The Desensitized Channelrhodopsin‐2 Photointermediate Contains 13 ‐cis, 15 ‐syn Retinal Schiff Base

Abstract Channelrhodopsin‐2 (ChR2) is a light‐gated cation channel and was used to lay the foundations of optogenetics. Its dark state X‐ray structure has been determined in 2017 for the wild‐type, which is the prototype for all other ChR variants. However, the mechanistic understanding of the channel function is still incomplete in terms of structural changes after photon absorption by the retinal chromophore and in the framework of functional models. Hence, detailed information needs to be collected on the dark state as well as on the different photointermediates. For ChR2 detailed knowledge on the chromophore configuration in the different states is still missing and a consensus has not been achieved. Using DNP‐enhanced solid‐state MAS NMR spectroscopy on proteoliposome samples, we unambiguously determined the chromophore configuration in the desensitized state, and we show that this state occurs towards the end of the photocycle.


Introduction
Microbial rhodopsins are heptahelical membrane proteins with ar etinal chromophore covalently bound to aconserved lysine in helix G. Al arge variety of light-driven functions is carried out by this protein family comprising proton and ion pumps,channels as well as sensors. [1] In the dark-adapted state the chromophore is either purely in the all-trans,15-anti configuration or am ixture of configurations is observed. Some microbial rhodopsins show light-adaption which refers to light-induced changes in the protein that remain even after the light has been switched off for some time.U sually,t he function of the protein is conveyed by the photoreaction that starts from the all-trans,15-anti chromophore configuration. Illumination leads to retinal isomerization around the C13 = C14 double bond resulting in a13-cis,15-anti configuration in the first photointermediate.T he system relaxes then via several photointermediates to the initial dark state with the all-trans,15-anti chromophore.T he details of this photocycle vary depending on the protein and its function. Knowing which (photo)-intermediates are adopted during the photocycle is aprerequisite for understanding the protein function as this varies between the different types of microbial rhodopsins.
Here,w ef ocus on Channelrhodopsin-2 from Chlamydomonas reinhardtii (ChR2). [2] ChR2 is acation channel and has found wide spread application in optogenetics. [3] Ac rystal structure of the dark state of ChR2 has been determined but not of the photointermediates so far. [4] These were investigated though with different spectroscopic techniques providing information about the changes of the protein and the chromophore compared to the initial dark state.Inthe darkadapted state (ChR2 470 ), the protein has an absorption maximum at 470 nm and the retinal Schiff base chromophore is in the all-trans,15-anti configuration. [5] Upon illumination, initial photoisomerization of the C13 = C14 bond leads to the first photointermediate (P 1 500 ). Schiff base de-protonation of P 1 500 results in at least two deprotonated states (P 2a 390 and P 2b 390 )and is accompanied by channel opening. [6] Thechannel remains open during Schiff base re-protonation (P 3 520 )b ut closes before the initial dark state is reached again. During continuous illumination, desensitization of the protein is observed and full photocurrents are only recovered after prolonged time in the dark. [2] This is unique to channelrhodopsins in the microbial retinal family and usually unwanted in optogenetic applications.U nderstanding the channelrhodopsin photocycle in general and desensitization in particular is therefore of general interest to the biophysical community.
Desensitization is associated with the non-conducting P 4 480 photointermediate which itself is photo active. [5,7] No consensus has been achieved with respect to the position of P 4 480 in the photocycle and its chromophore configuration. Based on FT-IR spectroscopy it was concluded that P 4 480 contains an all-trans,15-anti chromophore whereas aR esonance Raman study indicated that the chromophore configuration of P 4 480 is 13-cis,15-syn. [8] In this work, [8b] it was also postulated that P 4 480 is generated directly from the dark state via ap hoto reaction. This is in contrast to ap reviously published photocycle model in which P 4 480 is only formed during the open state decay. [7] Ac omparison of the different models and P 4 480 configurations is given in Figure S1. To resolve these contradicting views,i ti sn ecessary to characterize the chromophore in the desensitized state of ChR2 in detail. In principle such information can be obtained from crystallographic data. However,t odistinguish different chromophore configurations the resolution has to be extremely high and conditions have to be found under which the photointermediate can be studied. From the dark ChR2 X-ray structure,the configuration of the chromophore could not be determined. Retinal extraction experiments and Resonance Raman spectroscopy had hinted towards am ixture of alltrans,15-anti and 13-cis,15-syn chromophore. [9] However,t he first technique is invasive and the second often suffers from problems with band assignments.U sing solid-state NMR spectroscopy of ChR2 in lipid bilayers circumvents both problems.M aking use of signal-enhancement by dynamic nuclear polarization, we and others could unambiguously show that in the dark the chromophore is purely in the alltrans,1 5-anti configuration. [5,10] With this technique it is also possible to study the chromophore in photointermediate states as long as they can be cryo-trapped as demonstrated by us before. [5] As low temperatures are required for cryotrapping and ah igh detection sensitivity is desired, the experiments are ideally suited for sensitivity-enhanced MAS-NMR spectroscopy based on dynamic nuclear polarization (DNP). [11] In Figure 1a our experimental setup is shown where DNP enhanced solid-state NMR has been combined with in situ sample illumination in the optical range.
Cryo-trapping of aphotointermediate is possible when the energy barrier for its generation is lower (or can be overcome by photoexcitation) than the energy needed for its decay.W e have visualized this in Figure 1c for ChR2 with afree energy diagram based on our previous work. [5] Thep hotointermediate that can be trapped upon illumination at temperatures between 100 and 190 Kw as assigned to P 1 500 and always occurs in am ixture with ChR2 470 .T his assignment was confirmed by optical spectroscopy under cryogenic condition. Another photointermediate can be generated by rising the temperature above 200 K. Thes ame photointermediate,i n amixture with the dark state,isobtained by freeze quenching as ample that has been continuously illuminated at room temperature.T his intermediate was assigned to P 4 480 as it is the state that will be enriched during continuous illumination due to its long life time.T he same NMR signals are obtained when illuminating the sample at 245 K. Interestingly,i nt his case the ground state population is completely depleted but additional signals occur. This new photo intermediate is ap hoto product of P 4 480 .A si ti sn ot known to which of the several postulated P 4 480 photo intermediates the cryo-trapped intermediate corresponds,itwas termed P x .
The 13 Cchemical shifts of the retinal Schiff base chromophore are very sensitive to the configuration of the chromophore:T he C12 chemical shift is ar eadout for the configuration about the C13 = C14 bond whereas the C14 chemical shift is directly related to the C15 = Nb ond configuration. [13] C12 and C14 are significantly shielded in the 13-cis and 15-syn configurations,respectively.Inorder to resolve any ambiguity about the chromophore configuration in the photointermedi-ates,weanalyzed their chemical shifts,and we determined the distance between C12 and C15, which is significantly shorter in the 13-cis compared with the all-trans configuration ( Figure 1b). In addition, we performed at hermal relaxation experiment to elucidate the position of P 4 480 within the photocycle.All experimental details are given in the SI.

Results and Discussion
Ford etailed analysis of the retinal Schiff base chromophore in ChR2, [12,15-13 C 2 ]-all-trans-retinal-ChR2 proteoli- Microwave, optical and radiofrequency irradiation from agyrotron, an LED and the NMR console, respectively,c an be applied simultaneously to the sample spinning with 8kHz at the magic angle at cryogenic conditions. The sample consists of ap roteoliposome pellet containing channelrhodopsin-2,P DB code 6EID, [4] which is surroundedb ythe polarizing agent AMUPol. b) Illustration of the isotope labelling Scheme and comparison of the C12 and C15 distances in all-transretinal and 13-cis-retinal. Distances are taken from the X-ray structures of crystalline all-trans-retinal and 13-cis-retinal, [12] respectively. [ and P x is obtained. Heating the sample which was illuminated at 170 K to 245 Ki nthe dark causes thermal relaxation and leads to amixture of P 4 480 and ChR2 470 .T he color of the circles corresponds to the color of the respective NMR-spectrum in Figure 2.
posomes were prepared. Thep roteoliposomes were doped with the radical AMUPol to enable DNP-enhanced NMR experiments. [14] Theo btained signal enhancement was 50 ( Figure S2). Theassociated dramatic improvement turned out to be crucial for the NMR experiments on the cryo-trapped sample in order to deconvolute the signals of the different photointermediates.F urthermore,t he long-distance double quantum filtered experiments described below would not have been possible without the enhancement provided by the DNP experiment. 13 C-cross polarization spectra were recorded using three different illumination schemes:W ithout exposure to light, 470 nm illumination at 170 Ka nd 470 nm illumination at 245 K ( Figure 1d). Thes pectra are shown in Figure 2aand clear differences are observed around 165 ppm, 136 ppm, 132 ppm and 124 ppm. However,t he spectra are dominated by the natural abundance contribution from the protein, the lipids and spinning side bands of the glycerol signal. Double quantum filtering should be able to suppress these signals.The large distance between 13 C12 and 13 C15 and the large chemical shift anisotropy (CSA) of these atoms make these experiments challenging.T he standard Post-C7 experiment used in our previous study for aone bond distance did not work for this sample. [5,15] Therefore,weresorted to the CSA compensated SR26 sequence. [16] This experiment, to our knowledge,has so far been applied only to small molecules at ambient temperature.H ere we show that DNP-enhanced solid-state NMR enables the application of this sequence to apair of quite distant atoms in the chromophore incorporated in am embrane protein. In addition, we could also quantify these long distances.Although some natural abundance signal intensity remains,t he 13 C12 and 13 C15 signals can be much better identified in the spectra from the SR26 experiment compared to the CP spectra (Figure 2b-d), especially in the region around 123 ppm which shows ab ackground signal in the CP spectra which overlaps with the 13 C12 signal in the spectra of the illuminated samples.T he assignment of the 13 C12 and 13 C15 signals in the spectra is described in the SI and given in Table 1together with the assignment of 13 C14 which was obtained in our earlier work and recalled here to aid the analysis of the chromophore configuration. [5] The 13 C12 chemical shift clearly changes (À13.6 ppm) when trapping the first photointermediate,P 1 500 ,a sexpected for the isomerization around the C13=C14 bond. Thee ffect on the 13 C14 chemical shift in this intermediate is only minor showing that the C15 = Nb ond remains in an anti-configuration. Interestingly,the desensitized state,P 4 480 ,shows avery similar chemical shift for 13 C12 compared with P 1 500 indicating aC 13=C14 cis-configuration in P 4 480 .I na ddition, the 13 C14 chemical shift shows al arge shielding,w hich is expected for aC 15=N syn-bond. P x ,t he photointermediate of P 4 480 ,a lso shows an increased shielding of 13 C12 and 13 C14 but the effect is smaller compared to P 4 480 and interpretation is more ambiguous.
As additional support for the C13-cis configuration in P 4 480 and to further elucidate the chromophore configuration in P x , we wanted to take advantage of the sensitivity of the 13 C12-13 C15 distance to the C13=C14 configuration. This distance is shorter in the C13 = C14-cis confirmation (2.9 )than in the C13 = C14-trans configuration (3.8 )a nd can be used to distinguish them (Figure 1b). [12] To prove the feasibility of this approach we measured the 13 C12-13 C15 distance in free retinal using SR26 build-up experiments.W ed etermined the distance from fitting the experimental data to SR26 build-up curves generated using the Simpson software (Figure 3a). [17] Thecurves differ significantly from each other and the fitted distances correspond well to the expected distances in both molecules.W er epeated the experiment on ChR2 470 (Figure S3). As ufficient signal-to-noise ratio was difficult to obtain in this experiment and is an even greater challenge in the experiment with the illuminated samples as the signal  [a] 13 C14 chemical shifts were taken from our previous publication. [5] intensity is split between different states.Therefore,based on the free retinal build-up curves we picked only four different build-up time points (3, 5, 7a nd 9ms) by which the all-trans and 13-cis curves can be clearly distinguished. In this way we could record the SR26 spectra with asufficient signal-to-noise ratio to deconvolute the spectra and obtain the signal intensities (for the deconvolution see Figure S4 and supplementary section 1f or the details of the data analysis). Figure 3b shows the SR26 build-up curves of the 13 C12 signal. TheC hR2 470 signal in the dark sample as well as the ChR2 470 signal that remains in the 470 nm illumination at 170 Ks ample showed similar build-up curves which were fitted to similar distances corresponding to the all-trans configuration as expected. In contrast the curve of 13 C12-P 1 500 in the 170 Ki lluminated sample looked very different and fitted to 3.1 which is in good agreement with the expected 13-cis configuration. The 13 C12-P 4 480 curve in the sample illuminated at 245 Klooks similar to the 13 C12-P 1 500 curve and unambiguously confirms the 13-cis configuration of the chromophore in this state.T he P x 13 C12 signal overlaps with the natural abundance lipid signal and the build-up curve of this signal cannot be easily interpreted. Foranalysis of the P x state,w et herefore resorted to analysis of the 13 C15 signal. Figure 3c shows the results for 13 C15-ChR2 470 in the dark sample and 13 C15-P 4 480 and 13 C15-P x in the sample illuminated at 245 K. In agreement with the results on 13 C12, ChR2 470 and P 4 480 curves are fitted to 3.7 and 3.0 confirming the alltrans and 13-cis configuration, respectively.I nterestingly,t he 13 C15-P x build-up curve fits to ad istance of 3.4 which is neither in agreement with aplaner all-trans nor aplanar 13-cis configuration. Thus,t he retinal is in at wisted configuration which also explains the intermediate position of the chemical shift of the 13 C12 signal.
Theo bserved 13-cis,15-syn configuration of P 4 480 shows that the desensitized state does not correspond to the alltrans,15-anti O-state in the BR photocycle.T hus,torevert to the dark state,the chromophore configuration has to change which could explain the long lifetime of the desensitized state. Interestingly,t he observed configuration is the same as the population that arises during thermal equilibration of Bacteriorhodopsin in the dark. [13c] Thet wisted chromophore structure of the P x state has not been observed so far in any microbial rhodopsin. We speculate that the photoreaction of P 4 480 leads to aC13 cis-trans isomerization similar to the initial photoreaction. This would result in a1 3-anti,15-syn configuration in P x .T he shape of the chromophore thus resembles somewhat the K-like P 1 500 state.The energy stored in the P 1 500 state is transferred to the protein resulting in Schiff base deprotonation. ForP x no such reaction chain follows and the steric hindrances caused by this configuration can only be released by distorting the chromophore.F urther experimental evidence is needed to support this idea. In addition to time resolved experiments,c ryo-trapping photointermediates using different illumination and temperature schemes can provide insights into the sequence of photointermediates in aphoto cycle. We therefore compared our observations with the different models postulated (Figure S1). We observe that the only photo product obtained below 200 KisP 1 500 .Asthe signal-to-noise ratio of the spectra in Figure 2b-d) is limited due to the long distance of the atoms used for double quantum filtering,w er eplotted the double quantum filtered spectra on [14,15-13 C 2 ]-retinal-ChR2 from our previous work to unambiguously show that the P 4 480 photointermediate is not present in samples illuminated below 200 K( Figure S6). [5] This is in contrast to the idea that P 4 480 is also adirect product of the photo reaction. To generate P 4 480 additional thermal energy is needed and we could trap the state when illuminating at 245 K. We have shown that P 4 480 can also be reached by thermal relaxation of P 1 500 (without further illumination). This again is in contrast to the proposal that P 4 480 is generated directly from the dark state.O ur observations agree with ap hotocycle model where P 4 480 is generated at al ater state in the photocycle,f or example, during the open state decay as suggested before. [7] We did not discus these observation in detail in our previous work, [5] as the results were in agreement with the photocycle models at that time and amodel with the early P 4 480 state had not been postulated. To reproduce and confirm our results we performed at hermal relaxation experiment on the [12,15-13 C 2 ]retinal-ChR2 sample.T he proteoliposomes were illuminated at 170 Kt of orm the P 1 500 state and subsequently the sample temperature was increased to 245 Ki nt he dark (Figure 1d,  Figure 2e). The 13 C12-P 1 500 signal completely disappeared and instead the 13 C12-P 4 480 signal at 124.5 ppm is seen confirming  Figure 1d.The best fits of the build-up curves are shown together with the corresponding C12-C15 distances. The error ranges of the fits are shown in Figure S5.

Angewandte Chemie
Research Articles that P 4 480 follows P 1 500 in the photocycle.F igure 4s hows ap hotocycle model that agrees with our experimental data. TheC hR 470 population in the spectrum from the thermal relaxation experiment (Figure 2e)i si ncreased compared to the P 1 500 cryo-trapping condition (470 nm @1 70 K, Figure 2c), showing that the open state decays not exclusively towards P 4 480 but also ad irect shortcut to ChR2 470 may exist (dashed grey arrow in Figure 4). It can also be expected that P x can undergo athermal conversion to the dark state (dashed blue arrow in Figure 4).

Conclusion
In summary,for the first time,double quantum filtering of atoms at distances up to 3.7 has been shown in ar econstituted membrane protein. Based on this method, we have resolved the controversy of the chromophore configuration of the desensitized state of ChR2 (P 4 480 )u sing solid-state NMR spectroscopy as ar eadout which circumvents the assignment problems which pose ac hallenge to the interpretation of vibrational spectroscopy data. TheP 4 480 chromophore is in the 13-cis,15-syn configuration which fits well to its long lifetime. In addition, we could show that P 4 480 is not directly formed after light excitation but occurs at al ater state in the photocycle,p robably during channel closure.P 4 480 itself is photoactive and we could trap and analyze its photoproduct P x suggesting that it has anon-planar chromophore structure. Here,w eh ave shown that DNP-enhanced solid-state NMR spectroscopy in combination with cryo-trapping of photointermediates reveals details of the chromophore configuration in ChR2, which are difficult to obtain by other methods. In addition, we could show that reaction pathways can be deduced from the thermal relaxation pathways of cryotrapped samples.Therefore,wewould like to advocate the use of solid-state NMR spectroscopy in studying photoactive proteins. 390 and open channel P 3 520 states follow but no trapping protocol exists for these states. A1 3-cis,15-syn chromophore is observed in the desensitized P 4 480 state which is populated after channel closure. P 4 480 itself is photoactive and we term its photo product P x .P x containsatwisted chromophore structure.