Deciphering Design Principles of Förster Resonance Energy Transfer-Based Protease Substrates: Thermolysin-Like Protease from Geobacillus stearothermophilus as a Test Case

Protease activity is frequently assayed using short peptides that are equipped with a Förster resonance energy transfer (FRET) reporter system. Many frequently used donor–acceptor pairs are excited in the ultraviolet range and suffer from low extinction coefficients and quantum yields, limiting their usefulness in applications where a high sensitivity is required. A large number of alternative chromophores are available that are excited in the visible range, for example, based on xanthene or cyanine core structures. These alternatives are not only larger in size but also more hydrophobic. Here, we show that the hydrophobicity of these chromophores not only affects the solubility of the resulting FRET-labeled peptides but also their kinetic parameters in a model enzymatic reaction. In detail, we have compared two series of 4–8 amino acid long peptides, designed to serve as substrates for the thermolysin-like protease (TLP-ste) from Geobacillus stearothermophilus. These peptides were equipped with a carboxyfluorescein donor and either Cy5 or its sulfonated derivative Alexa Fluor 647 as the acceptor. We show that the turnover rate kcat is largely unaffected by the choice of the acceptor fluorophore, whereas the KM value is significantly lower for the Cy5- than for the Alexa Fluor 647-labeled substrates. TLP-ste is a rather nonspecific protease with a large number of hydrophobic amino acids surrounding the catalytic site, so that the fluorophore itself may form additional interactions with the enzyme. This hypothesis is supported by the result that the difference between Cy5- and Alexa Fluor 647-labeled substrates becomes less pronounced with increasing peptide length, that is, when the fluorophore is positioned at a larger distance from the catalytic site. These results suggest that fluorophores may become an integral part of FRET-labeled peptide substrates and that KM and kcat values are generally only valid for a specific combination of the peptide sequence and FRET pair.


Spectra of the fluorophores used
5,6-carboxyfluorescein (CF, Sigma-Aldrich), Cy5-monohydrazide (Cy5, GE Healthcare) and AlexaFluor647-hydrazide (AF647, Thermo Fisher) were used for measuring the absorbance and fluorescence emission spectra of the donor and acceptor fluorophores. The fluorophores were dissolved in DMSO to a stock concentration of 10 µM. For the measurements, they were diluted to a final concentration of 1 µM using MOPS buffer (20 mM MOPS pH 7.4, 5 mM CaCl 2 ), resulting in a final DMSO concentration of 10 %. Absorbance scans were measured in a 1 cm quartz cuvette using a UV/VIS spectrophotometer (V-630, Jasco) and emission scans were recorded in a fluorescence spectrometer (LS55, Perkin Elmer). For the fluorescence measurements, the samples were excited at 450 nm (CF) or 590 nm (Cy5 and AF647), and fluorescence emission was scanned from 470 nm to 750 nm (CF) or from 600 nm to 750 nm (Cy5 and AF647). The normalized spectra obtained from these measurements are shown in Figure S1.

Determination of the Förster radii
The Förster radius R 0 represents the donor-acceptor (D-A) distance where the FRET efficiency is 50 % of the maximal value. R 0 (expressed in Angstroms) can be calculated using the following equation: 1 " = 0.221 ( ) * + , - where k is the orientation factor between the donor and the acceptor, n is the refractive index of the solution, F D is the quantum yield of the donor in the absence of acceptor and J(l) is the spectral overlap integral between the normalized emission spectrum of the donor and the absorbance spectrum of the acceptor expressed in M −1 cm −1 nm 4 . The orientation factor k was set to be equal to 2/3, which is a commonly used assumption. The refractive index of the MOPS buffer is 1.3389, and the quantum yield of CF was measured and equals 0.91. The spectral overlap J(l) was calculated using a|e software from FluorTools (www.fluortools.com), according to the following equation: The R 0 values of the FRET pairs CF/Cy5 and CF/AF647 were calculated to be 50 Å and 46 Å, respectively. These values are comparable with the R 0 values of similar xanthene/cyanine FRET pairs such as Atto488/Atto647 (R 0 = 51 Å) and AlexaFluor488/AlexaFluor647 (R 0 = 56 Å) as stated by the fluorophore manufacturers, ATTO-TEC and Thermo Fisher, respectively.

Solubility of the FRET-peptides
To determine the accessible range of peptide concentrations, the solubility of intact (substrate) and fully cleaved (product) peptides was determined. The fluorescence intensity of all samples was measured at the donor wavelength and plotted against the peptide concentration. Deviations from a linear relationship indicate substrate or product aggregation. A range of substrate concentrations (0 µM -10 µM) was prepared in MOPS buffer with 10 % DMSO. The samples were transferred into black, flat bottom, 96well microtiter plates (Greiner) and fluorescence emission was measured at 520 nm (excitation at 450 nm) S5 with a microplate reader (Infinite 200 Pro, Tecan). Following the measurement of the intact peptides ( Figure S2a), 1 nM TLP-ste C288L/N181C was added to fully hydrolyze the substrate peptides (>16 hours incubation at room temperature in the dark). After complete hydrolysis of the substrates, the fluorescence measurement was repeated ( Figure S2b). Figure S2. Solubility of all substrate peptides (a) and their corresponding products (b). For all samples, the fluorescence intensity of the donor is shown (l ex = 450 nm, l em = 520 nm). For a better visualization, the data was recalculated so that the fluorescence intensity of all substrates displays the same value at a concentration of 2 µM. An intensity of 100 arbitrary units corresponds to a substrate concentration of 2 µM. Lines are drawn to guide the eye.
For all FRET-labeled substrate peptides, an almost linear relationship was observed at concentrations below 5 µM ( Figure S2a). Above this concentration, deviations from a straight line are visible for the Cy5-labeled substrates, indicating that the FRET-peptides are approaching the solubility limit. After hydrolysis of the substrates, even stronger deviations were observed for some of the peptides, especially for 4aa-Cy5 and 6aa-Cy5 ( Figure S2b). When the concentration of the hydrolyzed peptide exceeded 2.5 µM, the fluorescence intensity showed a plateau or even a slight decrease, suggesting that the cleaved S6 products form aggregates above the respective concentrations. It should be noted that the solubility data shown here was not corrected for the inner filter effect. For the AF647-labeled substrate peptides a linear relationship was observed up to 8 µM, suggesting that the contribution of the IFE is small in the range of concentrations tested.

Determination of the FRET efficiencies from relative quantum yields
In addition to the fluorescence intensity and lifetime also the quantum yield can be used to determine the FRET efficiency E: The quantum yields were determined from fluorescence intensities relative to a standard fluorophore.
Fluorescein (Sigma-Aldrich), dissolved in 0.1 M NaOH, was used as the standard. Its quantum yield in 0.1 M NaOH is equal to 0.925. 2 The substrate samples were diluted in MOPS buffer with 10 % DMSO.
Several solutions of fluorescein and the different substrates were prepared, spanning a concentration range between 0 nM and 1000 nM. Absorbance scans were measured in a 1 cm quartz cuvette using a UV/VIS spectrometer. Fluorescence emission scans were measured in the same cuvette, using a fluorescence spectrometer (l ex = 450 nm). For every concentration measured, the fluorescence emission peak was integrated from 470 nm to 620 nm. The obtained values were plotted against the absorbance at 450 nm and a linear fit was performed to obtain the slopes for all substrate samples and the fluorescein reference.
The quantum yield was calculated from these slopes using the following equation: where Φ is the quantum yield and n is the refractive index of the solution. The refractive index of NaOH is 1.8431 and the refractive index of MOPS was measured with a differential refractive index instrument (Optilab T-rEX, Wyatt Technology) and equals 1.3389. Table S1. Comparison of the FRET efficiencies determined in the intensity-based measurement (E I ) and

S7
in the measurement using relative quantum yields (E F ). Each measurement was performed in triplicate.
The values represent the mean ± the standard deviation. It should be noted that the relative quantum yields are based on an intensity-based measurement and are therefore not truly independent. We are reporting this measurement as it provides additional confirmation that the FRET efficiency values obtained from the microplate reader-based measurement are correct. In the microplate reader, only one intensity value is taken (narrow band pass at the emission maximum) at one concentration. In contrast, in the measurement based on relative quantum yields, the analysis is based on the complete donor emission spectrum obtained at several different concentrations. The values determined with the two different methods show excellent agreement (Table S1).

Fluorescence lifetime measurements
Time-resolved fluorescence measurements were performed to determine the FRET efficiency and the purity of the FRET-labeled peptides. The fluorescence lifetime of the donor is lowered when the acceptor is present and energy transfer is taking place. Comparing the fluorescence lifetime of the donor in the presence and absence of the acceptor consequently yields direct information about the FRET efficiency.
If the substrate is pure, all fluorescent species in the sample will be characterized by the same short lifetime. Fluorescent impurities (such as free donor) possess a different (longer) lifetime and will become visible during the time-resolved measurement.

S8
Donor fluorescence decay histograms were measured for all FRET-labeled peptide substrates. In addition, several product molecules were measured, i.e. the enzymatically cleaved substrates 4aa-Cy5 and 4aa-AF647 as well as a chemically synthesized peptide CF-PSVAG-NH 2 that mimics the cleaved Nterminal fragment of the substrates 8aa-Cy5 and 8aa-AF647.

Measurement setup
Fluorescence decays were measured using a custom-built confocal microscope equipped with an ultrafast avalanche photodiode (APD) for time correlated single-photon counting (TCSPC). 3 The

Characterization of the FRET-labeled peptides
All substrates were diluted to a concentration of 1 µM in MOPS buffer with 10 % DMSO. The cleaved peptides, the reference peptide CF-PSVAG-NH 2 and free CF were used in a concentration of 100 nM (in MOPS buffer with 10 % DMSO). All samples were freshly prepared before the measurement. The sample (100 µl) was pipetted onto a clean microscope coverslip placed into the microscope sample holder. The excitation light was focused into the sample solution at a distance of approximately 5 µm above the coverslip surface. Fluorescence decay histograms were recorded in a 50 ns time window (4 ps channel width) until 10000 counts were collected in any of the channels. In this way, the noise level (uncorrelated photon counts) remained below 50 counts per channel, thus providing a good signal-to-noise ratio.
The fluorescence lifetime t is defined as the average time it takes a population of excited fluorophores to reach 1/e of the initially excited population. The decay of the excited fluorophores typically follows first order kinetics, resulting in an exponential decay in the fluorescence intensity histogram I(t), that follows after the excitation pulse: where a j is the amplitude fraction of the j:th component with the corresponding lifetime t j .
In a TCSPC measurement, the measured fluorescence decay histogram I meas (t) is a convolution of the real decay I(t) with the IRF: The real decay and the fit parameters a j and t j are typically obtained using nonlinear fitting and a deconvolution procedure. Here, fluorescence lifetime analysis was performed using a free Matlab package developed by Enderlein et al. 4 The fitting algorithm is based on a Nelder-Meade simplex method and is capable of fitting multi-exponential decays to the measured data, including an experimentally measured IRF. The algorithm does not only fit the decay times, but also considers a time shift between the fluorescence photons and the IRF. with the IRF (scattered photons; t 1 ). As a result, it is hardly feasible to obtain an accurate lifetime for the donor in the presence of the acceptor.
It is interesting to note that the scatter component possesses a higher amplitude for the substrate samples when compared to free CF and the different product molecules. The amplitude for the scatter component is especially high for the AF647-labeled substrates, which have a shorter donor lifetime than the Cy5labeled substrates. This does not mean, however, that the absolute number of scattered photons has increased. Most likely, the AF647-labeled substrates possess a higher FRET efficiency so that the overall number of detected donor photons is lower. Assuming a constant number of scattered photons in all measurements, this directly leads to a relative increase in the amplitude of the scatter component. This observation suggests that the donor lifetimes become more and more difficult to determine with increasing FRET efficiencies.

S12
When comparing t 2 (short component, high FRET) and t 3 (long component, low FRET) for the Cy5and AF647-labeled substrates it becomes evident that the Cy5-labeled substrates possess a very high amplitude for the low FRET component characterized by the lifetime of CF (>85 %). This is also directly visible in the fluorescence decay histograms, which show a high similarity with the decay curve of the cleaved peptide ( Figure S3). Even though these amplitudes are not directly related to the concentration of the different species in the sample (different species emit a different number of photons), it can still be concluded that the Cy5-labeled peptides contain a higher fraction of donor only species (free donor or peptides lacking the acceptor). Clearly, the Cy5-labeled peptides are less pure than the AF647-labeled peptides.

Specificity of the enzymatic reaction
Two different control experiments were performed to prove that the chosen peptide sequence is a substrate of TLP-ste and that the change in donor emission is indeed the result of a TLP-ste catalyzed reaction: 1) incubation of the substrate under assay conditions in the absence of TLP-ste (autohydrolysis); 2) preincubation of the enzyme with the inhibitor ethylenediaminetetraacetic acid (EDTA).

Substrate autohydrolysis
The as the acceptor fluorophore. No increase in donor intensity is observed for the samples lacking the enzyme.

Enzyme inhibition with EDTA
As an additional test to prove that the peptide is specifically cleaved by TLP-ste, the enzyme was pre-

Correction for the inner filter effect
The inner filter effect (IFE) describes the influence of the fluorophores coupled to the substrate on the fluorescence intensity of the product formed during the catalytic reaction. To account for the IFE, the measured fluorescence intensities F meas need to be multiplied by an IFE correction factor CF IFE : F corr = CF IFE ´ F meas . 1,5-6 CF IFE was determined for all five substrates and every substrate concentration used in the kinetic measurements (0 µM -5 µM). As the IFE is sensitive to the overall concentration of chromophores in solution, CF IFE was also determined for different product concentrations (0 µM, 0.05 µM, 0.1 µM and 0.5 µM). Although only the donor-labeled product needs to be taken into account for the IFE, the solutions contained both the donor-and the acceptor-labeled product, for example CF-PSVAG-NH 2 and H-LAGGC(AF647)-NH 2 for the substrate 8aa-AF647. In this way, the conditions used S15 in the kinetic measurements can be mimicked more accurately. The magnitude of the IFE can be determined experimentally from absorbance measurements 1,6 or, alternatively, from fluorescence measurements. 5 In the following, a comparison of the two different methods is performed.

Absorbance measurement
The IFE correction factor CF IFE can be calculated from the measured absorbance of a substrate solution at the maximum excitation and emission wavelengths of the donor ( Figure S6a). 1,6 Using a microplate reader, the corresponding absorbance values at 490 nm (l ex ) and 520 nm (l em ) were determined, using the absorbance at 720 nm as a baseline. where A(l ex ) and A(l em ) are the absorbance values at the excitation and emission maxima of the donor, respectively.

Fluorescence measurement
The IFE correction factor CF IFE can alternatively be determined from the fluorescence intensity of the product solution in the presence and absence of fluorescent substrate molecules. The fluorescence-based equation is: 5 where F P is the fluorescence intensity of the donor-carrying product in the absence of substrate and F P+S is the fluorescence intensity of the donor-carrying product in the presence of substrate ( Figure S6b). When comparing the values of CF IFE determined from the absorbance and the fluorescence measurements, very similar correction factors are obtained. The correction factors obtained from the absorbance measurements were used for correcting the kinetic data. As the absorbance-based measurement is easier to perform, this method is preferred for correcting the kinetic measurements. S16 Figure S6. IFE correction factors CF IFE determined for all substrate concentrations used. a) The absorbance-based correction factor was determined for all substrates and is independent of the peptide length and the acceptor fluorophore. For the data shown, no product was added to the measurement. b) The fluorescence-based measurement was only performed for the AF647-labeled substrates. The resulting correction factors are very similar to the correction factors determined in the absorbance-based measurement. For the data shown, a product concentration of 0.5 µM was used. All measurements were performed in triplicate. The graphs show the mean ± the standard deviation.

Michaelis-Menten and Eadie-Hofstee analysis
Representative Michaelis-Menten and Eadie-Hofstee plots for the corrected data (FRET efficiency and IFE) are shown in Figure S7 for all substrates. The corresponding mean values of K M and v max obtained from three independent measurements are reported in Table S4.  The K M and v max values obtained from the Eadie-Hofstee fit are lower than the values from the Michaelis-Menten fit; however, the general trends are preserved. In the main text, we focus on the kinetic constants determined from fitting the data to the Michaelis-Menten equation.

Validation of k cat /K M values
Considering that substrate concentrations in the range of the K M value are not experimentally accessible, we have further obtained the k cat /K M values directly from the initial slope of the Michaelis-Menten plots and compared these values to the k cat /K M values calculated from k cat and K M determined independently.

Structure of the enzyme active site
TLP-ste is a relatively non-specific protease with a preference for hydrophobic amino acids in the S 1 ' position. Considering the possible influence of the hydrophobicity of the acceptor fluorophore (Cy5 vs. AF647), the distribution of hydrophobic amino acids surrounding the catalytic site of the TLP-ste homologue thermolysin was visualized ( Figure S8). This analysis suggests that the active site cleft is sufficiently large (~4 nm) to possibly interact with all residues of the longest peptides (8aa-Cy5 and 8aa-AF647). The S 1 '-S 4 ' part of the binding cleft possesses several hydrophobic amino acids that may interact with Cy5 (and the linkers), especially in the shorter substrates. With an increasing distance between Cy5 and this hydrophobic patch, this contribution may become weaker. Another possible explanation is that the different coupling site of AF647 orients the fluorophore differently with respect to the substrate peptide, preventing the fluorophore interaction with the enzyme. Furthermore, due to its sulfonation, the size of AF647 is larger than that of Cy5.