Supplementary Figure S1. Comparison of the On -minus- Off difference spectrum of Dronpa (black) with the GFP R minus GFP A difference spectrum 16, 17

Transcription

1 Supplementary Figure S1. Comparison of the On -minus- Off difference spectrum of Dronpa (black) with the GFP R minus GFP A difference spectrum 16, 17 (red) aids mode assignments. The inset highlight the chromophore changes, which involve the cis anion and trans neutral structure for Dronpa, and the cis anion and cis neutral state for GFP. The general frequency regions for assigned vibrational modes in both difference spectra are indicated with main their mode character as in Table 1 and Supplementary Table S1. 1

2 Supplementary Figure S2. Selected spectra at pump-probe delays of the Off state at 200fs, 3ps, 10ps, 30ps and 100ps. The pump-on minus pump-off difference spectra clearly change from early to late delay times, evident from the frequency shift of bleach and induced absorption features. This was modelled as a single sequential transformation species 1 -> species 2, with the corresponding two basis spectra (blue and red; Panel A) and concentration profiles (blue and red; panel B) shown in Figure 3 in the main manuscript. 2

3 Supplementary Figure S3. Selected spectra at pump-probe delays of the On state at 3ps, 10ps, 100ps and 1ns. The pump-on minus pump-off difference spectra change in amplitude from early to late delay times, whereas no changes were observed in the frequency positions of the features. This was modelled as a biphasic (parallel) decay, with the corresponding two nearly identical basis spectra (blue and red; Panel A) shown in Figure 4 in the main manuscript. 3

4 Supplementary Table S1 Unscaled harmonic frequencies (cm -1 ) of 4 -hydroxybenzylidene-2,3-dimethylimidazolinone (HBDI) calculated at the DFT B3LYP/6-311+g(d,p) level after geometry optimisation. Structures were either Neutral (N), with protonation of the phenolic oxygen, Zwitterionic (Z), with protonation of the imidazolinone nitrogen, Cationic (C) or Anionic (A). Calculations were done for protonated ( 1 H 2 O) or deuterated ( 2 H 2 O) chromophore in cis and trans configuration, referring to the exocyclic C=C bond. Restrained optimisation and harmonic frequencies (R) were performed using redundant coordinates, freezing the 15 degree angle between the phenol and imidazolinone planes, as observed in the off-state crystal structure (2POX.pdb). Calculated intensities are shown between brackets (km/mol). The first and second columns match mode numbering and label of general mode character of Table 1 in the main manuscript. cis-n- 1 H2 O cis-n- 2 H2 O cis-a cis-z- 1 H2 O cis-z- 2 H2 O trans-a trans-n- 1 H2 O trans-n- 2 H2 O trans-c- 1 H2 O trans-c- 2 H2 O (trans-n- 1 H2 O )R (trans-n- 2 H2 O)R (trans-c- 1 H2 O )R (trans-c- 2 H2 O)R 1 C=O 1767 (351) 1767 (351) 1697 (366) 1758 (696) 1758 (696) 1690 (358) 1738 (181) 1738 (182) 1792 (207) 1792 (208) 1752 (178) 1752 (178) 1806 (186) 1806 (187) 2 C=C 1687 (300) 1687 (300) 1656 (37) 1677 (295) 1676 (331) 1654 (127) 1679 (252) 1679 (252) 1663 (75) 1662 (70) 1684 (259) 1684 (257) 1663 (67) 1661 (187) 3 phe (473) 1643 (512) 1623 (1277) 1643 (18) 1642 (52) 1628 (699) 1646 (374) 1645 (389) 1620 (644) 1613 (1317) 1647 (284) 1646 (304) 1618 (1010) 1615 (1388) 4 phe (47) 1613 (22) 1609 (1903) 1604 (1925) 1593 (113) 1615 (46) 1609 (45) 1601 (894) 1587 (138) 1619 (36) 1614 (19) 1604 (474) 1589 (42) 5 C=N/C=C 1601 (59) 1596 (46) 1595 (8) 1585 (112) 1575 (108) 1551 (1603) 1597 (37) 1592 (28) 1590 (92) 1587 (138) 1608 (0.4) 1603 (0.7) 1596 (11) 1582 (59) 6 phe (127) 1540 (136) 1570 (21200) 1567 (85) 1568 (239) 1529 (134) 1543 (106) 1541 (129) 1544 (100) 1542 (146) 1542 (96) 1540 (116) 1543 (83) 1541 (118) 4

5 Supplementary Note 1 Both in solution and in crystals, the On state which is stable in the dark has a high fluorescence quantum yield of 0.85, and shows absorption and emission maxima at 503 and 522 nm 3. Illumination with nm light photoconverts the ground state with a quantum yield of 3.2*10-4 1, 9 to a weakly fluorescent (fluorescence quantum yield < 0.02) Off state which was suggested to be protonated at the phenolic oxygen and absorbs maximally at 380nm 1. This state has been referred to as the A2 ground state (from here on called the Off state), to distinguish it from a protonated dark ground state that is stably formed at ph 5, designated A1 10. In turn the Off state can be photochemically re-transformed with illumination at ~ nm 1, and will also thermally return to the On ground state, with time constants ranging from hours (an 18h time constant was determined for Dronpa 3 ) to seconds depending on specific mutations 27. QM/MM simulations have been reported that investigated the possible reaction mechanisms in Dronpa 12. Li et al (2010) invoke a twisted intramolecular charge transfer state in the photoisomerisation reaction, linking excited state proton transfer with trans-cis photoisomerisation in a concerted manner. It is however not clear how the ESPT reaction would be linked with an excited state potential that allows high yield excited state trans-cis photoisomerisation 12. An argument against ESPT occurring for the Off state excitation is the 440 nm wavelength of fluorescence from the short lived excited state. Habuchi et al reported a 12 ps time constant for fluorescence emission detected at 440 nm with 390 nm excitation of the Off state 9. Although no further spectral information was given, this indicates that the Stokes shift for the Off state is considerably smaller than the Stokes shift for the neutral cis A ground state of avgfp 28. The discovery of ESPT in the fluorescence photocycle of avgfp crucially explained the size of the Stokes shift, which is in excess of 100 nm, as an energy term related to the loss of proton affinity in the optically excited state 28. This spectroscopic behavior was shown 28 to follow the Förster cycle that describes the red shifted fluorescence emission of photoacid molecules 29, 30. The 4 ps time constant observed in 1 H 2 O would therefore appear unexpectedly short in relation to its possible ESPT, which if present, would be associated with only a weak energy term based on 5

6 the observed wavelength of emission. The possibility of ground state proton transfer upon phototransformation of the trans and cis chromophore of Dronpa may alternatively be considered on the basis of the structural environment of the phenolic group in the Off and On structures. The cis On state environment includes a hydrogen bonding interaction with Ser 142 OH group, stabilizing the anionic phenolate form, whereas the trans Off state may form hydrogen bonding interactions indirectly to the Glu 144 carboxylate 5. The GFP-like protein asfp595 from the sea anemone Anemonia sulcata is another example of photoswichable fluorescent protein, having a Met-Tyr-Gly derived chromophore. On the basis of theoretical work the imidazolinone nitrogen is implicated to form a zwitterionic chromophore in asfp595 13, 14. These studies proposed that the chromophore is present as a zwitterion in the trans form and that excited state proton transfer proceeds from the imidazolinone nitrogen (rather than the phenolic oxygen). Schäfer et al. (2008) propose that excited state proton transfer in asfp595 deactivates the otherwise fluorescent zwitter-ionic trans chromophore, explaining the low fluorescence quantum yield 13, 14. Theoretical calculations evaluated the potential energy of excited states of GFP chromophore models in the different protonation states 31, 32, which suggest that the zwitterion has a conical intersection in a twisted geometry involving the rotation of only a single bond 32. Electrostatic calculations performed for the neutral chromophore of avgfp showed the increased acidity of the phenol oxygen and increased basicity of the imino nitrogen in the excited-state 33. For the neutral chromophore this agrees with a picture of charge migration in the excited state typical of photoacid behaviour, and questions the proposal by Schäfer et al 13, 14 who invoke deactivation of the excited zwitterion state via proton transfer. In avgfp the zwitterion was shown not to be present. For the free chromophore model compound ethyl 4-(4-hydroxyphenyl)methylidene-2-methyl- 5-oxoimidazolacetate, Bell et al. determined a pka value of 8.0 for protonation of the imidazolinone nitrogen 34 and showed from Resonance Raman spectroscopy measurements that in avgfp the position of the Raman bands exclude the possibility of the zwitterionic form 34. In the case of asfp595, the carboxylate from Glu215 is in hydrogen bonding contact with the imidazolinone nitrogen Consequently, QM/MM simulations suggest that the zwitterion is the dominant species. Such interactions are not observed in Dronpa, yet the spectroscopic transitions for Dronpa 6

7 and asfp595 are very similar. Yet another proposal discussed the possibility of intersystem crossing in the cis to trans photoisomerisation in Dronpa 6, 9 Theoretical work by Olsen et al (2010) indicated that photoisomerisation occurs at different location depending on the protonation state of the chromophore 38. Specifically the quantum chemistry calculations indicated excited state torsion of the phenoxy-bridge in the neutral and the imidazolinone-bridge dihedral angles for the anionic chromophore 38. 7

8 Supplementary Note 2 Concerning the FTIR difference spectroscopy of the photoswitching reaction in 1 H 2 O and 2 H 2 O, samples of Dronpa in 1 H 2 O and 2 H 2 O at p 1 H/p 2 H 7.8 were subjected to repeated green (> 495 nm) and blue (400nm) illumination, and FTIR difference spectra were recorded and averaged to yield high S/N difference spectra (Methods) Harmonic frequency calculations of the planar, geometry optimised HBDI yields an unscaled calculated frequency of 1656 cm -1 for the vibration observed at 1628/1615 cm -1 in Dronpa/GFP R, assigned to the 63 rd mode (Supplementary Table S1), and displays a strong C=C stretching character having contributions from phenol and imidazolinone deformations, in agreement with previously reported calculations 19, The frequency increase therefore indicates an increased bond strength for the exocyclic C=C bond for the on-state of Dronpa relative to GFP R. This likely results from the non-planar conformation of the cis anion chromophore, as observed in the On -state structure, displaying a ~7 angle between the phenolate and imidazolinone rings 3. The C=N mode, proposed to match mode 61, has an unscaled calculated frequency of 1595 cm -1 (Supplementary Table S1) and shows a strongly delocalised character having contributions from C=N stretching and phenolate deformation in addition to C=C stretching. The increase of the correponding observed frequency from 1539 cm -1 in GFP to 1547 cm -1 in Dronpa similarly points to an increased localisation and bond strengthening in the imidazolinone ring arising from the nonplanar chromophore configuration in the Dronpa structure. Geometry differences do however not directly explain the higher frequency of the phenol 1 mode in Dronpa. The description of these mode characters is however approximate as harmonic frequency calculations (Supplementary Table S1), as well as isotopic substitution studies of the HBDI model compound 39-41, indicate significant delocalisation in this spectral region. Considering for instance the displacements for calculations of the cis anion of HBDI C=C and C=N modes (calculated 1656 and 1595 cm -1 ; Table S1), there are significant similarities. Both involve symmetric deformation of the phenolate ring as well as C=N stretching in the imidazolinone ring, but the 1656 cm -1 mode has dominant C=C stretching character. The frequency upshift of both modes is 8

9 indicative of the non-planar, distorted structure observed in the On state crystal structure 3. In contrast, the phenol 1 mode is downshifted relative to the corresponding mode in GFP. Displacements for the 1623 cm -1 mode also involve symmetric phenol deformation as well as imidazolinone C=N stretching, but C=C stretching and phenolic C-O - stretching is noticeably less, and C=O stretching is increased. A possible suggestion is that the hydrogen bonding to the phenolic oxygen and imidazolinone C=O group account for this mode softening in the On state of Dronpa relative to the GFP environment. As in GFP A, the Off state of Dronpa is characterised by relatively weak bands compared to the intensities associated with the on state (Supplementary Fig. S1). This is consistent with the neutral state of the chromophore in both fluorescent proteins. Furthermore, the band positions associated with chromophore modes also support the assignment to the protonated species (Supplementary Table S1). For this argument, of particular note is the expected band position of the C=O stretching mode. For the neutral trans chromophore, harmonic frequency calculation at the DFT level yields a frequency of 1738 cm -1 for this mode, compared to the 1697 cm -1 frequency for the cis anion, thus predicting a 41 cm -1 downshift for the off->on reaction (Supplementary Table S1), in reasonable agreement with the experimental observation of a 25 cm -1 downshift from 1690 to 1665 cm -1 (Table 1, Fig. 2). The C=O mode is calculated at 1690 and 1792 cm -1 for the trans anion and trans cation (Supplementary Table S1), which are too low and too high respectively to explain the experimental measurements. Band positions calculated for the trans neutral state (Supplementary Table S1) are in reasonable agreement with the proposed assignments (Table 1). It is noted, however, that in the crystal structure of the Off state (2POX pdb; 5 ) the chromophore is present in a twisted geometry, having a ~23 angle between the phenol and imidazolinone rings 5. The increased nonplanarity of the chromophore in Dronpa may contribute to a somewhat lower degree of correspondence between the IR signals from the neutral chromophore in Dronpa and GFP (Supplementary Fig. S1). Using redundant coordinates to frieze this angle, optimisation and frequency calculation indicates a general trend to upshift the band positions of chromophore group frequencies, as expected from the bond hardening generally (Supplementary Table S1). Finally, the neutral trans chromophore in the Off state is supported by TD-DFT calculations of the isolated electronic transitions, 9

10 providing 2.85, 3.31 and 2.97 ev for the trans anion, trans neutral and cis anion state, respectively. These values are in reasonable agreement with previously reported calculations at the CASPT2 level 12 2 H/ 1 H isotope substitution results in frequency shifts of both On state and Off state associated chromophore bands (Fig. 2, Table 1). A scaled subtraction of the onminus-off difference spectra recorded in 1 H 2 O and 2 H 2 O further visualises that both the On and the Off states show bands whose positions are sensitive to 1 H/ 2 H isotope substitution (Fig. 2, top panel). This is somewhat unexpected for the On state, which is assigned to a cis anion structure although the GFP R state of GFP, which is also assigned to the cis anion, also shows modest 2 H/ 1 H isotope substitution sensitivity 16. As explained in the main manuscript results section, an alternative assignment to a zwitterion is unlikely, as the bond order difference would downshift the C=O mode considerably, calculated at 1758 cm -1 (Supplementary Table S1). Therefore, the hydrogen bonding of the protein environment is put forward to explain the isotope substitution sensitivity, particularly the interactions with the imidazolinone carbonyl 3. Harmonic frequency calculations that include either an Arg sidechain or a water molecule interacting with the imidazolinone carbonyl, with either 1 H or 2 H isotopes, reproduce sensitivity of the chromophore mode frequencies to substitution (not shown). For the Off state, calculations generally agree with the substitution character, although a downshift of the C=O mode is not predicted (Supplementary Table S1). Hydrogen bonding may similarly explain the isotope substitution effect for the off state. The exclusion of the possibility that the On state is a zwitterion, with protonation of the imidazolinone nitrogen atom (Fig. 1), is of particular interest as the zwitterions was implicated in the very similar photoswitchable fluorescent protein asfp595 13, 14. It is noted that in the case of the cis anion in GFP, lesser 1 H/ 2 H isotope substitution sensitivity was observed, which was taken as support for the assignment of the anionic state 16. TD-DFT calculations result in an electronic transition energy of 2.90 and 2.97 ev for the zwitterions and anion, respectively, at the B3LYP/6-311+g(d,p) level, in reasonable agreement with previously reported calculations at the CASPT2 and SAC-CI level 12. In addition, similar oscillator strengths are calculated, and therefore the visible absorption frequency may not definitively decide on the 10

11 protonation state of the chromophore in the On state. However, harmonic frequency calculation of the zwitterion chromophore predict an anomalous upshifted frequency for the C=O mode at 1758 cm -1 (Supplementary Table S1), contrary to observations (Fig. 2). Similarly, the calculated band positions of other chromophore modes are upshifted relative to the anionic species. Considering the high level of similarity of the GFP and Dronpa spectra (Fig. 3), an assignment to the anionic cis chromophore is thus best supported. The apparent 1 H/ 2 H isotope substitution sensitivity must therefore result from the hydrogen bonding interactions with the protein environment. This could be verified from harmonic frequency calculations of a model that includes HBDI in hydrogen bonding contact with an Arg (CN 3 H + 5 ) group, but which required redundant coordinates to fix the orientations of the groups. Two isotope-sensitive induced absorption bands belonging to the On state may be identified as not originating from chromophore modes, at 1674 and 1609 cm -1 and 1655 and 1594 cm -1 in 1 H 2 O and 2 H 2 O, respectively (Fig. 2, Table 1). Considering all possible assignments to protein groups, for the 1674/1655 cm -1 ( 1 H 2 O/ 2 H 2 O) band are Gln,ν(C=O), Asn,ν(C=O) or Arg,ν asym (CN 3 H + 5 ) 20. One possible assignment for the 1609/1594 cm -1 ( 1 H 2 O/ 2 H 2 O) band is Arg,ν sym (CN 3 H + 5 ), whereas other possible assignments to protein sidechains are less likely considering particularly the 1 H/ 2 H isotope substitution effect 20. The 15 cm -1 downshift for this mode is very clearly observed due to the isolated nature of the induced absorption band (Fig. 2). Thus, since both features point to arginine, an assignment to the symmetric and antisymmetric stretching modes of the protonated form of an arginine sidechain to these bands is tentatively possible on the basis of the FTIR data and the isotope substitution behaviour alone. We propose that the 1655 cm -1 On state band in 2 H 2 O contains contributions from both the Arg,ν asym (CN 3 H + 5 ) and chromophore C=O, explaining also its increased intensity relative to the 1665 cm -1 band in 1 H 2 O (Table 1, Figure 2) Next, the crystal structures of the On state (2IOV; 3 ) and Off state (2POX; 5 ) can be considered. As is shown in Fig. 1, two aminoacid side chains undergo dramatic structural change as a result of photoswitching, identified as Arg 66 and His 193. In the On state, Arg 66 is hydrogen bonded to the imidazolinone carbonyl, whereas this 11

12 interaction is absent in the Off state. Whereas no possible signals for histidine may be identified in the FTIR difference spectra, an assignment of the 1674/1655 and 1609/1594 cm -1 ( 1 H 2 O/ 2 H 2 O) bands to ν asym (CN 3 H + 5 ) and ν sym (CN 3 H + 5 ) of Arg 66 are well supported already individually on the basis of the spectral characteristics and furthermore strongly suggested on the basis of the crystal structures. In addition, the hydrogen bonding interaction of Arg 66 with the imidazolinone carbonyl can be put forward to explain the 1 H/ 2 H substitution sensitivity of the 1665 cm -1 C=O frequency (Fig. 2). No corresponding bands may be identified for both modes in the Off state, + suggesting that either deprotonation of the CN 3 H 5 group occurs, or the intensity is reduced. Possibly, the sizeable 1547 (+)/1518(-) cm -1 difference feature in 2 H 2 O may correspond to an arginine NH bending mode, which is not observed in the corresponding spectral region in the 1 H 2 O measurements (Fig. 2). Inspection of the crystal structure in the Off state shows the side chain of Arg 66 in a conformation in 3.12 Å distance from the Glu 211 carboxylate as well as in 2.98 Å distance from the Glu 144 carboxylate 5. The Arg 66 CN 3 H + 5 group is therefore highly likely to be protonated in the Off state as well as in the On state, and the change of the environment is largely responsible for the increased intensity of both modes seen as induced absorptions associated with the on state. The absolute frequencies of both stretching modes identify the interaction of Arg 66 with the chromophore imidazolinone carbonyl to be only weakly ionic in character in the on state 20. Finally, the absence of bands in the cm -1 ( 1 H 2 O/ 2 H 2 O) indicates that protonation changes of carboxylates, either aspartate or glutamate, are not involved in the photoswitching reaction, in contrast with GFP 17-19, 42. It is therefore likely that the protonation and deprotonation of the phenolic oxygen occurs with exchange to bulk water, and is driven by pka changes resulting from the electrostatics changes around the phenol ring, as previously argued 5. 12

13 Supplementary Note 3 Concerning the ultrafast VIS-pump IR-probe spectroscopy of the Off state of Dronpa, under continuous illumination with 473nm light, the Off state is generated on the time-scale of minutes, allowing pump-probe measurements with excitation at 400nm For Off state transient absorption measurement with 400 nm pump wavelength with 4.2 W/cm 2 average power density was used. In this case special care had to be taken to compensate for photoconversion by pump light. Thus additional backlight consisting of cyan LED (495 nm central wavelength) and a continuous wave solid state laser (473 nm, 25 mw, approx. 125 mw/cm 2 ) were used. Furthermore, the sample was translated at the highest speed (approx. 75 µm/s) allowable to minimise subsequent pump pulse overlap. Yet in addition to that rest periods between separate delay points had to be taken to allow for the sample to return to dark state. After several tests it was found that 1 s data collection at one delay point and 60 s waiting period to allow recovery of the Off state, ensured a constant pump-probe amplitude of the Off state between acquisitions. As explained in the main manuscript results section, the primary photoproduct of the Off state is identified as the cis neutral ground state. In 2 H 2 O, the C=O stretching mode is predicted to be upshifted by 29 cm -1 for the neutral cis chromophore, and 41 cm -1 downshifted for the anionic cis chromophore (Supplementary Table S1). The C=C mode is predicted to be upshifted by 8 cm -1, and downshifted by 21 cm -1 for the neutral and anionic cis chromophore, respectively (Supplementary Table S1). Furthermore, the phenol 1 mode predicted to be downshifted by 2 cm -1 for both the neutral and anionic cis chromophore, respectively. Therefore, the upshifted frequency of the C=C mode is additionally be taken as an additional argument against the possibility of deprotonation in the 10 ps photoproduct. It is noted that the band positions of the ground state bleach features in the 10 ps excited state decay component match reasonably well with the band positions from steady state FTIR difference spectroscopy, except for the 1600 cm -1 bleach feature in the time-resolved difference spectrum (Fig. 3). The latter observation may suggest that this band could belong to the phenol 2 mode of the trans neutral chromophore, 13

14 while little signal is observed in the FTIR difference spectrum. Band positions for the C=O, C=C and phenol 1 mode are however seen at nearly the same positions. Spectral calibration of the dispersive ultrafast infrared measurements was matched to better than 1 cm -1 with the FTIR instrument (Fig. 3). Thus, the upshifted band position of the C=C mode, in addition to the lack of intensity which develops only on long time scale, support an assignment to a neutral cis chromophore in a singlet ground state configuration. Measurements at 1ms delay, from a negative pump-probe time delay, indicated little spectral evolution between 100ps and 1ms (not shown). Thus, the thermal ground state deprotonation of the phenolic oxygen likely takes place in the ms-s time scale, as real-time formation of the On state may be observed from ensemble photoconversion measurements with 100ms time resolution (not shown). 14

15 Supplementary Note 4 Concerning the ultrafast VIS-pump IR-probe spectroscopy of the On state of Dronpa, samples were prepared for pump-probe transient absorption spectroscopy by concentrating to 1.3 mm in 2 H 2 O and a 12 um path length, optimised for transient absorption spectroscopy. For On state transient absorption measurement the wavelength of pump was chosen to coincide with peak absorption at 503 nm. The pump radiation average power density on the sample was 1.8 W/cm 2. To make sure the sample was fully in light state additional UV LED backlight ( nm centre wavelength) was used. As photoconversion quantum yield is low compared to fluorescence channel, moderate sample translation speed was sufficient, mostly to avoid sample photodamage. Transient absorption measurements of the On state were preformed with delays ranging from -1 ps to 1 ns (Supplementary Fig. S3). Free Induction Decay analysis of the negative time data indicated dephasing times smaller than 500fs for the main signals, but were difficult to determine due to spectral congestion and complex instrument response and coherent artefact. As discussed in the main manuscript, it is proposed that eventual protonation of the phenolic oxygen to form the Off state follows cis-trans photoisomerisation, which occurs with low quantum yield of 3.2*10-4. It is not likely that the phenolic chromophore of Dronpa acts as a photo-base. We can consider whether it is alternatively possible that excited state cis-trans isomerisation of a transiently protonated chromophore is highly favoured to that of the excited anion such that the low quantum yield of On to Off switching is linked to infrequent transient protonation. We estimate an intrinsic time constant of 11 microseconds for the On to Off transition. Assuming an upper limit of 1 for the quantum yield of isomerisation of a protonated singlet cis chromophore, the pseudo first order rate of protonation would be 10 5 s -1. Assuming a pka of 8 this would suggest a thermal deprotonation rate of s -1 (100 fs), and this implies a second order bimolecular rate constant of M -1 s -1 if the sample is at ph 6.0. While the bimolecular rate constant would not be unreasonable, a 100 fs thermal deprotonation would be unrealistic. Considering that this assumes unity for the photoisomerisation quantum yield, the time constant 15

16 would be even shorter. Also, photoisomerisation should complete within this time in order for the above argument to hold. Finally, we note a steady state Raman spectroscopy study of the cis and trans forms of photoswitchable mutants of avgfp which presented measurements and harmonic frequency calculations of chromophore models in both forms 43. This work identified a prominent frequency dependence of a Raman active mode upon phototransformation in the 700 cm -1 region assigned to a planar mode mainly localized on the bridge, on the heterocyclic ring, and on the methyl groups and that it involves a deformation of the N-C=C-C dihedral angle. In contrast, it was difficult to resolve the crucial C=O modes especially in the trans form. The response of the 700 cm -1 feature presents additional opportunities for time resolved Raman and possibly infrared studies of photoswitching of Dronpa and other photoswitchable fluorescent proteins. 16

17 Supplementary References 27. Ando, R., Flors, C., Mizuno, H., Hofkens, J. & Miyawaki, A. Highlighted generation of fluorescence signals using simultaneous two-color irradiation on Dronpa mutants. Biophys J 92, L97-99 (2007). 28. Chattoraj, M., King, B.A., Bublitz, G.U. & Boxer, S.G. Ultra-fast excited state dynamics in green fluorescent protein: multiple states and proton transfer. Proc Natl Acad Sci U S A 93, (1996). 29. Förster, T. Fluorezenzspektrum und Wasserstoffionenkonzentration. Naturwiss 36, 186 (1949). 30. Förster, T. Die ph-abhangigkeit der fluoreszenz von naphthalinderivaten. Z. Electrochem 54, 531 (1950). 31. Weber, W., Helms, V., McCammon, J.A. & Langhoff, P.W. Shedding light on the dark and weakly fluorescent states of green fluorescent proteins. Proc Natl Acad Sci U S A 96, (1999). 32. Polyakov, I.V., Grigorenko, B.L., Epifanovsky, E.M., Krylov, A.I. & Nemukhin, A.V. Potential Energy Landscape of the Electronic States of the GFP Chromophore in Different Protonation Forms: Electronic Transition Energies and Conical Intersections. Journal of Chemical Theory and Computation 6, (2010). 33. Scharnagl, C. & Raupp-Kossmann, R.A. Solution pk(a) values of the green fluorescent protein chromophore from hybrid quantum-classical calculations. 108, (2004). 34. Bell, A.F., He, X., Wachter, R.M. & Tonge, P.J. Probing the ground state structure of the green fluorescent protein chromophore using Raman spectroscopy. Biochemistry 39, (2000). 35. Quillin, M.L. et al. Kindling fluorescent protein from Anemonia sulcata: darkstate structure at 1.38 A resolution. Biochemistry 44, (2005). 36. Wilmann, P.G. et al. The 2.1A crystal structure of the far-red fluorescent protein HcRed: inherent conformational flexibility of the chromophore. J Mol Biol 349, (2005). 37. Andresen, M. et al. Structure and mechanism of the reversible photoswitch of a fluorescent protein. Proc Natl Acad Sci U S A 102, (2005). 17

18 38. Olsen, S., Lamothe, K. & Martinez, T.J. Protonic gating of excited-state twisting and charge localization in GFP chromophores: a mechanistic hypothesis for reversible photoswitching. J Am Chem Soc 132, (2010). 39. Schellenberg, P., Johnson, E., Esposito, A.P., Reid, P.J. & Parson, W.W. Resonance Raman Scattering by the Green Fluorescent Protein and an Analogue of Its Chromophore. J Phys Chem B 105, (2001). 40. Esposito, A.P., Schellenberg, P., Parson, W.W. & Reid, P.J. Vibrational spectroscopy and mode assignments for an analog of the green fluorescent protein chromophore. J Mol Struct 569, (2001). 41. He, X., Bell, A.F. & Tonge, P. Isotopic Labeling and Normal-Mode Analysis of a Model Green Fluorescent Protein Chromophore. J Phys Chem B 106, (2002). 42. van Thor, J.J. & Sage, J.T. Charge transfer in green fluorescent protein. Photochem Photobiol Sci 5, (2006). 43. Luin, S. et al. Raman study of chromophore states in photochromic fluorescent proteins. J Am Chem Soc 131, (2009). 18