High frequency neural spiking and auditory signaling by ultrafast red-shifted optogenetics

Optogenetics revolutionizes basic research in neuroscience and cell biology and bears potential for medical applications. We develop mutants leading to a unifying concept for the construction of various channelrhodopsins with fast closing kinetics. Due to different absorption maxima these channelrhodopsins allow fast neural photoactivation over the whole range of the visible spectrum. We focus our functional analysis on the fast-switching, red light-activated Chrimson variants, because red light has lower light scattering and marginal phototoxicity in tissues. We show paradigmatically for neurons of the cerebral cortex and the auditory nerve that the fast Chrimson mutants enable neural stimulation with firing frequencies of several hundred Hz. They drive spiking at high rates and temporal fidelity with low thresholds for stimulus intensity and duration. Optical cochlear implants restore auditory nerve activity in deaf mice. This demonstrates that the mutants facilitate neuroscience research and future medical applications such as hearing restoration. Optogenetic applications would benefit from channelrhodopsins (ChRs) with faster photostimulation, increased tissue transparency and lower phototoxicity. Here, the authors develop fast red-shifted ChR variants and show the abilities for temporal precise spiking of cerebral interneurons and restoring auditory activity in deaf mice.

M icrobial-type rhodopsins, light gated cation channels (Channelrhodopsins, ChRs) and light-driven ion pumps are useful tools for multimodal optogenetic control of electrically excitable cells in culture, tissue and living animals [1][2][3] . Since the first description of the ChRs in 2002 and 2003, a set of different ChRs including red-shifted variants like VChR1, ReaChR and Chrimson have been described [4][5][6][7][8] . For different purposes ChRs were modified with respect to the kinetics, ion selectivity as well as light absorption [9][10][11] . ChR kinetics is a major issue, because the light sensitivity is regulated via the open lifetime of the channel 12 . Channels with a short open lifetime need correspondingly stronger light than channels with a long open lifetime for maximal photostimulation. This is due to the essential invariance of other channel parameters like single channel conductance and quantum efficiency. The mutual dependence between channel kinetics and light sensitivity accounts for the optimization of ChR expression and light delivery for successful experiments in the high frequency range. Although fast channels need stronger light for the activation, high speed is indispensable for many optogenetic applications in neurobiology because many types of neurons operate at high firing rates in the intact animal.
Prominent examples include spiral ganglion neurons (SGNs) of early auditory pathway and fast spiking interneurons in cortical areas, which fire action potentials at up to several hundred Hz 13,14 . However, light stimulation of ChR2-expressing SGNs indicated a strong limitation of the temporal response fidelity 15 . Therefore, fast ChRs are needed and their benefit for use in auditory research has already been indicated using Chronos, a "fast" blue light absorbing ChR, for stimulation of the cochlear nucleus 16 .
Electrical cochlea implants (eCI), to date, enable speech understanding in most of approximately 500,000 otherwise deaf users. However, the bottleneck of eCI is the poor frequency resolution of coding that results from wide current spread from each electrode contact and limits speech understanding in background noise 17 . Optical cochlear implants (oCI)-stimulating optogenetically modified SGNs, promise a fundamental advance of prosthetic sound coding by increasing frequency resolution, because light can be better confined than the electric field of electrodes 15 . For oCI eventually to be translated into the clinic, opsins need to be delivered into the SGNs by postnatal virus application to the ear and should endow SGNs with high light-sensitivity and temporal fidelity of spike generation, while light scattering and blue light induced phototoxicity should be minimized. Due to the aforementioned, adverse effects of optogenetic stimulation using blue light, the already available, fast blue light-activated ChR variants like ChETA (τ off = 4.4 ms 9 ) and Chronos (τ off = 3.6 ms 6 ) might have a limited applicability in animals and future clinical translation.
Here, we report that fast gating can be generally conferred to ChRs by helix 6 (helix F) mutation and demonstrate the utility of fast red-shifted ChRs for driving spiking of fast cerebral interneurons to the limit of their encoding range. Moreover, we established efficient virus-mediated delivery and expression of a fast Chrimson mutant in SGNs of mice, show that single-channel oCIs enable near-physiological spike rates and spike timing in SGNs and restore auditory activity in deaf mice. We demonstrate on several cell types in vitro and in vivo that the unfavorable low light sensitivity for activation is compensated by high expression levels of the fast Chrimson mutants.

Results
Fast helix F mutants and their calcium permeabilities. Closed to open state transition is associated with movement of helix F in several microbial-type rhodopsins [18][19][20] . Thereby helix F movement controls protonation reactions during vectorial proton transport and consequently the cycle time 21,22 . Closed to open state transitions of helix F have recently been verified for ChR 2 23,24 . Motivated by these findings we performed a systematic study about the effects of helix F mutations on the closing kinetics of ChR (Fig. 1). We heterologously expressed ChRs helix F mutants in neuroblastoma-glioma cells (NG cells) and performed whole-cell patch-clamp experiments. The helix F mutant F219Y significantly accelerated the closing kinetics of ChRs 2 ( Fig. 1c and Table 1). Mutations at the homologous positions of VChR1 (F214Y), ReaChR (F259Y) and Chrimson (Y261F) also accelerated the closing kinetics (Fig. 1b), albeit to a different extent (Fig. 1d-f and Table 1). The strongest effect on the lifetime of the channels was observed in ReaChR and VChR1, where the closing kinetics is accelerated by one order of magnitude (Table 1).
Interestingly, the relative calcium permeability of ChR2 F219Y P Ca /P Na = 0.30 ± 0.02 (n = 4) was increased compared to the relative calcium permeability of ChR2 wt P Ca /P Na = 0.13 ± 0.01 (n = 4). Permeability ratios were calculated according to the Goldman-Hodgkin-Katz equation 25 with the measured values of the reversal potentials after replacing external sodium by calcium. The critical role of a tyrosine at the homologous position on the calcium permeability is verified in ReaChR and Chrimson (Supplementary Table 1). Of note F219 (ChR2 numbering) points to L132 (ChR2 numbering) on helix C in the chimera C1C2 crystal structure (Fig. 1a). ChR2 L132C has an increased calcium permeability (CatCh, calcium translocating ChRs) 10 . In contrast to the FY mutations on helix F, which accelerate the closing kinetics the L132C mutation (helix C) as well as the corresponding mutations at the homologous positions of VChR1, ReaChR and Chrimson significantly slowed the closing kinetics (Table 1). Structural information, the effect on the kinetics and the effect on the calcium permeability indicate a probable interaction of helix C and helix F at those critical residues.
Chrimson mutants with accelerated closing kinetics. As shown above the Y261F mutation speeds up channel closing in Chrimson (Fig. 1f, Table 1). We identified two additional helix F mutations, which accelerated Chrimson's closing kinetics, namely S267M and Y268F (Fig. 1b, g and Table 1). The combination of the helix F mutations had a cumulative effect, further accelerating channel closing by up to one order of magnitude ( Fig. 1g and Table 1). Chrimson mutants carrying the Y268F mutation showed reduced expression in NG cells and a hypsochromic shift of their action spectra by 11 nm (Supplementary Fig. 1 and Supplementary Table 2). The hypsochromic shift might result from an interaction of F268 with the polyene chain of the retinal, as this was shown for F265 (F226, ChR2-numbering), located at the homologous position in the C1C2 structure (Fig. 1a).
Of special interest for optogenetic applications are the fast mutant Chrimson Y261F/S267M (f-Chrimson) and the very fast mutant Chrimson K176R/Y261F/S267M (vf-Chrimson), which carries the additional K176R mutation (Fig. 1h). As described earlier 6 , the closing kinetics of Chrimson K176R (ChrimsonR) is accelerated by a factor of ∼2 compared to wildtype ( Table 1). The closing kinetics of f-Chrimson were strongly accelerated from τ off = 24.6 ± 0.9 ms (wt-Chrimson, n = 5) to τ off = 5.7 ± 0.5 ms (f-Chrimson, n = 5). At the same time f-Chrimson was highly expressed in NG cells (Supplementary Table 2). Vf-Chrimson had ultrafast closing kinetics of τ off = 2.7 ± 0.3 ms (n = 7), which is at least as fast as the closing kinetics of Chronos (τ off = 3.6 ± 0.2 ms) 6 , the fastest ChR known to date. Of note, the action spectra of f-Chrimson and vf-Chrimson were not blueshifted ( Supplementary Fig. 1), thereby preserving the benefits of longer wavelength activation. Compared to Chrimson wt, the functional properties of the mutants were almost unaltered with respect to the linear voltage dependence of the photocurrents ( Supplementary Fig. 2), cation permeabilities (Supplementary  Table 3), the moderate slowing of the closing kinetics at positive voltages ( Supplementary Fig. 3) and peak current inactivation ( Supplementary Fig. 4). Of note, the measured cation permeabilities contradict a recent publication 26 , but are in accordance with a previous study 6 . At a temperature of 34°C we measured τ off values of 3.2 ± 0.2 ms (n = 3) for f-Chrimson and 1.6 ± 0.1 ms (n = 3) for vf-Chrimson ( Supplementary Fig. 5). Hence the ultrafast kinetics of the Chrimson mutants in principle enables neural photostimulation in an exceptionally high frequency range of up to ∼600 Hz.
Ultrafast red-shifted optogenetics. We heterologously expressed f-Chrimson and vf-Chrimson in primary cultures of rat hippocampal neurons by means of adeno-associated virus-mediated gene transfer (AAV2/1). Patch-clamp experiments proved a robust neuronal expression and confirmed the substantially faster kinetics of the mutants (Supplementary Table 4). The application of light pulses (λ 1 = 594 nm, λ 2 = 640 nm) triggered spiking with high reliability (Fig. 2 and Supplementary Fig. 7). The investigation of the dependence of spike probability on light pulse intensity showed that neural photostimulation via f-Chrimson (0.37-1.27 mW/mm 2 ) and vf-Chrimson (0.09-3.18 mW/mm 2 )   Table 4) is likely due to expression differences as well as the variability of membrane resistance, capacitance and spiking threshold of the investigated neurons. The primary culture of rat hippocampal neurons comprises a multitude of different neuronal subtypes, most of which have a maximal firing frequency of 40-60 Hz 9 . Therefore, in most cases spike failures occurred at a frequency of 60 Hz (Fig. 2c). In single cases a frequency of 100 Hz was achieved (Fig. 2d). The investigation of neural photostimulation in the high frequency range is impeded by the heterogeneity of the primary neuronal culture. Therefore, we conducted patch-clamp experiments on parvalbumin-positive interneurons heterologously expressing vf-Chrimson. Parvalbumin-positive interneurons display a fast spiking phenotype, and predominantly supply inhibition to the perisomatic domain of other neurons 14 . Heterologous expression of vf-Chrimson was achieved by intracerebroventricular injection of AAVs in transgenic mice that expressed tdTomato under the control of the parvalbumin promotor. Therefore, parvalbuminpositive interneurons could be identified in neocortical brain slices by their red fluorescence ( Supplementary Fig. 8a).
Using current injections we determined a maximal intrinsic firing frequency of 301 ± 29 Hz (n = 8) for the parvalbuminpositive interneurons, as expected for the fast spiking phenotype (Fig. 3a, b) 14 . Of note vf-Chrimson enabled neural photostimulation up to the intrinsic limit of the cells with high temporal precision (Fig. 3c-e and Supplementary Fig. 8c). As demonstrated, some cells followed photostimulation up to 400 Hz (Fig. 3c, d). In 2/7 cells the occurrence of extra spikes in response to the light pulse was observed ( Supplementary Fig. 8b) 9 , which compromised the fidelity of neural photostimulation in those cases. We note that similar to previous work using AAV transduction and single photon stimulation 6 , it was necessary to adjust irradiation intensity individually for each neuron to achieve optimal stimulation fidelity ( Supplementary Fig. 9). To our knowledge these results represent the fastest light triggered spiking measured to date, and indicate that vf-Chrimson opens new possibilities for the investigation of high frequency network events, such as sharp wave-ripples 27 .
f-Chrimson is a promising candidate for hearing restoration. Optogenetics bears great potential for improving the restoration of vision and hearing 28,29 . Future oCIs shall use tens to hundreds of microscale light sources to focally stimulate tonotopicallyordered SGNs in Rosenthal's canal (Fig. 4a) 28 . For deaf people, the lower spread of excitation from the light source in oCIs 15 , promises improved frequency and intensity resolution when compared to the eCI 28 . However, much remained to be done prior to a potential clinical translation of the oCI. For example, so far cochlear optogenetics was established using blue ChRs expressed in transgenic rodents or in mice following prenatal viral-gene transfer 15 .
Here, we tested the potential of f-Chrimson for optogenetic stimulation of SGNs. We established postnatal viral gene transfer by injecting AAV2/6-hSyn-f-Chrimson-EYFP into the scala tympani via the round window in 3-6-day-old mice ( Fig. 4b).
We readily observed photocurrents in patch-clamp recordings from isolated SGNs 30 in the second postnatal week (Fig. 4c), proving the basic functionality of f-Chrimson in the target cells. We then in depth analyzed expression and function 4-14 weeks after injection. Using confocal imaging of EYFP and parvalbumin immunofluorescence in cochlear cryosections we found a high transduction rate (near 80 %) in the injected ear, which was not significantly different between the cochlear turns (Kruskal-Wallis ANOVA followed by Dunn's test, P > 0.05, n = 5; Fig. 4d, f). SGN showed clear plasma membrane expression of f-Chrimson (insets in Fig. 4f) and survived the optogenetic manipulation as evident from the unaltered SGN density when compared to the noninjected ear (Mann-Whitney U test, P > 0.05, n = 5; Fig. 4e, f). The non-injected ear showed f-Chrimson expression in less than 5% of the SGNs (Fig. 4d, f), indicating minimal spread of AAV from the injected ear likely via the cochlear aqueduct.
We then established single-channel oCI stimulation by performing a posterior tympanotomy and inserting an optical fiber (50 µm diameter) through the round window to project the light of a 594 nm laser onto the SGNs of the basal cochlear turn of young mice (2-3 months, Fig. 5a). We could readily elicit optical auditory brainstem responses (oABR, Fig. 5b, c) that differed between animals in waveform and amplitude. For comparison we  recorded acoustic auditory brainstem responses (aABRs, Fig. 5b, c lower panels) that were similar in amplitude and waveform to oABR and also varied between animals (Fig. 5b). We note that the similarity to aABRs and the shorter latency (0.93 ± 0.13 ms vs. approximately 3 ms 31 ) and smaller maximal amplitude (10.7 ± 3.80 µV vs. approximately 1000 µV 31 ) of oABRs when compared to our previous report on transgenic mice 31 indicates more specific activation of the auditory pathway in the case of postnatal AAV-injection used in present study. We then characterized the oABRs in response to different light intensities, light pulse durations and light pulse rate (n = 5 mice). oABR amplitude grew and oABR latency got shorter with increasing light intensity (Fig. 5c, d, g). Stimuli as weak as 0.5 mW (Fig. 5c,d, duration: 1 ms, rate: 20 Hz) and as short as 80 µs (Fig. 5e, h, rate: 20 Hz, intensity: 11 mW) were sufficient to drive oABRs. Amplitudes typically varied for changes in light intensity of more than one order (Fig. 5c, d, output dynamic range >20 dB for oABR). oABR amplitudes declined when raising stimulus rates ( Fig. 5f, i). However, f-Chrimson-mediated oABRs remained sizable up to stimulus rates of 200 Hz, suggesting high temporal fidelity of light-driven SGN firing. Next, we used aged C57BL6/J mice (9 months-old, n = 5 mice) to explore the potential of oCI to restore activity in the auditory pathway of a mouse model of age-related hearing loss 32 , which is a major form of hearing impairment in humans. Auditory thresholds, estimated by aABR elicited by acoustic clicks were elevated to above 50 dB (SPL) (58 ± 3.3 dB SPL, Fig. 5j-l, typically 20 dB in young mice) and aABR amplitudes were reduced to 1/3 of those in young mice across all SPLs tested (Fig. 5l). oABR amplitudes measured in these aged mice were comparable to values obtained for young animals (Fig. 5l), but latencies tended to be shorter and less variable ( Fig. 5g and Supplementary  Fig. 10h). Interestingly, we found that light pulses as short as 40 µs were able to elicit oABRs (11 mW at 20 Hz, Supplementary Moreover, we were able to record oABRs at stimulation frequencies as high as 250 Hz (11 mW, 1 ms pulse duration), likely due to the lower average latencies found in these aged mice as compared to their younger counterparts ( Supplementary  Fig. 10g, j). Together the data indicate that optical activation of the auditory pathway proceeded with at least as high efficiency in aged C57BL/6J mice despite their profound age-related hearing impairment. f-Chrimson expression levels throughout the injected cochlea were homogeneous (one-way ANOVA followed by Tukey's test, P > 0.05, n = 5). Importantly, longterm f-Chrimson expression (9 months) did not seem to decay significantly (Kruskal-Wallis ANOVA followed by Dunn's test, P > 0.05, n = 5) ( Supplementary Fig. 10a-c) nor cause any significant loss of SGNs in the AAV-injected ear of these mice, when compared to the non-injected ear (t-test for comparison of When tested with constant current injection, the input-output curve of PV-interneurons plateaus at a maximum firing rate of 270 ± 33 Hz (n = 8). c Example traces of the vf Chrimson-expressing PV-interneuron from a activated by light pulses (565 nm, 0.5 ms) at frequencies ranging from 50-500 Hz. Note that this interneuron reliably followed frequencies of up to 400 Hz. d Spiking probabilities of PV-interneurons at different optical stimulation frequencies. On average (black), PV-interneurons followed stimulation up to 300 Hz reliably (94 ± 5% spiking probability), and could still encode input frequencies of up to 400 Hz with a reliability of 68 ± 16% (n = 7; three whole-cell, four cell-attached recordings). e Action potential latency (assessed at peak) and action potential jitter (s.d. of latencies) after light pulse onset for all stimulation frequencies with reliable spiking (>85%). Error bars are s.e.m. Scale bars: a 50 ms, 10 mV c 50 ms, 10 mV cell density across cochlear turns in the injected and non-injected ear, P > 0.05, n = 5) ( Supplementary Fig. 10d).
In order to scrutinize the temporal fidelity of stimulation, we turned to juxtacellular recordings from single neurons 13,33 . We established single-channel oCI stimulation via an optical fiber and targeted electrodes through a craniotomy to where the auditory nerve enters the cochlear nucleus (Fig. 6a) in order to measure the neural photoactivation. Those neurons could not be identified based on a response to acoustic stimulation, most likely due to impaired acoustic hearing following ear surgery and oCI. Therefore, we termed light-stimulated neurons putative SGNs. We found that the putative SGNs fired upon optogenetic stimulation with high temporal precision for stimulus rates of up to hundreds of Hz (Fig. 6b-e): some neurons followed stimulation to some extent even up to 1 kHz (Fig. 6b, d). The spike latency amounted to approximately 2 ms for stimulus rates of up to 400 Hz ( Supplementary Fig. 11a, b), which is in agreement with the data obtained on the interneurons (Fig. 3e).
Temporal precision of firing, evaluated based on vector strength (ref. 34 , see methods, Fig. 6c, d) and temporal jitter (i.e., standard deviation of spike latency across trials, Fig. 6e, Supplementary  Fig. 11c) varied between the recorded neurons and, generally, was good. The vector strength declined with increasing stimulation rate up to 1 kHz. For a comparison, we re-plot the median vector strength of firing driven by transposed tones in mouse SGNs (ref. 35 , Fig. 6d) used because phase-locking to pure tones is hard to achieve in the high frequency mouse cochlea 36 . Temporal jitter, evaluated for spikes occurring in the time window equal to the period stimulus, was typically below a millisecond and tended to increase when raising stimulus rates up to 300 Hz (Fig. 6e). At higher stimulus rates, the temporal jitter was higher than the values obtained for simulated Poisson spike trains (see methods, gray area, Fig. 6e), reflecting a reduced spike synchronization with the light pulses. Interestingly, the spike jitter increased significantly from 25 ms compared to the start of the stimulation (Supplementary Fig. 11c). Spike probability (Fig. 6c, d)    as the rate of stimulation increased, indicating that optogenetic coding by individual SGNs becomes less reliable as stimulus rate rises. This, however, is likely compensated at the population level, as several SGNs jointly encode information from each place of the tonotopic map 13 .

Discussion
As demonstrated, the investigation of the molecular properties of microbial-type rhodopsins is essential for the development of variants with superior properties for particular optogenetic applications. Our study reveals the critical role of helix F for the closing kinetics of various ChRs. Using site-directed mutagenesis we generated fast variants of four different ChRs, which, together, cover photoactivation over the visible spectrum. We deem the fast Channel opening/closing of ChR2 is based on a concerted movement of helices B, F and G 23,24 . Interestingly, we discovered a major impact of the interaction between the moving helix F and the virtually immobile helix C on ChR kinetics. High-resolution structures of the investigated ChRs are not available. However, the information of the high-resolution structure of the C1C2 ChR chimera in combination with the light-induced helix movement studies by electron spin resonance and the lowresolution structure by 2D cryoelectron microscopy allowed us to identify the crucial position F219 in helix F for the construction of a faster ChR2 mutant. This position is also conserved in ReaChR and VCR1. Analogous mutations lead to accelerated kinetics of channel closing. Of note, the surprisingly slow reacting Chrimson has already a tyrosine on this position (Fig. 1a, b). We suspected that the mutation of this tyrosine influences channel closing kinetics. Indeed, the mutation to a phenylalanine in the analogous position (Y261F) results to faster channel closing in Chrimson. Further inspection of the alignment of helix F shows at positions M225 and F226 in ChR2 the same analogy for VCR1 and ReachR but not for Chrimson. For Chrimson, mutations on these positions (S267M, Y268F) in addition to the Y261F mutation result in the ultrafast switching behavior (Fig. 1g,h). The surprising and peculiar phenotype of the back mutations in Chrimson is hard to explain without a high resolution structure of the protein.
This study achieved two important breakthroughs towards developing cochlear optogenetics for auditory research and future improved hearing restoration. First, we managed to achieve efficient, non-traumatic and neuron-specific expression of f-Chrimson in SGNs using postnatal AAV-injection into scala tympani through the round window. We found near 80% of the SGNs of the injected ear to express f-Chrimson at high levels and mostly in the plasma membrane of somas and neurites, which persists for at least 9 months after injection. These transduction rates were much higher than those achieved with transuterine injection of AAV2/6-hSyn-CatCh-YFP 15 and unlike there, independent from tonotopic position. We consider the minute transduction of the non-injected ear to reflect viral spread through the cerebrospinal fluid space, which calls for further optimization of the injection protocol 37 . Importantly, we did not find evidence for neuron loss even at 9 months after injection and we expect little, if any, risk of phototoxicity given the red-shifted action spectrum of f-Chrimson.
Secondly, using f-Chrimson, we overcame the likely biggest roadblock of current ChR2-based cochlear optogenetics: low temporal bandwidth of optical coding (<80 Hz) 15 . We found robust and fast photocurrents in cultured isolated f-Chrimsonpositive SGNs. In vivo, fiber-based stimulation resembling singlechannel oCI elicited activation of the auditory pathway in hearing and deaf mice. Using far-field neural population responses (oABR) as a readout we found low thresholds for radiant flux and energy (<0.5 mW, <0.5 µJ) as well as duration (<100 µs) and a wide dynamic range of coding (more than 20 dB, no saturation for most animals at maximal stimulation). This brings the oCI closer to the currently used eCI stimulation parameters (0.2 µJ and 80 µs per pulse) 38 and exceeds the eCI output dynamic range (<10 dB) 17 . Both, recordings of oABR and of firing in single putative SGNs indicated that f-Chrimson mediated oCI-enabled responses to follow at least 200 Hz, which corresponds to physiological steady-state firing rates of SGNs 13 . In fact, we found some neurons to follow stimulation to several hundreds of Hz nearly mimicking sound-evoked SGN activity. The closing kinetics of f-Chrimson and the resulting relative refractoriness probably also limits the temporal precision of f-Chrimson-mediated SGN firing. At 500 and 1000 Hz, vector strength, a measure commonly used to analyze the extent of phase-locking in SGNs 39 , was lower for f-Chrimson-mediated optogenetic stimulation than for mouse SGN firing with transposed tones 35 . We note that the SGNs recorded in the present study typically did not fire spontaneously probably due to the ear surgery. Besides the lack of spontaneous firing, the short (1 ms) and pulsatile optogenetic stimulation typically evoking a single spike likely explain why vector strength tended to be higher for low stimulus rates when compared to transposed tones, for which several spikes were generated per stimulus cycle. Moreover, vector strength and temporal jitter of f-Chrimson-mediated SGN firing in mice indicate a lower temporal precision than that of acoustic hearing and electric stimulation in species with prominent phase-locking of SGN firing 33,39 . Nonetheless, we reason that, even if the limited probability and temporal precision of single SGN firing for optogenetic stimulation at 100-500 Hz translated to species other the mouse, this will not impede the coding at the level of the auditory nerve population. Hence, we conclude that f-Chrimson is a good candidate opsin for the oCI. In fact, higher temporal jitter in response to optogenetic than electrical stimulation might render unnecessary the very high stimulation rates employed in eCI to avoid overly synchronized activity in the auditory nerve 17 . The favorable properties of the novel Chrimson mutants also facilitate multiple applications in basic neurosciences and in sensory restoration, such as the recovery of vision 29 .
In order to assess the permeability of potassium ions relative to the permeability of sodium ions (P K /P Na ), we measured photocurrents at voltages ranging from −120 to +40 mV in 20-mV steps. The P K /P Na ratio was determined from the difference of the reversal potentials of the photocurrents when replacing 90 mM NaCl, 2 mM MgCl 2 , 5 mM MOPS/TRIS pH 9 for 90 mM KCl, 2 mM MgCl 2 , 5 mM MOPS/TRIS pH 9 5 . The relative proton permeability was calculated from the photocurrent reversal potential in buffer containing 90 mM NMG, 5 mM KCl, 2 mM MgCl 2 , 5 mM MOPS/TRIS pH 9 using the Goldmann-Hodgkin-Katz equation 25 and assuming a cytoplasmic potassium concentration of 100 mM and an intracellular pH of 7.3 5 .
Electrophysiological recordings on NG108-15 cells. For the electrophysiological characterization of mutant ChRs whole cell patch-clamp were performed under voltage clamp conditions 42 using the Axopatch 200B amplifier (Axon Instruments, Union City, USA) and the DigiData 1322A interface (Axon Instruments, Union City, USA). Patch pipettes with resistances of 2-5 mΩ were fabricated from thin-walled borosilicate glass on a horizontal puller (Model P-1000, Sutter Instruments, Novato, USA). The series resistance was <10 MΩ and the input resistance ranged from 1.1 to 4.6 GΩ. The mean capacitance of the measured cells was 34.6 ± 24.3 pF (n = 61). If not stated differently the pipette solution contained 110 mM NaCl, 2 mM MgCl 2 , 10 mM EGTA, 10 mM HEPES, pH 7.4 and the bath solution contained 140 mM NaCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, pH 7.4.
In order to assess the permeability of calcium ions relative to the permeability of sodium ions (P Ca /P Na ), we measured photocurrent-voltage relationships and determined the reversal potential. The intracellular solution contained 110 mM NaCl, 10 mM EGTA, 2 mM MgCl 2 and 10 mM Tris (pH = 7.4) and the extracellular solution contained 140 mM NaCl, 2 mM MgCl 2 and 10 mM Tris (pH = 9). For the determination of the P Ca /P Na values, external 140 mM NaCl was exchanged with 90 mM CaCl 2. Permeability ratios were calculated according to the Goldman-Hodgkin-Katz equation 25 .
For determination and comparison of the off-kinetics and current densities, NG108-15 cells heterologously expressing aforementioned ChRs and channelrhodopsin mutants were investigated at a membrane potential of −60 mV. Photocurrents were measured in response to 3 or 500 ms light pulses with a saturating intensity of 23 mW/mm 2 using diode-pumped solid-state lasers (λ = 473 nm for ChR2 variants, λ = 532 nm for VChR1 and ReaChR variants, λ = 594 nm for Chrimson variants) focused into a 400-μm optic fiber. Light pulses were applied by a fast computer-controlled shutter (Uniblitz LS6ZM2, Vincent Associates, Rochester, USA).
The current density (J -60 mV ) was determined by dividing the stationary current in response to a 500 ms light pulse with a saturating intensity of 23 mW/mm 2 by the capacitance of the cell. In order to avoid an experimental bias, the NG108-15 cells for the electrophysiological recordings were chosen independent of the brightness of their EYFP fluorescence. The τ off value was determined by a fit of the decaying photocurrent to a monoexponential function. In order to investigate the dependence of the off-kinetics on the membrane potential τ off values were determined at membrane potentials ranging from −120 to +60 mV.
If not stated differently the off-kinetics was determined at room temperature (297 K). The temperature dependence of the off-kinetics of Chrimson-EYFP wt and Chrimson-EYFP K176R/Y261F/S267M (vf-Chrimson-EYFP) was investigated at temperatures ranging from 284 to 307 K. Photocurrents recorded at a temperature of 307 K were measured in response to 7 ns light-pulses with a wavelength of 594 nm in order to avoid tampering of the off-kinetics due to the opening/closing time of the shutter (700 µs). The ns light pulses were generated with the Opolette 355 tunable laser system (Opotek Inc, Carlsbad, USA). Thereby the pulse energy was set to value of 10 19 photons/m 2 .
The Opolette 355 tunable laser system was further used for the measurement of the action spectra of the Chrimson variants. For the recordings the pulse energies at the different wavelengths were set to values which corresponded to equal photon counts of 10 18 photons/m 2 for Chrimson wt and 10 19 photons/m 2 for the Chrimson mutants. The light pulses had a pulse width of 3 ms, a wavelength of λ = 594 nm and a saturating intensity of 11-30 mW/mm 2 . The τ off value was determined by a fit of the decaying photocurrent to a monoexponential function. The current density (J -70 mV ) was determined by dividing the stationary current in response to a 500 ms light pulse with a saturating intensity of 20-40 mW/mm 2 and a wavelength of 594 nm by the capacitance of the cell. In order to determine the lowest light intensity required to induce action potentials with a probability of 100% (J 100 ), 40 pulses (λ = 594 nm, pulse width = 3 ms, ν = 10 Hz) of varying light intensities were applied. The spike probability was calculated by dividing the number of lighttriggered spikes by the total number of light pulses.
Animals for recordings on parvalbumin-positive interneurons. Experimental mice were obtained from a cross of PV-ires-cre 43 and conditional tdTomato animals Ai9, male and female, 4-12 weeks 44 . Mice were maintained in a 12 h light/ dark cycle, with access to food and water ad libitum. All animal procedures were performed in accordance with institutional guidelines and were approved by the Regierungspräsidium Darmstadt.
Intracerebroventricular injections. Prior to pup injections 45 , the dam was habituated to the experimenter and the experimental room. Newborn mice (P2) were anesthetized using isoflurane (2-3%), and placed on a light source to reveal skull structures. Injections of 2 µl of AAV2/1-hSyn-vf-Chrimson-EYFP were performed into the right ventricle using a glass pipette (coordinates from bregma; rostral 0.75 mm, lateral 0.25 mm and ventral 2 mm). After injection pups were recovered for 5 min in a pre-warmed container with homecage-bedding before being placed back in the home cage.
Cloning for AAV2/6 production. pcDNA3.1(−)_f-Chrimson_EYFP was used as a starting material for cloning pAAV_hSyn_f-Chrimson_EYFP. The sequence of f-Chrimson_EYFP was amplified by means of a classical PCR. The resulting PCR fragment was then digested with BamHI/HindIII (Thermo Scientific, MA, USA), gel extracted (GeneJET Gel Extraction Kit, Thermo Scientific, MA, USA) and further used for ligation. At the same time the plasmid pAAV_hSyn_Chronos_GFP (Addgene, plasmid nr. 59170) was also digested using restriction enzymes BamHI/ HindIII and used as a backbone plasmid. All obtained ligation products were further tested by means of colony PCR and finally sequenced by an external company. The final product was then sent to the University of North Carolina Vector Core (Chapel Hill, NC, USA), and used to produce AAV2/6.
Postnatal AAV injection into the cochlea. All experiments were done in compliance with the German national animal care guidelines and were approved by the board for animal welfare of the University Medical Center Göttingen and the animal welfare office of the state of Lower Saxony. The calculation of animal number was performed prior to starting experiments. We planned to use the Wilcoxon Rank Sum Test and an error probability alpha smaller than 0.05, a power (1-beta) of 0.95 and effect size depending on the precise experimental protocol.
Postnatal AAV-injection into scala tympani of the left ear via the round window 46 was performed at p3-p6 on C57BL/6 wild-type mice, using AAV2/6 and the human synapsin promoter to drive transgenic expression of f-Chrimson-YFP in SGNs. In brief, under general isoflurane anesthesia and local analgesia achieved by means of xylocaine, the left ear was approached via a dorsal incision and the round window membrane was identified and gently punctured using a borosilicate capillary pipette that was kept in place to inject approximately 5 × 10 9 viral genomes. After virus application, the tissue above the injection site was repositioned and the wound was sutured and buprenorphine (0.1 mg/kg) was applied as pain reliever. Recovery of the animals was then daily tracked. Mice were randomly selected for injection in all experiments. No blinding was possible since injections have to be performed in the left ear leaving the right ear as an internal control. Hence, surgery prior to stimulation needed to be done in the injected ear. Animals were then kept in a 12 h light/dark cycle, with access to food and water ad libitum.
Animal surgery for recordings on the auditory pathway. Mice were anesthetized with i.p. administration of a mixture of xylazine (5 mg/kg) and urethane (1.32 mg/ kg) while analgesia was achieved with buprenorphine. The core temperature was maintained constant at 37°C using a custom-designed heat plate on a vibration isolation table in a sound-proof chamber (IAC GmbH, Niederkrüchten, Germany).
For auditory nerve recordings, a tracheostomy was performed before the animals were positioned in a custom-designed stereotactic head holder. Pinnae were removed, scalp reflected, portions of the lateral interparietal and of the left occipital bone removed, and a partial cerebellar aspiration performed to expose the surface of the cochlear nucleus.
Optical stimulation in vivo. The left bulla was reached using a retroauricular approach and opened to expose the cochlea. A 50 µm optical fiber coupled to a 594-nm laser (OBIS LS OPSL, 100 mW, Coherent Inc., Santa Clara, CA, USA) was inserted into the cochlea via the round window. Radiant flux was calibrated with a laser power meter (LaserCheck; Coherent Inc., Santa Clara, CA, USA).
Auditory brainstem responses. For stimulus generation and presentation, data acquisition, and off-line analysis, we used a NI System (National Instruments, Austin, TX, USA) and custom-written MATLAB software (The MathWorks, Inc., Natick, MA, USA). Optically evoked ABRs (oABRs) and acoustically evoked ABRs (aABRs) were recorded by needle electrodes underneath the pinna, on the vertex, and on the back near the legs. The difference potential between vertex and mastoid subdermal needles was amplified using a custom-designed amplifier, sampled at a rate of 50 kHz for 20 ms, filtered (300-3000 Hz) and averaged across 1000 and 500 presentations (for oABRs and aABRs, respectively). Thresholds were determined by visual inspection as the minimum sound or light intensity that elicited a reproducible response waveform in the recorded traces.
Juxtacellular recordings from single putative SGNs. For auditory nerve recordings 15 , a glass microelectrode (~25 MΩ) was advanced through the posterior end of the anteroventral cochlear nucleus, aiming toward the internal auditory canal using an Inchworm micropositioner (EXFO Burleigh). Extracellular action potentials were amplified using an ELC-03XS amplifier (NPI Electronic, Tamm, Germany), filtered (band pass, 300-3000 Hz), and digitized (TDT System 3) using custom-written Matlab (Mathworks) software. Data were further analyzed and prepared for display off-line using custom-written Python (Python Software Foundation, Delaware, USA) and Matlab software. Once light-responsive fibers were encountered, stimulation was performed by means of 400 or 900 ms-long light-pulse trains at varying stimulation rates, leaving 100 ms inter-train recovery over 20 repetitions. Responses within the first 400 ms were then used for analyses. Only recordings for which the fibers generate at least five spikes per light-pulse train (on average across the 20 iterations recorded for each frequency tested on each fiber) were included. Phase-locking was quantified using the vector strength 33 and its significance tested with the Rayleigh test. If L > 13.8, the null hypothesis was rejected at the 0.001 significance level: 47 insignificant VS were set to 0. The spike jitter, defined as the standard deviation of spike latency in one period of stimulation, was calculated using a time window equal to the stimulation period. The hazard function of the temporal jitter was evaluated for each stimulation rate by simulating Poisson spike trains at discharge rates from 10 to 300 spikes/s. The spike probability is the ratio between the number of spikes and the number of light-pulses. The temporal jitter is the standard deviation of spike latency across trials.
Data analysis. The data were analyzed using Matlab (The MathWorks, Inc., Natick, MA, USA), Excel, Igor Pro 6 (Wavemetrics, Portland, OR, USA), Origin 9.0 (OriginLab, Inc., Northampton, MA, USA), and GraphPad Prism (GraphPad Software, La Jolla, CA, USA). Averages were expressed as mean ± s.e.m. or mean ± s.d., as specified. References to data in the main text were expressed as mean ± s.e. m. For statistical comparison between two groups, data sets were tested for normal distribution (the D'Agostino & Pearson omnibus normality test or the Shapiro-Wilk test) and equality of variances (F-test) followed by two-tailed unpaired Student's t-test, or the unpaired two-tailed Mann-Whitney U test when data were not normally distributed and/or variance was unequal between samples.
For evaluation of multiple groups, statistical significance was calculated by using one-way ANOVA test followed by Tukey's test for normally distributed data (equality of variances tested with the Brown-Forsythe test) or one-way Kruskal-Wallis test followed by Dunn's test for non-normally distributed data.
Data availability. The data that support the findings of this study and code used for analysis are available from the corresponding author upon reasonable request.