| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Zoology and Physiology, University of Wyoming, Laramie, Wyoming
Submitted 10 January 2006; accepted in final form 13 June 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Erulkar et al. 1968; Gordon and O'Neill 1998; Heil et al. 1992a,b; Mendelson and Cynader 1985; Møller 1974; Nelson et al. 1966; O'Neill and Brimijoin 2002; Phillips et al. 1985; Poon et al. 1991; Shore et al. 1987; Vartanian 1974). Echolocating bats have taken such functional specializations a step further in that a significant percentage of auditory neurons responds selectively to FM signals (e.g., Casseday and Covey 1992; Casseday et al. 1997; Fuzessery 1994; Suga 1969; Vater and Schlegel 1979). Their auditory systems thus provide useful models for examining the mechanisms that shape neuronal selectivity for these important signal components.
The present study examines the mechanisms that shape response selectivity for the direction and rate of FM sweeps in the inferior colliculus (IC) of the pallid bat. The IC of this bat has an unusually high degree of response selectivity for the downward FM sweep of the echolocation pulse (Bell 1982; Brown 1976; Fuzessery et al. 1993), with over 50% of neurons responding selectively to the downward sweep direction and nearly 30% of these responding only to downward FM sweeps and not to individual tones within the sweep (Fuzessery 1994; Fuzessery and Hall 1996). The goals of this study are to determine the mechanisms that shape response selectivity for FM sweep direction and rate, to compare these mechanisms with those that create the same forms of selectivity in auditory cortex, and to create a reference for examining the ontogeny of these response properties.
The mechanisms proposed to shape neuronal selectivity for the direction of an FM sweep are similar to those suggested to shape directional selectivity for motion across visual and somatosensory receptor surfaces—that is, that there is an asymmetry in excitation and/or inhibition created in one direction that is not created in the other (Barlow and Levick 1964; Movshon et al. 1978; Reid et al. 1991; Sillito 1977). The asymmetry has been modeled as the result of interactions between networks of neurons at multiple levels of the auditory system (Gordon and O'Neill 1998; Suga 1965, 1973) or as the spatiotemporal integration of inhibitory or excitatory subthreshold postsynaptic events at single neurons (Rall 1964; Segev 1992). Phillips et al. (1985) suggested an excitatory summation that occurs when the tonal response area is approached from a given direction. An inhibitory mechanism often proposed is an asymmetry in the inhibitory sidebands flanking excitatory tuning curves (Britt and Starr 1976; Gordon and O'Neill 1998; Heil et al. 1992a; Shannon-Hartman et al. 1992; Suga 1965). This mechanism has recently received additional support from studies that used either patch clamping to measure the timing of inhibitory and excitatory inputs (Zhang et al. 2003) or two-tone inhibition paradigms designed to reveal the arrival times of these inputs (Gordon and O'Neill 1998). It has also been suggested that excitatory and inhibitory events work in concert to produce a directional asymmetry (Casseday et al. 1997); a neuron may receive sufficient excitatory drive only if excitatory inputs combine with a rebound from an earlier inhibitory input and this coincidence may occur only in one sweep direction.
Some of the mechanisms that shape direction selectivity may also contribute to a selectivity for FM sweep rate because auditory neurons are reported to display direction selectivity only at certain sweep rates (Gordon and O'Neill 1998; Heil and Irvine 1998; Tian and Rauschecher 1994). Gordon and O'Neill (1998) noted that it is important to consider both temporal and spectral differences in sideband inhibition to understand their roles in shaping selectivity for various FM sweep dimensions. A low-frequency inhibitory sideband, for example, would prevent a neuron from responding to an upward sweep if the inhibition arrived earlier than excitation, and thus create direction selectivity. However, if this inhibition were delayed relative to excitation, it would allow the neuron to respond to an upward sweep, but only until the sweep rate slowed to the extent that the delayed inhibition caught up with the excitation. Thus it is the relative timing of excitation and inhibition, as they occur during the course of an FM sweep, that determines the form of the resulting spectrotemporal filter.
The main focus of the present study is to understand the mechanisms that shape selectivity for FM sweep direction and rate in the region of pallid bat IC tuned to the echolocation pulse. We previously described the direction selectivity of this neuronal population (Fuzessery 1994); here we describe fast-pass and band-pass tuning for FM sweep rates that approximate those of the bat's echolocation pulse. A two-tone inhibition paradigm similar to that used by Gordon and O'Neill (1998), Faure et al. (2003), and Brimijoin and O'Neill (2005) is used to estimate the timing of excitatory and inhibitory inputs during the course of an FM sweep. The method was found to be effective in predicting how spectrotemporal asymmetries in low- and high-frequency inhibitions shape selectivity for sweep direction and rate.
An unexpected finding is that essentially identical forms of rate tuning are created by two different mechanisms: duration tuning and a delayed high-frequency inhibition. Over 50% of neurons in the high-frequency region of the pallid bat IC have been found to respond maximally to short-tone durations of 
5 ms (Fuzessery and Hall 1999). This duration tuning is reflected in their sweep rate selectivity; they prefer rates in which FM sweeps traverse their excitatory tuning curves in a time equal to their best duration. Neurons that lack duration tuning derive their rate selectivity for downward sweeps from a high-frequency inhibition that is delayed relative to excitation, imposing a fast-pass rate filter. Auditory systems can create similar spectrotemporal filters through different mechanisms.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Surgical procedures
Bats were prepared for surgery by first anesthetizing them with methoxyflurane (Metofane) inhalation, followed by an intraperitoneal injection of sodium pentobarbital (Nembutal, 30 µg/g body weight) and acepromazine (2 µg/g body weight). To expose the IC, the head was held in a bite bar and a midline incision was made in the scalp; the muscles over the dorsal surface of the skull were reflected. In preparation for securing an aluminum head pin to the skull, the skull over the neocortex was scraped clean and covered with a layer of glass microbeads in cyanoacrylate cement, followed by a layer of dental cement. Using skull and brain surface landmarks (the skull is semitransparent), a small hole (diameter 0.5 mm) was made over the IC.
Recording procedures
Experiments were conducted in a heated (85–90°F), soundproof chamber lined with anechoic foam. Bats were kept lightly anesthetized throughout the course of the experiments with sedation maintained by inhalation of Metofane and additional pentobarbital sodium (one third of presurgical dose) injections. Stimuli were generated using Modular Instruments and Tucker Davis Technologies digital hardware and custom software. The waveforms were amplified with a stereo amplifier and presented as closed-field stimuli through Infinity emit-K ribbon tweeters fitted with funnels that were inserted into the pinnae and sealed with petroleum jelly. This procedure attenuated speaker intensity level at the opposite ear by 
30 dB. The speaker–funnel frequency–response curve showed a gradual increase of 20 dB from 6 to 70 kHz, as measured with a Brüel & Kjær 1/8-in. microphone placed at the tip of the funnel.
Recordings were made with glass microelectrodes filled with 1 M NaCl, with an impedance of 3–7 M
. Electrodes were advanced remotely from outside the recording chamber using a Kopf Model 660 Micropositioner. Electrodes were advanced vertically down into the center one third of the dorsal surface of the IC and recording began when neurons with best frequencies (BFs) of >30 kHz were encountered. Based on the initial positioning of the electrode, it is assumed that all recordings were made in the central nucleus of the IC (ICC). The echolocation pulse of the pallid bat typically sweeps downward from 60 to 30 kHz (maximum range 80–25 kHz; Brown 1976) and previous studies (Fuzessery 1994; Fuzessery and Hall 1996) showed that the great majority of neurons that respond selectively to the echolocation pulse have BFs of 35 kHz. Response magnitudes and poststimulus time histograms were acquired and stored with the use of a Modular Instruments high-speed clock controlled by custom software. A response to a given stimulus was quantified as the total number of spikes elicited over 30 stimulus presentations. Because neurons in the pallid bat IC typically show little or no spontaneous activity, the isolation of neurons depended on responses to search stimuli consisting of tones, and downward and upward FM sweeps presented over a range of durations, from 1 to 10 ms.
Data collection
Sounds were presented at intervals of 400 ms. Envelope rise/fall times were set at 1 ms. When sound durations were <2 ms, the rise/fall times were each 50% of the duration. For each neuron, all sounds were presented at a single intensity 5–20 dB above the response threshold. This strategy allowed us time to conduct most of required stimulus manipulations before contact with the neuron was lost, but did not allow us to determine whether response properties changed with sound pressure level.
Data collection proceeded in the following sequence. If a neuron responded to tones, its excitatory tonal response area was mapped audiovisually as the combinations of sound pressure levels and tone frequencies evoking a response to each of ten consecutive stimulus presentations. Responses to the signals described below were quantified as the total number of spikes elicited by 30 stimulus presentations, collected 0–50 ms after stimulus onset. The great majority of responses were phasic and time-locked to stimulus onset.
Duration selectivity was tested at a neuron's BF. A neuron was considered duration selective if it exhibited a short-pass or band-pass duration function in which responses dropped to 50% of maximum response as tone duration increased or decreased relative to the best duration. The best duration was defined as the arithmetic center of the durations evoking a 80% maximum response. The shortest duration that was always tested was 0.5 ms, but in some neurons, durations as short as 0.1 ms were presented. If a neuron responded maximally to the shortest duration tested, this was assigned as its best duration. Duration tuning was defined only in response to tones and not in response to FM sweeps because response changes with FM sweep duration (with FM spectrum fixed) could be the result of a selectivity for sweep rate rather than sweep duration.
Responses to linear upward and downward FM sweeps of identical spectra were then tested to determine direction selectivity. The spectra of FM sweeps were centered on the neuron's BF. A neuron was termed selective for sweep direction if the maximum response in one direction had a directional selectivity index (DSI) of 0.6 (at which the response to the nonpreferred direction was 25% of the response to the preferred direction). The DSI = (D – U)/(D + U), where D and U are the maximum responses to downward and upward sweeps, respectively, regardless of the sweep rates at which they occurred (Britt and Starr 1976; Heil et al. 1992a; O'Neill and Brimijoin 2002). All direction-selective neurons preferred the downward direction; therefore preference for this direction was assigned a positive value, unlike the convention used in previous studies.
Downward FM–selective neurons were defined as responding similarly to both downward FM sweeps and tones, but giving 20% of maximum response to upward sweeps and noise. Downward FM specialists were more selective. They responded maximally only to downward sweeps and gave 20% to tones, upward sweeps, and noise.
Although the biosonar pulse of the pallid bat is an exponential downward FM sweep, linear sweeps were used in this study to maintain a constant rate of frequency change. This allowed us to more easily relate time and frequency within a sweep, which in turn allowed us to assign a single best sweep rate to rate-selective neurons and to more easily predict rate and direction selectivity (as described in RESULTS).
Band-pass noise with the same bandwidths as those of excitatory FM sweeps were presented as a control to determine whether a neuron actually required the orderly progression of frequencies present in a sweep or whether the simultaneous presentation of a spectrum that encompassed both the excitatory and inhibitory tonal response areas was also excitatory. Noise was presented at durations of 1 to 10 ms or, if the neuron was duration tuned, at the neuron's best duration. The band-pass filter was digitally generated and had a rolloff of >40 dB/octave.
Selectivity for FM sweep rate was then tested by presenting FM sweeps of at least three different bandwidths, centered on the neuron's BF and located, to the extent possible, within the bandwidth of bat's echolocation range of 25–80 kHz. A neuron was considered sweep rate selective if it exhibited a preferred rate when presented with a downward FM sweep that approximated the bat's echolocation pulse of 60 to 30 kHz. Some sweep bandwidths fell outside of this range, as low as 20 kHz, to generate a full range of sweep rates. The rates of each of these sweeps were varied by changing sweep duration. Sweep rates were calculated by dividing the bandwidth by the duration (kHz/ms). The sweep rate functions were then normalized to percentage of maximum response to compare the degree of overlap. If these normalized functions overlapped by within 1 kHz/ms along the slow-rate flank of the rate function peak, the neuron was considered rate selective. The upper limit of rate tuning was not reached in some neurons because, as a result of the decision to keep the sweep spectrum within the bat's biosonar range, the durations of the sweeps became so short (0.1 ms) as the sweep rate was increased that signal distortion became an issue. The best sweep rate was calculated by taking the arithmetic center of 80% maximum response for each rate function and averaging them.
Two-tone inhibition over time
The bandwidths and arrival times of low- and high-frequency inhibitions both play a role in shaping direction and rate selectivity. To estimate both the bandwidth and relative arrival times of inhibition and excitation using extracellular recording, we used a "two-tone inhibition over time" (TTI over time) paradigm in which intensity was held constant, and two tones, one excitatory and one inhibitory, were delayed with respect to one another. An example is shown in Fig. 6.
|
If the onset of the excitatory tone was advanced (negative delay, e.g., Fig. 3) relative to that of the inhibitory tone, and the response could still be suppressed, then inhibitory input was assumed to arrive before excitation. If the excitatory tone had to be delayed before the second tone suppressed the response, then it was assumed that inhibitory input arrived after excitation.
|
The duration of the inhibitory input—more specifically, whether it at least lasted the duration of the inhibitory tone—was determined by delaying the excitatory input until it no longer completely overlapped in time with the inhibitory tone (e.g., Fig. 12). The equation used to calculate whether the inhibitory input persisted for the duration of the inhibitory tone was: excitatory tone delay = (duration of the inhibitory tone) – (duration of the excitatory tone) + (minimum duration required for maximum response). If the result was equal to 0, inhibition persisted for this duration: if the value was <0 ms, inhibition did not last the duration of the inhibitory tone; if >0 ms, it persisted beyond this duration.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Fuzessery and Hall 1996). We exploited this feature to determine the mechanisms that shape response selectivity for the direction and rate of downward FM sweeps. The results are based on the responses of 64 neurons recorded in 20 bats. In all of these neurons, at least rate or direction selectivity was tested. All neurons had BFs within the spectrum of the bat's echolocation pulse (30–80 kHz). The majority had BFs of 35–50 kHz, near the spectral energy peak of the pulse. To determine selectivity for the spectrotemporal dimensions of FM sweeps, neurons were presented with four types of sounds: pure tones, upward and downward FM sweeps, and band-pass noise. Tones were used to define excitatory tuning curves (if present) and to determine whether neurons required a full FM sweep to be excited or whether a single excitatory tone within the sweep would suffice. FM sweeps of identical spectra, but different directions, were used to assess directional selectivity. Band-pass noise with a spectrum identical to that of an excitatory sweep was used to determine whether an orderly sequence of frequencies was required for excitation.
Response selectivity for FM sweep direction
Neurons were considered selective for FM sweep direction if their direction selectivity index (DSI) was 0.6 for the preferred direction, which corresponds to the nonpreferred direction evoking 20% of the response to the preferred direction. Of the 64 neurons, 42 (66%) were selective for the downward sweep direction. None responded selectively to the upward direction. Of the 42 direction-selective neurons, 30 (71%) of these (47% of the recorded population) responded to both downward sweeps and tones; these are termed downward FM (dFM) selective. The remaining 12 of the 42 (29%) direction-selective neurons (19% of the recorded population) responded only to downward FM; tones evoked 20% of the maximum response. These are termed downward FM specialists.
The remaining 22 of 64 recorded neurons (34%) were not selective for FM sweep direction. Tones and upward and downward sweeps all evoked 20% of the maximum response. None of the 64 recorded neurons responded to band-pass noise, indicating that simultaneous presentations of excitatory and inhibitory frequencies in the FM sweep spectra were either inhibitory and/or not excitatory. The percentages of downward FM selective and downward FM specialist neurons found in the present study are similar to those previously reported (53 and 31%, respectively; Fuzessery 1994).
Response selectivity for downward FM sweep rate
Most neurons (35 of 40 neurons tested, 88%) were selective for the rates (kHz/ms) of downward FM sweeps. Rate-tuned neurons are defined as showing a preference for the same sweep rate when presented with at least three different sweeps that varied in bandwidths ranging from 10 to 50 kHz. The rates of these sweeps were varied by changing their durations, from 0.25 to 50 ms. As seen in Fig. 1A, if a neuron is rate tuned, the sweep duration eliciting a maximum response will systematically increase with sweep bandwidth. The duration functions were converted to rate (kHz/ms) by dividing the bandwidth of the sweep (kHz) by the duration (ms) of the sweep (Fig. 1B). A neuron was considered rate tuned if the three rate functions fell within a 1 kHz/ms range along the "slow" flanks of the rate functions. Rate-tuned neurons exhibited either a fast-pass selectivity, in which responses dropped to 50% of maximum response at slower rates, or a band-pass selectivity, in which the responses dropped to 50% of maximum of response at faster and slower nonoptimal rates. In most cases, responses dropped to zero as the sweep rate decreased. There was most likely an upper rate limit for all neurons, but sweep rate was increased by shortening sweep duration. Once FM sweep durations were decreased to 0.25 ms to test responses to the highest rates, the fidelity of speaker output became suspect and thus higher rates were not tested. However, even at these very short FM sweep durations, these neurons were able to discriminate FM sweeps from band-pass noise with the same spectrum because none responded to the noise.
|
5 kHz/ms. The majority of neurons tested that were selective for the downward FM (dFM) sweep direction were also tuned to sweep rate. Of 11 dFM specialists tested for both direction and rate selectivity, 82% (n = 9) were rate tuned. Of 25 dFM-selective neurons tested for both, 76% (n = 19) were rate tuned. Thus of 36 direction-selective neurons tested, 78% were also selective for rate. Among neurons that were not direction selective, a greater percentage were also not rate tuned. Of 19 such neurons, only 42% were rate tuned.
Two different mechanisms create similar rate tuning
The mechanisms underlying rate tuning for downward FM sweeps were tested in 32 of the 40 rate-tuned neurons. Similar rate tuning appears to be created by two different mechanisms. The first is an early on-BF inhibition that creates duration tuning for excitatory tones, which is applicable to the 56% (n = 18) of rate-tuned neurons that were also selective for tone duration (Fig. 2, A and B). However, 44% (n = 14) of rate-tuned neurons were not duration tuned for tones (Fig. 2, C and D), so another mechanism must shape rate tuning. This appears to be a late-arriving, high-frequency inhibition, initiated by a dFM passing through frequencies higher than the excitatory frequency. This is termed delayed high-frequency inhibition. Because some duration-tuned neurons also have high-frequency inhibitory flanks, it is also possible that both mechanisms may work in concert.
|
), suggesting that this early inhibition creates a neuronal selectivity for short durations. This model for creating short-pass duration tuning has two components: an early inhibitory input and a delayed excitatory input, both generated by the same excitatory tone. The excitatory response is assumed to be a transient onset spiking of short duration and independent of sound duration. The inhibitory input lasts for the duration of the sound and, if the sound is long enough, the inhibition temporally overlaps with excitation and suppresses the response (Fig. 3C). In this hypothetical example (Fig. 3A, inhibitory domain 1), the low-frequency inhibition arrives 1 ms before excitation and has a bandwidth of 8 kHz. The high-frequency inhibition (inhibitory domain 2) arrives 1 ms after excitation and has a bandwidth of 5 kHz. The excitatory bandwidth is 5 kHz, as is the bandwidth of the on-BF inhibition, which arrives 1 ms before excitation. Figure 3B shows the excitatory band and inhibitory bands 1 and 2 on a time–frequency plot, with two downward FM sweeps from 60 to 30 kHz traversing these bands. Sweep 1 has a duration of 2 ms and a rate of 15 kHz/ms; sweep 2 has a duration of 6 ms and a rate of 5 kHz/ms. Given these parameters, sweep 1 will excite the neuron, whereas sweep 2 will inhibit it, regardless of whether the neuron's rate selectivity is created by early on-BF inhibition (Fig. 3C) or late high-frequency inhibition (Fig. 3D).
If rate selectivity were created by early on-BF inhibition (Fig. 3C), the following events would occur during the course of these FM sweeps. Sweep 1 (Fig. 3B) would traverse the excitatory band in 0.33 ms because it is traveling at a rate of 15 kHz/ms through a 5-kHz band. This would generate a short inhibitory postsynaptic potential (IPSP) of about 0.33-ms duration and, 1 ms later, a transient excitatory postsynaptic potential (EPSP) (Fig. 3C). The two events would not overlap temporally and the neuron would respond. Sweep 2 would traverse the excitatory band in 1 ms because it is traveling through the 5-kHz band at a rate of 5 kHz/ms. The early IPSP would now last about 1 ms and overlap with the EPSP, thus countering depolarization and suppressing the response. This mechanism thus creates a short-pass filter for tone duration and a fast-pass filter for FM sweep rate.
If rate selectivity were created by late high-frequency inhibition (Fig. 3D), the neuron would respond to sweep 1 (Fig. 3B) because excitation arrives before inhibition. The sweep, traveling at 15 kHz/ms, would take 0.33 ms to traverse the 5-kHz high-frequency inhibitory band and reach the excitatory band. This rapid traversing of the inhibitory band is not long enough to offset the 1-ms delay in the arrival of high-frequency inhibition. Consequently, there would be a brief EPSP followed 1 ms later by a brief IPSP (Fig. 3D). The two would not overlap in time and the neuron would respond. In sweep 2 (Fig. 3B), the sweep is traveling a 5 kHz/ms and requires 1 ms to cross the 5-kHz inhibitory band. The excitatory input is now triggered 1 ms after inhibition, offsetting the delay in the arrival of inhibition. The EPSP and IPSP now overlap in time, depolarization does not occur, and the neuron would not respond. Like early on-BF inhibition, this mechanism also creates a fast-pass filter for FM sweep rate.
Selectivity for a downward sweep direction is created by the low-frequency inhibitory flank (Fig. 3A, inhibitory domain 1), which has a broader inhibitory flank than that of high-frequency inhibition, and arrives earlier than or approximately the same time as excitation. An upward FM sweep will therefore cause inhibition to arrive before excitation.
Duration tuning and rate tuning
Duration tuning is defined here as a short-pass or band-pass selectivity for the duration of a tone (e.g., Figs. 2A and 4B). We previously reported (Fuzessery and Hall 1999) that the high-frequency region of the pallid bat IC contains a large percentage of neurons (59%) that respond preferentially or exclusively to sounds with durations of 5 ms. A similar percentage (60%) was obtained in the present study.
|
2 ms. The longest best duration was 5 ms. All neurons that were selective for tone duration were also selective for FM sweep rate and the following evidence suggests that, in the majority of these neurons, duration tuning shapes rate tuning. In 11 of the 18 neurons for which sufficient data were collected, a neuron's best rate could be predicted from its best duration and the bandwidth of its excitatory tuning curve. The underlying assumption is that a neuron will prefer a sweep rate in which the time spent within the excitatory bandwidth equals the best duration. The predicted best rate therefore equals the excitatory bandwidth divided by the best duration (kHz/ms). Figure 4 provides an example. The excitatory bandwidth (Fig. 4A) is 4 kHz at the test intensity (50 dB) and best duration is 0.5 ms (Fig. 4B), leading to a predicted best rate of 8 kHz/ms. The actual best rate is 12 kHz/ms (Fig. 4C). Figure 5 summarizes the predicted and actual best rates for 11 neurons. This high correlation (r2 = 0.797) suggests that duration tuning contributes to rate tuning in these neurons.
|
Delayed high-frequency inhibitory sidebands and rate tuning
Fourteen neurons that were selective for dFM rate were not selective for tone duration (e.g., Fig. 2C). Delayed high-frequency inhibition appears to be responsible for shaping their rate tuning. As noted above, the premise is that as the dFM rate decreases, the temporal advantage enjoyed by the faster excitatory input is lost and the response is suppressed. Delayed high-frequency inhibition thus creates a fast-pass rate filter.
Within the context of a spectrotemporally dynamic sound like an FM sweep, time and frequency are interchangeable. The arrival time of inhibition during a sweep depends on two factors: 1) its arrival time relative to excitation when tested using a two-tone inhibition paradigm and 2) the bandwidth of the inhibitory flank. The broader the bandwidth, the sooner the inhibitory flank will be encountered and the sooner inhibitory input will be triggered during a sweep.
The arrival times and bandwidths of inhibition were determined as follows. After the tonal response area was mapped, the TTI over time method (see METHODS) was used to quickly map inhibitory flanks in frequency and time. In the example shown (Fig. 6A), the excitatory tone had to be delayed 5 ms relative to a high-frequency tone before a high-frequency inhibitory sideband was apparent, indicating that high-frequency inhibition arrived 5 ms after excitation. A more detailed quantification of the arrival time of inhibition, relative to excitation, was obtained by measuring responses to pairs of tones, one from the excitatory tuning curve and one from the high-frequency inhibitory sideband (Fig. 6B). The timing of the excitatory tone (dark bar with arrow, Fig. 6B) had to be delayed 5 ms relative to the longer inhibitory tone (light bar, Fig. 6B) before the neuron's response was maximally suppressed. Note also that the inhibitory input triggered by low-frequency inhibition arrived earlier than that triggered by high-frequency inhibition (Fig. 6, A and B); this was typical of most neurons with both low- and high-frequency inhibitions.
The second feature of high-frequency inhibition needed to predict rate tuning is the bandwidth of the high-frequency inhibitory flank. In most neurons, the inhibitory and excitatory frequency domains were immediately adjacent to one another, but in some there was a small gap between them. To more accurately quantify the separation in frequencies at which inhibitory and excitatory inputs would be activated during an FM sweep, the spectral distance was measured in kilohertz (Fig. 6A, arrows). For high-frequency inhibition, the spectral distance is the bandwidth between the highest frequency in the inhibitory flank and the highest frequency in the excitatory band. For low-frequency inhibition, it was the bandwidth between the lowest inhibitory frequency and the lowest excitatory frequency.
The differences in arrival time and spectral distance seen for this neuron (Fig. 6) are typical of the recorded population (Fig. 7). As shown in Fig. 7A, the arrival times of low- and high-frequency inhibitions of all neurons tested (n = 36) were significantly different (t-test, P < 0.001). The average arrival time of low-frequency inhibition was 0.00 ms, for high-frequency inhibition, 2.37 ms. Note that while all 36 neurons in Fig. 7A had low-frequency inhibition, only 22 also had high-frequency inhibition. Figure 7C shows the arrival times of low- and high-frequency inhibitions in these 22 neurons. The trend is for the low-frequency inhibition to arrive earlier and the two populations are significantly different (paired t-test, P < 0.001).
|
The role of high-frequency inhibition in shaping rate tuning was tested in two ways. The first method is predictive, and calculates the sweep rate at which the delayed high-frequency inhibition will arrive at the same time as the excitatory input and suppress the response by 90% of maximum response. Note that this is different from predicting rate tuning created by duration tuning, where it is the best rate that is predicted. To calculate this 90% cutoff rate (kHz/ms), the "kHz" value used is the spectral distance and the "ms" value is the relative arrival time of inhibition that causes the response to drop to 10% of maximum response, when measured statically against an excitatory tone (e.g., Fig. 6B). In the example shown (Fig. 6), the spectral distance for high-frequency inhibition is 6 kHz (36–42 kHz, Fig. 6A) and the inhibition arrives 5 ms after excitation (Fig. 6B). The neuron is predicted to respond to downward FM sweeps until the rate slows to 6 kHz/5 ms = 1.2 kHz/ms. The actual rate at which 90% inhibition occurs is about 1.25 kHz/ms (Fig. 6C, arrow). Note that although the inhibition generated by a representative tone from the high-frequency sideband did not generate complete inhibition, the neuron is nonetheless completely inhibited at slower downward sweep velocities. This may indicate that the inhibition generated during a downward sweep is stronger than that generated by a single tone within the sweep.
Of the 14 neurons whose cutoff rate was assumed to be created by delayed high-frequency inhibition, sufficient data for prediction were collected in 12 and the rate was predicted in nine neurons whose high-frequency inhibition arrived later than excitation (Fig. 8). The correlation (r2 = 0.996) suggests that the properties of high-frequency inhibition do shape the 90% cutoff rates. However, in the remaining three of the 12 neurons, high-frequency inhibition arrived earlier than excitation. The predicted sweep rate cutoffs of these neurons could not be calculated because, according to our model, these neurons should not have responded to downward sweeps at all. Additional unknown mechanisms must therefore shape both the rate and direction selectivity in these three neurons.
|
|
Selectivity for the downward sweep direction observed in 66% of the recorded population was correlated with the finding (Fig. 7) that low-frequency inhibition in most neurons arrived earlier than or at the same time as excitation and that low-frequency inhibitory flanks were, on average, broader than the high-frequency inhibitory flanks (15 vs. 4.2 kHz). The role of low-frequency inhibition in shaping direction selectivity was tested directly by eliminating it from upward FM sweeps by starting the sweeps at the lowest excitatory frequency. As shown in Fig. 10, this neuron had a low-frequency inhibitory sideband that was broader than that of high-frequency inhibition and also arrived earlier than high-frequency inhibition (1 vs. 5 ms). It responded well to a downward FM sweep from 40 to 20 kHz, but barely responded to a spectrally identical upward sweep of 20–40 kHz (Fig. 10B). However, if the upward sweep started at 27 kHz, eliminating the low-frequency inhibitory sideband from the sweep, the neuron responded nearly as strongly as it did when presented with an excitatory tone, at short durations of 5 ms. Note also that the neuron is rate tuned for downward sweeps arising from the delayed high-frequency inhibition (note that its response to an excitatory tone indicates that it is not duration tuned), but that it was not rate tuned for the upward 27- to 40-kHz sweep because, in this sweep, the delayed high-frequency inhibition would be triggered after excitation. In 10 of 12 neurons tested, this manipulation eliminated direction selectivity. It did not do so in the remaining two, suggesting that additional mechanisms are shaping their direction selectivity.
|
One neuron (Fig. 11) whose low-frequency inhibition arrived very early, 3 ms before excitation, is of particular interest in terms of its direction selectivity. This neuron, as expected, did not respond to upward sweeps, but it also did not respond to downward sweeps if the downward sweeps entered the low-frequency inhibitory flank, whose highest frequency was 36 kHz (Fig. 11A). This neuron did not respond to standard downward FM sweeps of 60–30 and 70–20 kHz, but did respond to a 45- to 36-kHz sweep, which stopped just before entering the low-frequency inhibitory flank, with the same response magnitude evoked by an excitatory tone (Fig. 11B). However, if the downward sweeps were extended into the inhibitory flank (sweeps 45–34 and 45–32 kHz, Fig. 11B), the responses progressively decreased.
|
Duration of inhibition
Finally, the models of the mechanisms underlying rate and direction selectivity require that the inhibition flanking the excitatory band, once triggered during a sweep, lasts long enough to prevent a neuron from responding as the sweep enters the excitatory band. The two-tone paradigm provides information about whether inhibition lasts the duration of the inhibitory tone (tonicity) and how long it persists after that duration before the response returns to 90% of maximum response.
Using the equation described in METHODS, the tonicity and persistence of low-frequency inhibition were calculated for 34 neurons. In the great majority (94%, n = 32), inhibition lasted the duration of the 5- to 30-ms inhibitory tones used for two-tone inhibition, i.e., they had complete tonicity. In 73% (n = 25), inhibition persisted 1–10 ms beyond the duration of the inhibitory tone and in 23% (n = 8) persisted >10 ms, 43 ms. In one case, a range of inhibitory tone durations was tested (Fig. 12) and inhibition lasted the duration of the inhibitory tone as its duration was increased from 20 to 100 ms. In the remaining two neurons that lacked tonicity, inhibition was short-lived, returning to 90% maximum response in 3 and 7 ms when 20-ms inhibitory-tone durations were used.
High-frequency inhibition was measured in 15 of these 34 neurons. The majority (87%, n = 13) had inhibition that lasted for the duration of the inhibitory tone. Of these, seven had inhibition that persisted 5 ms beyond the duration of the inhibitory tone. Inhibition in the remaining six neurons persisted for >5 ms; the longest persistence observed was 23 ms. In the two neurons that did not display tonic high-frequency inhibition, inhibition lasted only 1 and 6 ms when tested with 20-ms inhibitory tones. The conclusion we draw is that the majority of neurons have inhibitory inputs that are sufficiently tonic and persistent so that, once activated, they will suppress responses during the brief 1.5- to 5-ms FM sweeps produced by an echolocating pallid bat.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
).
These caveats aside, the response selectivity for both sweep direction and rate in most neurons could be explained by inhibitory mechanisms. All neurons selective for sweep direction were selective for the downward direction, as was also found in a previous study of the pallid bat IC (Fuzessery 1994). This selectivity is created by two types of asymmetry in sideband inhibition. Low-frequency inhibition arrived earlier than high-frequency inhibition when tested with two-tone inhibition and the average bandwidth of low-frequency inhibition was significantly broader. The broader bandwidth further advances the arrival of low-frequency inhibition during the course of an upward FM sweep, assuming that the latency of an inhibitory tone (relative to the latency of an excitatory tone) accurately predicted its latency within the context of an FM sweep. Moreover, low-frequency inhibition was always present in downward direction–selective neurons (with one exception), whereas high-frequency inhibition was absent in 40% of the neurons tested.
There is, of course, nothing novel about the finding that sideband inhibition creates direction selectivity. This inhibitory mechanism has often been suggested since the earliest studies of directional selectivity for spectrotemporally dynamic sounds (e.g., Britt and Starr 1976; Heil et al. 1992a,b; Suga 1973). We can emphasize, however, that the pallid bat IC appears to regulate the arrival time of inhibition through a modulation of the bandwidth of inhibition. Because time and frequency are interchangeable during the course of a spectrotemporally dynamic signal like an FM sweep, this is a simple and effective mechanism for modulating the arrival time of inhibition. Inhibition can actually arrive later than excitation when tested with tone pairs, but still arrive earlier than excitation during the course of a sweep if the inhibitory bandwidth is appropriate. A practical implication, as noted by Gordon and O'Neill (1998), is that if the temporal relationship of the tones in a two-tone inhibition paradigm is not varied, the inhibitory flanks that shape direction and rate selectivity may not be apparent.
Different mechanisms, same rate selectivity
The great majority of neurons were tuned to sweep rate. An unexpected finding was that essentially identical rate tuning could be created by two different mechanisms: 1) an early on-BF inhibition, which creates duration tuning; and 2) a delayed high-frequency inhibition. Duration tuning is defined with tonal stimulation rather than FM sweeps because tones are of invariant frequency and the sensitivity cannot be confused with a sensitivity to FM sweep rate. Previous studies (Casseday et al. 1997; Fuzessery and Hall 1999; Gordon and O'Neill 1998) of rate or duration selectivity have noted that these two forms of sensitivity can be readily confused.
In a previous study of the pallid bat IC (Fuzessery and Hall 1999), roughly 50% of the neurons in the high-frequency region were found to have short-pass or band-pass selectivity for tone durations. In the present study, 60% of rate-tuned neurons exhibited this selectivity. The rate tuning of some duration-tuned neurons could be predicted through calculation of the time that FM sweeps spent within a neuron's excitatory tuning curve. In addition to this prediction, we also eliminated the possibility that high-frequency inhibition contributed to rate tuning by starting the downward sweeps in the excitatory tuning curve. In the majority of cases, this had no effect on rate tuning, indicating that it was indeed duration tuning that shaped rate tuning, and not the inhibitory sidebands.
Duration tuning—and the resultant sweep rate tuning—has been suggested (Fuzessery and Hall 1999) to be created by an on-BF inhibition in which excitatory frequencies first evoke an early inhibition that lasts the duration of the sound, followed by a later excitation with a fixed response latency. If the sound lasts long enough, inhibition and excitation will overlap in time and the response will be suppressed. In effect, the excitatory receptive fields of these neurons last only a few milliseconds. The mechanism underlying duration tuning has also been modeled as a coincidence mechanism, in which a rebound from the early inhibition contributes to excitation (Casseday et al. 1994). Present experimental design does not allow us to lend support to one model over the other. Faure et al. (2003) and present results provide evidence that this early inhibition typically lasts the duration of the sound, as required of these models, so both models are plausible.
These findings suggest that the purpose of early on-BF inhibition in the high-frequency region of the pallid bat IC is to shape neuronal selectivity for the rate of the bat's echolocation pulse and not a selectivity for durations of tonal stimuli. The great majority of duration-tuned neurons had best durations of 2 ms and some did not respond to tone durations >0.5 ms. It is unlikely that the role of these neurons is to process very short duration tones because the communication sounds used by this bat are largely of lower frequency and/or longer duration than the echolocation pulse and are also frequency modulated (Brown 1976).
Rate tuning for downward FM sweeps is also created by an "off-best frequency," delayed high-frequency inhibitory sideband. In addition to a delay in the arrival of this inhibition relative to excitation, the bandwidth of this inhibitory flank is typically narrower than the low-frequency flank. This ensures that high-frequency inhibition will not be activated too early during a downward FM sweep and thus suppress the response. This mechanism creates a "fast-pass" rate tuning in which the neuron responds to downward FM sweeps until the rate slows to the extent that the difference in arrival times of excitation and high-frequency inhibition is negated. In actuality, neurons influenced by this mechanism typically show a band-pass, rather than fast-pass, rate tuning and it is likely that the high-rate flanks of their rate functions are shaped simply by an inability to detect and respond to extremely short duration FM sweeps.
It is of interest that an early on-BF inhibition and a late high-frequency inhibition both create fast-pass filters for sweep rate. The fact that most of the rate-tuned neurons described in this study exhibited band-pass rate functions therefore needs explanation. At present, we can only suggest that the drop in response at faster sweep rates may be the result of a neuron's integration time or detection threshold.
Comparison of IC and auditory cortex
A comparison of present results and those of a parallel study in the pallid bat auditory cortex (Razak and Fuzessery 2006) is of general interest with regard to hierarchial processing in sensory systems. The rate selectivity at both levels is similar. Neurons with band-pass rate tuning do not have significantly different best rates, although the sharpness of tuning is significantly more narrow in the IC. Although the selectivity is generally similar, however, an unexpected finding is that one of the mechanisms that creates rate tuning in the IC seems reduced or absent at the cortical level. All rate tuning in the auditory cortex is shaped by delayed high-frequency inhibition. Duration tuning is rare and, when present, does not predict rate tuning. This apparent loss of an underlying mechanism is perhaps the result of a convergence of input onto thalamic and/or cortical neurons from IC neurons that are duration tuned and others that are not. The result would appear as a lack of duration tuning. This finding has an important implication. The presence of similar forms of selectivity at the midbrain and cortical levels does not necessarily imply that the auditory cortex simply inherits the selectivity. The auditory cortex may have to re-create, at least in part, forms of response selectivity created at lower levels of the system. This notion is supported by a recent patch-clamp study of FM sweep processing in the rat auditory cortex (Zhang et al. 2003), suggesting that intracortical processes sharpen neuronal selectivity for FM sweep direction.
Regarding behavioral relevance, the rate tuning of IC and cortical neurons is appropriate for processing the exponential FM sweep of the bat's echolocation pulse. The average best rate at both levels of the system was 3–4 kHz/ms, and ranged from 1 to 12 kHz/ms. The instantaneous rates of 2-ms echolocation pulses at 40–45 kHz (the best frequencies of most recorded neurons) is 7–8 kHz/ms. The average rate tuning is thus a little lower than might be expected. It is possible that if exponential, rather than linear, FM sweeps had been used as test stimuli, the sweeps may have traversed the high-frequency inhibitory flanks more rapidly (in neurons possessing this inhibition), resulting in a range of best sweep rates more in line with what the bat normally experiences.
Facilitatory mechanisms
The use of two or more tones that evoke excitatory and inhibitory events has proven a simple and useful tool for delineating the timing of these inputs and predicting responses to more complex sounds (e.g., Brosch and Schreiner 2000; Faure et al. 2003; Fuzessery and Feng 1983; Gordon and O'Neill 1998; Nataraj and Wenstrup 2005). Its use in the present study proved successful in revealing the properties of inhibitory mechanisms that contribute selectivity for dimensions of FM sweeps. However, it is likely that excitatory mechanisms, in the form of summation or facilitation, also contribute to response selectivity. This inference derives from several sources, perhaps the most obvious being the FM specialists described here and in previous studies (e.g., Casseday and Covey 1992; Casseday et al. 1997; Fuzessery 1994
