Potassium channel-based optogenetic silencing

Optogenetics enables manipulation of biological processes with light at high spatio-temporal resolution to control the behavior of cells, networks, or even whole animals. In contrast to the performance of excitatory rhodopsins, the effectiveness of inhibitory optogenetic tools is still insufficient. Here we report a two-component optical silencer system comprising photoactivated adenylyl cyclases (PACs) and the small cyclic nucleotide-gated potassium channel SthK. Activation of this ‘PAC-K’ silencer by brief pulses of low-intensity blue light causes robust and reversible silencing of cardiomyocyte excitation and neuronal firing. In vivo expression of PAC-K in mouse and zebrafish neurons is well tolerated, where blue light inhibits neuronal activity and blocks motor responses. In combination with red-light absorbing channelrhodopsins, the distinct action spectra of PACs allow independent bimodal control of neuronal activity. PAC-K represents a reliable optogenetic silencer with intrinsic amplification for sustained potassium-mediated hyperpolarization, conferring high operational light sensitivity to the cells of interest.

R eversible light-based silencing is a key requirement for investigations into the role of genetically defined cell populations, for example to explore effects of individual neurons in controlling network activity and behavior 1 , or to investigate the effects of local conduction block zones in the heart 2 . Current optogenetic inhibitors comprise inward-directed chloride pumps, outward-directed proton or sodium pumps, or chloride-conducting channelrhodopsins [3][4][5][6][7][8][9] . Their application, however, can suffer from paradoxical effects due to unintended changes in ion distributions [10][11][12][13] , and most of them require high expression levels and persistent high-intensity illumination for efficient inhibition of cell activity. The generation of a lightactivated potassium (K + ) channel has the potential to overcome many of these limitations, primarily because K + conductances underlie the resting state of excitable cells, so their activation can generate strong inhibition without major changes in transmembrane ion gradients. However, efforts to generate a widely usable light-gated K + conductance have not been successful so far. Available synthetic light-gated K + channels suffer from poor expression in mammalian cells 14 , the requirement for additional cofactors [15][16][17] , irreversibility of activation 18 , or the large size of the encoding genes 19 , all of which have hampered their widespread application.
In order to establish a light-controlled K + -based hyperpolarization tool, we conceived an optogenetic system consisting of a cyclic nucleotide-gated (CNG) K + channel and a photo-activated nucleotidyl cyclase (PAC). This concept requires a cyclic nucleotide-gated channel with high preference for K + over Na + and H + that shows high functional expression levels in mammalian cells. Classic mammalian CNG channels or hyperpolarizationactivated and cyclic nucleotide-gated (HCN) channels do not show a high K + selectivity over Na + and H +20, 21 . A cGMP-gated K + channel was found in Arbacia punctulata sperm 22 , but this channel bears a pseudotetrameric architecture and is therefore too large for applications relying on viral gene transfer. For the prokaryotic K + channels MLoK1 from Mesorhizobium loti 23 or MmaK from Magnetospirillum magnetotacticum 24 evidence for functional expression in mammalian cells is lacking 25 . In contrast, the small cAMP-gated K + channel SthK from Spirochaeta thermophila is well expressed in mammalian cells, and it carries an exceptionally large single channel current ensuring robust repolarization or hyperpolarization [26][27][28] , rendering it the most promising candidate for a silencer system.
PACs are blue-light-activated cyclases consisting of a blue light receptor using flavin (BLUF) domain coupled via a linker to the cyclase domain. They were identified in various species, providing a range of light sensitivities and enzyme kinetics 29 . The firstdiscovered PACα and PACβ from Euglenoid algae 30 are large proteins of 105 and 90 kDa, respectively, and only poorly soluble. In contrast, the small bPAC (33 kDa) of the soil bacterium Beggiatoa and similar small PACs from other species can be well expressed in a variety of host cells 29,[31][32][33] .
We characterized various combinations of the SthK channel with small PACs in cell lines to obtain robust light-activated K + currents. Among these, the combinations of bPAC or TpPAC with SthK showed most promising results, and were further validated for general applicability in excitable cells, including cardiomyocytes and different types of neurons. The 'PAC-K' silencer efficiently suppresses action potential (AP) generation and contraction in cardiomyocytes, and neuronal firing both in vitro and in vivo. We found that the combination of bPAC and SthK yields a highly light sensitive K + -based hyperpolarizer that requires minimal illumination for long lasting, fully reversible hyperpolarization, thereby providing an exceptionally efficient optogenetic silencer. Importantly, the K + currents induced by PAC-K activation mimic nature's own repolarization mechanism for resting cell polarization, which is energetically favorable and not associated with major alterations in other ion concentration gradients.

Results
Engineering of PAC-K. We tested two single-vector coexpression strategies for PACs and SthK (Fig. 1a): either with the PAC separated from the channel by a self-cleaving 2A peptide at the SthK C-terminus (split-PAC-K), or with the PAC N-terminally fused to SthK (fused-PAC-K). The design of the split system aimed to exclude any impairment of SthK function by the attachment of the PAC, while the fused system should achieve very short cAMP diffusion distances between enzyme and channel. In ND7/23 cells expressing split-bPAC-K with bPAC from Beggiatoa 31 , a 10-ms light pulse evoked intensity-dependent outward currents that lasted for several seconds (Fig. 1b, c;  Supplementary Fig. 1a). At 50% light saturation (EC 50 : 0.86 mW mm −2 ), currents reached a maximum within ∼12 s (Supplementary Fig. 1a), and decayed exponentially with τ off = 33 ± 17 s (median ± standard deviation). To achieve faster off-kinetics we tested less active PACs from Naegleria gruberi (NgPAC1) 34 , Turneriella parva (TpPAC) 35 , Oscillatoria acuminate (OaPAC) 36 , and Ilumatobacter coccineus (IcPAC) (Fig. 1b, c). Among these, TpPAC displayed four times faster off-kinetics and a 10 times reduced light sensitivity (Fig. 1c, e). Comparison of various split-PAC-K versions with the widely used light-driven H + pump Arch3 3,4 showed that only split-bPAC-K provided a higher effective light sensitivity at EC 50 than Arch3, whereas other PACexpressing cells were less sensitive ( Supplementary Fig. 1b). We surmise that this result might be due to a maximal activation of SthK at light intensities not saturating bPAC, because photon capture of flavins is lower compared to retinal, due to a smaller extinction coefficient 37 . However, all split-PAC-K versions transported considerably more charges per absorbed photon than Arch3 (up to a factor of 3 × 10 4 ) due to the long-lasting activation of currents ( Fig. 1c; Supplementary Fig. 1b). We simultaneously monitored the photoactivated K + current and intracellular cAMP levels using a red fluorescent cAMP reporter 38 (Fig. 1d). Current activation slightly trailed the cAMP increase (τ current /τ cAMP = 1.52 ± 1.26, n = 20 from seven cells) while current deactivation preceded the decline in cAMP (τ current /τ cAMP = 0.33 ± 0.22, n = 9 from three cells). Comparing the fused-PAC-K and split-PAC-K strategies for both bPAC and TpPAC revealed similar cellular light sensitivities and current kinetics for both PAC-K systems, with smaller photocurrents for fused-bPAC-K ( Fig. 1e-h), and poor photocurrents and onset kinetics for fused-TpPAC-K compared to split-TpPAC-K (Fig. 1f, g). We also assessed SthK activation by the bPAC-S27A variant with 15-nm red-shifted absorbance and further reduced dark activity 39 , but found no difference compared to activation by bPAC ( Supplementary Fig. 1d-g). In summary, both fused-PAC-K and split-PAC-K systems were identified as promising silencing tools.
Split-bPAC-K silences cardiomyocyte activity. To test the bPAC-K system in excitable cells, we first transduced rabbit primary ventricular cardiomyocytes with adenovirus for fused-bPAC-K or split-bPAC-K. While the fusion construct failed to elicit any light-induced response, split-bPAC-K-expressing myocytes (Fig. 2a) showed robust outward currents following short blue light exposure (10 ms, 460 nm). Amplitude and duration of currents were graded with light intensity (Fig. 2b-d). Single light pulses suppressed electrically evoked APs for extended time periods (283 ± 43 s at 4 mW mm −2 , Fig. 2e), with the duration of the AP-free interval scaling linearly with the K + current duration ( Fig. 2f). AP duration at 90% repolarization (APD 90 ) recovered to baseline values within 10 s after reoccurrence of excitability, and importantly, was not different from APD 90 measured in nontransduced control cells (Fig. 2g). Light-activation of split-bPAC-K also effectively silenced cardiomyocyte contractions (Fig. 2h). A complex illumination protocol (Fig. 2i) revealed a direct dependence of the contraction-free interval on light-pulse duration (Fig. 2j), and highly reproducible contraction-free intervals that are independent of previous inhibitory events (Fig. 2k).
PAC-K characterization in neurons. Next we assessed the expression of both PAC-K variants in cultured hippocampal neurons. The small size of the PAC-K-encoding DNA sequences allowed gene delivery by a single adeno-associated virus (AAV). Expression of either fused-bPAC-K or split-bPAC-K did not affect intrinsic electrophysiological membrane properties (Supplementary Table 1). In cells expressing split-bPAC-K, 100 ms pulses of 385 or 470 nm light with >1 mW mm −2 shifted the minimal current necessary to evoke an AP (rheobase) by more than 600 pA and suppressed spiking by hyperpolarization for up to 1 min (Fig. 3a-c). SthK-mediated changes in membrane voltage were in accordance with the reversal potential of K + (Fig. 3d). The underlying outward currents depended on the intensity of the applied light pulse ( Supplementary Fig. 2a-e), with peak amplitudes of 861 ± 119 pA and maximum duration of 81 ± 9 s. Importantly, split-bPAC-K functioned as an integrator of the activation light. When increasing the light pulse duration from 100 ms to several seconds, we were able to elicit sustained hyperpolarizing currents with light intensities as low as 0.4 or 0.04 mW mm −2 (Fig. 3e). Increasing the ambient temperature from 20 to 35°C shortened the current rise-time and duration, but did not change the amplitude (Fig. 3f). The hyperpolarization was specifically mediated by activation of the SthK channel, as blue light-exposure of cells expressing only bPAC caused small, depolarizing currents of −36.1 ± 7.4 pA ( Supplementary Fig. 2f, g). Split-TpPAC-K also elicited hyperpolarizing currents in neurons, but required 10× higher light intensities for efficient activation compared to split-bPAC-K ( Supplementary Fig. 3a). Split-TpPAC-K-mediated currents were four times faster compared to split-bPAC-K, but showed a more pronounced run-down during repetitive activation ( Supplementary Fig. 3b-e).
Since bPAC is specifically activated at wavelengths <500 nm ( Supplementary Fig. 1g), we tested whether one can optically elicit independent excitation and inhibition in cultured neurons coexpressing split-bPAC-K and the red-shifted channelrhodopsin bReaChES 40 . In whole-cell current-clamp mode we repetitively triggered APs by bReaChES activation with 550 nm light pulses. Single 100 ms pulses of 385 or 470 nm light abolished bReaChESdriven spiking in an intensity-dependent manner (Fig. 4a-c). Importantly, split-bPAC-K was completely insensitive to the green light used for bReaChES activation. In contrast, blue lighttriggered spiking via bReaChES activation, an effect that could be minimized by using UV light at low intensity for bPAC activation (Fig. 4d, e). In less invasive cell-attached recordings we further used independent optical excitation and inhibition by bReaChES and split-bPAC-K to demonstrate that endogenous ATP levels are sufficient for repetitive and stable silencing of bReaChES-driven AP firing ( Supplementary Fig. 4).
To test neuronal silencing in a more complex environment, we injected AAVs encoding either split-bPAC-K ( Fig. 5a-f) or fused-bPAC-K (Fig. 5g, h) into the hippocampus of wild type mice, or a Cre-dependent version of split-bPAC-K into the hippocampus of parvalbumin-Cre transgenic mice (Fig. 5a, c). Whole-cell recordings from both pyramidal and parvalbumin-positive interneurons in cornu ammonis area 1 (CA1) of acute hippocampal slices demonstrated that blue light triggers prolonged hyperpolarizing photocurrents (Fig. 5b, d). Hyperpolarization was accompanied by a strong reduction in the input resistance, demonstrating that opening of SthK channels creates a shunting inhibition that will further reduce the excitatory drive (Fig. 5b, d). The photoactivated SthK current efficiently suppressed ramp-induced neuronal firing in all neuron types tested (Fig. 5e, Supplementary Fig. 5). Of note, repetitive split-bPAC-K activation by 5 ms light pulses every 20 s persistently silenced AP firing for minutes (Fig. 5f).
Two-photon excitation laser-scanning microscopy in combination with fluorescent indicators allows high-resolution fluorescence imaging in intact tissues, and single-cell manipulation when applied for the activation of optogenetic actuators. It has therefore become a routine method for the interrogation of single-cell behavior in situ. In accordance with reported two-photon excitation cross sections of flavoproteins 41 , we assessed whether we could activate fused-bPAC-K at 930 nm in individual CA1 pyramidal neurons in acute hippocampal slices. Neuronal firing was efficiently suppressed by a 50 ms-long two-photon scan on the somata of individual neurons (Fig. 5g, h).
In vivo silencing of neuronal activity by bPAC-K. Finally, we assessed the efficacy of the bPAC-K silencer system in vivo. Silicon probe recordings in anesthetized mice expressing split-bPAC-K in principal cells revealed reliable inhibition of CA1 unit activity by a single light flash (Fig. 6a-d). We also recorded neuronal activity by two-photon in vivo imaging via the GCaMP6s was excited at 1000 nm to avoid co-activation of bPAC 41 . Baseline activity of hippocampal neurons in mice expressing both fused-bPAC-K and GCaMP6s was not different compared to neurons lacking the actuator (0.92 ± 0.05 vs. 1.06 ± 0.1 Hz), highlighting the absence of any identifiable dark activity. Wide-field illumination (488 nm) robustly and selectively reduced Ca 2+ transients in fused-bPAC-K-expressing neurons ( Fig. 6f-h). In order to assess whether the PAC-K silencer is able to alter behavior in a well-characterized model animal, we targeted bPAC-K to zebrafish motor neurons using the UAS/Gal4 system ( Fig. 7a-c), and studied spontaneously occurring muscle contractions (coils) of the embryos as a simple form of motor behavior. Under control conditions, embryos displayed coiling behavior at a frequency of 3-4 coils per 5 s. Blue light (6 µW mm −2 , 1 s) abolished coiling within 5 s after light on. About 20 s after light off, embryos started to move again and displayed transiently increased coiling behavior (ca. six coils per 5 s), compatible with a rebound from the inhibition state (Fig. 7d, g). Illuminating with further reduced intensity (2 µW mm −2 ) for 15 s instead of 1 s did not significantly alter the effectiveness of inhibition (n = 28; p > 0.05; Fig. 7e, h). Non-expressing control fish showed a slight reduction of coiling activity after blue light application, which can be explained by photoinhibition of embryonic motor responses through endogenous, non-visual rhodopsins 43 . Consistently, both non-expressing and transgenic fish showed a similar response to a 625 nm light pulse (Fig. 7f, i). Taken together, bPAC-mediated activation of SthK mediates robust silencing of neurons in the vertebrate central nervous system.

Discussion
Considerable efforts have been dedicated to developing tools for optogenetic inhibition of excitable cells 44 . Such tools are critical for understanding the roles of a particular cell type in the complex functional architecture of excitable tissues, such as the heart and the brain. For their effective use, optogenetic inhibitors must be (i) reliably targetable to the cells of interest, (ii) allow efficient and reversible inhibition of AP firing, ideally for variable durations, and (iii) have no dark activity or paradoxical effects on ion concentrations that may affect relevant cell and organ functions. The currently available inhibitory tools do not satisfy all of these conditions, in particular given that many experimental designs require prolonged inhibition. Different variants of chloride or proton pumps have been generated, with the general caveat that their prolonged activation may change intracellular ion concentrations, a phenomenon observed both for the proton pump Arch3 12 and the chloride pump eNpHR3.0 10 . This can lead to paradoxical effects, for instance through changes in the GABA reversal potential 11 . For both types of pumps, rebound depolarizations after termination of light exposition have been described 12 . Chloride-conducting channelrhodopsins (iChloc, iC++, and ACRs) have the advantage that, in contrast to ion pumps, they can produce inhibition by clamping the membrane potential to the chloride reversal potential [7][8][9] . However, activation of chloride channels suffers not only from the fact that prolonged chloride fluxes alter intracellular Cl − concentrations, but also from the observation that they can trigger depolarizing cation fluxes via voltage-dependent channels in neurons or neuronal compartments with a high intracellular Cl − concentration, based on depolarizing Cl − efflux upon light exposure 12,44,45 . Also the inhibition of myocardial excitability in this setting 46 is based on depolarization of cells above the excitation threshold, which is undesirable for rhythm management as it may lead to a potentially arrhythmogenic gain in cellular Ca 2+ uptake.
Optogenetic activation of K + conductances can potentially overcome these limitations, which has spurred the development of several light-gated K + channel systems [15][16][17][18][19] . The most promising published example is the engineered blue-light-induced consisting of a small viral K + channel connected to a LOV-Jα photoreceptor module 14 . BLINK1 is not, however, robustly expressed in mammalian cells. In contrast, we demonstrate here that the SthK channel is well expressed in different cell types, and that the PAC-K constructs allow highly efficient hyperpolarization via a light-gated K + conductance in cell lines, cardiomyocytes, and neurons. Whereas ion pumps such as halorhodopsin or archaeorhodopsin require continued photon flux due to the 1:1 photon:ion coupling, PAC-K-mediated hyperpolarization can be maintained by repetitive brief light exposure at low frequencies, or continuous illumination at very low intensities. This allows cell re-polarization or hyperpolarization over prolonged periods with a minimal photon budget, avoiding tissue warming and altered blood flow [47][48][49][50] . One should note though that illumination intensity determines the current kinetics ( Fig. 1, Supplementary Figs. 1 and 2), with higher intensity providing faster onset, but also longer duration of the hyperpolarization. Our tool is applicable on time scales from 100 ms up to minutes. However, the tool does not enable fast manipulations in the millisecond range that might be required for rapid closed-loop optogenetic control of network activities at >10 Hz. Shorter effect durations can be achieved by using TpPAC instead of bPAC ( Supplementary Fig. 3), but with similar onset kinetics at saturating light intensities (Fig. 1g). An additional advantage is the efficient activation of the PAC-K system by twophoton illumination. Since activation requires only milliseconds of illumination to trigger K + currents that last for seconds, PAC-K allows silencing of defined sets of individual neurons by sequential optical targeting. The PAC-K construct further provides the opportunity to combine optically controlled inhibition with excitation using red-shifted excitatory opsins, such as C1V1 51 , Chrimson 52 , ReaChR 53 , or variants of these. Dual-color optogenetic excitation/inhibition experiments so far suffered from spectral overlap of the rhodopsin-based actuators, and required calibration of the light intensities for optimal bidirectional control of membrane voltage 54  b Intensity-dependent spike suppression of green light-triggered APs by 385 nm light (n = 9-12, five cultures; no error bars shown for clarity). c The duration of the spike-free interval in two-color excitation/inhibition experiments was independent of the wavelength used for split-bPAC-K activation (385 nm: n = 9-12, five cultures; 470 nm: n = 9, three cultures). d Intensity and wavelength-dependent cross activation of bReaChES by blue light. Dashed lines indicate resting membrane potential before the light pulse. Scale bars: 200 ms, 20 mV. e AP firing evoked by activation of bReaChES with 385 or 470 nm light was minimized when using low light intensities for bPAC activation (385 nm: n = 10-13, five cultures; 470 nm: n = 9, three cultures; Bonferroni's multiple comparison test, * p < 0.05). Error bars represent SEM intensity. Still, the blue light used for PAC-K activation can potentially trigger an AP by activation of the red-shifted channelrhodopsin (Fig. 4e). However, given the extremely high light sensitivity of PAC-K (Fig. 1c) to blue and UV light, combined with the ability to integrate light over time (Fig. 3e), one can minimize activation of any co-expressed channelrhodopsin. Thus, bidirectional optical control of membrane voltage is feasible by expressing the actuator and inhibitor either in genetically distinct, but anatomically overlapping cell populations, or by coexpressing activator and inhibitor within the same cell type, as shown here for the combination of split-bPAC-K and bReaChES. Our new tool is useful in a variety of model systems and across disciplines exploring excitable cell systems, from the neurosciences to cardiovascular research.
An important aspect of a cAMP-based two-component optogenetic silencer is that one has to consider other cAMP-induced effects on intracellular signaling cascades, cellular metabolism, and gene expression. However, it should be noted that the PAC-K system was developed for silencing individual cells or cell populations in intact tissue, whether in neuronal networks or in the heart. These experiments allow one to study the functional circuitry constituted by the remaining non-affected cells, rather than the particular physiology of silenced cells. However, to exclude cAMP-mediated side effects in experiments involving multiple light applications over several days, it seems advisable to use PAC-only expression as an appropriate control for cAMP effects. bPAC-K represents a highly light-sensitive system due to the light-integration of bPAC and the intrinsic amplification by cAMP. Therefore, exposure of the PAC-K system to background light <500 nm should be minimized whenever possible. In cell types with low phosphodiesterase activity, baseline activity of the PAC might be of concern, although we found little dark activity across various experimental models, confirming previous reports 31,55 . If required, the bPAC(S27A) variant 39 with further reduced dark activity could be used, as the mutant enzyme is equally potent as wildtype bPAC in activating SthK (Supplementary Fig. 1). To conclude, the PAC-K silencer represents a novel genetically targetable light-controlled K + channel that is well expressed in a variety of model systems and cell types. It can thus be exploited for a wide range of scientific research questions for which efficient silencing of target cells is required.    (Fig. 1a). In the cleavable configuration (split-PAC-K), the SthK channel is separated from the PAC by a 2A ribosome skip peptide from porcine teschovirus-1 (P2A) 56 . In the fusion protein variant (fused-PAC-K), SthK is linked to the C-terminus of the PAC (TpPAC or bPAC) by a peptide sequence (PRTYETSQVAPAGAP). In all cases, mCherry served as fluorescent marker. The wavelength used to excite mCherry did not overlap with the excitation spectrum of the PACs. Expression cassettes for PAC-SthK-T2A-mCherry and SthK-P2A-PAC-Cherry were subcloned into a pCS2-CMV-based plasmid for characterization in ND7/23 cells. Different PAC variants (Fig. 1b, c,  Supplementary Fig. 1a, b) were compared using the cleavable PAC-K configuration. For expression in cardiomyocytes, SthK-P2A-bPAC(wt)-mCherry or bPAC (S27A)-SthK-T2A-mCherry were cloned into pAdeno-CMV (Clontech) using the Adeno-X Adenoviral System (Clontech). Adenovirus was provided by the Viral Core Facility of the Charité-Universitätsmedizin Berlin, Germany.

Methods
Similarly, we designed two viral expression constructs for neuronal expression: For the split-PAC-K configuration, we inserted SthK-P2A-bPAC(wt)-mCherry or SthK-P2A-TpPAC-mCherry into the CW3SL AAV expression vector 57 containing   kinase, pH 7.3, 300 mOsm. pH was adjusted with N-Methyl-D-glucamine NMG. Light stimulation was provided by a Mightex light source coupled to an Olympus IX-73 microscope. PACs were illuminated for 10 or 100 ms with a 470 nm LED through a LUCPLFLN ×60/0.7 Olympus objective. Illumination field was 0.102 mm 2 . Light intensities ranging from 0.11 to 100 mW mm −2 were applied for light titration curves and the data was fitted using a logistic function. Maximum current was defined as the maximum current amplitude obtained when saturation of the channel was achieved. The slope of each current was calculated from a linear fit made from the signal onset (3× signal noise) to the half peak amplitude. Latency was defined as the time between the beginning of the light stimulus and the signal onset. Data was analyzed using Clampfit 10, Origin 9, and GraphPad Prism 5. Two-group comparison statistics were performed using the Wilcoxon Rank Sum Test. Action spectra for bPAC, TpPAC, and bReaChES were measured on an Axiovert 100 Carl Zeiss microscope. Light was delivered through a W Plan-Apochromat ×40/1.0 DIC objective (Carl Zeiss) with a Polychrome V light source (TILL Photonics). Illumination field was 0.066 mm 2 . The half bandwidth was set to ±7 nm for all data points, and light exposure time was controlled with a shutter system (VS25 and VCM-D1, Vincent Associates). In order to obtain the same photon irradiance for all wavelengths (Photon count: 2.34 × 10 13 photons s −1 ), a motorized neutral density filter (NDF) wheel (Newport) was placed between the Polychrome V and the microscope. When necessary, a filter between the polychrome and the NDF was employed using a motorized 12-position filter wheel (Thorlabs).
After digestion cell suspensions were filtrated using a 1-mm mesh and centrifuged for 2 min at 22×g. Fibroblasts remaining in the supernatant were removed and cardiomyocytes were resuspended in plating medium (5 mM creatine, 2 mM L-carnitine hydrochloride, 5 mM taurine, 1 mM sodium pyruvat, 0.25 U/l insulin, 10 µM cytosine-β-D-arabinofuranoside, 5% FCS, and 0.05 mg/ml gentamycin). Cardiomyocytes were cultured on laminin-coated (100 µg/ml) coverslips. Medium was exchanged at 3-4 h after seeding cells, and adenovirus (type 5) coding for either of the two constructs was added immediately after medium exchange (MOI75 for 48 h). Medium was renewed after 48 h and cells were recorded using the patch-clamp technique.
Patch-clamp recordings and contraction tracking of cardiomyocytes. Wholecell patch-clamp measurements on single isolated cardiomyocytes were performed at room temperature using an inverted DMI 4000B microscope (Leica Microsystems), an Axopatch 200B amplifier, and an Axon Digidata 1550A (Molecular Devices). Activation light was delivered by a 460 nm LED using a 460-480 nm bandpass excitation filter and controlled via custom-built hardware (Essel Research and Development). External bath solution contained (in mM): 140 NaCl, 5.4 KCl, 1 CaCl 2 , 2 MgCl 2 , 10 glucose, 10 HEPES, pH 7.4, 300 mOsm; internal solution (in mM) 50 KCl, 80 K-aspartate, 2 MgCl 2 , 3 Mg-ATP, 10 EGTA, 10 HEPES, pH 7.4, 300 mOsm. In current-clamp mode APs were triggered by current injections of 50% more than the threshold to elicit an AP, using a current ramp from 0 pA to 50% above threshold within 10 ms. Light pulses (460 nm/10 ms) were applied after 15 current injections and recordings were continued until APs reappeared. The AP duration (APD 90 ) was determined using the 90th percentile of the signal above resting potential as described 60 . Depolarizations with peak amplitudes below 0 mV were marked as inhibited. In voltage-clamp mode cardiomyocytes were clamped to −60 mV and light (460 nm/10 ms) was applied after 5 s.
Cardiomyocyte contractions were followed by detecting changes in sarcomere length. To this end, cardiomyocytes were imaged with a MyoCam-S camera (IonOptix) on an inverted microscope (DM IRBE, Leica Microsystems). Sarcomere length was calculated in real time by a Fast Fourier Transform of the power spectrum of the striation pattern (IonWizard, IonOptix) 61 . Cells were field stimulated at 14 V/0.25 Hz with a Myopacer (IonOptix). A blue light pulse (460 nm/100 ms) was applied after 88 s. Data was analyzed with pClamp 11, OriginPro, GraphPad Prism 7 and Matlab.
Electrophysiology on neuronal cultures. Hippocampal neurons prepared from P0 C57/BL6-N mice of either sex were grown on a glial cell feeder layer at a density of 1.25 × 10 4 cells cm −2 in 24-well plates in Neurobasal-A supplemented with 2% B27 and 0.2% penicillin/streptomycin (Invitrogen). AAV particles were added 1-4 days after plating at a final concentration of 4.5 × 10 9 viral particles/ml for split-bPAC-K and 1 × 10 9 particles/ml for bReaChES. Cells were used for recordings after 15-21 days in vitro.
Whole-cell and cell-attached recordings were performed on an Olympus IX73 inverted microscope using a Multiclamp 700B amplifier (Molecular Devices) under the control of Clampex 10 (Molecular Devices). Data was acquired at 10 kHz and filtered at 3 kHz. Extracellular solution contained (in mM): 140 NaCl, 2.4 KCl, 10 HEPES, 10 glucose, 2 CaCl 2 , and 4 MgCl 2 (pH adjusted to 7.3 with NaOH, 300 mOsm). Synaptic currents were blocked by 5 µM NBQX and 4 µM gabazine in most recordings. The intracellular solution contained (in mM): 135 K + gluconate, 17.8 HEPES, 1 EGTA, 4.6 MgCl 2 , 4 Na 2 -ATP, 12 disodium creatine phosphate, and 50 U/ml creatine phosphokinase, pH adjusted to 7.3 with KOH, 300 mOsm. Membrane potential was set to −60 mV in whole-cell voltage-clamp recordings. In current-clamp recordings, membrane potential was manually adjusted when cells depolarized to more than −55 mV. Data was not corrected for a liquid junction potential, unless for the analysis of the K + reversal potential in Fig. 3d (liquid junction potential of 22.6 mV). Recordings were terminated when holding currents were >−200 pA. Pipettes contained extracellular solution for cell-attached recordings. A TTL-controlled LED system (pE4000, CoolLED) was coupled into the back port of the IX73 microscope by a single liquid light guide. Fluorescence light was filtered using single band pass filters of a "Pinkel" quadband filter set (AHF F66-415) and delivered through an Olympus UPLSAPO ×20, 0.75 NA objective. Neutral density filters were used for experiments involving low-intensity light stimulations. Illumination intensities above the objective were measured prior to the experiments using a photodiode power sensor (Thorlabs S170C). bPAC was activated by flashes from the 385 or 470-nm LED, while bReachES-triggered APs were elicited by 5-ms flashes of a GYR LED passed through a 554/24-nm filter.
Experiments involving temperature shifts were performed on an Olympus BX51 upright microscope equipped with a TC01 temperature controller (Multichannel Systems) and a CoolLED pE3 LED system. Excitation light of 470 nm was passed through a 470/40-nm filter (AHF F56-319) and applied via a 60× Olympus LumPlan FI 60 × 0.9 NA water immersion objective. Data was analyzed using AxographX 1.6.
A subset of electrophysiologically characterized neurons filled with 0.2% biocytin were fixed overnight in 4% paraformaldehyde dissolved in phosphatebuffered saline (PBS, pH 7.2). Recorded cells were visualized by 24 h incubation in PBS supplemented with streptavidin-coupled Alexa 488 (Invitrogen). After washing, the slices were mounted on slides and covered with Mowiol 4-88 (Carl Roth). Images were aquired using a Leica SP5 confocal laser scanning system. Two-photon photostimulation was performed using a microscope equipped with a galvometer-based scanning system (Prairie Technologies, Ultima). A ROI was defined covering the neuronal soma and stimulation was carried out by scanning the ROI with an ultrafast, Ti:sapphire pulsed laser (Chameleon Ultra, Coherent) tuned to 930 nm (50 ms scan duration, laser power <20 mW).
Cranial window preparations followed a former description 63 and were performed under isoflurane anesthesia (2% in oxygen) as follows. Eyes were covered with eye-ointment (Bepanthen, Bayer) to prevent drying and body temperature was maintained at 37°C using a regulated heating plate (TCAT-2LV, Physitemp) and a rectal thermal probe. The cranial bone was exposed, cleaned from connective tissue, and covered with optical adhesive dental cement (OptiBond FL, Kerr). A flat custom-made head post ring was fixed with UV curable dental acrylic (Tetric EvoFlow, Ivoclar Vivadent) to allow a close approach of the objective to the skull. A circular craniotomy (⌀ 3 mm) was opened above the right hemisphere hippocampus using a dental drill. Cortical tissue was aspirated using a 27-gauge needle until the fiber tracks of the corpus callosum became visible. A custom-made cone-shaped silicon inset (upper diameter 3 mm, lower diameter 2 mm, RTV 615, Movimentive) covered by a cover glass (⌀ 5 mm, thickness 0.17 mm) was inserted and fixed with dental acrylic. This special window design allowed easy implantation and maintenance, and the geometry was optimal for conserving the numerical aperture of the objective. The imaging session was conducted under anesthesia directly after surgery.
We used a commercially available two-photon microscope (A1 MP, Nikon) equipped with a 25× long working distance, water immersion objective (NA = 1, WD = 4 mm, XLPLN25XSVMP2, Olympus) controlled by the NIS-Elements software (Nikon). GCaMP6s was excited at 1000 nm using a Ti:sapphire laser system (~60 fs laser pulse width; Chameleon Vision-S, Coherent). At 1000 nm, flavoproteins are not excited by two-photon lasers 41 , allowing separate excitation of the GCaMP with no cross-talk to the silencer. Emitted photons were collected using gated GaAsP photomultipliers (H11706-40, Hamamatsu) that electrically protected the optical sensor during light stimulation and quickly (~µs) recovered sensitivity afterwards. Stimulation light was delivered using a 488-nm diode laser (Sapphire 488 LP, Coherent) through the microscope galvanometric scanning unit in parallel to the imaging. Activity was acquired with a resonant scanning system at a frame rate of 15 Hz and duration of 10 min per movie. A 20-ms stimulation light pulse was delivered in the middle of each movie and five movies were recorded for every field of view. For excitation of mCherry we used a fiber laser system at 1070 nm (55 fs laser pulse width, Fidelity-2, Coherent) and recorded a reference image for every field of view.
To remove movement artifacts caused by breathing, recorded movies were registered using the steady scan Matlab toolkit 64 . Individual cell fluorescence traces were identified and Ca 2+ events were deconvolved using a constrained nonnegative matrix factorization-based algorithm 65 . For later analysis, only components found in all five movies and showing a signal above threshold in the reference image were taken into account. We binarized Ca 2+ events by taking the onset of every event independent of event amplitude. Event probability was calculated by summing up the binarized events of all cells in all five movies within time bins of one second. We calculated the activity onset after stimulus for each individual cell by averaging the time to the first identified event after stimulus from all five movies.
In vivo silicon probe recordings. We injected 3× 200-300 nl of AAV9.CamKIIα: SthK-P2A-bPAC-mCherry into the left hemisphere of C57Bl6N mice (P26-P41, both sex) at coordinates (in mm relative to Bregma) AP −2.0, ML −1.6, DV −2.1, −1.7, −1.3. Four to six weeks after the stereotaxic injection, mice were anesthetized first with isoflurane (~2% in oxygen) and then by two i.p. injections (~10-15 min intervals) of 10% w/v urethane (total dose 1.8-2 g/kg) while lowering the isoflurane level under constant checking of breathing pattern and reflexes. Once a sufficient level of anesthesia was reached, mice were inserted into a stereotaxic frame and a craniotomy was performed at the same site as the original stereotaxic surgery. A 32-channel silicon probe (A1×32-poly2/3-10 mm-50-177-OA32LP, Neuronexus) with an integrated light guide (105 µm core, NA 0.22), attached to a reference wire with an Ag/AgCl pellet placed on the skull, was then inserted into the brain to a depth of 1550-2450 µm. Recordings were performed at depths of 1550-1800 µm using an RHD 2132 amplifier chip with an integrated 16-bit analog-to-digital converter and an RHD2000 USB interface board (Intan Technologies, Los Angeles, CA, USA). The light guide was coupled to an MBL-III-473 laser (100 mW, CNI, Changchun, China) producing pulses of 10 or 100-ms duration. Since no differences in response were observed, we pooled trials with both durations. Inter-pulse intervals were at least 3 min. At the end of each experiment, mice were killed with an overdose of urethane. We then performed cardiac perfusion with 0.1 M PBS followed by 4% paraformaldehyde (PFA) before removal of the brain and overnight fixation in PFA. Coronal sections (100 µm) were cut on a vibratome and imaged with a fluorescence microscope (Leica) to allow identification of mCherry-labeled cells, as well as to visualize the track left by the silicon probe. We confirmed that recordings from all mice were acquired in the hippocampal area CA1, and that the recording sites overlapped with the mCherry-labeled injection sites. As the aim of this experiment was to obtain a qualitative impression of the construct's effectiveness in inhibiting neuronal activity in mouse brain in vivo, and because we observed a very clear effect with the first three mice, no further animals were used. A more formal evaluation and quantification is beyond the scope of this study.
Generation of bPAC-SthK-tRFPt mpn157 transgenic zebrafish. Sequence encoding bPAC(wt)-SthK was fused to TagRFPT and placed downstream of the MCS of the upstream activating sequence (UAS) containing the Tol2 transposon system (pTol2-14xUAS, Baier laboratory). The assembly was performed with the In-Fusion ® PCR cloning system (In-Fusion HD Cloning Kit, TaKaRa Clontech, Cat. #639649). The gene encoding the fluorescent protein TagRFPT was amplified from the following plasmid: UAS:lyn-tagRFPT 67 . The construct (25 ng/μl) was injected into eggs with the TL genetic background at the one-cell stage together with zebrafish codon-optimized Tol2 transposase mRNA (50-100 ng/µl) synthesized from pCS-zT2TP 68 .
Confocal imaging of zebrafish embryos. F1 embryos (22-25 hpf; sex not determined) were dechorionated and mounted in 2.5% low melting agarose in Danieaus solution. Fish were imaged with a Zeiss LSM700 confocal microscope with a ×20 water immersion objective.
Inhibition of spontaneous movement in the coiling assay. Zebrafish embryos were placed in a 48-well plate well, inside a dark, temperature-controlled behavior box. After several minutes of adaptation, spontaneous coiling movements were recorded using an IR-sensitive camera (Mikrotron EoSens). Three recordings were carried out for each condition, with 2-min intervals in between. Activation light for bPAC was provided by a blue (455 nm) or red (625 nm) LED (Thorlabs) placed 15 cm above the samples and controlled by a custom-written Python code using a LabJack U3. Light intensity was measured at the location of the sample with an Energy Meter Console (PM100D, Thorlabs).
Image analysis was performed with custom-made python scripts (available upon request) following a published protocol as follows 69 . Single circular regions of interests (ROIs) were manually drawn around the chorion of each embryo. To identify movements of the embryo, every frame was subtracted from the previous frame. For each ROI, single coiling movements were identified from a series of frames in which the pixel values changed between frames above a pre-defined noise threshold. Embryos that did not show any coiling movement during the whole recording were removed from the analysis. The average coiling frequency was then calculated across embryos of a single trial and across the three trials, for each condition.
Data collection and analysis. Statistical analysis was performed in MATLAB, OriginPro, or GraphPad Prism v5-v7. Data are presented as mean ± standard error of the mean (SEM), if not stated otherwise, and were tested for normality and equal variance. Animal numbers were chosen due to electrophysiological standards to distinguish an in the forehand unknown effect size from chance level. Sample sizes were not predetermined by statistical methods but are similar to those reported in comparable previous publications 3,7 . For light titration experiments, different light intensities were applied in a randomized fashion to the recorded cells. Data collection and analysis were not performed blind to the conditions of the experiments.
Code availability. Custom python scripts for image analysis are available upon request.