Abstract
Anion channelrhodopsin GtACR1 is a powerful optogenetic tool to inhibit nerve activity. Its kinetic mechanism was interpreted in terms of the bacteriorhodopsin photocycle, and the L intermediate was assigned to the open channel state. Here, we report the results of the comparison between the time dependence of the channel currents and the time evolutions of the K-like and L-like spectral forms. Based on the results, we question the current view on GtACR1 kinetics and the assignment of the L intermediate to the open channel state. We report evidence for a red-absorbing intermediate being responsible for channel opening.
Significance
Anion channelrhodopsin GtACR1 is an important optogenetic tool. Understanding its photochemical mechanism is of primary importance to gain insights into its function. One of the crucial steps in this process is identifying the possible intermediates of the photocycle, particularly the one most tightly associated with the channel function. Combining spectroscopic data with electrophysiological records, we were able to accomplish this goal. Our results also showed that adopting the mechanism of bacteriorhodopsin to channelrhodopsin kinetics may produce a false picture of their functioning.
Introduction
The photoreactions of the anion channelrhodopsin from Guillardia theta (GtACR1) have received a great deal of attention recently, as this is an important tool for optogenetic studies in which light can be used to inhibit nerve activity (1). As is typical in studies of photoreactions of bacterial retinal proteins, interpretations of time-resolved spectral changes of GtACR1 have assumed a reaction mechanism that mirrors that of the bacterial proton pump bacteriorhodopsin (BR) (2). To suggest that GtACR1 and BR have similar photoreactions is questionable, especially since these two proteins have very different functions. However, within that framework, the previous studies have suggested that the state responsible for channel opening in GtACR1 is a state with a blue-shifted spectrum analogous to the L intermediate of BR. A photocycle scheme, containing steps analogous to the ones in the BR photocycle, was proposed (3). It was considered a simplification that captures the main events in the reaction sequence. Recently, an improved version of the model was introduced based on spectroscopic data in both the infrared and visible regions recorded on the wild-type (WT) protein (4). It suggests that several, early and late, consecutive L intermediates are involved in the photocycle, of which the late one is the open channel state. Recently, we conducted a detailed time-resolved absorption spectroscopy study on the WT protein, the A75E, and two low-conductance mutants, D234N and S97E, at physiological and acidic pH and found inconsistencies in the current interpretation of the GtACR1 kinetics (5). Here, we compare the time dependences of the spectral and channel current data and show that the open channel state is not a blue-absorbing L but a red-absorbing intermediate.
Materials and methods
The spectral information used in this study is a reproduction of the results derived from the analysis of time-resolved spectroscopic data in our previous work (5). A summary of the methods used in that work is given here. The proteins used in the experiments were solubilized in 0.03% DDM (dodecylmaltoside) detergent. Time-resolved absorption difference spectra after excitation by 532-nm laser pulse were recorded in the 350–725 nm wavelength and the 100 ns to seconds time range for the WT protein at pH 5.5, 7.4, and 8.5, for the A75E mutant at pH 7.4, for the D234N mutant at pH 4.5 and 7.4, and for the S97E mutant at pH 5.6 and 7.4. The data matrices were subjected to singular value decomposition (SVD): A = USVT, where matrix U contains the orthonormal spectral vectors, matrix V the orthonormal temporal vectors, and the diagonal matrix S is the measure of significance of the U and V vector pair contributions to the data. Global exponential fitting of the V vectors gave the apparent lifetimes and amplitude spectra, called b-spectra, for the samples studied. The b-spectra (bs) were converted into sequential spectra (In), also called spectra of the sequential intermediates: In = bs∗Ws−1, where Ws is the eigenvector matrix of the kinetic matrix constructed based on the apparent rates of the exponential fit. The sequential spectra were deconvoluted into spectral forms believed to represent the spectra of molecular states in the reaction sequence. The results of the decomposition were included in the composition matrix, Q(m,n), each column of which contains the spectral composition of the corresponding sequential intermediate. The time evolution of the spectral forms was calculated as the product of the composition matrix Q(m,n) and the matrix C(n,t) that contains the time-dependent concentrations of the sequential intermediates.
For comparison between the time-dependent spectral traces and the published current traces (6,7), the amplitude of the current is normalized to the concentration levels of the appropriate K-like and L-like spectral forms.
Results and discussion
Is the L intermediate the open channel state?
In our recent publication on anion channelrhodopsin GtACR1 photoreactions, the light-induced absorption difference spectra in the nanosecond to second time interval were analyzed for the WT, the A75E, and the poorly conducting S97E and D234N mutant proteins (5). Based on the results of that study, we suggested that an intermediate with a red-shifted spectrum is a better candidate for the open channel state than the blue-shifted L intermediate proposed earlier (3,4). Because this work is a continuation of our previous study, a brief presentation of the analysis used in that work is given here using the WT protein data at pH 7.4 as an example. For all the proteins, SVD of the data matrices was consistent with four independent spectra, whereas global exponential fitting delivered five apparent rates, indicating six intermediates present in the photocycles. The results of the global exponential fit are shown in Fig. 1 A. The two early microsecond apparent lifetimes and the corresponding b-spectra are analogous to the relaxation of the K intermediate in BR kinetics. However, they are not followed by any spectral change over two decades of time, the time window corresponding to the channel current onset. In the b-spectra, the positive components correspond to the decaying spectral forms, and the negative components belong to the spectral forms forming in the transition, which provides a clue to interpret the spectral changes corresponding to the apparent rates. The spectral forms involved in the reaction cascade, however, are more conveniently derived from the sequential intermediate spectra shown in Fig. 1 B. Except for the first and last spectra, In1 and In6, the sequential intermediate spectra do not meet the requirements set up for the spectra of molecular states (8); rather, they indicate the presence of a number of states in the sequential intermediates. The sequential scheme is clearly not the mechanism of the reaction. Deconvolution of the sequential spectra yielded four spectral forms, shown in Fig. 1 C. Each spectral form represents a particular state of the retinal chromophore, has a shape characteristic of the dark spectrum when plotted on the energy scale, and can be identified with one or multiple molecular states in the kinetic mechanism. The spectral forms are fairly similar for the different proteins (5). For comparison, the K-like and L-like spectral forms for the D234N mutant are also displayed, as dashed lines.
Figure 1.
Summary of the spectral analysis in (5) shown using the WT pH 7.4 data. (A) Results of the exponential fit showing the five apparent lifetimes and six b-spectra, as in Fig. 4 A of (5). (B) Sequential intermediate spectra In1–In6, as in Fig. 5 A of (5). The spectra are differences between the intermediate and dark spectra. (C) The four spectral forms deconvoluted from the sequential spectra. The difference spectra are analogous to the absolute spectra in Fig. 7 A. The L-like spectrum is taken from In3. The K-like and L-like spectra of D234N (dash) are shown for comparison.
The four spectral forms are a red-absorbing form, red-shifted from the dark spectrum by 38 nm for the WT, called the K-like spectral form, analogous to the spectrum of the K intermediate in the BR photocycle; a form, which is blue-shifted from the dark absorption spectrum by 18 nm for the WT, called the L-like spectral form, analogous to BR’s L intermediate; a very blue-shifted, around 120 nm from the dark absorption spectrum for the WT, form representing the deprotonated chromophore, called the M spectral form, similar to BR’s M intermediate; and a spectral form for the recovered state, called the R spectral form, which is blue-shifted from the dark spectrum by 4 nm for the WT, and thus has the smallest amplitude among the difference spectra presented. In addition to these forms, the S97E variant showed an early L-like form, the L′-like spectral form, which is blue-shifted from the dark spectrum only by 3 nm and thus almost identical to the R spectral form. The spectral shifts cited are estimates from the positions of the absolute spectra constructed for the WT and S97E mutant in our previous work (5). The deconvolution of the sequential spectra yields their composition in terms of spectral forms. For the WT protein shown above, this is as follows: In1 = 1.0 K-like, In2 = 0.72 K-like + 0.24 L-like, In3 = 0.56 K-like + 0.44 L-like, In4 = 0.51 K-like + 0.49 L-like, In5 = 0.14 K-like + 0.20 L-like + 0.20 M + 0.46 R, and In6 = 1.0 R. The composition is included in the composition matrix for each protein. The product of the composition matrix and the matrix of time-dependent concentrations of the sequential intermediates yields the time evolution of the spectral forms shown in Fig. 2 for the WT protein.
Figure 2.
The time-dependent concentration profiles of the K-like (solid), L-like (dash), M (dot), and R (solid) spectral forms for the WT at pH 7.4, as in Fig. 8 A of (5). The K-like and L-like forms of D234N at pH 7.4 are shown for comparison (dash-dot).
In addition to the K-like (solid line), L-like (dashed line), M (dotted line), and R (solid line) spectral forms of the WT protein, the K-like and L-like spectral forms of the D234N mutant are also shown for comparison (dash-dot lines). The D234N is a low-conductance mutant. Nevertheless, it displays a much higher concentration of the L-like spectral form than the WT does. It is the K-like spectral form that has low amplitude in the time window of the channel current.
Based on the lifetimes and b-spectra of the global exponential fit, together with the time evolutions of the spectral forms reported recently (5), we can draw the following conclusions.
-
(1)
The first two b-spectra in the submicrosecond and microsecond time range are consistent with the decay of a K-like intermediate into two L-like intermediates. This is similar to BR kinetics except the transition occurs in two steps. The analogy with BR kinetics, however, ends here because the K-like and L-like spectral forms are present during the entire time range of the photocycle, over four to five decades of time. The continuous presence of these forms supports the notion that multiple, structurally different isospectral molecules are present in the reaction cascade.
-
(2)
There are no obvious changes in the visible spectral range that could be connected to the channel current onset. This is unusual regardless of which intermediate is responsible for the channel opening. The opening of the anion channel presumably involves structural changes, perhaps even helical movements, which usually influence the chromophore visible spectra. We do not believe that the L formation, which is over in a few microseconds, is directly connected to the channel opening, which takes tens or hundreds of microseconds to complete. The longer, 18-μs conformational change reported by FTIR experiments is likely to be associated with channel opening; however, its connection to the fast K-to-L visible transition and the L intermediate is questionable. The longer lifetimes reported for L formation in visible absorption experiments, which were connected to the channel opening earlier (3,4), were possibly the result of poor time resolution in the early microsecond time range by the single-wavelength recording technique employed in those studies.
-
(3)
For all protein samples, the time evolution of the spectral forms shows high levels of the L-like form in the time interval of the open channel state. Moreover, it is the highest for those variants that reportedly display low or very low channel current, such as the D234N and S97E mutants. This finding clearly contradicts the current view on anion channel kinetics that presumes a BR-like photocycle and assigns the open channel state to one of the L intermediates (3,4). Consistent with the current view, one would expect low levels of the L-like form for the low-conducting mutants. Interestingly, it is not the L-like but the K-like spectral form whose concentration is consistent with the observed current levels.
The arguments listed above make the current interpretation of anion channelrhodopsin kinetics and the notion that the L intermediate is responsible for the open channel state highly questionable.
Comparison between current transients and spectral form evolutions: The open channel state is a red-absorbing intermediate
The question of which intermediate is the open channel state is of primary importance for understanding channel kinetics. In recent publications, a BR-like photocycle was postulated, and its L intermediate was proposed for the open channel state (3,4). We found no correlation between the channel current levels and the amplitudes of the blue-shifted L-like spectral form (5). Contrary to the current view, our results favor the red-absorbing K-like spectral form for the open channel state. Comparison between the published channel current transients and the time evolutions of the blue- and red-absorbing spectral forms presented below provides further evidence for the red-absorbing form being the open channel state.
There is no correlation between the channel current onset times and the lifetimes of the detected visible spectral transitions. Thus, the most probable spectral form responsible for channel opening has to be deduced from the rest of the current trace, from its plateau and characteristic decay regions. Accurate spectroscopic measurements require relatively high protein concentration and low level of light scattering, which cannot be achieved using cell suspension samples. The measurements are done using protein samples solubilized in detergent, whereas the current is recorded on proteins embedded in membrane; thus deviations in the kinetics measured by the two techniques are not unexpected. When appropriate, the gap observed between the current and spectral traces can be narrowed or completely eliminated by altering the current trace within acceptable limits. For this, the experimental current trace is approximated by sums of exponential functions: one or two exponentials for the current rise and two falling exponentials for its fast and slow decays. Altering the lifetimes and amplitudes of the exponential components obtained in the fit within reasonable limits, a new current trace can be produced by combining the altered exponential functions. The result of this procedure is a new, altered current trace, the fast and slow decaying components of which are shifted on the time axis, and their amplitudes are modified.
In Figs. 3 and 4, the time evolutions of the K-like and L-like spectral forms, solid and dashed lines, respectively, are compared with the respective current traces, dotted lines, normalized to the levels of the spectral traces. The time evolutions were calculated based on the composition matrices of the sequential intermediates, as described above and earlier (5).
Figure 3.
Comparison of the normalized channel current to the K-like (solid) and L-like (dash) spectral forms for the WT and A75E mutant. (A) WT at pH 7.4. The current recorded at positive (dot) and negative (dash-dot) potentials (6). (B) WT at pH 8.5. The current recorded at pH 7.4 (dash-dot) was shifted (dot) based on the published information (6). (C) A75E at pH 7.4. The current (dot) recorded at −30 mV (7) is normalized to the spectral traces.
Figure 4.
Comparison of the channel current to the K-like (solid) and L-like (dash) spectral forms for the D234N and S97E mutants. (A) D234N at pH 4.5. The published current (6) is normalized (dots) to the spectral traces. (B) D234N at pH7.4. The published current (6) (dash-dot) is slowed down (dots) for a match with the K-like form. (C) S97E at pH 7.4. Published current (6) is normalized (dots) for a match to K-like and L-like traces.
WT protein
For the WT at pH 7.4, Fig. 3 A, the current traces recorded at both positive and negative 60-mV potentials were published, dot and dash-dot, respectively (6). The K-like and L-like spectral traces correspond to Fig. 8 A in (5). Although the channel closing kinetics displayed by the current traces do not follow the K-like or L-like spectral forms perfectly, the mismatch is not significant. A very slight alteration of the fast and slow current decays is sufficient to bring the current and spectral traces together. Thus either of the K- and L-like spectral forms could be assigned to the open channel state based on the current and spectral form comparison. Indeed, in analogy with the steps seen in the BR photocycle, the open channel state was assigned to the L intermediate (3). The K-like and L-like spectral traces for WT at pH 8.5 correspond to Fig. 10 B in (5). The current trace for the WT at pH 8.5, Fig. 3 B, was not published. However, it was reported (6) that the fast decay rate of the current at pH 8.5 is three to four times higher than at pH 7.4. The current trace for pH 8.5 was created based on the average current trace recorded at pH 7.4, dash-dot, by multiplying its fast decay rate by a factor of 3.3 and reducing the amplitude of the slow decaying current component by a factor of 2.2. This resulted in a perfect match between the K-like spectral form and the created current trace, solid and dotted lines, respectively. Because the contribution of the slow component to the recorded current trace was already small, the amplitude reduction affected the current shape only slightly. It is very clear that accelerating the fast decay rate of the current recorded at pH 7.4 by any factor would bring the current trace closer to the K-like spectral form and further away from the L-like form. This strongly suggests that the intermediate associated with the open channel state must be a red-absorbing molecule.
A75E mutant
The time evolutions of the K-like and L-like spectral forms for the A75E mutant, shown in Fig. 3 C, are taken from (5), Fig. 8 C. Due to the presence of a carboxyl group near the extracellular end of the tunnel, the shape and the amplitude of the current trace for this mutant are dependent on the polarity of the applied voltage (7). At positive voltages, the current is suppressed and highly nonlinear, whereas at negative potential, it is more normal. The current trace shown, dash-dot, was recorded at negative 30-mV potential where the influence of the carboxyl group is not very significant, and reasonable comparisons between the current and the spectral traces can be made. The normalized current traces, dotted lines, follow the paths of both spectral forms reasonably well; thus both the K- and L-like forms are candidates for the open channel state.
D234N mutant
The amplitudes of the current traces shown in Fig. 3 are reasonably high, and the currents can be recorded with relative ease. The D234N mutant shown in Fig. 4 poses more difficulties and is regarded as a low-conductance mutant. The current traces for the D234N mutant are available at pH 5.4 and 7.4 (3), dotted lines in Fig. 4 A and B, respectively. The current for this protein has low amplitude compared with the WT and decays in two well-separated steps. The K-like and L-like spectral traces in Fig. 4 A and B correspond to Figs. 10 C and 8 B in (5), respectively. The K-like spectral forms have also low amplitude and follow this unique current shape reasonably well. For better comparison, their traces in Fig. 4 A and B are multiplied by a factor of two. The L-like spectral forms have high amplitude and do not follow the current shape. In Fig. 4 A, the normalized published current trace of pH 5.4, dotted line, matches the K-like spectral form of pH 4.5 almost perfectly, whereas the L-like spectral form follows a completely different decay path. Thus, the intermediate responsible for the open channel must be a red-absorbing intermediate.
The normalized current published for the D234N variant at pH 7.4, Fig. 4 B dash-dot line, and the K-like spectral form have similar shapes, though the current decays somewhat faster than the spectral form does. Shifting the current trace on the time axis by slowing down the fast and the slow current decay rates by factors of 3.7 and 1.9, respectively, results in a perfect fit to the decay of the K-like spectral form, dotted line. Note that the amplitudes of the fast and slow current decay components are not altered during their shift on the time axis, so their ratio is not affected. Reaction rate differences of this magnitude caused by environmental factors are not unusual. Note that a slight increase in the energy level of the activation barrier, around the value of the thermal energy kT, will slow down the rate of the reaction by a factor of 2.7. This is only around a 3% increase in the height of an average activation barrier of 20 kcal/mol. Since the L-like spectral form decays practically in one slow step, its shape is very different from the two-step current decay. Again, the intermediate responsible for the open channel state must be a red-absorbing intermediate.
S97E mutant
The S97E mutant has very low conductance (3), and its absorption kinetics is quite different from the rest of the proteins studied (5). The spectral traces are taken from Fig. 8 D of (5). It shows an early L′-like spectral form that converts into the normal L-like spectral form on the ∼1-ms timescale. The former is almost identical to the recovered R spectral form, whereas the latter is blue-shifted by ⁓14 nm. The normalized published current trace for the S97E variant at pH 7.4 (3) is shown in Fig. 4 C, dotted line. Its decay follows the decay of the L-like and also the final decay steps of the K-like spectral form. Based on this alone, both of these spectral forms could be assigned to the open channel state.
Assignment of the L-like intermediate, however, is problematic for the following reasons. First, there is a mismatch between the high amplitude of the L-like spectral form and the low amplitude of the channel current. Second, two spectrally different L-like forms are detected in the time window of the open channel state, and both of them have to be assigned to the open channel state: the early L′-like form at the onset of the current and the late L-like form at the closure of the channel. A spectral change of ∼14 nm in the L′ to L transition occurs on the ∼1-ms timescale, right at the peak of the channel current. Despite this, the channel current shows apparently no sign of disturbance that could be associated with this spectral transition. The big spectral change, presumably, reflects structural rearrangements around the chromophore, and it is expected to cause noticeable changes in the current profile if the L intermediate is the open channel state. Third, if the L′-like and L-like traces make up the open channel, their sum is expected to follow the current profile. This is not the case; the two shapes are significantly different. It is thus very unlikely that the L intermediate is the open channel state.
Under the assumption that the K-like spectral form represents a single intermediate spanning the entire time range, its assignment to the open channel state is also problematic. There is a big drop in its concentration in the time interval of the current rise. However, this obstacle can be easily eliminated by assuming that there are two isospectral red-absorbing intermediates that evolve simultaneously but only one of them is associated with the channel opening. The change in concentration of the K-like spectral form at the current onset belongs to the intermediate not connected to the channel directly.
The kinetics is inconsistent with the single photocycle mechanism
The simplest way to arrange intermediates of a reaction is to place them in a single chain. Parallel chains are more complex to deal with and are considered only if the single-chain concept fails. Our goal here is to test the single-chain concept. The kinetic results show a K intermediate converting into L intermediates reversibly within a few microseconds. This early equilibration between the K and L intermediates presets the ratio of their concentrations, and within the concept of a single reaction chain, this ratio should remain constant during the lifetimes of the intermediates. The time dependences of the spectral forms for the different proteins clearly show that the K-to-L ratio changes in time. This indicates that, as the reaction progresses, either K or L, or both of these earliest intermediates convert into Ki and Li molecules that are in different equilibria. This suggests a sophisticated kinetic network in which one of the putative K-like intermediates could, in principle, be associated with the open channel state. The situation becomes more complex in case of the S97E mutant in Fig. 4 C, which shows a well-defined transition into a late “normal” L-like spectral form, and this seemingly involves both the K- and L′-like forms. It would be a formidable task to arrange reaction steps of this complexity in a single chain. In addition, neither of the K- and L-like forms of a single chain follows the current trace within acceptable limits.
The single-chain concept faces additional and even more severe problems at the times of channel closure, as best illustrated by the WT kinetics at pH 8.5. At high pH, a considerable amount of the chromophore deprotonates and transiently produces an M intermediate. Within the single-chain concept, the Kn and Ln intermediate pair that produces M should maintain their concentration ratio unchanged. As the Kn and Ln intermediate pairs cannot act independently, the rise of M concentration must be accompanied by fall of both the Kn and Ln concentrations in a ratio determined by the equilibrium between them. As seen for the WT at pH 8.5, Fig. 3 B, the K-to-L ratio does not stay constant but changes dramatically during the M formation in the 10- to 100-ms time window. This cannot be reconciled with the single-chain concept. Similar changes are seen in all experiments where a sufficient amount of M is formed during channel closure.
The complications listed above make the single-chain concept untenable for a quantitative description of channelrhodopsin kinetics. Introducing parallel photocycles with two independent reaction chains, each starting with its own K intermediate, the restrictions and complications caused by placing the K-like and L-like forms into a single chain no longer apply. We propose that the anion channelrhodopsin GtACR1 follows parallel photocycles. In our future work, we will extend the kinetic analysis to include the concept of parallel reaction chains.
Conclusion
The arguments presented above do not favor the current view on GtACR1 kinetics. Instead, they strongly support the idea that an intermediate having a K-like absorption spectrum is responsible for channel opening. Thus, we propose a red-absorbing intermediate for the open channel state. This is not the K intermediate present at the earliest times but an isospectral intermediate emerging at later times.
Author contributions
I.S.: data analysis, primary manuscript writer; D.S.K.: manuscript preparation.
Acknowledgments
We thank Elena Govorunova for providing the electrophysiological records.
This work was supported by National Institutes of Health grant R01EY029343.
Declaration of interests
The authors declare no competing interests.
Editor: Gabriela Popescu.
References
- 1.Govorunova E.G., Sineshchekov O.A., et al. Spudich J.L. Natural light-gated anion channels: A family of microbial rhodopsins for advanced optogenetics. Science. 2015;349:647–650. doi: 10.1126/science.aaa7484. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lozier R.H., Bogomolni R.A., Stoeckenius W. Bacteriorhodopsin: a light-driven proton pump in Halobacterium Halobium. Biophys. J. 1975;15:955–962. doi: 10.1016/S0006-3495(75)85875-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sineshchekov O.A., Li H., et al. Spudich J.L. Photochemical reaction cycle transitions during anion channelrhodopsin gating. Proc. Natl. Acad. Sci. USA. 2016;113:E1993–E2000. doi: 10.1073/pnas.1525269113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Dreier M.A., Althoff P., et al. Gerwert K. Time-resolved spectroscopic and electrophysiological data reveal insights in the gating mechanism of anion channelrhodopsin. Commun. Biol. 2021;4:578. doi: 10.1038/s42003-021-02101-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schleissner P., Szundi I., et al. Kliger D.S. Isospectral intermediates in the photochemical reaction cycle of anion channelrhodopsin GtACR1. Biophys. J. 2023;122:4091–4103. doi: 10.1016/j.bpj.2023.09.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sineshchekov O.A., Govorunova E.G., et al. Spudich J.L. Gating mechanisms of a natural anion channelrhodopsin. Proc. Natl. Acad. Sci. USA. 2015;112:14236–14241. doi: 10.1073/pnas.1513602112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Sineshchekov O.A., Govorunova E.G., et al. Spudich J.L. Opposite charge movements within the photoactive site modulate two-step channel closing in GtACR1. Biophys. J. 2019;117:2034–2040. doi: 10.1016/j.bpj.2019.10.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Szundi I., Li H., et al. Kliger D.S. Platymonas subcordiformis Channelrhodopsin-2 Function: I. THE PHOTOCHEMICAL REACTION CYCLE. J. Biol. Chem. 2015;290:16573–16584. doi: 10.1074/jbc.M114.631614. [DOI] [PMC free article] [PubMed] [Google Scholar]