Proton Relay Reaction in Green Fluorescent Protein (GFP

Oct 6, 2006 - Calculation of transition dipole moment in fluorescent proteins—towards efficient energy transfer. Tamar Ansbacher , Hemant Kumar Sriv...
0 downloads 10 Views 338KB Size
Download PDF
J. Phys. Chem. B 2006, 110, 22009-22018

22009

Proton Relay Reaction in Green Fluorescent Protein (GFP): Polarization-Resolved Ultrafast Vibrational Spectroscopy of Isotopically Edited GFP Deborah Stoner-Ma,† Edward H. Melief,†,⊥ Je´ roˆ me Nappa,‡ Kate L. Ronayne,§ Peter J. Tonge,*,†,⊥ and Stephen R. Meech*,‡ School of Chemical Sciences and Pharmacy, UniVersity of East Anglia, Norwich NR4 7TJ, U.K., Departments of Chemistry and Biochemistry and Structural Biology Graduate Program, Stony Brook UniVersity, Stony Brook, New York 11794-3400, and Rutherford Appleton Laboratory, Central Laser Facility, CCLRC, Didcot, Oxon, OX11 0QX U.K. ReceiVed: August 17, 2006

The complex transient vibrational spectra of wild type (wt) GFP have been assigned through polarization anisotropy measurements on isotopically edited proteins. Protein chromophore interactions modify considerably the vibrational structure, compared to the model chromophore in solution. An excited-state vibrational mode yields information on excited-state electronic structure. The proton relay pathway is characterized in more detail, and the protonation of the remote E222 residue is shown to occur in a concerted step. Modifications to protein vibrational modes are shown to occur following electronic excitation, and the potential for these to act as a trigger to the proton relay reaction is discussed.

Introduction The green fluorescent protein (GFP) from the jellyfish Aequorea Victoria is a noninvasive genetically encoded sitespecific fluorescence marker that has had a major impact on cell biology.1 Fluorescent proteins have been widely used for imaging events such as gene expression and protein trafficking in living cells. There is also an interest in using GFP as an intracellular reporter of environmental factors such as pH and metal ion concentration.2-7 While selection methods provide a powerful means of generating fluorescent proteins (FPs) with novel properties, a detailed understanding of the photophysics of GFP will facilitate the rational modification of FPs in order to enhance their sensitivity and selectivity for a specific analyte. Only then will it be possible to harness fully the power of GFP as an intracellular sensor. Absorption and fluorescence of GFP arise from a covalently bound chromophore formed through a cyclization and oxidation reaction among the amino acid residues Ser65-Tyr66-Gly67.8-10 The photophysical properties of the resulting chromophore are dramatically influenced by the protein host. A synthetic analogue of the chromophore (4′-hydroxybenzylidene-2,3-dimethylimidazolinone, HBDI; Figure 1) is essentially nonfluorescent in fluid solvents at room temperature (quantum yield, Φf, < 4 × 10-4) and recovers fluorescence only on freezing in a glassy matrix at 77 K.11 The rapid radiationless decay arises due to a volumeconserving excited-state conformational relaxation, which promotes subpicosecond internal conversion.12-14 The precise nature of the coordinate(s) promoting internal conversion has (have) not been conclusively established, although experimental * Corresponding authors. Phone: 44-1603-593141 (S.R.M.), 631-6327907 (P.J.T.). Fax: 44-1603-592003 (S.R.M.), 631-632-7960 (P.J.T.). E-mail: [email protected] (S.R.M.), [email protected] (P.J.T.). † Department of Chemistry, Stony Brook University. ⊥ Department of Biochemistry and Structural Biology Graduate Program, Stony Brook University. ‡ University of East Anglia. § CCLRC.

Figure 1. HBDI and the calculated transition dipole moment.

data show there is a negligible barrier to motion along them12,14 and quantum chemical calculations suggest the involvement of a twisting about the exocylic bonds.15,16 In particular, recent CASSCF calculations suggest a role for torsional motion about the phenyl single bond.17 In sharp contrast, Φf for the chromophore in GFP is >0.8 and essentially independent of temperature, clearly illustrating the importance of the protein matrix in determining the photophysical properties of the chromophore. Understanding the nature of this interaction is central to the rational design of intracellular sensors. The structures of wtGFP and several of its mutants have been determined with high accuracy.18-22 In wtGFP and most mutants, intramolecular motion of the chromophore is suppressed both by packing of the surrounding residues and by an extensive hydrogen bonding network (Figure 2), which has the effect of both modifying the electronic spectra and suppressing radiationless decay. A second remarkable effect of the protein matrix is that it promotes an excited-state protontransfer reaction, which has not been observed in HBDI.23,24 In wtGFP, the ground-state absorption spectrum comprises two bands that are assigned to the neutral (protonated) and anionic (deprotonated) forms of the chromophore, the neutral form being dominant in wtGFP.25 Crystal structures of neutral and anionic forms suggest that in the anionic form the phenolic proton has been transferred to an adjacent protein residue and there has been a reorganization of the hydrogen bonds around the chromophore (Figure 2).20 Following optical excitation of the neutral form of wtGFP, the dominant fluorescence is seen to arise from the anionic form, with neutral fluorescence being barely observable (i.e., a proton-transfer reaction occurs with high efficiency).23,24,26 Careful studies of the picosecond timeresolved fluorescence and steady-state spectra of wtGFP has

10.1021/jp065326u CCC: $33.50 © 2006 American Chemical Society Published on Web 10/06/2006

22010 J. Phys. Chem. B, Vol. 110, No. 43, 2006

Stoner-Ma et al. characterize the proton relay mechanism in more detail, and to identify some unusually large frequency shifts between groundand excited-state modes in wtGFP. The implications of the mode assignments for the mechanism of the proton relay in wtGFP are discussed. Experimental Section

Figure 2. Three forms of the GFP chromophore, indicating the possible hydrogen-bonding interactions (adapted from Brejc et al.20).

established a three-state mechanism in which the neutral excited state (A*) rapidly donates its proton to an adjacent protein residue leaving the chromophore in the excited anionic state, with the surrounding matrix in the geometry appropriate for the neutral form (I* state). The I* state may either fluoresce to repopulate the A ground state via a reprotonation reaction27 or, more rarely, undergo a rearrangement of the matrix creating the stabilized anionic ground state (B).23,24 Recent picosecond time-resolved spectroscopy has shown that neither the A* nor I* fluorescence emission profiles exhibit any time dependence, suggesting a simple two-state (A* and I*) reaction without any additional excited-state stabilization.26 Experiments utilizing electronic spectroscopy have established the fundamental photophysics of wtGFP, but not the fate of the donated proton. This is an important question with regard to applications of GFP, which we have recently addressed using time-resolved infrared (TRIR) spectroscopy.28 It was established that the proton-transfer reaction causes protonation of the E222 residue, which is consistent with structural models, but surprising in that E222 is not adjacent to the phenolic OH of the chromophore. Rather, E222 protonation occurs as a result of a proton relay chain extending from the chromophore via a bound water molecule and residue S205 to E222 (Figure 2). This mechanism was supported by van Thor and co-workers’ study of the E222D mutant.29 It was also established that the rate of protonation of E222 was indistinguishable from the rate of decay of the A* state measured by time-resolved fluorescence.28 This led us to propose a mechanism in which the rate-determining step is the initial proton transfer from the chromophore, and all subsequent steps in the relay are ultrafast, in agreement with some molecular dynamics simulations.30 More recent quantum chemical calculations propose alternative sites as the trigger for a concerted proton relay reaction.31,32 In the following article, we extend our TRIR measurements of wtGFP by recording polarization-resolved transient vibrational spectra and applying the method to study proteins in which the chromophore has been isotopically labeled. The results are contrasted both with the polarization-resolved TRIR studies of the HBDI chromophore and its isotopes and with quantum chemical calculations of the vibrational spectrum of HBDI. The combination of these methods allows us to assign most of the modes in the complex transient IR spectrum of wtGFP, to

Plasmid containing His-tagged wtGFP was obtained from Prof. Rebekka Wachter (Arizona State University) and was used to transform BL21-DE3/pLysS cells (Stratagene). Protein was expressed overnight at 25 °C in the presence of 0.8 mM IPTG and purified by metal-affinity chromatography using Ni-NTA resin (Novagen). Protein was exchanged into 20 mM potassium phosphate, pH 7.5, and 300 mM sodium chloride and concentrated using Microcon YM10 centrifuge filters (Millipore). An overnight, room-temperature chymotrypsin digest (1 mg per 50 mg of GFP) cleaved the affinity tag, which was then removed by passing the reaction mixture through fresh Ni-NTA resin. Chromophore concentrations were calculated from the A395 using an extinction coefficient of 21 mM-1 cm-1. Expression of 13C tyrosine labeled GFP followed the procedure of Parsons and Armstrong.33 BL21-DE3 cells (Stratagene) were transformed with the GFP plasmid. Cells were grown in minimal media to mid-log phase, at which point 500-600 mg of either 1-13C tyrosine (Cambridge Isotopes) or 3-13C tyrosine (ICON Stable Isotopes) was added and the temperature was reduced to 25 °C. After 30 min, IPTG was added to 0.8 mM to induce protein expression. Cells were incubated for 20-24 h, and protein was purified as described for unlabeled protein. Yields were approximately half that obtained for unlabeled GFP and ranged from 4 to 8 mg/L cell culture. In these cases, the resultant chromophore was labeled at either the carbonyl (1-13C GFP) or bridge (3-13C GFP) carbon atom. The affinity tag was left intact in the labeled proteins. The protein samples were exchanged into D2O by being lyophilized and reconstituted a minimum of three times, followed by a final exchange into deuterated buffer using a Microcon YM10 centrifuge filter. Proteins were concentrated to 2-3 mM just prior to the experimental runs. The model chromophore compound, HBDI, and isotopically labeled HBDI were synthesized as previously described.34 Solutions of 30-40 mM HBDI in DMSO were used in these studies. Time-resolved IR spectra were obtained at the CCLRC Central Laser Facility. Descriptions of the system and data collection methods have been published.28,35 The laser system comprises an amplified titanium sapphire laser operating at a 1 kHz repetition rate, pumping a pair of optical parametric amplifiers. The pump beam was set at 400 nm (second harmonic of Ti:sapphire). The 150-fs pulse width allowed for roughly 400fs time resolution. The spot size and power of the pulse were adjusted so as to prevent photobleaching of the sample while retaining good signal-to-noise in the spectra. Typical settings included a 200-µm spot size and power levels invariably less than 1 µJ per pulse. Probe radiation in the IR region was generated by difference frequency mixing of the OPA output in a AgGaS2 crystal. Transmitted probe and reference spectra were collected on matched MCT detectors (32 pixels) in spectral regions approximately 150 cm-1 in width. Samples were contained in cells with CaF2 windows with spacers ranging from 6 to 50 µm. Sample rastering minimized overexposure to the pump beam. Runs typically consisted of 10-20 randomly ordered time delays, with a 10-s collection period for each delay. A minimum of three runs were taken for each sample at each spectral window, with the relative angle between the pump and probe

Proton Relay Reaction in GFP

J. Phys. Chem. B, Vol. 110, No. 43, 2006 22011

Figure 3. TRIR spectra of wtGFP. Spectra were recorded at 2 (yellow), 4 (turquoise), 10 (brown), 30 (green), 100 (blue), and 200 ps (red) after excitation. Spectra were recorded under magic angle conditions. The main transitions discussed are labeled.

Figure 4. IR spectra of HBDI in the 1500-1800 cm-1 region calculated by DFT. The transitions are labeled with the dominant displacement associated with the mode. Further details are given in Supporting Information.

beams varied in each run, using either magic angle (54.7°), parallel, or orthogonal orientations. These data permitted determination of the polarization anisotropy. The measured IR spectra were combined to generate the transient IR difference spectra at each delay time. The pixel to wavenumber conversion was based on the IR spectrum of water vapor. The reproducibility of spectral position was (3 cm-1, and the absolute accuracy was estimated as (5 cm-1. Vibrational spectra were calculated using density functional theory (B3LYP, 6-31++ G(d,p)) as implemented in the Gaussian 03 program.36 The structure of HBDI is shown in Figure 1, and the calculations pertain to the bare molecule. Results and Discussion TRIR Spectrum of wtGFP. The TRIR spectra of wtGFP as a function of time after excitation are shown in Figure 3. Three bands, at 1683, 1634, and 1595 cm-1, exhibit instantaneous bleaching (negative delta OD) and, within the signal-to-noise, no further temporal evolution of the maximum amplitude during 200 ps. The instantaneous bleach suggests the assignment of these bands to the ground state of the neutral form of the chromophore (A), directly excited by the 400-nm pump pulse. Preliminary assignments for these modes are based on earlier detailed studies of the Raman spectra of GFP, HBDI, and its isotopes,34 and on the calculated vibrational spectrum of HBDI (Figure 4). In order of decreasing energy, the calculated spectrum reports three strong IR allowed modes, assigned as follows: a mode mainly associated with the carbonyl stretch, but involving significant displacement of the bridging CdC mode; a mode involving the excocyclic CdC bond, but also associated with stretching of the CdN bond of the imidazolinone ring; and a CdC stretching mode essentially localized on the phenyl ring (see Supporting Information). These assignments of the IR difference spectra of wtGFP are consistent with earlier resonance Raman and IR studies of the protein and HBDI.34,37-39

Figure 5. Polarization-resolved TRIR for wtGFP. Parallel polarization (red) and perpendicular polarization (green). The anisotropy parameter (D, see text) is shown in black with the right-hand axis. (a) 4 ps after excitation. (b) 100 ps after excitation.

To reflect the major contribution to the displacement, these modes will subsequently be referred to as CdO, CdC, and phenyl 1. Two weaker modes appearing at slightly lower energies can be assigned to the CdN stretch of the imidazolinone ring and a second phenyl mode (phenyl 2), and at lower frequency still there is a third more intense phenyl mode (phenyl 3; Figure 4). In addition to the three prominent ground-state bleach modes, time-dependent changes in the TRIR spectra are seen, including a transient absorption at 1712 cm-1, a bleach that develops on the time scale of 10-100 ps in the complex band shape around 1560 cm-1, and a line shape change at 1615 cm-1. In addition to these time-dependent changes, there is a weak band of sigmoidal appearance at 1660 cm-1. Such a band shape can arise from combinations of bleaching and transient absorption, or a shift in frequency of a ground-state mode. It should also be noted that the shape of the second prominent bleach band (1634 cm-1) is also time-dependent (Figure 3). To aid in unraveling these complex time-dependent spectra, we will make use of polarization resolution and isotopic substitution (the latter being described in the next subsection). In Figure 5 are displayed the TRIR spectra for wtGFP recorded 4 and 100 ps after excitation, with the probe polarization oriented parallel (∆A||) and perpendicular (∆A⊥) to the pump polarization. The data may be compared with the same measurement for the neutral form of HBDI recorded 2 ps after excitation (Figure 6). Also shown in these figures is the ratio, D, where

D)

∆A|| ∆A⊥

from which the angle between the electronic transition moment of the chromophore and the transition dipole moment of the vibrational mode, θ, can be calculated using:40,41

cos θ )

(2DD +-21)

1/2

22012 J. Phys. Chem. B, Vol. 110, No. 43, 2006

Stoner-Ma et al.

Figure 6. Polarization-resolved measurement for HBDI in DMSO. The transient spectra were recorded 2 ps after excitation. Parallel (red) and perpendicular (green) measurements are shown along with the ratio D in black dotted line.

TABLE 1: Energies, Assignments, and Transition Dipole Orientations for TRIR Data of GFP and HBDI and Their Isotopically Substituted Derivativesa,b sample wtGFP wtGFP 1-13C GFP 3-13C GFP HBDI 1-13C HBDI 3-13C HBDI wtGFP wtGFP

assignment E222 (CdO) (A) Cro (CdO)

protein (S) Cro (CdC + CdO* + protein...)

wavenumber ((5) cm-1 1712 (A) 1681 obscured 1674 1698 1661 1696 1630 1638

D

θ (deg)

2.1 0.7

28 ( 4 67 ( 4

0.7 0.8

67 ( 4 62 ( 4

D (0.8-3.3) 1.1 51 ( 4 1.0 55 ( 4 D (1.5-2.6)

1-13C GFP 3-13C GFP HBDI 1-13C HBDI 3-13C HBDI wtGFP Cro (phenyl 1) 1-13C GFP 3-13C GFP HBDI 1-13C HBDI 3-13C HBDI wtGFP Cro (CdN, phenyl 2) weak 1-13C GFP weak

1626 1635 1635 1629 1621 1596 1594 1595 1601 1601 1601 1557, 1571

3-13C GFP 1-13C GFP Cro* (CdO) (A) 1-13C GFP E222 (COO-) wtGFP Cro phenyl 3 HBDI 1-13C HBDI 3-13C HBDI

1550, 1571 1530 1.1 1570 1510 2.9 1516 2.5 1517 1517

2.9 2.3 2.6 2.7

1559, 1568 1.3, 1.4

8(8 23 ( 6 17 ( 7 15 ( 7

46 ( 10, 43 ( 10 51 ( 8 8 ( 10 19 ( 7

a (A) ) Transient absorption. (S) ) Sigmoid shape. b Data are grouped according to mode assignment.

where D can take values between 0.5 (θ ) 90°) and 3 (θ ) 0°). The direction of electronic transition moment in HBDI was calculated by Usman et al.42 and is shown in Figure 1; the experimentally determined values for D and θ are presented in Table 1. The polarization-resolved data are a clear aid to assignment of the wtGFP TRIR spectra in that they both help to further distinguish the different vibrational modes and reveal complexities not immediately apparent in the transient spectra alone. An example of the latter is the 1637 cm-1 bleach in wtGFP, provisionally assigned to the CdC stretch of the chromophore on the basis of studies of HBDI and DFT calculations. This band is now revealed by the anisotropy measurements to be more complex (Figure 5). The strong dispersion observed for D within this band suggests contributions from excited-state absorption overlapping the ground-state CdC bleach (see below). In contrast, there is good agreement

Figure 7. TRIR spectra of isotopically labeled HBDI recorded 2 ps after excitation. HBDI (green), 1-13C HBDI (red), and 3-13C HBDI (blue).

between the anisotropy measured for the chromophore CdO and phenyl 1 modes in both wtGFP and neutral HBDI, supporting these assignments (Table 1). TRIR Spectroscopy of Isotopically Labeled GFP. The TRIR spectra for isotopically labeled HBDI are shown in Figure 7. These data may be compared with data for the labeled proteins and wtGFP measured at 4 and 100 ps after excitation (Figure 8). Both sets of spectra were recorded with a pumpprobe polarization set at the magic angle to eliminate the effect of relative orientation of the modes on the intensity. The effects of 13C labeling on the HBDI TRIR spectra are consistent with earlier static Raman and FTIR measurements as well as quantum chemical calculations.34,39 Substitution on the carbonyl group causes a large shift in the CdO mode and a smaller but readily observable shift in the CdC. This is consistent with the DFT calculations, which showed that this mode involved a significant stretch in the exocyclic CdC bond as well as the CdO stretch. Substitution on the bridging carbon atom causes a shift in the CdC mode, as expected, and a much smaller effect on the CdO mode, again in line with the DFT calculation. Neither isotopic substitution causes a significant shift in the 1600 cm-1 band, consistent with localization of this mode on the phenyl ring. This mode is therefore used to intensity normalize both the HBDI and the labeled GFP data (Figure 8). The comparison of Figures 7 and 8 reveals that the chromophore vibrational spectrum is substantially perturbed by the interaction with the protein matrix. First the mode dominated by the CdO stretch is shifted to lower energy by 15 cm-1 between HBDI in DMSO and wtGFP. This suggests a weakening of the bond, probably due to hydrogen-bonding interactions between the chromophore carbonyl and the side chain of R96 as suggested by X-ray structural data.20 Also noticeable is a large change in the intensity distribution among the three most prominent bleach modes between the wtGFP and the HBDI spectra. In HBDI, the relative strength of the bleach decreases in the order phenyl 1 . CdC > CdO (Figure 7). In wtGFP, phenyl 1 is the weakest of the three, with CdO and the complex 1638 cm-1 line shape being of nearly equal intensity. Thus, one consequence of the protein matrix is an alteration in the relative transition moment of these two high-frequency vibrational modes. This, coupled with similarly large effects on the electronic spectra,16,38 suggests a significant perturbation of the molecular structure of the chromophore in the protein compared to HBDI in solution. An important validation of current and future theoretical calculations on the electronic structure of GFP will be whether computational approaches can reproduce these protein-dependent changes in mode intensities. Indeed, the mode intensity distribution may act as a sensitive probe of chromophore host interaction. The isotope effects are also different between the labeled HBDI and GFP, and in a somewhat unexpected fashion. The

Proton Relay Reaction in GFP

J. Phys. Chem. B, Vol. 110, No. 43, 2006 22013

Figure 8. Comparison of the TRIR spectra of wtGFP (green) with 13C isotopically labeled protein where the label is at the 1-C (carbonyl) position (red) and the 3-C (bridge) position (blue). Data are shown for 4 and 100 ps after excitation. The spectra were normalized to have the same delta OD for the 1600 cm-1 phenyl mode. 13C

substitution of the CdO bond leads to the disappearance of the 1682 cm-1 mode, consistent with its assignment (except for some residual bleach, probably due to unlabeled protein). However, there is no obvious reappearance of the bleaching of this mode at lower frequency. On the basis of the HBDI data, a 37 cm-1 shift to lower frequency would place the 13CdO mode at 1644 cm-1. However, there is no instantaneous bleach at that wavenumber in the protein TRIR spectra. The bleach observed at 1638 cm-1 in wtGFP is, however, somewhat broadened and shifted to lower frequency in the 1-13C GFP spectrum, suggesting the possibility that the bleach associated with the 13CdO mode is shifted below 1640 cm-1 to merge with the lower frequency band. The effect of 13C substitution at the bridge bond of the GFP chromophore is less dramatic. The mode most consistent with a CdC stretch (1638 cm-1) is shifted to lower frequency by 6 cm-1 (compared to a 15 cm-1 shift in HBDI). The CdO mode is also shifted by 6 cm-1 on 3-13C substitution, similar to the shift in HBDI (Figure 7). Isotopic substitution has a particularly significant influence on the TRIR spectra below 1600 cm-1 (Figure 8). The complex band shape around 1565 cm-1 in wtGFP reveals a timedependent bleach (Figure 3).28 In comparing the wt and the 1-13C GFP spectra recorded at 4 ps, a new transient absorption appears instantaneously at 1530 cm-1 in the labeled sample accompanied by an increased bleach in a pair of bands around 1565 cm-1. The prompt appearance of the 1530 cm-1 transient in the isotope spectrum means that it can be assigned to a mode of the A* excited state of the chromophore. It will be shown below that the 1560 cm-1 band in the 4-ps spectrum of wtGFP is a composite of these bleaches and absorptions. In the 100-ps spectrum of the 1-13C GFP, the 1530 cm-1 transient absorption has largely decayed away (consistent with the assignment to A*). Importantly, the 1-13C GFP spectrum continues to reveal a time-dependent growth in the bleaching around 1560 cm-1 (Figure 8). Band Assignments in the TRIR Spectra of wtGFP. Assignments of transitions observed in TRIR spectra are often more difficult than that for conventional ground-state vibrational spectra. This is because of the possibility of the overlap of transient absorptions with ground-state bleaches. In analyzing the spectra of GFP, four possible contributions to the measured

time-dependent change in absorbance must be considered: the A ground-state bleach, the A* excited-state absorption, the I* excited-state absorption, and changes in protein vibrational spectra due to perturbation by chromophore excitation and proton transfer. Experimentally, we found that it is unnecessary to consider the contribution from I*. Our earlier transient study of the mutant T230V/E222Q which has an I ground state43 did not reveal any excited-state modes in the region of interest.28 The reason for this is not known precisely, but it is a common observation that vibrational modes of excited electronic states have a less readily detectable transient absorption than the corresponding ground-state bleach. In addition, TRIR studies show that the ground-state CdO and CdC transition moments are weaker (when normalized to the phenyl 1 mode) in the anionic than in the neutral form of HBDI (data not shown). This suggests that transient absorption of I* may drop below an observable threshold in the protein measurements, as was found for HBDI anion by Usman et al.42 Thus, we will consider the three remaining contributions to the wtGFP spectra and use the isotopically labeled polarization-resolved data described above to distinguish between them. The 1712 cm-1 transient absorption that develops on a 200ps time scale after excitation of wtGFP is important for characterizing the proton relay mechanism. We previously assigned this absorption to the protonation of the E222 residue as the product of the proton relay mechanism following photoexcitation of the chromophore.28 This was shown to be consistent with the absence of any similar bands in either HBDI or the T203V/ E222Q mutant.28 The frequency of the mode is also consistent with the carbonyl of a protonated glutamate residue in a hydrogen-bonding environment.44 van Thor and co-workers recently observed that a similar but shifted mode is found in the TRIR spectra of the E222D mutant which is also in agreement with this assignment.29 The present data serve to further support the assignment. First, the mode is unaffected by isotopic substitution, confirming that it is not associated with the chromophore CdO mode. Second, the polarization-resolved data lead to an angle between the vibrational and electronic transition moments of θ ) 28°. This result can be compared with predictions based on structural data. By assuming that the vibrational moment lies on the axis of the CO bond in E222

22014 J. Phys. Chem. B, Vol. 110, No. 43, 2006 and projecting that onto the face of the chromophore, we determine an angle of 34°, while projecting edgeways onto the chromophore we find an angle of 20° (see Supporting Information). Thus, the model that accounts for the TRIR spectra is in good agreement with the X-ray structure of the protein. The band appearing at 1681 cm-1 in wtGFP is assigned to the CdO mode of the A ground state of the chromophore bleached promptly by the excitation pulse. Support for this assignment comes from three observations: disappearance of the mode on 13C isotopic substitution (Figure 8), similarity with the HBDI isotope shift measurements, and the common angle of ca. 65° determined from the polarization dependence in wt and 3-13C GFP and HBDI (Table 1). It should be noted here that for both GFP and HBDI there is a disagreement between the measured θ and the value calculated from DFT; the calculated value is 85°. This overestimation appears to be a consistent error, with the DFT-calculated angle being between 5° and 25° larger than that measured (see Supporting Information). This could reflect an error in the determination of the direction of the electronic transition moment, or our DFT calculation may give too much weight to the contribution of the mode from vibrations other than CdO (the orientation of the CdO bond relative to the calculated transition moment is approximately 70°). The bleach at 1596 cm-1 in wtGFP can confidently be assigned to the intense phenyl 1 mode on the basis of the following observations. The bleach is instantaneous with no strong recovery observed in 200 ps, consistent with an A form chromophore ground-state mode. Both the steady-state IR and DFT calculations report an intense phenyl mode in the IR spectrum of HBDI at a frequency below that of the CdO and CdC modes (ref 34 and Figure 4), in agreement with the TRIR data for HBDI (Figure 7). In addition, the frequency of the mode is independent of isotopic substitution in the protein, within experimental error, as expected for a mode remote from the site of substitution, while the frequency of the equivalent mode in HBDI is also unaltered by isotopic substitution (Table 1). Finally, the measurement of the anisotropy for this mode in HBDI, wtGFP, and its isotopes yields a common angle between vibrational and electronic transition moment of 16° ( 8°, which is close to but once again below the angle calculated by DFT (34°). The DFT calculation predicts three phenyl-localized modes in the IR spectrum (Figure 4). The second phenyl mode (phenyl 2), which is weak and overlapped with other modes around 1560 cm-1, is discussed below. The third more intense mode is seen in wtGFP at 1510 cm-1 (Figure 5) and in HBDI at 1516 cm-1. The mode is unaffected by isotopic substitution (Table 1) consistent with assignment to the phenyl ring. The polarizationresolved measurements lead to θ between 0° and 26°, in agreement with the DFT calculation, which yields 10°. The remaining modes are less readily assigned and require a consideration of the temporal evolution of the spectrum. We consider first the complex line shape around 1565 cm-1, which evolves in both amplitude and shape (Figure 3). We previously assigned the time-dependent increase in the bleach to a protein mode involved in the proton relay reaction, specifically to the loss of the anti-symmetric carboxylate stretch as the E222 residue is protonated.28 The bleach develops with the same time constant as the 1712 cm-1 transient absorption, consistent with this model, and the frequency is in the expected range for a protein carboxylate mode. A more detailed resolution of the complex line shape is now possible from a study of the 1-13C substituted GFP (Figure 8). In this sample, a new transient

Stoner-Ma et al.

Figure 9. Comparison of the TRIR spectra of wtGFP (filled symbols) and the 1-13C GFP isotope (open), where the transient absorption at 1528 cm-1 in the isotope has been shifted by 37 cm-1 and added to the spectrum. Data are shown at 4 ps (circles) and 100 ps (squares).

absorption develops promptly at 1530 cm-1 and decays away on a tens of picoseconds time scale. This can thus be assigned to excited-state absorption (i.e., a vibrational mode of A*). Also in the 1-13C isotope, the two bands around 1560 cm-1 are more clearly resolved 4 ps after excitation and have a stronger bleach, compared to wtGFP (Figures 3 and 8). However, even in 1-13C GFP this band still exhibits increasing bleach as a function of time after excitation (Figure 8), characteristic of a reaction involving a protein ground-state mode. To demonstrate that part of the complex line shape seen in wtGFP at 1565 cm-1 arises from this newly observed excitedstate absorption, we take the transient absorption at 1530 cm-1, shift it by 37 cm-1, add it to the transient bleach of 1-13C GFP, and compare the result with the corresponding wtGFP spectrum. This has been done for data recorded at both 4 and 100 ps after excitation. In both cases, the agreement between the spectrum calculated from the 1-13C GFP data and the observed wtGFP spectra is good (Figure 9). Thus, the wtGFP line shape around 1565 cm-1 comprises three contributions: the excited-state (A*) absorption, which decays with time; a pair of prompt bleaches at 1560 and 1570 cm-1, which are constant on the 200-ps time scale; and a time-dependent bleach. The A* absorption at 1530 cm-1 in 1-13C GFP is the first excited-state mode to be unambiguously identified in GFP. This mode is shifted by 37 cm-1 on 13C substitution at the chromophore carbonyl, supporting the assignment to the CdO mode of A*. Above we suggested that, in the 1-13C labeled protein, the CdO mode of A was hidden under the 1638 cm-1 band. This in turn suggests that the change in frequency for the CdO mode between A and A* is on the order of 90 cm-1. If this assignment is correct, the CdO bond is dramatically weakened in the electronically excited state of A. Several quantum chemical calculations on HBDI have suggested large changes in bond order between the ground and first excited electronic states.15-17,45 The largest changes are predicted to be for the bridging CdC bond, which is consistent with models of the excited-state conformational change in the excited state, leading to ultrafast radiationless decay.15-17 However, changes are also predicted in electron density around the CdO bond, consistent with the experimental observation of a weakening of this bond.45 In addition, the DFT calculations of vibrational spectra show that the CdO mode of A also involves a significant component of the CdC stretch, which is predicted to be weaker in the A* state than in the A state. A combination of these factors may result in the large shift seen for this mode between ground and excited states. The two instantaneous bleaches prominent in the 4-ps spectrum of the 1-13C GFP can be assigned to modes of the groundstate A form. On the basis of the ordering of modes in HBDI

Proton Relay Reaction in GFP

J. Phys. Chem. B, Vol. 110, No. 43, 2006 22015

Figure 10. Time dependence of the bleaching of the protein mode in 1-13C GFP. The spectrum recorded at 4 ps was subtracted from the TRIR spectra recorded at the times indicated.

and the DFT calculation, we provisionally assigned these to the phenyl 2 mode (1570 cm-1) and the CdN mode of the imidazolidone ring (1560 cm-1). Neither mode shifts significantly on 1-13C substitution. The CdN mode is known to be somewhat delocalized and to include a CdC stretch component, and as a result shifts 10 cm-1 on 3-13C substitution (Table 1). This mode is the most intense in the Raman spectrum of GFP.38,39 Polarization-resolved data yield values of θ of 46° and 43°, respectively, again in the correct order predicted by DFT, but significantly smaller than the calculated values (70° and 67°, respectively). Significantly, the temporal evolution in the 1565 cm-1 band of the 1-13C GFP sample persists even in the absence of the decay of the excited-state absorption. There are no other bands observed in this region for HBDI, nor are any calculated by DFT. Thus, we assign this increasing bleach as indicative of the loss of a protein vibrational mode as a function of time after excitation. Since there are only two contributions to the 1565 cm-1 band in the 1-13C GFP sample, and the contribution from the pair of promptly bleached A modes can be treated as timeindependent, we can isolate the contribution of the protein to the overall bleach by subtracting the early time data from the time-dependent spectrum of 1-13C GFP. The result is shown in Figure 10: a single broad (20 cm-1) bleach centered at 1570 cm-1, which is initially of zero amplitude, becoming more intense with increasing time after excitation. The position of the mode is consistent with a protein carboxylate mode, and its time dependence is consistent with a proton transfer-induced carboxylate to carboxylic acid transformation. Thus, this mode’s assignment to protonation on the E222 residue is confirmed. On the basis of the ordering seen in HBDI and from DFT calculations, we predicted that the 1638 cm-1 bleach in wtGFP includes the CdC stretch of the A ground state. There is some support for this assignment from the GFP isotopes studied (Figure 8), where the peak frequency shifts on both 1- and 3-13C substitution. This was also seen in HBDI, but the shifts are appreciably smaller in the isotopically substituted GFP, and the effect of 1-13C substitution is unexpectedly larger than 3-13C (Table 1). The polarization-resolved data suggest a more complicated picture than a simple CdC bleach. For both wtGFP and HBDI, the value of D is frequency-dependent. This corresponds to a polarization-dependent line shape, which is seen particularly clearly in Figure 5, where the band measured with parallel polarization is significantly broader than that for perpendicular. This behavior cannot result from a single mode, suggesting additional contributions within this band. The underlying complexity becomes more apparent from a detailed investigation of the time dependence of the line shape, which is shown in Figure 11a for wtGFP. On the high wavenumber side, a shoulder is resolved in wtGFP, but not in the two isotopically substituted forms. On the low wavenumber side,

Figure 11. Time dependence of the line shape of the 1638 cm-1 bleach in (a) wtGFP (b) 3-13C GFP, and (c) 1-13C GFP. Times are in picoseconds.

a sigmoidal shape develops as a function of time after excitation in wtGFP. A transient absorption appears promptly around 1615 cm-1 and decays away on a tens of picoseconds time scale as a new absorption; superimposed on the bleach it appears at 1625 cm-1. Qualitatively, the same behavior is seen in the 3-13C and 1-13C labeled protein (Figure 11b,c). The isotope independence rules out an assignment to the A* CdO mode, although that is expected to contribute in this region of the spectrum for the wt sample. A plausible assignment for the 1615 cm-1 transient absorption is the A* mode of the CdC stretch, but no corresponding mode in the chromophore exists to account for the growing absorption at 1625 cm-1. Thus, we provisionally assign this band to a protein mode, presumably a carbonyl or amide I vibration which is sufficiently close to the chromophore to be

22016 J. Phys. Chem. B, Vol. 110, No. 43, 2006 influenced by the H-bond rearrangement which accompanies proton transfer. The remaining transient to be assigned is around 1660 cm-1. This sigmoid-shaped band appears promptly on electronic excitation and persists throughout the 200-ps observation period. The mode frequency is independent of isotopic substitution. Neither model studies nor DFT calculations predict any modes between the CdO and CdC stretches of HBDI. Thus, we also assign this transition to a protein mode, presumably a carbonyl. The polarization dependence of this mode is interesting, its amplitude being greatly enhanced for parallel orientation of pump and probe (Figure 5). This suggests a mode with its transition moment pointing along the direction of the electronic transition moment of the chromophore. The prompt appearance of this mode is significant, suggesting a protein mode that is perturbed (actually shifted to lower frequency) immediately upon electronic excitation. One possible explanation of such an effect of chromophore excitation on protein vibration before the proton transfer occurs is a Stark shift due to the changed dipole moment of the A* state, relative to A.46 The change in dipole moment will cause a change in the electric field strength experienced by the vibrational mode. If the oscillator is sufficiently close to the chromophore, this may give rise to a shift in vibrational frequency. Such shifts have been observed in the photosynthetic reaction center and on photolysis of disulfide linkages.47,48 However, the I* state is reported to have an even larger permanent dipole moment than the A* state,49 which then predicts an increasing shift with time after excitation. In fact, no significant temporal evolution is observed in either the amplitude or separation of the lobes in this sigmoidal line shape. Thus, a more plausible explanation is that a hydrogen-bonding interaction with the chromophore is perturbed simply by electronic excitation, and this perturbation is sufficient to cause a change in the vibrational frequency of a carbonyl mode. We will return to the potential mechanistic significance of this after considering the measured kinetics. In Figure 12, the kinetics of the spectral changes at particular wavenumbers are shown. In Figure 12a, the actual data are shown to indicate that in some cases the variation in OD is very small (0.2 mOD in the smallest case). In Figure 12b, the data have been normalized by the maximum absorbance change to allow a comparison, and the data are fit to a simple singleexponential function, with the results presented in Table 2. In fact, there is ample evidence from time-resolved fluorescence and optical transient absorption that the proton-transfer reaction has a nonexponential time course,23,24,50 but the TRIR data are of insufficient signal-to-noise to permit a meaningful multicomponent analysis. Thus, the time constants reported should be taken as mean values of the overall nonexponential kinetics. As in our earlier article, we find remarkable agreement between the kinetics associated with the increasing amplitude in the bleach around 1565 cm-1 and the rise of the transient at 1712 cm-1. This is entirely consistent with these changes reflecting the carboxylate to carboxylic acid transformation of E222. We do not find any evidence for transients faster than this reaction (i.e., there is no evidence for intermediates in the proton-transfer reaction). In fact, there are suggestions in the data that the complex relaxation seen around 1620 cm-1 may be somewhat slower than the proton relay reaction. The changes were in part ascribed to the temporal evolution of vibrational modes of the protein. It is possible that these changes reflect the relaxation of the protein environment to accommodate the new charge distribution following proton transfer. That such a relaxation is in response to protein charge rather than chro-

Stoner-Ma et al.

Figure 12. Kinetic traces of the main transient bands in wtGFP. Band positions are given in wavenumber. (a) Absolute change in delta OD as a function of time. (b) Same data normalized by the total magnitude of the absorbance change.

TABLE 2: Average Relaxation Times for wtGFP and Its Isotopes (from Fits to the Data in Figure 11)a wavenumber

wtGFP

3-13C GFP

1-13C GFP

1710 1626 1623 1615 1560

37 49 56 64 34

36 83 62 33 31

25 33 45 48 26

a

Estimated accuracy ) (20%.

mophore excitation is suggested by the observation of a timeindependent I* fluorescence emission.26 However, the differences in the kinetics are small compared to the error, and the more significant conclusion is the remarkable uniformity of the kinetics. Finally, we can discuss the mechanism of the proton-transfer reaction in light of the preceding data. We conclude that the final step on the time scale of 200 ps is protonation of E222, giving rise to a carboxylate to carboxylic acid transformation. There is no evidence for any resolvable kinetic intermediates, which suggests that, once triggered, the proton relay between the phenolic OH of the chromophore via a water molecule and S205 to E222 occurs on an ultrafast (subpicosecond) time scale. The rate-limiting step is the trigger. Previously, we proposed that the slow step was initial transfer from the chromophore to the water molecule.28 This is consistent with the fluorescence decay of A* being essentially indistinguishable from the E222 protonation. However, if the relay occurs in a concerted rather than stepwise fashion, any initial step could trigger the proton relay and simultaneously the A* quenching. The rationale for preferring the initial step as rate determining is that HBDI itself does not appear to exhibit proton transfer, suggesting a barrier to the reaction. However, it has been pointed out elsewhere that incorporation of HBDI into the protein causes dramatic changes in electronic and (see above) vibrational structures. Such strong

Proton Relay Reaction in GFP interactions may cause the suppression of the barrier to the proton-transfer reaction; indeed, suppression of the barrier may be the point of the hydrogen-bonding structure around the chromophore (Figure 2). In that case, any step may act as a trigger for a concerted proton relay reaction, as has in fact been proposed in some recent quantum chemical calculations.31,32 In this context, it is interesting to note that electronic excitation of A* perturbs the vibrational spectrum of the protein (notably around 1630 cm-1). These prompt photoinduced changes in the protein may themselves be sufficient to act as a trigger for the relay reaction. Conclusion Isotopic substitution and polarization anisotropy experiments have been employed to unravel the complex time-resolved ultrafast vibrational spectroscopy of GFP. Most of the transition frequencies observed have been assigned. A number of instantaneous and long-lived bleaches, which were shown to arise from the ground-state modes of the neutral chromophore of GFP, have been assigned. The intensity distribution among the vibrational modes was seen to be significantly different between the model compound and wtGFP, suggesting that strong specific interactions between the protein and chromophore perturb the vibrational spectra. An excited-state vibrational mode of the A* state was identified. Isotopic substitution showed that this mode includes a major contribution from the carbonyl stretch and that there is a large decrease in frequency between ground and excited electronic states. This suggested a change in electronic structure in the excited state, possibly associated with a more quinoidal form. In a previous publication, the excited-state proton-transfer reaction in GFP was followed in the time domain by the growth of a transient band at 1712 cm-1. The present data are entirely consistent with the assignment of this band to protonation of the E222 residue,28,29 via the proton wire illustrated in Figure 2. Detailed analysis allowed isolation of the disappearance of the anionic form of this residue. No intermediate states in the proton relay were observed, consistent with a concerted proton motion along the wire. Additional features of the transient vibrational spectra were observed which were assigned to a prompt modification of the vibrational structure of the surrounding protein matrix upon excitation of the chromophore. It was suggested that this is indicative of strong coupling of the electronic excitation to the protein structure. We speculated that this coupling may act as the trigger for the proton-transfer reaction. Acknowledgment. We express our gratitude to Dr. Mike Towrie of the CLF for his considerable assistance with the measurements. S.R.M. is grateful for support from EPSRC and CCLRC. J.N. thanks the Leverhulme Foundation for a research fellowship. This work was supported by NIH Grant GM66818 to P.J.T. Supporting Information Available: Displacements and orientation of dipole transition moments for CdO, CdC, CdN, phenyl modes 1-3, and E222. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Tsien, R. Y. Annu. ReV. Biochem. 1998, 67, 509. (2) Miyawaki, A.; Llopis, J.; Heim, R.; McCaffery, J. M.; Adams, J. A.; Ikura, M.; Tsien, R. Y. Nature (London) 1997, 388, 882.

J. Phys. Chem. B, Vol. 110, No. 43, 2006 22017 (3) Kneen, M.; Farinas, J.; Li, Y.; Verkman, A. S. Biophys. J. 1998, 74, 1591. (4) Elsliger, M. A.; Wachter, R. M.; Hanson, G. T.; Kallio, K.; Remington, S. J. Biochemistry 1999, 38, 5296. (5) Miyawaki, A.; Griesbeck, O.; Heim, R.; Tsien, R. Y. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 2135. (6) Truong, K.; Sawano, A.; Mizuno, H.; Hama, H.; Tong, K. I.; Mal, T. K.; Miyawaki, A.; Ikura, M. Nat. Struct. Mol. Biol. 2001, 8, 1069. (7) Nakai, J.; Ohkura, M.; Imoto, K. Nat. Biotechnol. 2001, 19, 137. (8) Shimomura, O. FEBS Lett. 1979, 104, 220. (9) Cody, C. W.; Prasher, D. C.; Westler, W. M.; Prendergast, F. G.; Ward, W. W. Biochemistry 1993, 32, 1212. (10) Reid, B. G.; Flynn, G. C. Biochemistry 1997, 36, 6786. (11) Niwa, H.; Inouye, S.; Hirano, T.; Matsuno, T.; Kojima, S.; Kubota, M.; Ohashi, M.; Tsuji, F. I. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 13617. (12) Litvinenko, K. L.; Webber, N. M.; Meech, S. R. J. Phys. Chem. A 2003, 107, 2616. (13) Webber, N. M.; Litvinenko, K. L.; Meech, S. R. J. Phys. Chem. B 2001, 105, 8036. (14) Mandal, D.; Tahara, T.; Meech, S. R. J. Phys. Chem. B 2004, 108, 1102. (15) Weber, W.; Helms, V.; McCammon, J. A.; Langhoff, P. W. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6177. (16) Voityuk, A. A.; Michel-Beyerle, M.; Rosch, N. Chem. Phys. 1998, 231, 13. (17) Martin, M. E.; Negri, F.; Olivucci, M. J. Am. Chem. Soc. 2004, 126, 5452. (18) Ormo, M.; Cubitt, A. B.; Kallio, K.; Gross, L. A.; Tsien, R. Y.; Remington, S. J. Science 1996, 273, 1392. (19) Yang, F.; Moss, L. G.; Phillips, G. N. Nat. Biotechnol. 1996, 14, 1246. (20) Brejc, K.; Sixma, T. K.; Kitts, P. A.; Kain, S. R.; Tsien, R. Y.; Ormo, M.; Remington, S. J. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 2306. (21) Wachter, R. M.; King, B. A.; Heim, R.; Kallio, K.; Tsien, R. Y.; Boxer, S. G.; Remington, S. J. Biochemistry 1997, 36, 9759. (22) Wachter, R. M.; Elsliger, M. A.; Kallio, K.; Hanson, G. T.; Remington, S. J. Structure 1998, 6, 1267. (23) Chattoraj, M.; King, B. A.; Bublitz, G. U.; Boxer, S. G. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 8362. (24) Lossau, H.; Kummer, A.; Heinecke, R.; Po¨llinger-Dammer, F.; Kompa, C.; Bieser, G.; Jonsson, T.; Silva, C. M.; Yang, M. M.; Youvan, D. C.; Michel-Beyerle, M. E. Chem. Phys. 1996, 213, 1. (25) Ward, W. W.; Prentice, H. J.; Roth, A. F.; Cody, C. W.; Reeves, S. C. Photochem. Photobiol. 1982, 35, 803. (26) Jaye, A. A.; Stoner-Ma, D.; Matousek, P.; Towrie, M.; Tonge, P. J.; Meech, S. R. Photochem. Photobiol. 2006, 82, 373. (27) Kennis, J. T. M.; Larsen, D. S.; van Stokkum, N. H. M.; Vengris, M.; van Thor, J. J.; van Grondelle, R. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 17988. (28) Stoner-Ma, D.; Jaye, A. A.; Matousek, P.; Towrie, M.; Meech, S. R.; Tonge, P. J. J. Am. Chem. Soc. 2005, 127, 2864. (29) van Thor, J. J.; Zanetti, G.; Ronayne, K. L.; Towrie, M. J. Phys. Chem. B 2005, 109, 16099. (30) Lill, M. A.; Helms, V. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 2778. (31) Wang, S.; Smith, S. C. J. Phys. Chem. B 2006, 110, 5084. (32) Vendrell, O.; Gelabert, R.; Moreno, M.; Lluch, J. M. J. Am. Chem. Soc. 2006, 128, 3564. (33) Parsons, J. F.; Armstrong, R. N. J. Am. Chem. Soc. 1996, 118, 2295. (34) He, X. A.; Bell, A. F.; Tonge, P. J. J. Phys. Chem. B 2002, 106, 6056. (35) Towrie, M.; Grills, D. C.; Dyer, J.; Weinstein, J. A.; Matousek, P.; Barton, R.; Bailey, P. D.; Subramaniam, N.; Kwok, W. M.; Ma, C. S.; Phillips, D.; Parker, A. W.; George, M. W. Appl. Spectrosc. 2003, 57, 367. (36) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; AlLaham, M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 98, revision A.7; Gaussian, Inc.: Pittsburgh, PA, 1998. (37) Esposito, A. P.; Schellenberg, P.; Parson, W. W.; Reid, P. J. J. Mol. Struct. 2001, 569, 25. (38) Bell, A. F.; He, X.; Wachter, R. M.; Tonge, P. J. Biochemistry 2000, 39, 4423. (39) Schellenberg, P.; Johnson, E.; Esposito, A. P.; Reid, P. J.; Parson, W. W. J. Phys. Chem. B 2001, 105, 5316.

22018 J. Phys. Chem. B, Vol. 110, No. 43, 2006 (40) Rubtsov, I. V.; Wang, J.; Hochstrasser, R. M. J. Phys. Chem. A 2003, 107, 3384. (41) Heyne, K.; Mohammed, O. F.; Usman, A.; Dreyer, J.; Nibbering, E. T. J.; Cusanovich, M. A. J. Am. Chem. Soc. 2005, 127, 18100. (42) Usman, A.; Mohammed, O. F.; Nibbering, E. T. J.; Dong, J.; Solntsev, K. M.; Tolbert, L. M. J. Am. Chem. Soc. 2005, 127, 11214. (43) Wiehler, J.; Jung, G.; Seebacher, C.; Zumbusch, A.; Steipe, B. ChemBioChem 2003, 4, 1164. (44) Tonge, P. J.; Moore, G. R.; Wharton, C. W. Biochem. J. 1989, 258, 599.

Stoner-Ma et al. (45) Laino, T.; Nifosi, R.; Tozzini, V. Chem. Phys. 2004, 298, 17. (46) Rosell, F. I.; Boxer, S. G. Biochemistry 2003, 42, 177. (47) Volk, M.; Gilch, P.; Kompa, C.; Haselsberger, R.; Ha¨rter, P.; Sto¨ckl, M.; Schere, W.; Latzel, K.; Michel-Beyerle, M.-E. J. Phys. Chem. A 2000, 104, 4984. (48) Volk, M.; Kholodenko, Y.; Lu, H. S. M.; Gooding, E. A.; DeGrado, W. F.; Hochstrasser, R. M. J. Phys. Chem. B 1997, 101, 8607. (49) Bublitz, G.; King, B. A.; Boxer, S. G. J. Am. Chem. Soc. 1998, 120, 9370. (50) Winkler, K. Phys. Chem. Chem. Phys. 2002, 4, 1072.