The dorsal cochlear nucleus receives input from the auditory nerve and relays acoustic information to the inferior colliculus. Its principal cells receive two systems of inputs. One system through the molecular layer carries multimodal information that is processed through a neuronal circuit that resembles the cerebellum. A second system through the deep layer carries primary auditory nerve input, some of which is relayed through interneurons. The present study reveals the morphology of individual interneurons and their local axonal arbors and how these inhibitory interneurons respond to sound. Vertical cells lie beneath the fusiform cell layer. Their dendritic and axonal arbors are limited to an isofrequency lamina. They give rise to pericellular nests around the base of fusiform cells and their proximal basal dendrites. These cells exhibit an onset-graded response to short tones and have response features defined as type II. They have tuning curves that are closed contours (0 shaped), thresholds ∼27 dB SPL, spontaneous firing rates of ∼0 spikes/s, and they respond weakly or not at all to broadband noise, as described for type II units. Their responses are nonmonotonic functions of intensity with peak responses between 30 and 60 dB SPL. They also show a preference for the high-to-low direction of a frequency sweep. It has been suggested that these circuits may be involved in the processing of spectral cues for the localization of sound sources.
The function of the dorsal cochlear nucleus (DCN) remains incompletely understood. The DCN has been shown to be multisensory (Itoh et al. 1987; Weinberg and Rustioni 1987; Wright and Ryugo 1996) receiving not only auditory-nerve input but facial and pinnae somatosensory inputs as well (Young et al. 1995). DCN units are spontaneously active and are inhibited by changes in pinna position. This suggests in animals with a mobile pinna, such as the cat, that pinna position may modulate responses to other sensory inputs. Much of this multimodal processing is thought to occur in the superficially located molecular layer of the DCN. Recently it has been demonstrated that the DCN may have a role in the acoustic startle reflex (Meloni and Davis 1998).
Based on an extensive series of studies of the DCN, it has been suggested that the detection of spectral notches, a cue used to localize the elevation of a sound, also may be a function of the DCN (Nelken and Young 1996). Fusiform or principal cells, the major class of DCN projection neurons, appear to be very sensitive to the presence or absence of a spectral notch in the stimulus spectrum. This sensitivity is likely a result of tonotopically organized inhibitory inputs. Fusiform cells project to the inferior colliculus providing key localization information to the ascending acoustic pathway.
In classifying the physiological responses of neurons in the CN, two principal schemes have been employed. The one employed more frequently in DCN uses of the shapes of the excitatory and inhibitory areas within a unit’s response field (Evans and Nelson 1973). The response field shapes have five categories: type I, a simple V-shaped excitatory; type II, a simple V-shaped excitatory but having inhibitory influences; type III, a V-shaped excitatory area with inhibitory sidebands; type IV, a low-level excitatory region and variable excitatory and inhibitory areas; and type V, almost all inhibitory influences. The definition of type II responses has been expanded over the years by Young and colleagues to include several additional features: units having little or no spontaneous activity, nonmonotonic rate-level functions, and a weak response to wideband noise (WBN) stimuli (Shofner and Young 1985; Young and Brownell 1976; Young and Voigt 1982). It has been suggested that type II units correspond to the vertical cells of the deep DCN.
The second classification scheme for physiological responses employs the shape of the poststimulus time histogram (PSTH) to categorize response patterns to tones at the best frequency (Pfeiffer 1966). For example, some of the self-descriptive response patterns are labeled: primarylike, chopper, onset, pauser, and buildup. These response patterns have been correlated with specific morphological cell types in the CN (e.g., Rhode et al. 1983a,b; Rouiller and Ryugo 1984). It will be shown here that the response pattern termed onset-graded or On-g is associated with the vertical cells of the DCN and that they satisfy the requirements for type II units.
Note that neither classification scheme is unambiguous (e.g.,Davis et al. 1996; Shofner and Young 1985). In fact, in barbiturate anesthetized gerbil, the response map technique (type I–IV) was judged not useful (Gdowski and Voigt 1997).
Multiple-unit recording allowed the study of correlated firing in simultaneously recorded units. These studies provided evidence that type II cells inhibited type IV cells with about a 0.7-ms delay between the spike in a type II and a reduction in the activity of the type IVs (e.g., Voigt et al. 1982). This is best explained by the existence of a monosynaptic inhibitory connection between the type IIs and the type IVs. Labeling and electrical stimulation studies have shown that the type IV are fusiform cells that project out of the DCN via the dorsal acoustic stria (Rhode et al. 1983b;Young 1980).
Unraveling the structure-function relations in the cochlear nucleus has been made possible by the development of intracellular recording and labeling. However, success with this technique has been limited to larger neurons that are easier to record from (Ostapoff et al. 1994; Rhode et al. 1983a,b; Rouiller and Ryugo 1984; Smith and Rhode 1985, 1987, 1989; Smith et al. 1991,1993). Many unanswered questions remain regarding the physiological correlates of smaller and intrinsic neurons. This is especially true in in vivo studies where intracellular recording and labeling is especially difficult. The present study describes structure and function for one class of these intrinsic cells, the vertical cells in the DCN that respond with an On-g pattern to best frequency tones in anesthetized cat.
Data were obtained from cats (n = 129) that were part of a larger study. Each cat was anesthetized with pentobarbital (50 mg/kg) delivered intraperitoneally. Additional doses were given intravenously to eliminate any withdrawal reflex to a paw pinch. A tracheotomy was performed, and the animal was placed on a ventilator to maintain expired CO2 at 4%. The left ear was removed surgically. The posterior skull was exposed and opened exposing the posterior cerebellum. Part of the cerebellum overlying the cochlear nucleus was aspirated with the remainder gently retracted to expose the CN. A 2% agarose solution was applied over the region to aid in stabilizing the CN. A plastic chamber then was secured with dental acrylic cement to the skull. The chamber was filled with warm mineral oil and covered with glass to further reduce brain pulsations. The animal was suspended from the hips and spine to reduce mechanical coupling of respiratory movements to the brain. The protocol meets National Institutes of Health standards for the care and treatment of animals and was approved by the University of Wisconsin Institutional Animal Care and Use Committee.
A RadioShack supertweeter was used to deliver the sound through an earpiece inserted into the external meatus. A calibrated probetube condenser microphone (Bruel and Kjaer) was placed within 2 mm of the tympanic membrane after visualization of the middle ear with an operating microscope. The acoustic signal was calibrated from 100 to 50,000 Hz in 100-Hz steps. The middle ear was vented by a 1-mm (ID) polyethylene tube 30 cm in length to relieve any static pressure buildup that could alter the acoustic calibration or the middle-ear transfer function (Guinan and Peake 1967).
Stimuli, response pattern, and labeling
While a glass electrode filled with a 1 M KCl and 1% Neurobiotin solution was advanced through the tissue in 1-μm steps using a Marzhauser micromanipulator, a frequency sweep (FM) was used as the search stimulus. The FM tone was linearly swept using a triangular-modulating waveform with 1-s rise and 1-s fall times. Units were characterized by collecting a response surface over the frequency-level plane, a PSTH, a tone-rate-level function (RLF), a noise RLF, and an FM sweep response. If conditions permitted, intracellular responses were recorded for: the response plane (1 stimulus repetition at each point in the frequency-level plane); the response to the short-tone at best frequency (BF = the frequency corresponding to the lowest stimulus level resulting in a increase in discharge rate above the spontaneous rate); a rate curve (2 repetitions at each of 11 levels, −10–90 dB SPL in 10-dB steps); and the response to a FM sweep. The Clampex program was used to capture a cell’s intracellular responses using a Digidata 1200 A/D system from Axon Instruments. During the intracellular condition (resting potential less than −30 mV), Neurobiotin was iontophoresed using a 1- to 2-nA pulsed current (2.5-Hz squarewave) for ≤10 min. A record was made of the unit’s intracellular response during injection to determine whether the current was driving the unit. The intracellular responses were sufficient to be certain that the cell characterized extracellularly was the one recorded intracellularly and labeled. That is, the nonmonotonic rate-level function remained and/or the asymmetric FM response. When possible the PSTH was repeated after the injection. The structure-physiology correlation was entirely consistent for all cells labeled.
The stimulus paradigm permits the characterization of the unit in terms of both response field properties and temporal response pattern (PSTH). The PSTH was determined by the response to 250 repetitions of a 50-ms tone repeated every 125 ms at the best frequency of the unit. The PSTH of cochlear nucleus neurons has been described as primarylike, onset, chopper, pauser, and buildup (Pfeiffer 1966).Bourk (1976) described subsets of these classes, in particular, of the onset units he studied, On-g units had the slowest (graded) decrease in firing rate after the initial response. This pattern is consistent with that found here for cells labeled in this study. A 1-ms binwidth is used to form the PSTHs during on-line data collection. Off-line the use of 0.2-ms binwidths reveals that many but not all of these response patterns would be classified as transient chopper (Ch-t describes those PSTHs in which ≥2 initial modes are present), at least at the frequency-level stimulus combination that results in the maximum discharge rate. For almost any other stimulus condition, the chopping pattern is usually not seen and the pattern is best described as On-g. Ch-ts have somewhat smaller coefficients of variation (CV = SD/mean of interspike intervals) than On-gs and their interspike interval histograms are more symmetric than those of On-gs, which appear more primarylike (see Fig. 5). The Ch-t pattern is absent in On-g units that have a low maximal discharge rate (Fig.4 A). The RLF, response map, and FM sweep and noise response features of these units are entirely different from those of Ch-ts.
Another distinguishing feature of this cell type is the relative noise–to-tone ratio (RNR). RNR = (NR − SR)/(TR − SR) where NR is the maximum rate to a wideband noise, TR is the maximum rate to a tone, and SR is the spontaneous spike rate (Young and Brownell 1976). However, because the SR of all the type II cells (=On-g response pattern in the present study) encountered was ∼0 spikes/s, the computation is reduced to RNR = NR/TR. An RNR of <0.3 has been taken to imply a type II response.
Twenty-five On-g cells were studied extracellularly, 7 intracellularly, and 5 were labeled.
After being given 1 ml heparin sodium, 0.5 ml lidocaine hydrochloride, and an overdose of pentobarbital (a full anesthetic dose), the animal was perfused with 1 l of saline to flush the blood followed by 1.5 l of a 2% glutaraldehyde and 0.5% paraformaldehyde solution followed by 1.5 l of a 4% glutaraldehyde and 1% paraformaldehyde solution. The brain was removed and stored in fixative for ≥1 day. It then was transferred to a sucrose solution for ≥1 day after which sections were cut on a freezing microtome at either 60 or 80 μm. Processing of the tissue followed the standard procedure for Neurobiotin. A standard ABC Elite kit was used to react the tissue for 2.5 h. Selected cells were reconstructed using the Neurolucida (MicroBrightField) anatomic reconstruction system.
A profile of the defining features of an On-g unit is shown in Fig. 1. These units have little or no spontaneous activity (SR < 2.5 spikes/s), a nonmonotonic rate-level function in response to a tone at BF (—) coupled with a weak response to noise (Fig. 1 A, - - -). The PSTH pattern in Fig. 1 B is labeled On-g because of the high initial discharge rate and graded decrease thereafter (Bourk 1976). The response surface is a mound rising above the frequency-level plane (Fig. 1 C) that can be represented as a series of closed contours in the response plane which shrink with increasing spike rate (line thickness is proportional to discharge rate in Fig. 1 D). The frequency extent of the response is relatively constant across all levels. At the lowest criterion level the contour is essentially a frequency threshold curve (FTC). The values associated here with response features of On-g units are in general agreement with previous reports for those of type II units.
An additional feature of On-g units is that the discharge rate is dependent on the direction of frequency change of a frequency-modulated (FM) stimulus. As shown in Fig. 2, more spikes occur during the high-to-low (HL) frequency sweep than during a sweep in the opposite direction (LH). When the sweep time was varied over a wide range, the ratio of LH/HL spikes was relatively stable, except for very short sweep times (0.2 s, Fig. 2), in which case LH/HL approached unity and the spikes per sweep decreased (LH/HL = 56, 22, and 6, respectively, for 3 sweep times of 10, 2, and 0.4 s). This FM response was very consistent for all On-g units and was a useful indicator for recording from On-g units because an FM signal served as the search stimulus.
Four sets of tone and noise rate-level functions (RLFs) for type II units with a range of BFs are shown in Fig.3. The peak discharge rate is achieved within 10–15 dB of threshold. In each case, the rate-level function for tone is strongly nonmonotonic with the firing rate reduced to ∼0 by 80–90 dB SPL. Additionally, the rate-level functions for wideband noise (WBN = 50–50 kHz) indicate the response to noise is either very weak or nonexistent (– – –). The RNR is <0.1 in each case, and the average RNR is 0.056 (n = 12) with seven units having an RNR ∼ 0.
Four PSTHs at BF are shown in Fig. 4, and in each instance, the PSTH is On-g in shape. However, note that an initial chopper pattern also is present (a chopper pattern is one for which the PSTH in response to a tone is multimodal with the mode separation unrelated to the stimulus period). The coefficient of variation (CV) of the interspike interval histogram is used to separate chopper units into sustained choppers (Ch-s) if CV < 0.3 and transient choppers (Ch-t) when CV > 0.3 (Young et al. 1988). The smaller the CV, the more regular is the discharge. The CV was calculated for the steady-state or last half of the response and was between 0.4 and 0.5 in each case. This places these units at the high end of the CV range for transient choppers and suggests that these units could be labeled Ch-t. These units phaselock to low-frequency tones, as measured by the synchronization coefficient (Goldberg and Brown 1969), as is seen for a 1,100-Hz tone in Fig. 4 D.
The response pattern varies with stimulus level although it remains On-g for a 60-dB range of levels (Fig.5, left). The corresponding interspike interval histograms (ISIHs) are similar in shape to those of primarylike units although more so for the 60 dB SPL case that is nearest the maximum response (middle). Regularity analysis (right) suggests that these units behave similar to Ch-t units (transient choppers). In response to a 50-ms tone, the CV is initially low but increases over the first 20 ms of the stimulus and remains relatively constant over the last half of the stimulus. At both low and high SPLs the late response is too weak to provide a valid regularity analysis as seen in the top and bottom panels. The final CV is ∼0.5, which is lower than primarylike and higher than most chopper units. Note, the range of CVs of Ch-t units overlaps with that of these type II units.
The pattern of the PSTHs varies relatively little with frequency. At high stimulus levels, an onset response remains irrespective of stimulus frequency indicating inhibitory events that allow at most an initial spike (Fig. 6). For low-level and non-BF stimuli, later spikes are largely absent, again suggesting a possible inhibitory input. The principal difference between low and high level responses is that low-level onset responses are less well timed than at the high levels where they remain well timed up to 80–90 dB SPL. This could be interpreted as suggesting the combined inputs of several auditory nerve (AN) fibers are sufficient to excite the cell. The loss of even the onset spike at high levels could be due to incomplete recovery from the inhibition due to prior stimuli. Long off times were not employed to test this latter possibility.
The response surfaces for four On-g units are combined in Fig.7 A. The remarkable aspect of each response surface is how limited it is in spectral extent. This is perhaps better captured in the contour plots shown in Fig.7 B. The low-frequency expansion of the response field at high levels or tail region found in most type I and type III FTCs is either absent or very limited.
The frequency threshold curves for seven On-g units were obtained from the response surface using a 10% of maximum rate criteria are shown in Fig. 8. Although some of the contours are not closed at this low discharge rate, all are closed at higher rates (e.g., Fig. 1 D). The high-frequency slopes are nearly vertical for all units. The low-frequency slopes (f < BF) are steep for the higher BF units while extending an octave or more for the low BF (<1–2 kHz) units. This may imply that low-frequency inhibitory inputs are fewer or less effective. Although there were exceptions, units typically have a threshold in the 20–40 dB SPL range with an average threshold of 27 dB SPL. The maximum spike rate occurs between 40 and 60 dB SPL.
Intracellular recordings from these cells has proven very difficult, a result likely due to their relatively small size. In a few instances recordings were made over the entire response field using one short tone per stimulus point as shown in Fig.9. There is a sustained depolarization during the stimulus followed by an after hyperpolarization. The use of KCl electrodes likely altered the chloride potential resulting in the inversion of any inhibitory postsynaptic potentials. However, some IPSPs appear to be present in Fig. 9 (2 bottom left panels). In addition to the afterhyperpolarization, a hyperpolarization was seen during the stimulus for some non-BF frequencies (e.g., 22 kHz between 50 and 90 dB SPL and 14 kHz at 60 and 70 dB SPL). Intracellular recording typically lasted <10 min. Resting potentials were at best −50 mV and often more than −30 mV and usually reduced to ∼0 mV during the recording.
The intracellular response of an On-g unit to the downward direction of a FM sweep shows that there is a sustained depolarization while the stimulus is in the response region of the unit (Fig.10). There is also a small hyperpolarization on each side of the excitatory region, suggesting inhibitory inputs, that is clearer in the filtered response shown in the inset. There is a depolarization in the lowest frequency region of the sweep that did not produce any discharges.
Five On-g cells reconstructed after injection were drawn using a computerized reconstruction system (Fig.11). Somas are elongate with longest diameter between ∼13 and 16 μm (average = 14.3 μm). The cells have a restricted dendritic and axonal field along the tonotopic axis of the DCN. The apical length of the dendritic field extended ∼200–270 μm reaching into the molecular layer. The basal dendritic field extends 50–230 μm. The axon (based on the bouton-like appearing swellings) of one of the vertical cells is labeled in red in Fig. 12. It is clear in the horizontal view that the axon projects rostral and toward the molecular layer while being restricted along the tonotopic axis. This polarization of the local axonal projection typically was seen when the three orthogonal views of the cells were examined. There is a partial-to-complete segregation of the local axonal field. Whether there is any physiological correlate of this morphology is not apparent at this time. The most that can be said is that the On-g dendrites appear to be limited to an isofrequency lamina.
Cells described here appear to be vertical cells or corn cells based on their depth in the nucleus and their dendritic and axonal morphology (Brawer et al. 1974; Lorente de Nó1981). Vertical cells are the most numerous cell type in layer III of the DCN (Lorente de Nó 1981) (Fig. 6-18). They have dendritic fields that are restricted along the tonotopic axis and are concentrated near fusiform cell bodies and have dendrites that don’t appear to enter the molecular layer. However, this latter feature could be a result of the use of the Golgi stain on neonatal cats (8-day-old cat by Lorente de Nó) versus the use of adult material here. Lorente de Nó further states the projection pattern of their axons justifies the existence of several subtypes of vertical cells. About one-third of the cells in the deeper part of layer III have axons that ramify and join the lateral ventrotubercular tract, whereas cells in the upper part of layer III only have local axons that often penetrate layer II (fusiform cell layer or FCL). Cells described here are likely members of the vertical cell group that lies in the upper portion of layer III of the DCN and that have only local axons. In fact, they bear no resemblance to any other cell type in this region e.g., the fusiforms, stellates, Golgi, or granule cells.
A transverse section of the DCN illustrates the relation of one vertical cell to the overlying fusiform cell layer (Fig.13 A). A cross-section of the CN with a composite of three labeled vertical cells is shown in Fig.13 B. The cross-section corresponds roughly to transverse section T130 in the Brawer et al. (1974) atlas. The cells shown were estimated to be located within ±160 μm of section T130. It is notable that the lowest BF unit (leftmost in figure) has a greater mediolateral extent than the higher BF units. This dendritic arrangement may underlie the wider tail region in the FTCs of low-BF On-g units and is similar to the those seen in low-BF fusiform cells (Fig. 15 A) (Rhode et al. 1983b).
Although the axonal arbor appears to be mostly limited to the fusiform cell layer, there were some boutons in the molecular layer. One feature of the axonal fields was the prominent clustering of boutons mostly on the basal side of the somas of fusiform cells (Fig.14). Numerous boutons also were seen in the neuropil. No axonal branches to the anteroventral cochlear nucleus (AVCN) were seen, a result that could be due to incomplete filling of the axon. The axon collateral, shown in the transverse section in Fig.12, could be directed toward the AVCN but faded rapidly. Therefore it is inconclusive whether it could be part of the tuberculoventral tract. Dendritic structures of these cells include mace-like formations similar to those reported by Brawer et al. (1974) for the vertical/corn cells in the DCN (Fig. 14 C).
The computerized anatomic reconstruction system allowed us to view the cells from all directions to determine whether there is any morphological polarization. The extent of the dendritic tree is 100 μm in the mediolateral direction for three neurons and 220 μm and 330 μm for the other two. The 100-μm extent corresponds to ∼1/3 octave based on a tonotopic map for the DCN (Spirou et al. 1993). Thus both the physiology and the morphology would support convergence of AN fibers with a restricted set of BFs on these cells. Furthermore the fact that wideband noise inhibits these cells at all levels indicates that other cells provide strong inhibitory inputs to these units.
On-gs: a subset of vertical cells
This study was based on the demonstration that units in the DCN that responded to short tones at BF with an On-g pattern fulfilled the criteria for type II neurons and that they arise from vertical cells located just below the fusiform cell layer. Cells having an On-g response pattern have a fairly uniform set of response features (see Table 1). These vertical cells are likely a subset of the vertical cell population in the DCN given that there are two or more vertical cell populations, one of which is located high in layer III and which only has a local axon (Lorente de Nó 1981). This is likely the population identified here.
Immunohistochemical studies also support the existence of more than one type of vertical cell. One population is glycinergic while both glycine and GABA are colocalized in another (Osen et al. 1990). Several studies have demonstrated that the vertical cells that give rise to the ventrotuberculo tract are glycinergic (e.g.,Saint-Marie et al. 1991; Wickesberg and Oertel 1990). Studies also have shown that the major inhibitory neurotransmitter affecting type IV neurons is glycine while GABA has a stronger effect on nontype-IV neurons (Caspary et al. 1987; Davis and Young 1998). The neurotransmitter of the cells studied here was not identified but is likely to be glycine.
Pentobarbital anesthetic effects
One concern in studies of the CNS is whether the anesthetic altered the response behavior to the point of rendering the classification scheme arbitrary relative to that used in the decerebrate state. Although in the ventral CN there appears to be relatively little difference between anesthetized and decerebrate conditions, this does not appear to be the case in the DCN (Evans and Nelson 1973). This was demonstrated when Young and Brownell showed that the response of a type IV unit could be converted to that of a type III unit by the application of a barbiturate during recording (Young and Brownell 1976). However, there remain considerable similarities as all the response patterns are present in both barbiturate anesthetized and awake states (Rhode and Kettner 1987).
Every On-g/type II unit in the present study exhibits a sharp reduction in discharge rate beyond the peak in their rate-level function while few do so in decerebrate cat. This leads one to speculate that GABA, an inhibitory neurotransmitter that is potentiated by pentobarbital anesthesia, is being potentiated in the CN (Potashner and Lake 1981; Schulz and Macdonald 1981). Bicuculline, a blocker of GABA, strongly enhances on-best frequency inhibition of type III units and has little effect on type IV units (Davis and Young 1998). This suggests that a disinhibitory process causes a glycinergic pathway to be expressed resulting in nonmonotonic rate-level function of the type III unit. It also suggests that if vertical cells provide inhibitory input to both type III and type IV neurons then there must be two groups with one having a strong GABAergic (inhibitory) input. In contrast, in both decerebrate and pentobarbital-anesthetized gerbils, the slopes of the rate-level functions are the same and less than seen here in cat. This could imply that there are species differences in DCN circuits. In addition to species differences, different electrodes, recording conditions, depth of the recording, search stimuli, and experimental goals could account for a sampling bias. There are instances of similar responses under both anesthetized and decerebrate conditions though the frequency of occurrence of different response patterns varies. Nevertheless, in cat, one hypothesis is that some of the differences seen in type II unit behavior in different studies are due to the effect of anesthesia.
Type II features
An RNR < 0.3, SR < 2.5 spikes/s, monotonic to slightly nonmonotonic, and a V-shaped FTC have been the standard features used to identify type II units (Joris 1998; Young and Voigt 1982). The use of this set of response features results in a more inclusive category than the present study that adds 0-shaped FTCs, strongly nonmonotonic RLFs, and an asymmetric response to FM signals as features of type II units.
The inclusiveness of the earlier definition of type II units is illustrated in two studies where several type II units displayed a pauser pattern (Joris 1998; Shofner and Young 1985). Pauser response patterns arise from fusiform cells based on labeling studies (Rhode et al. 1983b; Smith and Rhode 1985). Pauser units have been shown to have varying amounts of inhibitory inputs as measured by the width and depth of their inhibitory sidebands (Rhode and Greenberg 1994). The variable nature of the afferent and intrinsic connectivity resulting in varying amounts of excitation and inhibition may result in the response of some fusiform cells satisfying type II criteria. Fusiform cells are a distinct morphological type with axons projecting out of the CN, having local collaterals in the DCN, and containing round vesicles in their synaptic boutons, indicating they are excitatory projection neurons (Smith and Rhode 1985).
Type II units also have been described as having slightly nonmonotonic rate-level functions to BF tones (Shofner and Young 1985). Shofner and Young showed one example out of four of a type II unit in which the rate-level function decreased to zero at levels > 65 dB SPL, i.e., a strongly nonmonotonic RLF. This was the only unit for which the PSTHs shown would be classified as On-g. The other three units exhibited pauser and/or chopper response patterns with the hallmark features of type IIs [a nonmonotonic tonal-rate-level function and a weak (or no) response to WBN, along with little or no spontaneous activity]. They concluded that there doesn’t appear to be a characteristic PSTH for type II units.
Godfrey et al. (1975) defined an On-s response pattern in DCN units that likely corresponds to On-g units. Their spontaneous rate was <2.5 spikes/s, and for stimulus levels >25 dB SPL, the discharge rate decreased, which suggests these were type II units. The response patterns were intermediate between primarylike and chopper patterns.
Type II units exhibit little or no response to noise although noise has been shown to inhibit firing in On-g units. The effect of a simultaneously presented WBN masker during tone presentation demonstrated a reduced firing rate when compared with the response to the tone alone and the normal nonmonotonic RLF becomes a monotonic RLF (Fig. 6) (Rhode and Greenberg 1994). The response to noise varies in degree between studies and some of the variability may be due to the species used. Davis et al. (1996) showed a stronger response to noise in decerebrate gerbil (average RNR ∼ 0.3) than shown here for cat. The cat and gerbil data show almost no overlap when RNR and the normalized descending slope of the RLF are the coordinates of a scatter diagram (Fig.15). The gerbil tonal-rate-level nonmonotonicity is similar to that shown in decerebrate or unanesthetized cat by Shofner and Young (1985).
Frequency threshold curves
Type II FTCs have been characterized as similar to those of type I units, i.e., having a tip and tail region. However, a rapid reduction in firing with increasing stimulus level beyond the peak in the rate-level function results in narrow and closed contours for the FTCs, i.e., 0-shaped FTCs rather than V-shaped FTCs. This difference may result from the fact that in the early studies, stimuli were not presented at high enough levels to completely shut off the response. This cannot be the entire explanation because most of the slopes of the rate-level functions in these early studies are substantially less than those found for the On-g units here. Inhibitory input to these cells may be more effective under pentobarbital anesthetic than under decerebrate conditions as discussed earlier. Because there are instances in the earlier studies of nonmonotonic rate-level functions with slopes similar to those reported here, it is entirely possible that the explanation lies in there being more than one response pattern and/or morphological entity corresponding to type II designation. Different populations may have been sampled especially because different electrodes and sometimes different species were used in different labs.
Lorente de Nó (1981) described two types of vertical cells in mouse DCN: one with only a local axon and the other giving rise to the tuberculoventral tract. This tract has been shown to give rise to inhibitory synapses on bushy and stellate cells in the VCN and locally to neurons in the DCN (Wickesberg and Oertel 1990; Zhang and Oertel 1993). Young (1980) employed electrical stimulation of several regions of the CN to determine whether type II units project either out of the DCN via the DAS or to the AVCN via the tuberculoventral pathway. Only 1 of 30 type IIs could be activated antidromically from DAS stimulation, whereas 6 of 32 could be stimulated antidromically from the posterior AVCN. The fact that antidromic stimulation was effective in firing 6 of 32 type II neurons could be interpreted to mean that there is more than one type of type II neuron. However, it is also possible that the electrical stimulation was not properly positioned in those instances where it failed to fire the type IIs because retrograde labeling studies indicate that nearly all vertical cells in the deep DCN project to AVCN (Saint Marie et al. 1991).
Although the present study indicates that the vertical cells just below the fusiform cell layer are type II, Young found 6 of 29 type II/III units were located in the deep layer of the DCN. Saint Marie et al. also found the majority of the vertical cells that project to the AVCN lie in the deep layer. The present results are inconclusive with regard to a projection to the VCN. It is possible that only the short axon type of vertical cell described by Lorente de Nó was labeled in this study.
Profile of On-g/type II units
Units with the On-g response pattern form a well-defined category. Although a few features are sufficient to identify them, they have a multitude of features that distinguish them. As listed in Table 1, the spontaneous rate is ∼0 spikes/s, and the tonal-rate-level function is nonmonotonic with an initial rapid rise to a peak rate of discharge, followed by a rapid decrease in firing rate to 0 spikes/s at an average rate of decrease of −2.5% per decibel. The weak response to noise is expressed by a RNR that on average is 0.05. Another feature is the asymmetric response to a FM sweep with the high-to-low direction, resulting in nearly double the number of discharges than the opposite direction (at slow sweep rates). The response surface is a relatively circumscribed mound on the frequency-level plane that can be represented by a series of contours in the plane. This implies that the FTC is 0- rather than V-shaped. The PSTH also can be classified as Ch-t at the maximal response condition for many of the type II units studied, although outside of this restricted region it is On-g. That is, initially there is a prominent chopper pattern followed by a steady-state discharge where the coefficient of variation (of the ISIH) increases to between 0.4 and 0.5. The Q10s of these units are close to those of AN fibers. A difference is that the inhibitory inputs to On-g units prevent the widening of the FTC with increased stimulus level, thereby eliminating the tail region of the FTC. That is, Q40s are larger than those of AN fibers as also shown by Young and Voigt (1982).
Many of these response features are common to several studies of type II units (Davis et al. 1996; Ding and Voigt 1997; Godfrey et al. 1975; Joris 1998; Shofner and Young 1985). Whether those type II units that responded with a P/B pattern satisfy the expanded type II/On-g criteria, such as directional preference for FM sweep direction and closed contour FTCs, would be interesting to know. It is noteworthy that the discharge rate is much lower in the present study than in the others listed in Table 1.
Morphologically these cells fit the description of the vertical cells that have mace-like terminations of the dendrites (Brawer et al. 1974; Lorente de Nó 1981). Both the dendrites and axons are restricted to an isofrequency lamina and the somas are located just below those of the fusiform cells. The axons appear to locate rostral to the dendrites. Small pericellular nests, formed by axon collaterals, occur around the basal side of the soma and proximal basal dendrites of fusiform cells.
The cells labeled here are a subset of the total vertical cell population by virtue of only being located close to the fusiform cell layer. These units exhibit an On-g response pattern and consistently satisfy type II criteria. Because the PSTH often could be labeled Ch-t and an On-s response pattern is another name for On-g, there is general agreement of what constitutes the response pattern for the majority of type II units reported to date. Differences are likely due to sampling a different population of vertical cells and/or the effect of anesthesia.
Role of vertical cells
Several functional roles for the vertical cells have been described. A role in monaural echo suppression has been advanced in mouse (Wickesberg and Oertel 1988). It is unlikely that the properties of type II units shown here will support the hypothesis of a role in echo suppression for several reasons: their transient temporal response parameters, strong nonmonotonic rate-level behavior that means they would not suppress at high levels of stimulation, and a lack of response to noise that implies they could not participate in echo suppression of noise. However, there may be other populations of vertical cells that lie in the deeper DCN and that may have appropriate behavior.
Nelken and Young (1996) speculated that type IIs play a role in shaping the type IV neurons as threshold detectors. Type IVs have among the lowest thresholds in the CN and are often shut off at higher levels of stimulation. Given their nonmonotonic RLFs that peak at intermediate levels, type II units probably would serve to shut off the type IV units in the midrange of their response in the intensity domain. Other suggestions include a role in the determination of elevation of a sound source by participating in the detection of notches in the sound spectrum.
Models of DCN circuitry
A circuit is illustrated in Fig.16 that is similar to earlier models except for the addition of a second On-ch unit in the deep DCN (Davis and Voigt 1996; Nelken and Young 1996). As was shown in Fig. 14, there appears to be significant somatic connections on fusiform cells from vertical cells that are presumed to be inhibitory (indicated by open arrowheads in the figure). On-ch units, which are multipolar cells in the posteroventral cochlear nucleus (PVCN) that project to the DCN, have wide FTCs that indicate convergence of several AN fibers of varying BFs and a wide dynamic range that is necessary to inhibit both type II and type IV at high stimulus levels (Smith and Rhode 1985). These cells were shown to have pleomorphic vesicles in their terminals, which is taken to imply an inhibitory function. However, the terminations of their axons has yet to be determined. On-ch units have been suggested as a candidate for the role of wideband inhibitory (WBI) units due to their wide dynamic range and broad FTCs (Nelken and Young 1994; Winter and Palmer 1995a).
The nonmonotonic RLF of type II units easily could explain some type IV RLFs, e.g., the one illustrated in Fig.17. The type IV unit has a threshold near 0 dB SPL and its RLF rises rapidly to an interim maximum that is near the level at which the type II unit begins to discharge. As the maximum of the type II RLF is approached, the type IV RLF approaches an intermediate minimum. Then as the RLF of the type II unit decreases beyond its maximum, the type IV RLF begins to increase (>45 dB SPL) as would be expected if an inhibitory input was being reduced.
Despite the plausibility of type II-type IV RLF relation shown in Fig.17, the entire type IV response surface must be accounted for. This is an especially a difficult task when the large range of response behaviors of fusiform cells is considered. However, there is no reason that type IV response patterns should be entirely determined by a (subset) of type II neurons.
Although the data shown in Fig. 17 illustrate an instance of a type II–type IV interaction that may not require a WBI input to the type IV neuron, the response of the type IV neuron in Fig.18 does. Rate-level functions in Fig.18 A explain how the weak response to noise of the type II unit comes about. That is, the exuberant response to wideband noise of the On-ch neuron means there is a stronger inhibitory synaptic drive to the type II than the excitatory synaptic drive from the AN (Joris 1998; Winter and Palmer 1995b). The frequency threshold curves in Fig. 18 C illustrate why the frequency response of the type II is narrow. The frequency response on the On-ch unit is substantially wider than that of a typical AN fiber. This means that the inhibition is greater than the excitation especially off the center frequency of the AN input. The response features of a type IV unit are shown in Fig. 18, B andD. In this case, both the type II and type IV RLFs are nonmonotonic. Therefore an additional inhibitory input to the type IV neuron is required to reduce the firing of the type IV when the type II inhibitory drive itself is being reduced. The WBI easily can fulfill this role, in fact, the WBI is entirely adequate to explain the behavior of both the type II and type IV in this instance.
An additional reason for the WBI input directly to type IV units is that the latter often have wide inhibitory sidebands (0.5 to >2 octaves). Based on the narrow response field of the type IIs, this “wide” inhibitory bandwidth of the type IV unit cannot be explained easily. Lateral inhibitory sidebands could be implemented by convergence of several type II units of varying best frequencies, which would explain narrow inhibitory sidebands but not the high-level inhibition because the input from type II units would decrease to nearly nothing. Furthermore it is shown here that the axons of type IIs are limited largely to an isofrequency lamina. One or more On-ch units may provide the necessary inhibitory input. Furthermore the On-ch units in the deep DCN also may play a role.
One population of vertical cells has been shown here to likely provide one of the inhibitory inputs to fusiform cells. However, as noted earlier, there are at least two vertical cell types, one with a local axon and one that gives rise to a projection to the tuberculoventral tract. In addition, there are likely two populations of cells that could play the role of the wideband inhibitor. Given the normal random variations in synaptic connectivity, these candidates for members of this circuit provide adequate possibilities to explain the variety of response features seen in the projection neurons of the DCN without invoking the elaborate inputs from neurons in the molecular layer.
Special thanks go to histologists J. Ekleberry, J. Meister, and I. Sigglekow, to R. Kochhar for programming support, to K. Yentner for neuron reconstruction, and to A. Recio for participating in some of the experiments.
This study was supported by a National Institute of Deafness and Other Communications Disorders Grant NS-17590.
Address for reprint requests: Dept. of Physiology, 1300 University Ave., University of Wisconsin, Madison, WI 53706.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 1999 The American Physiological Society