Differential Central Projections ofPhysiologically Characterized HorizontalSemicircular Canal Vestibular Nerve

Afferents in the Toadfish, Opsanus tau

ALLEN F. MENSINGER,1* JOHN P. CAREY,2 RICHARD BOYLE,3

AND STEPHEN M. HIGHSTEIN1

1Department of Otolaryngology, Washington University School of Medicine,St. Louis, Missouri 63110

2Department of Otolaryngology, University of Washington School of Medicine,Seattle, Washington 98195

3Department of Otolaryngology/Head-Neck Surgery and Physiology and Pharmacology,Neurosensory Research Center, Oregon Health Sciences University, Portland, Oregon 97201

ABSTRACTAnatomical and neurophysiological studies were undertaken to examine the central

projection pattern of physiologically characterized horizontal semicircular canal vestibularnerve afferents in the toadfish, Opsanus tau. The variations in individual response character-istics of vestibular nerve afferents to rotational stimulus provided a means of typing theafferents into descriptive classes; the afferents fell into a broad continuum across thespectrum from low-gain, velocity-sensitive to high-gain, acceleration-sensitive responses(Boyle and Highstein [1990b] J. Neurosci. 10:1557–1569; Boyle and Highstein [1990a] J.Neurosci. 10:1570–1582). In the present study, each afferent was typed as a low-gain,high-gain, or acceleration fiber during rotational or mechanical stimulation (Rabbitt et al.[1995] J. Neurophysiol. 73:2237–2260) and was then intracellularly injected with biocytin.The axons were reconstructed, and the morphology, synaptic boutons, and projection patternof each axon were determined. The results indicated that the three descriptive classes ofvestibular nerve afferents have unique as well as overlapping central projection patterns anddestinations in the vestibular nuclei, with intranuclear parcellation in the anterior octavus,magnocellularis, tangentialis, posterior octavus, and descending octavus nuclei. In general,increased sensitivity and faster response dynamics were correlated with both a moreextensive central projection and a progressive increase in morphological complexity. Low-gain, velocity-sensitive fibers were the simplest morphologically, with the fewest number ofbranches (n 5 17) and shortest length (4,282 µm), and projections were confined to the middleportions of the vestibular nuclei. High-gain, velocity-sensitive fibers were morphologicallymore diverse than low-gain fibers, with a greater number of branches (n 5 26), longer length(6,059 µm), 29% greater volume, and amore widespread projection pattern with projections toboth the anterior and the middle portions of the vestibular nuclei. Acceleration fibers weremorphologically distinct from low- and high-gain fibers, with more elaborate branching (n 541), greatest overall length (17,370 µm) and volume (16% greater than high gains), anddisplayed the most extensive central projection pattern, innervating all vestibular nucleiexcept tangentialis. Thus, there are anatomically demonstrable differential central projec-tions of canal afferents with different response dynamics within the vestibular complex of thefish. J. Comp. Neurol. 384:71–85, 1997. r 1997 Wiley-Liss, Inc.

Indexing terms: teleost; synaptic bouton; vestibular nuclei

The hair cells of the semicircular canals of the vestibularlabyrinth detect angular motion of the head and modulatethe firing rate of the contacting primary afferents. Despitesimilarities and differences in morphology of the labyrinthamong vertebrates, a general feature is present in eachspecies: For an applied angular acceleration of the head,

Grant sponsor: NIH; Grant numbers: PO1-DC1837, PO1-NS17763-12.*Correspondence to: Dr. Allen F. Mensinger, Department of Otolaryngol-

ogy, Washington University School of Medicine, Box 8115, 4566 ScottAvenue, St. Louis, MO 63110. E-mail: mensina@tfneuron.wustl.eduReceived 2 December 1996; Revised 25 February 1997; Accepted 26

February 1997

THE JOURNAL OF COMPARATIVE NEUROLOGY 384:71–85 (1997)

r 1997 WILEY-LISS, INC.

the primary canal afferents do not respond uniformly.Individual canal afferents show differences in the magni-tude (gain) and time (phase) of firing-rate modulation withrespect to the stimulus and the degree to which theresponse is dependent on the frequency of stimulation(Fernandez and Goldberg, 1971; Blanks et al., 1975;O’Leary et al., 1976; Schneider and Anderson, 1976;Anderson et al., 1978; Landolt and Correia, 1980; Tomko etal., 1981; Boyle and Highstein, 1990a,b). It has beenreasoned that these diverse responses are functionallyrelevant to the processing and initiation of the differentvestibular reflexes and, thus, aremaintained after the firstsynapse in the vestibular nuclei of the brainstem. One ideais that certain afferent inputs are better matched thanothers to the dynamic load requirements of the individualcompensatory motor responses, from the vestibuloocularto the vestibular neck and limb reflexes (Goldberg andFernandez, 1971). Recent experimental evidence usingintracellular techniques indicates that at least a partialsegregation of afferent inputs occurs at the level of second-order neurons (Highstein et al., 1987; Boyle et al., 1992).Peripheral segregation of the origins of canal primary

afferents has been shown in the toadfish (Boyle et al.,1991), chinchilla (Fernandez et al., 1988), and squirrelmonkey (Fernandez et al., 1995; Lysakowski et al., 1995),with orderly maps residing within their crista of origin. Inthe toadfish, afferents with low-response sensitivity torotation, slow dynamics, and few terminal endings supplythe peripheral portions of the crista, whereas afferentswith higher sensitivities, faster dynamics, and greaternumbers of terminal endings innervate the more centralportions (Boyle et al., 1991). This suggested that centralanatomical maps might also be demonstrable within thevestibular nuclei. Preliminary work in the turtle showedthat the brainstem distribution of the descending limbs ofcentral axons of vestibular afferents vary systematicallywith parent axon diameter: Terminals of large-diameterafferents were concentrated in the rostral regions of themedial/descending nuclei, whereas terminals of progres-sively smaller diameter fibers were shifted toward thecaudal vestibular complex and adjacent brainstem struc-tures (Huwe and Peterson, 1995). Some parcellation ofeighth nerve activity, as mentioned above, has indeed beendemonstrated electrophysiologically in the squirrel mon-key (Goldberg et al., 1987; Highstein et al., 1987; Boyle etal., 1992). However, morphological studies in the mammalhave not provided many insights into the anatomicalbasis, if it is present, of afferent-specific subpopulations of

central vestibular neurons. For example, investigations inthe cat were unable to differentiate any parcellation ofaxon trajectories in the vestibular nuclei (Sato et al., 1989;Sato and Sasaki, 1993) but did find that regular andirregular types of horizontal canal afferents differ in theirmodes of terminal arborization (Ishizuka et al., 1982; Satoet al., 1989). The mammalian vestibular nuclei are notrigidly delineated with individual cell groups controllingparticular functions, and neurons participating in thevestibuloocular and vestibular neck reflexes can be mi-crons away from each other (Boyle et al., 1992). For thesereasons, the morphological tools available at the lightmicroscopic level might not be sensitive enough to detectany differences in afferent-cell synaptic formations. Alter-natively, this lack of discrete segregation of afferents inmammals might indicate that the unique afferent physi-ological signals are transmitted to secondary neurons viasynaptic sorting or by an as-yet-undefined mechanism.Teleosts possess a vestibular system that is remarkably

similar to that of higher vertebrates. However, the overallmorphology is simpler than in other phyla with a sparserand distributed neuropile (McCormick, 1982; Highstein etal., 1992), providing a promising opportunity to follow anddiscriminate primary afferents to their central termina-tions. The task is further facilitated in the teleosts (McCor-mick, 1982) by the anatomical separation of the lateral lineand vestibular systems.The present study confirms that there are anatomically

demonstrable differential central projections of canal affer-ents at the light microscopic level. A preliminary reporthas been presented (Carey and Highstein, 1990).

MATERIALS AND METHODS

Adult toadfish, Opsanus tau, of either sex were suppliedby the Marine Resources Center of the Marine BiologicalLaboratories in Woods Hole, Massachusetts. Fish weremaintained in running sea water or Instant Ocean(Aquarium Systems, Mentor, OH) at 15°C. All experimen-tal protocols were approved by the Washington UniversitySchool of Medicine Animal Use Committee and conformedto NIH guidelines. Fish were lightly anesthetized withMS-222 (Sigma, St. Louis, MO) and partially immobilizedby an intramuscular injection of pancuronium bromide(0.05 mg/kg; Sigma), secured in a plastic experimentaltank, and perfused through the mouth with sea water. Thewater level in the tank covered the fish’s premaxilla andmost of the operculum, and any exposed areas werecovered with moist tissues. Adequate stimulation of thehorizontal semicircular canal (HCN) was provided eitherby a rate table (rotation) or by a mechanical indenter inlater experiments. For rotation, the experimental tankwas placed atop a rate table (Contraves, Tampa, FL) withthe head of the fish centered on the intercanal axis andpitched nose down to visually align the horizontal canalwith the plane of rotation. The HCN and canal nerve wereexposed by a dorsal craniotomy, and the exposed cavitieswere filled with a fluorocarbon (FC-75; 3M Corp., Minne-apolis, MN). An external function generator (model 146;Wavetek, San Diego, CA) produced the sinusoidal stimula-tion for table rotation. The afferent’s response for rota-tional stimuli between 0.01 Hz and 1.0 Hz at 10° and25°/second was characterized.Details of the mechanical indentation of the membra-

nous labyrinth can be found in Rabbitt et al. (1995).Briefly, a piezoelectric microactuator (Burleigh, Fishers,

Abbreviations

AO nucleus anterior octavusDM, DOM rostral middle finger of nucleus descending octavusDO nucleus descending octavusDOD, DD dorsal nucleus descending octavusDOV, DV ventral nucleus descending octavusEG eminentia granularisIV fourth ventricleIX ninth nerveM1 first division of n. magnocellularisM2 second division of n. magnocellularisM3 third division of n. magnocellularisMA nucleus magnocellularisPO nucleus posterior octavusT, TA nucleus tangentialisVII seventh nerveVIII eighth nerveVIII a.r. anterior ramus of the eighth nerveX tenth nerve

72 A.F. MENSINGER ET AL.

NY, PZL 060-11) fitted with a 1.2-mm-diameter glass rodwas used to indent the long and slender limb of thehorizontal canal. The piezo was charged by a high-voltageamplifier (Trig-Tek 207A, Anaheim, CA) and was drivenby using an externally triggered function generator(Tektronix FG 501A, Beaverton, OR). Displacement of theindenter was monitored by using a linear voltage-displacement transducer (LVDT; Schaevitz DEC-050,Hampton, VA) mounted in series directly below the piezo.Afferent responses were characterized for sinusoidal inden-tations of 1–10 µm of half of peak-to-peak displacement,which is equivalent to rotational stimuli of 4–20°/second(Rabbitt et al., 1995). A glass microelectrode containing a2% solution of biocytin (Sigma or Molecular Probes, Eu-gene, OR) in 2MNaCl, 5 mMKCl, and 100 mMTris buffer,pH 7.2, with a DC impedance of 40–100 MV was insertedinto the nerve along its free end as it passed over thesaccule, usually within 1 mm of the horizontal ampulla.Axonal penetration was signaled by a 240 mV to 260 mVresting potential that stabilized after several seconds.Once the afferent was penetrated, stimulation sufficient tocharacterize it as either low-or high-gain, velocity- oracceleration-sensitive (Boyle andHighstein, 1990a,b; Boyleet al., 1991) was applied. Potentials were conventionallyamplified, idealized by an amplitude window-discrimina-tor circuit, and collected as events in the Spike-2 packageusing a Cambridge Electronic Design 1401 peripheralcomputer interfaced to an 80486 computer. Data wereanalyzed off-line by utilizing a library of Spike-2 user-written routines (Boyle and Highstein, 1990a,b; Boyle etal., 1991). After characterizing the fiber, it was labeled bypassing positive current pulses of 10–20 nA (1/second, 80%duty cycle) for 2–10 minutes through the electrode. Cur-rent injection was attempted in only one afferent fiber perfish.Immediately after injection, the electrode was with-

drawn from the nerve, and no further recordings wereattempted. The fish survived for 4–24 hours after theinjection. Fish were then deeply anesthetized withMS-222and perfused through the conus of the heart with 1 mlheparin, followed by 100 cc of saline and 200 cc of fix (3%paraformaldehyde and 1% glutaraldehyde in 0.1 M phos-phate buffer, pH 7.4). The brain, cranial nerves, andlabyrinth were removed en bloc, postfixed overnight, andtransferred to a solution containing 1% bovine serumalbumin (Sigma) in 0.1M phosphate-buffered saline (PBS).The tissue was then placed in 20% phosphate-bufferedsucrose. The brain was cut at 50 µm on a freezingmicrotome. Vestibular nerveswere straightened andwhole-mounted. Immunoperoxidase stainingwas done by incubat-ing the tissue in a solution of avidin biotin-horseradishperoxidase complex (ABC; 1:50 dilution; Vector Laborato-ries, Burlingame, CA) and 0.25% Triton X-100 in PBS for2-3 hours, followed by a 4–10 minute soak in 0.05%diaminobenzidine (DAB) in 0.1 M phosphate buffer (PO4)and 0.03% H2O2. Sections were examined under 325 highand 340 oil objectives on a Zeiss microscope. Light micro-graphs were made with a Kodak DSC-200 digital camera,imported into Photoshop (Adobe, San Jose, CA), andprinted with a Kodak (Rochester, NY) XLS-8300 printerwithout alteration. The injected fibers were reconstructedby camera lucida or Neurolucida (MicroBrightField, Inc.,Colchester, VT), a computer-assisted reconstruction sys-tem, and bouton numbers and location were determined.Fiber diameter of parent axons was calculated at the pointof fiber entry into the anterior ramus. The Neurolucida

program, MORPH, calculated total axon length, volume,and average fiber diameter from axon entry into theanterior ramus to all branch endpoints. NueRotate (Micro-BrightField, Inc.) enabled reconstructed axons to be ro-tated in three dimensions.

RESULTS

The response dynamics of toadfish horizontal canalafferents have been characterized and divided into threedescriptive groups (Boyle and Highstein, 1990a,b). Varia-tion in individual responses to either rotation or indentionacross the frequency band of 0.001–10 Hz provides ameans of typing the afferents into groups labeled low-gainand high-gain velocity-sensitive afferents and acceleration-sensitive afferents. Figure 1A–C shows averaged cyclehistograms to a 0.5 Hz, 610°/second, rotation (top trace)for three separate afferents. Low-gain afferents (for ex-ample, see Fig. 1A) have a relatively high discharge rate(and regular spacing of the interspike intervals) and aresponse characterized by a discharge that is maintainedthroughout the stimulus cycle, a moderate or low ampli-tude of modulation (gain), and a temporal relationship(phase) closely aligned with stimulus velocity at frequen-cies above about 0.01 Hz. High-gain afferents (for example,see Fig. 1B) generally have a lower resting discharge ratethan low-gain afferents, and the discharge is often silencedover a portion of the stimulus cycle, particularly forhigh-frequency or large-amplitude stimuli; these afferentsdemonstrate a pronounced gain and phase dependency onstimulus frequency and respond more closely in phase

Fig. 1. Cycle histograms of responses (impulses/second) of a low-gain (A), high-gain (B), and acceleration (C) afferent at 0.5 Hz(610°/second) during cycles of rotation. Top trace: Stimulus velocityprofile.

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 73

with velocity at frequencies between approximately 0.03–5Hz. Acceleration fibers (for example, see Fig. 1C) arereadily distinguished from the other afferents because oftheir response, which is more closely in phase with stimu-lus acceleration at all frequencies; furthermore, they typi-cally have the highest gain (note the different scale in Fig.1C) and also show a pronounced gain enhancement toincreasing stimulus frequency.Afferent responses to separate rotation and mechanical

indentation have been extensively studied in this animal,and individual afferents can be readily typed by usingeither stimulus (Boyle and Highstein, 1990a,b; Rabbitt etal., 1994, 1995). In this study, intracellular recordingswere taken from afferents during both rotational stimula-tion (nine afferents) and mechanical indentation stimula-tion (ten afferents). The data base for this report iscomprised of 19 injected and completely reconstructedfibers, consisting of six low-gain, eight high-gain, and fiveacceleration afferents. Fibers were classified as completelyfilled based on the extent of the label and the presence offilled synaptic boutons (n . 15) at the terminal ends of thefibers. In addition, numerous fibers (n 5 11) of each class(n $ 3) were partially filled (possessing ,15 boutons), andtheir projection patterns generally supported the conclu-sions presented below.The anatomical organization of the vestibular nuclei in

the toadfish has been studied (Highstein et al., 1992) andfollows the generic teleost plan (McCormick, 1982). Figure2 shows an exploded dorsal view of the toadfish brainstemfor illustrative purposes, delineating themaximummedial-lateral and anterior-posterior dimensions of the vestibularnuclei. Roman numerals indicate specific cranial nerves.Horizontal canal afferents enter the brainstem in theanterior ramus of the eighth nerve (VIII a.r.) and project tothe labeled nuclei described below and to the eminentiagranularis (EG) of the cerebellum. The anterior octavusnucleus (AO) is the most anterior subdivision of thecomplex bordered by the EG rostrally and the nucleusmagnocellularis (M1–3) caudally. EG overlies the rostralborders of theAO. The magnocellularis nucleus appears asa crescent-shaped nucleus subdivided into three sections,M1–3, and bordered rostrally by AO, medially by therostral finger of descending octavus (DO), caudally by DO,and laterally by VIII a.r. The DO is the largest of thevestibular nuclei and is subdivided into three regions: thedorsal descending (DOD), the rostral middle finger (DOM),and the ventral descending (DOV). DOD is located alongthe lateral border of the brainstem and is bordered ros-trally by M3 and caudally by the posterior octavus nucleus(PO). DOM diverges medially from dorsal descending nearthe medial ventral border of M3 and extends rostrally toapproximately the caudal border of AO. DOV is showndisplaced medially and caudally in the figure, as thedashed lines indicate its normal position ventral to theDOD. Nucleus tangentialis (T) is a rod-shaped nucleuslying tangential to the lateral edge of the brainstem, withthe dashed lines indicating its normal position ventral toM3 and to DO. PO is the most caudal nucleus, with itsrostral border adjacent to the caudal border of DO andextending past the entry of the tenth nerve.Figure 3A is a light micrographic montage of the dorsal

view of the brainstem with an intracellularly injectedhigh-gain afferent traveling within the eighth nerve (Fig.3A, upper right); the afferent enters the brainstem at theanterior ramus and projects ventrocaudally along the

Fig. 2. Dorsal view of the toadfish brainstem (exploded for illustra-tive purposes) delineating the maximum horizontal projections of thevestibular nuclei, the eminentia granularis (EG), and the cerebellum.Tangentialis (T) has been displaced laterally, and ventral descendingoctavus (DO ventral) has been displaced medially. The dashed linesindicate the normal position of these two nuclei ventral to magnocellu-laris (M3) and dorsal descending octavus (DO dorsal). Roman numer-als indicate specific cranial nerves. For other abbreviations, see list.

74 A.F. MENSINGER ET AL.

lateral edge of the brainstem. The same fiber is shownserially reconstructed at the right (Fig. 3B), with theobserved synaptic boutons indicated by solid circles. After

entering at the anterior ramus of the eighth nerve, thisfiber ran parallel to the lateral edge of the brainstem (Fig.3, a) and produced a short lateral branch, terminating in

Fig. 3. A,B: Light micrograph montage (A) of the dorsal view of thebrainstem with an intracellular-injected, high-gain vestibular nerveafferent traveling throughout the eighth nerve (VIII), entering thebrainstem at the anterior ramus (a.r.), and projecting ventrocaudallyalong the lateral right edge of the brainstem. The identical fiber is

shown serially reconstructed and in finer detail in B, with synapticboutons indicated by dots superimposed on the vestibular nucleishown inFigure 2. The lettered arrows correspond to identical sites in bothAandB.For further descriptions, see the text.Note that theDOdorsalwasremoved inB. Scale bars5 250 µm.

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 75

M3 (Fig. 3, b). The main fiber turned medially andgenerated a long medial branch (Fig. 3, c) that terminatedin M3 and DOV. The main fiber turned caudally and ranthe length of DOV, sending a single branch into T, andterminating in numerous branches (Fig. 3, d,e) throughoutDOV. The fiber had 51 branches and 60 synaptic boutonsdistributed as follows: M3 5 4, T 5 4, and DOV 5 52.Figure 4 is a series of light micrographs illustrating the

sagittal views of another high-gain fiber traveling caudallyfrom right to left. Figure 4A shows the thick main fiber(Fig. 4A, bottom) traveling caudally into the anteriorramus of the eighth nerve, with a thinner branch (Fig. 4A,top) that was generated farther caudally by the mainbranch traveling dorsally in the opposite (rostral) direction(Fig. 4A, rostral end indicated by arrow) to innervate theAO. Figure 4B depicts the fiber traveling caudally through

the DOV and generating several large branches. Figure 4Cis an enlargement of the terminal end of the fiber in theDOV illustrating a few of the synaptic boutons.Figure 5 is a reconstruction of the central projections of

separate fibers from each of the three types of afferents;circles represent observed synaptic specializations withinthe vestibular nuclei. The low-gain fiber (Fig. 5A) had asimple projection pattern (ten branches) with projectionslimited to T and DOV. The fiber bifurcated immediatelyupon entering the brainstem, producing a short, straightmedial collateral (Fig. 5A, a) that terminated in a cluster ofboutons in the anterior half of T. The lateral branch of themain fiber (Fig. 5A, b) continued ventrocaudally andproduced a medially directed collateral (Fig. 5A, c) withone branch terminating in the caudal half of T and theother branch (Fig. 5A, d) innervating the anterior portionof DOV. The main fiber continued caudally through theDOV, producing several branches before terminating atthe caudal border of the DOV. The fiber contained 92boutons, with 21 in T and 71 in DOV.The high-gain fiber (Fig. 5B) was moderately branched

with 25 branches and displayed a broader central projec-tion pattern than its low-gain counterpart. The fiberentered the brainstem and bifurcated approximately mid-way through M3. The medial branch quickly gave rise to asmall offshoot (Fig. 5B, a) terminating in M3 and thenturned anteriorly into AO, where the fiber bifurcated (Fig.5B, b) into a lateral branch (Fig. 5B, c) that terminated inAO and a medial branch (Fig. 5B, d) that terminated inEG. The lateral branch of the main fiber continued posteri-orly before producing a branch through the posterior halfof T that terminated in a single bouton (Fig. 5B, e) in theanterior half of DOV. The main branch (Fig. 5B, f) contin-ued caudally into the DOV, producing a dorsally directedbranch into the DOD (Fig. 5B, g) and numerous bifurca-tions and branches throughout the posterior half ofthe DOV. The bouton distribution (n 5 67) was as follows:EG 5 3, AO 5 5, M3 5 7, T 5 11, DOD 5 9, and DOV 5 32.The acceleration fiber (Fig. 5C) was extensively branched

and projected to all vestibular nuclei, with the exception oftangentialis. The fiber branched immediately after enter-ing the brainstem into three main branches (Fig. 5C, a–c).Branch a turned anteriorly and projected intoAO, where itproduced a short dorsal branch (Fig. 5C, d) into the middleof AO. The main trunk of branch a continued medially andbifurcated at the caudal border of AO, with one branch(Fig. 5C, e) continuing medially along the caudal border toAO, producing a dense concentration of boutons, andterminated in the medial caudal quadrant of AO, whereasthe second branch curved ventromedially. The fiberbranched (Fig. 5C, f) midway between M1 and DOM andproduced a short terminal projection into M1. The fibercontinued caudally and sent terminal branches mediallyinto DOM and laterally (Fig. 5C, g) throughout M2.Branch b continued posteriorly until the dorsal border ofM3, where it bifurcated into anterior and ventral branches.The anterior projection (Fig. 5C, h) terminated at thecaudal lateral border of AO, whereas the posterior projec-tion (Fig. 5C, b) continued medially before terminatinginto several short branches in the middle one-third ofDOD. Branch c traveled unbranched deep into the DOVbefore giving rise to a convoluted series of terminalbranches. One branch (Fig. 5C, i) continued deep andunbranched into the PO, past the innervation sites for thetenth nerve, and terminated near the posterior lateral

Fig. 4. Light micrographs of a sagittal view of an intracellular-injected, high-gain fiber. A: Sagittal view of parent fiber (bottom)entering the anterior ramus of the eighth nerve (VIII a.r.) with athinner branch (top) traveling to the nucleus anterior octavus (AO;arrow indicates the rostral end). B: Sagittal view of a fiber travelingthrough the ventral descending octavus (DOV). C: Sagittal view of theterminal end of a fiber in the DOV. Arrows indicate the locations ofsynaptic boutons. Scale bars 5 100 µm inA,B, 10 µm in C.

76 A.F. MENSINGER ET AL.

border of the PO. The bouton (n 5 212) distribution was asfollows:AO5 23, M15 7, M25 23, DOD5 25, DOV5 115,DOM 5 5, and PO 5 14.Figure 6 provides computer reconstructions of horizon-

tal (Fig. 6A) and sagittal (Fig. 6B) views of three differentfibers with their synaptic boutons (Fig. 6, solid circles) inthe vestibular nuclei. The fibers originated from the eighthnerve (Fig. 6, asterisk) and entered the anterior ramus onthe top right (Fig. 6A) or the far right (Fig. 6B) and

proceeded caudally (Fig. 6A, top to bottom; Fig. 6B, right toleft) into the vestibular nuclei. Both figures are aligned atapproximately the same location in M3 (Fig. 6, arrow) toillustrate the relative length and extent of each fiber.The low-gain (LG) afferent in Figure 6 had ten branches,

a total length of 2,752 µm, a volume of 39,754 µm3, and anaverage fiber diameter of 4.3 µm with 34 boutons (M3 5 2,T 5 4, and DOV 5 28). The fiber entered the anteriorramus of the eighth nerve and initiated a ventral course,

Fig. 5. Intracellularly injected and reconstructed vestibular nerve afferents: low gain (A), high gain(B), and acceleration (C) with synaptic boutons (circles) superimposed on the vestibular nuclei describedin Figure 2. Lower case letters refer to specific branches or segments that are further described in the text.

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 77

Fig. 6. Three vestibular nerve afferents: low gain (LG), high gain(HG), and acceleration (ACC) afferents with synaptic boutons (dots).Each fiber was reconstructed from horizontal sections (A) and thenrotated sagittally (B) to illustrate the relative depths of the centralprojections. The diameter of each fiber was tripled for illustrativepurposes; however, the relative diameters of portions of each axonremain accurate. Variation in diameter occurs along the length of each

axon. The reconstruction originates in the eighth nerve (asterisk) andproceeds caudally (A, top to bottom; B, right to left) into the vestibularnuclei. The sections are aligned (arrows) at the approximately thesame level in the anterior ramus to illustrate the relative length andextent of each fiber. Compass indicates orientation (R, rostral; C,caudal; L, lateral; M, medial; D, dorsal; V, ventral), with the horizontaland vertical lines equaling 200 µm.

producing a short branch terminating in a pair of boutonsin M3 (Fig. 6, arrow). Approximately 200 µm farthercaudal, the fiber produced a small ventral branch thatturned caudally and terminated in T. The main fiber

continued on a ventral course, producing numerous, thinterminal branches throughout the length of the DOV.The middle traces in Figure 6 show the morphological

features of a high-gain (HG) afferent. This is the same fiber

Figure 6 (Continued.)

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 79

that is described in Figure 3.Although the terminal sites ofthis afferent were similar to those of the above-mentionedlow-gain afferent, its axonal morphology was more exten-sive. The fiber had 51 branches, with an overall length of6,891 µm, a volume of 49,264 µm3, and an average fiberdiameter of 3.0 µm; the bouton (n5 60) distribution was asfollows: M3 5 4, T 5 4, and DOV 5 52.The acceleration (ACC) afferent illustrated in Figure 6

contained 66 branches, with a length of 20,342 µm, avolume of 141,282 µm3, and an average fiber diameter of3.0 µm. In contrast to the low- and high-gain afferents, theacceleration afferent supplied terminal boutons toAO (n 52), DOD (n 5 4), and PO (n 5 6); the remaining boutondistribution was as follows: M1 5 2, M3 5 3, and DOV 546. The fiber twisted throughout the eighth nerve beforeentering the ramus and producing two small branches intothe lateral border of M1. The main fiber produced a dorsalbranch that generated a short terminal branch (Fig. 6C,ACC, arrow) into M3 before diverging rostrally into AOand caudally into DOD. The main fiber continued caudallyand, in contrast to the LG and HG fibers, remainedunbranched through the rostral one-third of DOV. Thefiber generated numerous branches throughout the caudalhalf of DOV before terminating into PO.Table 1 summarizes the morphological data for the three

physiological types of afferent fibers. The averaged values(6S.E.) of boutons and branch number, parent axon diam-eter (at anterior ramus), the fiber’s total length (fromanterior ramus to central terminations), volume, andaverage fiber diameter (throughout the brainstem) aresummarized in each column for the three afferent types.The numbers in parentheses represents the minimum andmaximum values for each group. Figures 5 and 6 illustratethat the acceleration fibers were morphologically the mostcomplex, with the greatest bouton and branch numbers,fiber length, and volume. Acceleration fibers were signifi-cantly longer [P , 0.01; analysis of variance (ANOVA)]than low- and high-gain fibers and had significantly greatervolume than low-gain fibers (P , 0.05;ANOVA). High-gainfibers had greater volume, length, and average fiberdiameter than low-gain fibers. There was a great deal ofoverlap among the classes in parent fiber diameter (noteranges), and, although high-gain fibers averaged slightlygreater diameters than the other two classes, there was nosignificant difference among the three classes in parent oraverage fiber diameter.Figure 7 summarizes the synaptic bouton distribution

for the three afferent fiber classes. The figure superim-poses the synaptic bouton distribution within the vestibu-lar nuclei for each of the three types of afferents: low gain(Fig. 7A; n5 6), high gain (Fig. 7B; n5 8), and acceleration(Fig. 7C; n 5 5). Each dot represents a single bouton. Thethree classes displayed both distinct and overlapping

central terminations, with acceleration fibers averaging106.0 boutons/fiber, whereas low- and high-gain fibersaveraged 91.8 and 80.3 boutons/fiber, respectively. The DOreceived the majority of the projections from all primaryafferents, with 72% (acceleration afferents) to 82% (high-gain afferents) of boutons contained within its borders.The ventral subdivision of DO was the primary target forall three classes and contained significantly greater bou-tons within each class (P , 0.01; ANOVA) than othervestibular nuclei. Most of the boutons were confined to thelateral half of the nucleus, with only a few boutons foundnear themedial border. The bouton distribution of accelera-tion fibers was relatively evenly distributed throughoutthe DOV nucleus; however, the boutons of low- and high-gain fibers were heavily concentrated in the caudal half ofthe nucleus. The dorsal portion of the DO primarilyreceived significantly greater (P , 0.05; ANOVA) projec-tions from acceleration fibers, averaging 23.8 boutons/fibercompared with the input of low-gain (5.7 boutons/fiber)and high-gain fibers (1.3 boutons/fiber). The projectionsfrom the acceleration fibers were evenly distributedthroughout DOD, whereas the low-gain projections wereconcentrated in the caudal end of the nucleus. The DOMreceived sparse input, with only a single fiber in each classprojecting to the nucleus and with the low-gain fiberprojecting into the caudal and rostral border, whereas thehigh-gain and acceleration fiber projections were re-stricted to the rostral tip of the nucleus. Nucleus tangentia-lis (T), the only purely vestibular nucleus by virtue of theexclusive termination of semicircular canal afferents (High-stein et al., 1992), was preferentially targeted by low-gainafferents (8.3 boutons/fiber). High-gain afferents contrib-uted amoderate input (3.6 boutons/fiber); however, none ofthe five acceleration fibers was observed to issue a boutonto this nucleus. PO was exclusively (P , 0.05; ANOVA)innervated by acceleration fibers, with clusters in theanterior and posterior half of the nucleus separated by theinnervation sites of the tenth nerve. Magnocellularis re-ceived relatively equal projections from low-gain andacceleration fibers (7.0 and 8.8 boutons/fiber, respectively)and moderate innervation from high-gain fibers (3.0 bou-tons/fiber). Low- and high-gain fibers projected primarilyto M3, whereas acceleration fibers favored M2. Projectionsto the AO were almost exclusively (with the exception of asingle low-gain fiber) from high-gain and accelerationfibers. Boutons from acceleration fibers were found primar-ily near the caudal border, whereas high-gain fiber termi-nations were widely distributed throughout the nucleus.EG received surprisingly little innervation, as only asingle high-gain and acceleration fiber projected into thisarea.Figure 8 and Table 2 summarize the bouton distribution

of the primary afferent fibers shown in Figure 7. Figure 8A

TABLE 1. Primary Afferent Fiber Morphology1

Fiberclass (n)

Boutons(n)

Branches(n)

Length(µm)

Volume(µm3)

Parent fiberdiameter(µm)

Average fiberdiameter(µm)

Low gain (6) 91.8 6 30.7 17.2 6 4.3 4,282 6 791 55,786 6 12,551 5.10 6 0.51 3.6 6 0.14(16–247) (4–25) (2,852–6,452) (39,000–92,267) (3.67–6.07) (2.16–4.27)

High gain (8) 80.3 6 26.8 25.5 6 9.04 6,059 6 1,756 71,885 6 10,628 7.54 6 1.21 4.3 6 0.85(29–267) (10–51) (2,947–10,768) (48,736–99,634) (3.86–10.98) (2.96–6.56)

Acceleration (5) 106 6 28.3 41.3 6 16.1 17,370 6 3,306* 118,543 6 23,149* 5.33 6 0.96 2.94 6 0.20(30–212) (17–72) (10,768–21,000) (72,246–142,000) (3.40–6.29) (2.82–2.97)

1Mean values 6 S.E. with range indicated in parentheses.*Indicates significantly different mean (P , 0.05; ANOVA).

80 A.F. MENSINGER ET AL.

graphs the averaged number 6 S.E. of boutons containedin each vestibular nucleus and subdivision for low gain(Fig. 8A, open bars), high gain (Fig. 8A, hatched bars), andacceleration (Fig. 8A, solid bars). Note that magnocellu-laris is shown subdivided (M1, M2, and M3) as well ascombined (MA). Figure 8B shows the percentage of fibersfrom each class that had projections into the designatednuclei.

DISCUSSION

The toadfish is a highly specialized benthic ambushpredator. The fish will remainmotionless and allow prey to

move within range before quickly launching a terminalstrike. The eyes remain relatively fixed in contrast to otherteleosts that tract prey with independent eye and bodymovements. Although visual feedback maybe of littleconsequence once a terminal strike is initiated, it providesvital prey-localization and prestrike information for bodypositioning and alignment. The functional significance ofkeeping the eyes motionless is that the black pupil aper-ture is often the most conspicuous detail of a camouflagedbenthic predator, and restricting eye movements limits therisk of visual detection by potential prey (Lythgoe, 1979).The physiological result is the absence of spontaneoussaccades and a resultant weak optokinetic and vestibulo-

Fig. 7. Synaptic bouton distribution within the vestibular-related nuclei (for identification of thenuclei, see Fig. 2) for the three classes of fibers: A: Low gain (n 5 6). B:High gain (n 5 8). C:Acceleration(n 5 5). Each dot represents a single bouton.

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 81

ocular reflex (VOR) compared with visually active preda-tors (Dieringer et al., 1992). Although it is not certain towhat extent all or some of the reflex ocular and skeletomo-tor behaviors are determined by the pathways through thevestibular-related nuclei, it does provide a foundation onwhich to interpret the morphophysiology of the vestibularnerve afferents.The variations in individual response characteristics of

vestibular nerve afferents to rotation provided a means oftyping the afferents into descriptive classes; the afferentsfell into a broad continuum across the spectrum fromlow-gain, velocity-sensitive to high-gain, acceleration-sensitive responses rather than discrete partitions (Boyleand Highstein, 1990a,b). Because of this continuum ofafferent response, it is reasonable to assume that a particu-lar vestibular reflex uses the entire spectrum of availablecontrol signals from the vestibular nerve afferents; thus, itis more appropriate to consider the vestibular nerveafferent input to a reflex in terms of preferential distribu-tion or synaptic ratios (Goldberg et al., 1987; Highstein etal., 1987; Boyle et al., 1991). Low-gain, velocity-sensitiveafferents presumably supply a major input to relay neu-rons in the VOR pathways, providing the necessary linear-ity to drive the VOR, and are likely to contribute to spinalpathways maintaining motoneuron excitability (vestibu-lar tonus). High-gain, velocity-sensitive afferents are com-parable to their primate counterparts and exhibit a morenarrow bandwidth for velocity sensitivity that may pro-

vide the vestibular input necessary for fast spinal reflexes(Boyle et al., 1992). Because the acceleration fibers havenot been phylogenetically preserved through mammals,their contribution to the vestibular reflexes is less appar-ent; their temporal response to stimulus accelerationmake them well suited for supplying vestibulospinal path-ways involved, perhaps, in escape and avoidance behavior(Boyle and Highstein, 1990a,b).The present results indicate that the three descriptive

classes of vestibular nerve afferents have unique as well asoverlapping central projection patterns and destinationsin the vestibular nuclei. Some caution should be noted ininterpreting the data. Like any experiment involvingintracellular recording and injection techniques, theremay be a bias toward selecting larger diameter fibers. Inaddition, there may be certain fiber types segregated in anarea of the peripheral nerve that is not amenable torecording. A previous bulk-label study that mapped thebrainstem octavolateralis area of the toadfish showedterminal fields from horizontal canal afferents in the EGand in each of the octavus nuclei. EG (rostral), AO (ven-tral), M, DOV, and T received the heaviest projections;DOM and DOD contained moderate terminal fields,whereas light terminal fields were observed in PO and AO(dorsal). The study did not reveal any bias of smallerdiameter fibers toward a particular nucleus (Highstein etal., 1992).Despite the limitations of intracellular injection, the

present data support the results of the bulk-label studies,with the possible exception of the anterior nuclei (High-stein et al., 1992). Due to the heavy terminal field densityobserved in the previous study, it was expected that AOand EG would receive a greater number of projections. Amore detailed analysis of the horizontal canal nerve par-tially explains this discrepancy. The horizontal canal nerveconsists of 350 afferents (Boyle et al., 1991); however, onlyapproximately 90 fibers (25%) project into the AO, withmany of these fibers continuing through to the EG toproduce dense terminal fields in the ventralAO and rostralEG (Mensinger and Highstein, unpublished). Thus, itwould be predicted that, based on the present sample size(n 5 19), approximately five to six fibers would project tothe anterior nuclei. The present study found seven fibers(36%) projecting to the AO, indicating that the intracellu-lar labeling provided an adequate statistical representa-tive of anterior-projecting fibers. However, the surprisingresult was that only two fibers (11%) were observed tocontinue into the EG. This could be attributed to termina-tion or incomplete labeling of fibers within the boundariesof the AO or a sampling anomaly, such as EG-projectingfibers that were segregated in an area of the peripheralnerve not sampled with the intracellular electrode. Al-though it appears that the present study may have under-sampled the EG, it is hypothesized that, based on therepresentative sampling of anteriorly projecting fibers,EG-projecting fibers would consist primarily of high-gainor acceleration fibers.In general, increased sensitivity and faster response

dynamics were correlated with both an increase in centralprojections and a progressive increase in morphologicalcomplexity. Low-gain, velocity-sensitive fibers were thesimplest morphologically, with the fewest number ofbranches, and projections were confined mostly to themiddle vestibular nuclei. High-gain, velocity-sensitive fi-bers were morphologically more diverse than low-gain

Fig. 8. A: Bar graph of the average number 6 S.E. of boutonscontained in each vestibular-related nucleus and subdivision forlow-gain (open bars), high-gain (hatched bars), and acceleration (solidbars) afferents. Note that the nucleus magnocellularis is shownsubdivided (M1, M2, and M3) as well as combined (MA). B: Bar graphof the percentage of fibers from each class that projected into eachnuclei. For abbreviations, see list.

82 A.F. MENSINGER ET AL.

fibers, with a greater number of branches, longer lengthand greater volume, and a more wide-spread projectionpattern. Acceleration fibers were morphologically distinctfrom low- and high-gain fibers, withmore elaborate branch-ing and greatest overall length and volume, and displayedthe most extensive central projection pattern. In line withthe broad spectrum of physiological responses within theafferent population, variability inmorphological character-istics was evident within each afferent class, with overlap-ping characteristics apparent between afferent classes.Rostrally situated nuclei received relatively sparse affer-

ent input and were preferentially innervated by high-gainand acceleration fibers. Based on innervation patterns inmammals, it was hypothesized that the entire spectrum ofafferents might project into the EG to keep the cerebelluminformed of ongoing sensory activity, making the minimalafferent input to EG surprising. In the mammal, thecerebellar flocculus receives principally secondary vestibu-lar projections, whereas the nodulus and uvula receiveprimary vestibulocerebellar fibers from the vestibularnerve (Voogd et al., 1996). It is presently unknownwhetherthe EG corresponds to the floccular or nodulouvular granu-lar cells in mammals (Larsell, 1967). Perhaps most signifi-cant was the scarcity of low-gain fiber projection to eithertheAO or the EG, indicating that, if low-gain fibers projectto the cerebellum, then they may do so through secondaryprojections.TheAO receives input from both auditory and vestibular

organs, with the auditory portion concentrated in thedorsal two-thirds and the vestibular projections occupyingthe lower one-third of the nucleus (Highstein et al., 1992).It was hypothesized that, if this anatomical separationcontinued to secondary neurons, then the presumed targetnucleus of the ventral or vestibular portion of the AOwould be vestibular reflex-related nuclei, such as theoculomotor complex. However, if the projections do includethe oculomotor-related nuclei, then the results are con-trary to the expectation, generated by mammalian data(Highstein et al., 1987), that low-gain fibers would projectinto vestibuloocular reflex pathways and, thus, would beabundant in the AO. This lack of low-gain fibers in the AOsuggests that the AO may not be a VOR relay nucleus.Indeed, in the goldfish, it has been suggested that the DOperforms this function (Torres et al., 1992). The toadfishVOR does not seem to be highly developed, as quick phasesare rare, although slow phases are present (Dieringer etal., 1992). VOR circuitry in the toadfish remains a topic forfuture study. Retrograde transport of tracer from theoculomotor, trochlear, and abducens nuclei should resolvethis apparent inconsistency. Alternatively, it would seemmore plausible that, based on the response dynamics andanatomical position, the labeled fibers in the AO arerelaying signals to the cerebellum or into other pathwaysrather than to the oculomotor nuclei.

Magnocellularis received input from all three fiber types,with equal projections from low-gain and accelerationfibers that were approximately twice the number fromhigh-gain fibers. Intranuclear segregation was apparent,with low- and high-gain fibers projecting primarily to M3and acceleration fibers to M2. The magnocellularis isbelieved to be the homologue of the mammalian Deiter’snucleus, giving rise to the ipsilateral descending lateralvestibulospinal tract, which controls compensatory limbreflexes. The nucleus receives both octavus and lateralisinput, with chemical and putative electrotonic inputs fromboth sources (Korn et al., 1977; Highstein et al., 1992),suggesting that the toadfish needs comparably fast activa-tion of this nucleus and its output to extensor or antigrav-ity (buoyancy) axial muscles. The innervation patternindicated that all three classes play a role in this activity,with possible functional segregation between M2 and M3.The continuous firing of low-gain fibers might act as a fineadjustment mechanism to correct or stabilize small headrotations. Alternatively, the response dynamics may pro-duce ‘‘vestibular tonus’’ or the maintenance of a high levelof neuronal excitability as a result of a continuous input tointerneuronal and motoneuronal cell groups from fiberswith high discharge rates. The high-gain and accelerationfibers would provide fast positioning information duringdynamic movements, such as predation or escape.Nucleus tangentialis (T), the only purely vestibular

nucleus (Highstein et al., 1992), was preferentially tar-geted by low-gain fibers, with moderate input from high-gain fibers and no input from acceleration fibers. Intra-nuclear segregation again was apparent, as low-gain fibersterminated throughout the nucleus, whereas high-gainfibers were mostly concentrated in the lateral posteriorarea of the nucleus. Based on the physiological propertiesof the low-gain afferents and the distribution of thesefibers throughout the nucleus, it would appear that theentire nucleus may function in the VOR pathway. How-ever, the parcellation of high-gain fibers in the posteriorhalf of T, which is surrounded by anterior portions of theDOV, suggests that the secondary neurons of these high-gain fibers may exit through the DOV to form vestibulospi-nal or vestibulocolic reflex pathways. The lack of accelera-tion fibers in this nucleus combined with a low boutondensity in anterior DOV suggest that acceleration fibersavoid deep anteroventral placement, which would be farfrom putative extensor motor neuron targets.The DO was the primary target of afferent projections.

The subdivision of the nucleus into three regions revealeddifferential projection patterns, suggesting functional seg-regation. The rostral portions of the nucleus had beenhypothesized as the vestibuloocular relay nucleus basedon the heavy projection into this area of semicircular canalfibers having appropriate response characteristics to con-tribute to eye rotation. The canals also have slightly

TABLE 2. Synaptic Bouton Distribution of Primary Afferent Fibers1

Fiberclass

Nuclei

EG AO MA T DOD DOV DOM PO

Low gain 0 0.2 6 0.2 7.0 6 5.1 8.3 6 3.1** 5.7 6 2.7 66.0 6 30.5* 4.7 6 3.7 0High gain 0.4 6 0.4 4.4 6 2.2 3.0 6 0.8 3.6 6 1.2 1.3 6 1.0 66.6 6 23.9* 1.0 6 0.9 0Acceleration 1.4 6 1.3 5.0 6 4.0 8.8 6 4.9 0 23.8 6 11.0**† 59.8 6 14.4* 1.0 6 0.5 6.6 6 2.0**

1Mean values 6 S.E. EG, eminentia granularis; AO, anterior octavus; MA, magnocellularis; T, tangentialis; DOD, dorsal descending octavus; DOV, ventral descending octavus;DOM, rostral finger of descending octavus; PO, posterior octavus.*Indicates significantly different intraclass mean (P , 0.01; ANOVA), †(P , 0.05; ANOVA).**Indicates significantly different interclass means (P , 0.05; ANOVA).

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 83

differential central projections into this area, consistentwith putative separate secondary relay projections to thesubgroups of extraocular motor neurons (Carey and High-stein, 1990; Torres et al., 1992). The projections to therostral finger, although they were relatively light, weredominated by low-gain fibers, which was consistent with aVOR pathway. Additional support for this pathway wasgained by the projections of low-gain fibers to the adjacentrostral dorsal DO. The rostral low-gain projections werenot in an area overlapped by other DOD fibers, indicatingthat the rostral portion of the DOD may merge with therostral middle finger and form a VOR pathway.Dorsal DO received relatively heavy innervation from

acceleration fibers, moderate innervation from low-gainafferents, and little innervation from high-gain afferents.Low-gain fibers were segregated rostrally and caudally.The posterior cluster of low-gain boutons indicated thatthe low-gain fibers may provide some input into the spinalpathways. The projection pattern from acceleration fiberswas uniformly distributed throughout the remainder ofthe nucleus, consistent with targeting vestibulospinalpathways. The lack of high-gain fibers indicated that thereis intranuclear segregation of high-gain and accelerationfibers, suggesting that the latter may have a separatedorsal pathway to the spinal motor neurons.Ventral DO received the greatest number of projections,

with input evenly divided among the afferent populations.Approximately two-thirds of each fiber distribution werein the caudal half of the nucleus, with very heavy boutonconcentration in the caudal one-third. The overlappingprojection patterns prevented any differentiation of class.There was a heavy concentration of low-gain fibers in thecaudal one-third, again implying that at least some low-gain fibers target spinal circuits. The high-gain and accel-eration fibers were also concentrated, as expected, in thelower end of the nucleus close to the spinal neurons,consistent with their putative role in innervating thevestibulospinal and vestibulocolic pathways.PO was innervated exclusively by acceleration fibers.

The PO’s caudal location in the brainstem implies that theacceleration afferents target ipsilateral vestibulospinaltract neurons; this termination pattern of accelerationafferents is consistent with the notion that their responsedynamics are best tailored to the dynamic load require-ments of the spinal reflexes. The acceleration fiber termina-tions were separated into anterior and posterior clusters.Whether this is a physiological separation or simply amorphological partitioning, perhaps by the innervationsite of the tenth nerve, remains to be determined. Thepossibility exists that the acceleration fiber target neuronsof both the DOD and the PO unite to form a separateacceleration-sensitive spinal reflex pathway.The results show a clear difference in distribution and

morphology of the three physiologically typed vestibularnerve afferents. The questions remain: what dictates fibermorphology, and to what extent does the afferent’s struc-ture affect its intrinsic signal? Acceleration fibers are themost morphologically complex, followed, in order, by high-gain and low-gain fibers. Does the intrinsic signal carriedby each class determine the morphology, or is it simply thephysical area of the circuit? For example, because of theirmore extensive innervation patterns, acceleration fiberswere longer and, thus, developed more branches andboutons. However, physical distance does not necessarilycorrelate with bouton number or branching, because long,

unbranched acceleration fibers with few boutons couldhave also produced the same extensive innervation pat-tern. Therefore, it appears that themorphological complex-ity is correlated with the type of signal transmitted.Future studies of central neurons will help determine theeffect of fiber morphology on the transfer function.Based on the morphological trends, it was predicted that

bouton number would also increase with faster dynamicsand increased sensitivity. In the semicircular canal crista,acceleration afferents average a two- to fourfold greaternumber of synaptic contact sites on hair cells than thehigh-gain and low-gain afferents (Boyle et al., 1991).Although little difference was seen in bouton numberbetween low- and high-gain fibers, acceleration fibers didpossess 20–30% more boutons than the other two classes.Therefore, in contrast to the relatively localized pattern oflow-gain fibers, acceleration-related information is con-veyed not only over a wider area of the vestibular nuclei(fiber length) but also affects more cells (boutons).In summary, the three physiological classes of vestibular

nerve afferents had unique as well as overlapping centralprojections patterns, with intranuclear parcellation in themagnocellularis, tangentialis, and DO. The projectionpatterns demonstrated that there are anatomically demon-strable differential central projections of canal afferentswith different response dynamics within the vestibularcomplex of the fish. By comparing the central projections ofthe two most disparate classes of fibers and toadfishbehavior, an understanding of the underlying basis for thisdifferential parcellation may be achieved. Evolutionarypressure has selected for a relatively modest VOR inbenthic ambush predators (Dieringer et al., 1992). Al-though a reduced VOR maybe appropriately tuned for thistype of environment, the consequences of a weak VORmayhave resulted in a limited central projection pattern.Alternatively, the secondary neurons in the VOR pathwaymay simply be more limited or confined than in thevestibulospinal tract, resulting in a more limited projec-tion pattern. In contrast to low-gain fibers, the accelera-tion signals were distributed (with the exception of tangen-tialis) throughout the vestibular complex compared withthe relatively limited projection of low-gain fibers. Accelera-tion events, although they are relatively rare in thetoadfish compared with pelagic species, are necessary forpredation and defense. Consequently, the wide distribu-tion of acceleration fibers in the vestibular-related nucleiunderscores the vital role of these afferents in reflex headand body control.In early vertebrate phylogeny, the problem of equilib-

rium and balance was solved in an efficient and remark-ably conserved fashion. However, evidence of central segre-gation in higher animals has been elusive despite thehomology between the peripheral and at least some of thecentral vestibular components in fish and in land verte-brates. The concomitant complexities associated with ter-restrial evolution and an ever-increasing neuropil mayhave masked or overlapped this segregation in highervertebrates. This study represents the first clear evidencethat there is central parcellation of disparate peripheralphysiological vestibular classes.

ACKNOWLEDGMENTS

The authors thank Dr. A. Bass for a critical reading ofthe paper and P. Keller for histological support.

84 A.F. MENSINGER ET AL.

LITERATURE CITED

Anderson, J.H., R.H. Blanks, and W. Precht (1978) Response characteris-tics of semicircular canal and otolith systems in cat. I. Dynamicresponses of primary vestibular fibers. Exp. Brain Res. 32:491–507.

Blanks, R.H., M.S. Estes, and C.H. Markham (1975) Physiologic character-istics of vestibular first-order canal neurons in the cat. II. Response toconstant angular acceleration. J. Neurophysiol. 38:1250–1268.

Boyle, R., and S.M. Highstein (1990a) Resting discharge and responsedynamics of horizontal semicircular canal afferents of the toadfish,Opsanus tau. J. Neurosci. 10:1557–1569.

Boyle, R., and S.M. Highstein (1990b) Efferent vestibular system in thetoadfish: Action upon horizontal semicircular canal afferents. J. Neuro-sci. 10:1570–1582.

Boyle, R., J.P. Carey, and S.M. Highstein (1991) Morphological correlates ofresponse dynamics and efferent stimulation in horizontal semicircularcanal afferents of the toadfish, Opsanus tau. J. Neurophysiol. 66:1504–1521.

Boyle, R., J.M. Goldberg, and S.M. Highstein (1992) Inputs from regularlyand irregularly discharging vestibular nerve afferents to secondaryneurons in squirrel monkey vestibular nuclei. III. Correlation withvestibulospinal and vestibuloocular output pathways. J. Neurophysiol.68:471–484.

Carey, J., and S.M. Highstein (1990) Physiology and brainstemmorphologyof single horizontal semicircular canal afferents in the toadfishOpsanustau: A biotin dye study. Soc. Neurosci. Abstr. 16:372.

Dieringer, N., I. Reichenberger, and W. Graf (1992) Differences in optoki-netic and vestibular ocular reflex performance in teleosts and theirrelationship to different life styles. Brain Behav. Evol. 39:289–304.

Fernandez, C., and J.M. Goldberg (1971) Physiology of peripheral neuronsinnervating semicircular canals of the squirrel monkey. II. Response tosinusoidal stimulation and dynamics of peripheral vestibular system. J.Neurophysiol. 34:661–751.

Fernandez, C., R.A. Baird, and J.M. Goldberg (1988) The vestibular nerveof the chinchilla. I. Peripheral innervation patterns in the horizontaland superior semicircular canals. J. Neurophysiol. 60:167–181.

Fernandez, C., A. Lysakowski, and J.M. Goldberg (1995) Hair-cell countsand afferent innervation patterns in the cristae ampullares of thesquirrel monkey with a comparison to the chinchilla. J. Neurophysiol.73:1253–1269.

Goldberg, J.M., and C. Fernandez (1971) Physiology of peripheral neuronsinnervating semicircular canals of the squirrel monkey. I. Restingdischarge and response to constant angular accelerations. J. Neuro-physiol. 34:635–660.

Goldberg, J.M., S.M. Highstein, A.K. Moschovakis, and C. Fernandez(1987) Inputs from regularly and irregularly discharging vestibularnerve afferents to secondary neurons in the vestibular nuclei of thesquirrel monkey. I. An electrophysiological analysis. J. Neurophysiol.58:700–718.

Highstein, S.M., J.M. Goldberg, A.K. Moschovakis, and C. Fernandez(1987) Inputs from regularly and irregularly discharging vestibularnerve afferents to secondary neurons in the vestibular nuclei of thesquirrel monkey. II. Correlation with output pathways of secondaryneurons. J. Neurophysiol. 58:719–738.

Highstein, S.M., R. Kitch, J. Carey, and R. Baker (1992) Anatomicalorganization of the brainstem octavolateralis area of the oyster toad-fish,Opsanus tau. J. Comp. Neurol. 319:501–518.

Huwe, J.A., and E.H. Peterson (1995) Differences in the brainstem termina-tions of large- and small-diameter vestibular primary afferents. J.Neurophysiol. 74:1362–1366.

Ishizuka, N., S. Sasaki, and H. Mannen (1982) Central course and terminalarborizations of single primary vestibular afferent fibers from thehorizontal canal in the cat. Neurosci. Lett. 33:135–139.

Korn, H., C. Sotelo, and M.V.L. Bennett (1977) The lateral vestibularnucleus of the toadfish Opsanus tau: Ultrastructural and electrophysi-ological observations with special reference to electrotonic transmis-sion. Neuroscience 2:851–884.

Landolt, J.P., and M.J. Correia (1980) Neurodynamic response analysis ofanterior semicircular canal afferents in the pigeon. J. Neurophysiol.43:1746–1770.

Larsell, O. (1967) The Comparative Anatomy and Histology of the Cerebel-lum From Myxinoids Through Birds. Minneapolis, MN: University ofMinnesota Press.

Lysakowski, A., L.B. Minor, C. Fernandez, and J.M. Goldberg (1995)Physiological identification of morphologically distinct afferent classesinnervating the cristae ampullares of the squirrel monkey. J. Neuro-physiol. 73:1270–1281.

Lythgoe, J.N. (1979) The Ecology of Vision. Oxford: Clarendon Press.McCormick, C.A. (1982) The organization of the octavolateralis area in

actinopterygian fishes: A new interpretation. J. Morphol. 171:159–181.O’Leary, D.P., R.F. Dunn, and V. Honrubia (1976) Analysis of afferent

responses from isolated semicircular canal of the guitarfish usingrotational acceleration white-noise inputs. I. Correlation of responsedynamics with receptor innervation. J. Neurophysiol. 39:631–644.

Rabbitt, R.D., R. Boyle, and S.M. Highstein (1994) Sensory transduction ofhead velocity and acceleration in the toadfish horizontal semicircularcanal. J. Neurophysiol. 72:1041–1048.

Rabbitt, R.D, R. Boyle, and S.M. Highstein (1995) Mechanical indentationof the vestibular labyrinth and its relationship to head rotation in thetoadfish,Opsanus tau. J. Neurophysiol. 73:2237–2260.

Sato, F., andH. Sasaki (1993)Morphological correlations between spontane-ously discharging primary vestibular afferents and vestibular nucleusneurons in the cat. J. Comp. Neurol. 333:554–566.

Sato, F., H. Sasaki, N. Ishizuka, S. Sasaki, and H. Mannen (1989)Morphology of single primary vestibular afferents originating from thehorizontal semicircular canal in the cat. J. Comp. Neurol. 290:423–439.

Schneider, L.W., and D.J. Anderson (1976) Transfer characteristics of firstand second order lateral canal vestibular neurons in gerbil. Brain Res.112:61–76.

Tomko, D.L., R.J. Peterka, R.H. Schor, and D.P. O’Leary (1981) Responsedynamics of horizontal canal afferents in barbiturate-anesthetized cats.J. Neurophysiol. 45:376–396.

Torres, B., A.M. Pastor, B. Cabrera, C. Salas, and J.M. Delgado-Garcia(1992) Afferents to the oculomotor nucleus in the goldfish (Carassiusauratus) as revealed by retrograde labeling with horseradish peroxi-dase. J. Comp. Neurol. 324:449–461.

Voogd, J., N.M. Gerrits, and T.J.H. Ruigrok (1996) Organization of thevestibulocerebellum.Ann. N.Y. Acad. Sci. 781:553–579.

DIFFERENTIAL CENTRAL PROJECTIONS IN TOADFISH 85