Article pubs.acs.org/JPCB
Temperature Sensitive Singlet Oxygen Photosensitization by LOVDerived Fluorescent Flavoproteins Michael Westberg,† Mikkel Bregnhøj,† Michael Etzerodt,‡ and Peter R. Ogilby*,† †
Department of Chemistry, Aarhus University, DK-8000 Aarhus, Denmark Department of Molecular Biology and Genetics, Aarhus University, DK-8000 Aarhus, Denmark
‡
S Supporting Information *
ABSTRACT: Optogenetic sensitizers that selectively produce a given reactive oxygen species (ROS) constitute a promising tool for studying cell signaling processes with high levels of spatiotemporal control. However, to harness the full potential of this tool for live cell studies, the photophysics of currently available systems need to be explored further and optimized. Of particular interest in this regard, are the flavoproteins miniSOG and SOPP, both of which (1) contain the chromophore flavin mononucleotide, FMN, in a LOV-derived protein enclosure, and (2) photosensitize the production of singlet oxygen, O2(a1Δg). Here we present an extensive experimental study of the singlet and triplet state photophysics of FMN in SOPP and miniSOG over a physiologically relevant temperature range. Although changes in temperature only affect the singlet excited state photophysics slightly, the processes that influence the deactivation of the triplet excited state are more sensitive to temperature. Most notably, for both proteins, the rate constant for quenching of 3FMN by ground state oxygen, O2(X3Σg−), increases ∼10-fold upon increasing the temperature from 10 to 43 °C, while the oxygen-independent channels of triplet state deactivation are less affected. As a consequence, this increase in temperature results in higher yields of O2(a1Δg) formation for both SOPP and miniSOG. We also show that the quantum yields of O2(a1Δg) production by both miniSOG and SOPP are mainly limited by the fraction of FMN triplet states quenched by O2(X3Σg−). The results presented herein provide a muchneeded quantitative framework that will facilitate the future development of optogenetic ROS sensitizers.
■
INTRODUCTION Reactive oxygen species (ROS), such as hydrogen peroxide, the superoxide ion (O2·−), and singlet oxygen (O2(a1Δg)), play important roles in maintaining cell homeostasis.1−6 This includes their involvement in signaling events, particularly upon perturbation of the cell.1−6 Recently, the toolbox used to study these processes has been expanded with genetically encodable, protein-encased ROS photosensitizers.7−14 These optogenetic sensitizers are important because they facilitate ROS production with (1) molecular level spatial control via protein fusion, and (2) temporal and dose control that is readily defined by the incident light.15,16 Additionally, encapsulation of the photosensitizer in a protein matrix ideally ensures a local environment that is independent of the subcellular localization of the photosensitizing system and this, in turn, ideally ensures that the photophysics of the sensitizer is likewise location independent. However, to realize the full potential of optogenetic sensitizers to investigate mechanisms of ROS initiated cell signaling, it is necessary to provide a more complete and quantitative description of the photophysics of the protein-encased ROS sensitizers that are currently deemed most promising. In this way, one provides a useful starting point for further developments that rely on rational design. Of all the ROS, we have long been interested in O2(a1Δg), the lowest excited electronic state of molecular oxygen.17 Depending on its local concentration, O2(a1Δg) initiates a © 2017 American Chemical Society
broad spectrum of biological processes ranging from cell death (apoptosis and necrosis) to cell proliferation (stimulated mitosis).6,16−20 However, at this time, a genetically encodable photosensitizer that selectively produces O2(a1Δg) at the expense of other ROS does not exist, and this hampers detailed investigations of ROS-mediated mechanisms of cell signaling.16 This issue has recently been addressed by a number of research groups, and pronounced improvements in the O2(a1Δg) quantum yields, ΦΔ, of optogenetic sensitizers have been achieved.8,10,11,13,21−23 For example, after it was shown that the celebrated LOV2-derived flavoprotein miniSOG (for mini singlet oxygen generator) actually produces O2(a1Δg) with low efficiency (ΦΔ = 0.03 ± 0.01),24,25 we produced a Q103L mutant designed to create more O2(a1Δg) at the expense of the kinetically competing electron transfer reactions that ultimately produce the superoxide ion (Scheme 1).11 (Note: In designating this mutation, we now refer to the numbering system of residues used in the original miniSOG report.8) This new protein, named SOPP for singlet oxygen photosensitizing protein, also contains flavin mononucleotide (FMN) as the active chromophore, but the removal of a conserved hydrogen bond between FMN and the protein scaffold results in an appreciable increase in the O2(a1Δg) yield (ΦΔ = 0.25 ± Received: January 18, 2017 Revised: March 1, 2017 Published: March 3, 2017 2561
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B
investigated previously. We therefore set out to test our hypothesis stated above by systematically examining and quantifying how these photophysical properties of FMN in SOPP and miniSOG depend on temperature. We now show that the rate of oxygen diffusion through the protein scaffolds of miniSOG and SOPP, and the associated quenching of 3FMN by O2(X3Σg−), indeed increases as a function of temperature. However, with this same increase in temperature (a) the rates of competing and undesired triplet deactivation processes also increase, and (b) the concentration of dissolved oxygen in the solution decreases. The combination of these latter phenomena counteracts the effect of faster oxygen diffusion and, consequently, only a modest increase in the O2(a1Δg) quantum yield is observed as the temperature of these FbFPs is increased from 10 to 43 °C. These results allow us to establish a more solid platform for our understanding of the oxygen-dependent photophysics of SOPP and miniSOG and, consequently, define clear directions for future optimization of O2(a1Δg) production by FbFPs.
Scheme 1. Pertinent Photoinitiated Processes Involving FMN in SOPP and miniSOG
0.03).11 The improved efficiency of SOPP over miniSOG has recently been confirmed by other groups in vitro26 as well as in vivo.27 Nevertheless, SOPP still has its limitations with respect to ROS selectivity (i.e., the superoxide ion is still produced as a consequence of electron transfer reactions involving FMN11), and the O2(a1Δg) yield of ∼0.25 is certainly far from the value of ∼0.65 found for FMN that is not caged in a protein. With these observations in mind, and in light of the fact that the FMN triplet state, 3FMN, is the immediate precursor to O2(a1Δg), it is important to note that, under aerated conditions, the lifetime, τT, of the protein-shielded 3FMN in SOPP (τT ∼ 130 μs) and miniSOG (τT ∼ 40 μs) is much longer than that of freely dissolved 3FMN (τT ∼ 3 μs).11 The longer lifetimes of protein-encapsulated 3FMN can be attributed to inefficient quenching by ground state oxygen, O 2 (X 3 Σ g − ), as a consequence of hindered oxygen diffusion through the protein.11 This effect of protein encapsulation constitutes a challenge for the creation of an efficient O2(a1Δg) sensitizer since, in the kinetic competition of pathways by which 3FMN can be deactivated, one ideally wants the bimolecular collision with O2(X3Σg−) to dominate over other channels (Scheme 1). All of the available photophysical data on miniSOG and SOPP have been derived from experiments carried out at room temperature. In the least, it is easily argued that experiments performed at 37 °C should be more relevant for many studies involving live mammalian cells. However, it is also acknowledged that photophysical data acquired over a range of temperatures generally provides greater insight into systems in which multiple processes kinetically compete with each other. With this in mind, we hypothesize that the rate of oxygen diffusion through the protein matrix should increase with temperature due to increases in the extent of local and segmental motions of the protein. There is indeed appreciable precedence for this perspective.28−30 In turn, we hope that more efficient O2(X3Σg −) diffusion and the associated quenching of protein-shielded 3FMN will result in an increase of the O2(a1Δg) quantum yield at the expense of unwanted 3 FMN deactivation channels, such as protein-mediated electron transfer reactions and electronic-to-vibrational (e-to-v) energy transfer (Scheme 1). Of course, the caveat in this working hypothesis is that an increase in temperature may also promote the unwanted deactivation channels. Flavin-based fluorescent proteins, FbFPs, have been studied extensively in recent years,31,32 and aspects of the temperature dependent behavior of FMN fluorescence in these FbFPs have been qualitatively examined.33,34 However, to our knowledge, the temperature dependence of the photophysics of both the singlet and triplet states of FMN in FbFPs, particularly the photophysics that depend on oxygen, have not been
■
EXPERIMENTAL SECTION Chemicals. The reference O2(a1Δg) photosensitizer phenalen-1-one-2-sulfonic acid (PNS) was synthesized as described in the literature.35 Al(III) phthalocyanine chloride tetrasulfonic acid (AlPcS4) (Frontier Scientific), Coumarin 153 (99%, Sigma-Aldrich), 4-(dicyanomethylene)-2-methyl-6-(4-dimethylaminostyryl)-4H-pyran (>98%, Sigma-Aldrich), 4-dimethylamino-4′-nitrostilbene (>99.8%, Sigma-Aldrich), LDS 751 (Exciton), cyclohexane (HPLC-grade, Sigma-Aldrich), D2O (>99.9% D, Euriso-Top), methanol (HPLC-grade, SigmaAldrich), and o-dichlorobenzene (HPLC-grade, Sigma-Aldrich) were all used as received. Riboflavin-5′-monophosphate sodium salt hydrate (FMN) was used as obtained from Sigma-Aldrich. For most of the experiments involving freely dissolved FMN, the compound marketed as 73−79% pure was used. However, to test the accuracy of the data thus obtained, selected experiments were repeated using the compound marketed as 95% pure. Within the accuracy of our measurements, we could not discern any differences in the results obtained. The FMN bound in SOPP and miniSOG was produced endogenously by the E. coli BL21AI cells used for protein expression. The H2O-based (Milli-Q) and D2O-based 10 mM phosphate buffer saline (PBS) solutions were prepared by dissolving one PBS tablet (Sigma-Aldrich) in 200 mL of H2O or D2O to yield solutions with 2.7 mM KCl and 137 mM NaCl (pH 7.4 and pD 7.8, respectively). All measurements were performed in these buffered aqueous solutions. Protein Purification and Handling. The miniSOG and SOPP fusion proteins were expressed and purified as described previously.11 All measurements were performed on protein solutions that had been buffer-exchanged into PBS buffered H2O or D2O. Protein-containing solutions were saturated with oxygen or nitrogen at atmospheric pressure by passing the gas over the solution surface, while stirring the solution, for at least 60 min. This procedure was used to avoid protein denaturation at the air−water interface of bubbles that occur when passing the gas through the solution. Exposure of our solutions to the gas for periods longer than 60 min, and up to 24 h, did not yield any measurable difference in the determination of triplet lifetimes or O2(a1Δg) quantum yields (vide inf ra). Instrumentation and Methods. All optical experiments were performed in 1 cm path length quartz cuvettes. The 2562
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B
throughout the path length of the cuvette to avoid overlap associated distortion of the data.39 Probe light (710 nm, fwhm 10 nm) was isolated using a spectrograph placed after the sample and directed onto a PMT (Hamamatsu, H10721−20) operated at low gain. The signals were recorded on a 350 MHz oscilloscope (Tektronix TDS5032B). Time-resolved O2(a1Δg) → O2(X3Σg−) phosphorescence measurements were performed by irradiating the samples with fs laser pulses at a repetition rate of 1 kHz. The 1275 nm O2(a1Δg) phosphorescence signal was isolated using the combination of a 1064 nm long-pass filter and a 1290 nm band-pass filter (fwhm 80 nm) and monitored using a cooled near-IR PMT (Hamamatsu, R5509−42) operated in photon counting mode (Becker & Hickl, MSA-300). The sample absorbance at the irradiation wavelength did not exceed ∼0.1. Ground State Spectroscopy. Thermal release of FMN associated with protein denaturation was followed by circular dichroism (CD) spectroscopy performed on a JASCO-810 CD system, equipped with a Peltier temperature-controlled cuvette holder. The temperature dependent change in the CD signal of protein-bound FMN was followed at the 440 and 447 nm absorption maxima of miniSOG and SOPP, respectively. The temperature was continuously scanned from 40−80 °C with a ramp speed of 1 °C/min. The melting temperature of the holoproteins was determined by nonlinear fitting of the obtained data to a model for a two-state transition of a monomeric protein (see SI).40 Ground-state absorption measurements were performed using a dual-beam UV−vis−NIR spectrometer (Shimadzu UV3600) equipped with a temperature-controlled cuvetteholder. For all quantum yield measurements, the sample absorption spectrum was measured at the relevant temperature and the fraction of absorbed light was determined using the absorbance weighted by a spectrum of the fs-laser pulses.
samples under investigation were maintained at the desired temperature using cuvette holders through which temperaturecontrolled water was circulated. Warming/cooling units (Neslab RTE-101 and Haake C10−K10) were used to control the temperature of the circulated water. We estimate the precision with which the sample temperature was maintained to be ±0.2 °C, and the absolute accuracy with which the temperature was measured to be ±0.5 °C. Excited State Spectroscopy. A fs-laser system, described in detail elsewhere,36 was used as the excitation source for all measurements probing excited state photophysics. Either the direct 80 MHz output of the fs-oscillator or pulses emerging from the amplifier at a repetition rate of 25−1000 Hz were used. The laser power was adjusted using a combination of a half-wave plate and a Glan-Taylor polarizer attached to a motorized rotating mount controlled by a Labview program. For FMN excitation, the 840 nm laser output was frequencydoubled in a BBO crystal yielding pulses centered at 420 nm, while AlPcS4 was excited using 675 nm pulses obtained using an OPA. A software-controlled shutter was used to preclude sample irradiation when data were not being recorded. Fluorescence data were recorded on a home-built instrument. The detection path was perpendicular to the fs laser excitation path, and a magic angle polarization scheme was used to eliminate depolarization effects.37 Emitted light was collected and coupled into a spectrograph (Andor Technology, Shamrock 303i) using a 75 mm spherical lens in combination with a 50 mm cylindrical lens. The detectors attached to the separate ports of the spectrograph were an iCCD camera (Andor Technology, iStar 320T-73) and a PMT (Becker & Hickl, PMC-100-1). For these experiments, the sample absorbance at the irradiation wavelength did not exceed ∼0.05. Fluorescence lifetimes were determined from single photoncounting histograms that were recorded using the PMT connected to a TCSPC-module (Becker & Hickl, SPC-140). The frequency-doubled 80 MHz output of the fs-oscillator was used as the excitation source to facilitate time-efficient histogram build-up while avoiding pile-up effects.37 The instrument response function was determined at the excitation wavelength using a milk-powder solution to scatter light. The histograms were fitted using the open source FLIMfit 4.10.3 software that allows corrections for after-pulsing and incomplete decay of the fluorescence (see SI). Steady-state spectra used to determine fluorescence quantum yields were recorded with the iCCD camera. For these experiments, the laser amplifier was operated at 100 Hz to match the iCCD readout rate. The gate-width was set to 100 ns to allow for integration of photons from the full fluorescence decay (τs ∼ 5 ns, vide inf ra) associated with each laser pulse, and the gate opened ∼5 ns before the arrival of the laser pulses. The detection unit was intensity calibrated using the following standard fluorophore solutions: 1,1,4,4-tetraphenyl-1,3-butadiene in cyclohexane, coumarin 153 in methanol, 4-(dicyanomethylene)-2-methyl-6-(4-dimethylaminostyryl)-4H-pyran in methanol, LDS 751 in methanol,37 and 4-dimethylamino-4nitrostilbene in o-dichlorobenzene.38 Transient absorption decays were recorded with a “frontface” irradiation geometry as described in detail previously.11 The fs laser pump beam was collimated to a diameter of ∼3 mm yielding a pulse fluence of ∼2.5 mJ/cm2 at a repetition rate of 25 Hz. The probe beam consisted of the output of a Xe lamp that was passed through a water filter and a 400 nm long pass filter. The probe beam was kept smaller than the pump beam
■
RESULTS AND DISCUSSION For the work described herein, experiments were performed in both buffered H2O and D2O. The rationale for this approach is that the lifetime of O2(a1Δg) in D2O is appreciably longer than that in H2O, thereby (1) making it easier to optically detect O2(a1Δg) in D2O-based solutions, and (2) allowing us to readily change an important kinetic parameter that, in turn, facilitates a more accurate mechanistic interpretation of our data.17,41 As we have noted previously,11 this change in solvent also has subtle effects on the photophysics of both freely dissolved and protein-bound FMN, and this provides even further mechanistic insight. Thermal Stability. The CD signal of FMN bound to and immobilized by a protein is distinctly different from that recorded from FMN freely dissolved in solution.42 Thermally induced unbinding of FMN from SOPP and miniSOG was thus followed by measuring the CD signal at the absorption maximum of the respective samples (see Figure S1). These measurements indicate that, in the buffered aqueous systems used for our present study, the melting temperatures of the holoproteins are ∼65 °C. Similar thermal stability has also previously been observed for related FbFPs.33,34 On the basis of these results, we chose to investigate the photophysics of protein-bound FMN in both SOPP and miniSOG over the temperature range 10−43 °C; a range which covers physiologically relevant conditions and, at its maximum, is well below the melting temperature of the proteins. Specifically, we performed measurements at six temperatures: 2563
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B
Values of Φf in both H2O and D2O were determined against that of FMN in H2O (Φf (23 °C) = 0.25 ± 0.01).24,49,50 Temperature and solvent dependent changes in the refractive index, n, and thereby the collection efficiency of the emitted light, were taken into account by normalizing the recorded fluorescence intensities by n2, using published values of n at the pertinent temperature.37,51,52 The main results are presented in Figure 2, and a more extensive data compilation is presented in Table S1.
10, 16, 23, 30, 37, and 43 °C. The photophysical quantities determined were independent of whether the temperature was incremented or decremented over this range (vide inf ra). Temperature-Dependent Singlet State Photophysics. Absorption and Fluorescence Spectra. The influence of temperature on the potential energy surfaces of the ground (S0) and first excited (S1) singlet states of, respectively, FMN, SOPP, and miniSOG was probed via changes in their steadystate absorption and fluorescence spectra. When increasing the temperature from 10 to 43 °C, the S0−S1 absorption bands showed slight, but systematic changes (see Figures 1, S2, and
Figure 2. Fluorescence quantum yields, Φf, and S1 lifetimes, τs, of SOPP, miniSOG, and FMN plotted as a function of temperature for PBS buffered D2O and H2O solutions. The solid lines are linear fits to the data and serve only as a guide to the eye. The lifetimes shown are based on fluorescence emitted at 525 ± 20 nm. The error bars on Φf are associated with the estimated experimental precision of 3%. For τs, the error bars are the same size as the symbols shown.
Figure 1. Normalized absorption (full lines) and fluorescence (dashed lines) spectra of SOPP and miniSOG in PBS buffered H2O measured at the two extreme temperatures used for this study. Absorption spectra that have not been normalized are shown in Figure S3 to illustrate intensity changes, while the intensity changes in the fluorescence spectra are presented as Φf in Figure 2.
As previously observed in experiments performed at room temperature,11 FMN in SOPP and miniSOG has a larger Φf than FMN freely dissolved in bulk water. Our current value of Φf for miniSOG determined at 23 °C (Φf = 0.42 ± 0.02) is consistent with that reported by Wingen et al. (Φf = 0.41 ± 0.01).50 For all samples, we see a slight but systematic decrease in Φf as the temperature is increased. Additionally, unlike SOPP and miniSOG, Φf of free FMN exhibits a pronounced D2O/ H2O solvent isotope effect of ∼1.13 that is independent of temperature over the 33 °C range studied. A similar isotope effect has been observed at room temperature over a large pH/ pD range.53 We return to this observation below. To complement Φf measurements, values of τs were determined as a function of temperature. For freely dissolved FMN, the time-resolved fluorescence signal could always be modeled by a monoexponential decay, and lifetimes of ∼4−5 ns were obtained (Figure 2). It is important to note that quenching of the S1 state by O2(X3Σg−) is negligible under atmospheric pressure as we observe almost identical S1 lifetimes in air and oxygen saturated solutions (within 2%), an observation that is expected with τs values of ∼4−5 ns in aqueous solution. The lifetimes recorded were independent of the detection wavelength and decreased with an increase in the temperature (Figure 2). Our more encompassing results are in good agreement with data available from independent studies.50,54,55
S3). These effects are most apparent for SOPP and miniSOG where the vibronic transitions of FMN are better resolved.43−46 Only subtle temperature-dependent changes were likewise observed in the fluorescence spectra (see Figures 1 and S2). The response of the spectral profiles to temperature was the same in both D2O and H2O. The subtle, but distinct, temperature dependent changes observed in the steady-state spectra of the proteins are consistent with a more flexible protein matrix and larger configurational space for FMN at higher temperatures. An increase in configurational degrees of freedom may result in inhomogeneous broadening of the spectra and, at the same time, facilitate access to a more “relaxed” fluorescent state with red-shifted emission. Indeed, differences between the nuclear geometries of the S0 and the relaxed S1 state have been found in theoretical studies of flavin photophysics.44,47,48 Thus, we speculate that the higher temperatures may better enable the nuclear transformation associated with the S0−S1 transition of protein-encased FMN. S1 Deactivation. We investigated the radiative and nonradiative deactivation channels of S1 by quantifying fluorescence quantum yields, Φf, and S1 lifetimes, τs, in air-saturated solutions as a function of temperature. 2564
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B Our data show that Φf and τs for free FMN decrease to the same extent over the 33 °C temperature range investigated. On this basis, we determined that the radiative rate constant for S1 deactivation, kf, is independent of temperature (see Table S1). This is arguably expected based on the minor temperature dependent shifts in the fluorescence spectra (Figure S2) which imply that the potential energy surfaces and transition dipole moment of the S1 to S0 radiative transition will be relatively unaffected by temperature. This insensitivity of kf to temperature has previously been observed over much larger temperature ranges for structurally related isoalloxazines in apolar solvents.56 The solvent isotope effect on Φf observed for freely dissolved FMN is also observed on τs (Figure 2). This observation shows that kf is the same for FMN dissolved in H2O or D2O (see Table S1), as might again be expected based on the absence of a solvent isotope effect on the absorption and fluorescence spectra. When considered as a whole, our results indicate that freely dissolved FMN has one or more nonradiative deactivation channels (e.g., internal conversion) that are sensitive to both temperature and solvent isotope exchange. These nonradiative deactivation channels are discussed below. The S1 lifetime of the FbFPs was determined by monitoring the time-resolved FMN fluorescence at 470/475, 525, and 650 nm (see Figures S4−S5). The experiments were performed at multiple wavelengths to assess whether or not τs depends on which part of the fluorescence spectrum one monitors, a phenomenon that has been observed by others for related flavins.55,57,58 Within the precision of our measurements, we find systematic but small changes in τs across the spectrum, which may indicate that the fluorescence originates from nonrelaxed as well as relaxed configurations of S1, with the relaxation process occurring on a ns-time scale in the proteinencased fluorophore (see SI). To discuss the temperature dependence of τs for SOPP and miniSOG (Figure 2), we focus on the lifetimes recorded at 525 nm. For both proteins, τs decreases with an increase in temperature; however, the decrease is less pronounced than for FMN in bulk water (i.e., 5% versus 12% over the 33 °C temperature range investigated). Furthermore, as with Φf, the effect of solvent perdeuteration on τs is not as pronounced for the protein complexes as for free FMN. As observed for free FMN, our data yield values of kf for FMN in SOPP and miniSOG that are independent of both temperature and solvent isotopic substitution. The larger Φf values for the proteins compared to freely dissolved FMN are a combined effect of larger radiative rate constants, kf, and smaller nonradiative rate constants, knr, in SOPP and miniSOG (see Table S1). Consequently, it appears that the protein matrices (1) capture the S1 state of FMN in geometries that have larger transition dipole moments than the relaxed S1 state of freely dissolved FMN, and (2) do not facilitate nonradiative decay to the same degree as a surrounding environment of water. Temperature-Dependent Triplet State Photophysics. Triplet State Quenching by Oxygen. We now turn to the FMN triplet state, 3FMN, that is the direct precursor for the sensitized production of O2(a1Δg). We determined 3FMN lifetimes, τT, in freely dissolved FMN, SOPP, and miniSOG as a function of temperature and O2(X3Σg−) concentration by recording transient absorption signals at 710 nm. At this wavelength, 3FMN is the only FMN-derived transient species that absorbs light; the absorption spectra of the semiquinone
radicals of FMN are blue-shifted.11,59−62 For all samples and at all investigated temperatures, the decay of the transient absorption signal could be fitted to a single-exponential function (see Figures 3 and S6). This observation suggests
Figure 3. Representative normalized transient absorption traces recorded at 710 ± 5 nm and assigned to 3FMN. Data are shown for SOPP and miniSOG in air-saturated D2O PBS at two different temperatures. In all cases, the data could be fitted to a monoexponential decay function (black lines).
that, in SOPP and miniSOG, FMN remains bound to the protein during our experiment. The values of τT we determined fall in the range ∼0.5−4 μs for freely dissolved FMN, ∼ 15−50 μs for miniSOG, and ∼20−240 μs for SOPP, depending on temperature, solvent perdeuteration, and O2(X3Σg−) concentration (see Table S2 for the exact numbers). We obtain the bimolecular rate constant for quenching of 3FMN by O2(X3Σg−), kq, from plots of the reciprocal lifetimes against the O2(X3Σg−) concentration (Figures 4 and S7). The values of kq thus obtained as a function of temperature are shown in Table S2. Quenching of 3FMN, both in the proteins as well as in bulk solution, may be interpreted in terms of the processes shown in Scheme 2. In this regard, we first note that we limit ourselves to diffusion-dependent processes.28−30 Second, in the encounter between two triplet states, we invoke the standard statistical argument that a singlet state is one out of the nine possible spin states that can be formed.41,63 Hence, we multiply the bimolecular rate constant for diffusional encounter, kd, by 1/ 9. As outlined below, the evidence obtained for FMN justifies the fact that we only consider this singlet state reaction coordinate (i.e., the fraction of 3FMN states quenched by O2(X3Σg−) that produce O2(a1Δg) is unity, SΔ = 1). In the case of freely dissolved FMN, the rate constants, kd and k−d, are associated with diffusion of oxygen in the bulk water, whereas for protein-encased FMN, kd and k−d represent the rate-limiting diffusion of oxygen across the water−protein interface and through the protein matrix (i.e., the more facile diffusion in the aqueous medium is neglected in this case.). From Scheme 2, the overall quenching rate constant, kq, can be expressed as shown in eq 1.64,65 2565
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B
protein-dependent decrease in kq corresponds to a proteindependent decrease in kd, the rate constant that quantifies O2(X3Σg−) diffusion into and through the protein. Carrying this point further, we can approximate the duration of O2(X3Σg−) transit from the protein surface to the chromophore as τtrans = (9kq · 1 M)−1 (see SI for a detailed discussion). Over the temperature range examined in our study, the estimated transit times of O2(X3Σg−) through the protein cover the range of ∼3 to ∼60 ns. Upon inspection of the crystal structure of a flavoprotein closely related to SOPP and miniSOG, the isoalloxazine ring of FMN appears shielded from the solvent and no path for O2(X3Σg−) diffusion toward the chromophore is apparent (Figure S9). As such, values of kq for SOPP and miniSOG are likely best interpreted in terms of protein dynamics (i.e., small amplitude motions of the amino acids and polypeptide chain segments that result in transient channels for O2(X3Σg−) diffusion).28,70−73 These fluctuations that allow O2(X3Σg−) to move through the protein scaffold may be modeled as an activated process and analyzed in terms of the Arrhenius equation.28,65 The slopes of the Arrhenius plots of kq yield apparent activation energies, Ea, that can be used to discuss properties of the given system that influence O2(X3Σg−) diffusion (see Figure 5).74−76 The resultant values of Ea for SOPP, miniSOG, and FMN freely dissolved in water are listed in Table 1. As expected, Ea associated with the quenching of free 3FMN by O2(X3Σg−) in aqueous solution is markedly smaller than the corresponding values obtained for the protein-encased systems (Table 1). However, given the error limits on our data, attempts to comment on apparent differences in Ea values for the proteins are not justified. Rather, we revert to the following inequality, illustrated in Figure 5: kq(SOPP) > kq(miniSOG). This point simply documents the fact that the Q103L mutation changes the extent to which FMN is hydrogen bonded to the protein and results in a matrix through which O2(X3Σg−) diffusion is facilitated. Triplet State Quenching: Oxygen-Independent Processes. As discussed above and illustrated in Scheme 1, the deactivation of protein-encapsulated 3FMN can be described in terms of two competing processes: (1) quenching by O2(X3Σg−) that results in the production of O2(a1Δg), and (2) oxygen-independent channels, such as electron-transfer reactions that involve the surrounding amino acid residues.11,24 The latter processes are quantified by the rate constant k0 found as the intercept in the quenching plots shown in Figure 4. Arrhenius plots of these k0 values are likewise presented in Figure 5. Upon examining the magnitude of these k0 values, it is clear that, relative to miniSOG, the Q103L mutation characterizing SOPP reduces k0 by a factor of ∼4−5 with an associated increase in the activation energy (Table 1). This is consistent with the fact that SOPP is a better O2(a1Δg) sensitizer than miniSOG.11,26,27 Even though the Arrhenius plots of k0 are linear over the temperature range studied, our results do not allow assignment of k0 to a specific deactivation process. Further experimental
Figure 4. Plots of the reciprocal 3FMN lifetime for (top) SOPP and (bottom) miniSOG against O2(X3Σg−) concentration in buffered D2O. The solid lines are linear fits to the data recorded at the six different temperatures. The slopes of these fits are used to determine the bimolecular rate constant for quenching of 3FMN by O2(X3Σg−), kq, while the intercept yields the rate of oxygen-independent unimolecular deactivation of 3FMN, k0.
kq =
1 kdkET 9 k −d + kET
(1)
3
The value of kq for FMN that we determined in bulk H2O at 23 °C (kq = (1.16 ± 0.12) × 109 M−1 s−1) is in agreement with previously reported values.66−68 Furthermore, for data we recorded at all temperatures, the obtained values of kq are close to 1/9 of the diffusion-limited rate constant (see Figure S8).65 Our results therefore indicate that O2(X3Σg−) quenches 3FMN at the diffusion-controlled, as opposed to the reactioncontrolled, limit (i.e., kET ≫ k‑d).64,65 Our results are also consistent with the assumption set forth in Scheme 2 that all 3 FMN states quenched by O2(X3Σg−) result in the formation of O2(a1Δg), a point to which we will again return in our discussion below. Oxygen Diffusion in the Protein Matrix. For 3FMN in SOPP and miniSOG, the temperature dependent values of kq are ∼100−300 times smaller than the values found for freely dissolved 3FMN (Table S2). Similar reductions in kq have been observed upon monitoring the triplet states of other chromophores encased in a variety of proteins.28,29 Since the electrostatic environment inside FbFPs is comparable to the electrostatic environment of water,48,69 we infer that quenching of 3FMN by O2(X3Σg−) in SOPP and miniSOG also proceeds to result solely in the production of O2(a1Δg) (i.e., the assumption set forth in Scheme 2, with the spin statistical factor of 1/9, is likewise valid here). From this perspective, the
Scheme 2. Formation of the Encounter Complex between 3FMN and O2(X3Σg−), and the Subsequent Process of Energy Transfer that Results in O2(a1Δg) Formation
2566
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B I (t ) =
⎛ t ⎞⎞ Itot ⎛ ⎛ t ⎞ ⎜⎜exp⎜ − ⎟ − exp⎜ − ⎟⎟⎟ τT − τΔ ⎝ ⎝ τT ⎠ ⎝ τΔ ⎠⎠
(2)
Here the temporal evolution of the 1275 nm signal intensity, I(t), is determined by the lifetime of the sensitizer triplet state, τT, and the lifetime of O2(a1Δg), τΔ. The parameter Itot gives the total integrated intensity of O2(a1Δg) phosphorescence which, when normalized by τΔ, is used to calculate the quantum yield of O2(a1Δg) production, ΦΔ.77 For a given sensitizer, fits of eq 2 to the resultant O2(a1Δg) phosphorescence data may thus be used to obtain ΦΔ upon comparison to a corresponding O2(a1Δg) phosphorescence signal obtained from a O2(a1Δg) sensitizer for which ΦΔ is known. In this study, we used PNS (ΦΔ = 0.97 ± 0.06 at room temperature78) as the reference sensitizer. It has been shown that ΦΔ for the unsulfonated derivative of PNS, 1-phenalenone, is temperature independent over the range 20−60 °C.79 Likewise, we find that ΦΔ of PNS is temperature independent over the range 10−43 °C (see SI). For FMN and PNS in H2O and D2O, the O2(a1Δg) phosphorescence signals could indeed be accurately modeled at all temperatures using eq 2. The data show that τΔ decreases as the temperature increases, which is consistent with our previous reports of temperature sensitive O2(a1Δg) lifetimes.80,81 The data also indicate that ΦΔ for FMN is independent of temperature in both H2O and D2O over the range 10−43 °C (Table S5), yielding average numbers of ΦΔ(FMN, H2O) = 0.56 ± 0.05 and ΦΔ(FMN, D2O) = 0.64 ± 0.05. Solvent Isotope Effects on FMN Photophysics in Bulk Solution. The ΦΔ of FMN does not increase upon saturating the aqueous solutions with oxygen (see Table S5 and Baier et al.66). This indicates that all 3FMN states formed in bulk solution are quenched by O2(X3Σg−) under air-saturated conditions. Based on the assumption that the quenching of 3 FMN by O2(X3Σg−) results solely in the production of O2(a1Δg) (i.e., SΔ = 1, vide supra), we then propose that the inequality ΦΔ(FMN, H2O) < ΦΔ(FMN, D2O) reported above is a consequence of the inequality ΦT(FMN, H2O) < ΦT(FMN, D2O). In our previous room-temperature study, we indeed found a similar solvent isotope effect on ΦT of FMN.11 Furthermore, this solvent isotope effect of ∼1.14 on ΦΔ matches the solvent isotope effect on the fluorescence properties of freely dissolved FMN (vide supra). Taking all of these observations together, we propose that S1 of FMN in bulk aqueous solution undergoes internal conversion via a mechanism that is more efficient in H2O than in D2O. This has previously been observed for other fluorophores and can be interpreted in terms of several different mechanisms.82−85 Kinetics of O2(a1Δg) Phosphorescence in the ProteinBased Systems. For the flavoprotein-sensitized O2(a1Δg) phosphorescence signals, the kinetics are more complex than for FMN in bulk solution. This is illustrated in Figure 6 where it is seen that eq 2 no longer adequately models the time
Figure 5. Arrhenius plots for the rate constants associated with the competing processes that deactivate 3FMN in SOPP and miniSOG. (top) Arrhenius plots for kq, the second-order rate constant for O2(X3Σg−)-mediated deactivation of 3FMN. (bottom) Arrhenius plots for k0, the first-order rate constant that characterizes all oxygenindependent channels of 3FMN deactivation. Data recorded from both PBS buffered H2O and D2O solutions are shown. The solid lines are linear fits to the data. The error bars associated with the k0 data in miniSOG are comparable in size to the markers. Absolute values of the rate constants are provided in Table S2.
and theoretical investigations of the possible competing processes of intersystem crossing and electron transfer in these proteins are needed to determine which processes define k0 and to what extent. Nevertheless, we are able to show, as outlined in the following sections, how temperature dependent changes in kq, k0, and O2(X3Σg−) concentration lead to temperature-sensitive O2(a1Δg) quantum yields for SOPP and miniSOG. O2(a1Δg) Formation by FMN: Different Environments. Formation of O2(a1Δg) by FMN in Bulk Solution. For the triplet state photosensitized production of O2(a1Δg) in a homogeneous solvent, the time-resolved O 2 (a 1 Δ g ) → O2(X3Σg−) 1275 nm phosphorescence signal can invariably be fitted using the function shown in eq 2.17
Table 1. Activation Energies for the Deactivation of 3FMN via both Oxygen-Dependent and Oxygen-Independent Channels SOPP Ea(kq) [kJ/mol] Ea(k0) [kJ/mol] a
miniSOG
FMN
H2O
D2O
H2O
D2O
H2O
D2 O
51.4 ± 3.0 17.0 ± 1.2
54.9 ± 3.1 17.6 ± 1.6
47.7 ± 3.1 12.4 ± 0.5
50.7 ± 3.0 11.9 ± 0.6
25.0 ± 1.1
25.5 ± 1.6
a
a
Our data do not warrant a determination of these values (see SI for details). 2567
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B evolution of the O2(a1Δg) signal. Several points need to be considered as we analyze these data.
incident laser was then increased stepwise, successively collecting new O2(a1Δg) phosphorescence traces. If the kinetics of signals collected at higher power systematically deviated from the kinetics determined at lower powers, the recording was discarded. Second, the lifetimes of 3FMN bound in SOPP and miniSOG are long, certainly compared to the lifetime of freely dissolved 3 FMN (Table S2). As a consequence, in D2O solutions of SOPP and miniSOG, τΔ < τT or τΔ ∼ τT which makes the O2(a1Δg) phosphorescence fitting routine less stable and the assignment of the respective lifetimes less obvious. In H2O solutions, τΔ ≪ τT which yields a very low intensity O2(a1Δg) phosphorescence signal and the associated low signal-to-noise ratio likewise increases the uncertainty of the fitting routine (Figure 7). Third, the inability of eq 2 to adequately model the SOPP and miniSOG sensitized time-resolved O2(a1Δg) phosphorescence signals is readily seen by examining the associated residuals plots of the nonlinear regression (Figure 6). Upon using a 1200 nm band-pass filter (fwhm 30 nm) instead of the 1290 nm band-pass filter (fwhm 80 nm), the signals disappear. Thus, we confirm that our signal source is indeed O2(a1Δg), and the inadequacy of eq 2 to fit the data is therefore not due, for example, to the admixture of spectrally broad phosphorescence from 3FMN. On the basis of these points, more complex kinetic models that take into account the heterogeneous nature of the proteinbased systems were therefore explored. Escape of O2(a1Δg) from the Protein Matrix. In the photosensitized production of O2(a1Δg) by the flavoproteins, O2(X3Σg−) must diffuse through the protein matrix and quench the encapsulated 3FMN in a process that requires orbital overlap of the respective molecules.41,86 Therefore, O2(a1Δg) is initially formed inside the protein matrix, and one must consider all the competing channels for O2(a1Δg) deactivation/ removal from that point in time and space. Specifically, we must consider the possibility that some fraction of the O2(a1Δg) produced will be removed, by reaction and/or physical quenching, as it diffuses out of the protein to encounter the bulk solvent with a different set of O2(a1Δg) deactivation parameters. In this discussion, it is first useful to have an estimate of the transit time required for O2(a1Δg) to diffuse from its source of production in the protein out to the surrounding bulk solvent. Assuming that O2(a1Δg) and O2(X3Σg−) have the same diffusion coefficient inside the protein, this transit time will be equivalent to that estimated from our independent experiments of 3FMN quenching by O2(X3Σg−) (vide supra): τtrans ∼ 3−60 ns. In support of this estimate, recent molecular dynamics simulations indicate that oxygen will diffuse out of miniSOG on a time scale of ∼1 ns at room temperature.87 Likewise, for another small protein, myoglobin, extensive studies of oxygen transport in and out of the active site indicate that the transit time is ∼150 ns at room temperature.73,88 Clearly, these estimates all define a time constant that is short relative to (a) the lifetime τT of the O2(a1Δg) precursor, 3 FMN, in the protein (vide supra), and (b) the lifetime of O2(a1Δg) in the D2O solution surrounding the protein. The short time for O2(a1Δg) transit out of the protein, τtrans, and the resultant inequalities (i.e., τtrans ≪ τT and τΔ) are critical with respect to the models we invoke in our data interpretation. Model #1. We first consider a model developed by Lepeshkevich et al.88 for a Zn-based myoglobin system. This
Figure 6. Main windows show representative time-resolved O2(a1Δg) phosphorescence traces (gray) recorded upon excitation of SOPP and miniSOG in air-saturated D2O-based PBS buffer at 23 °C. Data recorded within the first 2 μs after excitation have been excluded from the analysis due to contributions from scattered excitation light, fluorescence, and optics luminescence. The solid orange and blue lines are fits to the data based on eqs 2 and 4, respectively. In these fitting routines, we used values for τT that were determined in independent transient absorption experiments (see text). The O2(a1Δg) lifetimes, τΔsolv, extracted using each of the two fitting functions are reported in the associated colors. Note that values of τΔsolv obtained from the fits are less than the value of 68 μs for neat D2O (see text). The black curves represent the contribution from O2(a1Δg) sensitized by freely dissolved FMN that we infer coexists in the solutions of SOPP and miniSOG. The normalized residuals associated with each of these fits are shown in the respective colors in the secondary windows. Normalization of the residuals was achieved by dividing by the square root of the signal intensity at each time bin. The signal intensity originating from O2(a1Δg) produced by free FMN is given by the area under the black curve, while the signal intensity from O2(a1Δg) that is produced by protein-bound FMN is given by the area between the blue and black curves.
First, the main protein folds of both SOPP and miniSOG contain six amino acid residues that are sensitive to oxidation by O2(a1Δg). During the process of collecting O2(a1Δg) phosphorescence signals, the samples are subjected to gradual oxidation by O2(a1Δg), which has been observed to change the signal kinetics.11,24,25 Thus, for the present study, we only examined the photosensitizing properties of freshly prepared protein solutions under conditions of minimal accumulated photons. We ensured that we worked in this limit by initially irradiating each sample at low incident power. The power of the 2568
DOI: 10.1021/acs.jpcb.7b00561 J. Phys. Chem. B 2017, 121, 2561−2574
Article
The Journal of Physical Chemistry B
inequality τΔptn ≪ τTptn. Specifically, it is important to note that the lifetime of O2(a1Δg) inside the protein, τΔptn, is short not as a consequence of protein-mediated removal reactions but, rather, as a consequence of the short transit time of O2(a1Δg) out of the protein. The second term in eq 3 is analogous to what we show in eq 2, where τΔsolv denotes the lifetime of O2(a1Δg) in the bulk solvent, including any quenching of O2(a1Δg) by the dissolved protein (vide inf ra). Although the O2(a1Δg) phosphorescence signals recorded from SOPP and miniSOG in D2O are nicely modeled by eq 3 (see Figure S10), absolute values of selected kinetic parameters that we extract from these fits are not particularly realistic (see SI for details). Moreover, and to better accommodate O2(a1Δg) data recorded in H2O (vide inf ra), we considered a second model. Model #2. The stimulus for invoking Model #2 principally derives from complementary SOPP- and miniSOG-sensitized O2(a1Δg) phosphorescence experiments performed in H2O instead of D2O, which drastically changes values of τΔsolv without appreciably altering values of τTptn. In short, we exploit the fact that τΔ(H2O) is appreciably shorter than τΔ(D2O) and τTptn for both proteins at all investigated temperatures (in neat H2O, τΔ ∼ 3.4 μs and in neat D2O, τΔ ∼ 68 μs at room temperature.)80,89 For experiments performed in H2O-based solutions, the equation that quantifies the precepts of Model #1 (i.e., eq 3) does not account for the data observed (Figures 7 and S13). Specifically, because τTptn > 20 μs in SOPP and miniSOG in airsaturated solutions, the deactivation of the 3FMN state will be rate limiting in H2O where τΔsolv ∼ 3 μs, and τTptn will thus define the decay of our observed signal. However, as clearly seen in Figure 7, the data recorded from H2O indicate that our model must be able to accommodate a second decay component with a lifetime
