11338
J. Phys. Chem. B 2010, 114, 11338–11352
The Protonation State of Glu181 in Rhodopsin Revisited: Interpretation of Experimental Data on the Basis of QM/MM Calculations Jan S. Fra¨hmcke,† Marius Wanko,‡ Prasad Phatak,† Maria Andrea Mroginski,§ and Marcus Elstner*,†,| Institute for Physical and Theoretical Chemistry, TU Braunschweig, Hans-Sommer-Str. 10, D-38106 Braunschweig, Germany, Nano-Bio Spectroscopy group and ETSF Scientific DeVelopment Centre, Dpto. Fı´sica de Materiales, UniVersidad del Paı`s Vasco, Centro de Fı´sica de Materiales CSIC-UPV/EHU, DIPC, and CIC nanoGUNE, AV. Tolosa 72, E-20018 San Sebastia´n, Spain, Max-Volmer-Laboratory for Biophysical Chemistry, TU Berlin, Strasse des 17 Juni 135, D-10623 Berlin, Germany, and Institute of Physical Chemistry, Karlsruhe Institute of Technology, Kaiserstrasse 12, D-76131 Karlsruhe, Germany ReceiVed: May 18, 2010; ReVised Manuscript ReceiVed: July 27, 2010
The structure and spectroscopy of rhodopsin have been intensely studied in the past decade both experimentally and theoretically; however, important issues still remain unresolved. Of central interest is the protonation state of Glu181, where controversial and contradictory experimental evidence has appeared. While FTIR measurements indicate this residue to be unprotonated, preresonance Raman and UV-vis spectra have been interpreted in favor of a protonated Glu181. Previous computational approaches were not able to resolve this issue, providing contradicting data as well. Here, we perform hybrid QM/ MM calculations using DFT methods for the electronic ground state, MRCI methods for the electronically excited states, and a polarization model for the MM part in order to investigate this issue systematically. We constructed various active-site models for protonated as well as unprotonated Glu181, which were evaluated by computing NMR, IR, Raman, and UV-vis spectroscopic data. The resulting differences in the UV-vis and Raman spectra between protonated and unprotonated models are very subtle, which has two major consequences. First, the common interpretation of prior Raman and UV-vis experiments in favor of a neutral Glu181 appears questionable, as it is based on the assumption that a charge at the Glu181 location would have a sizable impact. Second, also theoretical results should be interpreted with care. Spectroscopic differences between the structural models must be related to modeling uncertainties and intrinsic methodological errors. Despite a detailed comparison of various rhodopsins and mutants and consistently favorite results with charged Glu181 models, we find merely weak evidence from UV-vis and Raman calculations. On the contrary, difference FTIR and NMR chemical shift measurements on Rh mutants are indicative of the protonation state of Glu181. Supported by our results, they provide strong and independent evidence for a charged Glu181. Introduction Retinal proteins play an important role in different biochemical processes, such as vision (rhodopsin, cone pigments), bioenergetics (bacteriorhodopsin), and phototaxis (pharaonis sensory rhodopsin, channelrhodopsins). Rhodopsin (Rh) consists of seven transmembrane helices (opsin) and a retinal chromophore, which is covalently bound to a lysine residue via a protonated Schiff base (PSB). The PSB of retinal forms a lightabsorbing polyene of six conjugated double bonds, terminated by a β-ionone ring. Rhodopsin, a retinal protein responsible for dim light vision, is a prototype for class-A G-protein coupled receptors. Its ligand, retinal, is in an 11-cis configuration and embedded in a binding pocket built by ≈30 amino acids at the extracellular side of the helical region. The positive charge of the PSB is balanced by a counterion (Glu113).1-3 Other * To whom correspondence should
[email protected]. † TU Braunschweig. ‡ Universidad del Paı`s Vasco. § TU Berlin. | Karlsruhe Institute of Technology.
be
addressed.
E-mail:
amino acids in the binding pocket are important for the efficient photoisomerization or formation of the PSB-counterion complex. Light absorption in the dark state starts the primary photoreaction, an 11-cis-all-trans isomerization of retinal, which is followed by a conformational relaxation of the protein and transducin activation. This relaxation proceeds stepwise, forming several spectroscopically distinguishable intermediates with different lifetimes: photo (ps), batho (ns), lumi (µs), and meta I (ms). Meta I is in conformational equilibrium with meta II, which activates transducin.4-6 The bovine Rh dark-state structure is well characterized due to recently published high-resolution crystal structures7-12 except for the protonation state of active site Glu181, which still is a matter of debate. This residue plays an important role in the meta I to meta II transition, where the two proposed models assume different protonation states. The so-called counterionswitch mechanism13 assumes a protonated Glu181 in the dark state that transfers its proton to Glu113 during transition to meta I, where it becomes the new counterion of the PSB. This model is based on two observations: (1) In titration experiments of Rh and site-directed mutants,13-15 the primary counterion of the PSB was identified as Glu113 in the dark state and Glu181 under
10.1021/jp104537w 2010 American Chemical Society Published on Web 08/10/2010
The Protonation State of Glu181 in Rhodopsin meta I-like conditions. When one of these residues is replaced, the pKa of the Schiff base drops dramatically in the corresponding state. Hence, Glu113 must be charged in the dark state and Glu181 in meta I. (2) As discussed below, preresonance Raman and UV-vis spectra of the dark state appear more consistent with a protonated Glu181. On the basis of Fourier transform infrared (FTIR) measurements, Lu¨deke et al.16 suggested an alternative complexcounterion mechanism. They proposed both Glu113 and Glu181 to be unprotonated in the dark state, forming a complex counterion for the PSB. The conformational change toward meta I brings the PSB close to Glu181, which then becomes the predominant counterion. Glu113 is protonated by the PSB in the transition to meta II. As the formation of meta I and meta II depends on the initial protonation state of Glu181, it is important to understand not only the structure but also the activation mechanism of Rh. Over the years, this issue has been investigated using different spectroscopic techniques and theoretical methods, leading to conflicting findings with respect to the Glu181 protonation state, which shall be reviewed briefly here. The first attempts to gain insight into the structure of retinal’s binding site were based on optical and NMR spectroscopy. In 1985, Honig et al.17 suggested a series of simple point-charge models for the Rh binding site. Their favored model18 places one negative charge close to the PSB and another between C12 and C14 and thus reconciles the observed strong downfield shift of the 13C14 NMR peak in Rh19 with the assumption of a protonated Schiff base. Honig et al. also observed an enhanced opsin shift for 11,12-dihydroretinals in Rh, supporting the presence of a negative (partial) charge close to C12/C14. In 1985, Birge et al.20 measured the two-photon absorption spectrum of 11-cis-locked Rh and attempted to reproduce the one- and two-photon data in semiempirical CI calculations (INDO/PSDCI). They obtained the best results using a single counterion placed below the polyene between C13 and C15. They also claimed that, by introducing a second negatively charged group in the binding site, e.g., at the positions suggested by Honig et al.,18 the energy gap between the S1 and the onephoton forbidden S2 state would increase from 0.44 to more than 0.93 eV, whereas the measured gap is ca. 0.25 eV. Hence, they concluded that the binding site must be neutral. As is known from the X-ray structures, Glu181’s carboxyl group is located ca. 6 Å away from retinal’s backbone (distance from Glu181-Cδ to C12), whereas the point-charge models tested positions at 3 Å separation and neglected screening effects of surrounding polar groups.17,20 Nonetheless, this work is often interpreted to provide experimental evidence for a protonated Glu181. Also, NMR chemical shifts were used to determine the position of a single counterion by matching theoretical with experimental data.21-23 The resulting position is closer to C14 than to the Schiff base, suggesting the idea of a weak or delocalized counterion. In regard to the first X-ray structures, Verhoeven et al.24 remark that the structural models inferred from chemical shift constraints locate the counterion somewhat farther away from the PSB than indicated by the X-ray structures. This discrepancy was interpreted in terms of a complex counterion that may involve a charged Glu181 or Tyr268 residue.25,26 In 1993, Fahmy et al.27 studied the FTIR difference spectra (dark state-meta II) of Rh and the site-directed mutants of Asp83 and Glu122. They assigned all difference bands above 1700 cm-1 to Asp83 and Glu122, which therefore must be protonated
J. Phys. Chem. B, Vol. 114, No. 34, 2010 11339 in all intermediate states, and to Glu113, which is protonated in the formation of meta II. Therefore, they concluded that Glu113 is the only membrane-embedded carboxylic acid that is protonated in meta II. At the time being, it was unknown that the extracellular loop 2 (E2), which contains Glu181, folds deeply into the helical region close to the retinal. In another FTIR study, Nagata et al.28 see no vibrational changes for the dark state to batho transition that can be assigned to a protonated Glu181. This argues against a protonated Glu181 in the dark and batho states of Rh because the strong structural changes caused by the photoisomerization would also induce changes in the vibrational frequencies of a protonated Glu181. A more recent FTIR study by Lu¨deke et al.16 considers also site-directed mutants of Asp83, Glu122, and Glu181. In the difference spectra of dark state to meta I and dark state to meta II, peaks were measured at 1767 and 1735/1727 cm-1, which were assigned to protonated Asp83 and Glu122 via D83N and E122Q mutants. The double mutant D83N/E112Q shows a completely flat region above 1700 cm-1 in the dark state-meta I difference spectrum, whereas all peaks in this region from the WT spectrum are unchanged in the E181Q and E181M mutants. In the transition to meta II, one positive peak appears at 1713 cm-1. This peak must stem from the protonation event of Glu113. As no deprotonation of Glu181 can be detected and Glu181 acts as the principal counterion of the PSB in meta I, Lu¨deke et al. concluded that Glu181 must be unprotonated (charged) in all states. Mutation induced changes of protonation states were not discussed in ref 16 Yan et al.13 compared the preresonance Raman spectra of wild-type (WT) Rh and the E181Q mutant. The modes in the fingerprint, hydrogen out-of-plane (HOOP), and CdC-stretching regions were found to be conserved within a few cm-1. Also, the UV-vis absorption of Rh in the dark state is hardly altered by mutation of Glu181.14,15,29 As the substitution of a negatively charged Glu181 by a neutral Gln is expected to induce significant changes in the vibrational and UV-vis spectra, the authors concluded that Glu181 must be neutral. On the other hand, the 10 nm bathochromic shift in the E181Q mutant is removed in the presence of 140 mM NaCl.14 In retinochrome, where Glu181 serves as the primary counterion, a chloride ion has been shown to substitute the negative charge of Glu181 in the E181Q mutant, stabilizing the protonated chromophore. Hence, it appears plausible that, also in the bovine E181Q mutant, chloride can be stabilized by Gln181, without significant changes of the spectrum. Apart from the spectroscopic data, the analysis of rhodopsin’s reactivity with hydroxylamine may yield additional evidence on the matter. In all Glu181 mutants (except Asp181), the PSB shows a drastic increase in the hydroxylamine reactivity15 with respect to WT Rh. The accessibility of hydroxylamine to the Schiff base might depend on the stability of the hydrogenbonded network (HBN) that connects the loop E2 residues (including Glu181) with the remainder of the binding site, as the penetration of hydroxylamine might require the transient breaking of several hydrogen bonds of the HBN. Lewis et al.30 studied the effect of Glu181 mutation on the early photointermediates. They found a reduced stability of the batho intermediate for mutants other than E181D, indicating a less stable HBN in the binding pocket. Moreover, the blue shift in the blueshifted intermediate (BSI) is enhanced in the E181D mutant but reduced in the other mutants. In the E181Q mutant, this bathochromic mutation shift is partially undone in the presence of chloride, like in the case of the dark state. Assuming that the carboxyl group in E181D is shifted closer to the PSB, Lewis
11340
J. Phys. Chem. B, Vol. 114, No. 34, 2010
et al. noted that a charged Glu181 or Asp181 would explain their results. The different reactivities and batho lifetimes can thus be explained by (a) the reduced stability of the HBN between Glu181 and Glu113 due to the replacement of a charged Glu181 (or Asp181) by a neutral amino acid or (b) the changes in the HBN caused by the different hydrogen-bonding character of protonated carboxylic acids and, e.g., glutamine.30 A difference in HBN stability was indeed found in a theoretical MD study by Ro¨hrig et al.31 Assuming a protonated Glu181, they performed classical MD simulations of Rh in an explicit solvent and membrane environment and experienced a disruption of the HBN after 630 ps. Starting from an unprotonated Glu181, their HBN was stable during a 7 ns simulation. However, another classical MD simulation over 1.5 µs did not report such instabilities in either case.32,33 Janz et al.34 studied mutants of six amino acids in the HBN. All of them have a drastic effect on the energetics of Schiff base hydrolysis (flattening of the convex Arrhenius plot of WT Rh). This is attributed to a weakening of the HBN. Four of the mutants (T94I, S186A, Y192F, Y268F) clearly have a smaller HBN because of a missing hydroxyl group. Also, for the counterion mutant E113Q, this observed effect is expected. For the E181Q mutant, this result is surprising and indicates a strong disturbance of the HBN in this mutant, too. Previous quantum-chemical studies on Rh investigated optical and other spectroscopic properties with Glu181 modeled protonated,35-48 and several studies tested both protonation states31-33,49-53 but came to different conclusions. They are reviewed in the following. In an early QM/MM study, Schreiber et al.49 calculated the effect of a protonated or unprotonated Glu181 on the excitation energy of a reduced model chromophore in Vacuo. They obtained a surprisingly small shift (0.12 eV), and explained this with the position of Glu181 close to the center of the chromophore. They implied that the argument of Yan et al.15 for a protonated Glu181 (small mutation effect) does not hold. In classical molecular dynamics (MD) simulations,32,33 Martinez-Mayorga et al. simulated the batho to meta I transition, starting from either a protonated or an unprotonated Glu181. In the latter case, the predominant counterion spontaneously changes from Glu113 to Glu181 during 1000 ns of unconstrained MD simulation. This corroborates the model suggested by Lu¨deke et al., as it shows that such a switch is possible without any proton transfer. They also reported a better agreement between simulated and measured 2H NMR spectra54 for the meta I state obtained with the charged Glu181 in the setup, but they did not publish any details of the two meta I structures. Grossfield et al.33 reanalyzed these trajectories and found in both cases an increase in the number of water molecules inside the protein during meta I formation, consistent with 1H MAS NMR results. The MD study did not investigate the time scale or energetics of the proton transfer; the latter was simply established after 500 ns of simulation. Therefore, the authors consider the complex-counterion model more likely but do not exclude the proton-transfer-triggered counterion switch. In a recent QM/MM study, Hall et al.50 considered both protonation states and the E181Q mutant. They found differences in chromophore geometry, TDDFT excitation energies, and NMR chemical shifts. They reported a slightly better agreement in BLA for the protonated Glu181 model compared to NMR distance measurements. The bond length differences between the models (
