1

Femtosecond Structural Dynamics Drives the Trans/Cis Isomerization in Photoactive

Yellow Protein

Kanupriya Pande1,2, Christopher D. M. Hutchison3, Gerrit Groenhof4, Andy Aquila5, Josef S.

Robinson5, Jason Tenboer1, Shibom Basu6, Sébastien Boutet5, Daniel P. DePonte5, Mengning

Liang5, Thomas A. White2, Nadia A. Zatsepin7, Oleksandr Yefanov2, Dmitry Morozov4,

Dominik Oberthuer2, Cornelius Gati2, Ganesh Subramanian7, Daniel James7, Yun Zhao7, Jake

Koralek5, Jennifer Brayshaw1, Christopher Kupitz1, Chelsie Conrad6, Shatabdi Roy-Chowdhury6,

Jesse D. Coe6, Markus Metz2, Paulraj Lourdu Xavier2,8, Thomas D. Grant9, Jason E. Koglin5,

Gihan Ketawala6, Raimund Fromme6, Vukica Šrajer10, Robert Henning10, John C. H. Spence7,

Abbas Ourmazd1, Peter Schwander1, Uwe Weierstall7, Matthias Frank11, Petra Fromme6, Anton

Barty2, Henry N. Chapman2,12, Keith Moffat10,13, Jasper J. van Thor3, Marius Schmidt1*

1 Physics Department, University of Wisconsin-Milwaukee, Milwaukee, Wi 53211, USA 2 Center for Free Electron Laser Science, Deutsches Elektronen Synchrotron DESY, Notkestrasse 85, 22607 Hamburg, Germany. 3 Faculty of Natural Sciences, Life Sciences, Imperial College, London SW7 2AZ, UK. 4 Nanoscience Center and Department of Chemistry, University of Jyväskylä, P.O. Box 35, 40014, Jyväskylä, Finland. 5 Linac Coherent Light Source, SLAC National Accelerator Laboratory, Sand Hill Road, Menlo Park, CA 94025, USA. 6 Department of Chemistry and Biochemistry, Arizona State University, Tempe, AZ 85287, USA. 7 Department of Physics, Arizona State University, Tempe, AZ 85287, USA. 8. IMPRS-UFAST, Max Planck Institute for Structure and Dynamics of Matter, Luruper Chaussee 149, 22761 Hamburg, Germany. 9 Hauptman-Woodward Institute, State University of New York at Buffalo, 700 Ellicott Street, Buffalo, NY 14203, USA. 10 Center for Advanced Radiation Sources, University of Chicago, Chicago, IL 60637, USA. 11 Lawrence Livermore National Laboratory, Livermore, CA 94550, USA. 12 Center for Ultrafast Imaging, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany. 13 Department of Biochemistry & Molecular Biology and Institute for Biophysical Dynamics, University of Chicago, Chicago, IL 60637, USA. *Corresponding author, m-schmidt@uwm.edu

2

Abstract.

A variety of organisms have evolved mechanisms to detect and respond to light, in which the

response is mediated by protein structural changes following photon absorption. The initial step is

often the photo-isomerization of a conjugated chromophore. Isomerization occurs on ultrafast

timescales, and is substantially influenced by the chromophore environment. Here we identify

structural changes associated with the earliest steps in the trans to cis isomerization of the

chromophore in photoactive yellow protein. Femtosecond, hard X-ray pulses emitted by the Linac

Coherent Light Source were used to conduct time-resolved serial femtosecond crystallography on

PYP microcrystals over the time range from 100 femtoseconds to 3 picoseconds to determine the

structural dynamics of the photoisomerization reaction.

One Sentence Summary.

The trans to cis isomerization of the central chromophore in a protein is structurally characterized

on ultrafast time scales with TR-SFX.

3

Trans-cis isomerization constitutes a major class of chemical reactions of critical importance

to biology, where for example light-dependent isomerization of a retinal chromophore underlies

vision (1). Since isomerization occurs on the femtosecond (fs) to picosecond (ps) time scale,

ultrafast time-resolved methods are necessary to follow the reaction in real time. The spectral

response after photon absorption reveals the dynamics of the molecules involved (2-5) but does

not directly observe the associated structural changes, which have to be inferred by computational

approaches (6). Until recently it has been impossible to directly determine the structure of

molecules on ultrafast time scales. With the recent availability of hard X-ray pulses on the fs time

scale emitted by free electron laser (FEL) sources such as the Linac Coherent Light Source

(LCLS), the ultrafast fs to ps time scale has become experimentally accessible (7-11).

Photochemical reactions (12) are initiated by photon absorption, which promotes electrons into the

excited state. Thereafter, the nuclei experience - and the structure evolves on - the excited state

potential energy surface (PES) (13, 14). The shape of the surface controls the subsequent nuclear

dynamics. After returning to the ground state PES, the reaction continues and is driven thermally.

Although structures of longer-lived excited state intermediates have been characterized with ~100

ps time resolution at synchrotrons (15-19), the fs structural dynamics of ultrafast photochemical

reactions can only be investigated at an X-ray FEL (11). The photoactive yellow protein (PYP) is

an ideal macromolecular system with which to investigate ultrafast trans to cis isomerization. Its

chromophore, p-coumaric acid (pCA), can be photoexcited by absorbing a photon in the blue

region of the spectrum. Upon photon absorption PYP enters a reversible photocycle involving

numerous intermediates (Fig. 1A). The primary photochemical event that controls entry into the

photocycle is isomerization of pCA about its C2=C3 double bond (see Fig. 1B for the pCA

geometry). The pCA chromophore remains electronically excited for a few hundred fs (3, 5, 20).

Excited state dynamics is thought to drive the configurational change from trans to cis (3, 21). The

chromophore pocket within the PYP protein is sufficiently flexible to allow certain relatively large

atomic displacements, but also imposes structural constraints that may affect the pathway and

dynamics of isomerization (22, 23). In particular, the pCA chromophore is constrained by a

covalent bond to the Cys69 side chain of PYP (Fig. 1B), by unusually short hydrogen bonds

between its phenolate oxygen and nearby glutamate and tyrosine side chains (24), and by a

hydrogen bond between the carbonyl oxygen of its tail and the main chain amide of Cys69.

4

Previously, we showed that time-resolved pump-probe serial femtosecond crystallography

(TR-SFX) could be successfully carried out on PYP on the ns to microsecond (µs) time scales.

Difference electron density (DED) maps of very high quality, which compare the structures before

(dark) and after (light) absorption of a photon (25), were obtained to near-atomic (1.6 Å)

resolution. These experiments used a nanosecond (ns) laser pulse to initiate isomerization and

subsequent structural changes. An overall reaction yield as high as 40% (25) could be reached.

However, achieving fs time resolution requires that a fs pump laser pulse be used, which restricts

the reaction yield to the much lower value of the primary quantum yield (around 10%) and

correspondingly reduces the structural signal. The energy of fs pulses i.e. the number of photons

per pulse must also be limited to avoid damaging effects from their significantly higher peak

power. Here, we present results of TR-SFX experiments covering the time range from 100 fs to 3

ps. We directly follow the trans-cis isomerization of the pCA chromophore and the concomitant

structural changes in its protein environment in real time. Full details of the experiment and data

analysis are provided in the Supplementary Materials (SM). Light-initiated structural changes in

PYP were investigated at the Coherent X-ray Imaging (CXI) instrument of the LCLS (26).

Electronic excitation was initiated in microcrystals of PYP by fs pump laser pulses (λ=450 nm).

Permanent bleaching of the chromophore was avoided by limiting the laser pulse energy to 0.8

mJ/mm2 (5.7 GW/mm2). Laser pulse duration, spectral distribution and phase were characterized

by ‘Second Harmonic Generation Frequency Resolved Optical Gating’ (SHG-FROG) (27). The

pulse duration was 140±5 fs and had both positive group delay dispersion and third order

dispersion to maximize the conversion to the excited state (28). Offline spectroscopic experiments

on thin crushed crystals of PYP had established that photoexcitation with fs laser pulses under

comparable conditions could be as high as 10% without inducing damage (SM). The structural

changes induced by the laser pulse were probed with 40 fs X-ray FEL pulses at 9 keV (1.36 Å).

Both the pump-probe and the reference X-ray diffraction data were collected at the full 120 Hz

pulse repetition rate of the LCLS to a resolution of 1.6 Å and 1.5 Å, respectively. To address

concerns that the detector response might be influenced by the stray light of the intense fs laser

pulse, the reference data were collected as a negative time delay, where the fs laser pulse arrived

1 ps after the X-ray pulse.

To assess whether fs laser pulses excited a sufficiently large number of molecules under these

experimental conditions, we first performed a positive control experiment with a 200 ns pump-

5

probe time delay, where large structural differences between the light and dark states have been

well characterized (25, 29). From the pump-probe TR-SFX data and the reference data, DED maps

were calculated (SM). Fig. 1C shows that the fs laser pulses are able to initiate sufficient entry into

the photocycle to produce strong, chemically meaningful features. The 200 ns DED map is

essentially identical to maps determined earlier at both the LCLS (25) and at BioCARS (29) at a

time delay of 1 µs, and can be interpreted with the same mixture of intermediates, pR1 and pR2.

The extent of reaction initiation is 12.6 % as determined by fitting a calculated ‘pR1&pR2 minus

pG’ difference map to the 200 ns DED map, a value which agrees with the maximum extent of

excitation determined spectroscopically (7 – 10%). The fs time scale was explored by using

nominal settings for the time delay of 300 fs and 600 fs. The timing jitter between the 140 fs laser

pump and 40 fs X-ray probe pulses is ~280 fs (8). The jitter was measured for every X-ray pulse

by a timing tool (30, 31), which was combined with adjustments that take longer-term

experimental drift into account (see SM). Thus, each individual diffraction pattern was associated

with a definite “time stamp”. However, due to the drift, the time stamps were non-uniformly

distributed in time (Fig. S1). Since the quality of structure amplitudes and of the DED maps derived

from them depends on the number of diffraction patterns, indexed, time-stamped diffraction

patterns were binned into 8 different pump-probe delays with about the same number of patterns

(40,000) in each bin , spanning the time range from 100 to 1000 fs (Tab. S1B). A set of diffraction

patterns at a time delay of 3 ps was also collected. Since the jitter and drift are much smaller than

the delay, time stamping was not necessary for the 3 ps or the 200 ns delays. The values of R-split

for all datasets is 7.5 – 9.9% which indicates the high quality of the diffraction data, and results in

DED maps of comparable, good quality for all delays. Maps at 7 time delays are shown in Fig. 2.

Visual inspection of these maps reveals an important qualitative result. The features in all maps at

delays less than 500 fs are similar (compare Fig. 2, A-C); and features in all maps at delays greater

than 700 fs are also similar (compare Fig. 2, D-G), but differ from those in the first set.

Consequently, there must be a structural transition between the 455 fs and 799 fs time delays that

gives rise to the two distinct sets of features. To identify with more precision the time delay at

which this transition occurs, the time-stamped diffraction patterns were re-binned into 16 narrower

time bins with about 20,000 patterns in each bin (Tab. S1A). The resultant time series of 16 DED

maps in the fs time range (together with the map for the 3 ps time delay) were subjected to singular

value decomposition (SVD; Fig. S2B) (32). The volume occupied by the pCA chromophore, by

6

the Cys69 sulfur and the Glu46 carboxyl was included in the analysis. When a time series exhibits

a change, a corresponding change should be even more readily recognizable in the right singular

vectors (rSVs). This change is evident in the magnitude of both the first and second rSVs around

550 fs (red arrow in Fig. S2B). The substantial increase in the magnitude of the first right singular

vector after 155 fs (Fig. S2B) shows the earliest (fastest) evolution of the structure after excitation.

We tentatively associate the structural transition at around 550 fs, qualitatively evident by

inspection of the DED maps and more quantitatively in their SVD analysis, with the trans to cis

isomerization of the pCA chromophore. The transition occurs within ~180 fs (Fig. S2B), but its

exact duration needs to be further established. Rate kinetics would require that after a ~500 fs

dwell time the transition time would be stretched beyond the bandwidth limited rate. Yet the

observed transition time matches the experimental bandwidth of 3.15 THz. Therefore the ensemble

phase relation imparted by the optical pulse appears to be maintained for the duration of the dwell

time, which may be supported by coherent motion. Although no oscillatory motion was detected

in the TR-SFX data (they may be masked by the non-uniform data sampling), the time delay is

however within the vibrational dephasing time of the PYP S1 state (3) and ground state modes in

proteins (33). We further propose that at ~550 fs the system lies at or very close to a conical

intersection (20) (Fig. S8), a branch point from which molecules either continue towards the cis

configuration and enter the photocycle, or revert to the trans configuration and return to the resting

(dark) state.

To identify the isomerization, refined structures before and after the transition are required.

Initially, date in bins with 40,000 indexed diffraction patterns each were used, and preliminary

PYP structures refined against these data. Refinement details are in the SM. The 3 bins with the

shortest delays can be interpreted with chromophores in a twisted trans configuration (Fig. 2 A-

C). After 700 fs the configuration is near cis (Fig. 2 D-E). The time-course of the refined φtail

torsional angles can be fit with a transition time identical to that observed in the second rSV (Fig.

3). We took advantage of the similarity of the DED maps for extended time ranges before and after

the transition to further increase the accuracy of the refined structures. We combined the diffraction

patterns into two bins: the fast time scale (100-400 fs with 81,237 patterns) and a slower time scale

(800-1200 fs with 157,082 patterns) (Tab. S1C). We refined the structure denoted PYPfast against

the 100-400 fs data, and that denoted PYPslow against the 800-1200 fs data. The refinement

statistics are presented in Tab. S2. The DED maps are shown in Fig. 3, inserts (see also Fig.

7

S9B,D), with the corresponding, refined structures of PYPfast and PYPslow in pink and light green,

respectively. The 3ps DED map and the refined PYP3ps structure are shown in Fig. 2G. We used

as many diffraction patterns as possible to refine PYPslow (Fig. S12 B,D) and PYP3ps because at

the transition, roughly 30% of the excited molecules return directly to the dark state, no longer

contribute to the DED maps and reduce the signal. We emphasize that refinement of transient

structures populated on an ultrafast timescale is challenging, since these structures are very far

from equilibrium and likely to be highly strained. Restraints in standard libraries are derived from

structures at equilibrium and are therefore not applicable. In order to provide restraints more

appropriate for this refinement, we employed excited state quantum mechanics/molecular

mechanics (QM/MM) calculations on PYP (20, 34) (SM). In addition, we employed an iterative

procedure, in which improved difference phases φ∆F,calc were obtained and used with observed

difference structure factor amplitudes during refinement (SM). The structural results of the

refinement are summarized in Tab. 1. For the shortest time delays (up to about 450 fs), the PYP

chromophore tail adopts a highly strained, twisted trans configuration, in which the C1-C2=C3-C1’

torsional angle φtail (shown by the red line spanning these four atoms in Fig. 1B) is ~140o. The

position of the C2=C3 double bond in PYPfast is displaced by ~1Å behind the chromophore plane

(loosely defined by the Cys69 sulfur, the tail carbonyl oxygen and the atoms of the phenyl ring;

Fig. 2A-C). Hydrogen bonds to Glu46 and Tyr42, which are unusually short in the reference (dark)

structure (24), are substantially elongated from 2.5 Å to 3.4 Å (Tab. 1). This structure is primed

for the transition to cis. During the structural transition, substantial rotation about the double bond

takes place. The head of the chromophore pivots about tail atom C2 and thereby aligns the C2 =C3

bond along the tail axis. Simultaneously, the head rotates about the C3-C1’ single bond. (The

complex motions can be effectively illustrated by using an educator’s stick model set, see Fig. S3).

The phenolate oxygen (Fig. 1B, O4’) moves even further away (3.6 Å, Tab. 1) from Glu46 (Fig.

2D-F and Fig. S9C-D), thereby breaking the hydrogen bond. At time delays longer than about 700

fs, φtail has decreased to ~50o (PYPslow, Fig. 3), which is characteristic of a cis configuration.

PYPslow relaxes further towards the 3 ps structure (PYP3ps), in which the hydroxyl oxygen of the

head re-establishes its hydrogen bond with Glu46 (Fig. 2G). φtail changes slightly to ~35o. The

PYP3ps structure is already very similar to the early structures derived with 100 ps time resolution

by independent, synchrotron-based approaches (Tab. 1; PDB entries 4I38 and 4B90) (22, 23), and

8

has evolved only slightly from PYPslow by establishing shorter hydrogen bonds to Tyr42 and

Glu46.

The structures derived from the refinements confirm that the transition at around 550 fs is

indeed associated with a trans to cis isomerization. Theoretical considerations (20) (Fig. S8)

suggest that during isomerization the PYP chromophore relaxes through a conical intersection

between the electronically excited state PES and the ground state PES. Accordingly, structures

between 100 and 400 fs can be identified as electronically excited, whereas the structures at time

delays > 700 fs can be identified with the electronic ground state. In both the excited and ground

states, structural changes i.e. translation of atoms may also have occurred. Our experiments

identify the ultrafast dynamics of both the excited state structures and the ground state structures

(Figs. 2-3). Since we restricted our pump laser pulses to moderate power, we avoid damaging non-

linear effects (e.g. two photon absorption) and most excited molecules populate the excited state

surface S1 (5). Part of the stored energy is used to rapidly displace the chromophore by about 0.7

Å within the crowded molecular environment in the interior of PYP (Fig. 2A, Tab. 1). If this initial

displacement is complete after 250fs the chromophore must have experienced an acceleration of

~2x1015 m/s2 and attains a final velocity of 500 m/s (SM). Fig. 1B shows that 9 carbon atoms, two

oxygens and 7 hydrogen atoms (molecular mass = 147 g/mol) are displaced. During the first few

hundred fs the force on the chromophore is ~500 pN which is enormous compared to forces in

single molecules at thermal equilibrium which are usually only a few pN (35). The origin of the

force is due to the change of the potential energy surface when the chromophore is excited to the

electronic excited state which affects the intra and intermolecular interactions of the chromophore

as also inferred from ultrafast Raman spectroscopy(3). The energy required to displace the

chromophore is ~0.2 eV which is ~10% of the blue photon energy (2.76 eV) that starts the reaction.

It appears that by rapidly evolving down the excited state PES, part of the photon energy is initially

converted into kinetic energy which is then released by collision of the chromophore atoms with

the surrounding protein atoms comprising the chromophore pocket. The excited chromophore

loses 0.12 eV energy by intramolecular vibrational energy redistribution on the sub-100 fs time

scale (39) which can be roughly estimated from the Stokes shift by comparing absorption and

fluorescence spectra(3). Accordingly,~85% of the photon energy remains stored as strain and

electronic excitation in the chromophore before isomerization occurs. On passing through the

conical intersection (20), the molecules either revert towards the initial dark state (30% of the

9

excited molecules, Tab. 1, see also Tab. S3) or continue relaxing towards the cis isomer (70%),

gradually releasing the excess energy as heat. Because the chromophore pocket tightly restricts the

chromophore head displacements, further structural changes must be volume-conserving i.e. they

minimize the volume swept out by the atoms as they move. Accordingly, the chromophore

performs the complex motions described above (Fig. S3). Although the energy stored in the

chromophore is sufficient to break the hydrogen bonds (~0.1 eV), the spatial constraints imposed

by the chromophore pocket direct the reformation of the hydrogen bonding network at longer time

delays (Tab. 1). This is a ‘macromolecular cage effect’ reminiscent of the ‘solvent cage effect’ in

liquid chemical dynamics (36). The ‘macromolecular cage’ in PYP, however, is soft enough to

allow certain specific, relatively large (up to 1.3 Å, Tab. 1) structural changes. This contrasts with

crystals of small molecules, where the stronger crystal lattice constraints usually do not allow such

large displacements. Hence, biological macromolecular crystallography aimed at elucidating

biological function may also provide insight into the reaction mechanisms of small molecules.

To assess global conformational changes of PYP on the fs time scale, we calculated the radius

of gyration Rg from each refined structure (SM). Rg fluctuates by only 0.2% in all structures from

200 fs to 200 ns (Tab. 1). An increase of Rg by up to 1 Å determined by others using X-ray

scattering in solution upon photo-dissociation of CO from CO-myoglobin in solution (9) is not

observed in our PYP crystals. Concomitant, systematic large volume changes are also not apparent

in PYP crystals over the first 3 ps that our data span. Our data show no evidence for a protein

quake (9, 10, 37), characterized by an ultrafast and large change in Rg that occurs significantly

before a large volume change. The reason for this is unclear and will require further experiments.

Ultrafast fluorescence and transient absorption spectroscopy of PYP has shown that excited

state decay is multi-phasic (3, 5, 38). The fast (sub-ps) time constants are significantly more

productive in creating the cis-like photoproduct than the slow (ps) time constants; the long-lived

excited state population primarily decays back to the ground state (5, 39). With excitation at 450

nm, at least 50% of the total isomerization yield is generated with a dominant ~600 fs time constant

(5), which agrees with our observation of a transition at ~550 fs. It should be noted that a ‘ground

state intermediate’ with a 3-6 ps life time has been proposed by ultrafast spectroscopy (39).

However under the conditions employed here, the peak concentration of this intermediate is

expected to be small (5). In contrast to spectroscopic techniques that reported vibrational

coherence with 50 cm-1 and 150 cm-1 frequency (3, 40), we could not unambiguously detect

10

oscillations in our data (see above). Intense femtosecond optical pumping of PYP crystals

generates both excited state and ground state vibrational coherences within the 3.15 THz

experimental bandwidth(41). It will be an important goal of future experiments to structurally

characterize these coherences using fs TR-SFX. Nevertheless, our data show that before 400 fs

there are large distortions corresponding to a Franck-Condon (FC) excited state (42). The nuclear

dynamics of the FC excited state at 100-200 fs agrees with the conclusions from ultra-fast

spectroscopy (3, 42-45) that also suggest a distortion of the C2=C3 double bond on similar

timescales, as in the PYPfast structure. The isomerization at 550 fs through the conical intersection

between the excited state and ground state PES is in reasonable agreement with the timescales for

isomerization reported by others (3, 5, 42, 46). After passing through the conical intersection, the

chromophore is cis-like and still highly strained. The transiently-broken hydrogen bond is re-

established quickly as the structure relaxes, exemplified by the PYP3ps structure (Fig. 3). Further

relaxation on the ground state PES completes the initial phase of the isomerization.

11

References

1. G. Wald, Molecular basis of visual excitation. Science 162, 230-239 (1968). 2. Y. Mizutani, T. Kitagawa, Direct observation of cooling of heme upon photodissociation

of carbonmonoxy myoglobin. Science 278, 443-446 (1997). 3. R. Nakamura, N. Hamada, H. Ichida, F. Tokunaga, Y. Kanematsu, Coherent oscillations

in ultrafast fluorescence of photoactive yellow protein. The Journal of chemical physics 127, 215102 (2007); published online EpubDec 7 (10.1063/1.2802297).

4. P. M. Champion, Chemistry. Following the flow of energy in biomolecules. Science 310, 980-982 (2005); published online EpubNov 11 (10.1126/science.1120280).

5. C. N. Lincoln, A. E. Fitzpatrick, J. J. van Thor, Photoisomerisation quantum yield and non-linear cross-sections with femtosecond excitation of the photoactive yellow protein. Physical chemistry chemical physics : PCCP 14, 15752-15764 (2012); published online EpubDec 5 (10.1039/c2cp41718a).

6. A. Warshel, Bicycle-pedal model for the first step in the vision process. Nature 260, 679-683 (1976).

7. M. P. Minitti, J. M. Budarz, A. Kirrander, J. S. Robinson, D. Ratner, L. T.J, D. Zhu, J. Glownia, M. Kozina, H. Lemke, M. Sikorski, Y. Feng, S. Nelson, K. Saita, B. Stankus, T. Northey, J. B. Hastings, P. M. Weber, Imaging Molecular Motion: Femtosecond X-ray Scattering of an Electrocyclic Chemical Reaction. Physical review letters 114, 1-5 (2015)10.1103/PhysRevLett.114.255501).

8. J. M. Glownia, J. Cryan, J. Andreasson, A. Belkacem, N. Berrah, C. I. Blaga, C. Bostedt, J. Bozek, L. F. DiMauro, L. Fang, J. Frisch, O. Gessner, M. Guhr, J. Hajdu, M. P. Hertlein, M. Hoener, G. Huang, O. Kornilov, J. P. Marangos, A. M. March, B. K. McFarland, H. Merdji, V. S. Petrovic, C. Raman, D. Ray, D. A. Reis, M. Trigo, J. L. White, W. White, R. Wilcox, L. Young, R. N. Coffee, P. H. Bucksbaum, Time-resolved pump-probe experiments at the LCLS. Optics express 18, 17620-17630 (2010); published online EpubAug 16 (10.1364/OE.18.017620).

9. M. Levantino, G. Schiro, H. T. Lemke, G. Cottone, J. M. Glownia, D. Zhu, M. Chollet, H. Ihee, A. Cupane, M. Cammarata, Ultrafast myoglobin structural dynamics observed with an X-ray free-electron laser. Nat Commun 6, 6772 (2015).

10. D. Arnlund, L. C. Johansson, C. Wickstrand, A. Barty, G. J. Williams, E. Malmerberg, J. Davidsson, D. Milathianaki, D. P. DePonte, R. L. Shoeman, D. J. Wang, D. James, G. Katona, S. Westenhoff, T. A. White, A. Aquila, S. Bari, P. Berntsen, M. Bogan, T. B. van Driel, R. B. Doak, K. S. Kjaer, M. Frank, R. Fromme, I. Grotjohann, R. Henning, M. S. Hunter, R. A. Kirian, I. Kosheleva, C. Kupitz, M. N. Liang, A. V. Martin, M. M. Nielsen, M. Messerschmidt, M. M. Seibert, J. Sjohamn, F. Stellato, U. Weierstall, N. A. Zatsepin, J. C. H. Spence, P. Fromme, I. Schlichting, S. Boutet, G. Groenhof, H. N. Chapman, R. Neutze, Visualizing a protein quake with time-resolved X-ray scattering at a free-electron laser. Nature methods 11, 923-926 (2014); published online EpubSep (Doi 10.1038/Nmeth.3067).

11. T. R. M. Barends, L. Foucar, A. Ardevol, K. Nass, A. Aquila, S. Botha, B. Doak, K. Falahati, E. Hartmann, M. Hilpert, M. Heinz, M. C. Hoffmann, J. Koefinger, J. E. Koglin, G. Kovacsova, M. Liang, D. Milathianaki, H. Lemke, J. Reinstein, C. M. Roome, R. L. Shoeman, G. J. Williams, I. Burghardt, G. Hummer, S. Boutet, I. Schlichting, Direct observation of ultrafast collective motions in CO myoglobin upon ligand dissociation. Science, (2015).

12

12. P. Coppens, Perspective: On the relevance of slower-than-femtosecond time scales in chemical structural-dynamics studies. Structural Dynamics 2, (2015).

13. A. H. Zewail, Femtochemistry: Atomic-Scale Dynamics of the Chemical Bond Using Ultrafast Lasers (Nobel Lecture) Copyright((c)) The Nobel Foundation 2000. We thank the Nobel Foundation, Stockholm, for permission to print this lecture. Angew Chem Int

Ed Engl 39, 2586-2631 (2000); published online EpubAug 4 ( 14. M. Gao, C. Lu, H. Jean-Ruel, L. C. Liu, A. Marx, K. Onda, S. Y. Koshihara, Y. Nakano,

X. Shao, T. Hiramatsu, G. Saito, H. Yamochi, R. R. Cooney, G. Moriena, G. Sciaini, R. J. Miller, Mapping molecular motions leading to charge delocalization with ultrabright electrons. Nature 496, 343-346 (2013); published online EpubApr 18 (10.1038/nature12044).

15. S. Techert, F. Schotte, M. Wulff, Picosecond X-ray diffraction probed transient structural changes in organic solids. Physical review letters 86, 2030-2033 (2001).

16. C. D. Kim, S. Pillet, G. Wu, W. K. Fullagar, P. Coppens, Excited-state structure by time-resolved X-ray diffraction. Acta crystallographica. Section A, Foundations of

crystallography 58, 133-137 (2002). 17. J. B. Benedict, A. Makal, J. D. Sokolow, E. Trzop, S. Scheins, R. Henning, T. Graber, P.

Coppens, Time-resolved Laue diffraction of excited species at atomic resolution: 100 ps single-pulse diffraction of the excited state of the organometallic complex Rh2(mu-PNP)2(PNP)2.BPh4. Chem Commun (Camb) 47, 1704-1706 (2011); published online EpubFeb 14 (10.1039/c0cc04997b).

18. H. Ihee, M. Lorenc, T. K. Kim, Q. Y. Kong, M. Cammarata, J. H. Lee, S. Bratos, M. Wulff, Ultrafast x-ray diffraction of transient molecular structures in solution. Science 309, 1223-1227 (2005); published online EpubAug 19 (10.1126/science.1114782).

19. R. Neutze, R. Wouts, S. Techert, J. Davidsson, M. Kocsis, A. Kirrander, F. Schotte, M. Wulff, Visualizing photochemical dynamics in solution through picosecond x-ray scattering. Physical review letters 87, 195508 (2001).

20. G. Groenhof, M. Bouxin-Cademartory, B. Hess, S. P. De Visser, H. J. Berendsen, M. Olivucci, A. E. Mark, M. A. Robb, Photoactivation of the photoactive yellow protein: why photon absorption triggers a trans-to-cis Isomerization of the chromophore in the protein. Journal of the American Chemical Society 126, 4228-4233 (2004); published online EpubApr 7 (10.1021/ja039557f).

21. D. S. Larsen, M. Vengris, I. H. van Stokkum, M. A. van der Horst, F. L. de Weerd, K. J. Hellingwerf, R. van Grondelle, Photoisomerization and photoionization of the photoactive yellow protein chromophore in solution. Biophysical journal 86, 2538-2550 (2004); published online EpubApr (10.1016/S0006-3495(04)74309-X).

22. Y. O. Jung, J. H. Lee, J. Kim, M. Schmidt, K. Moffat, V. Srajer, H. Ihee, Volume-conserving trans-cis isomerization pathways in photoactive yellow protein visualized by picosecond X-ray crystallography. Nature chemistry 5, 212-220 (2013); published online EpubMar (10.1038/nchem.1565).

23. F. Schotte, H. S. Cho, V. R. Kaila, H. Kamikubo, N. Dashdorj, E. R. Henry, T. J. Graber, R. Henning, M. Wulff, G. Hummer, M. Kataoka, P. A. Anfinrud, Watching a signaling protein function in real time via 100-ps time-resolved Laue crystallography. Proceedings

of the National Academy of Sciences of the United States of America 109, 19256-19261 (2012); published online EpubNov 20 (10.1073/pnas.1210938109).

13

24. S. Anderson, S. Crosson, K. Moffat, Short hydrogen bonds in photoactive yellow protein. Acta Crystallogr D 60, 1008-1016 (2004); published online EpubJun (Doi 10.1107/S090744490400616x).

25. J. Tenboer, S. Basu, N. Zatsepin, K. Pande, D. Milathianaki, M. Frank, M. Hunter, S. Boutet, G. J. Williams, J. E. Koglin, D. Oberthuer, M. Heymann, C. Kupitz, C. Conrad, J. Coe, S. Roy-Chowdhury, U. Weierstall, D. James, D. Wang, T. Grant, A. Barty, O. Yefanov, J. Scales, C. Gati, C. Seuring, V. Srajer, R. Henning, P. Schwander, R. Fromme, A. Ourmazd, K. Moffat, J. J. Van Thor, J. C. Spence, P. Fromme, H. N. Chapman, M. Schmidt, Time-resolved serial crystallography captures high-resolution intermediates of photoactive yellow protein. Science 346, 1242-1246 (2014); published online EpubDec 5 (10.1126/science.1259357).

26. M. Liang, G. J. Williams, M. Messerschmidt, M. M. Seibert, P. A. Montanez, M. Hayes, D. Milathianaki, A. Aquila, M. S. Hunter, J. E. Koglin, D. W. Schafer, S. Guillet, A. Busse, R. Bergan, W. Olson, K. Fox, N. Stewart, R. Curtis, A. A. Miahnahri, S. Boutet, The Coherent X-ray Imaging instrument at the Linac Coherent Light Source. J

Synchrotron Radiat 22, 514-519 (2015); published online EpubMay 1 (10.1107/S160057751500449X).

27. R. Trebino, K. W. DeLong, D. N. Fittinghoff, J. N. Sweetser, M. A. Krumbugel, B. A. Richman, D. J. Kane, Measuring ultrashort laser pulses in the time-frequency domain using frequency-resolved optical gating. Review of Scientific Instruments 68, 3277-3295 (1997); published online EpubSep (Doi 10.1063/1.1148286).

28. C. J. Bardeen, Q. Wang, C. V. Shank, Selective Excitation of Vibrational Wave-Packet Motion Using Chirped Pulses. Physical review letters 75, 3410-3413 (1995).

29. M. Schmidt, V. Srajer, R. Henning, H. Ihee, N. Purwar, J. Tenboer, S. Tripathi, Protein energy landscapes determined by five-dimensional crystallography. Acta Crystallogr D 69, 2534-2542 (2013); published online EpubDec (Doi 10.1107/S0907444913025997).

30. N. Hartmann, W. Helml, A. Galler, M. R. Bionta, J. Grunert, S. L. Molodtsov, K. R. Ferguson, S. Schorb, M. L. Swiggers, S. Carron, C. Bostedt, J. C. Castagna, J. Bozek, J. M. Glownia, D. J. Kane, A. R. Fry, W. E. White, C. P. Hauri, T. Feurer, R. N. Coffee, Sub-femtosecond precision measurement of relative X-ray arrival time for free-electron lasers. Nat Photonics 8, 706-709 (2014).

31. M. R. Bionta, H. T. Lemke, J. P. Cryan, J. M. Glownia, C. Bostedt, M. Cammarata, J. C. Castagna, Y. Ding, D. M. Fritz, A. R. Fry, J. Krzywinski, M. Messerschmidt, S. Schorb, M. L. Swiggers, R. N. Coffee, Spectral encoding of x-ray/optical relative delay. Optics

express 19, 21855-21865 (2011); published online EpubOct 24 (10.1364/OE.19.021855). 32. M. Schmidt, S. Rajagopal, Z. Ren, K. Moffat, Application of singular value

decomposition to the analysis of time-resolved macromolecular X-ray data. Biophysical

journal 84, 2112-2129 (2003); published online EpubMar (10.1016/S0006-3495(03)75018-8).

33. K. D. Rector, M. D. Fayer, Myoglobin dynamics measured with vibrational echo experiments. Laser Chem 19, 19-34 (1999).

34. G. Groenhof, Introduction to QM/MM Simulations. Methods Mol Biol 924, 43-66 (2013)10.1007/978-1-62703-017-5_3).

35. M. Rief, H. Grubmuller, Force spectroscopy of single biomolecules. Chemphyschem 3, 255-261 (2002).

14

36. F. Patron, S. A. Adelman, Solvent Cage Effects and Chemical-Dynamics in Liquids. Chem Phys 152, 121-131 (1991); published online EpubApr 15 (Doi 10.1016/0301-0104(91)80039-K).

37. L. Genberg, L. Richard, G. Mclendon, R. J. D. Miller, Direct Observation of Global Protein Motion in Hemoglobin and Myoglobin on Picosecond Time Scales. Science 251, 1051-1054 (1991).

38. M. Vengris, M. A. van der Horst, G. Zgrablic, I. H. van Stokkum, S. Haacke, M. Chergui, K. J. Hellingwerf, R. van Grondelle, D. S. Larsen, Contrasting the excited-state dynamics of the photoactive yellow protein chromophore: protein versus solvent environments. Biophysical journal 87, 1848-1857 (2004); published online EpubSep (10.1529/biophysj.104.043224).

39. D. S. Larsen, I. H. van Stokkum, M. Vengris, M. A. van Der Horst, F. L. de Weerd, K. J. Hellingwerf, R. van Grondelle, Incoherent manipulation of the photoactive yellow protein photocycle with dispersed pump-dump-probe spectroscopy. Biophysical journal 87, 1858-1872 (2004); published online EpubSep (10.1529/biophysj.104.043794).

40. N. Mataga, H. Chosrowjan, Y. Shibata, Y. Imamoto, M. Kataoka, F. Tokunaga, Ultrafast photoinduced reaction dynamics of photoactive yellow protein (PYP): observation of coherent oscillations in the femtosecond fluorescence decay dynamics. Chem Phys Lett 352, 220-225 (2002); published online EpubJan 30 (Doi 10.1016/S0009-2614(01)01448-8).

41. A. T. N. Kumar, F. Rosca, A. Widom, P. M. Champion, Investigations of ultrafast nuclear response induced by resonant and nonresonant laser pulses. Journal of Chemical

Physics 114, 6795-6815 (2001). 42. M. Creelman, M. Kumauchi, W. D. Hoff, R. A. Mathies, Chromophore Dynamics in the

PYP Photocycle from Femtosecond Stimulated Raman Spectroscopy. Journal of Physical

Chemistry B 118, 659-667 (2014). 43. M. L. Groot, L. J. van Wilderen, D. S. Larsen, M. A. van der Horst, I. H. van Stokkum,

K. J. Hellingwerf, R. van Grondelle, Initial steps of signal generation in photoactive yellow protein revealed with femtosecond mid-infrared spectroscopy. Biochemistry 42, 10054-10059 (2003); published online EpubSep 2 (10.1021/bi034878p).

44. K. Heyne, O. F. Mohammed, A. Usman, J. Dreyer, E. T. Nibbering, M. A. Cusanovich, Structural evolution of the chromophore in the primary stages of trans/cis isomerization in photoactive yellow protein. Journal of the American Chemical Society 127, 18100-18106 (2005); published online EpubDec 28 (10.1021/ja051210k).

45. L. J. van Wilderen, M. A. van der Horst, I. H. van Stokkum, K. J. Hellingwerf, R. van Grondelle, M. L. Groot, Ultrafast infrared spectroscopy reveals a key step for successful entry into the photocycle for photoactive yellow protein. Proceedings of the National

Academy of Sciences of the United States of America 103, 15050-15055 (2006); published online EpubOct 10 (10.1073/pnas.0603476103).

46. S. Devanathan, A. Pacheco, L. Ujj, M. Cusanovich, G. Tollin, S. Lin, N. Woodbury, Femtosecond spectroscopic observations of initial intermediates in the photocycle of the photoactive yellow protein from Ectothiorhodospira halophila. Biophysical journal 77, 1017-1023 (1999).

47. H. Ihee, S. Rajagopal, V. Srajer, R. Pahl, S. Anderson, M. Schmidt, F. Schotte, P. A. Anfinrud, M. Wulff, K. Moffat, Visualizing reaction pathways in photoactive yellow protein from nanoseconds to seconds. Proceedings of the National Academy of Sciences

15

of the United States of America 102, 7145-7150 (2005); published online EpubMay 17 (10.1073/pnas.0409035102).

48. U. Weierstall, J. C. Spence, R. B. Doak, Injector for scattering measurements on fully solvated biospecies. The Review of scientific instruments 83, 035108 (2012); published online EpubMar (10.1063/1.3693040).

49. P. Hart, S. Boutet, C. Gabriella, M. Dubrovin, B. Duda, D. Fritz, G. Haller, R. Herbst, S. Herrmann, C. Kenney, N. Kurita, H. Lemke, M. Messerschmidt, M. Nordby, D. Schafer, M. Swift, M. Weaver, G. Williams, D. Zhu, N. Van Bakel, J. Morse, The CSPAD megapixel x-ray camera at LCLS. Proc. SPIE 8504, 85040C (2012).

50. C. D. M. Hutchison, J. Tenboer, C. Kupitz, K. Moffat, M. Schmidt, J. J. van Thor, Photocycle Populations with Femtosecond Excitation of Crystalline Photoactive Yellow Protein. J Chem Phys Lett, (submitted) (2016).

51. A. Barty, R. A. Kirian, F. R. N. C. Maia, M. Hantke, C. H. Yoon, T. A. White, H. Chapman, Cheetah: software for high-throughput reduction and analysis of serial femtosecond X-ray diffraction data. J Appl Crystallogr 47, 1118-1131 (2014); published online EpubJun (Doi 10.1107/S1600576714007626).

52. T. A. White, Post-refinement method for snapshot serial crystallography. Philosophical

transactions of the Royal Society of London. Series B, Biological sciences 369, 20130330 (2014); published online EpubJul 17 (10.1098/rstb.2013.0330).

53. T. A. White, R. A. Kirian, A. V. Martin, A. Aquila, K. Nass, A. Barty, H. N. Chapman, CrystFEL: a software suite for snapshot serial crystallography. J Appl Crystallogr 45, 335-341 (2012); published online EpubApr (Doi 10.1107/S0021889812002312).

54. E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt, E. C. Meng, T. E. Ferrin, UCSF chimera - A visualization system for exploratory research and analysis. Journal of computational chemistry 25, 1605-1612 (2004); published online EpubOct (Doi 10.1002/Jcc.20084).

55. H. M. Berman, T. Battistuz, T. N. Bhat, W. F. Bluhm, P. E. Bourne, K. Burkhardt, Z. Feng, G. L. Gilliland, L. Iype, S. Jain, P. Fagan, J. Marvin, D. Padilla, V. Ravichandran, B. Schneider, N. Thanki, H. Weissig, J. D. Westbrook, C. Zardecki, The Protein Data Bank. Acta crystallographica. Section D, Biological crystallography 58, 899-907 (2002).

56. M. D. Winn, C. C. Ballard, K. D. Cowtan, E. J. Dodson, P. Emsley, P. R. Evans, R. M. Keegan, E. B. Krissinel, A. G. W. Leslie, A. McCoy, S. J. McNicholas, G. N. Murshudov, N. S. Pannu, E. A. Potterton, H. R. Powell, R. J. Read, A. Vagin, K. S. Wilson, Overview of the CCP4 suite and current developments. Acta Crystallogr D 67, 235-242 (2011); published online EpubApr (Doi 10.1107/S0907444910045749).

57. G. N. Murshudov, P. Skubak, A. A. Lebedev, N. S. Pannu, R. A. Steiner, R. A. Nicholls, M. D. Winn, F. Long, A. A. Vagin, REFMAC5 for the refinement of macromolecular crystal structures. Acta crystallographica. Section D, Biological crystallography 67, 355-367 (2011); published online EpubApr (10.1107/S0907444911001314).

58. M. Schmidt, Structure based enzyme kinetics by time-resolved X-ray crystallography, in:

ultrashort laser pulses in medicine and biology. W. Zinth, M. Braun, P. Gilch, Eds., Biological and medical physics, biomedical engineering, ISSN 1618-7210 (Berlin ; New York : Springer, c2008, Germany, 2008).

59. A. Brown, F. Long, R. A. Nicholls, J. Toots, P. Emsley, G. Murshudov, Tools for macromolecular model building and refinement into electron cryo-microscopy

16

reconstructions. Acta Crystallogr D 70, 136-153 (2015); published online EpubJan (Doi 10.1107/S1399004714021683).

60. G. J. Kleywegt, Use of non-crystallographic symmetry in protein structure refinement. Acta crystallographica. Section D, Biological crystallography 52, 842-857 (1996); published online EpubJul 1 (10.1107/S0907444995016477).

61. F. Parak, H. Hartmann, M. Schmidt, G. Corongiu, E. Clementi, The hydration shell of myoglobin. European biophysics journal : EBJ 21, 313-320 (1992).

62. B. Lee, F. M. Richards, The interpretation of protein structures: estimation of static accessibility. Journal of molecular biology 55, 379-400 (1971).

63. N. R. Voss, M. Gerstein, 3V: cavity, channel and cleft volume calculator and extractor. Nucleic Acids Res 38, W555-562 (2010); published online EpubJul 10.1093/nar/gkq395).

64. D. Van Der Spoel, E. Lindahl, B. Hess, G. Groenhof, A. E. Mark, H. J. Berendsen, GROMACS: fast, flexible, and free. Journal of computational chemistry 26, 1701-1718 (2005); published online EpubDec (10.1002/jcc.20291).

65. N. Purwar, J. Tenboer, S. Tripathi, M. Schmidt, Spectroscopic studies of model photo-receptors: Validation of a nanosecond time-resolved micro-spectrophotometer design using photoactive yellow protein and alfa-phycoerythrocyanin. International Journal of

Molecular Sciences 14, 17 (2013). 66. J. Manz, H. Naundorf, K. Yamashita, Y. Zhao, Quantum model simulation of complete

S-0 -> S-1 population transfer by means of intense laser pulses with opposite chirp. Journal of Chemical Physics 113, 8969-8980 (2000).

67. D. J. Tannor, S. A. Rice, Control of selectivity of chemical reaction via control of wave packet evolution. J. Chem. Phys. 83, 5013-5018 (1985).

68. S. Ruhman, R. Kosloff, Application of Chirped Ultrashort Pulses for Generating Large-Amplitude Ground-State Vibrational Coherence - a Computer-Simulation. J Opt Soc Am

B 7, 1748-1752 (1990). 69. Y. Duan, C. Wu, S. Chowdhury, M. C. Lee, G. Xiong, W. Zhang, R. Yang, P. Cieplak, R.

Luo, T. Lee, J. Caldwell, J. Wang, P. Kollman, A point-charge force field for molecular mechanics simulations of proteins based on condensed-phase quantum mechanical calculations. Journal of computational chemistry 24, 1999-2012 (2003); published online EpubDec (10.1002/jcc.10349).

70. A. D. Becke, A New Mixing of Hartree-Fock and Local Density-Functional Theories. Journal of Chemical Physics 98, 1372-1377 (1993).

71. C. T. Lee, W. T. Yang, R. G. Parr, Development of the Colle-Salvetti Correlation-Energy Formula into a Functional of the Electron-Density. Phys Rev B 37, 785-789 (1988).

72. C. I. Bayly, P. Cieplak, W. D. Cornell, P. A. Kollman, A Well-Behaved Electrostatic Potential Based Method Using Charge Restraints for Deriving Atomic Charges - the Resp Model. J Phys Chem-Us 97, 10269-10280 (1993).

73. J. Tomasi, B. Mennucci, R. Cammi, Quantum mechanical continuum solvation models. Chemical reviews 105, 2999-3093 (2005).

74. S. Yamaguchi, H. Kamikubo, K. Kurihara, R. Kuroki, N. Niimura, N. Shimizu, Y. Yamazaki, M. Kataoka, Low-barrier hydrogen bond in photoactive yellow protein. Proceedings of the National Academy of Sciences of the United States of America 106, 440-444 (2009); published online EpubJan 13 (10.1073/pnas.0811882106).

17

75. W. L. Jorgensen, J. Chandrasekar, J. D. Madura, R. W. Impey, M. L. Klein, Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926-935 (1983)10.1063/1.445869 ).

76. G. Bussi, D. Donadio, M. Parrinello, Canonical sampling through velocity rescaling. Journal of Chemical Physics 126, (2007).

77. B. Hess, H. Bekker, H. J. C. Berendsen, J. G. E. M. Fraaije, LINCS: A linear constraint solver for molecular simulations. Journal of computational chemistry 18, 1463-1472 (1997).

78. S. Miyamoto, P. A. Kollman, Settle - an Analytical Version of the Shake and Rattle Algorithm for Rigid Water Models. Journal of computational chemistry 13, 952-962 (1992).

79. U. Essmann, L. Perera, M. L. Berkowitz, T. Darden, H. Lee, L. G. Pedersen, A Smooth Particle Mesh Ewald Method. Journal of Chemical Physics 103, 8577-8593 (1995).

80. B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, GROMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. Journal of chemical theory

and computation 4, 435-447 (2008). 81. B. O. Roos, P. R. Taylor, P. E. M. Siegbahn, A complete active space SCF method

(CASSCF) using a density matrix formulated super-Cl approach. Chem. Phys. 48, 157-173 (1980).

82. M. Boggio-Pasqua, C. F. Burmeister, M. A. Robb, G. Groenhof, Photochemical reactions in biological systems: probing the effect of the environment by means of hybrid quantum chemistry/molecular mechanics simulations. Physical chemistry chemical physics :

PCCP 14, 7912-7928 (2012); published online EpubJun 14 (10.1039/c2cp23628a). 83. E. Fabiano, G. Groenhof, W. Thiel, Approximate switching algorithms for trajectory

surface hopping. Chem Phys 351, 111-116 (2008); published online EpubJul 3 (10.1016/j.chemphys.2008.04.003).

84. A. A. Granovsky, Extended multi-configuration quasi-degenerate perturbation theory: The new approach to multi-state multi-reference perturbation theory. Journal of

Chemical Physics 134, (2011); published online EpubJun 7 (Artn 21411310.1063/1.3596699).

85. R. Trebino, personal communication. 86. X. Gu, L. Xu, M. Kimmel, E. Zeek, P. O'Shea, A. P. Shreenath, R. Trebino, R. S.

Windeler, Frequency-resolved optical gating and single-shot spectral measurements reveal fine structure in microstructure-fiber continuum. Optics letters 27, 1174-1176 (2002).

18

Acknowledgements

This work is supported by NSF-STC “BioXFEL” (NSF-1231306), by NIH grants R01GM095583

(PF), R01EY024363 (KM) and R24GM111072 (VS, RH and KM), Helmholtz Association

“Virtual Institute Dynamic Pathways” (HC), and NSF-0952643 (MS). KP is partly supported by

NSF-1158138 (to Dilano Saldin and MS) and BMBF 05K14CHA (to HC). JvT acknowledges

support from the Engineering and Physical Sciences Research Council via grant agreement

EP/M000192/1. GG and DM are supported by the Academy of Finland, DO by BMBF project

05K13GUK, and MM by the European Union through grant FP7-PEOPLE-2011-ITN NanoMem.

Use of the Linac Coherent Light Source (LCLS), SLAC National Accelerator Laboratory, is

supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences

under contract DE-AC02-76SF00515. Part of this work was performed under the auspices of the

US Department of Energy by Lawrence Livermore National Laboratory (LLNL) under contract

DE-AC52-07NA27344, and M. F. was supported by LLNL Lab-Directed Research and

Development Project 012-ERD-031. C.G. thanks the PIER Helmholtz Graduate School for

financial support. Parts of the sample injector used at LCLS for this research were funded by the

NIH grant P41GM103393, formerly P41RR001209. We thank the Moscow State University

supercomputing center and the CSC–IT Center for Science, Finland, for computing resources. We

thank Dr. Mark Hunter for valuable discussions, Dr. Timo Graen for help with the computer

simulations and Chufeng Li for assistance with injectors. The PYPref, PYPfast, PYPslow, PYP3ps and

PYP200ns structures are deposited in the protein data bank together with their respective weighted

difference structure factor amplitudes under accession codes 5HD3, 5HDC, 5HDD, 5HDS and

5HD5, respectively.

Author Contributions.

M.S. prepared the proposal with input from J.J.vT., K.M., V.S., J.C.H.S., H.N.C., A.O. and P.F.;

A.A., S.B., M.L., J.S.R. and J.E.K. operated the CXI instrument including the time-tool and the

fs-laser; K.P., A.B., J.T., S.B., T.A.W., N.Z., O.Y. and T.D.G. analyzed the SFX data. C.D.M.H

and J.J.vT. set up the FROG at the CXI instrument; G.G. and D.M. performed QM/MM

calculations; J.T., J.B., D.O., P.L.X., C.G., C.K. and M.S. prepared protein and grew nano- and

microcrystals; D.DeP., C.K., C.C., S.R-C., J.D.C., M.M., G. K., and U.W. provided and operated

the injector system; M.F., R.F., M.S., J.T., P.F., D.O. and C.G. wrote the electronic log; M.F., M.S,

19

J.T., J.S.R., J.J.vT. and K.M. discussed fs laser excitation; J.T., M.S, V.S, R.H, C.D.M.H. and

J.J.vT. performed preliminary ultra-fast experiments on crystals; M.S. calculated and analyzed the

difference maps; M.S., K.P., K.M., G.G., P.F. and J.J.vT. wrote the manuscript with improvements

from all authors.

Supplementary Materials.

Materials and Methods Supplementary Text References 48-86 Figs. S1 to S12 Tables S1 to S5

20

Figures

- Figure 1 -

Figure 1. A. The PYP photocycle from the perspective of a time-resolved crystallographer. Approximate time scales are given. The fs/ps time scale (in red) is structurally charted in this paper. B. The chemical structure of the pCA chromophore. The red line marks the four atoms that define the torsional angle φtail about the C2=C3 bond. C. Results of the positive control experiment at a 200 ns time delay. Reaction initiated by fs laser pulses. Negative (red) and positive (blue) DED features on the -3σ/3σ level. A mixture of the pR1 (magenta) and pR2 (red) structures is present. Main signature of pR1: features β1 and β2. Main signature for pR2: features γ1 and γ2. Structure of PYPref (dark) in

21

- Figure 2 -

Figure 2. Trans to Cis isomerization in PYP. Weighted DED maps in red (-3 σ) and blue (3σ); front (upper) and side view (lower). Each map is prepared from about the same number of diffraction patterns, except the 3 ps map (see Tab. S1 B-C). The reference, dark structure is shown in yellow throughout; structures before the transition and still on the electronic excited state PES are shown in pink; structures after the transition and on the electronic ground state PES are shown in light green. Important negative difference density features are denoted α, positive features as β in panels B and G. Pronounced structural changes are marked by arrows. A-C: time-delays before the transition. A. Twisted trans at 142 fs, φtail 154o. B. Twisted trans at 269 fs, φtail 140o some important residues are marked; dotted lines in B: hydrogen bond of the ring hydroxyl to Glu46 and Tyr42. C. Twisted trans at 455 fs, φtail 144o; dotted line in C: direction of C2=C3 double bond. D-G: time delays and chromophore configuration after the transition. D. Early cis at 799 fs, φtail 50o. E. Early cis at 915 fs; dotted line in E: direction of C2=C3 double bond. F. Early cis at 1023 fs; for E and F φtail ~65o. G. 3 ps delay; dashed line: direction of C2=C3 double bond, feature β1; φtail is 35o.

22

- Figure 3 -

Figure 3. Trans to cis isomerization in PYP. Pink: twisted trans on excited state PES; light green: cis on ground state PES. Torsional angle φtail (solid spheres) from structural refinement at various delays (see also Tab. S3). Gray region: not time-resolved. Dashed line: fit with eqn. S2, with a transition time of about 590 fs (see also Fig. S2). Inserts: structures of PYPfast (pink), PYPslow and PYP3ps (light green), and dark state structure PYPref in yellow. Difference electron density in red (-3σ) and blue (3σ).

23

Tables Table 1. Geometry of PYP structures. The PYPfast structure was refined using a data bin spanning 100-400 fs with 81327snapshots, and the PYPslow structure from a bin spanning 800 – 1200 fs with 157082 snapshots (Tab. S1b). Structures of IT, pR0 and pB1 from Protein Data Bank, code listed in brackets (22, 23, 47). Uncertainties of the torsional angles can be estimated to be +/-20o by displacing the 4 atoms that define the angle with the coordinate error (0.2 Å).

PYPref

(dark) PYPfast PYPslow PYP3ps PYP200ns

(fs-laser) pR1/pR2

IT (4I38)

pR0 (4B90)

pB1 (1TS0)

Time Delay 0 100- 400 fs

800 - 1200 fs

3 ps 200 ns 100 ps

100 ps ms

Torsional Angles [o] C1-C2=C3-C1’ (φφφφtail)

172 136 53 35 3/-8 90 33 -27

O1-C1-C2=C3 -15 -21 28 30 12/-6 11 29 -10 CB-S-C1-C2 -185 -171 -164 -137 163/-165 -136 -123 180 Hydrogen bonds [Å] pCA-O4’ - Glu46-Oεεεε

2.50 3.40 3.60 2.94 4.97/2.88 2.73 2.73 8.03

pCA-O4’ - Tyr42-Oηηηη

2.54 2.92 2.63 2.88 2.97/2.66 2.57 2.59 5.19

pCA-O1 – Cys69-N

2.77 3.11 2.50 3.12 3.37/4.29 3.04 3.05 2.88

others <pCA>a [Å] 0 0.66 0.78 0.60 1.55/0.81 0.67 0.68 2.39 <global>b [Å] 0 0.20 0.19 0.24 0.13 0.13 0.19 0.17 Radius of gyration c [Å]

13.32 13.33 13.30 13.34 13.29 - nd - - nd - - nd-

Volume [Å3] 17831 17856 17833 17838 17672 17830 17683 17807 ∆∆∆∆V to dark [Å3] 0 25 2 7 -159 -1 -148 -24 Photoactivation Yield [%]d

- na - 15.2 9.6 10.1 12.5c (5%)e (10%)e

(10%)e

aMean displacement of equivalent chromophore atoms relative to dark (SM). bMean displacement of equivalent cα atoms relative to dark (SM). cSee SM for the calculation. dDetermined by by fitting calculated DED maps to the experimental DED maps in the chromophore region. eEstimate

24

Supplementary Materials for

Femtosecond Structural Dynamics Drives the Trans/Cis Isomerization in

Photoactive Yellow Protein

Kanupriya Pande, Christopher D. M. Hutchison, Gerrit Groenhof, Andy Aquila, Josef S. Robinson, Jason Tenboer, Shibom Basu, Sébastien Boutet, Daniel P. DePonte, Mengning Liang,

Thomas A. White, Nadia A. Zatsepin, Oleksandr Yefanov, Dmitry Morozov, Dominik Oberthuer, Cornelius Gati, Ganesh Subramanian, Daniel James, Yun Zhao, Jake Koralek,

Jennifer Brayshaw, Christopher Kupitz, Chelsie Conrad, Shatabdi Roy-Chowdhury, Jesse D. Coe, Markus Metz, Paulraj Lourdu Xavier, Thomas E. Grant, Jason E. Koglin, Gihan Ketawala, Raimund Fromme, Vukica Šrajer, Robert Henning, John C. H. Spence, Abbas Ourmazd, Peter Schwander, Uwe Weierstall, Matthias Frank, Petra Fromme, Anton Barty, Henry N. Chapman,

Keith Moffat, Jasper J. van Thor, Marius Schmidt

correspondence to: m-schmidt@uwm.edu This PDF file includes:

Materials and Methods Supplementary Text References 48-86 Figs. S1 to S12 Tables S1 to S5

25

Materials and Methods Sample preparation, data collection

PYP micro-crystals were grown and SFX data were collected as previously described (25). Microcrystals were injected using a liquid jet injector (48). X-ray data were collected on a Cornell Stanford Pixel Area Detector (49) located at a nominal distance of 84 mm from the jet. An electronic low gain mask was used to 4.5 Å resolution to increase the dynamic range of the detector by a factor ~6. 40 fs X-ray FEL pulses (9.5 keV, 2 mJ/pulse) were focused to 1 µm. Laser excitation and timing

10 µJ of 450 nm pulses were generated using an TOPAS-prime pumped by a 800 nm CPA Ti:sapph system (Coherent Legend). These were focused onto the jet using a f=+300mm lens to give a 140(±10) µm (FWHM) spot corresponding to an energy density of 0.75 mJ mm-2. Due to `hot-spotting' in the output of the TOPAS, it was necessary to install a telescope to expand the beam prior to focusing in order to achieve the required power density. The telescope consisted of two AR coated lenses (f= -70 mm and +200 mm). When combined with the f=+300mm lens, the 40±5 fs TOPAS output pulses were stretched to 140±5 fs. The on-axis configuration ensures a stable laser beam position. We determined the exact laser pulse shape and pulse duration after all the optical components using Second Harmonic Generation Frequency Resolved Optical Gating (SHG-FROG), which was installed by the Imperial College team at the beamline (see further below).

The highest laser power that could excite as many molecules as possible without generating significant laser damage was determined spectroscopically beforehand using laser pulses of 100 fs, 300 fs and 1 ps duration (see below) (50). Optimally, a 7% - 10% photoactivation yield was reached with positively chirped 100 fs laser pulses directly on resonance and energy densities up to 0.8 mJ/mm2; single-shot photo-bleaching did not exceed 0.5% (50). For spatial alignment of the laser and X-ray FEL beams, the overlap was adjusted visually using a YAG screen. Time zero (T0) was determined by pumping a YAG screen at the interaction plane with the X-rays and measuring the change in the transmission of the optical laser as a function of delay between the two pulses. This determined T0 to an accuracy of < 150 fs. The nominal pump-probe time delay ∆set could then be set by offsetting the laser pulse timing relative to the X-ray pulses. To correct for the pump-probe jitter ∆jitter of individual pulses, a timing tool was used upstream of the vacuum chamber. By changing the phase delay between the laser oscillator and an accelerator RF reference signal, the 140 fs laser pulse could be precisely adjusted to any selected pump-probe delay. ∆jitter

measured by the timing tool was added to the selected time delay ∆set. Larger time drifts between the FEL X-ray pulses and laser pulses occurred over the duration of data collection. These drifts could be measured by changing the delay line of the timing tool, or corrected by changing the laser offset. Correction was preferred since this method left ∆set unchanged, thus simplifying subsequent timing analysis. Once the traces were detected again by the timing tool, ∆jitter was directly accounted for. If the drift was not corrected, it had to be added to the time-delay ∆set. Since we acquired data both with and without correction, our temporal density of diffraction patterns (snapshots) was skewed (Fig. S1). As a consequence, when sorting into timing bins containing equal numbers of patterns, the time axis was not sampled at equal intervals (Fig. S1). Jitter correction was done for all diffraction patterns in the fs time range. For the 3 ps and the 7200 ns delays, the jitter was less than 10% of the time delay and could therefore be disregarded. Data Binning, Common Scale

For the positive control, data with a pump probe time-delay of 200 ns were collected. Reference (dark) data were collected at 120 Hz in a separate data set at a -1 ps time delay i.e. the

26

X-ray pulse arrives before the laser pulse. The negative time delay was intended to account for potential effects of the intense fs blue laser stray light on the detector response. The fs time-range was covered by 2 nominal delays, 300 fs and 600 fs. Data were collected in the following order: First, 1.9 x 106 snapshots with a nominal delay of 200 ns were collected within 6 h. Then 3.3 x 106 snapshots with a nominal delay of 600 fs were collected for 13 h followed by collecting 1 x106 snapshots for the -1 ps reference dataset for 3 h. We continued to collect 1.6 x 106 snapshots with a 600 fs nominal delay for 8 h, and 1.4 x 106 snapshots with a delay of 3 ps for 7 h. Finally we collected 1.5 x106 snapshots with a nominal delay of 300 fs for 8 h. This arrangement is necessitated by the fact that the laser cannot be switched at 120 Hz between the nominal pump probe delays ∆set that we wished to use of 300, 600 and 3 ps. If we were able to achieve this switching and to add a fourth delay of -1 ps (dark), the combination of fast switching of ∆set with the delay jitter would have resulted both in a more uniform distribution of diffraction patterns with time delay than that achieved (Fig. S1), and in the association of each group of four delays with its own dark pattern, close in clock time. However, the fact that LCLS patterns are each obtained on a new crystal means that interleaving of dark and light patterns offers much less reduction in systematic errors than in the synchrotron case, where dark and light data are collected on the same crystal, close in clock time. In total we collected 10.3 x 106 detector readouts (snapshots) within 45 hours spread across five 12 h shifts. 9.4 x 105 snapshots (9.1 %) were hits and 6 x 105 (5.8 %) could be indexed (see also Tab. S1 for statistics).

To cover the fs and early ps time range, these three nominal pump probe delays ∆set were used. Individual diffraction patterns (snapshots) containing Bragg reflections were designated as hits and selected using Cheetah (51). The timing tool information was used to determine corrections to ∆set for each snapshot by locating the pulse-to-pulse transition edge in the time tool signal and converting this to a delay time. This resulted in a continuous distribution of delay times (Fig. S1). All hits were indexed and integrated using CrystFEL version 0.6.1 (52, 53). The highly partial intensities from all time-delays, around 500,000 indexed patterns in total (Tab. S1), were merged to obtain an average dataset. The indexing ambiguity was solved at this stage for all diffraction patterns together, as previously reported (25). The intensities were merged using the ‘partialator’ from CrystFEL (version 0.6.1+ab098ce1), applying individual linear and Debye-Waller scaling factors to each crystal to best fit the merged data in an iterative procedure. Partialities were not assigned. Diffraction patterns were scaled in one large scaling job to ensure that intensities from all diffraction patterns and all time delays were on the same scale. The indexed snapshots were finally merged into 16 bins with around 20000 snapshots each, spanning the time range from ~100 fs to 1.1ps (Fig. S1, red squares, and Tab. S1A); or into 8 bins (Tab. S1B); or into two bins (Tab. S1C).

Difference maps and Singular Value Decomposition

Difference maps and structural models are displayed by ‘UCSF-CHIMERA’ (54) throughout the manuscript. Difference structure factor amplitudes and weighted DED maps were calculated from the SFX amplitudes for each of the 16 time delays (Tab. S1A) as described(25) using the SFX data of the -1 ps negative time delay as the dark reference. Following visual inspection of the DEDs in 8 bins, the DEDt in 16 bins (thus offering better time resolution) were subject to singular value decomposition (32), solely to pin-point the time when the structural transition occurs. The volume of the asymmetric unit subjected to SVD was determined by the pCA chromophore atoms, the Cys69 sulfur and the Glu46 carboxyl. We did not attempt any interpretation of the singular vectors in terms of structure and/or kinetic models. Fig. S2B shows the result of the SVD analysis.

27

The time evolution of the overall signal in the DEDs is represented by the first right singular vector (rsv1, black circles). It shows how the total difference electron density evolves as a function of time from relatively small values around 100 fs to larger values on longer fs and ps time scales. The time course of rsv1 can be fitted empirically by a simple exponential function of the form

���1��� = � + � �1 − ��� � �����. (S1)

The signal in the DED maps rises with τ1=155 fs; larger structural displacements persist to the largest time delay studied, 3 ps. The dashed blue and red curves in Fig. S2B were calculated also by eqn. S1, by re-interpreting the deviations from the fitted curve. The second right singular vector (rsv2, red squares), shows a pronounced variation with time. It was fit by an additional empirical function to determine the time delay at which this variation occurred τct:

���2��� = ������� !" #

$� . (S2)

Note that Eqns. S1 and S2 are purely empirical and have no physical (predictive) power. The transition time τtc is about 550 fs. This is consistent with our earlier observation that on the fast fs time scale < 500 fs the difference maps (DEDfast) are qualitatively different from those (DEDslow) on the longer fs time scales > 700 fs (Fig. 2 and Fig. S9). The derivative of eqn. S2 is a bell-shaped curve whose full width half maximum (FWHM) depends on the fit parameter B2. After fitting eqn. S2 to rsv2, parameter B2 is ~50 fs and the FWHM of the derivative of eqn. S2 is ~180 fs (Fig. S2B, blue curve). Refinement and Phased Difference Maps 200 ns pump-probe delay as positive control and reference model.

The dark (reference) model (PYPref) was refined against the SFX structure factor amplitudes to 1.5 Å resolution (see above). To refine a model at the 200 ns pump-probe delay, we first inspected the difference map DED200ns. This map arises from a mixture of two structures, pR1 and pR2, available from Protein Data Bank (55) entries 4WLA and 4WL9, which were used as a starting model. The CCP4 (56) program ‘REFMAC5’ (57) was used to refine the structure (Tab. S2). The pCA chromophore was identified from the ‘HC4.cif‘ library, and linked to cysteine using the program ‘JLIGAND’. This construct was used then as an unusual amino acid in the refinement, with the advantage of offering full control over the restraints. The pR1 and pR2 models were refined against extrapolated structure factor amplitudes |&���| = |&',)*+,| + - × /∆& where w∆F are weighted difference structure factor amplitudes (between the 200 ns time delay and the reference) and n was 10, with the same restraints used to refine the dark model. Fig. 1C in the main text presents the refined 200 ns structure superimposed on the DED200ns map. 3 ps and below. On the ultrafast time scales (fs and ps), structural refinement was challenging. There are only limited diffraction data and any refined structures will be far from equilibrium (see main text). The goal of refinement is to obtain chemically and structurally plausible chromophore configurations that account for the most significant positive (features β and γ in Fig. 2) and negative difference density (features α in Fig. 2) around the chromophore. There is an obvious difference in the DED maps between the 3 ps delay and the fast fs time points (compare Fig. 2A-B with Fig. 2E). At 3 ps the positive DED that corresponds to the double bond is aligned along the axis of the chromophore tail (dashed line in Fig. 2E, and Fig. S12C), whereas on the fast fs time scale this feature is at an angle with the tail (dashed line in Fig. 2A, and Fig. S12A). The chromophore tail structures on the fast fs time-scale clearly differ from those at 3 ps. To derive the restraint for the C1-C2=C3-C1’ torsional angle (φtail), we rely on guidance by quantum molecular mechanics (QM/MM) computer

28

simulations (see below). A trajectory φtail is shown in Fig. S7. On the fast fs time scale this angle is > 100 deg, which corresponds to a twisted trans configuration. However, at longer time scales, this angle is close to 0o, which corresponds to a full cis configuration. In order to minimize bias towards the QM/MM simulations, we did not make use of the QM/MM force-field for our refinement, but accepted the torsional angle from the QM/MM calculations as a restraint to guide and corroborate our structural analyses. Refinement of the 3 ps structure.

To refine a structure and to explain the features in the 3 ps difference map (DED3ps), we restrained φtail to the QM/MM value of about 20o (Fig. S7C). Extrapolated structure factor amplitudes |Fext| were calculated as above (for the 200 ns time point) by adding the observed (measured) weighted differences structure factor amplitudes (w∆Fobs) to calculated structure factors (Fc,dark) derived from the refined reference (dark) structure: |&���| = |&',)*+,| + - × /∆&. Initially n = 11. The initial model was refined to |Fext| and a new model obtained. φtail observed in this new model is ~70o. New structure factors Flight with amplitudes and phases were calculated from this model. Difference phases φ∆F,calc were calculated by subtracting Fc,dark from Fc,light. φ∆F,calc were combined with w∆Fobs and phased, weighted difference structure factors w∆Fobs,φ were obtained. New extrapolated structure factors &1�2 = &',)*+, + - × /∆&345,6 were calculated, this time by adding nw∆F as a vector in the complex plane. The value of n was increased to 17. A new model was refined against |Fnew| and the procedure iterated until convergence, when φtail settled at ~35o. A difference map calculated from ∆Fcalc obtained by subtracting structure factors of the reference model from those of the final model essentially reproduces the observed DED features. Importantly, the more parallel alignment of the difference feature β1 in Fig. 2E is reproduced. The configuration of the chromophore in this structure (PYP3ps) is cis (Tab. S2). Refinement of structures on the fs time scale.

For the refinement on the fs time-scale the X-ray data were sorted into bins containing ~40,000 patterns, each (Tab. S1B). Up to the 455 fs bin, the strongly kinked positive feature β1 (see Figs. 2, S9 and S12A) indicates a trans configuration. φtail was restrained to the QM/MM value of 120o. At times larger than 700 fs the DED maps and features β1 are similar to that at 3 ps (compare Fig. 12, B and C), and φtail was restrained to 20o. Refinement commenced as described above. Fig. 3C and Tab. S3 show the resulting torsional angles. At times faster than 500 fs, the chromophore adopts a trans configuration, after 700 fs the configuration is cis. We could not successfully refine a single model into the 699 fs DED map time bin presumably because it arises from a mixture of trans and cis configurations which our data are inadequate to resolve. φtail determined at various time delays follows the time course derived by singular value decomposition (compare Fig. S2 A-B). A transition time τtc of 590 fs was obtained by fitting eqn. S2 to the φtail. In order to enhance the signal to noise ratio of the data and the credibility of the structures we then sorted our data in even larger bins (Tab. S1C).

The DED map determined from X-ray data in the 100-400 fs bin (81327 indexed diffraction patterns, Tab. S1C) is shown in Figs. 3A and S9B. For refinement of the corresponding structure PYPfast, φtail was restrained to the QM/MM value of 120o as for the fast time delays above. Upon convergence of the refinement, φtail settled at 140o. Feature β1 that aligns across the tail axis (Fig. 2A, dashed line) is exactly reproduced (see also Fig. S12A). Clearly, the configuration of the chromophore is trans. On time-scales > 550 fs a transition to the cis configuration occurs (Fig. 2D-F, Fig. 3B and Fig. S9C-D). The last three time bins were combined to further boost the signal to noise ratio in the DED map (Tab. S1C, 800-1200fs bin, 157082 indexed patterns). The result is

29

shown in Fig. 3B and Fig. S9D. Feature β1 largely determines the configuration of the double bond. β1 is more aligned along the tail axis (compare Fig. 2E, dashed line). The positive feature β3 in Fig. S9D indicates a shift in the ring-phenolic oxygen. As a result (Fig. 2D-F), the ring shifts and the hydrogen bond to Glu46 is broken (3.7 Å). φtail is around 50o, close to that of the 3 ps structure (40o). Fig. S3 shows a schematic view of the trans to cis transition that involves this structure. Overall, on the longer fs time scale the structure PYPslow is already cis but further relaxation towards PYP3ps is yet to occur. Notably, the hydrogen bond to Glu46 is re-established at 3 ps. With these structures at hand, the trans to cis isomerization can be structurally understood by concerted rotations about the C1-C2 and the C3-C1’ single bonds as well as the C2=C3 double bond as outlined in Fig. S3.

Overall, the refinement is robust and does not rely heavily on the QM/MM torsional angle restraints. Indeed, the models were reproduced with a variety of starting torsional angles φtail and φtail restraints. For example the 455 fs model (Fig. 2C, with φtail = 145o) is reproduced using either 0o or 180o for the initial torsional angle and 90o (the QM/MM value right before the hop, Fig. S7) or 180o as weak torsional restraints. Similarly, the 455 fs model (which is trans) can be used as a starting model to reproduce the 3 ps model (which is cis, φtail=35o). Note: B-factors refined against extrapolated amplitudes are not reliable partly because the time-dependent X-ray data are scaled against the reference (dark) data and partly because the differences amplitudes are added to the dark structure factors using dark state model phases rather than their true (but unknown) phases. For more information on phases of difference structure factors, see Schmidt 2008 (58). Difference-difference (residual) maps.

Most importantly, the principal difference electron density features are faithfully reproduced by difference maps calculated from the refined models. To obtain residual maps, difference maps calculated from the refined models (|Fcalc(t)|-|Fcalc(dark)| difference maps) were subtracted from the observed weighted |Fobs(t)|-|Fobs(dark)| difference maps. The resulting difference-difference maps are shown in Fig. S10. The absence of strong signal in residual maps shown in Fig. S10A-B,F indicates a match between the model structure and the observed difference density. Conversely, the presence of strong features in residual maps where a wrong model was deliberately used confirms the inadequacy of that model (Figs S10C-E). Structural Comparison, Volumes, Radius of Gyration, Energies.

The PYP structures refined here were compared with structures from the literature (PDB-codes 4I38, 4B90, 1TS0; see Tab. 1). Local structural changes were determined by averaging the distances between equivalent chromophore atoms of the dark structure and those at each time delay ∆t using Coot (59). Global structural changes were determined by averaging the Cα − Cα distances of equivalent residues of the dark and the structures at each time-delay ∆t using ‘lx_lsqman’ (60) (Tab. S4). For protein volume determination a water accessible surface (61, 62) must be calculated (Fig. S4). The water accessible surface and the PYP volume were determined by a probe of 1.5 Å radius using the ‘volume assessor’ located at http://3vee/molmovdb.org/volumeCalc.php (63). The radius of gyration of PYP was calculated by the standard equation (eqn. S3) (64).

789 = ∑ ;<��=< − �='�9< ∑ ;<<⁄ , (S3) where Rg

2 is the square of the radius of gyration, mi is the mass of atom i, �=< is the coordinate of atom i, and �=' is the center of mass of the PYP molecule. Results are listed in Tab. 1.

30

The structural displacements for 9 carbon, 2 oxygen and 7 hydrogen atoms of the chromophore

(147 g/mol) are <0.7 Å> on the average after 250 fs. Using the classic equation � = �9 ?�

9, the

acceleration must have been 2 x 1015 m/s2. The final velocity (starting with v=0) is � = ?� =500B

5 . The force on the chromophore is accordingly & = C.�EF,8/B3H� ∙ 2�10�J B

5� = 4.88�10M�CN

or ~500 pN, with A the Avogadro number (6.022 x 1023 N/mol). The energy is the force times the displacement which gives 3.2 x 10-20 J or 0.2 eV (same result is obtained when the kinetic energy is considered with v = 500 m/s).

Supplementary Text Spectroscopic investigations on crystals to determine the maximum permissible laser power.

Spectroscopic investigations were performed on thin (2 – 3µm) crystalline material produced by squeezing a large PYP crystal between two cover slides (65). 900 fs, 300 fs and 100 fs laser pulses with varying energy density were used to start the photocycle. The spectra were probed after the laser pulse with a fast spectrometer (Fig. S5A-B; 180 µs time delay) or with a continuous wave laser source and photodiode detector (Fig. S5C-D; 1 ms time delay). Irreversible bleaching is an indication for damage (Fig. S5A). The percentage of molecules that could be brought into the photocycle without (or with only small) irreversible bleaching was up to 10% (50). Power densities up to 0.8 mJ/mm2 could be used with positively chirped 140 fs pulses, while keeping single-shot photobleaching below 0.5%. FROG Measurements.

The 450 nm laser pulses were characterized using a custom built SHG-FROG setup (27). A schematic of the device is shown in Fig. S6A. A beam splitter was used to split the 450 nm pulse into two replicas which were sent on two independent delay lines. One line had a motorized delay stage to control the temporal delay (O) between the two pulses. In order to reduce the amount of dispersion introduced by the FROG setup itself, the beamsplitter was a 2 µm thick pellicle (Thorlabs BP245B1) and focusing was performed using reflective optics. The replicas were focused into a 20 µm substrate mounted BBO crystal (θ= 55º, φ = 90º) (Mesaphotonics) which was inclined at about 13º to phase match SHG at 450 nm. When spatially and temporally overlapped in the crystal, it is possible for SHG to result from the conversion of one photon from each replica pulse. Due to the non-collinear geometry and momenta cancellation, this 225 nm signal (ISHG(PO)) is emitted on axis and easily separated from the fundamentals; it was measured using a custom-modified UV spectrometer. When recorded as a function of relative delay, this FROG signal can be written as

QRST�PO� = U Q���Q�� + O��M<V�. (S4)

When plotted as a function of spectra and delay this produced a “FROG trace” (Fig. S6B)

which was analyzed using the FROG code software package (J.Wong et al. http://frog.gatech.edu/). This program creates a theoretical FROG trace (Fig. S6C) using a test pulse and then iteratively alters the simulated pulse parameters to produce the best fit of the replicated to the experimental trace. When fitting is achieved below a suitable error, the algorithm provides full reconstruction of the pulse’s spectral and temporal profile including phase. It is also possible to recover the second order dispersion or group delay dispersion (GDD) and third order dispersion (TOD) by fitting a 3rd order polynomial to the spectral phase. Phases associated with spectral intensities smaller than 1/e of the maximum intensity were ignored as they have little

31

meaning and are mostly artifacts of the fitting. The FROG reconstruction of the pulses that were used to pump the PYP crystals is shown in Fig. S6D-E. The pulses were found to be 140±5 fs long with a GDD of ≈1000 fs2 and TOD ≈+1300 fs3. The large dispersion was expected due to the 8 mm of BK7 and 4 mm of fused silica in the beam path. Since the measured GDD for the un-stretched TOPAS output was less than +50 fs2, the GDD of ≈1000 fs2 is very close to literature value for glass introduced (≈+1100 fs2).

Our positively chirped laser pulses favor the generation of the non-stationary excited state (28), hence optimizing the quantum yield of the reaction. Ultrafast spectroscopy literature is in good agreement with the time-dependence of the dynamics observed here. The on-resonance excitation and sign and magnitude of second order dispersion minimizes non-linear cross-sections of excited state absorption and stimulated emission (5, 66-68). While maximizing S0 to S1 population transfer, the intense femtosecond pulse also generates vibrational coherence(41). Ultrafast vibrational spectroscopy on PYP has established the chromophore response on ultrafast time scale (42). Using femtosecond stimulated Raman spectroscopy, a Franck-Condon (FC) intermediate is observed at ~100fs (42), distinct from the major S1 decay process with a ~450 fs time constant. The FC state already shows chromophore distortions. The decay of this FC state with a ~ 170 fs time constant agrees with the structural changes seen in our experiment prior to excited state decay, occurring with ~ 450 fs time constant. Furthermore, the chromophore tail carbonyl out-of-plane vibration downshifts from 665 to 640 cm-1 concurrent with excited state decay which was interpreted as resulting from motion along the photo-isomerization coordinate (42).

QM/MM Calculations Chromophore force field

All interactions were modeled with the Amber03 force field (69). Bonded and Lennard-Jones parameters to model the chromophore were determined based on the similarity of the chromophore atoms to the atoms already defined in the force field. Equilibrium bond lengths and angles were taken from an optimized chromophore model in vacuum (deprotonated thiomethyl p-coumaric acid, TMPCA). The optimization was performed at the B3LYP/6-31G(d) level of Density Functional Theory(70, 71) with the Gaussian09 program (Gaussian, Inc., Wallingford). The partial charges of the atoms were derived following the procedure recommended for the Amber03 force field (69, 72). First, the geometry of TMPCA was minimized at the HF/6-31G** level of ab initio theory. To account for the effect of a protein environment in these calculations, the IEFPCM continuum solvent model(73) was used with a relative dielectric of 4.0.(69) After geometry optimization, the electrostatic potential at 10 concentric layers of 17 points per unit area around each atom was evaluated using the electron density calculated at the B3LYP/cc-pVTZ level of theory, again using the IEFPCM continuum solvent model with a relative dielectric of 4.0. The atomic charges were obtained by performing a single-stage RESP fit to the electrostatic potential without symmetry constraints (72). The recommended two-stage fit with symmetry constraints on the phenol ring was considered unnecessary, because a high rotation barrier prevented sampling configurations with a flipped phenol ring in our classical MD simulations. Also, the non-symmetric charge distribution of the phenol ring yielded a more accurate fit to the quantum mechanical electrostatic potential than the symmetrized distribution. Molecular dynamics simulation setup.

The starting coordinates of the PYP crystal were taken from the x-ray structure (PDB entry: 2ZOH) (74). Six copies of the protein, including 120 crystal waters, were placed inside the unit

32

cell with periodic boundaries and soaked in 5M NaCl solution by adding 909 TIP3P waters (75),

59 Na+ and 23 Cl-. The total system contained 16,553 atoms and was simulated classically for 100 ns. The classical MD simulations were run at constant volume and temperature of 300 K (76), with a time constant of 0.1 ps for the temperature coupling. The LINCS algorithm was used to constrain bond lengths (77), allowing a time step of 2 fs in the classical simulations. SETTLE was applied to constrain the internal degrees of freedom of the water molecules (78). A 1.0 nm cut-off was used for non-bonded Van der Waals’ interactions, which were modeled by Lennard-Jones potentials. Coulomb interactions were computed with the smooth particle mesh Ewald method(79), using a 1.0 nm real space cut-off and a grid spacing of 0.12 nm. The relative tolerance at the real space cut-off was set to 10-5. All force field simulations were performed in single precision with the Gromacs-4.5.3 molecular dynamics program (80).

In the excited state dynamics simulations the QM subsystem, which included part of the backbones of Pro68, Cys69, the sidechains of Tyr42, Glu46 and Cys69, and the pCA chromophore (Fig. S7) in one of the asymmetric protein units, was described at the CASSCF/cc-pVDZ level (81), with an active space of 12 electrons and 11 orbitals, i.e., the complete π-system of the chromophore. Surface hopping was used to model the excited state decay. As explained elsewhere

(82), hops are triggered when the quantum mechanical probability for a transition from S1 to S0 equals unity, which can only occurs when the classical trajectory hits the conical intersection hyper-line. Although such strict hopping criterion could underestimate hopping (83), we consider this an advantage if only a few trajectories can be computed. An additional advantage of our diabatic surface hopping approach is that we obtain structural information about the location of the S1/S0 seam in the protein. The remainder of the system, consisting of the apo-protein, 1,629 TIP3P water molecules (75), 23 Cl- and 59 Na+ ions in the periodic unit cell, was modeled with the AMBER03 force field (69). The five bonds connecting the QM and MM subsystems were replaced by constraints (77) and the QM part was capped with hydrogen atoms. The force on the cap atoms was distributed over the two atoms of the bond. Because the QM subsystem was mechanically embedded into the MM system (Fig. S7A), electrostatic and Lennard-Jones interactions between MM and QM atoms were added. The time step was reduced to 1 fs. Photon absorption was modeled as an instantaneous, on-resonance transfer of the system from the electronic ground state surface (S0) to the first singlet excited state (S1) potential energy surface.

Prior to the excited state non-adiabatic QM/MM MD simulations, the system was simulated at the MM level for 100 ns. From this 100 ns trajectory, three snapshots were randomly selected and further equilibrated for 5 ps at the QM/MM level, using the Restricted Hartree-Fock (RHF) method to model the electronic wave function of the QM subsystem (Fig. S7A). All QM/MM simulations were performed with the QM/MM interface between (80) and Firefly version 8.1.0 (84).

Photo-isomerization trajectory. The main results of the simulations are summarized in Tab. S5. In all three simulations the

primary photo-dynamics initiated by photon absorption involves a rapid rotation of the C2=C3 torsion into a 90° twisted S1 minimum (Fig. S7B, Fig. S8). After a short period in this S1 minimum, the system reaches the nearby S1/S0 conical intersection seam and decays to the electronic ground state (S0) surface. Back on the ground state surface, the chromophore continues isomerizing towards the cis conformation (a, c, Tab. S5). During this process, the hydrogen bond between the carbonyl of the chromophore (O1) and the backbone amino group of Cys69 (N) remains intact (Fig. S7B, Fig. S8). The chromophore may relax also into the trans configuration, restoring the

33

dark state of the protein. The number of trajectories is statistically small, but yield a consistent mechanism that is also in full agreement with previous simulations (20).

34

Figure S1. Snapshot density (diffraction patterns per 50 fs interval). The influence of a substantially varying number of patterns per bin on data quality is avoided when bins with approximately equal number of patterns per bin are used (e.g. red squares illustrate use of 16 bins; see Tab. S1. Unavoidably, these 16 bins are of non-uniform width in time).

35

Figure S2. A. Torsional angle φtail from refinement at various time delays (from bins containing 40 k diffraction patterns, see also Tab. S2B and S5). Dashed line: fit by eqn. S2 with a 590 fs transition. B. Right singular vectors (rSV) from the singular value decomposition of the 17 difference electron density maps on the fs (with 20 k diffraction patterns in each bin) to 3 ps time scale. Black dots: rSV1; black line: fit with exponential (eqn. S1) with characteristic time τ = 155 fs; blue and red dashed lines: fit of two separate parts of the rSV1 with exponentials. Red arrow: transition. Red squares: rSV2, dashed black line: fit by an exponential function (eqn. S2) with a 560 fs transition. Circle: time point near the assumed conical intersection. Green triangles: rSV3. The values of rSV4 and rSV5 are not significant, and displayed by thin dashed dotted and dashed double dotted lines. Blue bell-shaped curve: duration of the transition, from 470 to 650 fs. Inset: magnitude of the singular values.

36

Figure S3. Schematic view of the trans to cis transition at 550 fs. The rotations can be easily understood using an educator’s stick model set for chemical structures. A. PYPfast structure. To produce PYPfast from the reference, PYPref, grab the C2=C3 double bond with the left hand and the C1-S bond with the right hand. Rotate the double bond clockwise by about 40o. The C1-C2 single bond points now backwards as observed, the configuration is still trans. Compared to the dark state, the hydrogen bonds to Glu46 and Tyr42 are elongated. To further the isomerization (and reach B), continue rotating the double bond clockwise by an additional 90o by keeping the atom C2 location fixed. The entire head-structure pivots about (the fixed) atom C2. Since the head cannot swing out, concerted anti-clockwise rotations about the C1-C2 and the C3-C1’ single bonds are necessary to minimize the volume swept out by the atoms that move during isomerization; see arrows in A. B. PYPslow structure (800-1200 fs), cis. The hydrogen bond to Glu46 is broken (head displacement exaggerated for clarity).

37

Figure S4. The solvent accessible surface (blue) of PYP3ps.

38

Figure S5. Entry into the photocycle in PYP crystals as evident by optical absorption changes initiated by 900 fs (A-B), 300 fs, and 100fs (C-D) laser pulses. Pump-probe delay 180µs for A-B, 1 ms for C-D. A. PYP absorption spectra of the dark state, following a sequence of 10 pump-probe cycles with 900 fs pulses at increasing laser pulse energy densities (shown in the legend). Irreversible bleaching (damage), as evident by the reduced OD of the dark state, occurs at 2.2mJ/mm2. B. Light-dark difference absorption spectra at various laser pulse energy densities. Spectra are recorded by accumulating 10 pump-probe cycles. Extent of (reversible) bleach at 450 nm is roughly proportional to number of molecules in photocycle. At 1.2 mJ/mm2 about 5% of molecules can be pumped into the photocycle. C. Permanent single-shot photo-bleaching (damage) measured after 100 fs (red squares) and 300 fs (blue diamonds) laser pulses of various energies (spot size 80 µm). D. Extent of reaction initiation with 100 fs and 300 fs laser pulses measured as the pB concentration at 1ms as the reversible ∆A450nm/A450nm. 100 fs laser pulses with 2 µJ (energy density: 400 µJ/mm2) result in up to 6% reaction initiation with small permanent damage.

39

Figure S6. A. A diagram of the SHG-FROG used to characterize the 450nm femtosecond pulses incident on PYP crystals. The incoming pulse (black) is split into two replicas that travel down individual paths (shown in blue and red for clarity) until they are overlapped in a space and time in the 20 µm thin BBO crystal. The resulting FROG signal (pink) is measured using the UV spectrometer. B. The experimentally measured FROG traces for the 140 fs stretched pulse. C. Reconstructed FROG traces. The “G error” for the fitting was <1%. The high order structure seen at the top of the reconstruction that is not seen in the experimental data is the result of fluctuations in the shot-to-shot pulse shape. This is commonly seen in multi-shot FROG measurements and does not greatly impact the recovered pulse length or phase. D-E. FROG reconstruction of the temporal and the spectral intensity (red) of the reconstructed pulse incident on the PYP microcrystals with the retrieved phase indicated (blue). High order time-structure is a reconstruction artifact due to beam intensity instability (85, 86). The pulses were found to be 140±5 fs long with a GDD of ≈1000fs2 and TOD ≈+1300fs3.

40

Figure S7. A. PYP configurations along the QM/MM trajectory. The atoms included in the QM region are shown in ball-and-stick representation and modeled at the CASSCF(12,11)/cc-pVDZ level of ab initio theory. The rest of the system, consisting of the protein, water, Cl- and Na+ counter ions (most of which is not shown here) are modeled with the Amber03 force field. B. Snapshots from a QM/MM trajectory after photo-excitation at t = 0 fs (Tab. S4, trajectory a). C. Torsional angles as a function of time in this trajectory. Left yellow shaded region, dynamics on the excited state potential energy surface. Vertical dashed line (surface hop): seam that separates excited state structural dynamics (left) from ground state dynamics (right). D. Color code for the torsional angles of the pCA chromophore in C.

41

Figure S8. Transition through conical intersection as conceived from theoretical consideration (lower scheme) and viewed from the perspective of the TR-SFX experiment (structures above).

42

Figure S9. Trans to Cis isomerization in PYP. Direct comparison of difference maps and structures of individual time-points determined from 40,000 diffraction patterns with those determined from averaging diffraction pattern in larger time-bins (compare Tab. S1 B-C). The 269 fs, 799 fs and 3 ps delays (panels A,C and E) delays are also shown in Fig. 2 in the main text. Weighted DED maps in red (-3 σ) and blue (3σ); front (upper) and side view (lower). The reference, dark structure is shown in yellow throughout; structures before the transition and on the electronic excited state PES are shown in pink; structures after the transition and on the electronic ground state PES are shown in green. Important negative difference density features are denoted α, positive features as β. Pronounced structural changes are marked by arrows. A. Representative time-delay (269 fs) before the transition; some important residues are marked. Dotted lines: hydrogen bonds of the ring hydroxyl to Glu46 and Tyr42. B. Chromophore configuration (PYPfast) by averaging 100 fs to 400 fs pump-probe delays. C. Representative time delay (799 fs) after the transition. D. Structure PYPslow on longer fs times obtained by averaging the 800 fs to 1200 fs pump-probe delays. E. 3 ps chromophore configuration. Dashed lines in A, D and E: direction of feature β1.

43

Figure S10. Difference-difference maps (residual maps): Calculated |Fcalc(t)|-|Fcalc(dark)| difference maps were subtracted from the observed |Fobs(t)|-|Fobs(dark)| difference maps. Pink: PYPfast; green: PYPslow and PYP3ps; yellow: PYPref. Residual features (blue: positive, red negative) on the +/-3 σ level. A and B. Fast and slow fs time scale, residual maps using the correct, refined models that correspond to the time-scales, twisted-trans for A, and near-cis for B. C and D. Residual maps using interchanged, deliberately incorrect models, near-cis for C and twisted-trans for D. The chromophore tail is much cleaner in A and B, but strong residual features denoted α and β in C and D indicate mismatch. Note: remaining features near the sulfur emerge from its strong electron density; its position is difficult to model exactly. E. Data binned from 400-800 fs across the transition. Feature β shows some mixing-in of the PYPfast structure in this time range. F. residual map at 3 ps time delay.

44

Figure S11. PYPref (yellow), PYPfs,fast (pink) and the theoretical PYP model from QM/MM simulations (orange) overlaid on the 269 fs difference map (red/blue -3σ/4σ). A. front view, B. side view. Nearby amino acid residues are marked in A. Arrow in B: similar chromophore tail displacements for the observed and theoretical PYP structures.

45

Figure S12. Full view of the DED maps (contour: red/green -3σ/3σ) at various time-ranges. The maps are overlaid on PYPref (yellow) and the refined structures. Contrary to all other figures that depict PYP structures, the view is from the opposite (back) of the chromophore, hence the Arg52 is on the right. This orientation allows for a detailed view of the positive feature β1 which indicates the configuration of the chromophore and the strong feature on the Cys69 sulfur (β2) that is also associated with the cis configuration. A. On the fast time scale, feature β1 is strongly bent, β2 is absent. The DED map is interpreted with a trans configuration (structure in pink). B. The slow time range, feature β1 is aligned along the tail axis, which indicates a cis configuration (structure in green). C. At the 3 ps delay feature β1 is similar to B (structure also in green). D. Consistency check by randomly selecting ~80,000 snapshots from the 157,000 snapshots in the bin. DED map with features β1 and β2 and structure (green) essentially the same as in B.

46

Table S1. TR-SFX data statistics. Resolution 1.6 Å; data in the highest resolution bin (1.6 - 1.67 Å) in brackets. A. Statistics of the data sorted into time bins with 32,690 snapshots with Bragg reflections (hits)/bin (approximately 20,000 indexed snapshots/bin). Times shown are the average of the time stamps in each bin and their standard deviation. B. Data statistics of bins with 65,380 snapshots/bin (approximately 40,000 indexed snapshots/bin). C. Statistics of the individual -1ps, 3 ps and 200 ns time points and when snapshots with time-stamps 100 – 400 fs and 800 – 1200 fs are combined. A

Time # indexed R-split CC* Completeness Multiplicity

100.8 ± 43 fs 19313 12.84

(37.42)

0.994

(0.926)

99.84

(100)

342.76

(25.2)

182.8 ± 16 fs 19659 12.58

(38.60)

0.994

(0.925)

99.84

(100)

348.41

(26.3)

236.3 ± 16 fs 19647 12.93

(41.08)

0.994

(0.896)

100

(100)

341.26

(26.0)

301.5 ± 22 fs 19480 13.19

(36.77)

0.993

(0.933)

99.76

(100)

330.86

(25.3)

390.7 ± 31 fs 19090 13.70

(43.08)

0.993

(0.904)

99.92

(100)

312.83

(22.2)

521.0 ± 38 fs 18905 14.30

(35.20)

0.992

(0.925)

99.97

(100)

295.09

(30.5)

657.5 ± 37 fs 21500 11.83

(25.12)

0.994

(0.968)

99.84

(100)

409.97

(59.1)

739.5 ± 15 fs 22554 10.99

(21.89)

0.995

(0.978)

99.76

(100)

459.67

(68.2)

782.4 ± 11 fs 22816 10.95

(23.76)

0.995

(0.969)

99.84

(100)

468.66

(69.0)

814.9 ± 9 fs 22560 10.83

(23.70)

0.995

(0.972)

100

(100)

467.46

(64.6)

842.8 ± 8 fs 22507 10.97

(24.63)

0.995

(0.962)

99.84

(100)

463.5

(63.0)

47

870.1 ± 8 fs 22416 10.76

(24.18)

0.995

(0.971)

99.6

(100)

465.45

(60.0)

898.8 ± 9 fs 22266 10.54

(23.59)

0.996

(0.972)

99.84

(100)

466.38

(61.8)

931.7 ± 11 fs 22409

10.49

(21.76)

0.996

(0.976)

100

(100)

472.31

(64.4)

978.4 ± 16 fs 22732 10.44

(23.46)

0.995

(0.964)

99.84

(100)

478.35

(67.2)

1064.8 ± 45 fs 23605 10.43

(20.94)

0.995

(0.975)

99.92

(100)

501.62

(75.9)

B

Time # indexed R-split [%] CC* Completeness Multiplicity

142 fs 38606 9.01

(26.05)

0.997

(0.962)

99.92

(100)

714.75

(48.5)

269 fs 38786 9.16

(26.08)

0.997

(0.959)

100

(100)

695.25

(49.6)

455 fs 37563 9.86

(26.58)

0.996

(0.957)

99.84

(100)

637.31

(57.7)

699 fs 43617 7.53

(16.94)

0.998

(0.984)

99.92

(100)

931.65

(124.9)

799 fs 44892 7.57

(19.84)

0.998

(0.974)

99.84

(100)

961.51

(121.1)

856 fs 44470 7.61

(17.60)

0.997

(0.983)

99.92

(100)

955.96

(112.9)

915 fs 44180 7.58

(18.75)

0.998

(0.981)

99.84

(100)

927.25

(106.3)

1023 fs 45880 7.53

(17.19)

0.998

(0.978)

99.76

(100)

971.3

(123.7)

48

C

Time # indexed R-split [%] CC* Completeness Multiplicity

Reference

(-1 ps)

82214 7.69

(14.83)

0.998

(0.987)

99.93

(100)

991.51

(22.1)

100 – 400 fs 81327 6.13

(18.13)

0.998

(0.981)

100

(100)

1546.81

(115.6)

800 – 1200 fs 157082 4.03

(9.11)

0.999

(0.989)

100

(100)

3322.0

(145.9)

3 ps 76411 5.43

(14.36)

0.999

(0.989)

100

(100)

1685.8

(145.9)

200 ns 102790 5.02

(12.04)

0.998

(0.993)

99.9

(100)

2287.0

(244.2)

49

Table S2. Difference maps and refinement for selected structures.

Structure Dark 100 - 400fs 800 – 1200 fs 3 ps 200 ns

signal +/-

(sigma)a

6.5/6.9 5.9/8.0 7.38/8.98 18.1/18.2

Rcryst/Rfree 15.5/18.2 20.6/26.1 18.5/23.9 20.6/26.0 28.9/37.5

completeness 98.5 99.0 98.9 98.9 98.9

rmsd bonds d[Å] 0.020 0.028 0.020 0.026 0.019

rmsd angles [o] 2.0 2.6 1.97 2.7 2.0

φφφφtail 170 140 53 38 3/-8

a max signal identified in weighted difference maps in multiples of the noise level.

50

Table S3. Torsional angles and distances from structural refinement at various, individual time delays.

Time Delay [fs] 142 269 455 699 799 856 915 1023 3ps C1-C2=C3-C1’ (φφφφtail)

154 140 144 naa 50 57 66 63 35

O1-C1-C2=C3 20 -7 15 na 36 23 32 43 30 CB-S-C1-C2 -135 -166 -154 na -152 -152 -157 -165 -137 pCA-O4’ - Glu46-Oεεεε 2.61 3.46 3.49 na 3.2 3.36 3.62 3.3 2.94

a refinement not successful presumably due to the existence of a mixture of molecules relaxing to cis towards the photocycle and those relaxing back to the reference state. The time course of the φtail in the 8 fs bins can be fit by eqn. S2 with a transition time of 590 fs, see Fig. S2A.

51

Table S4. Global structural changes [Å] as measured by averaging the distances between cα atoms of equivalent residues.

Time

Scale

0 100 to

400 fs

800 to

1200 fs

few ps 100 ps to

1 ns

µµµµs

PYPdark PYPfast PYPslow PYP3ps pR0/IT pR

0 PYPdark 0

100 to

400 fs

PYPfast, 0.20 (0.09) 0

800 to

1000 fs

PYPslow 0.19 (0.09) 0.23 (0.11) 0

few ps PYP3ps 0.24 (0.10) 0.26 (0.13) 0.25 (0.13) 0

100 ps to

1 ns

pR0/IT 0.19/0.13

(0.15/0.10)

0.27/0.22

(0.14/0.11)

0.23/21

(0.14/0.11)

0.26/0.25

(0.13/0.12)

0.15 (0.08)a

µµµµs pR 0.13 (0.09) 0.23 (0.12) 0.20 (0.1) 0.24 (0.11) 0.17/0.16

(0.12/0.09)

0

ms pB 0.17

(0.12)

0.25

(0.13)

0.24

(0.14)

0.28

(0.13)

0.19/0.15

(0.10/0.08)

0.18

(0.12) a distances between pR0 and IT

52

Table S5. Excited state lifetimes (2rd column), and products (3th column) in the excited state QM/MM trajectories.

trajectory S1 lifetime outcome

a 311 fs cis

b 221 fs trans

c 456 fs trans

<329 fs>