|Home | About | Journals | Submit | Contact Us | Français|
Auditory receptor cells rely on mechanically gated channels to transform sound stimuli into neural activity. Several TRP channels have been implicated in Drosophila auditory transduction, but mechanistic studies have been hampered by the inability to record subthreshold signals from receptor neurons. Here, we develop a non-invasive method for measuring these signals by recording from a central neuron that is electrically coupled to a genetically defined population of auditory receptor cells. We find that the TRPN family member NompC, which is necessary for the active amplification of sound-evoked motion by the auditory organ, is not required for transduction in auditory receptor cells. Instead, NompC sensitizes the transduction complex to movement and precisely regulates the static forces on the complex. In contrast, the TRPV channels Nanchung and Inactive are required for responses to sound, suggesting they are components of the transduction complex. Thus, transduction and active amplification are genetically separable processes in Drosophila hearing.
Mechanosensation is fundamental to all living organisms. However, the molecular identity of the channels that convert force into electrical current has been largely a matter of conjecture. Moreover, the molecular and cellular mechanisms that modulate the forces acting on these mechanosensitive channels are also poorly understood.
Studies in Drosophila melanogaster have made important contributions to our understanding of mechanosensation. In particular, a genetic screen in Drosophila identified the first member of the transient receptor potential (TRP) family to be implicated in mechanosensation (Robert and Hoy, 2007; Walker et al., 2000). That TRP channel—dubbed NompC or TRPN1—is thought to be a component of the transduction complex that converts mechanical force into an electrical signal in Drosophila auditory receptor neurons (Effertz et al., 2012; Effertz et al., 2011; Göpfert et al., 2006; Kamikouchi et al., 2009; Lee et al., 2010; Liang et al., 2011). Auditory receptor neurons in Drosophila are termed Johnston’s organ neurons (JONs), and are housed in the antenna. Sound stimuli cause the distal segment of the antenna to rotate on its long axis, and this rotation transmits forces into the more proximal portion of the antenna, just as rotating a key transmits force to a lock. This stretches JON dendrites, opening mechanosensitive channels (Göpfert and Robert, 2002; Göpfert and Robert, 2001; Kernan, 2007).
Multiple lines of evidence support the idea that NompC has a key role in mechanotransduction. Loss of the C. elegans homolog eliminates force-gated receptor currents in mechanosensitive cephalic neurons, and amino acid substitutions in the putative pore domain of the C. elegans channel can alter the ionic sensitivity of receptor currents (Kang et al., 2010). In Drosophila larvae, loss of NompC eliminates calcium signals in multidendritic mechanosensory neurons in the body wall during crawling (Cheng et al., 2010). In adult Drosophila, loss of NompC reduces sound-evoked electrical activity in the antennal nerve (Eberl et al., 2000; Effertz et al., 2012; Effertz et al., 2011), as well as evoked potentials in mechanosensitive bristles (Walker et al., 2000).
NompC has a particularly interesting role in the mechanics of the Drosophila antenna. Normally, motile elements in the auditory organ expend energy in order to augment sound-evoked motion (Göpfert et al., 2005). This process is called active amplification. Loss of NompC abolishes active amplification in the Drosophila antenna (Göpfert et al., 2006; Göpfert and Robert, 2003). Active amplification also exists in vertebrate hair cells, and a component of active amplification is linked to the gating of hair cell mechanotransduction channels (Hudspeth, 2008). By analogy with hair cells, active amplification in Drosophila has been proposed to depend directly on transduction channel gating (Nadrowski et al., 2008). NompC has been proposed to play a direct role in transduction chiefly because it is required for sound-evoked active amplification (Göpfert et al., 2006) and is also required for the normal mechanical compliance of the antenna in response to a force step (Effertz et al., 2012). However, loss of NompC does not entirely eliminate sound-evoked field potentials in the Drosophila auditory nerve (Eberl et al., 2000; Effertz et al., 2011, 2012), leading to the speculation that another gene might play a redundant function.
Two additional Drosophila TRP channels—Nanchung and Inactive—are also expressed in auditory receptor neurons (Gong et al., 2004; Kim et al., 2003), and likely function as a heteromer (Gong et al., 2004). These TRPV family members are not thought to be part of the transduction complex, because they localize to a subcellular region that is several microns away from the region occupied by NompC (Cheng et al., 2010; Gong et al., 2004; Lee et al., 2010; Liang et al., 2011). Nevertheless, both Nanchung and Inactive are required for sound-evoked field potentials in the antennal nerve, which houses the axons of JONs (Gong et al., 2004; Kim et al., 2003). These potentials are thought to reflect mainly spike-mediated currents in JONs. Thus, it has been proposed that Nanchung and Inactive are required to amplify subthreshold electrical signals generated by the transduction complex, thereby producing signals large enough to elicit spikes in JONs (Göpfert et al., 2006; Kamikouchi et al., 2009; Lee et al., 2010).
That said, it is not clear how Nanchung/Inactive might amplify a signal generated by the transduction complex. Amplification by second messengers is unlikely because these processes are much slower than the auditory transduction latency (Albert et al., 2007; Eberl et al., 2000). Electrical amplification also seems unlikely, as Nanchung and Inactive form channels in heterologous cells that are only weakly voltage-dependent (Gong et al., 2004; Kim et al., 2003).
A primary difficulty in resolving the roles of the TRP channels implicated in Drosophila auditory transduction has been the fact that recordings from individual auditory receptor neurons are not feasible. This is because JONs are very small cells embedded in a delicate antennal organ whose integrity is critical to their function. Thus, we lack any electrophysiological measure finer than field potential recordings from the auditory nerve.
Finally, the problem is compounded by the fact that the field lacks a consensus regarding what stimuli fall within the dynamic range of the Drosophila auditory system. On the one hand, active amplification of antennal motion can be observed in response to relatively weak sound stimuli (as low as 26 dB SVL; Göpfert et al., 2006). If active amplification is the hallmark of transduction, then Drosophila auditory sensitivity might rival that of humans. On the other hand, behavioral measures of auditory sensitivity suggest that Drosophila have a comparatively high threshold for hearing, variously reported as 92 dB (von Schilcher, 1976) or ~72 dB (Eberl et al., 1997; Inagaki et al., 2010).
In this study, we aimed to clarify these issues in three ways. First, we developed a novel behavioral assay to measure the sensitivity of Drosophila hearing, thereby establishing an upper bound for the most sensitive neural threshold that must exist among JONs. Second, we developed a non-invasive method for monitoring sound-evoked subthreshold signals in JONs. Third, using this recording method, together with genetic manipulations of transduction and spiking in JONs, we assessed the relative roles of TRP family members in specifying the sensitivity of auditory transduction. Our results show that Nanchung and Inactive are required for sound-evoked subthreshold signals in JONs. By contrast, NompC is not required for mechanotransduction, and indeed transduction can reach normal peak levels in the absence of NompC. Rather, our results imply that NompC modulates the forces that gate the transduction complex.
Prior electrophysiological, mechanical, and behavioral measures have led to different impressions of the sensitivity of the Drosophila auditory system (Eberl et al., 1997; Göpfert et al., 2006; Inagaki et al., 2010; Kernan, 2007). Therefore, we began by asking what sound intensities elicit a behavioral response. Behavioral measurements are important because they set an upper bound on the neural threshold.
Almost all studies to date have measured behavioral thresholds in the context of courtship, under conditions where it is difficult to precisely control the intensity of stimulation. We reasoned that a simple acoustic startle reflex might yield lower estimates of the threshold. We tethered flies and suspended them above a small plastic ball floating on a cushion of air (Figure 1A). The fly’s fictive running was measured by optically monitoring the movement of the ball. Calibrated sound stimuli were delivered from a speaker in front of the fly. In this apparatus, the flies tended to run spontaneously, alternating with brief bouts of standing still. In response to tone pips, the fly tended to transiently stop their forward running (Figure 1B).
We observed startle behavior in response to sounds with an intensity as low as 1.2 × 10−4 m/s, or 65 dB SVL (Figure 1C). This threshold is lower than that estimated previously using courtship behaviors (Eberl et al., 1997; Inagaki et al., 2010; Kernan, 2007) and is similar to that recently reported using a conditioned proboscis response reflex (Menda et al., 2011). This result means that the most sensitive JONs must have thresholds at or below this intensity. It also demonstrates that these intensities are behaviorally relevant.
We verified that startle behavior was abolished when we stabilized the most distal antennal segment with a drop of glue (Figure 1D). It was also attenuated when we suppressed spiking in JONs by selective RNAi-mediated knockdown of voltage-dependent sodium channels (Nagel and Wilson, 2011) under the control of a JON-specific Gal4 line (Figure 1E). Thus, the startle behavior requires sound-evoked spiking in JONs.
As an initial measurement of neural thresholds, we made field potential recordings from the antennal nerve. Sounds elicited field potential oscillations at twice the stimulus frequency (Figure 1F), as previously reported (Eberl et al., 2000). For the 300 Hz tone, 5.7 × 10−5 m/s (58 dB SVL) was the lowest intensity that elicited a response significantly above the response to background noise (Figure 1G; p < 0.05, t test, n = 6). As expected, the neural threshold is lower than the behavioral threshold.
We also used laser Doppler vibrometry to measure the sound-evoked rotational movement of the antenna. In agreement with previous studies (Göpfert et al., 2006; Göpfert and Robert, 2003), we observed a nonlinearity in the antenna’s movement as sound intensity increased. Specifically, antennal rotations (normalized to sound intensity) were largest for low-intensity sounds, and became smaller for high-intensity sounds (Figure 1G). This phenomenon is consistent with active amplification of movements produced by weak sounds (Göpfert et al., 2005). It is notable that active amplification is observable for intensities below the threshold for antennal field potential responses (Figure 1G). This suggests that active amplification may be a process distinct from transduction, rather than being a hallmark of transduction, and motivates the need for a sensitive measure of JON activity.
Attempts to record directly from individual JONs were unsuccessful due to the fact that these are small cells embedded in a delicate auditory organ. We therefore developed a method for recording signals noninvasively from JONs, with the ultimate goal of recording the signals that give rise to action potentials. We reasoned that we might be able to achieve this by recording from the giant fiber neuron (GFN), a single identifiable central neuron that extends dendrites into the region of the brain where JON axons terminate (Figure 2A; Kamikouchi et al., 2009). A recent study has shown that the GFN responds to auditory stimuli (Tootoonian et al., 2012). What distinguishes the GFN from other central auditory neurons is the finding that dye loaded into JONs can diffuse directly into the GFN, implying that it is coupled to the JON by gap junctions (Strausfeld and Bassemir, 1983). Consistent with this, electron microscopy has shown that JON axons form gap junctions with cells in the vicinity of the GFN dendrites (Sivan-Loukianova and Eberl, 2005). Thus, we made in vivo whole-cell patch-clamp recordings from the GFN to ask whether it receives direct electrical input from JONs via gap junctions. We made these recordings in voltage-clamp configuration to minimize cable filtering by the GFN dendrite, and to minimize the contribution of active conductances in the GFN. To target our electrodes to the GFN, we used specific Gal4 lines to drive GFP expression in this neuron.
In the absence of sound stimuli, we observed hundreds of spontaneous excitatory events in the GFN (Figure 2B) every second. Events that were well-isolated in time had a stereotyped profile within a fly and across flies, and were very fast (<1 ms half-width; Figure 2C). Pure tone stimuli caused excitatory currents to arrive in oscillatory bursts at twice the sound frequency (Figure 2B). This is similar to the frequency doubling observed in the antennal nerve field potential. When we prevented the distal antennal segment from rotating by fixing it with a drop of glue, we observed that spontaneous events persisted, but the response to sound was abolished (Figures 2B). Removing the antennae eliminated both spontaneous events and sound responses (Figures 2B). These results imply that spontaneous events arise in antennal neurons and—because they are modulated by sound—likely originate in JONs. The speed and stereotypy of these events suggest that they represent action potentials in JONs which then propagate into the GFN via gap junctions. (Note that, whereas we are voltage-clamping the GFN, we are unlikely to be voltage-clamping JONs across these gap junctions. This means that action potentials can arise in JONs and propagate across the gap junctions.)
We used pharmacological and genetic manipulations to verify that these events are JON spikes which propagate across gap junctions. We confirmed that blocking chemical synaptic transmission with bath application of Cd2+ had no effect (Figures 2D and 2E), although this manipulation blocks chemical synaptic transmission in the Drosophila olfactory system (Kazama and Wilson, 2008). We also confirmed that spontaneous events in the GFN were abolished by blocking spikes throughout the brain with bath application of tetrodotoxin (TTX). Similarly, events were virtually eliminated by RNAi-mediated knockdown of voltage-gated sodium channels selectively in JONs (Figures 2D and 2E). Finally, events were abolished by a null mutation in the gap junction subunit shakB (Figures 2D and 2E; Curtin et al., 2002; Phelan et al., 1996).
Together, these findings are strong evidence that events are individual JON spikes, rather than synaptic events. These results also demonstrate that the events propagate into the GFN via electrical synapses. Consistent with the conclusion that these synapses are electrical, there is a delay of <300 µs from JON spiking to the onset of currents in the GFN (Figure 2F).
Next we asked whether we could use GFN recordings as a way to monitor the subthreshold signals in JONs that give rise to spikes. To block spikes, we bath-applied TTX, which reduced sound-evoked currents to about 5% of their original level (Figure 3A). In experiments where we selectively knocked down voltage-gated sodium channels in JONs, we observed sound-evoked currents similar to those recorded in wild-type flies with TTX in the bath (Figures 3B and 3C). This argues that the effect of TTX on the sound-evoked currents is due to the blockade of spiking JONs. The currents recorded in TTX thus reflect the subthreshold depolarization in JONs that normally gives rise to JON spikes. The subthreshold depolarization propagates through gap junctions into the GFN, where it gives rise to currents in our voltage-clamp recording. We will use the term “generator currents” to refer to the currents we record in the GFN in the presence of TTX.
Both spike-mediated currents and generator currents were sensitive to weak sound intensities (Figures 3D and 3E). Notably, whereas the spike-mediated currents declined at high intensities, the generator currents showed a smooth monotonic dependence on sound intensity. This indicates that the decline in the spike-mediated currents is due to spike rate adaptation, and not adaptation in transduction. Also, whereas spike-mediated currents were selective for the frequency of the sound stimulus (with higher frequencies producing smaller responses), the generator currents were less so. This suggests that some of the frequency selectivity in spike-mediated currents is due to an inability to generate spikes efficiently at high pitches, again probably due to spike rate adaptation.
Next, we asked how transduction depends on antennal rotation. We measured rotations in response to these sound stimuli using laser Doppler vibrometry (see Figure S1 available online). We used these measurements to plot generator currents as a function of antennal rotation (Figure 3F). These plots show that different sound stimuli generated the same monotonic curve, regardless of frequency. This indicates that the apparent frequency selectivity of the generator currents is due to the frequency selectivity of the antenna, which has a resonant frequency around 160–300 Hz at low sound intensities (Göpfert and Robert, 2002, 2003). When we combined data from these two types of measurements to construct a current-rotation curve, it becomes clear that there is a single relationship between transduction and antennal movement. In the remainder of this study, we will focus on how TRP channels specify this relationship.
It has been proposed that Nanchung and Inactive amplify subthreshold transduction currents to the level of spike initiation (Göpfert et al., 2006; Kamikouchi et al., 2009; Lee et al., 2010). If so, then we should be able to measure generator currents in nanchung and inactive mutant flies. Contrary to this prediction, we found that both spike-mediated sound responses and sound-evoked generator currents were completely absent in null mutants of either gene (nan36a and iav1) (Figures 4A and 4B). The rate of spontaneous events was drastically reduced in both mutants, but events still had a normal size and shape (Figure 4A). This result suggests that these TRPV channels are required for a resting conductance that drives spontaneous JON spiking, but it also demonstrates that neither TRPV is required for JON spikes per se.
We observed no sound-evoked generator current in either mutant at any sound intensity in our test set. The meaningfulness of this observation depends critically on the sensitivity of our measurement, so we examined the recorded currents in the frequency domain where we expect signal detection to be optimal. The frequency representation shows a prominent peak at twice the frequency of the sound stimulus in wild-type recordings, but there is no corresponding peak in recordings from the TRPV mutants at this frequency (Figure 4C). Focusing on a narrow band around this frequency, we calculated the signal gain over background noise for the currents recorded in TTX. On average, the signal gain was >110-fold in wild-type, and indistinguishable from zero in both mutants (Figure 4D). If Nanchung and Inactive serve to amplify the transduction signal, then they would need to amplify that signal at least 110-fold to escape detection. If NompC were an essential component of the transducer, one might imagine that these phenotypes could arise if NompC were trafficked improperly in these mutants; however, we confirmed that NompC localizes correctly even in the absence of Nanchung (Figure 4E).
JONs were initially subdivided into types based on the observation that groups of JONs project to different brain regions (Kamikouchi et al., 2006). Calcium imaging studies have subsequently shown that type AB JONs have a lower threshold for sound stimuli than type CE JONs (Effertz et al., 2011; Kamikouchi et al., 2009). A calcium imaging study has also reported that NompC is absolutely required for sound responses in type AB JONs, whereas NompC is dispensable for sound responses in CE JONs (Effertz et al., 2011). Although the available evidence suggests that all JONs express NompC, Nanchung, and Inactive (Cheng et al., 2010; Gong et al., 2004; Lee et al., 2010), it remains possible that these TRPs might play different roles in different JON types (Effertz et al., 2011, 2012; Kamikouchi et al., 2009).
Given these considerations, we sought to clarify which JON types give rise to the signal that we record in the GFN. First, we filled the GFN with a biocytin marker in flies where distinct classes of JON axons were labeled with GFP. The GFN dendrite is likely to directly contact some JONs, given the short latency of the GFN response to sound stimuli (Figure 2F). Indeed, we observed apparent contacts between the GFN dendrite and type AB JONs, but no contacts for type CE JONs (Figure 5A). These results confirm an earlier study showing that the GFN dendrites arborize in the region where type AB axons terminate (Kamikouchi et al., 2009).
We next tested whether the GFN is functionally connected solely to type AB JONs, or whether type CE JONs also provide input to the GFN. This could be the case if an indirect connection existed between type CE JONs and the GFN. We created flies where just one of the two types of JONs is functional, by virtue of cell-specific rescue of inactive in an inactive mutant background. As a positive control, we confirmed that rescuing inactive expression in most or all JONs was able to rescue the mutant phenotype in GFN recordings (Figures 5B and 5C). When we rescued inactive selectively in type AB JONs, we also observed complete rescue, and these recordings were indistinguishable from wild-type or pan-JON rescue (Figures 5B and 5C). By contrast, rescuing inactive selectively in type CE JONs had no effect, equivalent to flies where the Gal4 driver was omitted (Figures 5B and 5C). Thus, the signals we record in the GFN arise exclusively in type AB JONs.
These results also place an upper bound on the number of JONs providing input to the GFN. The Gal4 line we used to rescue type AB JONs is expressed in a total of 145 neurons in each JO (Inagaki et al., 2010). Because this line produced complete rescue, our recorded signals arise from this number of JONs, or a subset thereof.
We next examined generator currents in a mutant that lacks functional NompC (specifically, nompC3/nompC1 trans-heterozygotes; Walker et al., 2000). Sound stimuli still evoked generator current in the nompC mutant, meaning that transduction is still present in the type AB JONs that provide input to the GFN. However, responses were systematically smaller than normal (Figure 6A). Sound-induced antennal rotations were also smaller in nompC mutants at some of these particle velocities, due to a loss of active amplification (Göpfert et al., 2006), so we controlled for this by measuring sound-evoked antennal movements in wild-type flies and mutants, and plotting the sound response data relative to antennal rotations. This showed that currents in nompC mutants are smaller even if we control for the size of antennal rotations (Figure 6B), and this was consistent across a range of sound frequencies (data not shown).This implies that the loss of NompC reduces the sensitivity of the transduction complex to antennal rotation.
In nompC mutants, the maximal amplitude of the sound-evoked currents that we recorded was lower than wild-type levels (Figure 6B). However, nompC mutant responses to sound stimuli did not saturate, and so the true maximum amplitude is not clear from these recordings. Therefore, we extended our observations by measuring generator currents while using a rigid, piezoelectrically-actuated probe to rotate the distal antennal segment. In this recording configuration, step rotations produce short-latency currents which decay over tens of milliseconds; the rapid adaptation to the static step is further evidence that the GFN is postsynaptic to the fast-adapting AB JONs and not the slow-adapting CE JONs (Kamikouchi et al., 2009). Step rotations of the antenna produce currents which increase monotonically with step amplitude and saturate for the largest steps (Figures 6C and 6D).
In the nompC mutants, responses to small steps were systematically weaker as compared to wild-type (Figures 6C and 6D). For small steps, the rise time was also slower in the nompC mutant than in wild-type. In effect, mutant responses are similar to wild-type responses to smaller steps, suggesting that the transduction complex is experiencing less force than normal. Importantly, however, the rise time and amplitude of the currents evoked by the largest steps were essentially identical in both genotypes (Figures 6C–6E). These results imply that NompC does not alter properties inherent to the transduction channels. Rather, NompC is required for normal sensitivity of the transduction complex to antennal movement.
In general, the sensitivity of transduction relies on the existence of adaptation. During a sustained displacement, adaptation shifts the operating range of the system so that it maintains maximal sensitivity (Albert et al., 2007; Fettiplace and Ricci, 2003). We therefore wondered if adaptation was normal in the nompC mutant. To investigate this, we again used the piezoelectric probe to apply steps of various amplitudes. Steps in either the medial or lateral direction produced generator currents (Figures 7A and 7B). This is probably because type AB JONs are stretched by both medial and lateral steps (Figure S2).
To measure adaptation, we applied test steps either from a starting point corresponding to the antenna’s resting position, or from a new starting point that was offset from the resting position. We chose an offset step that was large enough to evoke a nearly-saturating generator current (Figure 7A). We observed that, within <30 ms after the offset step, currents had regained sensitivity to small rotations in both directions (Figures 7A and 7B). In other words, the region of maximal sensitivity to rotation was recentered around the new static position (Figure 7B). This implies that the lateral and medial resting forces on the transduction complex have been re-equalized.
In nompC mutants, we observed the same process. Although the overall sensitivity of the system was lower in the mutants, the region of maximal sensitivity still migrated by an amount that equal to the magnitude and direction of the static offset (Figures 7C and 7D). These findings demonstrate that NompC is not required for transducer adaptation.
In addition to the loss of sensitivity in the nompC mutant, we noted another striking phenotype: mutant responses lack the bilateral symmetry of wild-type responses. In wild-type recordings, responses to medial and lateral steps were similar in magnitude and kinetics, and both the onset and offset of a step elicited a response (Figure 8A). In nompC mutants, responses to medial step rotations less than 10−2 radians were systematically smaller than responses to small lateral steps (Figures 8A and 8B).
We saw similar results in response to sound stimuli. During each cycle of a sound stimulus, the antenna rotates both medially and laterally, and so type AB JONs are likely stretched twice per cycle, once medially and once laterally (Figure S2). Consistent with this, sound normally elicits oscillations in the generator current at twice the sound frequency (Figures 8C and 8D). By contrast, in nompC mutants, sound elicited oscillations predominantly at the sound frequency, rather than twice this frequency (Figures 8C and 8D).
To see how a small medial-lateral asymmetry can produce a large effect on the dominant frequency of sound responses, it is useful to consider a simple simulation. In this simulation, the relationship between transduction current and rotation is given by a pair of curves —one for medial movement, and the other for lateral movement (Figure 8E). In our simulation, we took the shape of these curves from fits to our data (compare with Figure 7B). We computed the transduction currents at each moment in time as the amount of current specified by these curves, given a sinusoidal stimulus. In JONs, the voltage response to a sound must be a low-pass filtered version of the transduction currents, because of the capacitance of the cell membrane. (Recall that JONs themselves are not voltage-clamped in our recordings; the currents we record are due to voltages propagating through gap junctions into the GFN.) We therefore low-pass filtered the simulated currents.
Our simulations showed that, when the curves are symmetrical around the resting position of the antenna, currents oscillate at twice the frequency of the sound stimulus (Figure 8E). By contrast, when the curves are shifted away from the resting position of the antenna, the dominant frequency of the output signal drops by half, provided that the shift is sufficiently large (Figure 8F). A relatively small shift in the curves (compared to their dynamic range) is required for this behavior. This small shift in the simulated mutant curves mirrors the asymmetry in our mutant data (Figure 7D).
We found that shifting the simulated curves—and also decreasing their steepness—recapitulates the key features of the nompC mutant sound responses (Figure 8F). Namely, sound-evoked responses are reduced, the dominant frequency of phasic oscillations drops by half, and the tonic component of the response shrinks relative to the phasic component. In addition, the level of resting current increases, in agreement with the increased rate of spontaneous spiking in the nompC mutant data (Figure S3).
In this study, we showed that relatively low-intensity sounds (i.e., lower-intensity than previously used to study courtship behavior) can elicit a behavioral response in Drosophila. This provides a motivation for investigating Drosophila auditory transduction near absolute threshold and in particular the mechanisms that specify the sensitivity of the transduction complex. This in turn requires developing a sensitive method for measuring transduction currents from type AB JONs, the receptor neurons that are most sensitive to sound (Kamikouchi et al., 2009; Yorozu et al., 2009). Our anatomical and genetic data demonstrate that GFN currents are a selective measure of spiking and generator currents in type AB JONs.
Although this approach involves recording JON activity indirectly via the GFN, the currents we record are nevertheless relatively fast. Indeed, they have latencies and rise times that are similar to (and even faster than) currents that are recorded directly from the cell bodies of mechanosensitive neurons (e.g., Geffeney et al., 2011). Thus, although the signals we record are likely smoothed by cable filtering, the degree of filtering is not necessarily larger than in the case where signals are recorded directly from mechanosensitive neurons. We could observe generator currents in the GFN in response to the smallest step stimulus we used, and this stimulus is essentially identical to the threshold stimulus for evoking calcium responses in JONs (Effertz et al., 2011). The threshold for evoking GFN currents was also essentially the same as the threshold for evoking an antennal nerve field potential response. Finally, these thresholds are just below the threshold for Drosophila auditory behavior. Taken together, these comparisons argue that our approach is sensitive enough to report generator currents evoked by near-threshold auditory stimuli.
Our results confirm and extend what is known about the fundamental properties of transduction in Drosophila JONs. First, our measurements show that the transduction complex in type AB JONs is gated by antennal rotations as small as 5 × 10−4 radians. This rotation corresponds to a 74 nm displacement of the distal end of the “lever”; (the arista) which projects from the most distal segment of the antenna (see Supplemental Experimental Procedures). This measurement of the transduction threshold is consistent with that obtained by a previous study (Effertz et al., 2011). We should emphasize that the displacement that actually gates the transduction complex is certainly much smaller than this (on the order of a few nm), but because this displacement occurs within the interior of the antenna itself, we cannot measure it directly.
Second, we show that the type AB JONs that provide input to the GFN are depolarized by both lateral and medial rotations. Our data suggest that bidirectionality is probably a property of individual JONs of this type, and not just the population as a whole (Figure S2). Indeed, the geometrical arrangement of type A (and perhaps B) JONs within the auditory organ suggests that individual JONs of this type should be stretched by both medial and lateral movements, and thus should respond twice per sound cycle (Kamikouchi et al., 2006).
Finally, we find evidence that some transduction channels are open at rest, even in the absence of sound. This conclusion relies on our observation that JONs spike spontaneously, and that the rate of spontaneous activity is substantially reduced by loss of either Nanchung or Inactive. This conclusion is consistent with previous studies which used other techniques to make inferences about JON activity (Albert et al., 2007; Kamikouchi et al., 2009).
We have shown that loss of either Nanchung or Inactive abolishes generator currents. Our findings are consistent with previous reports that loss of either Nanchung or Inactive completely eliminates antennal field potential responses to sound (Gong et al., 2004; Kim et al., 2003). However, antennal field potentials are thought to reflect the spiking activity of JONs rather than subthreshold activity (Eberl and Kernan, 2011). Thus, it was not clear from this result whether Nanchung and Inactive were required for transduction or merely spike generation.
Previously, it has been proposed that the role of Nanchung and Inactive is to amplify the transduction signal (Göpfert et al., 2006; Kamikouchi et al., 2009; Lee et al., 2010). However, the latency and speed of the generator currents we record implies that the transduction complex is directly gated by force, rather than gated indirectly by a second messenger. Given this, the Nanchung/Inactive complex is unlikely to merely amplify the transduction signal, because amplification would need to occur within microseconds (which rules out a role for diffusible second messengers), and amplification would need to be >100-fold in magnitude. This level of amplification seems unlikely, given the weak voltage dependence of the channels formed by Nanchung and Inactive (Gong et al., 2004; Kim et al., 2003). Finally, because the Nanchung/Inactive complex does not colocalize with NompC in the JON dendrite (Cheng et al., 2010; Lee et al., 2010; Liang et al., 2011), no amplification mechanism could rely on direct protein-protein interactions between these components.
Given these considerations, it seems more likely that Nanchung and Inactive form part of the transduction complex itself. Consistent with this conclusion, both Nanchung and Inactive confer calcium responses to hypo-osmotic stimuli in heterologous cells (Gong et al., 2004; Kim et al., 2003). However, more work will be needed to test the idea that Nanchung and Inactive could function as force-gated ion channels. An alternative possibility is that Nanchung and Inactive are required for the trafficking or function of an unknown channel.
Previous work has shown that the loss of Nanchung or Inactive results in abnormally large sound-driven antennal movements, as well as spontaneous oscillatory movement in the absence of sound (Göpfert et al., 2006). Our results show that this phenotype goes hand-in-hand with loss of all measurable transduction in JONs (Figure 9). Together, these findings imply that transduction in JONs inhibits the active amplification of antennal movements, possibly because the transduction complex represents a mechanical load on the amplifier element. The presence of active movements in the absence of transduction is also incompatible with the idea that the active amplification of antennal movement is a direct consequence of transduction channel gating.
Our results demonstrate that NompC is not required for mechanotransduction in the type AB JONs that provide input to the GFN. Moreover, the maximal level of transduction current is essentially normal in the absence of NompC, and the rise time of the current is normal at this maximal level. This result argues that NompC does not specify the intrinsic properties of the transduction channel, such as conductance or ionic selectivity. This result also implies that NompC is not required for the proper trafficking or localization of the transduction complex. These conclusions differ from that of a previous study. That study reported that sound-evoked calcium signals are lost in nompC mutant type AB JONs, and concluded that NompC is absolutely required for transduction in these JONs (Effertz et al., 2011). The basis for this discrepancy is not clear, but is likely related to the differences between calcium imaging and electrophysiological recordings. It is possible that the calcium indicator does not report the entirety of the generator current, but rather a small and slow component that does require NompC (Figure S4).
Our results imply that the principal role of NompC is not to transduce force into an electrical signal, but rather to modulate the forces on the transduction complex. Specifically, we find that generator currents are more sensitive to movement when NompC is present, which implies that NompC effectively amplifies mechanical input to the transduction channel, given a fixed amount of antennal movement. Thus, NompC is likely to generate force, or to be permissive for a process that generates force, within the interior of the antenna.
Previous studies have shown that loss of NompC abolishes active amplification of sound-evoked antennal movement, and also reduces spontaneous oscillatory antennal movement (Göpfert et al., 2006; Göpfert and Robert, 2003; see also Figure S1). Thus, loss of NompC appears to eliminate or occlude a process that exerts force on the antenna. This is broadly consistent with our conclusion that NompC is involved in a process which generates force within the interior of Johnston’s organ. Recent studies have proposed that NompC is part of the transduction channel, or channel gating spring, or is otherwise required for the function of either of these components (Effertz et al., 2012; Göpfert et al., 2006); however, our observation that transduction persists in the absence of NompC is not consistent with these ideas. Rather, we propose that NompC is permissive for the function of a mechanical amplifier operating between the antennal sound receiver and the transducer. In other words, we propose that the force generated within Johnston’s organ is exerted on the transduction apparatus as well as the distal antennal segment.
In addition to amplifying mechanical input to the transduction complex, NompC appears to be required for balancing the medial and lateral resting forces on the transduction complex. In the presence of NompC, JONs are equally sensitive to medial and lateral movements, suggesting that medial and lateral resting forces on the transduction complex are balanced. By contrast, in the absence of NompC, JONs are less sensitive to medial movements than to lateral movements. Our simulations show that this phenotype can result from asymmetrical medial and lateral resting forces on the transduction complex. Thus, a single NompC-dependent process may be responsible for balancing resting forces, as well as actively amplifying stimulus-evoked forces. Adaptation appears to be a separate process, because it does not require NompC.
In sum, we propose that NompC functions in a manner analogous to the role of prestin in the mammalian cochlea (Dallos, 2008). Prestin is expressed by outer hair cells in the cochlea, and is essential for the ability of outer hair cells to mechanically amplify sound-evoked movements of the basilar membrane. In this manner, prestin increases the sensitivity of the transduction apparatus of the inner hair cells to sound stimuli. However, like NompC, prestin is not absolutely required for transduction, and is not colocalized with the transduction apparatus.
On the basis of its subcellular location, NompC is well-positioned to act as a modulator of mechanical forces. Whereas Nanchung/Inactive are localized to the proximal dendrite, NompC is localized to the distal dendrite, closer to the point where the dendrite inserts into the connective structures that link it to the moving segment of the antenna (Cheng et al., 2010; Gong et al., 2004; Lee et al., 2010; Liang et al., 2011). A bundle of microtubules runs longitudinally through the dendrite (Todi et al., 2004), and this could provide a substrate for adjustments of tension that propagate from the distal to the proximal dendrite. We propose that transduction occurs in the proximal dendritic segment (where Nanchung and Inactive are localized), and this would place NompC in series between the moving segment of the antenna and the transduction complex.
How might NompC be involved in modulating mechanical force? One possibility is that NompC itself generates force that adjusts the longitudinal tension within a JON. NompC contains an unusually large number of ankyrin repeats (Walker et al., 2000). Ankyrin repeats can act as elastic elements, and can generate a refolding force when unfolded (Serquera et al., 2010; Sotomayor et al., 2005). If, for instance, calcium entry into JONs were to modulate the energetics of the unfolded state on a cycle-by-cycle basis, then the refolding force could augment transduction. An alternative possibility is that NompC does not itself generate force, but it is permissive for a process that generates force. For example, calcium influx through NompC might change the state of motor proteins that adjust longitudinal tension within a JON.
Assuming that NompC forms part of a channel, this channel appears to carry relatively little current, or is otherwise ineffective at exciting the JON. We found no detectable generator current in the absence of either Nanchung or Inactive, meaning that any current must be below the limit imposed by noise in our recording. That limit is about 100-fold smaller than the generator currents we measure. Moreover, a previous study reported that sound-evoked calcium signals in JONs are essentially eliminated when Nanchung is absent (Kamikouchi et al., 2009). Together, these findings argue that any ionic flux through NompC is far less than the flux through the transduction complex itself. This conclusion relies on the idea that NompC can still function when Nanchung is absent. In support of this, we have shown that NompC localizes properly in the absence of Nanchung. Moreover, active amplification of antennal movements is intact when Nanchung is absent (Göpfert et al., 2006; see also Figure 9). Because the active amplification of sound-evoked movements requires NompC, this implies that NompC can function without Nanchung. Interestingly, we observe a slow current that persists for hundreds of milliseconds after sound offset, and which absolutely requires both Nanchung and NompC (Figure S4).
Future studies will be required to fully elucidate the mechanism of NompC’s action. What makes this mechanism intriguing is the implication there may be two functionally distinct types of TRP channels involved in Drosophila hearing (Figure 9). One of these (the transduction channel) evidently carries most or all of the current, and requires Nanchung and Inactive. The other—which requires NompC—carries comparatively little current, and controls the active generation of force within the auditory organ.
Procedures are summarized below; see Supplemental Experimental Procedures for details on all sections.
Fly stocks and genetic manipulations were as follows:
Sound particle velocities were measured using a calibrated pressure gradient microphone. All electrophysiological, behavioral, and laser Doppler vibrometry recordings were made in a sound isolation chamber which reduced background noise to 23 dB SPL (unweighted). The duration of the tone was 250 ms, and the first and last 10% of the stimulus was cosine theta squared ramped. The speaker was ~230 mm from the fly.
The fly was glued to a tether that was positioned above a ball floating on a cushion of air. The fly’s locomotion was recorded by measuring the motion of the ball using an optical mouse sensor below the ball.
Field potential recordings were performed using a saline-filled quartz electrode. The electrode was inserted between the first and second antennal segments.
Sound-driven antennal movements were measured using a laser Doppler vibrometer with the laser spot focused on the most distal branch point of the lever-like structure (the arista) which protrudes from the most distal antennal segment (Figure S1). The laser beam was positioned orthogonal to the plane of the arista, and so our measurements quantified the displacement of the arista along this axis. Measurements of aristal displacement were converted into measurements of aristal rotation by measuring the distance from the laser spot on the arista to the midline of the distal (third) antennal segment, and then taking the small angle approximation. The arista is rigidly coupled to the distal antennal segment, and so aristal and antennal rotation are the same. We report rotation (rather than antennal displacement) because this measure should not depend on the position of the laser measurement spot on the arista.
Currents were recorded from the Giant Fiber Neuron (GFN) in vivo in whole-cell voltage-clamp mode under visual control on an upright compound microscope. The fly was suspended in a piece of titanium foil such that the upper side of the fly’s head was bathed in oxygenated saline, while the lower side of the head and both antennae remained dry. The GFN was identified based on GFP expression under the control of the G0117-Gal4 or G0066-Gal4 lines.
The second antennal segment was glued to the titanium foil, leaving only the distal (third) antennal segment free to rotate. A piezoelectric stack was used to rotate the third antennal segment via a tungsten probe attached to the arista. Laser Doppler vibrometry was used to measure the displacement of the piezoelectric stack. These measurements showed that the rise time of step stimuli (from 10% of maximum to 90% of maximum) was 300–400 µs. The tip of the tungsten probe was placed on the distal-most branch point of the arista, the same location targeted in the laser Doppler vibrometry measurements (Figure S1). The measured displacements of the probe (and thus the arista) were converted to rotations by measuring the length of the arista (Figure S1).
We thank John Assad, David Corey, Barry Dickson, Tom Middendorf, Daniel Robert, Bernardo Sabatini, Gary Yellen, and members of the Wilson lab for helpful comments throughout the project. Elizabeth Hong provided expert advice. The following individuals kindly supplied us with flies (see Supplemental Experimental Procedures for original citations): “Dickinson wild-caught”; stock (Michael Dickinson), iav1 and UAS-iav (Craig Montell), UAS-dicer2 (Heather Broihier), JO-AB-Gal4 and JO-CE-Gal4 (David Anderson), nompC3, nompC1, and UAS-nompC-L:GFP (Yuh-Nung Jan), nan-Gal4 (Larry Zipursky), and shakB2 (Robert Wyman). This work was supported by a Program Grant from the Human Frontiers Science Program. R.I.W. is an HHMI Early Career Scientist. B.P.L. and A.E.B. were supported by National Science Foundation Graduate Research Fellowships. Additional support to B.P.L. was provided by the Sault Ste. Marie Tribe of Chippewa Indians Self-sufficiency Fund. B.P.L. and R.I.W. designed the experiments, except for the behavioral experiments, which were designed by A.E.B. and R.I.W. B.P.L. performed the experiments and analyzed the data, except for Figures 1A–1D, where the experiments and analyses were performed by A.E.B. with assistance from Q.G. B.P.L. and R.I.W. wrote the paper. A.-S.C. provided the G0117-Gal4 and G0066-Gal4 lines.
Supplemental Information includes four figures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi.org/10.1016/j.neuron.2012.11.030.