THE JOURNAL OF COMPARATIW NEUROLOGY 293:347-376 (1990)

Structural and Functional Organization of a Diencephalic Sensory-MotorInterface in the Gymnotiform Fish, Eigenmannia CLJFFORD H. KELLER,LEONARD MALER, AND WALTER HEILIGENBERG Scripps Institution of Oceanography, University of California San Diego, La Jolla, California 92093 (C.H.K., W.H.); Department of Anatomy, School of Medicine, Universit,y of Ottawa, Ottawa, Canada K1H 8M5 (L.M.)

ABSTRACT The diencephalic nucleus electrosensorius (nE) of gymnotiform fish comprises a series of finely tuned neuronal filters for control of the jamming avoidance response (JAR) and probably other electromotor tasks as well. The nE receives electrosensory input from the dorsal torus semicircularis (TSd) and octavolateral input from the ventral torus (TSv). The nE, in turn, projects to various hypothalamic and thalamic nuclei, including the prepacemaker nucleus (PPn), which can modulate the frequency of electric organ discharges (EODs) via its unique input to the medullary pacemaker nucleus. Four subdivisions of the n E can now be recognized: I ) The beat-related area (nEb)-a rostra1 cluster of tightly packed cells which receives TSd input and project.s to the inferior lobe, anterior tuberal nucleus, anterior thalamic nucleus, central posterior thalamic nucleus, and PPn. The nEb contains neurons responsive to beat patterns caused by jamming stimuli. Stimulation of the nEb with L-glutamate, however, fails to induce any EOD-frequency shift. 2) The area causing EOD-frequency rises (nEt)-a horizontal band of cells at the dorsal aspect of the caudal nE which receives TSd input and projects to the PPn and vicinity and to the cerebellum; nEt stimulation induces slow EOD-frequency rises characteristic of the JAR. Responses of these cells to jamming stimuli are not yet known. 3) The area causing EOD-frequency falls (nEl)-a horizontal band of cells at the ventral aspect of the caudal nE which receives TSd input and projects only to the PPn and vicinity; nE! stimulation induces slow EOD-frequency falls characteristic of the JAR. The responses of these cells to jamming stimuli are not yet known. 4) The acousticolateral region (nEar)-a complex medial region of the nE which receives input predominantly from the ventral torus and projects to the inferior lobe, anterior tuberal nucleus, central posterior thalamic nucleus, PPn, and cerebellum; the sensory and motor properties of this region are not known in detail, although auditory and mechanosensory responses have been recorded here. Projections to the P P n and its vicinity suggest direct control of electromotor behaviors by the nE, whereas thalamic and hypothalamic projections may provide a substrate for electrosensory influences on neuroendocrine and motivational control centers. The optic tectum projects strongly to the pretectum and various other diencephalic nuclei in the vicinity of the nE, but it does not innervate the nE itself. Accordingly, ablation of the tectum does not affect the performance of the JAR. Key words: electroreception, neuroethology, pretectum, teleost, thalamus

Accepted September 29,1989.

0 1990 WILEY-LISS, INC.

348

C.H. KELLER ET AL.

The teleost diencephalon is a complex of many small and interconnected nuclei that presents a formidable challenge to the understanding of its functional organization. Despite a large literature identifying diencephalic cell groups, their patterns of afferent and efferent connectivity, and possible homologies across thp fishes (e.g., Braford and Northcutt, ’83). ascribing functional significance to these patterns has met with little success. Recent work demonstrating sensory and motor characteristics of two nuclei within the diencephalon of weakly electric gymnotiform fish (Eigenmannza, Apteronotus) suggests that the nucleus electrosensorius ( n E , Keller a n d Heiligenberg, ’89) a n d t h e prepacemaker nucleus (PPn, Kawasaki et al., ’88)together compose a sensory-motor interface for control of a reflexlike electromotor behavior, the jamming avoidance response (JAR, for review see Heiligenherg, ’86). Physiologically and anatomically distinct subnuclei have been identified within the n E and PPn, which play well-defined roles for the control of this and other electromotor behaviors. The ability to examine these roles experimentally has been instrumental for this study of the functional organization of the diencephalon. Gymnotiform fish use their electrosensory/motor systems for object detection (rlectrolocation) and intraspecific communication. Natural objects moving relative to the fish cause phase and amplitude modulations in the fish’selectric field that provide cues for object detection. Electric fields of neighboring fish of similar electric organ discharge (EOD) frequencies mix to cause similar electric field modulations, and such “jamming” is deleterious to the fish’s electrolocating abilities (see review by Bastian, ’86).By shifting its EOD frequency away from similar interfering frequencies, a fish can minimize the deleterious effects of jamming. The simplicity and robustness of this jamming avoidance response have made this behavior a valuable model system in neuroethology. The present work demonstrates the pattern of afferent input to and efferent pathways from the n E complex, thus confirming and expanding upon earlier findings of Carr et al. (’81) and Rastian and Yuthas (’84).We also identify cell types within functionally defined subregions of the nE that are responsible for these efferent pathways. Intracellular labelling of physiologically defined cells in studies now underway will allow correlation of functional, and anatomical properties of neuronal types identified in the present study.

MATERIALS AND METHODS The glassknife fish, Eigenmannia, used throughout this study were of either sex, were 10 to 15 cm in total length, and were bought from local dealers or raised in the laboratory. Fish were immobilized by injection of Flaxedil (circa 5 pl 0.0002‘ c gallamine triethiodide) or immersion in urethane (2’0 in aquarium water), held in the center of an experimental tank within a foam-lined forceps and respirated with aquarium water or a dilute urethane solution. The fish’s stable maintenance was assured by monitoring the attenuated EOD with a suction electrode placed over the fish’s tail. After topical application of lidocaine, a small hole (1-4 mm) was drilled in the skull above the optic tectum to allow access to stimulation and injection sites. Following injection of a neuronal tracer, a small piece of gelfoam was placed in the wound which was then closed with Vetbond (3M). The fish was allowed to recover, and it survived for 1 to 4 days

before it was immersed in a solution of MS-222 and then perfused with heparinized fish Ringer containing lidocaine (0.02 “ 0 ), followed by 4 Yo paraformaldehyde in phosphate buffer. Following postfixation of 5 to 24 hours, the brains were cut on a vibratome into 50 wm sections, which were then incubated, mounted, and counterstained with neutral red or methylene green.

Abbreviations A

a ALLNG B APT ATh AVI CANS Cb CE CP DFI DLTh DMTh D Pn DTn ELL EOD eTs Fb G H Hd,HI IRA JAR LMRa

MgT

MLF nB nE nEar nEb nE 1 nET nLT nPC nPPv nRLl,nRLm

PC PrG PG ,PGl,PGm PPn, PPn-C,PPn-G PT,PTI PTh PTPv RMT RI, R SC SE 1‘A TeO TL TP TPP TSd TSv V VLTh VMTh

pretectal nucleus “A” axon anterior lateral line nerve ganglion pretectal nucleus “B” area pretectalis anterior thalamic nucleus area ventrolateralis (thalamus) commissnra ansiilata cerebellum central nucleus of the inferior lobe central posterior nucleus nucleus diffusus lateralis of the inferior lobe dorsolateral thalamus dorsoinedial thalamus dorsal posterior nucleus (thalamus) dorsal tegmental nucleus electrosensory lateral line lobe electric organ discharge efferents of the torus semicircularis forebrain glomerular nucleus habcnula hypothalamus, dorsal and lateral subdivisions interstitial reticular area jamming avoidance response lateral mesencephalic reticular area magnocellular tegmental nucleus medial longitudinal fasciculus nucleus a t the base of the optic tract nucleus electrosensorius nucleus electrosensorius-acousticolateralis region nucleus electrosensorius-beat related area nucleus electrosensorius-area causing EOD-frequency falls nucleus electrosensorius-area causing EOD-frequency rises lateral tuberal nucleus niicleus of the posterior commissure posterior periventricular nucleus nucleus recessus lateralis, lateral and medial subdivisions posterior commissure periglomerular nucleus preglomerular nucleus, lat.era1 and medial subdivisions prepacemaker nucleus, chirp and gradual subdivisions pretectal nucleus, lateral subdivision prethalamic nucleus periventricular pretectal nucleus rostra1 mrsencrphalic tegmental nucleus lateral recess of the third ventricle red nucleus suprachiasmatic nucleus nucleus sub-electrosensorius anterior tuberal nucleus optic tectum torus longitudinalis posterior tuberal nucleus prriventricular nucleus of the posterior tuberculum torns semicircularis dorsalis torus semicircularis ventralis vent.ricle ventrolateral thalamus ventromedial thalamus

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EPFERENTS

P

.

I .

f-.Ij

TeO

349

T a

Ch

Imm

TSV

Fig. 1. A: Lateral view of the brain of Eigenmannia indicating the location (*) of the nNt. Approximate plane of sectioning indicated by arrows. B: Transverse section of the brain of Eigennnnnin a t the level of the nE1 (equivalent to the level of Fig. 4). More detail is presented in subsequent figures. Methylene green stain, 50 urn section thickness.

350

Fig. 2.

Transverse section through the caudal thalamus of Eigen(level is 150 pm caudal t o the center of the nEf as depicted in Figs. 1 and 4). A T h e major nuclear groups seen a t this level. B: An enlarged view of the area enclosed by the rectangle in A. This enclosed area is shown schematically in C and D with efferents labelled by PHAL injection into the ventral torus (TSv; in C) or the dorsal torus (TSd; in W L Q ~ ~ ~ L Z

C.H. KELLER ET AL.

D). C: A strong projection from TSVto the interstitial reticular area (IRA) is seen, whereas D: T S d efferents primarily pass through this level in a rostra1 direction and form only sparse terminals. Lines represent fibers, dots denote terminals, filled circles represent retrogradely labelled somata. Scale bars = 200 Wm. Methylene-green stain, 50 fim section thickness.

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

/ RMT

\

Fig. 3. Transverse section 100 pm caudal to the center of nE1 (Figs. 1,4).Same presentation as in Figure 2. C : T h e rostral-most extent of IRA receives a strong projection from TSV.Additionally, a few TSv fibers and terminals are seen medially adjacent to the TSd projection (D). A few cells mark the caudal extension of the band of cells which forms the nEf more rostrally. These cells are found in an area receiving TSd and/or TSv terminals.

The known spatial relationship of intended injection sites to areas whose stimulation causes specific EOD-frequency modulations facilitated accurate placement of relatively small injections of neuronal tracers (Keller and Heiligenberg, '89). EOD modulations, monitored via the tail electrode, were induced by local iontophoresis of the ubiquitous

excitatory transmitter L-glutamate from one barrel of a multibarrel electrode. L-glutamate is known to excite somata and dendrites but not axons (Stone, '85). Following localization of an area by glutamate stimulation, a neuronal tracer, horseradish peroxidase (HRP, Sigma, ca. 10% in 0.1 M Tris buffer, pH 7.6), wheat germ agglutinin-conjugated

352

Fig. 4. Transverse section at the center of the nET (Fig. 1; see Keller and Heiligenberg, '89) presented as in Figure 2. Borders for the nE are drawn to coincide with cell groups receiving TSVor TSd terminals. C: TSVprojection generally fills the nEar without discernible concentra.

C.H. KELLER ET AI,.

tions of terminals; sparser TSVterminals are found within both the nE' and nEl. D: Two distinct proliferation zones of TSd-efferent terminals

coincide with nE+ and nE1.

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

Fig. 5. Transverse section 50 Frn rostra1 to the center of nK':T(Pigs. 1,4J presented as in Figure 2. C: The TSVprojection is primarily medial to that of TSd with sparse TSVterminals within the medial nEt and nE!. D: Two distinct zones of TSd-efferent terminals coincide with nET and nEj.

353

354

C.H. KELLER ET AL.

7 C:TSv Fig. 6. Transverse section 100 fim rostral to the center of nET (Figs. 1,4)presented a s in Figure -. projection is quite evenly spread throughout nEar, impinging only slightly on the medial aspect of nEi. Several cells within the deeper portions of the nE apparently project to TSV.D: Dense TSd-efferent terminals fill the nEl and are also seen outside of this cell cluster within a ventrally directed arc of fibers.

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

355

Fig. 7. Transverse section 200 pm rostral to the center of nET (Figs. 1, 4) a t the caudal edge of nEb. Presented as in Figure 2. C: TSv-efferent terminals are evenly spread throughout much of the nE. A retrogradely labelled soma with major dendrites is also shown. D: T S d efferents continue within a ventrally directed arc, terminals proliferate on the caudal edge of nEb.

HRP (WGA-HRP, Sigma, ca. 2% in 0.1 M Tris buffer, pH 7.6) or Phaseolus-leucoaglutininlectin (PHA-L, Vector, ca. 2.55 in 50 mM Tris buffer, pH 8.5), was iontophoresed locally through another barrel of the same electrode. The usual technique for achieving small, dense injections was to fill the tip of a micropipette with the desired amount of con-

rentrat,ed tracer and to “top-off’ the capillary with a 3 % solution of Alcian-blue (in 0.2 M acetate-acetic acid buffer, pH 4.0). Several microamps of positive current applied briefly to the Alcian-blue solution forced it down the capillary to act as a plunger, pushing the tracer out ahead of it. HRP and WGA-HRP were visualized with either the TMB

C.H. KELLER ET AL.

356

Fig. 8. Transverse section 300pm rostral to the center of nET (Figs. 1,4). Presented as in Figure 2 . C: TSV efferents are confined t o the more medial aspect of the nucleus and the edges of nEb. D: TSd efferents encircle the nEb; terminals appear to lie around but not within nEb; fihers continue ventrally towards the postoptic commissure.

method (Mesulam and Rosene, '79) or the DAB and glucoseoxidase reaction (Metcalfe, '85). PHA-L was visualized by using a protocol modified from Gerfen and Sawchenko ('84), which uses a primary antibody (rabbit anti-PHA-L, Dakopatts) followed sequentially by a secondary antibody (goat anti-rabbit, Vector) an avidin-biotin complex (Vector ABC

kit) and the DAB-glucose-oxidase reaction of Metcalfe ('85). Portions of the optic tectum were aspirated in seven fish held under urethane anesthesia. Following survival times of 1, 2 , 3 , 4 (2 fish), 5, or 6 days the fish were immersed in MS222 and perfused with Ringer followed by 10% formalin.

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

D

Fig. 9. Transverse section 350 pm roatral to the center of nEt (Figs. 1, 4).Presented as in Figure 2. C: TSv-efferentprojection is relatively sparse within the more medial aspect of nE. D: TSd efferents encircle but do not enter the nEb. Fibers continue ventrally to decussate in the postoptic comrnissure.

357

C.H. KELLER ET AL.

358

c

I Fig. 10. Transverse section 400 wrn rostra1 to the center of nEf (Figs. 1,4).Presentation as in Figure 2. C TSvlefferent terminals are very sparse. D TSd efferenh encircle b u t do not enter the nEb. Fibers continue ventrally to decussate in the postoptic commissure.

359

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

Fig. 11. Transverse section 550 fim rostra! to the center of nEt (Figs. 1 , 4 ) and rostral of the nEb. Presentation as in Figure 2. C There are nn TSv-efferent terminals laterally. Some fibers pass ventrally t o decussate within the postoptic commissure. D. TSd efferents pass through this

level to decussate within the postoptic commissure; only very sparse terminals are seen. A separate area of termination, however, extends rnedially to the midline within and near to the VLTh.

Brains were postfixed for approximately 1 week, then sectioned at 60 wm on a vibratome and postfixed for another one to several weeks. Degenerating axons and terminals were visualized with the Fink-Heimer silver-impregnatinn technique as outlined in Heimer ('70). T o test the role of the optic tectum (TeO) in the JAR, large portions of the tectum of one fish of the genus Apteronotus and one Eigenmannia were aspirated. Apteronotus is more robust than Eigenmannia and was better able to survive these extensive lesions. Each fish's JAR was tested before and after lesioning the Te0. A sinusoidal jamming st,imulus was held in transverse orientation and frequency-

clamped by computer alternately 4 Hz above and 4 Hz below the frequency of the fish's own EOD or a mimic (after Heiligenberg and Bastian, 'SO). Following TeO lesions and retesting of the ,TAR, each fish was heavily anaesthetized in MS222 and perfused for later histological confirmation of the extent of the lesions.

RESULTS Background The anatomical substrate of the electrosensory system is well characterized within gymnotiforrn fish, particularly in

C.H. KELLER ET AL.

360

B

\

-/’--/I

,--

/’ -

x

.-

Fig. 12. A. Injections into the ventral torus semicircularis (TSv) retrogradely label a small number of cells of several cell types within the nE. Three such cells are shown in the left column; right column sets the context (top and bottom drawings at the level of Fig. 6, middle drawing

at the level of Fig. 4). Upward- and downward-pointing arrowheads

regard to the control of the JAR in Eigenmannia and A p t e ronotus (for review see Carr and Maler, ’86). Electrosensory information enters the diencephalon via the dorsal torus semicircularis (TSd) and the optic tectum (TeO). Initial processing for control of the JAR is accomplished up to and within the dorsal torus which has a strong efferent projection to the n E (Carr et al., ’81). The tectum also receives a dorsal torus efferent projection and plays a major role in spatial electrosensory tasks (Carr e t al., ’81; Bastian, ’86). Earlier reports suggested that the TeO is also necessary for performance of the JAR (Rose and Heiligenberg, ’86); this role is reconsidered in the present paper. Cells have been recorded within both the dorsal torus and the tectum which are responsive to modulations of’ the amplitude and phase of the fish’s electric organ discharge (EOD) or its mimic. Such modulations are caused by objects moving in the fish’s environment as well as by E o n s of neighbors, which cause interference patterns that beat a t a rate equal to the frequency difference between the two EODs. Certain of these “beat-sensitive” cells, so-called “sign-selective” cells, respond maximally to specific combi-

nations of amplitude and phase modulation that reflect the sign of the difference frequency. Sign-selective and other beat-sensitive cells have also been recorded within the n E (Bastian and Yuthas, ’84; Keller, ’88) where responses of some of these cells display strong selectivity for JAR-related stimuli. This degree of specifity and robustness is not seen within the torus or tectum (Keller, ’88).Many of these beatrelated cells are found within an isolated cell-cluster located within the rostra1 n E (the “nucleus electrosensorius-beatrelated area,” nEb; see below). Beat-related responses are occasionally recorded throughout much of the nE, however, suggesting that “beat-relatedness’’ is not an exclusive property of cells within nEb. More caudally within the nE are found two areas which control JAR-related frequency rises ( n E f ) or frequency falls (nEl), respectively (Keller and Heiligenberg, ’89). Stimulation of either of these latter two areas by local iontophoresis of L-glutamate elicits a corresponding EOD-frequency shift, while bilateral lesion of either one eliminates the corresponding frequency shift in the ,JAR. Numerous other physiological cell types are found within the nE. They respond to rapid onset or offset of

denote the nET and nEl, respectively. B Injections into the nE label many cells of various types within the TSV.Scale bars 50 pm. ~

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

36 1

Fig. 13. A A medial subdivision of the periventricular pretectal nucleus (PTPv) projects to the TSd. Chartings of cells backlabelled from large HRP injections into the TSd. Fibers extend into several thalamic and pretectal nuclei and especially strongly into the magnocellular tegmental nucleus (MgT). B: Cells backlabelled within the central nucleus of the inferior lobe (CE) after HRP injection to TSd. Scale bars = 200 um.

EOD-related electric fields. to low-frequency electric signals, to mechanosensory, or to acoustic stimuli. These various single-unit responses suggest that the n E consists of an array of finely tuned filters for specific sensory information. The prepacemaker nucleus, lying at the border of the diencephalon and mesencephalon, provides the only known input to the medullary pacemaker nucleus, which controls the fundamental frequency of the EOD (Heiligenberg et al., '81). Two subnuclei have been identified within the PPn, from which either smooth JAR-like or abrupt chirp-like EOD-frequency rises can be elicited by local stimulation (Kawasaki et al., '88). Sign-selective cells have been recorded within the more anterior of these subnuclei, the PPn-G, from which smooth EOD-frequency rises can be elicited. Responses of these cells closely follow characteristics of the intact behavior and are thought to be the final common pathway for JAR-related pacemaker modulations (Rose et al., '87). The more caudal and lateral sub-nucleus, the PPn-C, contains large multipolar cells with extensive dendritic fields. PPn-C neurons are responsible for rapid

chirp-like EOD-frequency rises. The P P n thus appears to integrate various inputs for final control of electromotor behaviors. A possible direct link between the nE and the P P n was suggested by rather large HRP injections into the nE (Bastian and Yuthas, '84). Injections into the vicinity of the PPn, however, did not retrogradely label somata within the nE. The large numbers of fibers passing through the n E present problems for interpreting these results. The present work further elucidates this connection and identifies additional pathways and targets for transmission of information from the nE.

The complex of the nucleus electrosensorius The n E lies ventral to the TeO a t the lateral extreme of the diencephalon (Carr et al., '81; Fig. 1).Rostrally, the n E lies dorsal to the rostra1 aspect of the preglomerular nucleus (PG). More caudally, the preglomerular nucleus is displaced medially and the nE lies dorsal to the nucleus diffusus lateralis (DFI) of the inferior lobe. A detailed description of

1 0 0 u m CAUDAL Figure 14

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS intranuclear connectivity and cell morphology is not yet possible; a more general description follows. Three functional areas within the nucleus can be identified as described above. The nEb consists of a distinctive group of 400 to 500 somata of 6 to 8 pm in diameter, tightly packed rostrally and dispersing somewhat caudally. The rostral end of nEb marks the rostral end of the nucleus as defined by TSd injections (see below). The n E / begins caudal to the nEb. This area is defined by loci where L-glutamate stimulation causes EOD-frequency falls and bilateral ablation eliminates JAR-related EODfrequency falls (Keller and Heiligenberg, ’89). These loci closely coincide with a dense band of 8 to 1 2 gm wide somata. Whether these cells are themselves responsible for EOD-frequency falls is not yet known. This band extends medio-laterally for 200 to 250 pm and rostro-caudally circa 250 pm. L-glut,amate stimulation of the most medial edge of the band does not cause EOD-frequency falls. Approximately 150 pm above the caudal end of nE1 lies another medio-lateral band of somata, the nE1. The lateral portion of this band closely coincides with loci whose stimulation with L-glutamate causes EOD-frequency rises and whose bilateral ablation eliminates JAR-related EOD-frequency rises. The nEt is much smaller in rostro-caudal extent (100 to 150 pm) than is the nEl. Somata within the nE1 are 10 to 15 Fm in diameter and extend dendrites laterally or ventrolaterally. Medially this band enlarges and disperses into the nEar (see below). Medially the nE1 and nE, give way to a fourth area, the acousticolateral region of t,he nE (nEar), which is defined by its innervation from the ventral torus (TSv, see below). Cells within the nEar are only loosely aggregated. The medial border of this area and of the nE as a whole is only loosely defined by the ext,ent of TSVand TSd terminations. The lateral aspect of the nF,, wherein lie the n E f , nEl, and nEb appears reticulate due to the dense plexi of toral and tectal efferent fibers coursing through the area. Agerent input to nucleus electrosensorius. The pattern of innervation of the n E by efferent fibers of the dorsal torus, originally described by Carr e t al. (’81), is mapped in detail in Figures 2-11 (parts D). These figures represent the projection pattern seen in one brain after two large P H A - L injections which together filled most of the TSd unilaterally without spilling outside of the TSd. Additional cases with similar or smaller injections were used to corroborate these patterns. In agreement with the original description, efferent fibers leave the dorsal torus along the ventrolateral aspect passing through or above the lateral mesencephalic reticular area (LMRa) travelling between the LMRa and the base of the tectum. Most toral efferent fibers are at this point caudal to the nE; within the vicinity of the LMRa, fibers may branch sending one ramus ventrorostrally towards the nE. Travelling rostrally, the first effer-

Fig. 14. Chartings of terminals (dots) and fibers (squiggly lines) seen within the diencephalon of Eigenmannia in silver-impregnated degeneration material after partial lesions of the optic tectum. Densest terminals are fnund in a band along t h e dorsal aspect of the pretectum, within and near to the periventricular pretectal nucleus (PTPv) and the nucleus of the posterior commissure (nPC), within the periventricular nucleus of the post,erior tuberculum (TPP), and within the magnocellular tegmental nucleus (MgT). Very few terminals are seen within the nE. Scale bars 200 pm. ~

363

ent terminals within the n E form a dense proliferation within and about the nE7. The caudal-most beginnings of this terminal field are seen in Figure 3, continuing into the nE: delimited in Figures 4 and 5. The density of terminals diminishes as fibers sweep ventro-rostrally. A second area of proliferation lies in and around the n E l (Figs. 4-6). Farther rostral, a third area of proliferation envelopes the tightly packed cluster of cells designated “nEb” (Figs. 7-10). Same fibers continue medially from the level of Figures 8-10, extend into the ventrolateral thalamic nucleus (VLTh) and its vicinity, and apparently make terminals en route (not drawn). Hostrally, at the level of Figures 9-11, the remaining fibers decussate via the postoptic commissure to end within the contralateral nE. No topographic ordering of inputs to the nE from the TSd was seen in the current material. The ventral torus semicircularis (TSv) receives lateral line and auditory input from the mechanoreceptive portion of the octavolateral system (nucleus medialis, Maler et al., ’74) and is not part of the electroreceptive system (Matsuhara et al., ’81). The TSv projects to the n E as one of several efferent targets. Part C of Figures 2-10 demonstrates the pattern of TSVinnervation of the nE. The figures represent a moderately-sized (ca. 150 pm diameter) injection of PHAL into TSVwith slight labelling of the most ventral TSd. Areas of possible overlap with TSd projections, however, are only drawn if they have also been seen in much smaller injections restricted to the TSV.The TSVlies directly ventral to the most ventral laminae of the TSd, which send strong projections to n E (Fig. 1). Additionally, some TSd efferent fibers course through the TSV before joining the main TSd-efferent plexus, and some TSVcells extend processes into the TSd. It is essential, therefore, to avoid the “fiber-of-passage” problem when tracing TSV enlerent fibers to areas that also receive TSd input. Very small injections of YHA-L into electrosensory-motor areas of n E retrogradely label somata within TSV,further corroborating the overlap in efferent projections from TSVand TSd. Efferent fibers exit the TSValong the ventrolateral aspect adjoining medialiy with the TSd efferent plexus. A dense terminal field within the interstitial reticular area (IRA) is depicted in Figure 2 at the caudal-most level, where TSd axonal terminals occur within the nE. Progressing rostrally (Figs. 3-11), TSVefferent fibers enter the nE along its dorsal and medial aspects and course ventrally. The medial nE is generally filled with a moderate density of efferent terminals. This area, which receives TSv but not TSd efference, is designated the “nE acousticolateral region” (nEar). TSV innervation of the more lateral region, which also receives electrosensory input from the TSd, is characterized by tufty patches of local axonal arborizations which coarsely overlap much of the TSd projection. Although the medial portions of all three major TSd-terminal proliferations also receive TSVefferent fibers, these TSVterminals are not as tightly or densely clustered as are those of TSd-efferent fibers. TSvefferent fibers continuing through the n E decussate with those of the TSd via the postoptic commissure to terminate within the contralateral nE. Retogradely labelled somata appeared in the more rostral levels of n E (Figs. 6-9), generally near to, but not within the n E l , and often within the nEar. A complete categorization of backlabelled cell types within the n E is not yet possible, but several representatives are shown in Figure 12A. A number of cell types within the TSVwere backlabelled after HRP injection to the nE; several examples are shown in Figure 12B.

P91

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFPERENTS

MLF

Fig. 15. Nucleus electrosensorius efferents. A. Camera lucida drawing of all labelled fibers in six consecutive 50 pm sections after PHA-L injection into the nE (injection site is circled). B: Charting of the same case as in A, demonstrating fibers projecting to the inferior lobe as well as some projecting from the TSV.

The original description of Eigenmannia’s TSd efferents to the diencephalon (Carr et al., ’81) showed a band of fibers and terminals across the rostral-dorsal aspect of the diencephalon. In the present study, however, a small number of tightly packed cells within the pretectal periventricular nucleus (PTPv) were backlabelled, whose dendrites and/or axons extend across this same area (Fig. HA). These P T P v cells also send fibers densely into the magnocellular tegmental nucleus (MgT). The projection suggested by Carr et al. (’81) cannot be completely ruled out; however, it seems likely that improved techniques have allowed more complete visualization of neurons only incompletely labelled in the earlier report. Two additional sources of TSd input have also been identified: the cenlral nucleus of the inferior lobe (CE, Fig. 13B) and the nucleus subelectrosensorius (SE, Fig. 16). On the basis of intracellular labelling o f projection neurons, the optic tectum (TeO) has also been suggested as a source of inputs to the nE (Heiligenberg and Rose, ’87). We reexamined this putative input anatomically and behaviorally. Some TSd cell-types projecting to the nE also send axonal collaterals to the TeO and the nucleus praeeminentialis (Carr et al., ’81).Injections of WGA-HRP into the TeO can, therefore, label not only tectal efferents but also the collaterals of TSd cells which project to the tectum. In order to demonstrate only tectal efferents, we therefore mapped

365

silver-impregnated degenerating terminals after lesions confined to the TeO. Survival times after lesion of 4 and 5 days provided clear punctate argyrophilic debris (herein interpreted as degenerating terminals) with minimal degeneration o f fibers, thus allowing mapping of terminals within and around the nE. Figure 14 illustrates the results of such lesions. Degenerating tectal efferent fibers were been entering the diencephalon mediorostrally within the dorsal medial optic tract and more caudolaterally within a dense plexus of fibers between the torus semicircularis and the tectum extending ventral to the tectum. The latter fiber tract extends through the nE to become the tecto-bulbar tract more caudally. This tract may sparsely innervate the n E in transit. There was, however, no dense accumulation of terminals. The densest terminal fields within the diencephalon were found in four areas: 1) a lateral-to-medial band extending from ventral of the tectum through various cell groups medially, which is largely coincident with a band of somata retrogradely labelled by cerebellar injections (see below); 2) medially, within and near to the nucleus of the posterior commissure (nPC) and the pretectal periventricular nucleus (PTPv); 3) within the periventricular nucleus of the posterior tuberculum (TPP); and 4) within the magnocellular tegmental nucleus (MgT, identified as “N?” in Heiligenberg and Rose, ’87, Fig. 1). Several possible additional sources of afferent input to the nE could be identified by injection of HRP into the nE. These injections, however, labelled many fibers which transit the nE but which did not directly connect with it. Cells backlabelled within the nucleus subelectrosensorius (SE) and the central posterior thalamic nucleus (CP), however, could be reconstructed to demonstrate connections with the nE itself. The SE connection appears to be via dendrites belonging to somata within the SE. Cells within the central posterior thalamic nucleus send a dense hand of fibers sweeping into the ventral portion of the nE; whether these are axons or dendrites is not yet known. Both nuclei are dis cussed in more detail below. Cells within the nE also appeared to extend dendrites into at least two other nearby nuclei: the lateral subdivision of the pretectal nucleus (PT1, Fig. 11) and the prethalamic nucleus (PTh, rostra1 to Fig. 11; shown in Sas and Maler, ’86, their Figs. 2 and 7C). These nuclei. and possibly other diencephalic nuclei, thus may serve the n E as sources of afferent information. It has not yet been possible to confine injections to either the lateral pretectal or the prethalamic nucleus to demonstrate the source of these fibers conclusively. m e r e n t pathways from the nucleus electrosensorius. Figure 15A illustrates fibers labelled by an injection of PHA-L, 100 hm in diameter, to the n E to demonstrate its efferent projections within the diencephalon. All fibers labelled in six consecutive 50 fim sections have been drawn, with the exception of those within the nE itself, where the high density of labelled fibers allowed only the fibers of largeqt diameter to be mapped. The major efferent pathways extend niedially into the anterior thalamic nucleus (ATh), the prepacemaker nucleus (PPn). and the central posterior thalamic nucleus (CP), and medio-ventrally into the anterior tuberal nucleus (TA). Additional efferents project to the inferior lobe (Fig. lSB), the cerebellum (see below and Fig. 17) and the ventral torus (Fig. 12). These projections will be discussed in more detail below. The preglomerular nucleus (PG) also contained small numbers of labelled fibers after nE injection. Most of these fibers continued on

366

C.H. KELLER ET AL.

A

I 0 IH

B

C

Fig. 16. A: HRP injections to the nE label many cell., within the nucleus subelectrosensorius (SE). Scale bar = 200 pm. B,C: Individual cells within the SE, reconstructed after HRP injection to the CP. Simi-

lar cells are backlabelled after injections to nE, TA, DFL, or TSd. Scale bars = 50 fim.

NUCLEIJS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

367

Fig. 17. A band of cells extending across the roof of the pretectum projects to the cerebellum. This band covers the dorsal third of the nE, including the nET. Chartings of somata retrogradely labelled by large

injections of WGA-HRP into the cerebellum. Areas caudal to the dienrephalon are not presented. Scale bar = 200 rm.

to the inferior lobe; injections confined within the preglomerular nucleus failed to label cells within the nE. We also investigated a reported projection from the nE to the forebrain (Bastian and Yuthas, ’84). Injections of PHAL, WGA-HKP, HRP, or Di-I (Godement e t al., ’87) to the forebrain, however, failed to backlabel somata within the nE. High densities of fibers passing through the n E on their way from the forebrain to the TeO may have suggested a projection from the nE to the forebrain. There was some indication of forebrain-efferent terminals within this dense fiber tract as it transits the nE. The high density of passing fibers, however, precluded definitive mapping of this putative terminal field with the current techniques. A large number of cell types within the diencephalon extend dendrites into the nE, and it is not yet possible or useful to identify all of these. Two such cell types, one within the nucleus subelectrosensorius and another within the P P n , will be discussed. The nucleus subelectrosensorius comprises a number of different cell types extending caudally from a position between the P P n and the caudal nE,

and lying dorsal to the glomerular nucleus (Figs. 4-6). Nucleus subelectrosensorius cells innervate the ventral aspect of the nE (Fig. 16) with a dense band of fibers which overlaps with fibers from the central posterior thalamic nucleus, discussed below. In addition to its connection with the nE, the nucleus subelectrosensorius makes connections with TSd, TeO, anterior tuberal nucleus, inferior lobe, and thalamic areas medial to the nucleus subelectrosensorius. Lglutamate stimulation of the nucleus subelectrosensorius had no effect on the frequency of EODs, and the nucleus’s functional role is unknown. Cells within the PPn-C also deserve mention in the present context. The anatomy and physiology of these cells have heen studied extensively (Kawasaki and Heiligenberg, ’88; Kawasaki et al., ’88; Zupanc and Heiligenberg, ’89). PPn-C cells lie within the ventrocaudal subnucleus of the P P n and are responsible for abrupt EOD-frequency increases, or “chirps,” seen during aggression and courtship. Some of these cells, apparently a small percentage. send long dendrites into the ventral aspect of the nE near to the nEl. One such cell can be seen within the

368

C.H. KELLER ET AL.

I

TeO

TSV

TeO i

CP

& &

F&.

,

4 nEb

MLF-

PGI PGm

sc

* a

--

Figure 18

369

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS P P n in Figure 15 (confirmed as a PPn-C cell by tracing of its axon to endings on the relay cells within the pacemaker nucleus). The more usual dendritic fields of PPn-C cells broadly overlap with the projections of n E cells to the P P n and central posterior thalamic nucleus (see below). Injection of neuronal tracers into each of the efferent targets identified above allowed preliminary identification of cell types and subnuclei within the n E responsible for each projection. These will be considered in turn.

Cerebellum A number of cell groups lying along the dorsal aspect of the rostral diencephalon project to various subdivisions of the cerebellum (Fig. 17; Sas and Maler, ’87; Striedter, ’87). This roof-like band of cells includes somata within several other pretectal nuclei besides the n E and is largely coincident with a strong tectal terminal field as described above. Figure 17 demonstrates the extent of this band of cells as seen in large WGA-HRP cerebellar injections: beginning rostrally within the nucleus at the base of the optic tract (nB, anterior to most rostral section in Fig. 17), extending caudally and laterally to include pretectal nucleus “A” and “B,” the pretectal nucleus (PT), and the dorsal tegmental nucleus (DTn), and continuing more caudally and laterally into the dorsal third of the nE, including, but not limited to, the nEf. The caudal-most extent of these cells lies caudal to the nE as a small cell group within the toral and tectal efferent plexus (seen in Fig. 2). A number of different cellular morphologies are represented across this band of cells, and each cell type extends dendrites laterally or ventrolaterally within the area included by the band.

Inferior lobe Injections of HRP into the nucleus diffusus lateralis and the central nucleus of the inferior lobe labelled cells in a rather continuous strip from the mid-thalamus caudally to the habenular level rostrally (Fig. 18). A number of cell types are represented. Rostrally, a t the level of Figures 10 and 1 1 and further rostral, these include a cell type clustered within the lateral pretectal nucleus (PT1) and the prethalamic nucleus; this cell type is densely multipolar with highly varicose local dendrites and soma diameters of circa 10 pm (Fig. 18D). The other retrogradely labelled cells have more diverse morphologies and disperse within the strip; they can probably be subdivided into several types. The rostral-most of these lie within the nEar and are separated from the PT1 by a small but consistently recognizable fiber

Fig. 18. HRP injections into the inferior lobe retrogradely label cells in the nEar, medial n E and the PT (lateral subdivision). A: Chartings progressing rostrally from the injection site in the inferior lobe. Symbols as in Figure 2. Scale hars = 200 Fm. R: Camera-lucida drawing of a retrogradely labelled cell located within the nEar a t the level of nE4 (second panel from left in A, a t the level of Fig. 6) which sends dendrites into the electrosensory region of nE. Cells drawn in C and E are representatives of another cell type located within the nEar more rostrally a t the level of nEh (third panel from left in A, at the level of Fig. 9). These cells send long dendrites t o the immediate edge of nEb. Locations of similar somata are indicated by asterisks, loci of another cell type (that of D, within the PTl) are indicated by stars; small hut recognizable fiber tract is circled by dots. D. A third cell type, located within the lateral subdivision of the pretectal nucleus, is shown (right-most panel in A, at the level of Figs. 10, 11). This cell type extends large numbers of local dendrites with high densities of varicosities. Cell type in F lies just dorsal to nEb. Scale hars for B-F = 50 Wm.

tract. One of these cell types is hi- to multipolar, has an oblong soma 15-20 p m in length, and extends primary dendrites laterally to the edge of nEb (Fig. 18C,E) at the level of Figures 9, 10. Branching from these dendrites are many local, highly varicose dendrites which fill the nE. Cells of a second cell type with soma diameters of 10-12 pm lie slightly medial and dorsal to nEb (Fig. 18F, also at the level of Figs. 9, 10). A small number of cells within the nEb are also labelled. More caudally, labelled cells within this strip are scattered and of a broadly branching unipolar type. Some of these cells lie within the nEar and extend dendrites near to the n E l (Fig. 18I3, level of Figs. 5,6), while others lie caudal to the nE. Additionally, retrogradely labelled somata were seen within the cerebellum and the perilemniscal (subvalvular) area.

Anterior tuberal nucleus Injections confined within the anterior tuberal nucleus (‘L’A) labelled two cell types within the n E a t the level of nEb (Pig. 19). Many cells of the first type were labelled. This cell type lies within the nEb; some or all of these cells send axonal collaterals to the central posterior nucleus (CP), the prepacemaker nucleus (PPn) and, possibly, to the anterior thalamic nucleus (ATh). Their somata are circa 6-8 pm in diameter and send two or more primary dendrites ramifying within the nEb (Fig. 19B,D). Few of these dendrites could be seen to extend beyond the tightly packed cell cluster that comprises the nEb. A second labelled cell type comprises bipolar cells with soma diameters of 10-12 pm, located medially to the nEb. These cells send one primary dendrite into the nEb (Fig. 19C).

Anterior thalamic nucleus Figure 20 illustrates a small HRP injection confined within the laterally opening horseshoe-shaped band of cells that demarcates the anterior thalamic nucleus (ATh). Two separate areas and cell types within the nE were retrogradely labelled by this injection: I) somata within the cluster identified as nEb that are indistinguishable from cells labelled by injections to the anterior tuberal nucleus (Fig. 19B,D),and 2) a band of cells extending into the dorsal third of the nE and coincident with somata backlabelled by cerebellar injections. Three other areas sending projections to the cerebellum also contained retrogradely labelled somata: the nucleus at the base of the optic tract (nB), the dorsal tegmental nucleus (DTn), and the area pretectalis (APT). Although axonal collaterals could not be traced to the cerebellum, the locations and morphologies of these cells backlabelled by ATh injection suggest their identity with cells backlabelled from the cerebellum. A cluster of cells lying laterally within the nucleus subelectrosensorius was also labelled. These cells project strongly to the anterior tuberai nucleus as well as to the anterior thalamic nucleus and extend fibers into the ventral nE, near to the nEl. Additionally, fibers and somata were labelled in a lateroventrally extending band marking the continuation of the anterior thalamic nucleus and merging caudally with the lateral extent of the central posterior thalamic nucleus. A strong axonal projection sweeps through this band and joins with the projection of the subelectrosensorius cell cluster to terminate massively within the anterior tuberal nucleus. A small number of somata were also labelled within the anterior tu beral nucleus and caudally within several cell clusters of the reticular formation. A similar but very weak pattern

C.H. KELLER ET AL.

370 . .

.. ,

. , , . . . .. .

2 Fig. 19. HRP injections into the anterior tuheral nucleus (TA) retrogradely label somata within the nEb and a second cell type medial to nEb, within nEar. Cells labelled within nEb send axon collaterals to the medial thalamic nuclei (CP, PPn, possibly ATh). A Chartings of a 180 pm diameter H R P injection to TA. Scale bar = 200 pm. Symbols as in

Figure 2. B, and D: Cells backlabelled in nEh. C Cell backlabelled within the medial nE. Stars indicate positions of labelled somata with similar morphology; asterisks denote loci of labelled cells similar to those in B and D. Scale bars for B-D = 50 pm.

Fig. 20. A HRP injections confined within the anterior thalamic nucleus (ATh) retrogradely labelled cells within the nEb (indistinguishable from those labelled after TA or CP injection; see Fig. 19B,D), as well a s a band of cells lying dorsally within various pretectal nuclei and the nE, and a cell cluster within the nucleus subelectrosensorius (SE). A

strong projection extends to t h e anterior tuberal nucleus (TA). Scale bar = 200 pm. B Cells lying within the dorsal band. This band is congruent with cells backlabelled from the cerebellum, which also share a similar cell morphology. Scale bar = 50 wm.

372

C.H. KELLER ET AL.

,

Figure 21

373

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS of projections was seen contralaterally.

Central posterior thalamic nucleus and the prepacemaker nucleus The central posterior nucleus of the thalamus (CP) curves ventrolaterally from the ventricle. In gymnotiform fish, the prepacemaker nucleus lies a t the ventrolateral boundary of the central posterior thalamic nucleus (Heiligenberg et al., '81, '88; Kawasaki et al., '88; Keller et al., '88; Kawasaki and Heiligenberg, '89). Medially within the CP, it is possible to inject small volumes of neuronal tracers without spillage to neighboring nuclei or danger of filling many passing fibers. At more lateral sites these problems intensify such that tracing of connections to these areas can be problematic. Similarly, although rostral P P n injections fill primarily the PPn-C and caudal injections fill primarily the PPn-C, it has not yet been possible to confine injections to only one subnucleus of the PPn. Figure 21A illustrates an HRP injection of 80 pm in diameter which was well confined to the medial portion of the central posterior thalamic nucleus at the rostral-most level of the PPn. Such injections retrogradely labelled somata within the nEb. Fibers course laterally through the central posterior thalamic nucleus and rostrolaterally around the anterior thalamic nucleus towards the nE. Some of these fibers are the axons of nEb somata, and others are part of an extensive fiber field which extends along the ventromedial aspect of the nEar to the lateral edge of the brain just below the nEb and nEl. Additionally, a large number of periventricular cells within the CP were usually labelled as were a few somata within a dense terminal field medial to the ventral lateral (VLTh), dorsal lateral (DLTh), and dorsal medial (DMTh) thalamic nuclei (at the level of Fig. 10). Another projection extends through the central posterior thalamic nucleus and P P n ventrally to end in a rich field between the glomerular nucleus ( G ) and the anterior tuberal nucleus (TA), a few fibers extending into TA. Figure 21B presents a more lateral (and larger, 150 pm) injection centered within the P P n but extending into the central posterior thalamic nucleus and surrounding areas. Injections of this sort retrogradely labelled many somata within the nEb and other somata in a band extending across the dorsal third of the nE including nEt (Fig. 21E). A few cells were labelled caudal to the nEb continuing to the nE1. Some similarly placed injections labelled a distinct group of cells within the medial nEt as shown in Figure 21C. Generally, as injections were made more laterally within the central posterior thalamic nucleus and the PPn, increasing numbers of labelled somata were found within the nEb, and additional areas of the nE were labelled. Very small injections confined to the rostral P P n labelled only a few somata within the nEb. Labelled cells located in the dorsal band (including n E f , Fig. 21B,E) resemble those backlabelled by injections to the anterior thalamic nucleus and the cerebellum. They have soma diameters of circa 7-12 pm and are hi- to multipolar,

Fig. 21. A Small injections of HRP into the medial portions of the central posterior thalamic nucleus (CP) label somata within the nEb. B: HRP injections located more laterally within the CP and the PPn label larger numbers of cells within the nEt (shown in parts C and E: asterisks denote filled cells of similar morphology), nEb (shown in part D; arrowheads point to axons), nEar, and elsewhere within the nE. Scale bars in A and H = 200 pm; in C-E = 50 pm.

JAR

BEFORE

\

AFTER I

TeO

A, LESION

LESION ,/

2 H z L 10 s

J Fig. 22. Removal of ca. 90% of the optic tectum (TeO) does not affect performance of the JAR. A Camera lucida drawing of the hrain of Eigenmannia after TeO lesion. This 50 pm section contained the largest remaining portion of the TeO (at the level of n E f , Fig. 5 ) . Scale bar = 500 pm. B: The JAR is indistinguishable before (top trace) and after (middle trace) the lesion shown in A. The JAR is shown as a shift in instantaneous EOD-frequency in response to jamming stimuli (bottom trace) placed alternately 4 Hz above (Df 0) or below (Df < 0) the frequency of the mimic of the fish's EOD.

extending dendrites along this same band, primarily laterally. Figure 21C presents an nEb cell backlabelled after a very small injection (50 p m diameter) confined within the rostral PPn. This cell is very similar to those backlabelled by larger injections into the central posterior nucleus or P P n or by injections confined to either the anterior thalamic or the anterior tuberal nucleus.

Behavioral effects of optic tectum lesions In light of the minimal projection from the optic tectum (TeO) to the nE, we re-examined the behavioral role of the tectum in the JAR. Previously, large tectal lesions had been

C.H. KELLER ET AL.

374

Fig. 23. Schematic drawing of major afferent (left side) and efferent (right) pathways of t h e nF:. Line thickness indicates relative strength of pathway. Dotted lines represent dendritic connections.

found to abolish the ,JAR (Rose and Heiligenberg, '86). Sections from these brains were re-examined and found to contain extensive damage to TSd efferents and parts of the nE, pathways not well-defined a t the time of the earlier study. We aspirated most of the optic tectum in one Eigenmannia and one fish of the genus Apteronotus (a close relative of Ezgenmannia which is more robust and better able to survive the extensive lesions required). In each fish, the JAR remained completely intact despite removal of more than 90' of the optic tectum (Fig. 22).

DISCUSSION The present work comprises a fine-grained description of afferent and efferent innervation of the diencephalic nucleus electrosensorius, a nuclear complex which functions as a set of finely tuned neuronal filters for specific sensorymotor tasks. The simplest criteria for the definition of the nE is that of Carr et al. ('81): the lateral dorsal aspect of the diencephalon which receives input from the torus semicircularis. Caudally, the nE is continuous with the interstitial reticular area (IRA), a region which receives octavolateral

input strictly from the TSV. Rostrally and medially, nE is delimited by pretectal nuclei that receive predominantly tectal input or direct retinal input (Sas and Maler, '86), but no toral input. I t is important to note that the nE itself receives little or no tectal input (see below). In contrast to earlier suggestions by Carr et al. ('81), we find no evidence for somatotopy in the toral projection to the nE. The clear somatotopic organization of the dorsal torus semicircularis (TSd) provides a substrate for comparisons of amplitude and phase information between different parts of the body surface and is reflected in the strong dependence on stimulus field orientation of beat-sensitive units within the TSd. This somatotopy is not preserved within the TSd projection to nE; rather, the pattern of TSd terminations appears to be related to the function of the target area within the nE. As a likely consequence of spatial integration and the ensuing loss of somatotopic order, many n E units respond to beat stimuli independently of their orientation. The fact that the entire electric organ is controlled as one obviates the need for topographic organization beyond a separation of the controls for EOD-frequency rises and falls. More complex systems, involving antagonistic or synergistic control of several motor elements may

375

NUCLEUS ELECTROSENSORIUS AFFERENTS AND EFFERENTS

require a more detailed topographic organization of premotor areas. A number of cell types located within laminae VII, VIII, IX of the TSd are backlabelled by nE injection (Carr et al., '81). I t is likely, but not yet certain, that terminals of these cell types are represented differentially within the subareas of the nE. The nucleus electrosensorius itself can be subdivided on the basis of local cell clustering, as well as input and output pathways (Fig. 23), and the validity of this parcellation is supported by functional distinctions between structurally defined subdivisions. The four subdivisions recognized by present criteria are 1) nEb-a rostra1 cluster of tightly packed cells which receives TSd input and projects to the inferior lohe, anterior tuberal nucleus, anterior thalamic nucleus, central posterior thalamic nucleus and the prepacemaker nucleus. This cell group contains neurons which respond to heat patterns caused by jamming stimuli, but its stimulation by iontophoresis of L-glutamate does not induce any EOD-frequency shift in Eigenmannia. 2) n E f - - a horizontal band of' cells found a t the dorsal aspect of the caudal nE, which receives TSd input and projects to the prepacemaker nucleus and its vicinity as well as to the corpus cerebelli. Stimulation of this area induces the slow EOD-frequency rises characteristic of the JAR, and bilateral lesion of this area eliminates JAR-related EOD-frequency rises. The responses of these cells to jamming stimuli are not yet known. 3) nE1-a horizontal band of cells found at the ventral aspect of the caudal nE, which receives TSd input and projects only to the virinity of the PPn. Stimulation of this area induces the slow EOD-frequency falls characteristic of the JAR, and bilateral lesion of the nEJ eliminates JARrelated EOD-frequency falls. The responses of these cells to jamming stimuli are not yet known. 4) nEar-a complex medial region of the nE which receives predominantly TSV input and projects to the inferior lobe, anterior tuberal nucleus, central posterior thalamic nucleus, prepacemaker nucleus and the corpus cerebelli. The sensory and motor properties of this region are not known in detail although auditory and mechanosensory responses have been recorded here. Functional properties allowing further subdivision of the nE are still under investigation as are the interconnections between the subareas defined so far. Our preliminary data and those of Bastian and Yuthas ('84), however, suggest extensive connections within the nE. The patterns of efferent projections suggest that the nEb influences the electromotor system directly, via its projection to the PPn, while its output to other thalamic (anterior and central posterior thalamic nuclei) and hypothalamic nuclei (inferior lobe and anterior tuberal nucleus) may provide specific electrosensory information to areas imparting motivational and hormonal influences on behavior. The nEt and nE4 have more limited projections, confined primarily to the P P n and its vicinity and may thus be dedicated primarily to the electromotvr system. The projection of nEt to the cerebellum would be less surprising if the nEl shared a similar projection. The corpus cerebelli of gymnotiforms receives additional massive electrosensory input via a pathway from the tectum to the nucleus lateralis valvulae and, hence, to the cerebellum (unpub. obs.). Recordings in the cerehellum will be required to determine the role of these electrosensory inputs. Recently, nEar neurons responsive to chirp-like EOD-modulations have been labelled intracellularly and shown to send extensive dendritic arborizations into the electrosensory portions of the nE (Keller et

al., '89). Thus nEar neurons may supply efferent targets with integrated electrosensory and acousticolateral information pertaining to socially relevant stimuli. Carr et al. ('81) pointed out that the TSd appears to contain many parallel processing streams which eventually converge upon three targets: the nucleus praeeminentialis (nP), TeO, and nE. This study further emphasizes the relative independence of these three outputs. The nucleus praeeminentialis is strongly dedicated to feedback within the electrosensory system (Sas and Maler, '87; Bastian and Bratton '88). The TeO does not appear to influence the electromotor system directly since no direct projections to the nE or P P n could be identified. The TeO obviously is involved in the somatic motor system; this is apparent both from the projection of the tectum to nuclei such as the magnocellular tegmental nucleus. which, in turn, projects to the spinal cord, as well as from the somato-motor effects of tectal stimulation (Yuthas, '85). Furthermore, the lesion studies reported here demonstrate that the tectum is not necessary for the JAR; earlier contrary findings (Rose and Heiligenberg, '86) apparently resulted from unnoticed damage to the toral efferents or to the nE. The electrosensory portions of the nE are involved in electrocommunication such as the JAR and chirping, and perhaps in neuroendocrine and motivational controls through electrosensory input to the hypothalamus and thalamus. The lack of' a topographic representation in the nE and the absence of nE projections to any obvious somatic motor regions suggest that this nucleus is not engaged in electrolocation per se. Finely tuned units such as those found within the nE might, however, be useful for object identification or similar tasks requiring convergence of inputs from many areas of the body surface. Recordings from the n E of Eigenmannia's close relative Sternopygus, which lacks a JAR, might be instructive in this regard. The electrosensory portion of the nE cannot he present in non-electrosensory teleosts, which raises the question of its evolutionary origin. On the basis of the direct continuity of the electrosensory portions of nE with the octavolateral n E as well as the continuity of the nEar with the pretectal nuclei, we hypothesize that the nE has evolved from the pretectum. This view is supported by the location of cells in this region which project to the cerebellum; these cells cover a continuous band stretching from pretectal nuclei (nuclei A and B, DTn, and PT) into the nEar and then into the n E f . The role of this complex multisensory pretectal projection to the cerebellum is still unknown.

ACKNOWLEJXMENTS We thank Grace Kennedy and Georgia Malan for expert technical assistance, and T.H. Rullock, C.E. Carr, J.T. Enright, T. Finger, and G. Striedter for helpful criticism of the manuscript. Special thanks to T. Finger for assistance with the silver-staining technique. This work was supported by NSF grant 8716781, NIMH arant 2 ROMH26149-13. NINCDS grant 5 R 0 1 NS22244-63 to W.H. and MRC grant MT6027 to L.M.

LITERATURE CITED Bastian, J. (1986) Electroreception: behavior, anatomy, and physiology. In T.H. Bullock and W. Heiligenberg (eds): Electroreception. New Y o r k Wiley, pp. 577412. Bastian, J., and B.O. Bratton (1988)The nucleus Praeeminentialis: proper-

C.H. KELLER ET AI,. ties of neurons providing descending inputs to the electrosensory lateral Neurosci. Abstr. 14:311. line lobe of weakly electric fish. SOC. Bastian, J., and J. Yuthas (1984) T h e jamming avoidance response of Eigenmannia: Properties of a diencephalic link between sensory processing and motor output. J. Comp. Physiol. 154:895-908. Braford, M.R. Jr., and R.G. Northcutt (1983) Organization of the diencephalon and pretectum of the ray-finned fishes. In R.E. Davis and R.G. Northcutt (eds): Fish Neurobiology. Ann Arbor: University of Michigan Press, Val. 2, pp. 117-163. Carr, C.E., and I.. Maler (1986) Electroreception in Gymnotiform fish Central anatomy and physiology. In T.H. Bullock and W. Heiligenberg (eds): Electroreception. New York: Wiley, pp. 319-374. Carr, C.E., L. Maler, W. Heiligenherg, and E. Sas (1981) Laminar organization of the afferent and efferent systems of the torus semicircularis of Gymnotiform fish: Morphological substrates for parallel processing in the electrosensory system. J. Comp. Neurol. 203:649-670. Gerfen, C.R., and Sawchenko, P.E. (1984) An anterograde neuroanatomical tracing method that shows the detailed morphology of neurons, their axons and terminals: Immunohistochemical localization of an axonally transported plant lectin, Phaseolus uulgark leucoagglutinin (PHA-L). Rrain Res. 290:219-238. Godement, P., J. Vanselow, S. Thanos, and F. Bonhoeffer (1987) A study in developing visual systems with a new method of staining neurones and their processes in fixed tissue. Development 101:697-713. Heiligenberg, W. (1986) Jamming avoidance responses: model systems for neuroethology. In: T.H. Bullock and W. Heiligenherg (eds): Electroreception. New York Wiley. pp. 613-650. Heiligenberg, W., and J. Bastian (1980) The control of Eigenmannia’s pacemaker by distributed evaluation of electroreceptive afferences. J. Comp. Physiol. A 136:113-133. Heiligenberg, W., M. Kawasaki, and J. Nissanov (1988) Suppression of the electric organ discharge by desynchronization of pacemaker neurons in the electric fish Hypopornus. Soc. Neurosci. Abstr. 14:204. Heiligenberg, W., T. Finger, J. Matsubara, and C. Carr (1981) Input to the medullary pacemaker nucleus in the weakly electric fish, Eigenmannia (Sternopygidae, Gymnotiformes). Brain Res. 211:41%423. Heiligenberg, W., and G. Rose (1987) The optic tectum of the gymnotiform electric fish, Eigenmannia: lahelling of physiologically identified cells. Neuroscience 22331-340. Heimer, L.(1970) Selective silver-impregnation of degenerating woplasm. In W.J.H. Nauta and S.O.E. Ehhesson (eds): Contemporary Research Methods in Neuroanatomy. New York Springer Verlag, pp. 1 6 1 3 1 . Kawasaki, M., and W. Heiligenberg (1989) Distinct mechanisms of modulation in a neuronal oscillator generate different social signals in the electric fish Hypopornus. J. Comp. Physiol. A 165:731-741. Kawasaki, M., and W. Heiligenberg (1988) Individual prepacemaker neurons can modulate the pacemaker cycle of the gymnotiform electric fish, Eigenmannia. J. Comp. Physiol. A 16213-21. Kawasaki, M., L. Maler, G.J. Rose, and W. Heiligenberg (1988) Anatomical

and functional organization of the prepacemaker nucleus in Gymnotiform electric fish: The accommodation of two behaviors in one nucleus. J. Comp. Neurol. 276:113-131. Keller, C.H. (1988) Stimulus discrimination in the dienceDhalon of Eirenmannia: The emergence and sharpening of a sensory filter. J. Comp. Physiol. A 162:747-757. Keller, C.H., M. Kawasaki, and W. Heiligenherg (1988) Pacemaker modulations in the electric fish Sternopygus. Soc. Neurosci. Abstr. 14204. Keller, C.H., and W. Heiligenberg (1989) From distributed sensory processing to discrete motor representations in the diencephalon of the electric fish, Eigenmannia, J. Comp. Physiol. 164~565-576. Keller, C.H., M. Kawasaki, and W. Heiligenberg (1989) Intracellular labelling of physiologically defined cells within a diencephalic sensory-motor interface of weakly electric knifefish. Soc. Neurosci. Abstr. 151136. Maler, l,., T. Finger, and H. Karten (1974) Differential projections of ordinary lateral line and electroreceptors in the gymnotid fish, Apteronotus albrifrons. J. Comp. Neurol. 158:363-382. Matsubara, J., T. Finger, L. Maler, and C. Carr (1981) Electrosensory, auditory and lateral line nuclei in the midbrain of a gymnotiform weakly electric fish, Sternopygus. Soc. Neuro. Abstr. 7:84. Mesulam, M.M., and D.L. Rosenc (1979) Sensitivity in horseradish peroxidase neurohistochemistry: A comparative and quantitative study of nine methods. J. Histochem. Cytochem. 27:763-773. Metcalfe, W.K. (1985) Sensory neuron growth cones comigrate with posterior lateral line primordial cells in Zebra fish. J. Comp. Neurol. 238:21%224. Rose, G., and W.Heiligenberg (1986) Neural coding of difference frequencies in the midbrain of the electric fish Eigenmannia: Morphological correlates of phase and amplitude sensitivity. J. Neurosci. 5:2269-2280. Rose, G., M. Kawasaki, and W. Heiligenberg (1987) ‘Recognition units‘at the top of the neuronal hierarchy? Prepacemaker neurons in Eigenrnannia code the sign of frequency differences unambiguously. J. Camp. Physiol. 162759-772. Sas, E., and L. Maler (1986) Retinofugal projections in a weakly electric gymnotid fish (Apteronotus leptorhynchus). Neuroscience 18:247-259. Sas, E., and L. Maler (1987) T h e organization of afferent input to the caudal lobe of the cerebellum of the gymnotid fish Apteronotus leptorhynchus. Anat. Emhryol. 177.55-79. Stone. T.W. (1985) Microiontophoresis and pressure ejection. In A.D. Smith (ed.):IBRO Handbook Series: Methods in the Neurosciences. Val 8. New York: Wiley. Striedter, G.F. (1987) Does nucleus electrosensorius of gymnotoids have a homologue in the dienrephalon of non-electroreceptive teleosts? SOC. Neurosci. Abstr. 13:130. Yuthas, J. (1985) Motor patterns evoked by stimulation of the optic tectum in two species of weakly electric fish. Soc. Neurosci. Abstr. 11:1021. Zupanc, G.K.H., and W. Heiligenberg (1989) Sexual maturity-dependent changes in neuronal morphology in the prepacemaker nucleus of adult weakly electric knifefish, Eigenrnannia. J. Neurosci. 9:3816-3827.