Enlightening the photoactive site of channelrhodopsin-2by DNP-enhanced solid-state NMR spectroscopyJohanna Becker-Baldusa,b, Christian Bamannc, Krishna Saxenab,d, Henrik Gustmanne, Lynda J. Brownf,Richard C. D. Brownf, Christian Reiterg, Ernst Bambergc, Josef Wachtveitle, Harald Schwalbeb,d,and Clemens Glaubitza,b,1

aInstitute of Biophysical Chemistry, Goethe University Frankfurt, 60438 Frankfurt, Germany; bCentre for Biomolecular Magnetic Resonance, GoetheUniversity Frankfurt, 60438 Frankfurt, Germany; cMax-Planck-Institute of Biophysics, 60438 Frankfurt, Germany; dInstitute of Organic Chemistry andChemical Biology, Goethe University Frankfurt, 60438 Frankfurt, Germany; eInstitute of Physical and Theoretical Chemistry, Goethe University Frankfurt,60438 Frankfurt, Germany; fDepartment of Chemistry, University of Southampton, Southampton SO17 1BJ, United Kingdom; and gBruker Biospin GmbH,76287 Rheinstetten, Germany

Edited by G. Marius Clore, National Institutes of Health, Bethesda, MD, and approved June 25, 2015 (received for review April 21, 2015)

Channelrhodopsin-2 from Chlamydomonas reinhardtii is a light-gated ion channel. Over recent years, this ion channel has attractedconsiderable interest because of its unparalleled role in optogeneticapplications. However, despite considerable efforts, an understand-ing of how molecular events during the photocycle, including theretinal trans-cis isomerization and the deprotonation/reprotonationof the Schiff base, are coupled to the channel-opening mechanismremains elusive. To elucidate this question, changes of conformationand configuration of several photocycle and conducting/nonconduct-ing states need to be determined at atomic resolution. Here, weshow that such data can be obtained by solid-state NMR enhancedby dynamic nuclear polarization applied to 15N-labeled channelrho-dopsin-2 carrying 14,15-13C2 retinal reconstituted into lipid bilayers. Inits dark state, a pure all-trans retinal conformation with a stretchedC14-C15 bond and a significant out-of-plane twist of the H-C14-C15-Hdihedral angle could be observed. Using a combination of illumina-tion, freezing, and thermal relaxation procedures, a number of in-termediate states was generated and analyzed by DNP-enhancedsolid-state NMR. Three distinct intermediates could be analyzedwith high structural resolution: the early P5001 K-like state, the slowlydecaying late intermediate P4804 , and a third intermediate populatedonly under continuous illumination conditions. Our data providenovel insight into the photoactive site of channelrhodopsin-2 duringthe photocycle. They further show that DNP-enhanced solid-stateNMR fills the gap for challenging membrane proteins betweenfunctional studies and X-ray–based structure analysis, which is re-quired for resolving molecular mechanisms.

channelrhodopsin | retinal | solid-state NMR | DNP | freeze trapping

Since their discovery (1), channelrhodopsins (ChRs) havegenerated enormous interest because of the rapid development

of their applications in optogenetics (2–7). Commonly, ChR2 fromChlamydomonas reinhardtii (8) and its variants are used thanks totheir favorable expression levels. They are the only proteins knowntoday functioning as light-gated ion channels (Fig. 1A). Like othermicrobial retinal proteins, they undergo a periodic photocycle. InChRs, this photocycle is coupled to channel opening and closing asrevealed in electrophysiological recordings (8). A chimera of ChR1and ChR2 has been crystallized to yield a structure at 2.3-Å reso-lution (9). However, little is known on how this coupling functionson a molecular level, and a large number of studies based on visible(10–13), IR (11, 14–19), resonance Raman spectroscopy (20, 21),and EPR spectroscopy (22, 23) has been performed to addressthis question.The photocycles of microbial rhodopsins are usually compared

with bacteriorhodopsin, the first discovered and most studied light-driven proton pump (24). Without any illumination, microbialretinal proteins thermally equilibrate into a dark state (25). In thecase of bacteriorhodopsin, for example, this state contains a mixtureof two species termed bacteriorhodopsin568 (all-trans,15-anti retinal

Schiff base) and bacteriorhodopsin548 (13-cis,15-syn conformation)(26, 27). On illumination, light adaption occurs from the dark stateto the ground state, which contains only the all-trans,15-anti con-former as the photocycle starting point (28). A similar light–darkadaption has been found in halorhodopsin from Halobacteriumsalinarium (29). However, such a light/dark adaption in conjunctionwith a conformer mixture does not seem to be a general property ofmicrobial membrane proteins. Other systems have been describedwhere the ground state contains only an all-trans,15-anti retinal Schiffbase chromophore [e.g., green proteorhodopsin (30), Anabaenasensory rhodopsin (31), Oxyrrhis marina proteorhodopsin (32),sensory rhodopsin I from H. salinarum (33) and Salinibacter ruber(34), and sensory rhodopsin II from Natronobacterium pharaonis(35, 36) and H. salinarum (37)].In ChR2, the retinal is covalently bound to the lysine residue

257 conserved in all retinal proteins through a Schiff base linkage(38). The X-ray structure of the ChR chimera shows the retinalin an all-trans configuration (9), although other conformationscannot be excluded at the obtained resolution. Results of retinalextraction in conjunction with resonance Raman studies wereinterpreted as an isomer mixture containing 30% of a 13-cisretinal in dark- and light-adapted ChR2 (20). In addition,nanosecond IR spectroscopy on the E123T mutant of ChR2indicated the presence of some 13-cis retinal in the dark state

Significance

Channelrhodopsin-2 is a dimeric membrane protein functioningas a light-gated ion channel, which has triggered numerousoptogenetic applications. We present the first NMR study, toour knowledge, by which structural details of the retinal co-factor could be resolved. This study was only possible by en-hancing the detection sensitivity 60-fold through dynamicnuclear polarization (DNP), a highly promising hybrid methodlinking EPR with solid-state NMR spectroscopy. Our data showthat ground-state channelrhodopsin-2 contains the retinal co-factor in its all-trans configuration with a slightly perturbedpolyene chain. Three different photointermediates could betrapped and analyzed. Our study shows that DNP-enhancedsolid-state NMR is a key method for bridging the gap betweenX-ray–based structure analysis and functional studies toward ahighly resolved molecular picture.

Author contributions: J.B.-B., E.B., J.W., H.S., and C.G. designed research; J.B.-B., C.B., K.S.,and H.G. performed research; C.B., K.S., L.J.B., R.C.D.B., C.R., E.B., and H.S. contributednew reagents/analytic tools; J.B.-B., H.G., and C.G. analyzed data; and J.B.-B. and C.G.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: glaubitz@em.uni-frankfurt.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1507713112/-/DCSupplemental.

9896–9901 | PNAS | August 11, 2015 | vol. 112 | no. 32 www.pnas.org/cgi/doi/10.1073/pnas.1507713112

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

using a similar spectroscopic assignment as in the resonanceRaman study (39). In contrast to bacteriorhodopsin, no lightadaption was observed using resonance Raman techniques (20)or visual spectroscopy (12). The occurrence of a conformermixture in the ground state without light adaption would makeChR2 unique among the microbial retinal proteins, but addi-tional data are needed to confirm these observations more directlyat improved atomic resolution.The current model of the ChR2 photocycle is shown in Fig.

1B (14, 40). According to this model, blue light excitation leadsto a retinal all-trans to 13-cis isomerization, resulting in ared-shifted first intermediate P500

1 (12) resembling a K-likestate, which most likely contains a 13-cis,15-anti retinal Schiffbase chromophore similar to Bacteriorhodopsin (27). To ourknowledge, such red-shifted K-like intermediates occur in allmicrobial retinal proteins (38). Schiff base deprotonation leadsto the M-like state P390

2 (10, 11). This state is followed by the red-shifted intermediate P520

3 , which has previously been correlatedwith the open state (10). However, later data confirmed thatchannel opening occurs before P520

3 formation and might happenduring a spectroscopically silent transition between P390

2a andP3902b states (41). The last photocycle intermediate is the long-

lived intermediate P4804 state (τ = 24 s), which is referred to as

the desensitized state with a spectral characteristic similar to theground state (11, 42). In addition, time-resolved FTIR spec-troscopy indicated that P520

3 could partially convert directly tothe ground state (14).The situation becomes more complicated under continuous

light illumination (40, 43). Under these conditions, a high tran-sient current is observed first that is quickly reduced to a muchlower steady-state current. After turning off the irradiation, thesteady-state current decays biexponentially. This observation canonly be explained by a branching of the photocycle. Two openstates and two closed states are required to quantitatively de-scribe the observed behavior under continuous light conditions.The two closed states are most likely the ground state and thedesensitized state P480

4 that accumulates under continuous illu-mination and is identical to the same intermediate from a singleturnover (18). One of the open states is probably the open stateP5203 observed in single-turnover experiments. However, little is

known about the identity of the second open state, which onlyoccurs under continuous light conditions. It might be an M-likeP3902 state, another P520

3 state, or another unknown state. Lightexcitation of probably P480

4 creates this additional state. This stateor group of states here is referred to as Px containing at least oneopen state (Fig. 1B). It is also likely that the open states P520

3 and

Px to some extent can convert directly to the ground state, whichis indicated by dashed lines in Fig. 1B.All of the above-described states were detected by visible and

FTIR spectroscopy, and assignments of spectroscopic signaturesto conformational and configurational states of the retinal werebased on analogous data previously studied. However, detailedinformation on bond lengths or torsion angles that would alsolink to quantum chemical calculation is still missing. To fill thisgap between static crystallographic data on the one hand andkinetic and functional data based on optical spectroscopy andelectrophysiology on the other hand, we applied solid-statemagic angle spinning NMR on isotope-labeled ChR2 and retinalto obtain site-resolved structural data directly in a membraneenvironment under various experimental conditions. In this way,fine details of the chromophore conformation during the pho-tocycle could be resolved, which will be important to understandthe link between channel and photocycle activity in ChR2. Alimitation using proteoliposomes is the amount of sample thatcan be studied, because the protein-to-lipid ratio cannot be in-creased too much without compromising protein integrity. Inaddition, trapping photointermediates works best using sam-ples with low optical density, which reduces further the usableamount of protein, resulting in a poor NMR signal-to-noise ratio.Therefore, cross-effect dynamic nuclear polarization (DNP) en-hanced magic angle spinning (MAS) NMR [review in the workby Maly et al. (44)] was indispensable in overcoming these sen-sitivity problems (Fig. 1C). This technique requires temperaturesaround 100 K that are also compatible with trapping of photo-intermediates as outlined below. DNP-enhanced MAS NMR isnot yet a routine method but is applied increasingly to complex,mechanistic studies on retinal proteins (45–48) and other mem-brane proteins (49–51).Here, DNP-enhanced solid-state NMR spectroscopy has been

applied to 15N-labeled ChR2 carrying 14,15-13C2 retinal recon-stituted into lipid bilayers and incubated with the DNP polarizingagent AMUPOL (52) in a glycerol–water mixture. The labelingscheme adopted here is shown in Fig. 2A. The 13C14 chemicalshift is sensitive to the configuration of the C13-C14 bond. To-gether with the neighboring 13C15 atom, the two 13C-labeled

blue l

ight

blue l

ight

microw

ave

B0

ChR2470

O

N NH

NO

O

NO

O

O O

O

4

out

in

H+

K+Na+

Ca+

Na+

Ca+

H+

K+

(M-like)P2b390

P3520 P2a

390

channel

opening

P1500

(K-like)(desensitized state)P4

480Px

chan

nel

closin

g all-trans13-cis

A B C

D

Fig. 1. (A) Visualization of dimeric ChR2 reconstituted into the lipid bilayeras used in this study [cartoon based on the crystal structure of the ChR1/2chimera (data from ref. 9)]. Blue light illumination activates ChR2. (B) Singleturnover (black arrows) and continuous illumination photocycle (blue ar-rows) (14, 40). (C) Schematic view of the experimental setup for generatingand measuring different photointermediates. (D) The DNP enhancement isgenerated by magnetization transfer from the biradical AMUPOL to ChR2.

15N+ Lys257

H13C14

13C15

13C chemical shift (ppm) 200 150 100 50

A

0

CP

CP + DNP

DQF + DNP

CP at RT

C15 C14

0.0 0.5 1.0 1.5 2.0

experimental data

D = 2200 Hz / 1.51 Å D = 2100 Hz / 1.53 Å

D = 2300 Hz / 1.49 Å

inte

snity

a.u

.

DQ build-up time in ms

0.0 50.0 100.0 150.0 200.0 250.0

experimental data 154° 156° 158° 160° 162°

inte

nsity

a.u

.

HCCH evolution time in µs

B

C

D

lipid

glycerolC’

lipidC=C

aromatic

Fig. 2. (A) DNP-enhanced MAS NMR has been applied to U-15N-ChR2 con-taining 14,15-13C all-trans–retinal. (B) A 62-fold signal enhancement isachieved for 13C cross-polarization (CP; CP vs. CP + DNP). The 13C naturalabundance background can be efficiently suppressed by a double quantumfilter (DQF; DQF + DNP), resulting in a spectrum with only the resonances ofC14 and C15. As a control, one additional CP spectrum was acquired at 850MHz close to room temperature (CP at RT). The gray bar indicates where the13C14 signal would be expected for a chromophore in the 13-cis,15-syn con-formation. (C) 13C14-13C15 double-quantum (DQ) build-up curve. (D) HCCHdephasing curves for the C14-C15 spin system in ChR2 during two rotorperiods reporting on the HCCH dihedral angle.

Becker-Baldus et al. PNAS | August 11, 2015 | vol. 112 | no. 32 | 9897

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

spins can be used for double quantum filtering of this spin pairagainst the natural abundance background and at the same time,offer the possibility to study the length and the dihedral angle ofthe C14-C15 bond. Furthermore, the chemical shift of the Schiffbase nitrogen is also sensitive to the chromophore conformation,reports on the protonation state of the Schiff base, and reflectscounterion interactions. Using this approach, we were able toprovide a first analysis, to our knowledge, of the retinal–Schiffbase chromophore in ChR2 in its ground state as well as threedifferent photointermediate states at atomic resolution.

Results and DiscussionDark State—Ground State. Fig. 2B shows the 13C-DNP–enhancedMAS NMR spectrum of ChR2. Using AMUPOL as the polarizingagent, a 62-fold signal enhancement was achieved under our ex-perimental conditions. The observed resonances mainly stem fromthe 13C natural abundance background of protein, glycerol, andlipid. To suppress these signals and identify the 13C-labeled retinalsites, a double quantum filter has been applied, revealing just twopeaks from retinal carbons C14 and C15 at 126.3 and 166.5 ppm,respectively (Fig. 2B). The C14 chemical shift is very sensitive tothe conformation of the chromophore (26), and the value of 126.3ppm is, therefore, a very strong indicator for an all-trans,15-antichromophore conformation as observed for bacteriorhodopsinand proteorhodopsin (SI Appendix, Table S1). By contrast, theC14 signal of the 13-cis,15-syn chromophore in dark-adaptedbacteriorhodopsin appears at 111 ppm. A signal resonating at asimilar chemical shift has been observed for a 13-cis,15-syn sub-population in the A178R mutant of green proteorhodopsin (SIAppendix, Table S1). In ChR2, no signal at or near 111 ppm couldbe observed (Fig. 2B, gray area). We, therefore, conclude that thechromophore in ChR2 is present in a single all-trans,15-anti con-formation. To exclude that this result is an artifact of the sampleconditions required for DNP, a 13C-CP spectrum of ChR2 withoutthe addition of radicals or cryoprotectants was recorded at am-bient temperature using 832 times the number of scans comparedwith the DNP experiment (Fig. 2B, CP at RT vs. CP + DNP). Inboth cases, the retinal resonances compare well. Only the C14signal is shifted slightly upfield, which can be attributed to tem-perature effects. The small additional intensities observable in theambient temperature spectrum result from spinning side bandsand can be moved by changing the MAS frequency (SI Appendix,Fig. S1). In addition, keeping the sample in the dark at 4 °C for24 h did not change the appearance of the DNP NMR spectra (SIAppendix, Fig. S2).To further study the conformation of the chromophore, the

C14-C15 retinal distance has been determined using double-quantum build-up experiments (Fig. 2C). The obtained value of1.51 ± 0.02 Å is significantly longer than the 1.42 Å observed forgreen proteorhodopsin (47). This increase in bond length cor-responds well with a lower double-bond character of the bond asexpected from the blue shift of the absorption maximum com-pared with green proteorhodopsin. Measurements of the H-C14-C15-H retinal torsional angle revealed a significant out-of-planetwist with an angle of 158° ± 2°. A similar out-of-plane twist hasalso been observed for bacteriorhodopsin (164°) (53) and greenproteorhodopsin (161°) (47), indicating that this out-of-planetwist is a general property of microbial retinal rhodopsins andmight help to provide a favorable orientation of the Schiff baseduring the subsequent photocycle steps.The 15N chemical shift of the protonated Schiff base (pSB) was

detected as single resonance at 196.5 ppm using 1H-13C/13C-15Ndouble cross-polarization magnetization transfer from the termi-nal retinal carbon (C15) to the directly bonded Schiff base ni-trogen (Fig. 3). Hu et al. (54) established a relationship betweenthe Schiff base chemical shift and the wavelength of the absorp-tion maximum using Schiff base–counterion model complexes,which is shown in Fig. 3, Inset together with the values for

bacteriorhodopsin, green proteorhodopsin, and ChR2. ChR2(470 nm/196.5 ppm) agrees well with the predicted trend but de-viates more strongly from model behavior than bacteriorhodopsin.Based on the model compound geometries, it can be concludedthat the distance from the Schiff base to its counterion in ChR2 isshorter than in bacteriorhodopsin. As for C14- or C15-retinalatoms, no peak doubling or splitting is observed, confirming thatretinal is present in only one conformation in the ChR2 groundstate. The chemical shift difference of the all-trans,15-anti and the13-cis,15-syn 15N signals observed in dark-adapted bacteriorho-dopsin is 8.3 ppm (46), which would be well-resolved under theexperimental conditions applied here.The observation of a 100% all-trans,15-anti conformation of

the retinal cofactor in ground-state ChR2 is in line with manyother microbial rhodopsins but in contrast to previous reports onChR2. Previously, a population of 30% 13-cis retinal based onretinal extraction with subsequent HPLC analysis and resonanceRaman experiments has been reported (20).The reason for this discrepancy is the invasive character of

retinal extraction in the previously reported work (20). Thisprocess required breaking the Schiff base linker by subjecting theprotein to EtOH followed by extraction of retinals into hexanesand HPLC purification (55). For bacteriorhodopsin, the resultsobtained in this way (56) agree well with data from other methods(26, 27). In other cases, however, this protein treatment led to lessconsistent results. For example, retinal extraction studies of greenproteorhodopsin suggested values between 5% and 20% cis-reti-nal (57, 58). Later, resonance Raman (59) and solid-state MASNMR (30) studies confirmed that the ground state contains verylittle (if any) cis-retinal. Similarly, for Anabaena sensory rhodop-sin, retinal extraction experiments showed 24% of the 13-cis iso-mer in the dark-adapted state (60), whereas solid-state MASNMR experiments showed a purely all-trans configuration (31).Resonance Raman experiments are noninvasive and should givereliable information on the retinal chromophore. However, as-signment of the vibrational bands is very challenging and cannoteasily be transferred between different systems. In the case ofbacteriorhodopsin, assignment was based on differently isotope-labeled retinals (27). Such data are missing for ChR2, resulting inambiguous interpretation of Raman data (20) (more detaileddiscussion is in SI Appendix).Our data, therefore, show unambiguously that ground-state

ChR2 contains an all-trans,15-anti chromophore, which also ex-plains the previously reported monoexponential decay of thephotoexcited state (12) and the absence of a light adaption stepas observed in bacteriorhodopsin, because the amount of thisconformer is already close to 100% in the dark.

0 200 150 100 50 15N chemical shift (ppm) 15N (ppm)

22

20

16

18

165 185175

BR568

mod

el co

mpo

unds 1/λ

max (10

3 cm-1)GPR

195

ChR2

amide

DCP+DNP

CP+DNP

pSB

HisArg Lys

Fig. 3. DNP-enhanced 15N cross-polarization (CP) spectrum (CP + DNP) anddouble cross-polarization (DCP) –filtered 15N spectrum (DCP + DNP) of ChR2.(Inset) Comparison of the correlation of the absorption maximum with the15N Schiff base chemical shift for ChR2, green proteorhodopsin (GPR), bac-teriorhodopsin (BR), and model compounds. Modified from ref. 54.

9898 | www.pnas.org/cgi/doi/10.1073/pnas.1507713112 Becker-Baldus et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Observing ChR2 Photointermediates by DNP-Enhanced MAS NMR. Forthe analysis of photointermediates, sample illumination has beenused in the past [e.g., combined with solution-state NMR toaccess, for example, the photokinetics of rhodopsin (61, 62) andsolid-state NMR for thermal trapping of bacteriorhodopsin orrhodopsin intermediates (46)]. Special consideration has to begiven to an efficient illumination setup, because light penetrationof the sample significantly decreases with sample thickness (62).The amount of ChR2 proteoliposomes within the opticallytransparent sapphire rotor used for DNP NMR was, therefore,reduced to about 20% and evenly distributed across the innerrotor surface by short sample rotation at room temperature. Inprinciple, sample illumination can be done in two ways. Onepossibility is to illuminate the sample outside of the NMRmagnetand then, trap the generated state by quickly freezing it andinserting it in the cold probe. However, temperature control usingthis method can be difficult. Another option is to equip the DNPsolid-state probe with a light guide and illuminate the sampledirectly while it is spinning in the stator. Such an approach hasbeen shown by a number of laboratories using different designswith DNP (46) or standard MAS probes (62–64). The latter ap-proach offers better temperature control, but quick freezing ofthe spinning samples is not possible. In this study, both methodshave been used and yielded similar results as discussed below.Significant spectral changes are observed on illumination with

blue light at 110 K (Fig. 4). For the retinal 13C14 signal, one ad-ditional peak at 124.2 ppm is observed in addition to the ground-state signal, whereas the 13C15 resonance broadens slightly. Sim-ilarly, a new 15N signal upfield of the pSB ground-state resonanceis detected at 181 ppm. It can be concluded that illumination atlow temperatures leads to a mixture of two states, one of whichcorresponds to the ground state.For assigning the second state to one of the photointermediates

(Fig. 1B), optical spectra on thermally trapped samples preparedunder conditions very close to those used for DNP have beenrecorded. The difference between spectra of a dark sample andthose of an illuminated sample shows negative intensities below

480 nm (ground-state bleaching) and positive absorption with amaximum at 500 nm (Fig. 4C). These values are characteristicfor the K-like intermediate P500

1 , which strongly suggests that thisstate has been trapped and observed by DNP-enhanced MASNMR in agreement with earlier UV-visible and FTIR spectro-scopic studies (11).The NMR data are also in good agreement with the thermally

trapped K state of bacteriorhodopsin, which was accumulated byillumination at 532 nm at 90 K (45, 46). The authors in refs. 45 and46 reported upfield shifts for both 13C14-retinal and 15N pSB signalswith respect to the ground state by 4.9 and 8.7 ppm, respectively.The same trend is observed here, and the newly generated state is,therefore, assigned to the K-like intermediate P500

1 containing a13-cis,15-anti chromophore. Because P500

1 is also photoactive andcan be converted to the ground state by light, only a mixture ofstates can be obtained as reported for bacteriorhodospin (65).To reach other intermediates during the photocycle, ChR2 was

now subjected to a thermal relaxation and a thermal trappingprotocol (Fig. 5A). In the first case, the P500

1 state was created asdescribed above followed by thermal relaxation and equilibrationat temperatures between 194 and 273 K, after which the samplewas again cooled to 110 K for DNP detection (Fig. 5A). In this way,the thermal energy should allow ChR2 to overcome the energybarrier between its P500

1 state and subsequent photocycle inter-mediates. The second protocol involved continuous illuminationat temperatures between 234 and 254 K followed by freezing (Fig.5B). In addition, a sample was illuminated at room temperatureand then quickly frozen in liquid nitrogen. The second procedureshould make especially those photointermediates accessible, whichonly occur under continuous illumination (Fig. 1B).The thermal relaxation approach results in altered line shapes

of the 13C14- and 13C15-retinal resonances (Fig. 5A). With in-creasing relaxation temperatures, the 13C14-P500

1 signal broadens

100 180 160 140 120 13C chemical shift (ppm)

dark

blue light at 110 K

200 100 15N chemical shift (ppm)

C15 C14 pSB

P1500

P1500

300 350 400 450 500 550 600 wavelength (nm)

Δ ab

s.A

B

C

Fig. 4. 13C double quantum filter- and 15N double cross-polarization–fil-tered, DNP-enhanced spectra of ChR2 in a sapphire rotor (A) in the dark and(B) after illumination with blue light at 110 K. (C) Optical difference spec-trum of ChR2 at 150 K of dark- and blue light-illuminated ChR2 (Δ abs refersto the absorption difference).

110 K

194 K

207 K

217 K

227 K

245 K

255 K

265 K

273 K

140 100 180 13C chemical shift (ppm)

blue light RTfast freeze

blue light 254 K

blue light 245 K

blue light234 K

140 100 180

C15C14

C15

C14

13C chemical shift (ppm)

P4804

ChR2470P500

1

ChR2470P480

4

Px

blue light

thermal relaxation

110 K

detectionblue light

thermal trapping

110 K

detection

(a)

(b)

A B

RT

RT

Fig. 5. (A) 13C double quantum filter (DQF) DNP-enhanced spectra of ChR2obtained by thermal relaxation. (B) 13C DQF DNP-enhanced spectra of ChR2obtained by thermal trapping. Dashed lines indicate the observed peakposition. RT, room temperature.

Becker-Baldus et al. PNAS | August 11, 2015 | vol. 112 | no. 32 | 9899

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

initially, converts into a new signal at 119.3 ppm between 217 and255 K, and disappears above 265 K. In principle, the 13C15 reso-nance follows this trend, but changes are less pronounced,because only some line broadening can be detected. At highertemperatures, all additional signals vanish, and the sample re-turns into the ground-state ChR2470.The thermal trapping protocol results in a different spectral

characteristic of the chromophore as shown in Fig. 5B. Con-tinuous illumination at room temperature is known to accu-mulate the long-lived intermediate P480

4 (10), which can befurther stabilized for DNP NMR detection by rapid freezing.Therefore, the two signals observed under these conditions forthe 13C14 site can be assigned to the ground state (126.0 ppm)and P480

4 (119.3 ppm). The latter is identical to the 13C14 chemicalshift that was observed during the thermal relaxation experi-ment. We, therefore, conclude that this peak belongs to a P480

4population, which can also be generated when lowering theillumination temperature. The ground-state subpopulation isalmost completely depopulated at 245 K, which proves thatthe whole sample can be illuminated sufficiently with the appliedsetup. At this temperature, an additional signal at 122.7 ppmcan be detected, which is also seen in the 234- and 254-Kspectra. Because this species is only generated under continu-ous light, it is tentatively assigned to one of the Px states(Fig. 1B).The most pronounced photointermediates have been pre-

pared at 245 K in both thermal relaxation as well as trappingapproaches. We have, therefore, recorded 15N double cross-polarization–filtered spectra at this temperature using bothprotocols (SI Appendix, Fig. S3). Under thermal relaxation,only one signal for the pSB is observed at a chemical shiftidentical to the ground state but with slightly reduced intensity.Furthermore, no evidence for a deprotonated Schiff base spe-cies (i.e., an M-like state) has been found. A possible expla-nation could be that the Schiff base signal of this photo state issimilar to the ground state or broadened beyond detection. Theexperiment was repeated under thermal trapping conditions.Here, the ground state is depleted at 245 K, which results in aloss of the pSB signal at 196.5 ppm and shows that the P480

4 wasnot hidden underneath. A new signal occurs at 185 ppm, which isnot visible under thermal relaxation conditions and therefore, isassigned to the Px state.Thermal relaxation and the thermal trapping protocols at

245 K were also compared using optical spectroscopy undernearly the same experimental condition (SI Appendix, Fig. S4).These data confirm that the thermal relaxation and the thermaltrapping protocols at 245 K result in a different population ofphotointermediates.In summary, three photointermediates (P500

1 , P4804 , and Px)

could be generated using different illumination/relaxation pro-tocols. This assignment is also in agreement with UV-visiblespectra, which were obtained under cryogenic conditions similarto those used here for DNP on ChR2 samples and subjected tothe same illumination and freeze-trapping protocols. Similar tothe ground state, we have recorded double-quantum filter build-up and HCCH torsion angle data for all of them (SI Appendix,Fig. S5). The results are given in Table 1 together with theground-state data. Our data show that the twisting and stretchingof the C14-C15 bond observed in the ground state are conservedin all three trapped states.At first glance, it seems surprising that the retinal isomeriza-

tion in the K-like state does not have a strong effect on the C14-C15 bond, because a hydrogen-out-of-plane band indicative ofbond twisting has been reported to occur from the K to the Lstate of bacteriorhodopsin (66, 67). However, direct solid-stateNMR experiments on bacteriorhodopsin, as discussed above, haveshown that this bond is already twisted in the ground state, whichwas observed here for ChR2, and its out-of-plane orientation

increases in the M state (53). The latter effect could also beexpected for ChR2, for which the deprotonation of the chromo-phore in the M-like state is accompanied by channel opening (53).However, such a state could not be trapped in this study. Thestable conformation of the C14-C15 bond during the ChR2 pho-tocycle shows a strong coupling between retinal and channelopsin-2 and indicates that, most likely, other parts of the retinal cofactorrespond more strongly.Our data show that the K-like state P500

1 is relatively stable andthat thermal relaxation only allows accumulating the long-livedP4804 intermediate. This finding is in contrast to bacteriorhodopsin,

for which K, L, and late M states could be generated in this way(46). The observed differences imply that the energy barriers be-tween these states are significantly lower in ChR2 compared withbacteriorhodopsin. Furthermore, the detection of a new inter-mediate Px under continuous illumination shows that this branchof the photocycle indeed exists, because it is required to explainthe electrophysiological data for the WTs (40, 43) and mutantswith slow photocycles (17, 68).

Conclusion and OutlookHere, we presented the first NMR study, to our knowledge, ofChR2 using DNP-enhanced solid-state MAS NMR. Our datashow a pure all-trans,15-anti retinal Schiff base with a stretchedC14-C15 bond length and a significant out-of-plane twist of theH-C14-C15-H dihedral angle in the ground state of ChR2. Threedifferent photostates could be generated using thermal relaxation/trapping protocols, including a so-far unknown intermediate thatonly occurs under continuous light conditions. Additional inter-mediates will become accessible using ChR2 variants like C128Tand D156A with long-lived P520

3 and P3902 states (68). Our data

provide novel insight into the photoactive site of ChR2 and showthat DNP-enhanced solid-state NMR fills the gap between func-tional and X-ray–based structure analysis, which is required toresolve its molecular mechanism. Additional studies using exten-sively labeled retinals incorporated into isotope-labeled opsin formore structural insight during channel-opening and -closing eventswill be reported in the future.

Methods15N-ChR2 was expressed in Pichia pastoris, generated with 14–15-13C2–all-trans retinal, and after purification, reconstituted into liposomes. Solid-state MAS NMR was performed under cryogenic conditions (100 K) usingcross-effect DNP enhancement provided by doping the proteoliposomes withAMUPOL and applying high-power microwave irradiation to the sample.Trapping of the different photointermediates was achieved using protocolscombining illumination, freezing, and thermal relaxation. Optical data wererecorded under conditions that resembled the NMR samples as closely aspossible. SI Appendix has a detailed description of the applied methods.

ACKNOWLEDGMENTS. Oliver Ouari and Paul Tordo are acknowledged forproviding the polarizing agent AMUPOL. The work was funded by DeutscheForschungsgemeinschaft/Sonderforschungsbereich 807 Transport and Com-munications across Membranes. The dynamic nuclear polarization experi-ments were enabled through DFG Equipment Grant GL 307/4-1 and theCluster of Excellence Frankfurt: Macromolecular Complexes Frankfurt. Workat the Center for Biomolecular Magnetic Resonance is supported by theState of Hesse.

Table 1. Parameters obtained for the retinal C14-C15 bond inChR2 ground state and its photointermediates

State δ(C14)/ppm δ(pSB)/ppm R/Å Φ/°

ChR2470 126.0 ± 0.5 196.5 ± 0.5 1.51 ± 0.02 158 ± 2P1

500 124.2 ± 0.5 181.0 ± 1.0 1.56 ± 0.04 156 ± 4P4

480 119.3 ± 0.5 — 1.53 ± 0.04 152 ± 4Px 122.7 ± 0.5 185.0 ± 1.0 1.52 ± 0.02 156 ± 2

PHI = C14-C15 HCCH torsion angle; R = C14-C15 bond length.

9900 | www.pnas.org/cgi/doi/10.1073/pnas.1507713112 Becker-Baldus et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

1. Nagel G, et al. (2002) Channelrhodopsin-1: A light-gated proton channel in greenalgae. Science 296(5577):2395–2398.

2. Fenno L, Yizhar O, Deisseroth K (2011) The development and application of opto-genetics. Annu Rev Neurosci 34(1):389–412.

3. Nagel G, et al. (2005) Light activation of channelrhodopsin-2 in excitable cellsof Caenorhabditis elegans triggers rapid Behavioral responses. Curr Biol 15(24):2279–2284.

4. Liewald JF, et al. (2008) Optogenetic analysis of synaptic function. Nat Methods 5(10):895–902.

5. Zhang F, et al. (2007) Multimodal fast optical interrogation of neural circuitry. Nature446(7136):633–639.

6. Hegemann P, Moglich A (2011) Channelrhodopsin engineering and exploration ofnew optogenetic tools. Nat Methods 8(1):39–42.

7. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005) Millisecond-timescale,genetically targeted optical control of neural activity. Nat Neurosci 8(9):1263–1268.

8. Nagel G, et al. (2003) Channelrhodopsin-2, a directly light-gated cation-selectivemembrane channel. Proc Natl Acad Sci USA 100(24):13940–13945.

9. Kato HE, et al. (2012) Crystal structure of the channelrhodopsin light-gated cationchannel. Nature 482(7385):369–374.

10. Bamann C, Kirsch T, Nagel G, Bamberg E (2008) Spectral characteristics of the pho-tocycle of channelrhodopsin-2 and its implication for channel function. J Mol Biol375(3):686–694.

11. Ritter E, Stehfest K, Berndt A, Hegemann P, Bartl FJ (2008) Monitoring light-inducedstructural changes of Channelrhodopsin-2 by UV-visible and Fourier transform in-frared spectroscopy. J Biol Chem 283(50):35033–35041.

12. Verhoefen MK, et al. (2010) The photocycle of channelrhodopsin-2: Ultrafast reactiondynamics and subsequent reaction steps. ChemPhysChem 11(14):3113–3122.

13. Stehfest K, Ritter E, Berndt A, Bartl F, Hegemann P (2010) The branched photocycle ofthe slow-cycling channelrhodopsin-2 mutant C128T. J Mol Biol 398(5):690–702.

14. Lórenz-Fonfría VA, et al. (2013) Transient protonation changes in channelrhodopsin-2and their relevance to channel gating. Proc Natl Acad Sci USA 110(14):E1273–E1281.

15. Neumann-Verhoefen MK, et al. (2013) Ultrafast infrared spectroscopy on channelr-hodopsin-2 reveals efficient energy transfer from the retinal chromophore to theprotein. J Am Chem Soc 135(18):6968–6976.

16. Radu I, et al. (2009) Conformational changes of channelrhodopsin-2. J Am Chem Soc131(21):7313–7319.

17. Ritter E, Piwowarski P, Hegemann P, Bartl FJ (2013) Light-dark adaptation of chan-nelrhodopsin C128T mutant. J Biol Chem 288(15):10451–10458.

18. Eisenhauer K, et al. (2012) In channelrhodopsin-2 Glu-90 is crucial for ion selectivityand is deprotonated during the photocycle. J Biol Chem 287(9):6904–6911.

19. Kuhne J, et al. (2015) Early formation of the ion-conducting pore in channelrhodopsin-2.Angew Chem Int Ed Engl 54(16):4953–4957.

20. Nack M, Radu I, Bamann C, Bamberg E, Heberle J (2009) The retinal structure ofchannelrhodopsin-2 assessed by resonance Raman spectroscopy. FEBS Lett 583(22):3676–3680.

21. Bruun S, et al. (2011) The chromophore structure of the long-lived intermediate ofthe C128T channelrhodopsin-2 variant. FEBS Lett 585(24):3998–4001.

22. Krause N, Engelhard C, Heberle J, Schlesinger R, Bittl R (2013) Structural differencesbetween the closed and open states of channelrhodopsin-2 as observed by EPRspectroscopy. FEBS Lett 587(20):3309–3313.

23. Sattig T, Rickert C, Bamberg E, Steinhoff HJ, Bamann C (2013) Light-induced move-ment of the transmembrane helix B in channelrhodopsin-2. Angew Chem Int Ed Engl52(37):9705–9708.

24. Oesterhelt D, Stoeckenius W (1971) Rhodopsin-like protein from the purple mem-brane of Halobacterium halobium. Nat New Biol 233(39):149–152.

25. Lozier RH, Bogomolni RA, Stoeckenius W (1975) Bacteriorhodopsin: A light-drivenproton pump in Halobacterium Halobium. Biophys J 15(9):955–962.

26. Harbison GS, et al. (1984) Dark-adapted bacteriorhodopsin contains 13-cis, 15-syn andall-trans, 15-anti retinal Schiff bases. Proc Natl Acad Sci USA 81(6):1706–1709.

27. Smith SO, Lugtenburg J, Mathies RA (1985) Determination of retinal chromophorestructure in bacteriorhodopsin with resonance Raman spectroscopy. J Membr Biol85(2):95–109.

28. Stoeckenius W, Bogomolni RA (1982) Bacteriorhodopsin and related pigments ofhalobacteria. Annu Rev Biochem 51(1):587–616.

29. Kamo N, Hazemoto N, Kobatake Y, Mukohata Y (1985) Light and dark adaptation ofhalorhodopsin. Arch Biochem Biophys 238(1):90–96.

30. Pfleger N, Lorch M, Woerner AC, Shastri S, Glaubitz C (2008) Characterisation of Schiffbase and chromophore in green proteorhodopsin by solid-state NMR. J Biomol NMR40(1):15–21.

31. Wang S, et al. (2013) Solid-state NMR spectroscopy structure determination of a lipid-embedded heptahelical membrane protein. Nat Methods 10(10):1007–1012.

32. Janke C, et al. (2013) Photocycle and vectorial proton transfer in a rhodopsin from theeukaryote Oxyrrhis marina. Biochemistry 52(16):2750–2763.

33. Yan B, Nakanishi K, Spudich JL (1991) Mechanism of activation of sensory rhodopsin I:Evidence for a steric trigger. Proc Natl Acad Sci USA 88(21):9412–9416.

34. Kitajima-Ihara T, et al. (2008) Salinibacter sensory rhodopsin: Sensory rhodopsin I-likeprotein from a eubacterium. J Biol Chem 283(35):23533–23541.

35. Hirayma J, Kamo N, Imamoto Y, Shichida Y, Yoshizawa T (1995) Reason for the lack oflight-dark adaptation in pharaonis phoborhodopsin: Reconstitution with 13-cis-reti-nal. FEBS Lett 364(2):168–170.

36. Imamoto Y, et al. (1992) Chromophore configuration of pharaonis phoborhodopsinand its isomerization on photon absorption. Biochemistry 31(9):2523–2528.

37. Scharf B, Hess B, Engelhard M (1992) Chromophore of sensory rhodopsin II fromHalobacterium halobium. Biochemistry 31(49):12486–12492.

38. Ernst OP, et al. (2014) Microbial and animal rhodopsins: Structures, functions, andmolecular mechanisms. Chem Rev 114(1):126–163.

39. Lórenz-Fonfría VA, et al. (2015) Pre-gating conformational changes in the ChETAvariant of channelrhodopsin-2 monitored by nanosecond IR spectroscopy. J Am ChemSoc 137(5):1850–1861.

40. Nikolic K, et al. (2009) Photocycles of channelrhodopsin-2. Photochem Photobiol85(1):400–411.

41. Lórenz-Fonfría VA, Heberle J (2014) Channelrhodopsin unchained: Structure andmechanism of a light-gated cation channel. Biochim Biophys Acta 1837(5):626–642.

42. Nack M, et al. (2012) Kinetics of proton release and uptake by channelrhodopsin-2.FEBS Lett 586(9):1344–1348.

43. Hegemann P, Ehlenbeck S, Gradmann D (2005) Multiple photocycles of channelrho-dopsin. Biophys J 89(6):3911–3918.

44. Maly T, et al. (2008) Dynamic nuclear polarization at high magnetic fields. J ChemPhys 128(5):052211.

45. Bajaj VS, Mak-Jurkauskas ML, Belenky M, Herzfeld J, Griffin RG (2009) Functional andshunt states of bacteriorhodopsin resolved by 250 GHz dynamic nuclear polarization-enhanced solid-state NMR. Proc Natl Acad Sci USA 106(23):9244–9249.

46. Mak-Jurkauskas ML, et al. (2008) Energy transformations early in the bacteriorho-dopsin photocycle revealed by DNP-enhanced solid-state NMR. Proc Natl Acad Sci USA105(3):883–888.

47. Mao J, et al. (2014) Structural basis of the green-blue color switching in proteo-rhodopsin as determined by NMR spectroscopy. J Am Chem Soc 136(50):17578–17590.

48. Mehler M, et al. (2013) The EF loop in green proteorhodopsin affects conformationand photocycle dynamics. Biophys J 105(2):385–397.

49. Jacso T, et al. (2012) Characterization of membrane proteins in isolated native cellularmembranes by dynamic nuclear polarization solid-state NMR spectroscopy withoutpurification and reconstitution. Angew Chem Int Ed Engl 51(2):432–435.

50. Ong YS, Lakatos A, Becker-Baldus J, Pos KM, Glaubitz C (2013) Detecting substratesbound to the secondary multidrug efflux pump EmrE by DNP-enhanced solid-stateNMR. J Am Chem Soc 135(42):15754–15762.

51. Reggie L, Lopez JJ, Collinson I, Glaubitz C, Lorch M (2011) Dynamic nuclear polari-zation-enhanced solid-state NMR of a 13C-labeled signal peptide bound to lipid-reconstituted Sec translocon. J Am Chem Soc 133(47):19084–19086.

52. Sauvée C, et al. (2013) Highly efficient, water-soluble polarizing agents for dynamicnuclear polarization at high frequency. Angew Chem Int Ed Engl 52(41):10858–10861.

53. Lansing JC, et al. (2002) Chromophore distortions in the bacteriorhodopsin photo-cycle: Evolution of the H-C14-C15-H dihedral angle measured by solid-state NMR.Biochemistry 41(2):431–438.

54. Hu J, Griffin RG, Herzfeld J (1994) Synergy in the spectral tuning of retinal pigments:Complete accounting of the opsin shift in bacteriorhodopsin. Proc Natl Acad Sci USA91(19):8880–8884.

55. Scherrer P, Mathew MK, Sperling W, Stoeckenius W (1989) Retinal isomer ratio indark-adapted purple membrane and bacteriorhodopsin monomers. Biochemistry28(2):829–834.

56. Pettei MJ, Yudd AP, Nakanishi K, Henselman R, Stoeckenius W (1977) Identification ofretinal isomers isolated from bacteriorhodopsin. Biochemistry 16(9):1955–1959.

57. Friedrich T, et al. (2002) Proteorhodopsin is a light-driven proton pump with variablevectoriality. J Mol Biol 321(5):821–838.

58. Imasheva ES, et al. (2005) Formation of a long-lived photoproduct with a de-protonated Schiff base in proteorhodopsin, and its enhancement by mutation ofAsp227. Biochemistry 44(32):10828–10838.

59. Dioumaev AK, et al. (2002) Proton transfers in the photochemical reaction cycle ofproteorhodopsin. Biochemistry 41(17):5348–5358.

60. Vogeley L, et al. (2004) Anabaena sensory rhodopsin: A photochromic color sensor at2.0 A. Science 306(5700):1390–1393.

61. Stehle J, et al. (2014) Characterization of the simultaneous decay kinetics of meta-rhodopsin states II and III in rhodopsin by solution-state NMR spectroscopy. AngewChem Int Ed Engl 53(8):2078–2084.

62. Concistrè M, et al. (2009) Light penetration and photoisomerization in rhodopsinstudied by numerical simulations and double-quantum solid-state NMR spectroscopy.J Am Chem Soc 131(17):6133–6140.

63. Daviso E, Diller A, Alia A, Matysik J, Jeschke G (2008) Photo-CIDNP MAS NMR beyondthe T1 limit by fast cycles of polarization extinction and polarization generation.J Magn Reson 190(1):43–51.

64. Tomonaga Y, et al. (2011) An active photoreceptor intermediate revealed by in situphotoirradiated solid-state NMR spectroscopy. Biophys J 101(10):L50–L52.

65. Xie AH (1990) Quantum efficiencies of bacteriorhodopsin photochemical reactions.Biophys J 58(5):1127–1132.

66. Ames JB, et al. (1989) Bacteriorhodopsin’s M412 intermediate contains a 13-cis, 14-s-trans,15-anti-retinal Schiff base chromophore. Biochemistry 28(9):3681–3687.

67. Rödig C, Chizhov I, Weidlich O, Siebert F (1999) Time-resolved step-scan Fouriertransform infrared spectroscopy reveals differences between early and late M in-termediates of bacteriorhodopsin. Biophys J 76(5):2687–2701.

68. Bamann C, Gueta R, Kleinlogel S, Nagel G, Bamberg E (2010) Structural guidanceof the photocycle of channelrhodopsin-2 by an interhelical hydrogen bond.Biochemistry 49(2):267–278.

Becker-Baldus et al. PNAS | August 11, 2015 | vol. 112 | no. 32 | 9901

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020