Gating mechanisms of a natural anion channelrhodopsin

Significance

The recently discovered natural anion channelrhodopsins (ACRs) have exceptional properties for use as membrane-hyperpolarizing tools for neural silencing, but our knowledge about them is only rudimentary. We have gained, to our knowledge, first insights into the molecular mechanisms of their unique light-gated anion conductance by photocurrent measurements and mutant analysis. We identified a mutant constitutively open in the dark—to our knowledge, the first such mutant of any light-gated channel, invaluable for probing the structure of the channel open state—and a “step function” ACR mutant useful in some optogenetics protocols. Our results will help decipher the mechanism of channelrhodopsin ion conductance and facilitate rational design of inhibitory tools tailored to the needs of optogenetic studies and potential clinical applications.

Abstract

Anion channelrhodopsins (ACRs) are a class of light-gated channels recently identified in cryptophyte algae that provide unprecedented fast and powerful hyperpolarizing tools for optogenetics. Analysis of photocurrents generated by Guillardia theta ACR 1 (GtACR1) and its mutants in response to laser flashes showed that GtACR1 gating comprises two separate mechanisms with opposite dependencies on the membrane voltage and pH and involving different amino acid residues. The first mechanism, characterized by slow opening and fast closing of the channel, is regulated by Glu-68. Neutralization of this residue (the E68Q mutation) specifically suppressed this first mechanism, but did not eliminate it completely at high pH. Our data indicate the involvement of another, yet-unidentified pH-sensitive group X. Introducing a positive charge at the Glu-68 site (the E68R mutation) inverted the channel gating so that it was open in the dark and closed in the light, without altering its ion selectivity. The second mechanism, characterized by fast opening and slow closing of the channel, was not substantially affected by the E68Q mutation, but was controlled by Cys-102. The C102A mutation reduced the rate of channel closing by the second mechanism by ∼100-fold, whereas it had only a twofold effect on the rate of the first. The results show that anion conductance by ACRs has a fundamentally different structural basis than the relatively well studied conductance by cation channelrhodopsins (CCRs), not attributable to simply a modification of the CCR selectivity filter.

Recently we reported a class of rhodopsins from the cryptophyte alga Guillardia theta that act as anion-conducting channels when expressed in cultured animal cells and therefore can be used to suppress neuronal firing by light (1). These proteins—named anion channelrhodopsins (ACRs)—show distant sequence homology to cation channelrhodopsins (CCRs) from chlorophyte (green) algae (2), but completely lack permeability for protons and metal cations. This property, as well as large current amplitudes and fast kinetics, makes ACRs superior hyperpolarizing optogenetic tools, compared with proton and chloride pumps (3, 4), or engineered Cl⁻-conducting CCR variants (5, 6). Two G. theta (Gt) ACRs differ in their spectral sensitivity and channel kinetics (1). GtACR1, with its absorption peak at 515 nm (Fig. S1), has an advantage over a more blue-shifted GtACR2 with maximal efficiency at 470 nm, because it allows using light of longer wavelengths that is less scattered by biological tissue.

Fig. S1.

The action spectra (dashed lines) of WT GtACR1 photocurrents recorded at: −60 mV, bath pH 7.4; 40 mV, bath pH 7.4; −60 mV, bath pH 9; and photocurrents from the GtACR1_E68Q mutant (−60 mV, bath pH 7.4). The solid line shows the absorption spectrum of detergent-purified Pichia-expressed GtACR1.

Because a CCR could be altered to exhibit Cl⁻ channel activity by mutation of a single amino acid residue (5), a fundamental question about ACRs is whether they differ from CCRs only by their selectivity filter. In our search for an answer, we analyzed photocurrents generated by GtACR1 in HEK293 cells under single-turnover conditions in which secondary photochemistry does not complicate the photocycle. We found that GtACR1 conductance comprises two distinct mechanisms with opposite dependencies on the membrane voltage and bath pH, but the same ionic selectivity. The first mechanism, observed by a fast component of the photocurrent decay, depends on Glu-68, the position occupied by a positively charged residue in the engineered Cl⁻-conducting CCR mutant. Substitution of Glu-68 with Arg in GtACR1 did not change anion selectivity of the channel, but, rather, inverted its gating, rendering it constitutively open in the dark and closed in the light. The second mechanism, reflected by a slow component of the current decay, involves Cys-102. Our results demonstrate that the structural basis for anion conductance in ACRs is fundamentally different from that of Cl⁻-conducting CCR mutants.

Results

The rise and decay of photocurrents generated by GtACR1 response to a 6-ns laser flash under our standard ionic conditions (150 mM NaCl in the bath and 126 mM KCl in the pipette, pH 7.4; for other components, see SI Materials and Methods) were both at least biphasic (Fig. 1A). Exponential functions were fit to the data to derive amplitudes and time constants (τ) of each component. At the holding voltage (E_h) −60 mV, τ of the major rise component (∼95%) was 86 ± 15 μs, accompanied by a minor component with τ 1.8 ± 0.3 ms (mean ± SEM; n = 5). In some cases, this minor rise component could be further deconvoluted in two, but this additional complexity was not addressed in the present study. The two decay components had comparable amplitudes and τ values of 48 ± 2 ms and 245 ± 20 ms (mean ± SEM; n = 5). A third, very slow decay component could also be resolved, but its contribution was ≤0.5% and was not considered here.

Fig. 1.

The voltage dependence of the kinetic components of the current. (A) Typical photocurrents generated by GtACR1 in a HEK293 cell in response to a laser flash (6 ns, 532 nm) at the holding voltages (E_h) indicated. Ten individual sweeps recorded with 5-s intervals at each E_h were averaged, smoothed by logarithmic adjacent averaging, and fit with four exponential functions to derive the time constants (τ) shown for the trace recorded at −60 mV in this particular cell. For mean τ values for an ensemble of cells, see text. The experimental data are shown as black solid lines and the fitted curves as red dashed lines. (B) The voltage dependence of the amplitudes of the two decay components determined by exponential fit as shown in A, measured in a typical cell in the standard bath solution (filled symbols) and upon substitution of F⁻ for Cl⁻ (open symbols). (C) The voltage dependence of τ values of the two decay components. The data points are mean values ± SEM (n = 14). (D) The voltage dependence of the contributions of the fast decay (left axis) and slow rise (right axis) to the total amplitude. The data points are mean values ± SEM (n = 9).

Previously we reported that the voltage dependence of the peak GtACR1 current recorded in response to a pulse of continuous light is linear (1), in contrast to inward rectification typical of the current–voltage relationships of CCRs (7). Deconvolution of laser-flash–evoked currents revealed that the voltage dependence of the amplitude of the fast decay showed outward rectification, whereas that of the slow component showed inward rectification (Fig. 1B). The fast decay phase strongly accelerated upon shifting E_h to more positive values, but the slow phase became even slower (Fig. 1C). The rates of the two components of the rise were nearly voltage-independent (Fig. S2A). However, the contribution of the slow rise component to the total amplitude increased at more positive voltages, as did that of the fast decay (Fig. 1D). These results indicate that GtACR1 channel conductance comprises two mechanisms, one with the slow rise and fast decay and another with the fast rise and slow decay, which are altered in opposite way by changes in membrane voltage. The biphasic rise and decay cannot be explained by two preexisting populations of GtACR1 undergoing different photocycles, because the relative contributions of the rise components do not match those of the corresponding decay components (Fig. 1D). We measured the action spectra of photocurrents under conditions that dramatically alter relative contribution of the two gating mechanisms. Within the accuracy of our measurements, the spectral sensitivity did not depend on the voltage, the bath pH, or even on a mutation of the key residue (see below) that controls fast channel closing (Fig. S1). These findings confirm the absence of spectrally identifiable heterogeneity of the unphotolyzed state of GtACR1.

Fig. S2.

(A) The voltage dependence of τ values for the two rise components of photocurrent generated by GtACR1 in HEK293 cells. The data points are mean values ± SEM (n = 14). (B) The dependence of the slow decay rate measured at 0 mV on the pH of the bath. The data points are mean values ± SEM (n = 4–10).

The two conducting forms have the same relative permeability for Cl⁻: When Cl⁻ in the bath was replaced with F⁻, the current–voltage relationships for the amplitudes of both decay components shifted to the same extent (Fig. 1B, open symbols). However, replacement of Cl⁻ in the bath with F⁻ or Asp⁻ strongly accelerated the fast current decay (Fig. S3), whereas the slow decay was not changed. A similar acceleration of closure when F⁻ ions pass through the channel has been observed in neuronal gamma-aminobutyric acid (GABA)-activated chloride channels (8).

Fig. S3.

The influence of bath anions on the decay of laser-evoked photocurrents generated by the GtACR1 in HEK293 cells. The experimental data (thin solid lines) were fit with two exponentials (dashed lines), from which indicated τ values were derived.

The two conducting forms of GtACR1 also showed opposite dependencies on the bath pH. The fast decay strongly accelerated (Fig. 2A and black filled symbols in Fig. 2B), whereas the slow decay was much less sensitive and slowed upon an increase in bath pH (Fig. S2B).

Fig. 2.

The influence of pH on the fast current decay. (A) The voltage dependence of the fast decay τ at the indicated pH of the bath. The data points are mean values ± SEM (n = 4–10 cells). (B) The dependence of the fast decay rate measured at 0 mV on the pH of the bath (filled symbols, black for WT and red for the E68Q mutant) and the pipette (black open symbols, WT). The data points are mean values ± SEM (n = 4–10 cells).

The fast decay rate was nearly completely insensitive to the intracellular pH (Fig. 2B, black open symbols), whereas it exhibited a strong dependence on the extracellular pH with a pK_a ∼9.7 (Fig. 2B, black filled symbols). Therefore, we hypothesized that deprotonation of a specific amino acid residue accessible from the extracellular side of the membrane is required for fast channel closing in GtACR1. Although the pK_a values of the Asp and Glu side chains in solution are much lower than 9.7, it is known that the protein environment can significantly increase them (9). Therefore, we started a mutagenesis scan for this critical residue by individually substituting Asn for Asp and Gln for Glu in the transmembrane domain of GtACR1. The positions of the mutated residues in our GtACR1 homology model are shown in Fig. S4. Glu-60 was also included in this analysis, although its predicted location is close to the cytoplasmic surface, because this position corresponds to Glu-82 of Chlamydomonas reinhardtii channelrhodopsin 2 (CrChR2), in which its replacement with Ala strongly inhibited photocurrents (10). All but one tested GtACR1 mutant exhibited biphasic current rise and decay, as did wild type (WT), although the rates and contributions of the phases to the total amplitude quantitatively varied (Fig. S5).

Fig. S4.

GtACR1 homology model built by the Robetta server (robetta.bakerlab.org/) using the 3ug9 structure as a template. The side chains of amino acid residues, mutations of which have been tested in this study, plus that of Lys-238, which forms the Schiff linkage with retinal, are shown in color corresponding to the color of the helix in which the residue is located.

Fig. S5.

The decay of laser-evoked photocurrents generated by GtACR1 mutants containing carboxylate-neutralizing mutations Asp → Asn and Glu → Gln (solid red lines) compared with that of the normalized WT signal from Fig. 1A (black dashed line). None of these mutations caused monoexponential closing of the channel. The indicated τ values were derived by two-exponential fit of the decay.

In contrast, both the rise and decay of photocurrents recorded at the bath pH 7.4 from the GtACR1_E68Q mutant were monoexponential (Fig. 3A, red line). At higher pH values, however, a biphasic current rise and decay were observed also in this mutant (Fig. 3A, blue line). The contributions of fast-closing and slow-opening components in the mutant strongly diminished in parallel until vanishing upon a decrease in pH (Fig. 3B, circles and solid lines). Therefore, the remaining monoexponential opening and closing of the channel observed in the mutant at neutral pH corresponds to the fast rise/slow decay gating mechanism in WT. The rate of fast decay in the mutant was severalfold slower (Fig. 2B, red symbols) than in WT at the same pH. The pH dependence of the slow decay rate was similarly weak as in WT, but changed in the opposite direction (Fig. 3B, squares and dashed line). The suppression of the fast decay component by E68Q mutation was also strongly supported by measurements in the double E68Q_C102A mutant (see below). Our interpretation of these results is that, in addition to the Glu-68 residue, fast closing of the channel involves another yet unidentified pH-sensitive group X. The E68Q mutation led to a substantial suppression of the gating mechanism associated with the slow opening/fast closing of the channel. However, this mutation did not influence the current amplitude (Fig. S6A) and anion permeability of GtACR1: The E_rev of photocurrents recorded from the mutant were close to that of WT in both the standard bath solution and upon replacement of Cl⁻ with F⁻ (Fig. S6B).

Fig. 3.

(A) Typical photocurrents generated at −60 mV by the GtACR1_E68Q mutant at the bath pH 7.4 (solid red line) and 9.4 (solid blue line). The current traces were normalized at the peak value and fit with multiexponential functions (dashed lines) that yielded the indicated τ values for the current rise and decay. The normalized WT current from Fig. 1A was fitted with four exponentials (black lines), and its τ values are shown for comparison. (B) The influence of the bath pH on the slow rise (open circles) and fast decay (filled circles) contribution (left axis) and the slow decay rate (filled squares; right axis) in the GtACR1_E68Q mutant. The data points are mean values ± SEM (n = 3–5 cells).

Fig. S6.

(A) Mean peak amplitudes of laser-evoked photocurrents generated by WT GtACR1 and its E68Q, C102A, and S130A mutants. The data are mean values ± SEM (n = 3–5 cells). (B) The reversal potentials of peak photocurrents by WT GtACR1 and its E68Q mutant with indicated anions in the bath. The data are mean values ± SEM (n = 3–5 cells).

To test the effect of a positive charge in the Glu-68 position, we replaced it with Arg. The GtACR1_E68R mutant demonstrated a unique “inverted” behavior, which is best illustrated by application of a continuous light pulse: Illumination caused currents to change in the opposite direction than that observed in any thus-far-tested ChRs (Fig. 4A). This behavior implies that switching on the light caused a decrease, rather than an increase in GtACR1_E68R conductance—i.e., the GtACR1_E68R channel is constitutively open in the dark and closes upon illumination.

Fig. 4.

Photocurrents from the inverted GtACR1_E68R mutant. (A) Typical photocurrents generated in response to a 1-s light pulse by the GtACR1_E68R mutant in a HEK293 cell at voltages from −60 mV (bottom trace) to 60 mV (top trace) changed in 20-mV steps. (B) The current–voltage relationships measured in the GtACR1_E68R mutant in the dark. The data are mean values ± SEM (n = 3 or 4 cells) normalized to the value measured at −60 mV in the standard (Cl⁻) bath. (C) The E_rev shifts measured for currents generated by WT GtACR1 and the GtACR1_E68R mutant in the dark upon switching from the standard (Cl⁻) bath solution to that with F⁻ or Asp⁻ (black bars) and upon illumination in the same indicated bath (red bars). The data points are mean values ± SEM (n = 3 or 4 cells). (D) A typical photocurrent trace generated by GtACR1_E68R in response to a laser flash at −60 mV (solid line). The dark current level was subtracted. The red dashed line is a multiexponential fit of the signal yielding indicated τ values. (E) The voltage dependencies of the amplitudes of the rise (channel closure) components of the laser-evoked signals generated by GtACR1_E68R in a typical cell. (F) The voltage dependence of the decay τ values of the laser-evoked signals. The data points are mean values ± SEM (n = 3 cells).

To further test this interpretation, we measured current–voltage relationships (IE curves) and compared E_rev shifts observed upon substitution of poorly permeable F⁻ or nonpermeable Asp⁻ for Cl⁻ in the bath for the WT GtACR1 and its E68R mutant. First, we analyzed current responses to voltage steps applied in the dark. As expected of leak currents, in the WT, the IE curves were linear and insensitive to the bath exchange. In contrast, the IE curves in the mutant changed their shape and shifted to more positive voltages upon the bath exchange (Fig. 4B and black solid bars in Fig. 4C), which showed that in the mutant, a specific anion conductance was present in the dark, as expected from an open GtACR1_E68R channel.

Next, we examined the behavior of the IE curves in response to illumination. In the WT, illumination caused a positive shift of the E_rev, which was greater for nonpermeable Asp⁻ in the bath than for poorly permeable F⁻ (Fig. 4C, red hatched bars), as expected (1). In other words, illumination caused the appearance of Cl⁻ conductance in the WT (channel opening). In contrast, in the mutant illumination caused a negative shift of E_rev shift with either F⁻ or Asp⁻, indicating closing of Cl⁻ channels (Fig. 4C, red solid bars).

Photocurrents from the E68R mutant elicited by a laser flash could be fit with only three exponentials (two for the rise—i.e., closing of the channel—and one for the decay—i.e., reopening of the channel; Fig. 4D), but a better fit was obtained when the slower rise component was split into two components. The E_rev of both rise components was the same (Fig. 4E). All current components accelerated upon the membrane depolarization, but the decay was most sensitive to it (Fig. 4F). The light-induced closing of the channel (the current rise) was preceded by a fast (τ ∼40 μs) negative component, the amplitude of which varied from cell to cell. Although its nature is not yet clear, it may reflect a fast electrical response related to retinal isomerization.

Cys-102 in GtACR1 corresponds to Cys-128 in CrChR2, the mutation of which leads to a dramatic decrease of the current decay rate [so-called step-function tools (11)]. In the GtACR1_C102A mutant, the fast current decay (30% contribution) was only twofold slower (τ 91 ± 2 ms; n = 5 cells) than that in the WT, indicating that Cys-102 was relatively unimportant for the fast decay conductance mechanism (Fig. 5, red line). In contrast, the slow decay in this mutant was dramatically slower than that in the WT and comprised two phases with 38% and 32% contributions and τ values 7.5 ± 0.9 and 53 ± 9 s, respectively (n = 5 cells). In the double mutant GtACR1_E68Q_C102A, the contribution of the fast current decay was reduced approximately fourfold, and its rate slowed compared with the single C102A mutant (Fig. 5, green line). This result further confirms that Glu-68 is a key residue that controls fast channel closing.

Fig. 5.

The decay of typical photocurrents generated by GtACR1_C102A (solid red line), GtACR1_E68Q_C102A (solid green line), and GtACR1_S130A (solid blue line) at −60 mV in the standard bath solution. The traces were normalized at the peak amplitude. The corresponding dashed lines show two- or three-exponential fits used to determine the decay time constants for the mutants, as shown on the graph.

Cys-128 in CrChR2 has been proposed to form a hydrogen bond with Asp-156 in the fourth transmembrane helix [the so-called “DC” gate (12)]. The corresponding residue in GtACR1 is Ser-130, which, according to our homology model, is predicted to form a hydrogen bond with Cys-102 (Fig. S4). However, replacement of Ser-130 with Ala did not greatly influence the rate of channel closing, with that of both phases becoming only ∼50% reduced in the mutant (Fig. 5, blue line) compared with the WT, in a striking contrast to the effect of the corresponding D156A mutation in CrChR2 (13).

SI Materials and Methods

Molecular Biology.

A DNA polynucleotide encoding the seven-transmembrane (7TM) domain (residues 1–295) of G. theta ACR1 optimized for human codon use was synthesized (Genewiz) and cloned into the mammalian expression vector pcDNA3.1 (Life Technologies) in frame with an enhanced YFP tag. Mutations were introduced by using a QuikChange XL site-directed mutagenesis kit (Agilent Technologies) and verified by DNA sequencing.

HEK293 Electrophysiology.

HEK293 (human embryonic kidney) cells were transfected by using the ScreenFectA transfection reagent (Waco Chemicals USA). All-trans-retinal (Sigma) was added as a stock solution in ethanol at a final concentration of 5 µM. Measurements were performed 48–72 h after transfection with an Axopatch 200B amplifier (Molecular Devices). The signals were digitized with a Digidata 1440A by using pClamp 10 software (both from Molecular Devices) at the sampling rate of 4 µs per point. Patch pipettes with resistances of 2–5 MΩ were fabricated from borosilicate glass and filled with the following solution (in mM): KCl 126, MgCl₂ 2, CaCl₂ 0.5, EGTA 5, and Hepes 25, pH 7.4. The bath solution contained (in mM): NaCl 150, CaCl₂ 1.8, MgCl₂ 1, glucose 5, and Hepes or Tris 10; pH was adjusted with N-methyl-d-glucamine to the desired value. For measurements of E_rev and changes in signal kinetics, NaCl in the bath was replaced with NaF or sodium aspartate. A 4 M salt bridge was used in all experiments. All measurements were carried out at room temperature (25 °C). Laser excitation was provided by a Minilite Nd:YAG laser (532 nm, pulsewidth 6 ns, energy 12 mJ; Continuum). Continuous light pulses were provided by a Polychrome IV light source (T.I.L.L. Photonics GMBH) in combination with a mechanical shutter (Uniblitz Model LS6; Vincent Associates; half-opening time 0.5 ms). Maximal quantum density at the focal plane of the 40× objective lens was 8.5 mW/mm². A laser artifact measured with a blocked optical path was digitally subtracted from the recorded traces. For further analysis, the signals were logarithmically averaged with a custom-created computer algorithm. All current–voltage dependencies were corrected for LJPs calculated by using the ClampEx built-in LJP calculator. Numerical data in the text are presented as the mean values ± SEM. Curve fitting was performed by using Origin 7 software (OriginLab Corp.).

Purification of GtACR1 from Pichia and Absorption Spectroscopy.

The expression construct contained the 7TM domain of GtACR1 and a nine-His tag at the C terminus in the pPIC9K vector backbone (Invitrogen). Geneticin-resistant Pichia pastoris clones were selected according to the manufacturer's instructions. Protein expression and purification on a Ni-NTA column (Qiagen) were carried out as described (9). For absorption measurements, the protein samples were concentrated in 300 mM NaCl, 0.01% DDM, and 20 mM Hepes (pH 7.4). The absorption spectrum of partially purified GtACR1 in the UV-Visible range was recorded on a Cary 4000 spectrophotometer (Varian).

Discussion

We show that both rise and decay of photocurrents generated by GtACR1 upon excitation with short laser flashes (single-turnover conditions) proceed in two phases. Qualitative analysis of the dependencies of these phases on the membrane voltage and the external pH suggests that GtACR1 exhibits two conduction-gating mechanisms. One of these mechanisms is characterized by a fast rise and slow decay, and the other one by a slow rise and fast decay. Our observations raise two possibilities: (i) The same anion gate is controlled by two distinct mechanisms—i.e., there are two different ways to remove the structural block defining the gate. (ii) The anion channel diverges into two branches, each with its own gate, so that if either gate opens, the ions can flow. Distinguishing these possibilities will likely require structural analysis of ACRs and their mutants altered in gating properties.

Selective influence of specific mutations on each of the two components of the photocurrent decay strongly supports the existence of two distinct gating mechanisms of GtACR1. We have identified Glu-68 in the second transmembrane helix of GtACR1 as an essential residue that controls the slow rise/fast decay mechanism, most probably by a protonation change. In contrast, the E68Q mutation has essentially no effect on the fast rise/slow decay mechanism. According to the primary sequence alignment, Glu-68 corresponds to Glu-90, the residue that determines cation selectivity of CrChR2 (7, 14, 15). Mutation of Glu-90 to Lys or Arg converted CrChR2 to an anion channel with residual permeability for protons (5). However, the selectivity filter of ACRs appears to be different from that of the CrChR2_E90K/R mutants, because the presence of the Glu-90 homolog in WT GtACRs is obviously not a barrier to anion permeation. Furthermore, replacement of Glu-68 with Gln or Arg in GtACR1 did not change anion permeability of the channel, which confirmed that this residue does not determine the selectivity filter in ACRs.

Although the E68R mutation did not change the ionic selectivity of the channel, it inverted its gating. The channel became constitutively open in the dark, whereas illumination caused it to close. A similar inversion of protein function was previously observed in haloarchaeal sensory rhodopsin I (SRI), in which a single point mutation either of the photoreceptor itself or of its cognate transducer converted SRI from an attractant to a repellent photoreceptor (16, 17). The SRI inversion is attributable to a switch in its conformation from the C (retinylidene Schiff base accessible from the cytoplasm) to E (Schiff base accessible from the extracellular space) conformer (18–20). The stable open channel of GtACR1_E68R is likely to be particularly valuable for structural comparisons with the WT by enabling identification of residue determinants and structural changes distinguishing the closed and open conformations of the channel by structural methods such as molecular spectroscopy and crystallography.

Our mutant analysis showed that Cys-102 controls only the slow phase of the photocurrent decay, corresponding to the fast opening/slow closing gate, whereas its fast decay phase was practically unaffected by the GtACR1_C102A mutation. This residue (Cys-128 in CrChR2) is conserved in all thus-far-known ACRs and CCRs, but is substituted with Val in the third homologous sequence from G. theta that generated no photocurrents, despite good expression in HEK293 cells (1). The effects of C128X mutations on channel activity have been best studied in CrChR2 (11, 13), but a similar dramatic decrease in the current decay rate was also observed in the corresponding mutants of other CCRs, such as Mesostigma viride channelrhodopsin 1 (MvChR1) (21). This result has been attributed to a disruption of the hydrogen bond (“DC gate”) that Cys-128 forms with Asp-156 in CrChR2 (12). Indeed, mutation of Asp-156 yielded comparable or even greater extension of the channel open time, as did that of Cys-128 (13, 22). In GtACR1, the position of Asp-156 is occupied by Ser-130, which, according to our homology model, forms a hydrogen bond with Cys-102. However, in the S130A mutant, the slow decay phase was practically unchanged, which suggests that the effect of the C102A mutation in GtACR1 was not caused by a disruption of the putative hydrogen bond.

The two gating mechanisms of GtACR1 identified by kinetic analysis of laser-evoked signals control conduction with the same anion selectivity. None of the tested mutations that selectively influenced the kinetics of each form changed anion selectivity of GtACR1. These observations show that in GtACR1 channel gating is structurally independent of the selectivity filter. In contrast, in CrChR2 two conducting states have been recognized by their different relative permeability to H⁺ and Na⁺ (23, 24). A model of two interconnected photocycles, each of which contains a closed (nonconducting) state and an open (conducting) state, has been developed for CCRs (25–27). However, single-turnover photocurrents generated by CrChR2 upon laser excitation decay monoexponentially (28, 29), which means that the second conducting state accumulates only under continuous light, indicating that the second state requires secondary photochemistry. Two exponential phases have been resolved in the decay of laser-evoked photocurrents from the Volvox carteri channelrhodopsin 1/Volvox carteri channelrhodopsin 2 (VcChR1/VcChR2) hybrid (30), but this phenomenon has not yet been incorporated into the four-state model.

The importance of the Cys-128 homolog is not the only similarity between the slow phase of the decay of GtACR1 photocurrent and the monoexponential decay of the current generated by CrChR2. The latter also became slower with membrane depolarization (29, 31). However, in CrChR2, this voltage dependence was eliminated by the E123T mutation (29), whereas in WT GtACR1, the voltage dependence persisted in the presence of a noncarboxylic residue (Ser-97) in the position of Glu-123 (corresponding to Asp-85 in bacteriorhodopsin).

The many differences between ACR and CCR conductance mechanisms evident in our results clearly indicate that ACRs differ from CCRs, not simply by their selectivity filter, but they represent a distinct class of light-gated channels that evolved to conduct anions.

Materials and Methods

A DNA polynucleotide encoding the seven-transmembrane domain (residues 1–295) of GtACR1 optimized for human codon use was synthesized and mutated as described in SI Materials and Methods. Photocurrents were evoked by a Nd:YAG laser (532 nm, pulsewidth 6 ns) and recorded by whole-cell patch clamp from HEK293 cells transiently transfected with WT GtACR1 or its mutants fused to enhanced YFP. A laser artifact measured with a blocked optical path was digitally subtracted from the recorded traces. For further analysis, the signals were logarithmically averaged with a custom-created computer algorithm. For multiexponential curve fitting, Origin 7 software was used. Monochromatic (10 nm half-band) light from a Polychrome IV light source was used in experiments with continuous light pulses. All current–voltage dependencies were corrected for liquid junction potentials (LJPs) calculated by using the ClampEx built-in LJP calculator. Action spectra were constructed by correction of measured currents for quantum density of the stimulation light. SI Materials and Methods provides detailed description of experimental procedures.