Optical Stimulation of Zebrafish Hair Cells Expressing Channelrhodopsin-2

Vertebrate hair cells are responsible for the high fidelity encoding of mechanical stimuli into trains of action potentials (spikes) in afferent neurons. Here, we generated a transgenic zebrafish line expressing Channelrhodopsin-2 (ChR2) under the control of the hair-cell specific myo6b promoter, in order to examine the role of the mechanoelectrical transduction (MET) channel in sensory encoding in afferent neurons. We performed in vivo recordings from afferent neurons of the zebrafish lateral line while activating hair cells with either mechanical stimuli from a waterjet or optical stimuli from flashes of ∼470-nm light. Comparison of the patterns of encoded spikes during 100-ms stimuli revealed no difference in mean first spike latency between the two modes of activation. However, there was a significant increase in the variability of first spike latency during optical stimulation as well as an increase in the mean number of spikes per stimulus. Next, we compared encoding of spikes during hair-cell stimulation at 10, 20, and 40-Hz. Consistent with the increased variability of first spike latency, we saw a significant decrease in the vector strength of phase-locked spiking during optical stimulation. These in vivo results support a physiological role for the MET channel in the high fidelity of first spike latency seen during encoding of mechanical sensory stimuli. Finally, we examined whether remote activation of hair cells via ChR2 activation was sufficient to elicit escape responses in free-swimming larvae. In transgenic larvae, 100-ms flashes of ∼470-nm light resulted in escape responses that occurred concomitantly with field recordings indicating Mauthner cell activity. Altogether, the myo6b:ChR2 transgenic line provides a platform to investigate hair-cell function and sensory encoding, hair-cell sensory input to the Mauthner cell, and the ability to remotely evoke behavior in free-swimming zebrafish.


Introduction
In the auditory, vestibular, and lateral-line systems, the hair cell is the sensory receptor responsible for encoding mechanical stimuli into trains of action potentials in afferent neurons [1,2]. This mechanotransduction process is accomplished with incredible sensitivity and precision, involving a complex interplay of intrinsic hair-cell mechanisms [3][4][5]. At least two important features of hair cells ensure the speed and precision of sensory transduction, the mechanoelectrical transduction (MET) channel and the ribbon synapse. While much is known about the properties of these two features, how the two contribute to the overall encoding of sensory stimuli is not as well understood.
Located at the tips of hair-cell stereocilia, the MET channel opens within 50 ms following deflection, making it one of the fastest gating channels in the nervous system [3,[6][7][8]. In addition to its rapid gating kinetics, the MET channel has a very large single channel conductance [9]. Combined, these properties of the MET channel allow for graded deflections of stereocilia to result in quick and robust changes in the hair-cell membrane potential. This change in receptor potential rapidly activates voltage-gated calcium channels at the hair-cell ribbon synapse and results in precisely timed fusion of synaptic vesicles and release of glutamate into the synaptic cleft [5].
Due to the rapid and continuous nature of hair-cell stimuli, the ribbon synapse is required to provide the requisite release of glutamate [7]. The ribbon synapse is characterized by the ribbon body, an electron dense structure that is thought to both coordinate synaptic vesicles and allow for their sustained fusion with high temporal fidelity [10][11][12][13]. Together, the ribbon synapse and the MET channel function to encode sensory information rapidly and faithfully. Here, we sought a means to examine the MET channel's contribution to the encoding of action potentials in afferent neurons. To accomplish this, we turned to optogenetics in order to depolarize the receptor potential without activating MET channels and then compared optically evoked activity with responses to mechanical stimuli.
In order to depolarize hair cells while circumventing activation of the MET channel, we generated a transgenic zebrafish line with hair-cell specific expression of Channelrhodopsin-2 (ChR2) using the myo6b promoter [14]. ChR2 is a light-gated ion channel maximally excited by ,470-nm wavelength light [15,16]. When expressed in neurons, flashes of ,470-nm light open ChR2 channels, which then depolarizes the cell membrane and evokes action potentials [15][16][17]. Previously, ChR2 has been expressed in various zebrafish neurons, including those to elicit escape responses and control eye moment [18][19][20]. Here, to examine the contribution of hair-cell mechanisms to the encoding of action potentials in afferent neurons, we performed in vivo recordings from afferent neurons of the lateral line in transgenic myo6b:ChR2 zebrafish. We compared afferent neuron activity during hair-cell excitation by mechanical activation of MET channels with a waterjet and by optical activation of ChR2, effectively circumventing the MET channel.
In addition to studying the role of the MET channel in afferent spike encoding, we examined behavioral escape responses of transgenic larvae during remote activation of transgenic hair cells. In teleost fish, the fast-start escape response is generated by the Mauthner cells (M-cells), a pair of large reticulospinal neurons in the hindbrain that are excited by sensory inputs including afferent neurons from the ear and lateral line [21][22][23][24][25]. Optical stimulation of ChR2-expressing hair cells with flashes of ,470-nm light generated an escape response that was observed coincidentally with field potentials indicating an M-cell response evoked from free-swimming fish. By shortening the duration of optical stimuli, we found that the escape response and the coincident field potential disappeared when stimuli were less than 100 milliseconds (ms). Altogether, our findings highlight both the vital role of MET channels in the faithful encoding of hair-cell stimuli into action potentials in afferent neurons, and the use of optophysiology to study hair-cell evoked behaviors in zebrafish.

Ethics statement
All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at Amherst College under assurance number 3925-1 with the Office of Laboratory Animal Welfare. Zebrafish were raised and maintained according to standard procedures [26]. Adults and larvae were maintained under a 14-hour light cycle at 28.5uC.

Generation of myo6b:ChR2-YFP construct and transgenic line
The myo6b:ChR2-YFP plasmid was generated using standard cloning procedures. The myo6b-GFP plasmid containing an I-SceI meganuclease cut site was obtained from Dr. Teresa Nicolson at Oregon Health and Science University and the Vollum Institute. The ChR2-YFP plasmid was obtained from Dr. Karl Deisseroth at Stanford University. We identified restriction enzyme sites in the myo6b plasmid and removed the GFP sequence via restriction digest. We then ligated the ChR2-YFP sequence into the meganuclease plasmid, resulting in ChR2-YFP expression controlled by the hair-cell specific myo6b promoter. The myo6b:ChR2-YFP plasmid, along with I-SceI enzyme, was injected into zebrafish embryos at the one cell stage [27]. DNA injections were performed using a pressure injector (MPPI-3, Applied Scientific Instruments, Eugene, Oregon). After DNA injection, F 0 embryos were identified by visualization of YFP fluorescence, and then raised and outcrossed to wild-type stock. The resulting F 1 larvae that were screened for hair-cell specific YFP expression were subsequently raised. Adult F 1 fish were then outcrossed to wildtype fish and ChR2-YFP-positive F 2 larvae were used for experiments.

Microscopy and imaging
Afferent neuron recordings and imaging were performed using 10X and 40X objectives on a fixed-stage upright microscope (BX51WI; Olympus, Center Valley, Pennsylvania) equipped with a FITC filter cube (Chroma Technology, Bellows Falls, Vermont) and a SOLA fluorescent light source (Lumencor Inc., Beaverton, Oregon). Images were captured using a Spot Pursuit Camera (1.4 MP Monochrome w/o IR; Diagnostic Instruments, Sterling Heights, Michigan) and either the upright or a dissecting microscope (Nikon SMZ 1500; Melville, New York). YFPfluorescent images were pseudocolored yellow using ImageJ (NIH, Bethesda Maryland) and were adjusted for brightness and contrast using Photoshop CS5 (Adobe, San Jose, California). For the adult transgenic image in Figure 1D, the background surrounding the fish body was removed for clarity using Photoshop CS5. Escape responses were recorded with a Chromachip II Camera, (Javelin Electronics, High Wycombe, United Kingdom) mounted on a dissection microscope (Olympus SZ40; Center Valley, Pennsylvania). To filter the bright flash of blue excitation light from the video, incoming light to the camera was passed through an orange acrylic filter (Pearl Biotech, San Francisco, California). Video output from the camera was collected on a PC computer using a USB-capture device and software (Elgato Video, San Francisco, California).

Lateral-line electrophysiology
Our recording procedures were previously described in detail [28]. Briefly, larvae were anesthetized, mounted, and microinjected in the heart with 125 mM a-bungarotoxin to block muscle activity (Abcam, Cambridge, Massachusetts). Larvae were then rinsed and returned to normal extracellular solution (in mM: 130 NaCl, 2 KCl, 2 CaCl 2 , 1 MgCl 2 and 10 HEPES, pH 7.8, 290 mOsm). Extracellular recordings were performed at room temperature with borosilicate glass recording electrodes (Sutter Instruments, Novato, CA) fabricated with long tapers and resistances between 5 and 15 MV in extracellular solution (P-97 Puller; Sutter Instruments, Novato, CA). Extracellular action currents were recorded from an individual lateral line afferent neuron in the loose-patch configuration (seal resistances ranged from 20 to 80 MV). Recordings were done in voltage-clamp mode, sampled at 50 ms/pt, and filtered at 1 kHz with an EPC 10 amplifier and Patchmaster software (Heka Electronic, Bellmore, New York).

Mechanical stimulation
Stimulation of neuromast hair cells was performed as previously described [28]. Briefly, mechanical stimuli were delivered to hair cells using a pressure clamp (HSPC-1; ALA Scientific, New York) attached to a glass micropipette (tip diameter ,30 mm) filled with normal extracellular solution. This waterjet was positioned approximately 100 mm from a given neuromast and the displacement of the neuromast kinocilia was verified by eye. The waterjet pressure clamp was driven by a step voltage command delivered by the recording amplifier via the Patchmaster software. The stimulus pressure was monitored and recorded via a feedback sensor located on the pressure clamp headstage. After establishing a recording from a given afferent neuron, its primary innervated neuromast was identified by progressively stimulating from neuromast to neuromast until phase-locked spiking was observed.

Optical stimulation
Hair cells of wild type and transgenic zebrafish larvae were optically stimulated using flashes of light from a fluorescent light source (SOLA Light Driver; Lumencor, Beaverton Oregon). White light flashes were subsequently filtered via a narrow-pass FITC excitation filter (460 to 490 nm; Chroma Technology, Bellows Falls, Vermont) and transmitted through a 40X water immersion lens (Olympus, Center Valley, Pennsylvania) onto the mounted larva. Optical flashes were triggered via a 5-volt TTL output from the EPC10 amplifier and Patchmaster software (HEKA Electronik, Bellmore, New York) to a remote control accessory (RCA; Lumencor, Beaverton Oregon) on the SOLA light source. Light intensity at the level of the sample was measured at 6.9-klux using a Light Meter Probe (MLT331; AD Instruments, Colorado Springs, Colorado). For behavioral experiments, light flashes were delivered using a blue LED light (470 nm; LEDSupply.com, Randolph, Vermont) with a Tight Spot LED Optic lens (Carclo, Latrobe, Pennsylvania) connected to a 1,000 mA BuckPuck driver (LEDSupply.com, Randolph, Vermont). Flashes were triggered with a TTL signal from the Powerlab 26T amplifier (AD Instruments, Colorado Springs, Colorado) used for the hindbrain recordings. Light intensity for this behavioral preparation was measured at 3.1-klux.

Field Recordings
M-cell field recording procedures were based on those from previously published studies [29,30]. All recordings were performed with larvae in a drop of distilled water at room temperature. For touch stimulation, larvae were stimulated with a waterjet while either embedded in 2% low-melt agarose or free swimming, and we did not observe a significant difference in field potentials between the two (data not shown). For optical stimulation, all larvae were free swimming. Field potentials were recorded with stainless steel insect pins embedded in the bottom of a Sylgard-coated dish (Dow-Corning, Midland, Michigan). Potentials were recorded using an extracellular amplifier (Model 3000; A-M systems Inc, Carlsborg, Washington) with 5000X gain, a 300-Hz high pass filter, and 1-kHz low pass filter and were then collected with a PowerLab 26T (AD instruments, Colorado Springs, Colorado).

Signal analysis
Data were analyzed using custom software written in Igor Pro (Wavemetrics, Lake Oswego, Oregon) and were plotted with Igor Pro and Prism 5 (Graphpad, La Jolla, California). Data sets were tested for normality using Kolmogorov-Smirnov normality tests. Statistical significance between conditions was determined using paired two-tailed Student's t tests. In order to evaluate the fidelity of phase locking for evoked spiking, we converted individual spike times, with time = 0 coinciding to stimulus onset, to unit vectors with appropriate phase angle for the given stimulus and then calculated vector strength (r) [31].

Results
In order to examine the contribution of the MET channel to the rapid and precise encoding of action potentials in afferent neurons, we sought a means to depolarize hair cells without activating MET channels. Therefore, we generated a stable transgenic line of zebrafish expressing the light-activated ion channel, Channelrhodopsin-2 (ChR2), in hair cells.
Hair cell ChR2 expression in the myo6b:ChR2-YFP transgenic zebrafish To express ChR2 in hair cells, we generated transgenic zebrafish expressing ChR2 tagged with yellow fluorescent protein (YFP) driven by the myo6b promoter (myo6b:ChR2-YFP; Fig. 1A). In order to confirm hair-cell expression of ChR2 in larval zebrafish, we analyzed YFP fluorescence from six stable transgenic lines at 5 days post fertilization. For our experiments, we selected the line with the strongest fluorescence in all hair cells of the ear and lateral line ( Fig. 1B and 1C). Importantly, we did not observe any obvious misexpression of YFP including any in the larval eye ( Fig. 1B' and 1B''). In addition, YFP fluorescent hair cells were observed as soon as they are deposited along the lateral line (data not shown). Finally, the transgenic line continued to express ChR2-YFP in hair cells during development and throughout adulthood (Fig. 1D). This stable expression provides an exciting opportunity for optogenetic experiments with adult animals. Importantly, we have not observed any changes in adult zebrafish mating, fecundity, or life span. Using our newly created transgenic line, we performed in vivo electrophysiological recordings from single afferent neurons of the lateral line while alternately stimulating transgenic hair cells with either a mechanical waterjet or optical flashes of ,470-nm light. Both mechanical and optical stimulation using 100-ms squarewave pulses produced similar trains of action potentials (spiking) in afferent neurons (Fig. 2A). Recordings from wild-type larvae showed no responses during presentation of similar flashes of light (data not shown). In order to characterize spike encoding during repeated pulses, we recorded 60 consecutive sweeps during both mechanical and optical stimulation (Fig. 2B). Interestingly, there was a significant increase in the number of spikes per stimulus with optical stimulation (Fig. 2C; mechanical spikes per sweep = 4.460.9, optical spikes sweep = 5.761.0; n = 9; p,0.05).
Auditory features are encoded in afferent neuron spiking using multiple parameters, including total spike number and first spike latency [32]. Total spike number conveys information about the amplitude, frequency, and spatial location of a sound [33,34]. However, it has been reported that the nervous system must average the total number of spikes across multiple stimulus presentations in order to encode these features accurately [32,34]. First spike latency, which is the time from stimulus onset to when the first action potential is observed, is also known to encode the amplitude, frequency, and spatial location of stimuli [35,36]. In addition, first spike latency is not thought to require averaging across repeated stimuli in order to encode these features accurately [32,34,36]. Thus, animals may use first spike latency to respond quickly to auditory stimuli [37]. Presumably, in order for this fast encoding to be meaningful, these initial spikes should occur with minimum variability from stimulus to stimulus.
To determine the fidelity of first spike latencies across repeated stimuli, we analyzed recordings of 60 consecutive sweeps during mechanical and optical stimulation. Our analysis revealed no significant difference between mean first spike latency (FSL) during mechanical versus optical stimulation ( Fig. 2D; mechanical FSL = 9.060.9 ms; optical FSL = 11.260.7 ms; n = 9; p = 0.11). This result was somewhat surprising, as ChR2 gating has been shown to be significantly slower than MET channel gating [3,38], though the overexpression of ChR2 may confound our results. Further analysis revealed that the coefficient of variation (C.V.) of first spike latencies was significantly greater with optical stimuli (Fig. 2E; p,0.05). One possibility for this result is that depolarization via ChR2 activation is not as consistent as depolarization through collective, and concerted activation of MET channels [39,40].

Comparison of phase-locked spiking during mechanical and optical stimulation of hair cells
Lateral-line hair cells are capable of encoding higher frequency mechanical stimuli with great accuracy (see [41]). Given the increased variability of first spike latency seen with optical stimuli, we examined the temporal fidelity of spike encoding during 20-Hz mechanical and optical stimulation. Visual inspection of single sweeps revealed that spiking was more closely phase locked to the stimulus onset for mechanical stimuli than for optical stimuli (Fig. 3A). To examine the relationship between the time of stimulus onset and the subsequently evoked spikes, we constructed polar plots with spikes from 60 consecutive sweeps at 20 Hz. Figure 3B shows that the majority of spikes occurred near stimulus onset with mechanical stimulation, whereas when stimulated optically, spiking occurred throughout the duration of the stimulus with some spikes occurring after stimulus offset (Fig. 3B). This experiment provides further evidence that depolarization of hair cells through ChR2 activation is less accurate than the rapid and concerted activation of MET channels with mechanical stimuli. A reduction in phase locking of spikes in response to delivered stimuli is often quantified by calculating vector strength, r [31]. Here, we quantified vector strength during delivery of 10, 20, and 40-Hz mechanical and optical stimuli (Fig. 3C). Across all frequencies, optical activation of hair cells (10 Hz r = 0.6460.04; 20 Hz r = 0.6560.04; 40 Hz r = 0.6360.03; n = 7) led to significantly less accurate encoding than mechanical stimulation (10 Hz r = 0.8060.06, p,0.05; 20 Hz r = 0.8260.03, p,0.001; 40 Hz r = 0.7760.03, p,0.01). Altogether, these results highlight the importance of the MET channel for the faithful encoding of relevant sensory stimuli into action potentials in afferent neurons of the lateral line.

Hair-cell activation of the escape response
Hair-cell sensory information is vital to the startle and escape responses in vertebrates. We determined whether remote activation of hair cells with optical stimuli could evoke an escape response in transgenic larvae. Delivery of both touch stimuli with a waterjet and optical stimuli with flashes of ,470-nm light evoked similar escape responses. In addition, we recorded field potentials in order to determine whether a pattern generated by M-cell activity was similar during escape responses from the two modes of stimulation ( Fig. 4B-C). Waveforms from field potential recordings indicated initial M-cell responses, as well as activity from other hindbrain neurons and subsequent contraction of axial muscles [29,30]. Both wild type and myo6b:ChR2 transgenic larvae displayed similar field potentials in response to touch stimuli delivered via a waterjet directed at the head (Fig. 4D). However, wild-type larvae did not respond to a 100-ms flash of ,470-nm light ( Fig. 4E; Video S1; n = 18), while myo6b:ChR2 larvae displayed a robust escape response along with concomitant field potentials ( Fig. 4E; Video S2; n = 55). If hair-cell inputs were bringing the M-cell membrane potential to threshold, we predicted that shortening the duration of the optical flash would decrease spiking of afferent neurons and thus lower the probability of an Mcell action potential. Consistent with our prediction, we found that by shortening optical stimulus duration, we reduced the frequency of observed escape responses and coincident field potentials (Fig. 4F).

Discussion
Here, we present an initial characterization of transgenic zebrafish with hair-cell specific ChR2 expression. We successfully evoked stimulus-dependent spiking in afferent neurons with optical activation of transgenic hair-cells using flashes of ,470-nm light. Additionally, we evoked escape responses with optical stimulation of transgenic hair-cells in free-swimming zebrafish larvae. ChR2 expression in hair cells of the ear and lateral line provide an opportunity for investigation of hair-cell sensory encoding and hair cell stimulus-induced activation of the escape response.
First spike latencies are more variable during optical stimulation Utilizing our myo6b:ChR2 transgenic line, we found that flashes of ,470-nm light evoked trains of spiking in afferent neurons that were remarkably similar to the spiking evoked by mechanical stimuli. Unsurprisingly, the total number of spikes in response to each mode of stimulation was significantly different. One logical explanation for the increase in spikes during optical activation is that the overexpression of ChR2 throughout the hair-cell membrane, combined with a high intensity flash of light resulted in an increased amount of depolarization compared to the mechanical stimulus. The greater depolarization of the hair-cell receptor potential would result in increased neurotransmitter release and subsequent spikes in the afferent neuron. Interestingly, despite the increased number of spikes, mechanical stimulation still resulted in similar first spike latencies.
First spike latency is important for encoding the onset of sound [32,34,36]. We observed that the mean first spike latency in response to either repeated mechanical or optical stimuli was not different. However, the variability of the first spike latencies during optical stimulation was significantly greater. For mechanical stimulation, a fast, temporally precise change in receptor potential is provided in part by the large conductance of the MET channels, and their localization to the tips of stereocilia, which ensures their simultaneous activation during deflection of the hair bundle [39,40,42,43]. In contrast to this localized, large depolarizing current source, we observed ChR2 channels, via YFP fluorescence, along the entire plasma membrane of the hair cell. Thus, a flash of ,470-nm light most likely results in stochastic activation of ChR2 channels along the entire cell membrane. This widespread, random activation of a large number of relatively small conductance ion channels could explain the increased variability of first spike latencies.
The increased variability during optical stimulation results from an increase in both faster and slower first spike latencies, which may explain the lack of difference between the mean first spike latency for the two modes of stimulation. The increase in temporal jitter of first spikes during ChR2 activation further highlights the importance of MET-channel activation for the temporal fidelity of spike encoding [3,[6][7][8]. We also observed a difference in the rate of spike adaptation between the two modes of stimulation (see Fig. 2B). ChR2 channels are known to adapt during prolonged stimulation [44]. In addition, there are many sources of adaptation described within the hair cell, including mechanisms involving MET channels [8,45]. We are currently investigating these mechanisms of hair-cell adaptation.
The temporal fidelity of phase-locked spiking is reduced during optical stimulation Given the increased variability of first spike latencies that we observed, we predicted that this would result in decreased vector strength during sustained stimulation. Therefore, we delivered optical stimuli at 10, 20, and 40 Hz, which did result in a significant reduction in the fidelity of phase-locked spiking. Presumably, this occurred for many of the same reasons as for why the first spike latencies were more variable with ChR2 activation. We note that newly optimized versions of ChR2 may be capable of higher fidelity phase locking [46]. For instance, ChETA is capable of following stimuli up to 200 Hz [44]. Altogether, our findings provide evidence that the direct mechanical gating of MET channels plays a role in the faithful encoding of sensory stimuli.

Escape responses and M-cell activity are evoked by optical activation of hair cells
In teleost fish, the M-cell receives sensory information from the eighth cranial nerve as well as from the lateral line [24,25,47,48]. Here, we show that optical stimulation of hair cells with flashes of ,470-nm light evoked an escape response in transgenic larvae. Furthermore, reducing the duration of optical stimuli reduced the number of observed escape responses. We also found that recorded field potentials consistent with M-cell activity were lost concomitantly with the loss of the escape response. This result provides strong support that hair cell stimuli evoke escape responses through activation of the M-cell. Our experiments were based on activation of both ear and lateral-line hair cells. While afferent activity from zebrafish auditory hair cells is known to evoke M-cell responses [22,23,49], a role in the larval zebrafish is less well described [50]. The precise activation of lateral-line hair cells using the myo6b:ChR2 transgenic line will allow for examination of the lateral line contribution to the escape response. Furthermore, the ability to control hair-cell activity remotely in free-swimming larvae provides many opportunities for physiological and behavioral studies in zebrafish.

Conclusion
Here, we demonstrate the potential for using optogenetics in the study of sensory hair cells and their role in eliciting an escape response. We described transgenic zebrafish with hair-cell specific expression of ChR2. We drove hair-cell activity remotely and recorded encoded spikes from afferent neurons of the lateral line. We compared phase-locked spiking between remote optical and direct mechanical stimulation. We found that temporal precision of phase locking, which is essential to sensory encoding, was lost when the receptor potential was depolarized through ChR2 activation instead of through MET channels. Our findings provide in vivo support for the role of the MET channel in the temporally precise encoding of sensory stimuli into spikes in afferent neurons. We also demonstrate how remote activation of ChR2 in hair cells can elicit behavioral escape responses in free-swimming fish. The optical transparency of the zebrafish larva, combined with the power of optophysiology, provide a robust and reliable way to observe and activate hair cells in both the ear and lateral line. Altogether, hair-cell specific ChR2 expression will be a useful tool for research on hair cells and neuronal circuits that involve encoded information from hair cells.

Supporting Information
Video S1 Wild-type larval responses to 100-ms flashes of ,470-nm light. Three flashes of ,470-nm light (100 ms each) are delivered to a wild-type larva (5 dpf