Channel gating in kalium channelrhodopsin slow mutants

Abstract

Kalium channelrhodopsin 1 from Hyphochytrium catenoides (HcKCR1) is the first discovered natural light-gated ion channel that shows higher selectivity to K⁺ than to Na⁺ and therefore is used to silence neurons with light (optogenetics). Replacement of the conserved cysteine residue in the transmembrane helix 3 (Cys110) with alanine or threonine results in a >1,000-fold decrease in the channel closing rate. The phenotype of the corresponding mutants in channelrhodopsin 2 is attributed to breaking of a specific interhelical hydrogen bond (the “DC gate”). Unlike CrChR2 and other ChRs with long distance “DC gates”, the HcKCR1 structure does not reveal any hydrogen bonding partners to Cys110, indicating that the mutant phenotype is likely caused by disruption of direct interaction between this residue and the chromophore. In HcKCR1_C110A, fast photochemical conversions corresponding to channel gating were followed by dramatically slower absorption changes. Full recovery of the unphotolyzed state in HcKCR1_C110A was extremely slow with two time constants 5.2 and 70 min. Analysis of the light-minus-dark difference spectra during these slow processes revealed accumulation of at least four spectrally distinct blue light-absorbing photocycle intermediates, L, M₁ and M₂, and a UV light-absorbing form, typical of bacteriorhodopsin-like channelrhodopsins from cryptophytes. Our results contribute to better understanding of the mechanistic links between the chromophore photochemistry and channel conductance, and provide the basis for using HcKCR1_C110A as an optogenetic tool.

Introduction

Channelrhodopsins (ChRs) are the only known proteins capable of passive ion transport upon illumination [1]. As such, they are widely used as optogenetic tools to control neuronal excitability [2]. Several ChR families are currently known, different by both structure and function. Upon expression of their genes in neurons, cation channelrhodopsins (CCRs) generate H⁺ and Na⁺ influx and activate neuronal firing [3]. Anion channelrhodopsins (ACRs) conduct Cl⁻ and are used as optogenetic silencers of neurons [4]. Recently discovered kalium channelrhodopsins (KCRs) show higher permeability to K⁺ than to Na⁺ and only negligible permeability to H⁺ [5]. They lack the “K⁺ channel signature sequence” universally found in “classical” voltage or ligand-gated K⁺ channels [6, 7], which reveals an alternative mechanism of K⁺ selectivity.

The first two KCRs (HcKCR1 and HcKCR2) were discovered in the stramenopile fungus-like protist Hyphochytrium catenoides, and later their homologs were found in other stramenopile [8] and alveolate [9] microbes. When expressed in mammalian cells and activated by illumination, KCRs generate K⁺ efflux and hyperpolarize the membrane, as do native mammalian voltage-gated K⁺ channels during the repolarization phase of the action potential. Therefore, KCRs have been used as optogenetic tools to inhibit spiking in mouse cortical [5] and hippocampal [8, 10] neurons, and in human induced pluripotent stem cell-derived atrial cardiomyocytes [8]. Unlike ACRs, KCRs do not depend on the Cl⁻ gradient and therefore are not expected to cause neurotransmitter release at the axonal terminals that accumulate more Cl⁻ than the soma [11].

KCRs are composed of seven transmembrane helices (TM1-TM7), as are all microbial rhodopsins [12–14] except enzymerhodopsins that have an additional N-terminal helix. Out of all ChRs, KCRs show the closest protein sequence homology to bacteriorhodopsin-like cation channelrhodopsins (BCCRs) from cryptophytes, although none of the latter exhibit K⁺ selectivity [15, 16]. Our recently obtained high-resolution structures of HcKCR1 and its close Na⁺ selective homolog [17] revealed that these proteins form trimers like BCCRs [18, 19] and haloarchaeal ion-pumping rhodopsins [20] rather than dimers like chlorophyte CCRs [21] and cryptophyte ACRs [22]. As in all microbial rhodopsins, in KCRs the retinal chromophore forms a Schiff base (SB) linkage with a conserved Lys residue in the middle of TM7. In most microbial rhodopsins, the SB is protonated in the unphotolyzed state. Photoexcitation initiates a cycle of photochemical conversions (a photochemical reaction cycle, or “photocycle”) that returns the molecule to the initial state. Spectrally distinct intermediates of the photocycle are traditionally designated by the letters K, L, M, N and O [23]. Formation of the red-shifted K intermediate manifests photoisomerization of all-trans to 13-cis retinal. Deprotonation and subsequent reprotonation of the SB are observed as formation and dissipation, respectively, of the blue-shifted M intermediate. In ion-pumping rhodopsins the photocycle is tightly coupled to active ion translocation across the membrane [24, 25]. In KCRs, as in other ChRs, photon absorption leads to formation of a continuous aqueous pore, in which the ions move passively along the electrochemical gradient. Temporal correlation between photochemical conversions and channel gating is different in different ChRs, and the question of their mechanistic link remains open. As those of other known ChRs, KCR sequences form a large cytoplasmic C-terminal domain in addition to the 7TM domain. The cytoplasmic C-terminal domain is not required for channel function.

Optogenetic tools with extended lifetime of the channel open state allow using lower light intensities and durations and thus prevent overheating of the tissue. Although natural ChRs vary in this respect, stronger effects can be achieved by mutations. In particular, replacement of a highly conserved cysteine residue in TM3 with serine, threonine or alanine yields variants with the long-lived open state in Chlamydomonas reinhardtii channelrhodopsin 2 (CrChR2) [26]. HcKCRs show little sequence homology to chlorophyte CCRs such as CrChR2 [5, 8, 9] and lack a specific interhelical hydrogen bond (H-bond) known as the “DC gate”, mutagenetic disruption of which stabilizes the open state of the channel [21, 27]. Here we show that the C110A and C110T mutations nevertheless cause pronounced changes in channel gating kinetics of HcKCR1.

Results

Closing of the channel.

To analyze the channel kinetics in the HcKCR1 mutants under single-turnover conditions, we expressed the corresponding genetic constructs in HEK293 (human embryonic kidney) cells and recorded photocurrents elicited by 6-ns laser flashes. Using short laser flashes minimized photoexcitation of the thermal intermediates of the photocycle, as occurs under illumination with longer pulses of continuous light.

Mutation of HcKCR1 at the position C110 to Ala or Thr did not suppress photocurrents, which remained in the same nA range as in the wild type, but caused a dramatic (>3 orders of magnitude) slowing of channel closing (Figure 1a). In contrast to the wild type, in which only one exponential component with the time constant (τ) 38 ± 4.6 ms (mean ± sem; n = 4 cells) at 0 mV could be resolved in the photocurrent decay [5], in the mutants closing was biphasic. In HcKCR1_C110A, τ of the fast component was 6.7 ± 1.3 s, and constituted 23 ± 1.6% of the full closing amplitude; τ of the slow component was 75 ± 10 s (mean ± sem; n = 5 cells for each mutant). The kinetic parameters of the HcKCR1_C110T mutant were similar: 12.2 ± 1.8 s, 27 ± 5% and 55.5 ± 0.5 s, respectively (Figure 1b). Closing of the channel could also be followed by a parallel increase in the membrane resistance within ~1 min (Supplemental Figure 1). Both components of the channel closing accelerated upon shifting the holding potential to more positive values (Supplemental Figure 2).

Figure 1.

Analysis of channel closing in the slow HcKCR1 mutants compared to the wild-type. (a) Photocurrent traces recorded from C110A (red), C110T (blue) and the wild type (black). The thin solid lines are experimental data, the thick dashed lines are multiexponential computer approximations. (b-d) Ensemble data for the fast closing τ (b), slow closing τ (c), and fast closing relative amplitude (d). The data are the mean values ± sem (n = 5 cells).

In wild-type HcKCR1, ion selectivity did not change during the open state produced by laser flash excitation [5]. In contrast, in the slow mutants the reversal potential of the slow closing was 37 ± 8 mV (mean ± sem; n = 4 cells) more positive than that of the fast closing, indicating a decrease in the K⁺ selectivity before final closing (Figure 2). The shift under 1-s stimulation appeared to be even larger. We cannot exclude that even upon laser excitation the slow decay of photocurrent might reflect a contribution of the product of two-photon absorption, considering the existence of an extremely long-lived photocycle intermediate that did not dissipate during a 5-min interval between laser flashes (see below). In any case, it means that three conformational states can be observed in both slow mutants: the closed state, a high K⁺-selective open state, and a lower K⁺-selective open state.

Figure 2.

The voltage dependence of channel closing in HcKCR1_C110A upon photoexcitation with 6-ns laser flashes and 1-s pulses of continuous light. (a, b) Photocurrent traces recorded at the indicated voltages upon laser flash (a) and 1-s pulse (b) excitation. In (b), the thin solid lines are experimental data, the thick dashed lines are multiexponential computer approximations. (c, d) The current-voltage relationships of the fast (black) and slow (red) closing components upon laser flash (c) and 1-s pulse (d) excitation.

Empirical calculations [28] yield the value 13.83 for the pK_a of Cys110 and predict its Coulombic interactions with the SB, both aspartates in the photoactive site (Asp105 and Asp229), Cys77 in TM2, and Cys232 in TM7. The Asp105 and Asp229 mutants have been characterized by us and others previously [8, 29], so here we created and tested the C77A and C232A mutants. Neither of these mutations reproduced the C110A/T phenotype (Supplemental Figure 3), which confirmed our hypothesis that the slowing of channel closure in these cases is caused by mutagenetic disruption of the direct interaction of Cys110 with retinal (see Discussion).

Opening of the channel.

As described above, the C110A and C110T mutations eliminated the fast channel closing observed in the wild type HcKCR1. This led to the appearance of the third slow phase of channel opening, in addition to the two phases resolved in the wild type (Figure 3). The main fast opening phase with τ ~0.3 ms that comprised ≥ 85% of the amplitude in the wild type did not change significantly in the mutants, as compared to the wild type. The τ values of the two minor opening components in HcKCR1_C110A and C110T were in the range of 2-12 and ~150 ms.

Figure 3.

Photocurrent rise in the C110A (a) and C110T (b) mutants (red) compared with wild-type KCR1 (black). The thin solid lines are experimental data, the thick dashed lines are multiexponential computer approximations.

Absorption spectroscopy and the photocycle.

To analyze photochemical conversions in the HcKCR1_C110A mutant, we expressed and purified it from the methylotrophic yeast Pichia pastoris. The absorption maximum of fully dark-adapted purified HcKCR1_C110A was at 509 nm, 7 nm blue-shifted compared to that of the wild type (Figure 4a). As absorption changes in this mutant proceeded in two very different time windows, we used two different experimental setups to follow them. Fast absorption changes with τ up to several tens of seconds were measured using a flash photolysis setup. Slow absorption changes that needed recording for more than an hour were followed by successive scans of the sample on a Cary spectrophotometer.

Figure 4.

Absorption spectroscopy and photochemical conversions of purified HcKCR1_C110A. (a) Absorption spectra of wild-type HcKCR1 (black) and C110A mutant (red). (b) Transient absorption changes recorded at the maximal absorption wavelength of the unphotolyzed state during the fast photocycle. (c, d) Comparison of the M-like intermediate kinetics during the fast photocycle (black) with that of photocurrent (red) in the time domain of the current rise (c) and decay (d). The thin solid lines in b-d are experimental data, and thick dashed lines are multiexponential computer approximations.

Since the largest absorption changes were observed in the range of an M-like intermediate (400 nm), formation of which manifests deprotonation of the SB, we present flash photolysis results only for 400 nm (M) and 510 nm (unphotolyzed state). Recovery at the wavelength of the maximal absorption of the unphotolyzed state was biphasic. The minor component of the fast recovery with τ ~5.5 s (Figure 4b) closely correlated with that of the fast channel closing (Figure 1). The major component of the recovery proceeded with τ ~22 s (Figure 4b), which was smaller than that of the slow channel closing, but in the same time window. This apparent difference can be explained by the fact that a small fraction of the protein recovers during the slow photocycle (see below), which could not be taken into account by our fitting procedure. The overall conclusion is that the state of purified pigment is very similar to that in the membrane. The M-like intermediate dissipated in parallel with the fast recovery of the unphotolyzed state with nearly identical kinetic components (τ 2.5 and 22 s), indicating that it is present during the entire open state lifetime (Figure 4c). Three phases could be observed in M rise, as in the channel opening. However, there was no quantitative correlation between the amount of the deprotonated SB and the conductance of the channel. Major opening proceeded slightly faster than the small initial rise in absorption at 400 nm. It is quite possible that this small rise in absorption reflects accumulation of a late L-like intermediate. Two slow phases of M accumulation are 4- and 8-fold slower than slow components of channel opening. The peak of channel opening was reached 6-fold later than the peak of M accumulation.

Slow absorption changes.

Illumination of purified HcKCR1_C110A with a long pulse of continuous light or a series of laser flashes caused bleaching of the main band in the visual range (Figure 5a). The maximal depletion of the unphotolyzed state took place at 100 ms after photoexcitation, and its recovery during the slow photocycle proceeded in two stages with τ ~5.2 and 70 min (Figure 5b). Theoretically, a train of laser flashes with a 100-ms interval should have produced the highest fraction of the open state. Each successive flash would convert the remaining unphotolyzed protein, thus reaching the maximal conversion. The saturation rate depends on the quantum yield of primary photoconversion. In bacteriorhodopsin it is 0.64 ([23]), whereas in the chimeric channelrhodopsin C1C2 it is ~0.3 [30]. Our data show that the quantum yield in HcKCR1_C110A is 0.53 ± 0.5 (mean ± sem; n = 3 cells), i.e., closer to bacteriorhodopsin (Supplemental Figure 4), consistent with the protein sequence homology between KCRs and the latter protein.

Figure 5.

Long-lasting absorption changes in the C110A mutant. (a) Absorption spectra of the dark adapted (black) and light-activated (red) samples. (b) Kinetics of slow absorption changes at the wavelength of maximal absorption after a train of laser flashes. (c) The difference (light minus dark) absorption changes after the laser (red) or continuous light pulse (black) excitation. (d) Successive difference absorption changes upon transition in the minute range.

Depletion of the unphotolyzed state was accompanied by a shift of the absorption peak to shorter wavelength (Figure 5a, red). Regardless of whether a pulse of continuous light or a series of laser flashes was used, the first spectral transition observed within 10-30 s after photoexcitation was an increase in absorption at ~400 nm (M₁ intermediate) that accompanied depletion of the unphotolyzed state (Figure 5c). However, the shape of the difference (light minus dark) curve indicated that a long-lived L-like intermediate and probably the M₂ intermediate (370 nm) also at least partially contribute to the blue-shifted spectrum. This hypothesis was confirmed by measuring sequential difference spectra during the dark recovery (Figure 5d). A strong decrease in the absorption at ~450 nm characteristic of an L-type intermediate was observed within 10 min in the darkness. This decrease was accompanied by the appearance of absorbance at ~337 nm (a UV form). During the following 30 min the UV form converted to a product absorbing at 370 nm (the M₂ intermediate).

The extremely slow dark recovery of the unphotolyzed state in the HcKCR1_C110A mutant did not allow us to accumulate enough data for global fit analysis. Nevertheless, our results have revealed at least 5 spectral intermediates in its photocycle: the unphotolyzed state, the UV-form, L, M1 and M2.

Discussion

The slow-cycling mutants obtained by replacement of the conserved Cys in TM3 with other residues have been best studied in CrChR2. In CrChR2_C128T, channel opening in response to a 10-ns laser flash was fit with a single exponential with τ 5.5 ms, whereas in the wild type it was 0.2 ms [31]. From the current trace plotted on a linear scale it is not clear whether the channel opening in this mutant was indeed monophasic, or reflected averaging of several kinetic components, as no attempts at their deconvolution have been made. Therefore, it is difficult to compare the effects of the corresponding mutations in HcKCR1 and CrChR2. In the wild-type HcKCR1 [5] and CrChR2 [32] the kinetics of channel opening is very similar (τ in the range of several hundred μs), but that of accumulation of the M-like intermediate is dramatically different (4.7 ms and 25 μs). In both proteins, the τ values of M rise were not affected by the mutations, remaining ~200-fold different (this study and [33]).

In CrChR2, Cys128 forms a water-mediated interhelical H-bond with an aspartic acid residue in TM4 (Asp156) known as the “DC gate” [21, 27]. Asp156 serves as the proton donor to the SB in the CrChR2 photocycle [34]. Mutagenetic disruption of the DC gate retards reprotonation of the SB and thus slows channel closing [33]. A high-resolution HcKCR1 structure that we have obtained recently [17] shows that the residue structurally corresponding to Asp156 is the non-polar Val133 that cannot form side chain H-bonds (Supplemental Figure 5). Therefore, the slow closing kinetics in the HcKCR1_C110A mutant requires a different explanation. The Cys110 side chain is located within 5 Å from the retinal chromophore, so the effect of its mutations on the channel kinetics might be directly related to the interactions between the chromophore and the apoprotein. In contrast to CrChR2, in which substitution of Cys128 with non-polar Ala caused a more dramatic decrease in the photocurrent decay rate than with polar Thr [26, 33], in HcKCR1 the effects of these two substitutions were comparable, which suggests that their nature is steric rather than electrostatic. In the HcKCR1 structure that we obtained [29], Cys110 is predicted to have long-distance (9 and 7.8 Å, respectively) Coulombic interactions with Cys77 in TM2 and Cys232 in TM7. We show that photocurrent kinetics is not much affected by the C77A and C232A mutations that are expected to disrupt these interactions. These observations corroborate our conclusion that the phenotype of the C110A/T mutations is determined by disruption of the interaction of Cys110 with the retinal chromophore. Models of the C110A and C110T mutants are shown in Supplemental Figure 6, but their high-resolution structures by X-ray crystallography or cryo-electron microscopy are still highly desirable.

One of the main observations in this paper is that absorption changes in purified HcKCR1_C110A proceeded on two dramatically different time scales. Slow components in a single photocycle have been observed in other microbial rhodopsins, but to the best of our knowledge, a 200-fold difference between the τ values of the fast and slow recovery of the unphotolyzed state has never been reported. An additional argument in favor of the existence of the two photocycles is that in the slow photocycle a considerable amount of the M-like intermediate persists for at least 10 min, whereas in the fast photocycle it disappears within 20 s. A possible explanation for these observations in HcKCR1_C110A is a branching photoconversion as suggested for the slow mutants of CrChR2 [31]. Alternatively, a two-photon process can be involved in generation of the extremely slow photocycle, as we have reported for RubyACRs [35].

In addition to the three blue-shifted intermediates absorbing in the visible range, analysis of the difference light minus dark absorption spectra of purified HcKCR1_C110A revealed formation of a UV-absorbing state. Earlier we reported a UV-absorbing photointermediate with a characteristically structured spectrum in the photocycles of BCCRs from the cryptophyte algae of the genus Rhodomonas from which also the BCCR known as ChRmine was derived [16]. In Rhodomonas BCCRs, formation of this long-lived intermediate is the main reason for their unusually fast and complete desensitization (i.e., reduction of photocurrent during illumination). A possible role of this intermediate in desensitization of KCRs needs further investigation.

In proton-pumping channelrhodopsins, de- and reprotonation of the SB are parts of the vectorial process of proton translocation across the membrane [36, 37]. In HcKCR1_C110A, both deprotonation of the SB and opening of the channel exhibit three exponential components, but there is no correlation between their respective τ values. As our flash-photolysis studies have been carried out in detergent-purified protein, we cannot exclude a possible influence of detergent on the photocycle kinetics. However, in anion channelrhodopsin GtACR1 deprotonation of the SB both in detergent and membranes was ~50-times slower than the photocurrent rise [38], which indicates that at least in this protein deprotonation of the SB is not required for channel opening.

We show that the HcKCR1_C110A channel exhibits at least three functional states: closed, open with high K⁺ selectivity, and open with decreased K⁺ selectivity. In combination with very slow channel closing, this result is important for elucidation of structural determinants of K⁺ selectivity in KCRs that differ from well-known “selectivity filters” in voltage- and ligand-gated K⁺ channels. Based on our understanding of photoelectric and absorption characteristics of HcKCR1_C110A, protocols for the enrichment of specific channel states for structural studies can be designed. Also the long lifetime of the channel open state in HcKCR1_C110A and HcKCR1_C110T may be beneficial for using these mutants as optogenetic tools for prolonged inhibition of neuronal activity with only short light pulses.

Materials and Methods

Molecular biology, HEK293 transfection and patch clamp recording

The wild-type synthetic mammalian-codon optimized DNA construct encoding amino acid residues 1-265 of the transmembrane domain of HcKCR1 (GenBank accession no. MZ826861) in frame with an mCherry fluorescent tag was cloned into the pcDNA3.1 vector backbone (Addgene plasmid no. 177336). Mutations were introduced using a QuikChange XL site-directed mutagenesis kit (Agilent Technologies) and verified by DNA sequencing. The resultant plasmids were used to transfect HEK293 (human embryonic kidney) cells obtained from the American Type Culture Collection (ATCC; catalog #CRL-1573) and tested for mycoplasma contamination by PCR analysis. All-trans-retinal (Sigma) was added immediately after transfection (3 µM final concentration). Photocurrents were recorded 48-96 h after transfection in the whole-cell voltage clamp mode with an Axopatch 200B amplifier and digitized with a Digidata 1440A using pClamp 10 software (all from Molecular Devices). The pipette solution contained (in mM): KCl 130, MgCl₂ 2, HEPES 10, pH 7.4, and the bath solution contained (in mM): NaCl 130, CaCl₂ 2, MgCl₂ 2, glucose 10, HEPES 10, pH 7.4. Laser excitation was provided by a Minilite Nd:YAG laser (532 nm, pulsewidth 6 ns, energy 5 mJ; Continuum). Continuous light pulses were provided by a Polychrome V 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). The shutter was controlled by pClamp 10 software to allow illumination of a cell for 1 s. Continuous illumination for 1 s is referred to as “1-s stimulation” or “1-s pulse” in the description of results. The current traces were logarithmically filtered using Logpro software freely available from Zenodo [39]. Curve fitting and descriptive statistics were performed using Origin Pro 2019 software (OriginLab Corporation).

Expression and purification of HcKCR1 from Pichia pastoris

The HcKCR1 expression construct was fused in frame with a C-terminal 8-His tag and subcloned into the pPICZalphaA vector (Invitrogen). The resultant plasmid was linearized with SacI and used to transform P. pastoris strain SMD1168 (his4, pep4) by electroporation according to the manufacturer’s instructions. Resistant transformants were selected on 0.5 mg/ml Zeocin. Transgene expression was induced by the addition of 0.5% methanol in the presence of 5 µM all-trans retinal. After 24-30 h growth, the cells were harvested and disrupted in a bead beater (BioSpec Products). Membrane fragments were collected by ultracentrifugation and solubilized by incubation with 1.5% dodecyl maltoside (DDM) for 1.5 h at 4°C. The supernatant was mixed with nickel-nitrilotriacetic acid agarose beads (Thermofisher) and loaded on a column. The protein was eluted with buffer containing 300 mM imidazole, which was removed by repetitive washing using YM-10 centrifugal filters (Amicon).

Absorption spectroscopy and flash photolysis

Absorption spectra of purified HcKCR1 were recorded using a Cary 4000 spectrophotometer (Varian). Light-induced absorption changes were measured with a laboratory-constructed crossbeam apparatus. Excitation flashes were provided by a Minilite II Nd:YAG laser (532 nm, pulsewidth 6 ns, energy 5 mJ; Continuum). Measuring light was from a 250-W incandescent tungsten lamp combined with a McPherson monochromator (model 272, Acton). Absorption changes were detected with a Hamamatsu Photonics photomultiplier tube (model R928) combined with a second monochromator of the same type. Signals were amplified by a low noise current amplifier (model SR445A; Stanford Research Systems) and digitized with a GaGe Octopus digitizer board (model CS8327, DynamicSignals LLC), maximal sampling rate 50 MHz. Logarithmic filtration of the data was performed using the GageCon program [40].

Computational biology

Empirical pK_a calculations were carried out by PROPKA [28] implemented at the PDB2PQR server [41]. In silico mutagenesis was carried out using our wild-type model (PDB accession: 8GI8) and Coot software [42]. The mutated residue was autofit according to the wild-type electron density map. The models of the mutants were further refined using the real-space refinement function of Phenix [43] according to the wild-type electron density map.

Abbreviations:

ACRs

anion channelrhodopsins

BCCRs

bacteriorhodopsin-like channelrhodopsins

CCRs

cation channelrhodopsins

ChRs

channelrhodopsins

KCRs

kalium channelrhodopsins

SB

Schiff base

Data availability

Data will be made available on request.