Proton transfer in the photocycle of the photoactive yellow protein

Promotion on 12 September 2008

PhD promotion of
Elske Leenders
Prof. Dr. P.G. Bolhuis
10:00, 12 September 2008
Agnietenkapel, UvA, Amsterdam

Summary of thesis

The photoactive yellow protein (PYP) is a widely used model protein. It is a typical, easy to study, example of photoactive and other signalling proteins. For this reason, PYP has been studied extensively using experiments and simulations. PYP is triggered by UV light: this starts the photocycle, in which the protein changes its shape to transduct the signal to its bacterial host. The photocycle consists of several steps and starts with the excitation and isomerisation of the chromophore. This chromophore, p-coumaric acid (pCA), is attached to the protein through the cysteine residue at position 69 via a thioester bond. When the protein is at rest in its ground state (pG), pCA is in the trans configuration. It is deproto- nated; its negative charge is stabilised inside the protein by hydrogen bonds donated by tyrosine at position 42 and glutamic acid at position 46. A positively charged arginine at position 52 is in plane with the chromophore ring and provides extra stabilisation as well.

After the excitation by UV light, pCA isomerises, while maintaining the hydrogen bonds with Tyr42 and Glu46. This new configuration has a red-shifted absorption peak in UV-vis spectroscopy and is hence called the pR state. The photocycle continues to the pB (blue- shifted) state. In this state, pCA is protonated, Glu46 is deprotonated and the protein is partially unfolded. This state is considered the signalling state of PYP. The pG to pR step is very fast (picoseconds) and local. The pR to pB step takes microseconds, as more extended configurational and chemical changes occur. This step is not as well understood. The initial stage of this process, particularly the proton transfer to pCA, is the topic of this thesis.

The photocycle of PYP finishes with the last millisecond step in which the protein folds back from the pB state into its ground state. Experimentalists still debate on the nature of the proton donor in the pR-to-pB step and on whether it donates the hydrogen directly or indirectly. To get better knowledge of the mechanism, I studied this proton transfer using Car-Parrinello molecular dynamics, classical molecular dynamics (with the Gromos96 force field) and computer simulations combining these two methods (quantum mechanics / molecular mechanics, QMMM). The simulations reproduce the chromophore structure and hydrogen bond network of the pro- tein measured by X-ray crystallography and NMR well. When the chromophore is protonated, it leaves the assumed proton donor, Glu46, with a negative charge in a hydrophobic environment.

I showed that the stabilisation of this charge is a very important factor in the mechanism of protonation. Protonation frequently occurs in simplified ab initio simula- tions of the chromophore binding pocket in vacuum, where amino acids can easily hydro- gen bond to Glu46. When the complete protein environment is incorporated in a QMMM simulation of the complete protein, no proton transfer is observed within 14 ps. The hydrogen bond rearrangements in this time span are not sufficient to stabilise the new pro- tonation state. Force field molecular dynamics simulations on a longer time scale showed 130 Summary which internal rearrangements of the protein are needed.

Combining these simulations with more QMMM calculations enabled me to check the stability of protonation states and clarify the initial requirements for the proton transfer in PYP. I found that Glu46 needs to accept at least three hydrogen bonds to stabilise its negative charge inside the hydrophobic protein environment after a proton transfer to pCA. These new hydrogen bonds are made available by internal hydrogen bond rearrangements inside the protein. These rearrangements can be the onset to further structural changes towards the pB state. They occurred faster (within picoseconds) when the protein was simulated with a force field than with the QMMM method.

To study the important hydrogen bond changes in more detail, I looked at two model systems: pCA in water and Glu in water. For both cases, these molecules had not been studied with dynamical quantum simulations before. Experiments and static simulations of these systems focused on spectroscopic properties, whereas I studied structural and dy- namical features that are important to the hydrogen bonds around pCA and Glu. I studied four different configurations of pCA solvated in water with Born-Oppenheimer molecular dynamics simulations. I researched the influence of the protonation and isomerisation state of pCA on its hydrogen bonding properties and on the Mulliken charges of the atoms in the simulation. The chromophore’s isomerisation state influenced the hydrogen bonding less than its protonation state did. In general, more charge yielded a higher hydrogen bond coordination number.

Whereas deprotonation increased both the coordination number and the residence time of the water molecules around the chromophore, protonation showed a somewhat lower coordination number on two of the three pCA oxygens, but much higher residence times on all of them. This could be explained by the increased polarisation of the OH groups of the molecule. The presence of the chromophore also influenced the charge and polarisation of the water molecules around it. This effect was different in the four systems studied, and mainly localised in the first solvation shell.

I also performed a proton transfer reaction from hydronium via various other water molecules to the chromophore. In this small charge-separated system, the protonation occurred within 6.5 ps. I identified the transition state for the final step in this protonation series. In this configuration, the hydronium molecule had two elongated OH bonds. I also conducted molecular dynamics simulations on glutamic acid and glutamate sol- vated in water, using both density functional theory (DFT) and the Gromos96 force field. I focused on the microscopic aspects of the solvation - particularly on the hydrogen bond structures and dynamics - and investigated the influence of the protonation state and of the simulation method.

Radial distribution functions showed that the hydrogen bonds are longer in the force field systems. I found that the partial charges of the solutes in the force field simulations are lower than the localised electron densities for the quantum simula- tions. This lower polarisation decreases the hydrogen bond strength. Protonation of the carboxylate group renders glutamic acid a very strong and stable hydrogen bond donor. The donated hydrogen bond is shorter and lives longer than any of the other hydrogen bonds. The solute molecules simulated by the force field accept on average three hydro- gen bonds more than their quantum counterparts do. The life times of these bonds show the opposite result: the residence times are much longer (up to a factor 4) in the ab initio simulations.

In relation to the simulation of the proton transfer in the photoactive yellow protein, it is interesting to see that the residence time changes upon protonation in Glu are larger 131 than the residence time differences upon protonation of the phenolic oxygen of pCA. This means that deprotonation of glutamic acid has a larger residence time decreasing effect than the opposite effect on the chromophore, and hence, that it is likely that the hydro- gen bond between the two molecules in the protein breaks easier after the proton transfer. This indicates that a direct proton transfer from Glu to pCA can be a starting point for the breaking of hydrogen bonds and ultimately unfolding of the protein. I also saw that upon deprotonation of Glu, glutamate attracts three extra hydrogen bonds to the now negatively charged carboxylate group - independent of the simulation method. In force field simula- tions of glutamate in the protein as well, the number of hydrogen bonds increased from 1 to 4 in less than 10 ps. The need to stabilise the negative charge on Glu makes hydrogen bond rearrangements easier. Exactly these rearrangements are part of the process of unfolding the protein into its signalling state.

Possible pathways for the protonation reaction in the PYP photocycle are those in which Glu46, the likely proton donor, receives at least two hydrogen bonds before deprotonation. A third hydrogen bond can then be donated to Glu46 as a result of - or simultaneously with - the deprotonation. Two likely candidates are pathways in which extra stabilisation is provided by water or by a hydrogen bond from Thr50. This lowers the free energy barrier for the proton transfer by more than 20 kBT. I studied these possible pathways with dynamical QMMM (quantum mechanics / molecular mechanics) simulations. To speed up the processes and to estimate the free energy barriers involved, I used direct metadynamics. I found that a proton transfer directly after the excitation and isomerisation of the PYP chromophore has such a high barrier, that this is an unlikely reaction path. The same accounts for a protonation of pCA when it is solvated. But when in the pR state a water molecule or Thr50 donates a hydrogen bond to Glu46 (which happens spontaneously within nanoseconds at room temperature), only one extra hydrogen bond from Tyr42 is needed to make the new protonation state stable. The barrier for the Tyr42 hydrogen bond switch plus the proton transfer from Glu46 to pCA is then low enough to occur within only a couple of nanoseconds. This is well within the time measured experimentally for the pR-to-pB transition.