Proc. Natl. Acad. Sci. USAVol. 81, pp. 940-944, February 1984Neurobiology

The goldfish nervus terminalis: A luteinizing hormone-releasinghormone and molluscan cardioexcitatory peptide immunoreactiveolfactoretinal pathway

(retina/olfactory system)

WILLIAM K. STELL, STEVEN E. WALKER, KULDIP S. CHOHAN, AND ALEXANDER K. BALLDepartment of Anatomy and Lions' Sight Centre, The University of Calgary Faculty of Medicine, 3330 Hospital Drive, N.W.,Calgary, AB, Canada T2N 4N1

Communicated by John E. Dowling, August 12, 1983

ABSTRACT Antisera to two putative neurotransmitters,luteinizing hormone-releasing hormone (LHRH) and mollus-can cardioexcitatory tetrapeptide (H-Phe-Met-Arg-Phe-NH2;FMRF-amide), bind specifically to neurites in the inner nucle-ar and inner plexiform layers of the goldfish retina. Retro-grade labeling showed that intraocular axon terminals origi-nate from the nervus terminalis, whose cell bodies are locatedin the olfactory nerves. Double immunocytochemical and ret-rograde labeling showed that some terminalis neurons projectto the retina; others may project only within the brain. Allterminalis neurons having proven retinal projections wereboth LHRH- and FMRF-amide-immunoreactive. The activityof retinal ganglion cells was recorded with microelectrodes inisolated superfused goldfish retinas. In ON- and OFF-centerdouble-color-opponent cells, micromolar FMRF-amide andsalmon brain gonadotropin-releasing factor ([Trp7, Leu8jLHRH) caused increased spontaneous activity in the dark, lossof light-induced inhibition, and increased incidence of light-entrained pulsatile response. The nervus terminalis is there-fore a putatively peptidergic retinopetal projection. Sex-relat-ed olfactory stimuli may act through it, thereby modulatingthe output of ganglion cells responsive to color contrast.

The nervus terminalis (terminal nerve), the anteriormost ofthe paired cranial nerves, comprises ganglia and slendernerve trunks closely associated with the olfactory nerve andbulb (1). Largely forgotten for over half a century, it has sud-denly regained prominence in neuroendocrinology, olfactionand sexual behavior, comparative neuroanatomy, peptidephysiology, and vision (2-8). Particularly significant are re-ports that the so-called nucleus and nervus olfactoretinalis ofseveral bony fishes share with their mammalian counterpart,the terminal nerve, the property of binding antibodies to thepeptide luteinizing hormone-releasing hormone (LHRH) (3-5, 8, 9). Such neuroactive peptides, which occur widelythroughout the nervous system, are of great interest as syn-aptic transmitter candidates, although their modes of actionand roles in neural information processing remain largelyspeculative (10, 11).We report here the results offurther neuroanatomical, im-

munocytochemical, and neurophysiological studies on theterminal nerve of goldfish. Of the bony fishes, the structureand function of the retina and the neuroendocrine regulationof reproduction and sexual behavior are best known in gold-fish. Further studies of the terminal nerve in goldfish there-fore seemed likely to clarify its roles in reproductive behav-ior and the efferent modulation of visual input to the brain. Italso seemed important to reexamine its structure and neuro-chemistry, since it has been observed in goldfish by neuroan-atomical methods (1, 7, 12, 13) but not by LHRH immunocy-

tochemistry (3). Finally, because the retina is a well-knownmodel for the central nervous system and a convenient phys-iological test system, the assembly of terminal nerve and ret-ina seemed likely to yield useful data on the function of pep-tidergic neural pathways. In the course of our studies welearned that the terminal nerve contains a second peptideimmunochemically similar to the molluscan cardioexcitatorytetrapeptide (H-Phe-Met-Arg-Phe-NH2; FMRF-amide) (14-18). We thus had the further experimental opportunity toconsider the functional role for coexistence of two or moresynaptic transmitter candidates in the same neuron (11, 19-21).

MATERIALS AND METHODSGoldfish (Carassius auratus) were obtained from varioussources in all seasons and kept at '20'C under ordinary fluo-rescent lights (light/dark cycle, 10:14 hr).Immupocytochemistry. Fish 3-10 cm long were anesthe-

tized with tricaine methanesulfonate (ethyl m-aminoben-zoate, Sigma). Eye cups, optic nerves, and brains were dis-sected and fixed by immersion for 2 hr at 40C with 4% para-formaldehyde in 0.05 M phosphate buffer (pH 7.4) with 3%sucrose, washed in buffer, infiltrated with 30% sucrose at40C, and either sectioned on a cryostat (15-40 ,um) or flatmounted.

Flat mounts were prepared by isolating retinas from fresheye cups of dark-adapted fish, fixing them as describedabove, removing the vitreous, and drying them flat (receptorside down) on a gelatinized slide. The whole mounts werethen frozen and thawed and treated for immunocyto-chemistry as if they were sections.

All primary antisera were from rabbits. We used antiserato LHRH (U-705 and U-706 from G. Kozlowski, lot 22279from Immuno Nuclear, Stillwater, MN, BOON-7 and BE4Dfrom S. A. Joseph), FMRF-amide [L-135 from G. Dockray(see ref. 22)], and assorted others.

For immunofluorescence, sections were treated with pri-mary antisera diluted 1:100 in phosphate-buffered saline (pH7.4) containing 0.25% Triton X-100, for 16 hr at 40C; fol-lowed by fluorescein isothiocyanate- or tetramethyl rhoda-mine isothiocyanate-conjugated goat anti-rabbit gammaglobulin (Sigma) and fluorescence microscopy (23-25).For immunoperoxidase, sections or whole mounts were

treated with primary antiserum diluted 1:1,000 in phosphate-buffered saline (pH 7.4) with 0.25% Triton X-100, for 14-24hr at room temperature, then with unlabeled goat anti-rabbitgamma globulin, and finally with peroxidase-antiperoxidase(Boehringer Mannheim), visualized using 3,3'-diaminobenzi-dine (26) or Hanker-Yates reagent (27).

Abbreviations: LHRH, mammalian luteinizing hormone-releasinghormone; fish GnRF, salmon brain gonadotropin-releasing factor[Trp', Leu8]LHRH; FMRF-amide, H-Phe-Met-Arg-Phe-NHI2 (mol-luscan cardioexcitatory peptide); YGGFMRF-amide, H-Tyr-Gly-Gly-FMRF-amide.

940

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. NatL Acad. Sci. USA 81 (1984) 941

Controls for specificity used primary antisera that hadbeen preabsorbed overnight with synthetic peptides (Penin-sula Laboratories, San Carlos, CA; final concentration, 100,Ag/ml). Anti-LHRH was also preabsorbed with FMRF-am-ide, and anti-FMRF-amide with LHRH, to rule out unex-pected crossreactivity. Immunocytochemical localizationwith antisera to peptides other than LHRH and FMRF-am-ide served as further controls.Double immunocytochemical labeling used mainly se-

quential staining, first with anti-LHRH followed by tetra-methyl rhodamine isothiocyanate-conjugated goat anti-rab-bit gamma globulin, then anti-FMRF-amide followed by flu-orescein isothiocyanate-conjugated goat anti-rabbit gammaglobulin (or reverse order). This procedure showed whethersome structures were immunoreactive with the second pri-mary antiserum but not the first (28). Colocalization was alsoassessed, less critically, by comparing staining for differentpeptides, and by restaining after elution of the first antigen-antibody complex (29).Retrograde Labeling. We injected 2.5 A.l of either 30%

horseradish peroxidase (Sigma, type VI), 2% nuclear yellow(courtesy of Heinz Loewe, Hoechst, Frankfurt; see ref. 30),or 2% true blue (150/129, from Dr. Illing, KG, Postfach1150, D-6114 Gross-Unstadt, F.R.G.; see ref. 30) into thevitreous body. Usually we injected nuclear yellow into oneeye and true blue into the other so that bilaterally as well asunilaterally projecting neurons could be identified. Controlinjections were made into the orbits. After survival times of4-7 days for horseradish peroxidase, 2 days for nuclear yel-low, and 7 days for true blue, eye cups and brains were fixedand sectioned as for immunocytochemistry. Horseradishperoxidase was localized by either the 3,3'-diaminobenzi-dine (26) or the Hanker-Yates method (27) and visualized ina bright field. Both nuclear yellow and true blue were visual-ized by fluorescence in a dark field (30).To determine the total number of cells in the nervus ter-

minalis we fixed heads of two small fish in Bouin's solution,embedded them in parafin, sectioned them serially at 8 pm,and stained them with hematoxylin and eosin. All terminaliscells were drawn with the aid of a camera lucida and count-ed.To determine whether nervus terminalis cells labeled with

intraocularly injected nuclear yellow and true blue were alsoLHRH- or FMRF-amide-immunoreactive, sections contain-ing cells labeled with nuclear yellow and true blue were pho-tographed completely, treated to localize LHRH-immunore-active or FMRF-amide-immunoreactive cells immunocyto-chemically, and photographed again for comparison.

Electrophysiology. Retinas were isolated from large (15-20cm) dark-adapted goldfish in dim light, mounted receptorside up in a chamber and perfused with oxygenated teleostsaline at 1 ml/min (31). Ganglion cell nerve impulses wererecorded extracellularly with metal microelectrodes (Haer,Brunswick, ME), amplified and displayed conventionally,and stored on analog magnetic tape. Light stimuli, deliveredfrom a dual-beam optical bench, were either spots (0.8-mmdiameter) or annuli (i.d., 2.0 mm; o.d., 4.4 mm) centered onthe receptive field, 600 msec in duration, repeated every 3sec. Their wavelength (red, 650 nm; green, 500 nm; blue, 430nm; interference filters with bandwidths 10-20 nm) and in-tensity were controlled by manually exchanging fixed filters.A computer terminal, keypad, and custom interface allowedus to enter parameters onto the analog tape in digitally read-able form, for analysis on a PDP-11/34 minicomputer using acustom software package (D. White, Datek Services, Calga-ry). Spontaneous activity was usually computed during theperiod starting 2.2 sec after the end of each light stimulus andending at the beginning of the next stimulus. The criterionfor change was a 15% difference in firing rate from 1 min tothe next; base-line fluctuations were <15%. Activity in

-60% of the cells treated with mammalian LHRH or its ana-logs was not analyzed by computer, because it was obviousthat the peptides were ineffective. Changes were considereddrug related if they occurred 1-4 min after the drug enteredthe chamber and reversed in a similar time on washout; theonset latency was closely related to dose. The time for an80% exchange of fluid in the chamber was about 3 min.The perfusion medium could be switched from normal to

altered (e.g., high Mg2', low Ca2+) with a valve and could besupplemented by slow addition of concentrated peptide solu-tion from a microliter syringe pump. Final dilutions were cal-culated, assuming no loss and uniform mixing in tubing andchamber, and represent the maximum concentration pre-sented to the photoreceptor surface of the retina.

Synthetic peptides added to the medium including mam-malian LHRH decapeptide (Peninsula 7201); the "superac-tive" LHRH agonists, [D-Ala6]LHRH (Peninsula 7202) and[D-Ala6, des-Gly10]LHRH-ethylamide (Peninsula 7206); theLHRH antagonists [D-Phe2, D-Ala6]LHRH (Peninsula 7203),[D-pGlul, D-Phe2, D-Trp36]LHRH (Peninsula 7207), and (N-acetyl-D-p-Cl-Phel 2, D-Trp3, D-Phe6, D-Alal0]LHRH (gift ofDavid H. Coy; see ref. 32); salmon brain gonadotropin-re-leasing factor (fish GnRF; [Trp7, Leu8]LHRH) (90-158-20;gift of Jean Rivier; see ref. 33); and the molluscan cardioex-citatory peptide, FMRF-amide (Peninsula 8755) and its en-kephalin-like extended agonist, H-Tyr-Gly-Gly-FMRF-am-ide (YGGFMRF-amide; Peninsula 8764).

RESULTSLHRH-like Immunoreactivity in Goldfish Retina. Certain

antisera to LHRH showed an immunoreactive fiber systemin frozen-sectioned or flat-mounted goldfish retina. As inother species (2, 3), the fibers form a dense plexus in theinterface between inner plexiform and inner nuclear layers(Fig. 1 E, G, and H). Slender beaded processes occasionallyleave the plexus and extend through the inner nuclear layer,sometimes as far as the outer nuclear layer. Some enlargedfibers wrap around (Fig. 1H) or course along putative ama-crine, bipolar, and ganglion cells and may make synapses onthem or their processes (5). As seen in whole-mounts, fibersin the plexus are randomly oriented near the retinal center,but parallel to the retinal margin in the periphery, as are gan-glion cell dendrites (34) and horizontal cell axons (35) in cyp-rinid retinas.These fibers appear not to originate within the retina, as

retinal cell bodies were never stained. Furthermore, a fewstained fibers were observed in the optic fiber, ganglion cell,and inner plexiform layers. In whole-mounts 2-3 dozenthick, smooth immunoreactive fibers could be followed fromthe optic nerve head to their termination in the inner plexi-form layer. These observations suggest that the LHRH-im-munoreactive fibers are efferent-i.e., of central origin.

Retinal LHRH-Immunoreactive Fibers Arise from the Ner-vus Terminalis. Immunoreactive fibers are evident not onlyin the retina but also, as observed elsewhere (3), in the opticnerve (Fig. 1C) and brain regions including olfactory bulb,telencephalon, and optic tectum. LHRH-immunoreactivecell bodies are seen in the hypothalamus (nucleus lateralistuberis) and olfactory nerve. Immunoreactive cells in the ol-factory nerve are loosely clustered along the rostromedialaspect of the olfactory bulb (Fig. LA). They appear bipolar tomultipolar, with perikarya elongated along the rostrocaudalaxis, and measure (32.3 + 4.5 ,um) x (18.7 + 3.1 ,um) (mean± SD, n = 20). LHRH-immunoreactive fibers extend ros-trally from these cells within the olfactory nerve, but wehave not yet been able to identify their distal terminals.LHRH-immunoreactive fibers extend caudally in a loosebundle within the medial olfactory tract, through the basaltelencephalon to the region of the anterior commissure (theventral commissure of Goldstein; ref. 7). Some fibers cross

Neurobiology: Stell et aL

Proc. Natl. Acad. Sci. USA 81 (1984)

*i-. ;|. <

<ica

izaion of peptides

opIc.ne1. (Bart (s)oLHRH-immunoreaoctie and (F)al

FMFamide-immunoreactive immunoploeroxdae-taneterminalisnuosifilserinvralsections ofolatretinereandbl,showingamenenleu dofubead

end (arros)atthe bmordeeatweefinnersi splexiformllayers(ipf

and inner nuclear layer (mnl). (Bar = jim.) (G and H) Higher mag-

nification of immunoperoxidase preparation for LHRH-immunore-

active terminals in retina, as in E and F. (G) The plexus of fibers

(arrows) can be seen near a large cell (a) in the amacrine cell portion

of the mnl. (H) At a different focal plane, branches from the plexus

can be seen enveloping the cell in a basket-like arrangement. (Bar =

50 tim.)

here, while others continue caudally without crossing. At

least some of the crossed and uncrossed fibers join the base

of the optic nerves, near the chiasm, and continue distally,

randomly dispersed, in their superficial layers. We observed

20-30 LHRH-immunoreactive cell bodies in each olfactory

nerve and a comparable number of LHRH-immunoreactive

fibers in each optic nerve.

Retrograde labeling after intraocular injection of tracers

proved that the LHRH-immunoreactive retinal fibers arise

from cells associated with the olfactory system in the brain

and perhaps from no other brain center. Intraocularly inject-

ed horseradish peroxidase, nuclear yellow, and true blue

were transported to cells in the olfactory nerve. These mark-

ers also labeled cells in the trigeminal ganglion and mesen-

cephalon (cranial nerve V), oculomotor complex (cranial

nerves III, IV, and VI) and hypoglossal nucleus (cranialnerve XII), but probably because of leakage into the orbitsince cells in these locations were labeled also after intraor-bital injection. In two animals in which one eye was injectedwith nuclear yellow and the other with true blue, 12.0 + 1.2cells per olfactory nerve (mean ± SD, n = 4 nerves) project-ed only to the contralateral retina, 3.3 ± 0.4 cells projectedonly to the ipsilateral retina, and 1.5 ± 1.1 cells projected toboth retinas. These cells appeared similar in size and shapeto the LHRH-immunoreactive cells.Having thus proven independently that (some) cells in the

olfactory nerves are LHRH-immunoreactive and that (some)cells in the same region project to the retina, it remained tobe established to what degree the two classes of cells areidentical. This was shown by double retrograde and im-munocytochemical labeling. In four animals a total of 23.29+ 2.71 (mean ± SD; n = 8 nerves) cells per olfactory nervewere LHRH-immunoreactive, whereas 14.38 + 2.50 cellsper nerve were labeled retrogradely from one or both eyes.All of the retrogradely labeled cells were LHRH-immunore-active. The difference, 8.91 ± 3.68 cells per nerve, repre-sents cells that are immunoreactive but have no demonstra-ble projection to the retina. In serial paraffin sections coun-terstained with hemotoxylin and eosin, however, wecounted a total of 62 ± 4.08 (mean + SD; n = 3 nerves fromtwo fish) large neurons in each olfactory nerve. Thus asmany as 40 (nearly 2/3) of these neurons appear not to beimmunoreactive for LHRH.The location of these cell bodies and fibers within the ol-

factory nerve and tracts of goldfish fits precisely the descrip-tion of the nervus terminalis in the goldfish (1, 7), carp (12),and catfish (13). Furthermore, the caudal course of the fi-bers, including the projection along the optic nerves, fits pre-cisely the description of the nervus terminalis in smelt (36)and goldfish (7). We feel justified, therefore, in identifyingthe goldfish LHRH-immunoreactive olfactoretinal efferentsystem (with its projections within the brain) as the nervusterminalis and the cluster of nerve cell bodies in the olfactorynerve as its ganglion.

Goldfish Nervus Terminalis Is Immunoreactive for TwoPeptides. During this investigation several reports appearedin which FMRF-amide-immunoreactive neurons were local-ized to various regions in the vertebrate brain (15-17). In oneof these (16), FMRF-amide-immunoreactive cell bodies wereillustrated in the ventral telencephalon of the teleost Poecil-ia. FMRF-amide was also of interest because of its sequencehomology to the enkephalins, in view of our previous studieson opioid pathways in goldfish retina (37).We find that the localizations of FMRF-amide-like immu-

noreactivity in olfactory nerve (Fig. 1B), optic nerve (Fig.1D), and retina (Fig. 1F), not to mention olfactory bulb, tel-encephalon, and optic tectum (not shown), are identical tothose of LHRH-immunoreactivity. Furthermore, mapping ofLHRH-immunoreactive and FMRF-amide-immunoreactivefibers in serial sections of the optic nerves yields similar pat-terns, and sequential staining of individual sections revealsno singly labeled LHRH-immunoreactive or FMRF-amide-immunoreactive optic nerve fibers. Finally, in serial sectionsof the terminalis ganglion stained separately (Fig. 1 A and B)or in individual sections restained after elution of the firstprimary antibody (not shown), all cell bodies immunoreac-tive for one peptide are seen to be immunoreactive for theother as well. Controls confirm that the nervus terminaliscontains two independent immunoreactive species, closelyresembling LHRH and FMRF-amide. Cross-preabsorption(anti-LHRH with FMRF-amide, anti-FMRF-amide withLHRH) does not diminish the intense staining usually ob-served with these antisera.

Action of Exogenous Peptides in the Retina. The effects ofexogenous LHRH, FMRF-amide, and analogous peptideson the activity of retinal ganglion cells might indicate how

942 Neurobiology: Stell et aL

op, 4*1

Proc. Natl. Acad. Sci USA 81 (1984) 943

the nervus terminalis influences visual function in the retina.Most of the cells recorded from were either red-ON-centeror red-OFF-center double-color-opponent ganglion cells(38). The effects of FMRF-amide (or YGGFMRF-amide)were similar in ON and OFF cells. In an OFF cell (Fig. 2), ared spot elicited a strong transient ON inhibition and a strongtransient OFF excitation; weaker sustained responses fol-lowed the transient responses. YGGFMRF-amide (orFMRF-amide) produced (0 an increase in spontaneous firingrates: 70% for OFF cells (control rates, 4-8 Hz), 30%o for ONcells (control rates, 12-20 Hz); cells with control rates closeto zero (ON or OFF types) developed spontaneous firing at1-5 Hz; (h) loss of the sustained inhibition normally elicitedby light-ON (or light-OFF, not seen here); and (iii) an in-crease in the extent of pulsatile or periodic response to light-ON or light-OFF. About 2/3 of OFF cells and 1/3 of ONcells showed these effects (Table 1).FMRF-amide was not effective in the presence of blockers

of synaptic transmitter release (1 mM Co2+, one cell; noCa2+, 20 mM Mg2+, one cell). This finding, which suggeststhat FMRF-amide acts through one or more interneurons, isconsistent with our observation that immunoreactive fibersusually do not contact ganglion cells (5).LHRH, its superactive analogs, and its antagonists were

rarely effective in altering ganglion cell function. LHRH waseffective in only 4 of 27 units tested (Table 1). The peptidase-resistant agonists, [D-Ala6]LHRH, which is effective at 0.01-0.001 times the concentration of LHRH in bullfrog sympa-thetic ganglion (39), and [D-Ala6, des-Glyl0]LHRH-ethyl-amide, reported to be 100 times more potent than LHRH toinduce spawning in fish (40), were no more effective thanLHRH. The inactivity of mammalian LHRH and its analogsmay be due to their inappropriate structure. This is suggest-ed by our preliminary results with fish GnRF (33), a plausibletransmitter candidate in this retinopetal system.The effects of fish GnRF were not distinguishable from

those of FMRF-amide. As shown in Fig. 3, GnRF producedin green-ON-center cells: (i) an increase in spontaneous fir-ing, (it) a loss of the sustained inhibition normally seen at

250T

250-c)u)

0.

4)

V) A250-

250-T

Control

YGGFMRFamide

Recovery

i

I

Table 1. Proportion of cells showing peptide-induced effects

Mammalian Mammalian Salmon FMRF-amideLHRH and LHRH brain and YGGFMRF-analogs antagonists GnRF amide

OFF cells 4/17 1/9 7/8 24/33ON cells 1/12 0/9 4/5 12/23The three effects considered are those listed in the text or in Fig. 2

or 3. Almost all affected cells showed increases in spontaneous ac-tivity and loss of light-induced inhibitions. A few showed only in-creased pulsatility. Three units that showed decreases in rate havenot been included.

green-OFF, and (iii) an increase in pulsatility at green-OFF.It is noteworthy that neither the balance between green-

and red-center responses nor threshold intensity (not shown)is altered by either GnRF (Fig. 3) or FMRF-amide (notshown).

Preliminary experiments indicate that the effects of thetwo peptides are not additive. In three cells that were excitedby GnRF, the addition of YGGFMRF-amide did not produceany additional effect. In one cell excited by GnRF, the addi-tion of YGGFMRF-amide caused a rapid and reversible de-crease in the firing rate.

DISCUSSIONThese experiments confirm that in goldfish, as in other tele-ostean species (1-3, 41), there is a centrifugal pathway to theretina from cell bodies associated with the forebrain. Thisretinopetal system is clearly identical to the nervus termina-lis in carp as described by Sheldon (12) and the nervus ter-minalis and tractus terminalis-opticus in smelt as describedby Holmgren (36), and homologous to the nucleus and trac-tus olfactoretinalis of various teleosts according to Munz andcolleagues (2, 3, 41). Because its neurons may be primaryolfactory receptors (36), the nervus terminalis may thereforeconstitute a direct pathway by which specific olfactory stim-uli can alter the visual responsiveness of specific retinal out-flow neurons. Although we have been unable to observe any

0.4)

ALIS, LI-

&-L f. iLi=jkE

0 1Time, sec

FIG. 2. YGGFMRF-amide affects spike activity in a red-OFF-center, double-color-opponent ganglion cell (excited by red centeror green surround at light-OFF and inhibited at light-ON; the re-verse for green center and red surround). A moderately bright redlight (4 ,uW/cm2) was flashed every 3 sec for 600 msec (horizontalbar at top). Each of the three traces is the average of 20 sweeps (i.e.,-1 min). YGGFMRF-amide increases spontaneous activity (i.e.,just before the light stimulus); increases activity during the sustainedlight-ON phase, which is normally inhibited; and markedly enhancesthe pulsatile activity induced by either light-ON or light-OFF.

Time, sec

FIG. 3. Fish GnRF affects spike activity in red-OFF-center dou-ble-color-opponent ganglion cell. Stimulus was a green spot of lowintensity (0.02 ,uW/cm2); other features are as in Fig. 2. The greenstimulus produces a transient and sustained ON excitation and OFFinhibition, in addition to a transient OFF response by activation ofthe red system. Fish GnRF, like FMRF-amide, increases spontane-ous activity, increases activity during the normally inhibited OFFphase, and markedly enhances the pulsatility induced by light-ON orlight-OFF. The balance between the green- and red-center influ-ences is not altered, however.

Neurobiology: Steil et aL

A. __&J~-ML

I

Proc. Natl. Acad. Sci. USA 81 (1984)

other retinopetal system in goldfish, several authors have de-scribed multiple retinopetal systems in various teleosts (41,42). The retina is not likely to be the only destination of ner-vus terminalis fibers, which are said to project also to olfac-tory nuclei (36) and to centers near the anterior commissurethat may be concerned with olfactory (pheromonal?) regula-tion of sexual behavior and reproduction (6, 36, 43, 44).The efficacy offish GnRF and FMRF-amide at micromolar

concentrations supports the candidacy of a LHRH-like and aFMRF-amide-like peptide as neurotransmitters of the nervusterminalis in the retina. In view of the homology of the ex-tended peptide, YGGFMRF-amide to opiate peptide [Met]-enkephalin, it is noteworthy that at threshold concentrationsFMRF-amide agonists enhance, while opiate agonists de-press (37), the activity of goldfish OFF-center ganglion cells.The behavior of mammalian LHRH in goldfish retina is

consistent with the observation that mammalian LHRH andsuperactive agonist analogs are effective only at high dosesand are equipotent in releasing gonadotropin from trout pitu-itary (45). Because salmon brain GnRF is highly potent in theretina, our LHRH agonist and antagonist data suggest thatthe teleostean GnRF and its postsynaptic membrane recep-tor differ from their counterparts in mammals.The most optimistic interpretation of our data is that the

retinopetal nervus terminalis fibers in goldfish contain bothLHRH- and FMRF-amide-like peptides, that specific olfac-tory stimulation of these fibers causes release of both pep-tides from axon terminals in the distal part of the inner plexi-form layer and inner nuclear layer, and that both the LHRH-and FMRF-amide-like peptides act as neurotransmitters.These peptides might be expected to be excitatory from theiraction in other systems (14, 18, 39), but almost any effectcould result either from differences in postsynaptic responseto the peptides or from sign inversions in polysynaptic path-ways to the ganglion cells. We speculate that the major in-puts to sustained ganglion cells are a sign-conserving inputfrom bipolar cells and a sign-inverting input from sustainedamacrine cells, working in a push-pull manner (46, 47). Ourdata are consistent with effects of both FMRF-amide andGnRF on sustained amacrine cells, because (i) immunoreac-tive fibers terminate on amacrine cells or their processes,and (it) sustained amacrine cells are probably responsible forthe 10-Hz pulsatility (48), which was markedly enhanced byboth peptides, especially during sustained light-evoked ac-tivity (Figs. 2 and 3).

We thank Dr. R. Glenn Northcutt for introducing us to the nervusterminalis; Drs. Gerald Kozlowski, Graham Dockray, Shirley Jo-seph, and others for gifts of antisera; Dr. Jean Rivier for salmonbrain GnRF; Dr. David Coy for LHRH antagonists; and Dr. NinaTumosa for preparing some of the illustrations in Fig. 1. Mrs. MaryPollock patiently and skillfully typed the manuscript. The researchwas supported by an operating grant from the Medical ResearchCouncil of Canada (MRC) and an establishment grant from the Al-berta Heritage Foundation for Medical Research (AHFMR) toW.K.S.; U.S. Public Health Service National Research ServiceAward 3 F32 EY-05520 to S.E.W.; and a postdoctoral fellowshipfrom the MRC and Research Allowance from the AHFMR toA.K.B.

1. Demski, L. S. & Northcutt, R. G. (1983) Science 220, 435-437.

2. Munz, H., Stumpf, W. E. & Jennes, L. (1981) Brain Res. 221,1-13.

3. Munz, H., Claas, B., Stumpf, W. E. & Jennes, L. (1983) CellTissue Res. 222, 313-323.

4. Stell, W. K., Chohan, K. S., Lam, D. M. K. & Kozlowski,G. P. (1982) Invest. Ophthalmol. Vis. Sci. Suppl. 22, 278.

5. Ball, A. K. & Stell, W. K. (1983) Invest. Ophthalmol. Vis. Sci.Suppl. 23, 66.

6. Stacey, N. E. & Kyle, A. L. (1983) Physiol. Behav. 30, 621-628.

7. Springer, A. D. (1983) J. Comp. Neurol. 214, 404-416.8. Silverman, A. J., Antunes, J. L., Abrams, G. M., Nilaver, G.,

Thau, R., Robinson, J. A., Ferin, M. & Krey, L. C. (1982) J.Comp. Neurol. 211, 309-317.

9. Schwanzel-Fukuda, M. & Silverman, A. J. (1980) J. Comp.Neurol. 191, 123-225.

10. Iversen, L. L. (1983) Annu. Rev. Pharmacol. Toxicol. 23, 1-27.

11. Hckfelt, T., Johansson, O., Ljungdahl, A., Lundberg, J. M. &Schultzberg, M. (1980) Nature (London) 284, 515-521.

12. Sheldon, R. E. (1912) J. Comp. Neurol. 22, 177-339.13. Brookover, C. & Jackson, T. S. (1911) J. Comp. Neurol. 21,

237-257.14. Price, D. A. & Greenberg, M. J. (1977) Science 197, 670-671.15. Dockray, G. J., Vaillant, C. & Williams, R. G. (1981) Nature

(London) 293, 656-657.16. Boer, H. H., Schot, L. P. C., Veenstra, J. A. & Reichelt, D.

(1980) Cell Tissue Res. 213, 21-27.17. Weber, E., Evans, C. J., Samuelson, S. J. & Barchas, J.

(1981) Science 214, 1248-1251.18. Gayton, R. J. (1982) Nature (London) 298, 275-276.19. Hokfelt, T., Lundberg, J., Schultzberg, M., Johansson, O.,

Skirboll, L., Anggard, A. Fredholm, B., Hamberger, B., Pen-now, B., Rehfeld, J. & Goldstem, M. (1980) Proc. R. Soc. Lon-don B Ser. 210, 63-77.

20. Pelletier, G., Steinbusch, H. W. M. & Verhofstadt, A. A. J.(1981) Nature (London) 293, 71-72.

21. Erichsen, J. T., Reiner, A. & Karten, H. J. (1982) Nature(London) 295, 407-410.

22. Dockray, G. J. & Williams, R. G. (1983) Brain Res. 266, 295-303.

23. Coons, A. H. (1958) in General Cytochemical Methods, ed.Danielli, J. D. (Academic, New York), pp. 399-422.

24. Hartman, B. K. (1973) J. Histochem. Cytochem. 21, 312-332.25. Hbkfelt, T., Fuxe, K. & Goldstein, M. (1975) Ann. N. Y. Acad.

Sci. 254, 407-432.26. Sternberger, L. A. (1974) Immunocytochemistry (Wiley, New

York).27. Hanker, J. S., Yates, P. E., Metz, C. B. & Rustioni, A. (1977)

Histochem. J. 9, 789-792.28. Greenwood, R. S. & Winstead, K. K. (1981) Soc. Neurosci.

Abstr. 7, 100.29. Tramu, G., Pillez, A. & Leonardelli, J. (1978) J. Histochem.

Cytochem. 26, 322-324.30. Kuypers, H. G. J. M., Bentivoglio, M., Catsman-Berrevoets,

C. E. & Bharos, A. T. (1980) Exp. Brain Res. 40, 383-392.31. Ishida, A. T. & Fain, G. L. (1981) Proc. Natl. Acad. Sci. USA

78, 5890-5894.32. Erchegyi, J., Coy, D. H., Nekola, M. V., Coy, E. J., Schally,

A. V., Mezo, I. & Teplan, I. (1981) Biochem. Biophys. Res.Commun. 100, 915-920.

33. Sherwood, N., Eiden, L., Brownstein, M., Spress, J., Rivier,J. & Vale, W. (1983) Proc. Natl. Acad. Sci. USA 80, 2794-2798.

34. Kock, J.-H. & Reuter, T. (1978) J. Comp. Neurol. 179, 549-567.

35. Stell, W. K. (1975) J. Comp. Neurol. 159, 503-519.36. Holmgren, N. (1920) Acta Zool. 1, 137-315.37. Djamgoz, M. B. A., Stell, W. K., Chin, C.-A. & Lam,

D. M. K. (1981) Nature (London) 292, 620-623.38. Daw, N. W. (1968) J. Physiol. (London) 197, 567-592.39. Jan, Y. N., Jan, L. Y. & Kuffler, S. W. (1979) Proc. Natl.

Acad. Sci. USA 76, 1501-1505.40. Cooperative Team for Hormonal Application in Pisciculture

(1977) Sci. Sin. 20, 469-474.41. Munz, H. & Claas, B. (1981) Neurosci. Lett. 22, 223-226.42. Ebbesson, S. 0. E. & Meyer, D. L. (1981) Science 214, 924-

926.43. Finger, T. E. (1975) J. Comp. Neurol. 161, 125-142.44. Kim, Y. S., Stumpf, W. E. & Sar, M. (1978) J. Comp. Neurol.

182, 611-620.45. Crim, L. W., Evans, D. M., Coy, D. H. & Schally, A. V.

(1981) Life Sci. 28, 129-135.46. Marc, R. E. & Lam, M.-K. (1981) J. Neurosci. 1, 152-165.47. Belgum, J. H., Dvorak, D. R. & McReynolds, J. 5. (1982) J.

Physiol. (London) 326, 91-108.48. Chan, R. Y. & Naka, K. I. (1976) Vision Res. 16, 1119-1129.

944 Neurobiology: Steil et aL