Report

Engineering of a light-gated potassium channel

See allHide authors and affiliations

Science  08 May 2015:
Vol. 348, Issue 6235, pp. 707-710
DOI: 10.1126/science.aaa2787

An optogenetic tool to silence neurons

Potassium channels in the cell membrane open and close in response to molecular signals to alter the local membrane potential. Cosentino et al. linked a light-responsive module to the pore of a potassium channel to build a genetically encoded channel called BLINK1 that is closed in the dark and opens in response to low doses of blue light. Zebrafish embryos expressing BLINK1 in their neurons changed their behavior in response to blue light.

Science, this issue p. 707

Abstract

The present palette of opsin-based optogenetic tools lacks a light-gated potassium (K+) channel desirable for silencing of excitable cells. Here, we describe the construction of a blue-light–induced K+ channel 1 (BLINK1) engineered by fusing the plant LOV2-Jα photosensory module to the small viral K+ channel Kcv. BLINK1 exhibits biophysical features of Kcv, including K+ selectivity and high single-channel conductance but reversibly photoactivates in blue light. Opening of BLINK1 channels hyperpolarizes the cell to the K+ equilibrium potential. Ectopic expression of BLINK1 reversibly inhibits the escape response in light-exposed zebrafish larvae. BLINK1 therefore provides a single-component optogenetic tool that can establish prolonged, physiological hyperpolarization of cells at low light intensities.

Potassium ion (K+) channels have a modular structure with sensor domains connected to a central ion-conducting pore (1). The pore integrates signals coming from the sensors and translates them into opening or closing the channel (2). This allows K+ channels to alter the membrane potential of cells in response to a variety of physiological stimuli. Extending the range of signal inputs recognized by K+ channels can be achieved by grafting exogenous sensor domains onto the pore module (3, 4). With this modular interplay between sensor and pore, it is possible to engineer synthetic channels that respond to any signal by ex novo coupling of sensors to pores. This strategy provides new tools for the investigation and manipulation of biological functions (5). An attractive synthetic channel in this context is a light-gated K+ channel, which is important because of the ability of K+ to terminate excitatory currents within cells. This device would allow remote manipulation of the membrane potential with high temporal and spatial resolution and would represent an efficient control mechanism for many cellular processes, including neuronal firing and hormone release.

Several attempts have been made to create synthetic light-gated K+ channels (69); however, these systems suffer from several shortcomings in that they require the addition of cofactors (6, 7), are irreversible (8), or rely on multiple components (9). To overcome these obstacles, we engineered a single-component light-gated K+ channel by fusing the LOV2-Jα photosensory region of a plant blue-light receptor (10) to the miniature K+ channel pore Kcv (11). Rational design and directed evolution were employed to ultimately generate a blue-light–inducible K+ channel that functions reversibly to drive cell membrane potentials to K+ equilibrium in the absence of exogenous cofactors. The LOV2-Jα photoswitch from Avena sativa phototropin 1 (hereafter LOV) can be used to control protein activity by light-induced conformational changes (12). We therefore adopted this strategy to place Kcv under light control. LOV was fused to various regions of Kcv known to be mechanically important for channel gating (fig. S1A and constructs 3 to 12 in table S1).

A functional complementation approach based on the growth rescue of Δtrk1 Δtrk2 potassium transport–deficient yeast (strain SGY1528) (13) was adapted to screen for light-gated channel activity after replica plating (fig. S1B). One Kcv variant with LOV fused at the N terminus (LK) showed light-induced growth on selective agar (4 mM K+) and liquid culture (fig. S1, B and C). LK was expressed in Xenopus oocytes and tested by a two-electrode voltage clamp. LK currents showed modest but reproducible increases in conductance after transfer from darkness to blue light (455 nm, 80 μW/mm2) (fig. S2A). However, photostimulation of LK currents required tens of minutes to develop and appeared to be irreversible. In an attempt to enhance coupling of LOV to Kcv, the soluble photosensory region of LK was tethered to the plasma membrane. Introducing a putative myristoylation/palmitoylation sequence (MGCTVSAE) (14) at the N terminus of LK resulted in improved, but unexpected, properties. The new variant myLK (Fig. 1A and fig. S2B) showed an enhanced response to light compared with LK, but, in this case, light was found to inhibit rather than activate the channel conductance (fig. S2B). Moreover, the effect of light was reversible and it did not decrease after repetitive exposures. The light sensitivity of the channel was wavelength-specific, elicited by blue but not by red light (fig. S2B). The dynamic range of the light effect (DR), i.e., the ratio between light and dark current, was approximately 1.3. To improve the performance of myLK, three point mutations (G528A, I532A, and N538A) known to augment LOV-effector protein interactions (15) were introduced into the construct, singly or in combination (constructs 15 to 21, table S1). Several of these variants exhibited robust differential growth over a range of selective conditions (fig. S3). Notably, two mutants, myLKI532A and myLKN538A, had increased DR values, 1.5 and 1.9, respectively (Fig. 1B), confirming that myLK architecture enables rational protein design.

Fig. 1 Engineering and functional characterization of light-gated myLK channels.

(A) Cartoon representation of myLK engineered by fusing LOV to Kcv with an additional N terminal myristoylation/palmitoylation sequence; myLK is shown as a monomer in the membrane. Kcv comprises slide helix (SH), pore-helix (PH), turret (T), and transmembrane domains (TM1 and TM2) (23). The LOV domain includes LOV2 and Jα (10). Jagged lines indicate lipid anchoring to the membrane. (B) Currents recorded at –60 mV in 100 mM [K+]out, from oocytes expressing myLK mutants: I532A, N538A, and myLK33. Note that myLK I532A and N538A conductances are inhibited by blue light (blue bars, 455 nm), whereas that of myLK33 is activated. (C) Repetitive photoactivation of myLK33 shows the reproducibility of the effect. (D) Red light (red bars, 617 nm) at two light intensities, 13 and 50 μW/mm2, does not activate myLK33 current, whereas blue light does. (E) Current response of myLK33 to increasing light intensities recorded at –80 mV showing threshold (>7 μW/mm2) and saturating (35 μW/mm2) values of light intensity. (F) Dose-response curve obtained from n = 4 oocytes. Line indicates data fitting by a Hill-type equation, yielding a dissociation constant k = 25 μW/mm2 and a Hill coefficient n = 6. (G) Kinetics of myLK33 activation by different intensities of blue light and (H) subsequent inactivation in the dark. An oocyte expressing myLK33 was exposed to a repetitive light/dark regime with increasing light intensities ranging from low (21 μW/mm2) to medium (29 μW/mm2) to saturating (40 μW/mm2) intensities. The current responses to the three light treatments are normalized to common starting values I = 0 for the onset of channel activation and I = –1 for the start of current decay in the dark; numbers on current traces indicate the light intensities in μW/mm2. Fitting of data with single exponential equation (red lines) yields similar τon values for low (21 μW/mm2, τ = 49 s), medium (29 μW/mm2, τ = 52 s), and high (40 μW/mm2, τ = 59 s) light intensity. Currents decay in the dark with the same velocity, irrespective of pretreatment with a high (τ = 148 s), medium (τ = 137 s), or low (τ = 158 s) light. (I) τon (blue symbols) and τoff (black symbols) from a large number of independent experiments plotted as a function of light intensity. (J) Photoadduct formation kinetics for myL (construct 13, table S1) expressed and purified from Escherichia coli. Light-induced absorption changes were recorded at 450 nm (ΔAbs450 nm) in response to blue-light irradiation (455 nm, 90 μW/mm2) and show exponential kinetics (τon = 7.6 s). (K) Dark recovery of Abs450 nm after light excitation shows exponential kinetics (τoff = 38 s). (L) Representative light-induced absorption spectra of myL recorded at 2-s intervals (upper panel) and light-minus-dark difference spectra recorded every 10 s (lower panel).

To screen for additional DR improvements, random mutagenesis was performed using myLKN538A to generate a library of myLK-encoding sequences. Potassium transport–deficient yeasts, transformed with this mutant library, were grown initially on nonselective agar plates and then replica-plated onto selective medium before exposure to darkness or light. Yeast expressing the parental myLKN538A channel did not grow on media with [K+]out below 4 mM. Thus, variants growing below this concentration were selected for further characterization. Thirty-five variants were obtained showing strong differential growth in the light (either activation or inhibition), 13 of which survived a second round of selection (to eliminate false positives) and were sequenced. The recovered mutations mapped throughout the randomized portion of myLK, but a subset clustered around proline 13 of Kcv (fig. S4), a residue known to affect channel gating (16). Notably, one particular variant, myLK33 (construct 33, fig. S4), harboring a mutation in P13 (P13L) and in the myristoylation/palmitoylation sequence (A7T), was activated rather than inhibited by light, with a relatively large DR value (DR = 3) (Fig. 1B). MyLK33 was therefore chosen for extensive functional characterization.

The channel could be repeatedly activated by light when expressed in oocytes without undergoing apparent inactivation (Fig. 1C). Activation was blue-light specific because red light had no perceivable effect on channel activity (Fig. 1D). By exposing the oocyte to increasing light intensities (Fig. 1E), we obtained a sigmoidal current response with a distinct activation threshold (Fig. 1F). The mean dose-response curve was best fitted with a Hill function yielding a value for half-activation, k = 25 μW/mm2, and a Hill coefficient n = 6. This operational light sensitivity of the cell is less by a factor of about 500 than that of cells expressing light-sensitive pumps such as NpHR (17).

To obtain information on the kinetics of channel activation/deactivation in light/dark, cells were irradiated with different intensities of blue light and transferred back into darkness. Activation and deactivation kinetics could be fitted with a single exponential function (Fig. 1, G and H). The time constants did not change over a wide range of light intensities, from suboptimal to saturating (Fig. 1I). Activation (τon, 87 ± 28 s, n = 52) was about twice as fast as deactivation (τoff 168 ± 31 s, n = 43). To further examine the relationship between channel gating and the photocycle of the sensor, we performed spectroscopic measurements on the sensor portion of the protein (myL), which was used to construct the channel. Global photoadduct formation at saturating light intensity (90 μW/mm2) and its recovery in the dark (Fig. 1, J to L) show exponential kinetics with values of 7.6 s for τon and 38 s for τoff. Hence, myL activates, under the same light conditions, about 10 times as fast as the channel, whereas photoadduct decay is about 4.5 times as fast as channel deactivation in the dark. Thus, the primary light sensing by the photosensor seems to be only loosely correlated with channel gating. This agrees with the observation that other constructs harboring the same photoswitch as myLK33 (see, for example, I532A and N538A) responded to light with a different kinetics (Fig. 1B). Together, these findings suggest that light-regulated channel activity is mostly dominated by slow conformational changes that follow LOV2-Jα photo activation..

Experimental evidence that myLK33 retains the pore properties of the parental K+-selective channel Kcv is summarized in fig. S5. Barium, a known Kcv channel blocker, completely inhibited both dark- and light-induced currents of myLK33 (fig. S5A). The current-voltage (I-V) curve (fig. S5B) shows that light activation of myLK33 is voltage-independent. The reversal potentials (Vrev) of myLK33 photocurrent shifted according to the Nernst equation by 51 mV for a 10-fold increase in external K+ concentration, a value similar to that of Kcv (11) (fig. S5C). Moreover, exchanging 100 mM external K+ with Na+ shifted Vrev to the left by 110 mV, indicating a very low Na+ permeability (PNa/PK = 0.015 ± 0.001 (n = 4) (fig. S5D).

Despite these desirable features, in vivo application of myLK33 is hampered by its high dark activity. We have previously shown in other synthetic channels that the linker between the sensor and the effector module influences the coupling of the two (4). We therefore progressively reduced the linker region within myLK in an attempt to improve the control of the photosensor over the channel pore (constructs 23 to 37 in table S1 and fig. S6). Best results were obtained with myLJSK, which lacks the final nine amino acid residues within the Jα helix (fig. S7A). This variant showed stringent differential growth in the yeast complementation assay (fig. S7B) but was poorly expressed in oocytes. Immunolabeling showed that this channel is present on the membrane of human embryonic kidney 293T (HEK293T) cells, albeit at a moderate level, when compared with the parental channel Kcv (fig. S7C). Subsequent electrophysiological characterization in HEK293T cells demonstrated that myLJSK is activated by blue light (Fig. 2, A and B). Light activation is fully reversible in the dark (Fig. 2, A and B) and inhibited by Ba2+ (Fig. 2, C and D). Importantly, myLJSK lacks any channel activity in darkness, as evident from the comparison of the amount of current in the dark and in the presence of Ba2+ (fig. S7D). Channel opening moves the reversal potential of the cell with EK, the K+ equilibrium voltage (Fig. 2E). The light sensitivity of myLJSK occurs in the same range as that of myLK33, being saturated at ~60 μW/mm2 (fig. S7E).

Fig. 2 Properties of BLINK1 in HEK293T cells.

(A) Whole-cell currents recorded in HEK293T cells expressing myLJSK (BLINK1), in dark, blue light (455 nm, 40 μW/mm2) and after returning to the dark. Voltage steps from +60 to –120 mV, tails at –80 mV (selected traces are shown every 40 mV). A measurable current, above the background, was found in about 8 to 10% of the GFP-positive cells. (B) Steady state I/V relation from currents in (A): dark (black squares), after 5 min in blue light (blue circles) and return in dark (triangles). (C) Currents recorded from an HEK293T cell as in (A) but with 5 mM of BaCl2 added in light. (D) Mean current-voltage relation (n = 4) from experiments as in (B): dark (black squares), blue light (blue circles), blue light + 1 mM BaCl2 (inverted triangles). (E) Mean Vrev values (n = 3) of light-activated BLINK1 current plotted as a function of [K+]out; slope = 58 mV/log [K+]. Pipette solution contained 130 mM K+. (F) Repetitive activation/deactivation of BLINK1 current at –70 mV by blue light/dark transitions. (G) Single-channel fluctuations recorded at –70 mV from the same BLINK1-expressing cell in dark (left) and in blue light (right). Dashed line indicates zero current level; dotted lines indicate apparent single-channel opening levels. Calculated single-channel conductance is 70 pS.

Light gating of myLJSK is best appreciated in gap-free recordings. Repetitive exposure to blue light caused rapid activation followed by deactivation in darkness (Fig. 2F). Data recorded from single-channel fluctuations (Fig. 2G) showed a light-induced increase in channel activity and indicated a unitary conductance of about 70 pS, consistent with the high conductance of Kcv and its synthetic variant KvSynth1 (4). The high single-channel conductance and the relatively low macroscopic currents (≤400 pA at –100 mV) are consistent with the low number of channels, which were detectable by immunolocalization (fig. S7C). A low number of channel proteins with a large unitary conductance offers the advantage of an efficient control over the membrane voltage with minimal disturbance of the cell. These properties of myLJSK ultimately fulfilled our criteria for successful light gating of Kcv. Hence, we renamed this variant BLINK1, blue-light–induced K+ channel 1. Expression of BLINK1 in Sf9 insect cells (fig. S8A) produced fluorescence properties that are characteristic for LOV-containing proteins (fig. S8, B and C). Photoadduct formation and decay for BLINK1 was therefore analyzed by fluorescence spectroscopy and showed kinetics similar to those obtained for myLOV in vitro (14.5 s for τon and 51s for τoff) (fig. S8, D and E).

To determine the in vivo applicability of BLINK1 for optogenetics, we examined its ability to regulate the escape response of zebrafish embryos. Two-day old embryos respond to touch with a burst of swimming (18). We reasoned that BLINK1 photoactivation in zebrafish neurons (either somatosensory or motor) and/or in myocytes would prevent or drastically impair this behavior. Embryos injected with either BLINK1 or green fluorescent protein (GFP) RNA showed robust escape motions when kept in darkness [over 90% of embryos in both cases (BLINK1, n = 199; GFP, n = 151)]. By contrast, when embryos were exposed to blue light, 37% of BLINK1-expressing embryos exhibited a reduced escape response (n = 230) compared with just 9% in control larvae (n = 170) (Fig. 3A and movie S1). The blue-light effect was fully and repetitively reverted in darkness (Fig. 3B), as expected for a BLINK1-driven effect. The light-driven effect on embryonic escape motion developed with a half-time of 15 to 20 min and was reverted by dark with a similar kinetics (Fig. 3C).

Fig. 3 Light controls the behavior of zebrafish expressing BLINK1.

(A) Altered escape response in 2-day-old zebrafish, expressing BLINK1 or GFP. The embryos were injected at t0, kept 24 hours in the dark, and then either exposed to blue light (80 μW/mm2) (blue) or kept in the dark (black). The escape response was tested by gentle mechanical stimulation with a pipette tip (see movie S1). Number of embryos (n) is indicated. (B) Reversibility of the effect of blue light on the escape response. Each data point represents the response of a batch of 2-day-old BLINK1-injected embryos, preselected for positive light response (n = 15). The embryos were repetitively exposed to blue light (blue circles) and dark (black circles) treatments (30 to 45 min each). (C) Kinetics of the light effect on the escape response of 2-day-old embryos. At time zero, blue light was turned on and the response to mechanical stimulation was checked every 15 min. After 60 min, the effect reached the maximum and the light was turned off to monitor the deactivation kinetics. Data are from three experiments in which the n of responding embryos was 62 over a total number of 163 (38%). Data were normalized to the maximum number of responding embryos at plateau in each experiment (after 60 min of light).

Several additional observations underscored that BLINK1 was expressed in zebrafish and that altered behavior of the larvae was driven directly by its activation in blue light. First, the presence of the channel is detectable by Western blotting in BLINK1-injected embryos (fig. S9A). Second, the success of eliciting escape behavior was highly dependent on the wavelength of light with red light (617 nm) being ineffective at evoking altered behavior (fig. S9B). Third, in the dark, viability and morphology were similar in embryos expressing BLINK1 or GFP (wild-type embryos at 2dpf BLINK1 = 77%, n = 81; GFP = 69%, n = 59), confirming that BLINK1 is tightly closed in the absence of light. Taken together, these data demonstrate the ability of BLINK1 to modulate behavioral responses in vivo.

In conclusion, we have created a light-activated K+ selective channel by combining a blue-light sensor and a simple K+ channel pore. The resulting BLINK1 channel is fully genetically encoded and does not depend on external factors for its light regulation, as the flavin mononucleotide chromophore is ubiquitously present in cells. BLINK1 is therefore a promising tool in optogenetics and has several desirable properties relative to other light-gated pumps (17, 19, 20) and channels (21, 22) used for inhibiting the functions of excitable cells. First, in contrast to light-gated pumps, which move H+ or Cl ions, it moves K+, a physiological ion, down its electrochemical equilibrium. Thus, BLINK1 does not expose the cell to unphysiological hyperpolarizations or ion gradients. Second, its large unitary conductance guarantees that a small number of channels can efficiently decrease the input resistance of cells and hyperpolarize the membrane. In addition, the channel does not inactivate in light, which allows long-term control of channel activity by light. The apparent sensitivity of BLINK1 to light, combined with its slow kinetics, makes it a powerful tool for long-term inhibition of cells at low light intensities. Our pilot experiments in zebrafish demonstrate that BLINK1 can be successfully used as an in vivo optogenetic tool. Besides an obvious application for inhibiting neuronal activity, this channel will also find applications in the control of cellular processes, which require long-term stabilization of the membrane voltage such as cell cycle regulation or the control of hormone secretion.

Supplementary Materials

www.sciencemag.org/content/348/6235/707/suppl/DC1

Materials and Methods

Figs. S1 to S9

Table S1

Movie S1

References (2431)

References and Notes

  1. Acknowledgments: We thank G. Romani for providing the antibody to Kcv; D. Minor for the Saccharomyces cerevisiae SGY1528 strain, ; U. P. Hansen, I. Schroeder, and A. Bertl for helpful discussion; and M. Ascagni for technical help with confocal microscopy. This work was supported by Fondazione Cariplo grant 2009-3519, PRIN (Programmi di Ricerca di Rilevante Interesse Nazionale) 2010CSJX4F, and MAE (Ministero Affari Esteri) 01467532013-06-27 to A.M.; by the UK Biotechnology and Biological Sciences Research Council (BB/J016047/1 and BB/M002128) to J.M.C.; and by the Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz (LOEWE) initiative Soft Control to G.T.
View Abstract

Navigate This Article