In the course of this study, intracellular recordings were made from 78 neurons of which 53 were labeled with neurobiotin. Of these recovered neurons, 17 were identified as fusiform cells based on the bipolar arrangement of the dendrites and, in most cases, on the trajectory of the axon.
Confidence in the association between physiology and anatomy
In all experiments, the stability of the resting membrane potential was monitored during iontophoresis of neurobiotin. Injections were halted if the resting membrane potential became greater than 20 mV above the starting value. In the cases for which the impalement was stable after iontophoresis, some portion of the data collection protocol was repeated.
Despite precautions, there were ambiguities in some cases. Table 1 details the types of ambiguities encountered and the basis for identification for each of the 17 fusiform cells in this study. The left side of Table 1 lists the number of neurobiotin deposits attempted and the number of neurons recovered for each experiment. Six cases were unambiguous in that only one injection of neurobiotin was attempted and only one neuron was recovered.
For the cases in which multiple deposits were made, recovered neurons were identified on the basis of location. First, neurons were identified on the basis of the separation of the electrode tracks in the rostral-caudal dimension. When this was not clear, identifications were made by comparing the relative depths of the labeled neurons to the micropositioner readings recorded during the experiment.
A second class of ambiguity occurred when more neurons were recovered than deposits were made. A hierarchy of criteria was used to make identifications in these cases. 1) Faintly or incompletely labeled neurons were assumed to result from neurobiotin leakage during brief intracellular contacts and were eliminated from consideration. 2) Small current pulses used to study responses were often sufficient to label neurons. These were identified on the basis of position as described above. 3) Recovered cartwheel cells could usually be matched up with experiment notes indicating brief impalements of complex-spiking neurons. 4) In one case (97005), two fusiform cells were recovered in different portions of the tonotopic axis and an identification was made on the basis of best frequency. In this instance, the association of physiology with a specific neuron is perhaps a bit tenuous, but the classification of fusiform cell is certain. Finally, responses of two fusiform cells were not identifiable and were rejected from consideration in this study.
These fusiform cells are anatomically similar to those described by earlier studies in a variety of species, as illustrated in Figs. 1-6. Spinous, thickly branched apical dendrites extend into the molecular layer, typically reaching all the way to the ependyma. The basal dendrites are less densely branched and spine-free and descend into the deep layer of the DCN. Planarity of the dendritic fields is particularly conspicuous in Figs. 2 and 6. Axons were identified in 15 cases, arising from either the soma (Figs. 1, 3, and 6) or a proximal basal dendrite (Figs. 2 and 5) and joining the DAS to leave the nucleus at its dorsomedial border. The labeling of the axon was typically intermittent, consistent with the presence of a myelin sheath, which hinders the access of histochemical reagents (Zhang and Oertel 1993b). No axon collaterals were ever observed, a consistent observation in rodents (Ding et al. 1999; Manis 1990; Zhang and Oertel 1994) that stands in contrast to observations in the cat (Lorente de Nó 1981; Rhode et al. 1983; Smith and Rhode 1985).
The physiological characteristics of the 17 fusiform cells are summarized in Table 2. The responses are quantified using a variety of common measures, including spontaneous firing rate (SR), PSTH shape, normalized slope of the BF rate-level curve, relative noise response, input resistance, and resting potential (RP). By any of these metrics, the fusiform cell population demonstrates a wide range of physiological properties. Specific examples are illustrated in Figs. 1-6.
The responses of cell 97073 (RP = 66 mV, SR = 0.0 spikes/s, Rin = 8 M) were weak and of high threshold (Fig. 1D). In response to sideband tones, it did not fire action potentials, although at high levels did respond with subthreshold depolarizations of approximately 5 mV in amplitude (Fig. 1, G and I). The cell discharged in a buildup pattern (Fig. 1E). The PSTH data were obtained shortly before the impalement was lost at a point when the threshold had decreased. Thus the PSTH in Fig. 1E exhibits a larger rate and shorter latency than the membrane potential record of Fig. 1H obtained earlier in the impalement.
Cell 97022a (RP = 62 mV, SR = 0.2 spikes/s, Rin = 29 M) was a pauser unit with regular interspike intervals (Fig. 2, E and F). The rate-level curves for BF tones, below-BF tones, and broadband noise were highly nonmonotonic (Fig. 2D). The action potentials had deep undershoots and were accompanied by large sustained depolarizations, approximately 10 mV in size. Depolarizations during the stimulus were present even as the number of spikes decreased at high levels (Fig. 2J). The presence of sideband inhibition is suggested by hyperpolarization of the membrane during the stimulus in response to high-frequency tones (Fig. 2G).
Cell 97055a (RP = 76 mV, SR = 15.2 spikes/s, Rin = 32 M) was classified as a pauchopper, which is characterized by a long first interval followed by regular spiking in the steady state (Fig. 3, E and F). The firing rate was a monotonic function of level when the cell was driven with BF tones, below-BF tones, or broadband noise (Fig. 3D). The broadband noise series was obtained shortly after impalement before the membrane potential stabilized, and thus the noise responses have a larger spontaneous rate (Fig. 3D) and a larger resting potential (Fig. 3J) than the remainder of the data. Stimulation with above-BF tones evoked inhibition of the spike rate (Fig. 3D) and hyperpolarization of the membrane (Fig. 3G).
Cell 98001 (RP = 57 mV, SR = 11.9 spikes/s, Rin = 26 M) was excited by BF-tones, sideband tones, and broadband noise. In contrast to the previous examples, this cell showed no appreciable sustained depolarization during the stimulus, but did exhibit prolonged afterhyperpolarizations. The PSTH data clearly show a pause following the first spike (Fig. 4E) and regular discharges in the steady state (Fig. 4F).
Cell 97002 (RP = 66 mV, SR = 0 spikes/s, Rin = 14 M) is classified as a pauser unit, since the mean interspike interval was initially long and then declined to a constant steady-state value (Fig. 5F). The BF rate-level curve is characterized by a low threshold and firing rates that decreased steadily for levels above 40 dB SPL (Fig. 5D). The cell was excited by broadband noise and sideband tones without a rollover in the rate at high levels. During excitation, action potentials with deep undershoots were superimposed on a large sustained depolarization followed by little or no afterhyperpolarization (Fig. 5, H-J). Figure 5G shows that in response to a tone more than an octave above BF, the cell still experienced sustained, albeit subthreshold, depolarizations. Thus in contrast to some of the previous examples, there was no indication of sideband inhibition.
Cell 97054b (RP = 54 mV, SR = 20.2 spikes/s, Rin = 16 M) was excited by BF tones and more weakly excited by below-BF tones and broadband noise (Fig. 6D). Excitation was accompanied by minimal sustained depolarization, but substantial afterhyperpolarizations (Fig. 6, H-J). The above-BF rate-level curve in Fig. 6D shows inhibitory responses at high levels. Under such stimulus conditions, the membrane depolarized weakly and then underwent a large hyperpolarization that outlasted the stimulus by several milliseconds (Fig. 6G). This cell is a pauchopper unit, as evidenced by the long first interspike interval and subsequent chopping pattern (Fig. 6E).
Sustained stimulus-evoked changes in membrane potential
Fusiform cells often exhibited sustained depolarizations and afterhyperpolarizations in response to acoustic stimulation, as illustrated in Fig. 7. For five cells, responses to BF tone bursts of increasing level are shown. Three responses were averaged at each sound level and the plots scaled to emphasize the slow potential changes; consequently, the action potentials have been attenuated and in some cases, clipped. Also, note that the data were collected in 5-dB increments, but for purposes of clarity, only every other level is shown.
In general, both sustained depolarizations and afterhyperpolarizations grow monotonically in amplitude with increasing sound level, as is the case for fusiform cells in the anesthetized cat (Rhode et al. 1983). Figure 7A shows an example of a cell that showed sustained depolarizations during the stimulus, but no afterhyperpolarizations. Stimulus-evoked depolarization was apparent at 20 dB SPL, but did not reach spike threshold. The depolarization grew larger with increasing sound level, reaching spike threshold at 50 dB SPL, after which it remained relatively constant in amplitude. This cell was one of the few in our survey that showed conspicuous postsynaptic potentials.
Figure 7, B and C, shows two cells that did not have spontaneous activity and for which the sustained depolarizations were accompanied by pronounced afterhyperpolarizations. The latter grew in both amplitude and duration with increasing sound pressure level. Afterhyperpolarizations were often followed by rebounds, indicated by the arrowheads in Fig. 7, B and C. In most cases, the rebound did not reach the resting potential, but a nonmonotonic membrane potential trajectory was nevertheless apparent.
Figure 7 also demonstrates that afterhyperpolarization had a profound effect on spontaneous firing. For the cell in Fig. 7D, the accumulated effect of afterhyperpolarization resulted in a 5-mV decrease in the resting potential and eliminated spontaneous firing altogether.
Current-clamp results using the Manis paradigm
A current-clamp paradigm was previously used in vitro to demonstrate the ability of a fast, inactivating K+ current to alter fusiform cell firing patterns (Kanold and Manis 1999; Manis 1990). This current was elicited by first deinactivating it with a hyperpolarizing pulse of injected current, then activating it with a depolarizing pulse. The resulting outward K+ current inactivated with a time constant of about 11 ms and opposed membrane depolarization, turning a chopper discharge pattern into a buildup or pauser pattern.
We used this current-clamp paradigm on nine fusiform cells to demonstrate the existence of the inactivating K+ current in vivo. Four of the cells showed results consistent with those described in the slice. Membrane potential waveforms for three such cells are shown in Fig. 8. For the lowest holding currents and smallest pulse amplitudes (bottom left of each panel), each cell showed a buildup pattern. The long-latency portion of the buildup is characterized by a gradual increase in the membrane potential toward spike threshold. The latency shifted to shorter values as the pulse amplitude increased (toward the top of each column). The latency also shortened at a fixed pulse amplitude when the holding current became less negative. Large pulse amplitudes sometimes evoked an onset spike before the buildup, creating a pauser discharge pattern (Fig. 8, B and C). Finally, at the largest pulse amplitudes and smallest holding currents, the cells responded with a chopping pattern for the duration of the pulse (top righthand plot in each panel).
Only four of the nine fusiform cells tested showed the results represented in Fig. 8. These had larger input resistances (Rin 26 M vs. Rin 16 M) than the remainder, which showed chopper responses unaffected by prior hyperpolarization. It is likely that, for the holding currents used ( 1 nA), the latter group could not be sufficiently hyperpolarized to deinactivate the underlying K+ currents.
Temporal response patterns
For PSTH data from the DCN of anesthetized gerbil, Gdowski (1995) quantified the shape of the regularity histogram by simultaneously fitting two lines to the mean ISI and noting the transient (mT) and steady-state (mSS) slopes. In the present study, however, there were several instances that seemed to call for a three-line fit (e.g., Fig. 9M). These cases typically had a sharp initial decrease in the mean ISI, followed by longer periods of transient and steady-state firing. Since the initial decrease was a common feature of these units, it was not a useful means of characterizing their behavior, and so a three-line fit was not attempted. Rather, the data were recomputed using 5-ms bins and the two-line fit repeated after eliminating the first point. The slopes of these two lines were called m1 and m2 to distinguish them from mT and mSS.
For the data of Fig. 9M, the slope value m1 captures the gradual increase in ISI over the first 50-75 ms, whereas the transient slope mT as computed in earlier studies would be dominated by the very brief initial decrease. The positive value of m1 is interpreted as weak inhibition, in that the trend of increasing ISIs is reminiscent of the behavior of auditory nerve fibers and so does not reflect significant modification by inhibitory processes. This is in contrast to Fig. 9C, for example, for which the slope value m1 is steeply negative. The negative value indicates that the ISI trend is opposite that of the underlying excitatory drive from auditory nerve fibers. It is thus interpreted as reflecting an inhibitory input that is initially strong and gradually weakens over the first half of the stimulus.
The regularity histograms in Fig. 9, A-P, are plotted in order of increasing m1 values. Across this sample of the fusiform cell population, the interval varies from steeply negative (A) to steeply positive (P). This variation is also apparent in the distribution of m1 values plotted in Fig. 10A.
Another metric for quantifying the extent to which inhibition contributes to DCN physiology is the normalized tone slope (Davis et al. 1996; Young and Voigt 1982). This quantity is measured by first finding the level at which the BF rate-level function either saturates or becomes nonmonotonic. From this level up, a line is fit to the rate data and the slope divided by the maximum firing rate to give a normalized tone slope in units of dB1. More negative tone slopes indicate more highly nonmonotonic rate-level functions and therefore responses characterized by greater inhibitory influence at higher levels. Figure 10B shows the distribution of fusiform cell normalized tone slopes. In general, the values are not distributed as evenly as those of the regularity slope. A majority of the values (9/17) are clustered between 4 · 103 and 8 · 103/dB.