We labeled eight MNTB cells intraaxonally after obtaining their physiological response properties. In agreement with previousqualitative anatomic descriptions (Jean-Baptiste and Morest 1975;Lenn and Reese 1966; Morest 1968, 1973), our measurements showthat calyceal coverage can take up one-fourth to one-half of thetotal surface of the cell body and primary dendrites. In addition,we have shown that nonexcitatory inputs also concentrate on thecell body and consistently cover around one-fifth of the surface,but the source of this input or means of activation is unknown.Our data also show that inputs with both excitatory and inhibitoryfeatures are rare on the distal dendritic tree. We also have shownfor the first time at the ultrastructural level that the terminalsof these MNTB axons contain nonround vesicles and, when observedin the LSO, terminate on the somata and primary dendrites of LSOprincipal cells.
As has been previously described in cat and other species (Banks and Smith 1992; Kuwabura and Zook 1991, 1992; Schofield andCant 1992; Sommer et al. 1992; Spangler et al. 1985), MNTB cellssend their axonal projections only to areas of the ipsilateralbrain stem, namely (in cat) the ventral nucleus of the laterallemniscus, the lateral and medial superior olives, the dorsomedialand ventromedial periolivary nuclei, and the MNTB. The only previousin vivo intracellular study (Sommer et al. 1992) indicated thatthe rat LSO receives MNTB inputs with best frequencies from 47kHz down to 14 kHz. Our labeled cells show that, in the cat,this frequency range extends down to 635 Hz. In addition, despitean admittedly small sample, our labeled cell data provides evidencethat the MNTB projection to the MSO may be frequency specific-cellswith CFs below <13 kHz innervate the MSO, cells with CFs >13 kHzdo not.
Physiologically our short tone data from cells, positively identified by intracellular staining, show that MNTB cell responsesare like those of globular bushy cells with similar best frequencies.We also have included physiological data from cells in the MNTBrecorded extracellularly that displayed prepotentials. Prior tothis, the only evidence that the complex spikes seen in the MNTBarose from the MNTB principal cell and its calyx was providedby Guinan and Li (1990). By stimulating either LSO or DMPO whilerecording a complex spike unit in the MNTB, they could antidromicallyactivate a spike, designated the "A" spike. This A spike resembledthe portion of the complex waveform, designated the "C2" component,that was believed to be the postsynaptic MNTB cell spike. Whenthe complex spike was activated by stimulation of the trapezoidbody at the midline, thus driving the globular bushy axon/calycealsynapse and presumably activating the MNTB principal cells, theantidromic A spike was refractory and could not be driven for~1 ms, implying that the A spike and C2 component were both generatedby the MNTB cell. We felt that this evidence, together with ourpositive identification of MNTB principal cells with physiologycomparable to the complex waveform units, was sufficient to permitour inclusion of the extracellular data.
Globular bushy cell inputa comparison of responses
The data available from our MNTB cell sample show the results of what is apparently a good transmission of temporal informationacross the synapse. Globular bushy cells carry a variety of accuratelytimed information about low- as well as high-frequency auditorystimuli. Our previous reports (Joris et al. 1994a; Smith et al.1991) have shown that low-frequency globular bushy cells can phaselock to the stimulus waveform of a short tone at their CF withremarkable precision (as measured by the synchronization coefficient)and often will entrain (generate a spike for every cycle of thestimulus waveform). Because our sample of MNTB cells with CFs<1 kHz is so small, it is difficult to compare GBC and MNTB cellpopulations in terms of phase locking capabilities. Neverthelessthe positively identified MNTB cell with a CF <1 kHz showed "enhanced"synchrony, i.e., its maximum synchrony (0.89) was high relativeto that of most auditory nerve fibers at that frequency. Also,maximum synchronization for the one MNTB cell in our sample witha CF between 1 and 3 kHz was at the lower edge of the range observedin the auditory nerve. These observations are consistent withsynchronization in GBCs, which also show enhanced synchronizationat CFs <1 kHz and poor synchronization at intermediate CFs (1kHz < CF < 3 kHz). GBCs with CF >3 kHz show enhanced synchronizationto tones in the low-frequency tail of the tuning curve (Joriset al. 1994b), but we did not test this in MNTB cells. As seenin Fig. 10, the onset component of the MNTB cell responses waswell timed, although as a population not as well timed as globularbushy cells (Fig. 11C). Finally, GBC and MNTB cells show a slightimprovement in envelope phase-locking compared with the auditorynerve, and overall the two populations are indistinguishable intheir envelope phase-locking properties (Joris and Yin, unpublisheddata).
As described in RESULTS, the percentage of MNTB cells we recorded from, with CFs <3 kHz, was significantly lower (4%) thanthe percentage of globular bushy cells (34%) we previously recordedfrom with similar CFs. We feel that, even though the frequencymap of Guinan et al. (1972b) indicates that low frequencies areless well represented than higher frequencies in MNTB, 4% is probablynot a true reflection of the actual number of MNTB cells withCFs <3 kHz but is probably biased by our surgical approach andelectrode placement. In these experiments, the electrode was insertedinto the brain stem just lateral to the pyramidal tract. In someinstances, the electrode entered the brain near the rootlets ofthe abducens nerve perpendicular to the surface of the brain.Using this approach the more laterally situated, lower frequencyregions of the MNTB are not traversed. Even when we angled theelectrode so that it diverged from the midline with increasingdepth the lateral aspect of the MNTB was often still not reliablysampled. This would probably also explain the disparity in meanspontaneous rates between our GBC (12.4 spikes/s) and MNTB cell(27 spikes/s) populations. Many of the low CF globular bushy cellswe labeled with CFs <3 kHz had spontaneous rates of 0 spikes/s.Our two examples of MNTB cells with CFs <3 kHz also had spontaneousrates of 0 spikes/s, and it is likely that, had we sampled a higherpercentage of low CF MNTB cells, the means would be closer.
Globular bushy cell calyceal and other inputs
In our previous report (Smith et al. 1991), we noted that the labeled calyx of a globular bushy cell axon covered ~24% ofits postsynaptic MNTB cell body surface, while the total (labeledand unlabeled) round vesicle coverage of the cell body was 26.4%.In this case, the great majority of the excitatory input was thelabeled calyx, providing evidence that this cell received onlyone calyx. In the present study, the total round vesicle terminalcoverage for three of our labeled MNTB cell bodies was 39.6, 43.8,and 48%. What does this say about the excitatory synaptic inputto the MNTB cell body? 1) In the four cases, we have examinedin detail the extent of somatic coverage by excitatory terminalshas varied considerably, ranging from slightly more than one-fourthto slightly less than one-half. Thus if each MNTB cell is receivingonly one calyx, then there may be considerable variation in amountof surface of MNTB cells covered by one globular bushy calycealterminal. 2) Alternatively, some MNTB cells receive two calycesor 3) some MNTB cells receive one calyx and a lot of other noncalycealround vesicle inputs. These could potentially arise from the small"precalycean" terminals of other globular bushy cell axons thatmake smaller terminal endings in the MNTB (Morest 1968; Smithet al. 1991). The axons of other projection neurons of the cochlearnucleus, with round vesicle containing terminals, do not ventureinto the MNTB so they are unlikely sources for additional excitatoryinputs. These include the spherical bushy cells (Smith et al.1993b), the stellate cells the axons of which exit the cochlearnucleus via the trapezoid body and that respond in a chopper fashionto short tones (Smith et al. 1993a), the octopus cells of theposteroventral cochlear nucleus (PVCN) and the giant and fusiformcells of the dorsal cochlear nucleus (DCN) (Joris et al. 1992).Little information is available on the other potential sourcesof excitatory inputs to MNTB in the brain stem. Labeled MSO principalcells in the guinea pig (Smith 1995) never gave off collateralsto MNTB, and there is only one report of an occasional projectionto MNTB from labeled rat LSO principal cell the axon of whichcrossed the midline (Banks and Smith 1992).
We also measured the nonround vesicle coverage of the same four cell bodies. The nonround vesicle coverage of the cell receivingthe labeled calyx reported previously (Smith et al. 1991) was21.6%, and 21.3, 22.8, and 20.4% for the three cells in the presentstudy. Thus it would appear that the presumably inhibitory "nonroundvesicle" input to the soma is quite heavy and, at least in termsof surface coverage, is constant from cell to cell. Anatomically,it has been shown in cat and guinea pig that glycine immunoreactiveterminals are rather sparse in the MNTB (Adams and Magnaini 1990;Helfert et al. 1989; Wenthold et al. 1987) as is glycine receptorlabeling (Zarbin et al. 1981) while -aminobutyric acid (GABA)immunoreactive terminals are quite numerous (Adams and Magnaini1990). This agrees with our observations here and similar observationsin the rat (Banks and Smith 1992) that the MNTB cells, which arethe major source of glycinergic innervation to the auditory brainstem, only give off occasional sparsely branching collateralswithin the home nucleus. Little evidence is available as to thesource of any of the GABAergic terminals nor is there evidenceas to whether they are activated by auditory stimuli. Kuwaburaet al. (1991) described axon collaterals of cells in the rodentventral nucleus of the trapezoid body that could branch in MNTB,but it is not known what transmitter these cells manufacture.In vitro physiology in the mouse has shown that bath applicationof either GABA or glycine decreases the input resistance in onlyabout half of the mouse MNTB cells tested (Wu and Kelly 1995).Further in vitro data indicated that shock stimulation of thetrapezoid body at the midline or ipsilateral to the MNTB in therat (Banks and Smith 1992) but not the mouse (Wu and Kelly 1991)would occasionally evoke inhibitory postsynaptic potentials (IPSPs)that could be blocked by the glycine antagonist, strychnine. Ourin vivo data on the response of MNTB cells to binaural stimulationas the sound to the contralateral ear was held constant whilethe sound to the ipsilateral ear was increased (Fig. 2) showedno inhibitory affects on the contralaterally driven output. Inother in vivo experiments, Guinan and Li (1990) also could notdemonstrate any sort of inhibitory affect on the calyceal drivenMNTB cell output either by auditory stimulation to either earor by shock stimulation of the trapezoid body. Some of our MNTBshort tone responses show an apparent sag in the sustained response(Fig. 10), but it would be premature to propose an inhibitory inputas the cause. Thus inhibitory inputs on MNTB cells appear to bepresent and functional but there is no evidence of activationby simple auditory stimuli.
Calyceal and postsynaptic physiology
In addition to the anatomic specializations, some physiological specializations also apparently have developed both pre- andpostsynaptically to aid in the secure and rapid transfer of accuratelytimed information. On the basis of whole cell patch recordingsfrom calyceal terminals, the calyx possesses a large, high-voltage-activatedcalcium conductance with fast activation and deactivation kinetics(Borst et al. 1995) that would be required for rapid synaptictransmission. In addition, preliminary reports (Forsythe 1994)showed that, in current clamp, the presynaptic terminal can spikeat high rates in response to depolarizing current pulses becauseof a fast delayed rectifier potassium current that allows rapidspike repolarization. Such a rapid response capability also mightbe requisite for rapid synaptic transmission. The synaptic currentgenerated postsynaptically by the calyx is fast and very large,averaging ~5 nA (Borst and Sakmann 1996), but this current amplitudediminished significantly during repetitive stimulation such thatthe calyceal synapse could not make the MNTB cell fire reliablyto a 100-Hz shock stimulation of the bushy cell input. It wouldappear, however, that such poor postsynaptic following is probablya function of the immature preparation used (8- to 10-day-oldrats) because shock rates of the bushy cell input generating MNTBcell spikes were reliably followed to well over 500 Hz in a maturemouse slice preparation (Wu and Kelly 1993). Postsynapticallythe MNTB cell has specialized potassium conductances (Banks andSmith 1992). One has been designated ITEA, which might allow forrapid spike repolarization, the other IDTX, which might permitonly a single spike to occur in response to the huge calycealinput (Brew and Forsythe 1995).
MNTB cell output
Our anatomic data on the projections of MNTB principal cells correspond well with Spangler et al. (1985). With one exception,all axons that could be followed to LSO innervated that structurein a fashion that was consistent with the proposed frequency mapof the LSO (Guinan et al. 1972b). Our one example of a low-frequencyMNTB cell sending a projection to the low frequency limb supportsthe contention (Finlayson and Caspary 1991; Joris and Yin 1995)that some low-frequency LSO cells are inhibited by contralateralstimulation.
The terminals we examined at the EM level had nonround vesicles and terminated on large dendritic profiles or the cell bodiesof principal cells. Cant (1984) reported that almost three-fourthsof the surface of the LSO principal cell body and proximal dendritesare covered with synaptic terminals that are almost exclusivelythose containing small vesicles many of which are flattened orcylindrical. Our labeled synaptic terminals from identified MNTBprincipal cells correspond to the description both in locationand vesicle content (Cant 1984) and provide the first direct evidencethat these terminals are derived from axons of MNTB cells. Inthe rat, Moore and Caspary (1983) showed that the contralaterallyevoked inhibition of ipsilaterally evoked excitation could beblocked by application of the glycine antagonist strychnine. Similarresults were obtained in the slice preparation, by Wu and Kelly(1991), with IPSPs evoked by electrical activation of MNTB. Thusour anatomic data, showing that the axon terminals of MNTB cellsto the LSO have nonround synaptic vesicles, provide an importantpiece of data linking the presumed inhibitory function of MNTBonto ipsilateral LSO to the anatomic data of Cant (1984) showingflattened vesicles on the soma and proximal dendrites of LSO cells.
For axons that could be followed through MSO the ones with the three lowest CFs sparsely innervated the nucleus in appropriatefrequency regions, whereas the three with the higher best frequenciesdid not. Clark (1969) and Schwartz (1980) have reported the presenceof terminals with flattened vesicles on principal cells in catMSO. The innervation of MSO by MNTB cells also has been notedanatomically in rat, bat, mouse, gerbil, and guinea pig (Banksand Smith 1992; Kuwabura and Zook 1991, 1992; Smith 1995), andIPSPs, evoked by MNTB shock, in MSO principal cells were blockedby strychnine (Grothe and Sanes 1993, 1994; Smith 1995), confirmingthe glycinergic nature of the input. The function of such an activelydriven inhibitory input to the MSO from MNTB remains speculative.Several possibilities include a MNTB input the CF of which isslightly off the CF of the innervated MSO cell could serve tosuppress coincident excitatory inputs at non-CF frequencies; mediationof precedence-like effects; and changing the MSO cells restingmembrane potential during contralateral ear stimulation.
All MNTB neurons with darkly labeled axons had a collateral that could be followed rostrally to VNLL. Unfortunately nothingis known about the distribution of synaptic terminals on cellsin the VNLL or their ultrastructure nor is there any informationon why VNLL cells might require a contralaterally driven inhibitoryinput when much of the excitatory input to this nucleus is alsofrom the contralateral side. Collaterals also were sent to thetwo nuclei medial to the MSO, the DMPO and VMPO. In the cat, GBCsoften send an excitatory collateral to synapse on cells in theDMPO (Smith et al. 1991), so, as with the VNLL, we again notea region receiving both a contralateral excitatory input fromcochlear nucleus and an inhibitory input from MNTB, a nucleusthat is driven by contralateral excitatory input. The rodent superiorparaolivary nucleus (SPN) is a large periolivary nucleus dorsomedialto the MSO that is considered the homolog of the cat DMPO. Itreceives input from both cochlear nuclei (Thompson and Thompson1991) as well as a strong topographic input from MNTB principalcells (Banks and Smith 1992), and its cells project primarilyeither to the inferior colliculus or cochlear nucleus (Schofield1991). Although some extracellular recordings were done in vivoon cells in the gerbil SPN and cat DMPO (Guinan et al. 1972a;Spitzer and Semple 1995), the physiology of a large populationof cells in this nucleus has never been explored in great detailand consequently the possible effects of the MNTB input, on auditoryresponse properties of cells here, are not known. We also occasionallynoted a collateral being given off within the MNTB and, as notedabove, inhibitory synaptic events can be elicited in rat MNTBcells by shocking their bushy cell input (Banks and Smith 1992),but it is not clear that these were due to collaterals of MNTBcells.