Abstract
Many auditory, vestibular, and lateral-line afferent neurons display spontaneous action potentials. This spontaneous spiking is thought to result from hair-cell glutamate release in the absence of stimuli. Spontaneous release at hair-cell resting potentials presumably results from CaV1.3 L-type calcium channel activity. Here, using intact zebrafish larvae, we recorded robust spontaneous spiking from lateral-line afferent neurons in the absence of external stimuli. Consistent with the above assumptions, spiking was absent in mutants that lacked either Vesicular glutamate transporter 3 (Vglut3) or CaV1.3. We then tested the hypothesis that spontaneous spiking resulted from sustained CaV1.3 activity due to depolarizing currents that are active at rest. Mechanotransduction currents (IMET) provide a depolarizing influence to the resting potential. However, following block of IMET, spontaneous spiking persisted and was characterized by longer interspike intervals and increased periods of inactivity. These results suggest that an additional depolarizing influence maintains the resting potential within the activation range of CaV1.3. To test whether the hyperpolarization-activated cation current, Ih participates in setting the resting potential, we applied Ih antagonists. Both ZD7288 and DK-AH 269 reduced spontaneous activity. Finally, concomitant block of IMET and Ih essentially abolished spontaneous activity, ostensibly by hyperpolarization outside of the activation range for CaV1.3. Together, our data support a mechanism for spontaneous spiking that results from Ca2+-dependent neurotransmitter release at hair-cell resting potentials that are maintained within the activation range of CaV1.3 channels through active IMET and Ih.
Introduction
Spontaneous action potentials (spikes) in sensory neurons are critical for information processing (Douglass et al., 1993; Lestienne, 2001; Yu et al., 2004). In both auditory and vestibular systems, spontaneous spiking of afferent neurons is required for sensitivity and frequency selectivity (Kiang et al., 1965; Manley and Robertson, 1976; Köppl, 1997; Jones and Jones, 2000). Spontaneous spiking is not intrinsic to the afferent neuron, but rather the result of Ca2+-dependent glutamate neurotransmission from hair cells (Moser and Beutner, 2000; Keen and Hudspeth, 2006; Li et al., 2009). Furthermore, recent studies are consistent with Ca2+-dependent vesicle fusion requiring the L-type calcium channel CaV1.3 (Spassova et al., 2001; Robertson and Paki, 2002; Brandt et al., 2003; Sueta et al., 2004; Nemzou N et al., 2006). Spontaneous activity is predicted to result from resting membrane potentials that are within the activation range for CaV1.3 (Jørgensen and Kroese, 2005). However, the contribution to spontaneous afferent spiking by depolarizing hair-cell currents active at rest has not been elucidated.
In the absence of external stimuli, mechanotransduction (MET) channels have a resting level of activation (Corey and Hudspeth, 1983; Ohmori, 1987; Hudspeth and Lewis, 1988). At rest, this nonzero open probability allows the hair-cell receptor potential to report both positive (depolarizing) and negative (hyperpolarizing) deflections of stereocilia (Roberts et al., 1988). In addition, the resting MET activation results in depolarization of the resting membrane potential (Farris et al., 2006). Therefore, one would predict that the magnitude of the resting MET current (IMET) will directly influence CaV1.3 activity and spontaneous neurotransmitter release.
Two other depolarizing current sources that influence the hair-cell resting potential are the inward rectifier currents Ih and IKir (Holt and Eatock, 1995; Marcotti et al., 1999). The hyperpolarization-activated potassium-sodium current, Ih, is composed of both K+ and Na+ cations that result in an Ih reversal potential that is more positive than that for strictly K+-selective currents, such as IKir. Functionally, mixed selectivity coincides with hair-cell resting potentials that are more depolarized than non-Ih-expressing hair cells (Sugihara and Furukawa, 1989; Holt and Eatock, 1995). In contrast, IKir is associated with hyperpolarizing resting potentials and is often absent from cells that express Ih currents (Holt and Eatock, 1995; Goodman and Art, 1996; Sugihara and Furukawa, 1996; Marcotti et al., 1999; Jørgensen and Kroese, 2005). Thus, Ih expression may depolarize the resting potential within a range suitable for sustained CaV1.3 activity.
Here, we examined the contribution of depolarizing current sources to spontaneous spiking of lateral-line afferent neurons in intact zebrafish larvae. Collectively, our data suggest that spontaneous spiking of lateral-line afferent neurons results from CaV1.3 activity at hair-cell resting potentials that are depolarized by both IMET and Ih. Functionally, the presence of Ih when combined with a resting IMET appears to prevent the resting potential from hyperpolarizing below the activation range for CaV1.3 channels.
Materials and Methods
Fish strains and reagents.
Zebrafish larvae, Danio rerio, were raised in E3 embryo media at 30°C. The cdh23tc317e allele causes a premature truncation of Cdh23 at the 15th cadherin repeat and has been described by Nicolson et al. (1998) and Söllner et al. (2004). The cav1.3aR1250X allele has been described by Nicolson et al. (1998) and Sidi et al. (2004). The vglut3484+2T>C allele has been described by Obholzer et al. (2008).
Neuromast RT-PCR.
Neuromasts were isolated from day 5 larvae as previously described (Trapani et al., 2009). Five neuromasts from a single larvae were used to amplify cDNA. We designed primers for the hcn2 and hcn4 genes using submitted sequences (Gene ID 557285 and 557843, respectively). The hcn1 (chromosome 12) and hcn3 (chromosome 25) primers were designed using predictions found on the UCSC Genome Browser on Zebrafish (ENSDART00000113932 and ENSDART00000088249, respectively).
Pharmacology.
For all experiments, drugs were brought to final concentration in normal extracellular solution (in mm: 130 NaCl, 2 KCl, 2 CaCl2, 1 MgCl2, and 10 HEPES, pH 7.8, 290 mOsm) and bath applied to the whole animal. For microphonics experiments, we used the same solution in the bath and both the recording and waterjet pipettes. Amiloride, dihydrostreptomycin (DHS), BAPTA, and EGTA were purchased from Sigma-Aldrich. For all experiments that required intracellular or basolateral drug access, we included 0.1% DMSO. The L-type calcium channel antagonist isradipine; the AMPA receptor antagonist 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX; Ascent Scientific); and the Ih antagonists DK-AH 269 and ZD7288 (Tocris Bioscience) were brought to final concentration in extracellular solution with 0.1% DMSO. For experiments in which we blocked transduction and Ih, we included 0.1% DMSO with DHS and then applied DK-AH 269 with both DHS and 0.1% DMSO.
Electrophysiology and lateral-line afferent recordings.
In an earlier study, we described our recording setup for both microphonics and action currents (Obholzer et al., 2008; Trapani and Nicolson, 2010). Briefly, larvae were anesthetized, mounted, and microinjected in the heart with 125 μm α-bungarotoxin to suppress muscle activity. Larvae were then rinsed and maintained in normal extracellular solution. For extracellular current recordings, borosilicate glass pipettes were pulled (P-97, Sutter Instruments) with a long taper, and resistances were between 5 and 15 MΩ. Microphonic pipettes resembled standard patch pipettes with resistances from 1 to 5 MΩ. Signals were collected with an EPC 10 amplifier and Patchmaster software (Heka Elektronik). Extracellular currents were acquired from an individual lateral-line afferent neuron in the loose-patch configuration (resistances ranged from 20 to 80 MΩ). Recordings were done in voltage-clamp mode, sampled at 50 μs/point, and filtered at 1 kHz. Microphonic potentials were obtained in current-clamp mode and sampled at 100 μs/pt. An additional amplifier (Model 440, Brownlee Precision) further amplified (total 10,000×) and filtered (50 Hz, eight-pole Bessel) the voltage signal. For microphonics, the recording electrode was positioned (MPC-385, Sutter Instruments) adjacent to the neuromast at the height of the hair-cell stereociliary bundles. For extracellular current recordings, the recording electrode was positioned within the posterior lateral-line ganglion against a given cell body. The cell's innervated neuromast was identified by progressively moving the waterjet from neuromast to neuromast until phase-locked spiking was seen.
Mechanical stimulation.
Stimulation of neuromast hair cells was performed as described previously (Trapani et al., 2009). Briefly, we used a pressure clamp (HSPC-1, ALA Scientific) attached to a glass micropipette (tip diameter ∼30 μm) filled with normal extracellular solution to stimulate hair cells. The waterjet pipette was positioned (MP-265, Sutter Instruments) ∼100 μm from a given neuromast, and displacement of the kinocilia was verified by eye. The pressure clamp was driven by a sinusoidal voltage command delivered by the EPC 10 via an analog output. Waterjet pressure was monitored via a feedback sensor located on the HSPC-1 headstage and was collected in Patchmaster alongside the recording.
Electrical stimulation.
For experiments directly stimulating the afferent fibers with a stimulating pipette, we visualized the lateral-line neuron using neuroD:GFP transgenic fish. Stimuli (0.5 ms) were generated using an ISO-Flex stimulus isolator (A.M.P.I.) and delivered via an electrode pipette similar to that used for extracellular current recordings. After establishing an extracellular action current recording with spontaneous spiking, the stimulating pipette was advanced beneath the skin just anterior to primary neuromast 1, and placed against the afferent fiber bundle. Stimulus voltage was then increased until a stimulus-dependent spike was seen in addition to the background, spontaneous spikes.
Signal analysis.
Data were analyzed using custom software written in Igor Pro (WaveMetrics) and were plotted with Prism 5 (GraphPad). Average spike rate (in spikes per second) was computed from the total number of spikes in 60 consecutive 1 s traces. Spontaneous interspike intervals (ISIs) were determined from the concatenation of 400 consecutive 1 s traces without stimulation. Exponential decay equations were fit to the peak of the histogram, which was either the 10–20 ms bin for wild type (WT) or the 20–30 ms bin for IMET- or Ih-blocked histograms. The initial ISI bin (0–10 ms) had fewer events that likely resulted from a combination of the synaptic delay between the hair cell and neuron and the absolute and relative refractory periods of the neuron. For all histograms, both single-phase and two-phase exponential equations were compared and the best-fit equation (p < 0.05) was determined using an extra sum-of-squares F test. For ISI histogram distributions best fit by a single-phase exponential decay, the equation used was y = y0 × exp(−K × x). For two-phase decays, data were fit using y = PhaseFast × exp(−KFast × x) + PhaseSlow × exp(−KSlow × x), where PhaseFast = y0 × ProportionFast and PhaseSlow = y0 × (1 − ProportionFast). Note that ProportionFast was multiplied by 100 and changed to %τfast for the text and figures. Time constants, τ, τfast, and τslow, are reported as the inverse of the rate constants K, KFast, and KSlow. For experiments with DHS application, 400 s of activity 100 s after application was analyzed for the postdrug response. Mean ISI time was calculated as the sum of all ISIs divided by total spike number. Representative microphonic traces were the average of 200 consecutive 500 ms traces. Mean power spectral density (PSD) was calculated from the sum of individual PSDs from 200 traces using the stimulus portion (200 ms) of each trace. Values in the text are expressed as mean ± SEM. Where appropriate, data for statistical significance tests were confirmed for normality using Kolmogorov–Smirnov normality tests and for equal variances using F tests comparing variances. Statistical significance between two conditions was determined by either paired or unpaired, two-tailed Student's t tests, as appropriate.
Results
Lateral-line neurons display phase-locked spikes in response to deflection of neuromast hair cells. Recently, we observed that neurons in the zebrafish afferent lateral-line organ return to spontaneous activity within seconds after stimulation (Trapani et al., 2009). In immature systems, including the mammalian auditory organ, afferent neurons fire spontaneous bursts of spikes as the result of spontaneous Ca2+ spikes in immature hair cells (Jones et al., 2001; Tritsch et al., 2007; Sonntag et al., 2009; Tritsch and Bergles, 2010). Mature afferent neurons, however, display stochastic spontaneous spiking in the absence of external stimuli, suggesting a Poisson process of neurotransmitter release from hair cells (Hoagland, 1933; Kiang et al., 1965; Manley and Robertson, 1976; Sonntag et al., 2009). Our experiments were done on day 5 and 6 zebrafish larvae, which display behaviors indicative of mature auditory, vestibular, and lateral-line organs (Kimmel et al., 1974; Sidi et al., 2003; Zeddies and Fay, 2005; Obholzer et al., 2008). Therefore, we monitored action currents to determine the nature of the spontaneous spiking seen in lateral-line afferent neurons.
Stochastic spontaneous spiking of lateral-line afferent neurons
Shown in Figure 1A is a representative 60 s trace of spontaneous spiking from a lateral-line afferent neuron. If spontaneous spiking occurs from a stochastic process, then the probability for generation of a spike is independent of any previous spikes. Typically for neurons, this behavior means that the intervals between successive spikes will be Poisson distributed—that is, ISIs of increasing lengths occur with exponentially decreasing probability. The resulting ISI histograms will therefore be best fit by an exponential decay equation. For a population of 10 recordings from lateral-line afferent neurons, we found that 8 of the ISI histograms were best fit by two-phase exponential decay equations (p < 0.0001) (mean values in Table 1), while the remaining two were best fit by single-phase exponentials (p < 0.0001) (Fig. 1B,C). We did not observe bursting, rhythmic, or spike patterns other than stochastic during 30 min of continuous recording of spontaneous activity (supplemental Fig. 1, available at www.jneurosci.org as supplemental material).
Spontaneous spiking of lateral-line neurons in zebrafish larvae. A, Single 60 s trace with 525 spikes (mean ISI = 113 ms) from an extracellular recording of a single WT lateral-line neuron (vertical calibration: 50 pA). B, ISI histogram from the neuron in A. Data are from 400 s of activity (mean ISI = 117 ms; 3416 spikes) binned in 10 ms intervals and best fit with a two-phase exponential decay (black line; τfast = 55 ms; τslow = 123 ms; %τfast = 39). Inset, ISI histogram with 100 ms bins. C, Time constants from two recordings fit by single-phase exponential decay equations (τ; white circles; n = 2) and eight recordings that were best fit by two-phase exponential decay equations (τfast = white circles; τslow = gray circles; n = 8). Green circles represent the recording in B. Mean ± SEM for τfast and τslow values with %τfast are shown in Table 1. Plus symbol denotes a τslow value with a weight of just 8%τslow.
Values for spontaneous spiking under different conditions
Each lateral-line neuromast contains two populations of hair cells (5–7 cells in each) with opposite hair-bundle polarity and directional sensitivity (Flock and Wersall, 1962). Thus, bidirectional stimulation of a neuromast produces microphonic potentials with a characteristic 2f frequency component that results from alternating activation of the two hair-cell populations (see below, Acute block of IMET eliminated all transduction current). In addition, the neuromast is innervated by two afferent neurons, each of which makes synaptic contacts with all hair cells of the same hair-bundle polarity (Nagiel et al., 2008; Faucherre et al., 2009). Thus, the variability of mean ISI times may be due in part to the variability in number of innervated hair cells and synaptic inputs for a given afferent neuron. Furthermore, that ISI histograms were best fit by exponential equations with two time constants suggests that more than one stochastic process occurs. Given the complexity of our in vivo system, we cannot know the exact nature of the processes involved. However, one potential process may result from instability of the hair-cell resting potential, which is predicted to affect the activation of CaV1.3 channels (Jørgensen and Kroese, 2005). We therefore examined the contribution to spontaneous activity of CaV1.3 channels and currents implicated in setting hair-cell resting potentials using genetic and pharmacological approaches.
Hair-cell microphonics and postsynaptic action currents in zebrafish mutant lines
Similar to auditory and vestibular afferent neurons, spontaneous spiking in the lateral-line organ is thought to be generated by hair-cell transmission (Zimmerman, 1979; Sewell, 1990; Bailey and Sewell, 2000; Dawkins and Sewell, 2004). Here, we took advantage of two zebrafish lines with mutations in either the CaV1.3 calcium channel or the Vesicular glutamate transporter 3 (Vglut3) and assayed for their requirement for spontaneous lateral-line activity.
The vglut3 and cav1.3a mutant zebrafish lines have severe defects in hair-cell synaptic transmission (Nicolson et al., 1998; Sidi et al., 2004; Obholzer et al., 2008). In wild-type zebrafish, both in situ hybridization data and immunohistochemistry indicate that vglut3 and cav1.3a are expressed in hair cells and not in afferent fibers (Sidi et al., 2004; Obholzer et al., 2008). In addition, CaV1.3 channels are clustered at hair-cell synapses (Sidi et al., 2004; Brandt et al., 2005), and Vglut3 is present throughout the basal half of hair cells (Obholzer et al., 2008). For this study, we used the alleles vglut3484+2T>C and cav1.3aR1250X, which result in premature truncation of their respective protein products and are presumably functional nulls.
As previously reported, microphonic potentials were normal in vglut3484+2T>C mutants and were reduced but still present in cav1.3aR1250X mutants (Fig. 2) (n = 4) (Nicolson et al., 1998; Obholzer et al., 2008). However, lateral-line recordings revealed that spontaneous activity was absent in both mutants (Fig. 2) (n = 11 and 7, respectively). During the course of our lateral-line recordings with the cav1.3aR1250X mutant, we noted that an occasional spike (<10 spikes per 1000 s of recording) occurred in 5 of 7 recordings (supplemental Fig. 2A, available at www.jneurosci.org as supplemental material); however, we never saw spikes in recordings from vglut3484+2T>C mutants (Obholzer et al., 2008). In addition, we confirmed the requirement for CaV1.3 channels with the L-type Ca2+ channel antagonist, isradipine, which essentially blocked spontaneous spiking, reducing it to a pattern similar to cav1.3aR1250X mutants (supplemental Fig. 2B, available at www.jneurosci.org as supplemental material) (n = 4). This finding argues that the vesicle machinery was still operational in cav1.3aR1250X mutants despite a lack of functional CaV1.3. Furthermore, the selective AMPA receptor antagonist NBQX abolished spontaneous spiking (supplemental Fig. 2C, available at www.jneurosci.org as supplemental material) (n = 3). Altogether, these results support a model in which spontaneous spiking of lateral-line afferent neurons is generated by Ca2+-dependent glutamate transmission from hair cells.
Spontaneous activity was absent in transmission mutants. Representative microphonics and lateral-line recordings from WT and three hair cell mutants are shown. Average microphonic potentials (left panel; 20 Hz stimulus) and single traces of evoked spikes (middle panel; 20 Hz) and spontaneous spikes (right panel) in WT (top traces; vertical calibration: 40 pA), vglut3484+2T>C (upper middle traces), cav1.3aR1250X (lower middle traces), and cdh23tc317e (bottom traces; vertical calibration: 20 pA) are shown. n.a., Not applicable; due to a lack of transduction in cdh23tc317e mutants, it is not possible to obtain evoked spiking.
The above results with synaptic transmission mutants illustrated a requirement of hair-cell input for spontaneous spiking of the afferent neuron. We next assayed the contribution of the hair-cell resting IMET to spontaneous activity using two zebrafish mutants that lack mechanotransduction. Mutation of either of the zebrafish genes cadherin 23 (cdh23) or protocadherin 15 (pcdh15) results in disruption of vital dye labeling via permeation through MET channels (Seiler and Nicolson, 1999; Söllner et al., 2004; Seiler et al., 2005). In addition, tip links, which are a necessary component of the transduction apparatus (Assad et al., 1991) are absent in cdh23 mutants (Söllner et al., 2004). As described previously, microphonic potentials were completely absent in mutants with either the cdh23tc317e (see Figs. 2, 4A) (n = 4) (Nicolson et al., 1998) or pcdh15t288 allele (Nicolson et al., 1998). Despite the absence of IMET, we observed spontaneous spikes in recordings from both cdh23tc317e and pcdh15t288 lateral-line neurons. As seen in wild-type larvae, the selective AMPA receptor antagonist NBQX abolished spontaneous spiking in the transduction mutants (data not shown).
The spontaneous spiking of lateral-line neurons in the transduction mutants revealed increased periods of inactivity (Fig. 3A). We recorded 400 s of spiking from cdh23tc317e mutants, determined their mean ISI time (mean values in Table 1) (n = 5), and generated ISI histogram distributions. All histograms were best fit (p < 0.0001) by two-phase exponential decay equations (Fig. 3B,C; mean values in Table 1). We hypothesized that the increase in magnitude of the τslow component resulted from loss of the steady depolarizing current through MET channels and consequent hyperpolarization of the resting potential away from the activation voltage of CaV1.3. Thus, the τslow may now reflect oscillation of the resting potential into and out of the activation voltage for CaV1.3 channels. Several lines of evidence are consistent with this hypothesis. First, visual inspection of all recordings revealed that the long ISI times fit by the τslow component were intermixed with faster ISI times. Second, the shorter-length ISI times fit by the τfast component were not significantly different from wild-type τfast (Fig. 3C, Table 1) (p = 0.18). Third, the longer-length ISI times described by the τslow were significantly different from the wild-type τslow (p < 0.005). And finally, the pcdh15t288 mutant showed a similar phenotype of interspike intervals (mean values in Table 1) (n = 4).
Spontaneous spiking was reduced in lateral-line neurons of the cdh23tc317e transduction mutant. A, Single 60 s trace of spontaneous activity (mean ISI = 395 ms; 152 spikes; vertical calibration: 30 pA) from a cdh23tc317e mutant larva. B, ISI histogram from 400 s of activity (mean ISI = 352 ms; 1135 spikes) best fit by a two-phase exponential decay (black line; τfast = 67 ms; τslow = 530 ms; %τfast = 90). Black arrowhead highlights the increased number of long duration ISI intervals. Inset, ISI histogram with 100 ms bins. C, Plot of exponential time constants from ISI histograms for all recordings (brown circles from B) of spontaneous activity from cdh23tc317e mutants (gray circles; n = 5). Wild-type data are from the 8 recordings in Figure 1 with two-phase exponential fits (white circles). Mean values with SEM are shown in Table 1. D, Box plot of ISIs from WT and cdh23+c317e mutant larvae. The median ISI time is represented by the horizontal line within the box; the whiskers represent minimum to maximum.
Acute block of IMET eliminated all transduction current
Because indirect, developmental effects of cdh23 or pcdh15 mutations on currents and other hair-cell properties cannot be ruled out, we examined spontaneous spiking in wild-type larvae during acute block of IMET. DHS is one of several well characterized compounds known to rapidly block MET channels (Jørgensen and Ohmori, 1988; Kroese et al., 1989; Ricci, 2002; Beurg et al., 2009). We quantified the extent of IMET block by measuring the reduction of microphonic potentials following bath application of DHS. Within seconds of exposure to 2 mm DHS, neuromast microphonic potentials were eliminated (Fig. 4A). Averaged power spectra from analysis of the stimulus portion of individual traces revealed an expected loss at both the 20 Hz and 40 Hz (2f) frequency components (Fig. 4B) (n = 4).
Acute block of MET channel by 2 mm DHS. A, Representative trace of average microphonics before (middle; control) or 60 s after application of 2 mm DHS (bottom) in response to 20 Hz stimulus (top). B, Power spectra for control (black; n = 4), 2 mm DHS-treated (red; n = 4), and cdh23tc317e mutant (brown; n = 3) larvae. Note the characteristic 2f frequency component representing the neuromast's two direction-selective populations of hair cells. Error represented by shaded area above and below colored line. Power of averaged noise traces is indicated by dashed black line. C, A single recording with each trace showing an overlay of 10 consecutive sweeps (vertical calibration: 30 pA). Top, Twenty hertz stimulus trace; middle, control spikes (198 spikes); bottom, following 2 mm DHS application (35 spikes). D, Period histogram for all spikes from control (black bars) and following application of 2 mm DHS (red bars) from sixty 1 s sweeps of 20 Hz stimulation (from the cell in C). E, Average vector strength (gray circles; red and black circles are from the cell in C and D) for spikes during control and following 2 mm DHS application. The mean 20 Hz vector strength (error bar, SEM) for spontaneous activity is depicted by the white bar (data from Fig. 1).
Despite the apparent loss of microphonic potentials, it was possible that there was slightly less than 100% block and that a small amount of IMET remained. We predicted that if there was any remaining MET conductance, spiking in response to mechanical stimulation would display some degree of phase locking. We stimulated neuromast hair cells at 20 Hz before and during application of 2 mm DHS (Fig. 4C). We then quantified the degree of phase locking by calculating the vector strength (r) of 60 s of activity in both conditions. Vector strength is a measure of the synchrony between response and stimulus, where a value of 1 represents perfect synchrony and 0 represents no relationship (Goldberg and Brown, 1969). While control vector strength indicated a strong degree of phase locking (0.81 ± 0.04; n = 4), the average vector strength in 2 mm DHS (0.09 ± 0.02; n = 4) was not significantly different from spontaneous activity analyzed for vector strength at 20 Hz (r = 0.14 ± 0.02; n = 10; p = 0.28) (data from Fig. 1). In addition, following DHS application, the average ISI times during 20 Hz stimulation (370 ± 100 ms; n = 4) coincided with the mean ISIs for cdh23tc317e mutants and for spontaneous activity recorded in 2 mm DHS (Fig. 5, Table 1).
Spontaneous activity was reduced during acute block of mechanotransduction with 2 mm DHS. A, Single 60 s trace (mean ISI = 81 ms; 735 spikes; vertical calibration: 20 pA) and corresponding ISI histogram (4908 spikes) from 400 s of continuous spiking (mean ISI = 81 ms; τfast = 13 ms; τslow = 87 ms; %τfast = 64) before application of DHS (Pre-DHS). B, Single 60 s trace (mean ISI = 209 ms; 282 spikes) and ISI histogram from the same recording in A after application of 2 mm DHS (Post-DHS). For Post-DHS, the ISI histogram from 400 s of spontaneous spiking (mean ISI = 226 ms; 1325 spikes) was best fit by a two-phase exponential decay (τfast = 40 ms; τslow = 325 ms; %τfast = 89). Inset, ISI histogram with 100 ms bins. C, Diary plot of spike rate (each symbol represents a 10 s bin) before and after application of 2 mm DHS (gray bar; n = 7). D, Comparison of time constants from pre-DHS histograms best fit by single-phase exponentials (white circles) to post-DHS (gray circles) best fit by two-phase exponentials (n = 3). E, Comparison of time constants from pre-DHS ISI histograms best fit by two-phase exponentials (white circles) to post-DHS (gray circles, red circles from B; n = 4). F, Box plot of ISIs from pre-DHS (white bar) and post-DHS (gray bar) from all recordings (n = 7). The median ISI time is represented by the horizontal line inside the box; the whiskers represent minimum to maximum; the mean ISI for post-DHS treatment is listed in Table 1.
Spontaneous spiking was reduced during acute block of IMET
To examine spontaneous spiking during acute block of IMET, we recorded 400 s of spontaneous spiking from wild-type larvae and then bath applied DHS to a final concentration of 2 mm (Fig. 5A,B). The ISI histograms before DHS application were best fit by either single- (Fig. 5D) (n = 3) or two- (Fig. 5E) (n = 3) phase exponential decays (p < 0.0001). However, during IMET block by DHS, all ISI histograms were now best fit by two-phase exponential equations (p < 0.0001) (Fig. 5B,D,E; mean values in Table 1). Similar to the cdh23tc317e mutants whose τfast values were not different from wild-type values, the DHS-τfast values were not significantly different from pre-DHS values (p = 0.25). Furthermore, the DHS-τslow values were similar to the cdh23tc317e mutant values for τslow (p = 0.87). We noted that in the presence of DHS, spiking persisted for as long as recordings were maintained and was unaffected by addition of 0.1% DMSO (data not shown). Spontaneous spiking was similarly affected by the MET antagonist amiloride (mean ISI = 170 ± 13 ms; τfast = 62 ± 24 ms; τslow = 422 ± 191 ms; %τfast = 83 ± 12; n = 4); however, spiking was lost over time, and incubation with 0.1% DMSO resulted in rapid loss of spiking within seconds of application. This finding may reflect the ability of amiloride to block CaV1.3 channels (Garcia et al., 1990), an effect that would be markedly accelerated by the DMSO application. Collectively, our results are consistent with both MET channel activity at rest and a role in spontaneous postsynaptic activity observed in vivo.
Spontaneous spiking was reduced during block of Ih
The above results from wild-type larvae were consistent with resting potentials that are maintained within the activation range for CaV1.3 channels. We also observed that in the absence of the depolarizing IMET, spiking persisted. This finding suggests that an additional depolarizing current is active at rest. Previous reports have shown that expression of the hyperpolarization-activated cation current, Ih, contributes to depolarized resting potentials in certain hair cells (Sugihara and Furukawa, 1989; Holt and Eatock, 1995; Biel et al., 2009). Therefore, we postulated that at rest, Ih currents depolarize the hair-cell resting potential in the zebrafish lateral-line organ.
The channels that mediate Ih currents are members of the hyperpolarization-activated cyclic nucleotide-gated channel family (HCN1-4) (Santoro and Tibbs, 1999). A recent study showed that HCN1 immunolabel was localized to the basolateral end of mouse inner ear hair cells (Horwitz et al., 2010). To confirm HCN expression in lateral-line neuromasts, we performed RT-PCR on isolated neuromasts. We amplified a product corresponding to HCN1, but were unable to amplify HCN2, HCN3, or HCN4 (supplemental Fig. 3D, available at www.jneurosci.org as supplemental material). To study HCN1 function in hair cells, genetic disruption of HCN1 would be ideal; however, mutations in zebrafish HCN1 have not been reported. Furthermore, knockdown of HCN1 using morpholinos would not be specific to hair cells, nor complete at the larval stage (day 5) used in our experiments. Therefore, we turned to a pharmacological approach.
We examined the role of Ih in spontaneous activity using the HCN antagonist, ZD7288. Because ZD7288 is known to have nonspecific effects on synaptic transmission (Chevaleyre and Castillo, 2002), we also performed our experiments with a second HCN blocker, DK-AH 269. After 5 min in the presence of either 1 mm ZD7288 or 1 mm DK-AH 269, spontaneous spiking displayed increased periods of inactivity (Fig. 6A). Similar to spontaneous spiking during block of IMET, following block of Ih all ISI histograms were best fit by two-phase exponential decay equations (p < 0.0001) (Fig. 6B–E; mean values in Table 1). As with the loss of IMET, this finding suggests that suppression of the depolarizing influence of Ih led to hyperpolarization of the resting potential.
Spontaneous activity was markedly reduced by Ih antagonists. A, Top, Single 60 s trace of spontaneous activity before drug (mean ISI = 62 ms; 967 spikes; vertical calibration: 40 pA). Bottom, Single 60 s trace of spontaneous activity 300 s after application of 1 mm ZD7288 to the same recording above (mean ISI = 344 ms; 178 spikes). B, ISI histogram from 400 s of activity (mean ISI = 805 ms; 519 spikes) in 1 mm ZD7288 best fit by a two-phase exponential decay (τfast = 33 ms; τslow = 409 ms; %τfast = 94). Inset, ISI histogram of same data with 100 ms bins. C, Comparison of time constants from a pre-ZD7288 histogram best fit by a single-phase exponential (white diamond) compared to post-ZD7288 (gray diamonds) best fit by two-phase exponentials (n = 1). Circles represent pre- and post-ZD7288 histograms best fit by two-phase exponentials (white and gray, respectively; data from B in blue; n = 3). D, Top, Single 60 s trace of spontaneous activity pre-drug (mean ISI = 77 ms; 830 spikes; vertical calibration: 20 pA). Bottom, Sixty-second trace of spontaneous activity 300 s after application of 1 mm DK-AH 269 (mean ISI = 192 ms; 316 spikes). E, ISI histogram from 400 s of activity (mean ISI = 805 ms; 519 spikes) in 1 mm DK-AH 269 best fit by a two-phase exponential decay (τfast = 34 ms; τslow = 197 ms; %τfast = 75). Inset, ISI histogram with 100 ms bins. F, Comparison of time constants from pre-DK-AH 269 histograms fit by single-phase exponentials (white diamonds) and post-DK-AH 269 best fit by two-phase exponentials (gray diamonds; n = 2). Circles represent pre- and post-DK-AH 269 histograms best fit by two-phase exponentials (white and gray, respectively; data from E in orange; n = 2). Plus symbols indicate two pre-DK-AH 269 recordings that had apparently large τslow but with weights of 3 and 7%τslow, compared to the same recording post-DK-AH 269, which now had weights of 17 and 25%τslow.
A recently described postsynaptic role for Ih (Yi et al., 2010) suggests that the Ih antagonists likely had a postsynaptic effect in our system. Furthermore, in addition to blocking HCN channels, Ih antagonists are known to block the MET channel (Farris et al., 2004; Horwitz et al., 2010). Therefore, we characterized the Ih antagonists in several ways. First, we determined whether ZD7288 affected general excitability of the lateral-line neuron. We used a stimulating electrode to directly evoke action currents from afferent fibers in a transgenic larva (neuroD:GFP). While we observed that ZD7288 did not affect the stimulus-dependent action potential threshold (supplemental Fig. 3A–C, available at www.jneurosci.org as supplemental material), we cannot account for the effect of ZD7288 on HCN channels located at dendritic boutons (Yi et al., 2010). Second, we measured microphonic potentials and found that a 1 mm concentration of either drug blocked transduction by >90% (supplemental Fig. 3D,E, available at www.jneurosci.org as supplemental material). Thus, during application of Ih antagonists, postsynaptic HCN channels, Ih, and a portion of IMET were blocked.
Combined block of IMET and Ih suppressed lateral-line spontaneous spiking
We predicted that the spontaneous spiking we recorded in the presence of Ih antagonists resulted from the remaining IMET, and that complete block of both IMET and Ih would abolish all spontaneous activity. Therefore, we simultaneously blocked both IMET and Ih using two different approaches. First, we recorded from larvae where transduction was first blocked with 2 mm DHS for 400 s and then we coapplied 1 mm DK-AH 269 (Fig. 7A). We characterized the time course of reduction in spike rate by calculating the average spike rate for every 100 consecutive spikes (Fig. 7B). In the presence of both drugs, the average spontaneous ISI time was 7053 ± 1912 ms (Fig. 7C, Table 1) (n = 4). At this spike rate, we were unable to collect enough data to generate ISI histograms (see bottom panel of Fig. 7A). The second approach took advantage of chelating external calcium with BAPTA, which disrupts transduction by breaking the stereociliary tip links necessary for mechanical gating of the MET channel (Assad et al., 1991; Zhao et al., 1996). Following a 5 min preincubation with 5 mm BAPTA and several wash steps, we observed that microphonics potentials were abolished (supplemental Fig. 3D, available at www.jneurosci.org as supplemental material). We then established a lateral-line recording and applied 1 mm ZD7288, which reduced spontaneous spiking to an average ISI time of 4340 ± 490 ms (Fig. 7C, Table 1) (n = 3). We found that ISI histograms following BAPTA treatment were no longer exponentially distributed and that this phenomenon also occurred after EGTA treatment (data not shown). In both experiments where IMET and Ih were simultaneously blocked, the residual spontaneous spikes continued to appear for the extent of recordings that lasted >30 min. These results argue strongly for a role of Ih in the hair-cell-dependent spontaneous activity of lateral-line afferent neurons.
Spontaneous spiking was reduced by >98% following block of IMET and Ih. A, Top, Continuous 60 s trace of spontaneous spiking before drugs (mean ISI = 46 ms; 1313 spikes; vertical calibration: 40 pA). Middle, During application of 2 mm DHS (mean ISI = 147 ms; 407 spikes). Bottom, Trace of 60 s of continuous spiking 5 min after application of 1 mm DH-AK 269 in the continued presence of 2 mm DHS (mean ISI = 6000 ms; 10 spikes; vertical calibration: 10 pA). B, Time course of the average (100 consecutive spikes per bin) spike rate following application of 2 mm DHS (red bar) and then coapplication of 2 mm DHS and 1 mm DH-AK 269 (red and orange bar, respectively). C, Bar graph representing the average mean-ISI time for all WT (black bar), IMET-blocked (red bar), Ih-blocked (orange and blue bars), and IMET and Ih-blocked conditions (hatched and gray bars) (all values from Table 1).
Discussion
Similar to neurons described in the auditory and vestibular systems of other species, our recordings of spontaneous spiking in zebrafish lateral-line neurons revealed stochastic patterns of activity (Heil et al., 2007; Neubauer et al., 2009). Spontaneous activity has been shown to be dependent on L-type calcium channel activity (Robertson and Paki, 2002). As we reported previously, spontaneous spiking was completely absent in a hair-cell-specific mutant that lacked Vglut3 (Obholzer et al., 2008). Here we have shown that spontaneous spiking requires CaV1.3, indicating that spontaneous transmitter release does not occur without CaV1.3 channels. In contrast, spontaneous spiking still occurred in two mechanotransduction mutants carrying the alleles cdh23tc317e or pcdh15t288. The spiking pattern was, however, distinct from that seen in wild-type larvae. Further characterization of spontaneous activity revealed that spiking was dependent on the depolarizing currents, IMET and Ih. Overall, our results suggest that spontaneous neurotransmission in lateral-line hair cells requires the contribution of these two currents to maintain the resting membrane potential within the activation range of CaV1.3.
Spontaneous activity and the resting potential of hair cells
Our recordings from wild-type afferent neurons in the absence of stimulation were characterized by stochastic activity that was best described by either single-phase or two-phase exponential decays with both components within the range of the mean ISI times seen across all recordings. Due to the innervation of multiple hair cells by a single afferent neuron, and the multiple synaptic contacts made at each hair cell, we cannot determine what is responsible for the variability both in mean ISI times and in the time constants of exponential decay equations describing histograms of the ISI times. It seems plausible that part of the variability may arise from electrical resonance (Roberts et al., 1988; Rutherford and Roberts, 2009). In contrast to Ca2+ channels at conventional synapses, CaV1.3 displays fast activation at negative voltages (Schnee and Ricci, 2003; Grant and Fuchs, 2008; Johnson and Marcotti, 2008; Zampini et al., 2010). We envision that the resting potential oscillates entirely within the voltage range where CaV1.3 channels are active, but that their level of activation varies with voltage. Thus, two time constants in the wild-type data might result from a combination of the Poisson nature of vesicle fusion (Del Castillo and Katz, 1954), the variation in activity of CaV1.3 channels, and the innervation of multiple hair cells. Evidence for electrical resonance in lateral-line hair cells would require stable recordings of membrane potentials, a technique that has not yet been established with zebrafish hair cells.
MET channel activity at rest contributed to spontaneous spiking
MET channels have a resting level of activity in the absence of direct stimulation that is dictated by calcium-dependent adaptation (Farris et al., 2006). We examined spontaneous afferent spiking under conditions where this depolarizing IMET current was blocked. Either the cdh23tc317e or pcdh15t288 mutations or saturating concentrations of MET channel blockers eliminated microphonic potentials and significantly reduced afferent spiking. When IMET was blocked either acutely or genetically, ISI histograms were always best described by two-phase exponential decay equations (τfast and τslow). Interestingly, the τfast component was similar to control recordings that were best fit by either single-phase τs or the τfast component from two-phase equations. This finding suggests that the ISI times described by the τfast component resulted from a similar process. The reason for the increase in the τslow component is not clear, but we speculate that it arises from hyperpolarization of the membrane potential to a voltage range where the activation of CaV1.3 channels is infrequent, resulting in increased periods of inactivity and in much longer ISI times.
It is possible that indirect effects of the MET antagonists could have changed the nature of the spontaneous spiking. However, we think this is unlikely: first, we repeated the DHS experiments with a second MET antagonist, amiloride, which blocks transduction via a different mechanism (Ricci, 2002), and saw a similar effect on spontaneous spiking. Second, the phenomenon of spontaneous spiking with τfast components similar to wild-type τfast but with significantly larger τslow components was seen in all of the following conditions: cdh23tc317e, pcdh15t288, 2 mm DHS, and 1 mm amiloride.
Ih at rest contributed to spontaneous spiking
Following IMET block, spontaneous spiking was reduced but still present, indicating that there was remaining CaV1.3 channel activity. This result suggested that an additional depolarizing current was present—otherwise one would predict that the resting membrane potential should be near the K+-reversal potential—resulting in resting potentials suitable for the activation of CaV1.3. Morphologically, lateral-line hair cells are similar to type II vestibular hair cells in mammals (Nicolson, 2005), which often express the hyperpolarization-activated current, Ih (Holt and Eatock, 1995; Sugihara and Furukawa, 1996; Brichta et al., 2002). In addition, Ih is known to control excitability, and, due to its activation at resting membrane potentials, contribute to the resting potential in many cell types (Biel et al., 2009). Application of the HCN antagonists ZD7288 or DK-AH 269 resulted in reduced spontaneous afferent spiking with ISI histograms always best fit by two-phase exponential decay equations. One caveat of using HCN antagonists is the potential block of HCN channels in the afferent neuron. The contribution of dendritic HCN channels to afferent spiking has been explored in cochlear afferent fibers (Yi et al., 2010). Application of ZD7288 slowed the decay of EPSPs and resulted in hyperpolarization of the afferent neuron by ∼4 mV, but did not affect the frequency of synaptic events (Yi et al., 2010). Although we did not see a change in excitability of the afferent neuron in the presence of ZD7288, we cannot rule out an effect on dendritic HCN channels.
Comparison of our data from control and either IMET- or IH-blocked conditions provides further evidence that the membrane potential at rest is primarily within the activation range of CaV1.3 channels. Consistent with this idea, we observed that blocking either depolarizing current resulted in 100% of ISI histograms being described by two-phase exponentials (τfast and τslow), with a significant increase in the magnitude of the τslow component. The significant increase of τslow following block of either depolarizing current source argues for the requirement of both currents for maintenance within this range.
We abolished essentially all spontaneous activity in afferent neurons when we coapplied saturating concentrations of both MET and Ih antagonists to wild-type larvae. This finding is consistent with a concomitant depolarizing influence of both IMET and Ih at rest. When both currents are blocked, the resting potential appears to hyperpolarize to outside the activation voltage range of the CaV1.3 channel.
Given the functional requirement for spontaneous activity in many hair-cell types, it seems likely that depolarizing currents that are active at rest would be necessary to ensure that the resting potential is within the activation range of CaV1.3 channels. Altogether, our results support such a mechanism for spontaneous activity in the lateral-line organ. Specifically, MET channels and hyperpolarization activated HCN channels are active at rest and serve to depolarize the resting potential of hair cells. Consequently, at rest, CaV1.3 channel activity then leads to Ca2+-dependent neurotransmitter release and, ultimately, spontaneous spiking of afferent neurons.
Footnotes
-
This study was supported in part by National Institutes of Health Grant R01 DC006880. We thank L. Trussell and K. Bender for detailed discussions throughout this study. We also express gratitude to A. Ricci for insightful comments on the manuscript. We thank W. Mo for assistance with the RT-PCR experiments.
- Correspondence should be addressed to Josef G. Trapani, Howard Hughes Medical Institute, Oregon Hearing Research Center, and Vollum Institute, 3181 SW Sam Jackson Park Road, Oregon Health and Science University, Portland, OR 97239. trapanij{at}ohsu.edu
