Introduction

Olivocochlear neurons form a descending system that originates in the superior olivary complex of the brainstem and projects to the cochlea. Two subsystems of olivocochlear neurons exist: a medial olivocochlear (MOC) system that originates in the medial portion of the superior olive and projects mainly to outer hair cells, and a lateral olivocochlear (LOC) system that originates in the more lateral portion of the superior olive and projects mainly to the region near inner hair cells (reviewed by Warr, 1992). Most experimental data on the function and physiology of the olivocochlear neurons pertains to the MOC system (reviewed by Guinan, 1996). The action of MOC neurons on the periphery is to shift the responses of auditory-nerve fibers to higher sound levels, suggesting that they can control the dynamic range of hearing (Wiederhold and Kiang 1970; Guinan and Gifford 1988). Another action of MOC neurons is to reduce the effects of noise masking on tone burst responses, suggesting an antimasking role (Winslow and Sachs 1987; Kawase et al. 1993). Finally, MOC neurons protect the cochlea from acoustic overstimulation (Rajan 1988; Reiter and Liberman 1995).

The MOC neurons respond to sound and thus constitute the efferent limb of a reflex that we will call the MOC reflex. Single-unit recording studies (Fex 1962, 1965; Robertson and Gummer 1985; Liberman and Brown 1986; Brown 1989) have identified two main populations of MOC neurons based on response laterality: a population of ipsilateral (Ipsi) neurons that are excited by sound in the ipsilateral ear, and a population of contralateral (Contra) neurons that are excited by sound in the contralateral ear. A much smaller additional population of MOC neurons are excited by sound in either ear. Labeling studies (Liberman and Brown 1986) suggest that Ipsi neurons originate in cell bodies on the side of the brainstem opposite to the cochlea that they innervate, whereas Contra neurons originate on the side of the brainstem that is the same as the cochlea that they innervate. The relative positions of these two populations of MOC neurons is shown in Fig. 1 for the MOC neurons innervating the cochlea on the right side of the figure, the “ipsilateral” cochlea. Another brainstem auditory reflex, the stapedius reflex, is also consensual and is also organized in populations of efferent neurons that are defined by response laterality (McCue and Guinan 1988; Vacher et al. 1989).

Fig. 1
figure 1

Overview schematic indicating the pathways of the sound-evoked medial olivocochlear (MOC) reflexes to one cochlea, the ipsilateral cochlea on the right. The reflex pathway in response to ipsilateral sound begins in the ipsilateral cochlea with the response of outer and inner hair cells (OHC, IHC) and the type I afferent fibers of the auditory nerve. These fibers project centrally into the cochlear nucleus and synapse there. An unspecified population of cochlear nucleus neurons form the MOC reflex interneurons. These interneurons send axons across the midline (solid pathway) to innervate MOC neurons having a response to ipsilateral sound. These Ipsi neurons send axons back across the midline to innervate outer hair cells (OHC) in the ipsilateral cochlea. The reflex pathway in response to contralateral sound begins in the contralateral cochlea and continues with type I nerve fibers. From the contralateral cochlear nucleus, reflex interneurons send axons that cross the midline (gray pathway) to innervate Contra neurons. In addition to these dominant inputs, both types of MOC neurons receive inputs that are not strong enough to produce a response on their own but do facilitate the response to the dominant ear (small arrows, Facilitatory Input). (IVN inferior vestibular nerve)

Full size image

The pathways that mediate the MOC responses to sound are known only in very general terms. The pathways must be different for Ipsi and Contra neurons (Fig. 1), because the dominant ear that evokes a response is different. For Ipsi neurons, the reflex pathway begins with hair cells in the ipsilateral cochlea. These hair cells synapse on afferent nerve fibers that project centrally to the cochlear nucleus, an obligatory synaptic site (Fekete et al. 1984). Cochlear nucleus neurons are thus the second-order neurons, or interneurons, in this reflex. The reflex pathway must project across the midline to the Ipsi neurons; in fact, anatomical studies have shown direct projections from neurons in the cochlear nucleus that cross the midline to reach MOC neurons in the periolivary areas (Warr 1969; Robertson and Winter 1988; Thompson and Thompson 1991; Ye et al. 2000). These studies and others (Adams 1996) suggest that most of the input to MOC neurons is onto their dendrites. Finally, the Ipsi neurons project axons that again cross the midline to reach the outer hair cells of the cochlea. For the Contra neurons, information from interneurons of the contralateral cochlear nucleus must reach the Contra neurons via a crossing pathway (Fig. 1, gray shading). The Contra neurons then project to the cochlea without crossing the midline. Finally, MOC neurons in both the Ipsi and Contra groups receive facilitatory inputs from the nondominant ear (Fig. 1, small arrowheads). On their own, these inputs are not strong enough to produce a response to sound and will not be further considered (Liberman 1988; Brown et al. 1998).

The present study seeks to more precisely specify the neural pathway responsible for the MOC response to sound. The approach taken is to study the responses of MOC neurons, because their response characteristics will place constraints on the types of inputs that could generate them. Also, recent results from cochlear-nucleus lesion studies have greatly assisted in defining the sites of the cochlear-nucleus interneurons. When combined, information from studies of responses and lesions gives a much clearer picture of the MOC reflex pathway and also points to future work needed in this area.

Methods

All experimental procedures on animals were in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals, and were performed under approved protocols at the Massachusetts Eye & Ear Infirmary. Methods for recording single MOC neurons from guinea pigs are described in previous publications (Brown 1989; Brown et al. 1998). Briefly, single MOC neurons were recorded with glass micropipettes in the intraganglionic spiral bundle of one cochlea, the “ipsilateral” cochlea. The recordings were made from guinea pigs anesthetized with either a combination of Nembutal (15 mg/kg), fentanyl (0.2 mg/kg), and droperidol (10 mg/kg), or with a combination of urethane (1,500 mg/kg), fentanyl, and droperidol. Responses were measured to sound stimuli that were presented in either the ipsilateral ear, contralateral ear, or both. Sound stimuli were tone or noise bursts with rise/fall times of 2.5 ms and duration 50 ms repeated at 10/s.

Methods for lesion studies of the cochlear nucleus have been described previously (de Venecia et al. 2001). Guinea pigs were anesthetized with urethane (1500 mg/kg) and fentanyl (0.2 mg/kg). Using methods described previously (Liberman et al. 1996), the 2f1-f2 distortion-product otoacoustic emission (DPOAE) was recorded from the ear canal in response to high frequencies (f2=10 or 12 kHz; f2/f1=1.2). Samples were taken at time intervals of approximately 10 ms, beginning with the onset of stimuli. Level combinations (60–85 dB SPL) that resulted in large adaptations of the DPOAE were used (Kujawa and Liberman 2001) and the adaptations to a small matrix of L1/L2 combinations in this region were recorded (Boyev et al. 2002). Kainic acid (5 mM in PBS, pH=7) was pressure-injected into the ipsilateral cochlear nucleus via a micropipette (total injection volume 0.5–1.0 µl) and the DPOAE measurements were recorded for the following 10–13 h. The region of cell loss in the cochlear nucleus was confirmed histologically using frozen sections (Melcher et al. 1996).

Results

Reflex pathways have rapidly conducting axons

An important constraint on the reflex pathways is made by the short response latency of some MOC neurons (Fig. 2A). There is a wide range of minimum latencies across neurons, but the shortest are approx. 5 ms in guinea pigs (Fig. 2B; Robertson and Gummer 1985; Brown 1989) and in cats (Fex 1965; Liberman and Brown 1986; Liberman 1988). The latencies depend greatly on sound level. They are minimal for high-level sound although are typically much longer at lower sound levels. Minimum latency does not have a large dependence on other physiological characteristics of the neuron such as characteristic frequency (CF) or whether the neuron responds to ipsilateral or contralateral sound. A complementary measure of MOC response is the group delay in response to amplitude-modulated tones, which averages approx. 8 ms in the guinea pig (Gummer et al. 1988). Such short latencies are consistent with the proposed three-neuron reflex illustrated in Fig. 1 if all neurons have axons that conduct fairly rapidly. Thus, the population of auditory-nerve fibers that must be involved in the reflex are type I fibers (afferent fibers contacting the inner hair cells; Spoendlin 1971; Kiang et al. 1982), since these fibers convey impulses to the cochlear nucleus within a few milliseconds after sound onset. The type II fibers (afferent fibers contacting the outer hair cells) have thin, unmyelinated fibers that would be expected to convey impulses to the cochlear nucleus only after approx. 10 ms in the guinea pig (Brown 1993). It is also not clear whether type II auditory-nerve fibers respond to low and moderate levels of sound as would be required to drive the MOC reflex (Robertson 1984; Brown 1994; Robertson et al. 1999). Thus, we can rule out proposals (Kim 1986) in which type II fibers provide the sole or dominant input to the MOC reflex. Similarly, the MOC reflex interneurons of the cochlear nucleus must be a population of neurons that have relatively rapidly conducting axons. Such a consideration makes it unlikely that “small cells” of the cochlear nucleus are the dominant MOC reflex interneurons, because they probably do not have the thick, myelinated axons required to provide such short-reflex latency.

Fig. 2
figure 2

A Poststimulus time (PST) histogram obtained from a single MOC neuron showing the response to a tone burst. The minimum latency as indicated was determined as the time from tone burst onset to the beginning of the first bin that contained a response greater than spontaneous firing. Stimuli were 200 repetitions of characteristic frequency (CF; 15.6 kHz) tone bursts at 80 dB SPL. B Minimum latencies for MOC neurons recorded in the guinea pig. Stimuli were noise bursts or CF tone bursts at high levels (60–100 dB SPL). Data set was partly from (Brown 1989) and other recordings. Dashed line represents auditory-nerve fiber latencies to click stimuli from (Evans 1972)

Full size image

Lesion studies identify PVCN as the site of the reflex interneurons

Recent studies using kainic acid lesions have explored which of the three subdivisions of the cochlear nucleus (anteroventral, posteroventral, or dorsal cochlear nuclei; AVCN, PVCN, or DCN) contain the MOC reflex interneurons (de Venecia et al. 2001). Each cochlear nucleus subdivision contains a unique and varied assortment of anatomical cell types (Osen 1969; Brawer et al. 1974; Hackney et al. 1990) and physiological unit types as defined by response to sound (reviewed by Rhode and Greenberg 1992). By lesioning only one of the subdivisions in each animal, the hypothesis tested was that neurons in a particular subdivision were necessary to evoke the reflex. The reflex metric was the adaptation of the 2f1–f2 distortion product otoacoustic emission (DPOAE). The large, fast component of the adaptation (Fig. 3) appears to be produced by MOC neuron activity elicited by the f1 and f2 primary tones, which causes changes at the outer hair cells that affect the DPOAE over a time course of several hundred milliseconds. Such adaptation results from OC neuron activity, as demonstrated by the fact that it is greatly decreased by severing the axons of OC neurons in the brainstem (Liberman et al. 1996). Data showing a large effect of a kainic-acid lesion on the adaptation are shown in Fig. 3 (de Venecia et al. 2001). The most effective lesions were located in a rather small region in the dorsal and caudal PVCN. Lesions in other subdivisions had only a transient effect that recovered. Note that these experiments have so far tested only the response to ipsilateral sound. These lesion/physiology studies clearly define the location of the physiologically important interneurons, which are those interneurons that provide input to produce the sound-evoked reflex. They are consistent with previous anatomical studies indicating that PVCN is a source of projections to periolivary areas containing MOC neurons (Warr 1969) and to the MOC neurons themselves (Thompson and Thompson 1991). There are other cochlear-nucleus areas such as the marginal shell of the AVCN that project to the MOC neurons (Ye et al. 2000), but lesions in these areas in the guinea pig do not block the MOC reflex (de Venecia et al. 2001). Thus, marginal shell to MOC neuron projections presumably mediate other functions, perhaps providing the facilitatory inputs shown in Fig. 1.

Fig. 3
figure 3

Effect of a kainic-acid lesion on the efferent-mediated adaptation of the distortion-product otoacoustic emission (DPOAE). Before the lesion, there was a prominent, positive-going adaptation of the DPOAE that took place in the first several hundred milliseconds after onset of the two primary stimulus tones L 1=76 dB, L 2=80 dB SPL). The dots are the data points and the line is a smoothed curve fit to the points (using a Stineman function in Kaleidagraph 3.0). After the injection of 0.4 µl kainic acid into the ipsilateral cochlear nucleus, the adaptation was greatly reduced (data points, triangle; smoothed curve, thin line). Postexperiment histology showed a region of cell loss that was centered in the dorsal and caudal portions of the posteroventral subdivision of the cochlear nucleus (PVCN; de Venecia et al. 2001)

Full size image

Responses of MOC neurons suggest that PVCN choppers are the reflex interneurons

We now consider which PVCN neurons are the likely interneurons of the MOC reflex. The PVCN contains at least five types of neurons: granule cells, small cells, giant cells, octopus cells, and stellate cells (Osen 1969; Brawer et al. 1974; Hackney et al. 1990). Granule cells are not the interneurons because their axons do not project out of the cochlear nucleus—rather, their thin, unmyelinated axons project to the DCN (Mugnaini et al. 1980). The projections of small cells have not been well studied. They may be locally projecting interneurons or they may project with such small axons that they cannot be the MOC interneurons on the basis of the latency arguments made above. Similarly, giant multipolar cells are considered unlikely to be the interneurons, because many of them project to the contralateral cochlear nucleus (Cant and Gaston 1982; Schofield and Cant 1996) and because these neurons are glycinergic (Wenthold 1987). Such inhibitory neurons and others are unlikely to be the interneurons of the MOC reflex, because the interneurons must cause an excitatory response to sound. Thus, we are left with the two main types of PVCN neurons, octopus and stellate cells, as being possible MOC reflex interneurons.

The sound-evoked responses of MOC neurons suggest that stellate cells are the most likely candidates to be the interneurons. MOC neurons have a sustained firing pattern that continues throughout a tone burst (Fig. 4; Fex 1962; Robertson and Gummer 1985; Liberman and Brown 1986). This sustained firing is present at low, moderate, and high sound levels. In contrast, octopus cells produce an onset pattern of activity in response to a short tone burst (Rhode et al. 1983; Rouiller and Ryugo 1984). This onset pattern usually consists of a single spike and then no activity even though the tone burst continues. The transient nature of this response rules out the possibility that onset units are the interneurons, because they could not provide the sustained input needed to generate the ongoing firing pattern of MOC neurons. In contrast, the firing pattern of cochlear nucleus stellate cells is sustained like that of MOC neurons. Stellate cells have been correlated with chopper units, named because their regular interspike intervals produce chopping peaks in their poststimulus time histograms (Rhode et al. 1983; Rouiller and Ryugo 1984). Two subtypes of chopper units have been observed, chop-S or “regular” choppers and chop-T or “irregular” choppers (Young et al. 1988). Chop-S units have firing rates that adapt only a small amount over the first 10 or 20 ms after stimulus onset, whereas chop-T units have firing rates that have significant amounts of adaptation. A study of MOC responses (Brown 2001) indicates minimal adaptation: for 500-ms-duration tone bursts, the MOC neurons show short-term adaptation that is quantitatively about half that of auditory nerve fibers. For 10-s tone bursts, the response of MOC neurons is sustained, showing none of the long-term adaptation seen in auditory-nerve fibers. Long-term adaptation characteristics of chopper units have not been published. Overall, the firing patterns suggest that PVCN chopper units, possibly chop-S units, are the most likely candidates to be the reflex interneurons.

Fig. 4
figure 4

Poststimulus time histogram from a single MOC neuron that demonstrates its sustained response to sound. At the beginning of the histogram, the peaks are large but the firing between the peaks is small, whereas toward the end of the stimulus, the peaks are small and the firing between them is larger. Overall, the average rate at the beginning of the histogram is not much different from the average rate at the end, a demonstration of minimal adaptation (Brown 2001). Stimuli were 60 repetitions of 93-dB SPL tone bursts at the characteristic frequency (2.6 kHz) in the contralateral ear

Full size image

The frequency tuning characteristics of MOC neurons are also consistent with the idea that they receive inputs from chopper units as the reflex interneurons. MOC neurons have sharp frequency tuning curves (Fig. 5). In the guinea pig, MOC neurons are tuned as sharply as auditory-nerve fibers (Robertson and Gummer 1985; Brown 1989), although in the cat they are somewhat more broadly tuned than auditory-nerve fibers (Liberman and Brown 1986). PVCN chopper units also have sharp tuning, approaching the sharpness of auditory-nerve fibers, whereas onset units have generally broader tuning (PVCN: Godfrey et al. 1975; overall VCN: Jiang et al. 1996). In addition to their sharp frequency tuning, at least some MOC neurons have inhibition of spontaneous firing resulting from a single tone just outside the unit’s excitatory response area (Brown 1989). This property has not been fully explored in MOC neurons, since most of them do not have spontaneous activity. PVCN chopper units have been reported to have such inhibition (Godfrey et al. 1975; Rhode and Kettner 1987). Onset units have not been tested extensively, because they generally do not have spontaneous activity. However, one study of onset units (Jiang et al. 1996) has shown that, rather than inhibition, the units often had facilitatory responses to an excitatory tone by a second tone just outside the unit’s response area. Overall, the frequency tuning and presence of inhibition in MOC neurons is consistent with the idea that chopper units are the dominant MOC reflex interneurons in the cochlear nucleus.

Fig. 5
figure 5

Excitatory tuning curves from two MOC neurons that show sharp frequency tuning of the response to sound. Thin line, Contra neuron; thick line, Ipsi neuron

Full size image

Discussion

Our consideration of the responses of MOC neurons, along with results of lesion studies, has led us to propose that the MOC reflex pathway is composed of the following chain of elements, each of which responds in turn to sound: hair cells, type I auditory-nerve fibers, PVCN chopper units, MOC neurons, and outer hair cells. The idea that type I nerve fibers form the afferent limb of the reflex pathway is an inescapable conclusion given the short latency of MOC response. That PVCN chopper units are the interneurons is a somewhat more tentative conclusion, because it hinges on several pieces of knowledge that come from studies in either guinea pigs or cats. This knowledge comes from studies of lesions, cell types, response types, projections, and MOC response. Our conclusion about the interneuron identity could be better supported by studies of the physiological characteristics of MOC neurons and PVCN chopper units in the same species and experimental conditions. Specifically, adaptation characteristics of chopper units and MOC neurons need to be compared for both 500-ms and 10-s time scales. Also, a careful study of tuning and inhibition in both types of units would be helpful. If inhibition is difficult to study because of lack of spontaneous activity, two-tone paradigms could be used in which the first tone provides excitation at CF and the second tone is used to probe areas of inhibition. Such an approach might be successful if the inhibitory areas were much larger than two-tone suppression regions that are also likely to be present.

The idea that PVCN neurons provide input to MOC neurons originated with the studies of Warr (1969), who showed that PVCN neurons project to the periolivary areas in cats. More specifically, studies in guinea pig (Thompson and Thompson 1991) have shown that PVCN neurons project directly to MOC neurons. We postulate a simple, three-neuron reflex pathway for the response to sound because of this direct pathway, even though the MOC minimum latencies are not so short as to exclude an additional synapse between the cochlear nucleus and the MOC neurons. Neither of the projection studies was able to identify the type of PVCN neuron whose projections were being studied. In rats, individually labeled multipolar/stellate cells of the PVCN send branches to the rostral periolivary area (Friauf and Ostwald 1988), a nucleus that contains many MOC neurons. Projections of PVCN stellate/multipolar cells are now receiving attention (Doucet and Ryugo 2002). Certainly projections of those cells corresponding to chop-S units would be particularly interesting to investigate, since those cells are possible interneurons of the MOC reflex. It is important to remember, though, that projection studies do not always reveal the neurons that are important to evoke a response, and thus they must be accompanied by physiological and/or lesion studies (de Venecia et al. 2001) to link the underlying anatomical circuitry with physiological effects.

The arguments outlined above hold for the MOC reflex to ipsilateral sound. It is not yet clear that the cochlear nucleus interneurons for the reflex to contralateral sound are located in PVCN. However, the latency constraints of the contralateral reflex are similar to those of the ipsilateral reflex, and most physiological properties of Ipsi and Contra neurons are similar. These reasons lead us to suspect that the type of cochlear-nucleus interneuron is similar for the two reflex pathways. On the other hand, the facilitatory pathways to MOC neurons (small arrows in Fig. 1) have only been minimally explored. These inputs are not strong enough to elicit a response to sound and may well have different characteristics, perhaps because they are mediated by different interneurons or pathways. Further physiological and lesion studies of the facilitatory pathways are clearly needed for a complete understanding of the MOC reflex. Finally, although we have concentrated on the relatively simple pathway that elicits the MOC response to sound, anatomical studies have shown that higher centers provide several inputs to MOC neurons (Faye-Lund 1986; Mulders and Robertson 2000a). How these inputs modulate the response to sound mediated by the simple pathway awaits future studies.

The type of studies outlined above for the MOC reflex could also be applied to an analogous reflex, the stapedius reflex, a reflex for which the cochlear nucleus interneurons are also not known. Although there is one early study on this subject (Borg 1973), the cochlear nucleus lesions in that study were large and affected axons of passage such as those of auditory-nerve fibers. Such an approach severely limits the conclusions that can be drawn. Studies such as those we have done for the MOC reflex (de Venecia et al. 2001), using more focal lesions made with kainic acid to spare axons of passage, are needed to define the locations of interneurons for the stapedius reflex. For the tensor tympani, one study (Rouiller et al. 1986) has demonstrated inputs from the superior olivary complex, indicating a more complex pathway than for the MOC reflex. Other studies (Itoh et al. 1986; Ito and Honjo 1988) suggest that the tensor tympani receives direct projections from the cochlear nucleus. Additional projection studies like those used to study the OC inputs (Thompson and Thompson 1991) need to be done for both the stapedius and tensor tympani reflex pathways. Results of these studies, along with the known physiological characteristics of the stapedius motoneurons (Kobler et al. 1987, 1992; Vacher et al. 1989), might also more clearly elucidate the middle-ear muscle reflex pathways.