 289, 77–81. Boute, N., Jockers, R., Issad, T., 2002. Trends Pharmacol. Sci. 23 (8), 351–354. Chang, J.-Y., 1985. Eur. J. Biochem. 151, 217–224. Coughlin, S.R., 1999. Proc. Natl. Acad. Sci. U.S.A. 96, 11023–11027. Coulon, V., Audet, M., Homberger, V., Bockaert, J., Fagni, L., Bouvier, M., Perroy, J., 2008. Biophys. J. 94, 1001–1009. Cox, A.J., DeWeerd, A.J., Linden, J., 2001. Am. J. Phys. 70 (6), 620–625. De, A., Loening, A.M., Gambhir, S.S., 2007. Cancer Res. 67 (15), 7175–7183. Eglen, R.M., 2007. Biochem. Soc. Trans. 35 (4), 746–748. Felber, L.M., Cloutier, S.M., Kundig, C., Kishi, T., Brossard, V., Jichlinski, P., Leisenger, H.-J., Deperthes, D., 2004. Biotechniques, 878–885. Forster, T., 1959. Discuss. Faraday Soc. 27, 7–17. Gales, C., Rebois, R.V., Hogue, M., Trieu, P., Breit, A., Herbert, T.E., Bouvier, M., 2005. Nat. Methods 2 (3), 177–184. Gill, S.C., Von Hippel, P.H., 1989. Anal. Biochem. 182, 319–326. Grant, S.A., Lichlyter, D.J., Lever, S., Gallazzi, F., Soykan, O., 2004. Sensors Letters 2 (3–4), 164–170. Kroeger, K.M., Hanyaloglu, A.C., Eidne, K.A., 2002. Lett. Pept. Sci. 8, 155–162. Levi, J., De, A., Cheng, Z., Gambhir, S.S., 2007. J. Am. Chem. Soc. 129, 11900–11901. Li, I.T., Pham, E., Truong, K., 2006. Biotechnol. Lett. 28 (24), 1971–1982. Mahajan, N.P., Harrison-Shostak, D.C., Michaux, J., Herman, B., 1999. Chem. Biol. 6, 401–409. Marsh Jr., H.C., Meinwald, Y.C., Thannhauser, T.W., Scheraga, H.A., 1983. Biochemistry 22, 401–4174. Meddahi, S., Bara, L., Fessi, H., Samama, M.M., 2004. Thrombosis Res. 114 (1), 51–56. Medintz, I.L., Clapp, A.R., Brunel, F.M., Tieffenbrunn, T., Tetsuo Uyeda, H., Chang, E.L., Deschamps, J.R., Dawson, P.E., Mattoussi, H., 2006. Nat. Mater., 581–589. Miller, J.N., Miller, J.C., 2000. Statistics and Chemometrics for Analytical Chemistry, 4th edition. Pearson Education Ltd, Harlow, England. Mitra, R.D., Silva, C.M., Youvan, D.C., 1996. Gene 173, 13–17. Molinari, P., Casella, I., Costa, T., 2008. Biochem. J. 409, 251–261. Nguyen, A.W., Daugherty, P.S., 2005. Nat. Biotech. 23 (3), 355–360. Ohmiya, Y., 2005. Jpn. J. Appl. Phys. 44 (9A), 6368–6379. Overton, M.C., Blumer, K.J., 2002. Methods 27, 324–332. Palmier, M.O., Van Doren, S.R., 2007. Anal. Biochem. 371, 43–51. Parker, C.A., Rees, W.T., 1962. Analyst 87, 83–111. Pfleger, K.D.G., Eidne, K.A., 2006. Nat. Methods 3 (3), 165–174. Piston, D.W., Kremers, G.-J., 2007. Trends Biochem. Sci. 32, 407–414. Pollock, B.A., Heim, R., 1999. Trends Cell Biol. 9, 57–60. Sheraga, H.A., 2004. Biophys. Chem. 112, 117–130. Tyas, L., Brophy, V.A., Pope, A., Rivett, A.J., Tavare, J.M., 2000. EMBO Rep. 1 (3), 266–270. Vinkerborg, J.L., Evers, T.H., Reulen, S.W.A., Meijer, E.W., Merkx, M., 2007. ChemBioChem 8, 1119–1121. Ward, W.W., Cormier, M.J., 1979. J. Biol. Chem. 254, 781–788. Waud, J.P., Bermundez Fajardo, A., Sudharan, T., Trimby, A.R., Jeffrey, J., Jones, A., Campbell, A.K., 2001. Biochem. J. 357, 687–697. Zhang, B., 2004. Biochem. Biophys. Res. Commun. 323, 674–678.