Brain Research Bulletin. Vol. 29, pp. 58 l-587, 1992 0361s9230/92 $5.00 + 00 Printed in the USA. All rights reserved. Copyright 0 I992 Pqamon Press Ltd.

Inhibitory Inputs to the Subfornical Organ From the AV3V: Involvement of GABA

TOSHIMASA OSAKA,* HIROSHI YAMASHITA* AND KIYOMI KOIZUMIt’

*Department of Physiology, University of Occupational and Environmental Health, Kitakyushu 807, Japan, and f Department of Physiology, State University of New York, Health Science Center at Brooklyn, NY I I203

Received 22 November 199 1; Accepted 16 April 1992

OSAKA, T., H. YAMASHITA AND K. KOIZUMI. Inhibitory inputs to the subfornical organ from the AMV: Involvement of GAEA. BRAIN RES BULL 29(5) 581-587, 1993.-Action potentials were recorded extracellularly from single neurons in the subfomical organ (SFO) of the pentobarbital-anesthetized cat following stimulation of the regions surrounding the anteroventral third ventricle (AV3V). Of 328 SF0 neurons studied, 103 were antidromically activated, showing direct projections from the SF0 to the AV3V. However, the major effects of stimulations of the AV3V on SF0 neurons were orthodromic inhibition; almost 30% of SF0 neurons were inhibited by various sites in the AV3V, while a smaller proportion of cells were excited. Local application of bicuculline, an antagonist for GABA, attenuated the inhibitory responses induced by stimulation of the AV3V in seven out of eight neurons tested. Application of GABA inhibited 16 out of 24 neurons, while that of bicuculline alone excited I 1 out of 26 neurons, suggesting the tonic inhibitory action of GABA on some SF0 neurons. On the other hand, application of kynurenic acid, a nonspecific antagonist for the excitatory amino acids, did not affect the excitatory responses induced by stimulation of the AV3V, but kynurenic acid itself inhibited 6 out of 18 neurons tested. Application of glutamate excited most SF0 neurons. This suggests that the excitatory amino acids may be the transmitter(s) of interneurons in the SF0 but may not mediate the excitation from the AV3V.

Subfornical organ AV3V GABA Bicuculhne Excitatory amino acids Water balance

THE subfornical organ (SFO), one of the circumventricular or- gans that have no blood-brain barrier, is thought to be involved in body water homeostasis. The organ responds to changes in osmotic pressure of plasma or tissue fluid as well as to blood- borne substances, particularly angiotensin II and natriuretic polypeptides (3,10,11,20,23). Changes in activity of the SF0 neurons are transmitted to the hypothalamus where drinking, vasopressin release, and cardiovascular responses are integrated to maintain water balance. Since the SF0 is generally considered to be a sensory organ, only a little attention has been paid to its neural inputs. However, histological studies (I ,6) showed that many synapses and nerve fibers containing various putative neurotransmitters are present in the SFO, suggesting the presence of abundant neural inputs. In our previous studies, mainly in- hibitory inputs to the SF0 from the paraventricular nucleus of the hypothalamus (PVN) were found in the cat (26). Others (9) reported inputs from the median preoptic (POMn), supraoptic, and paraventricular nuclei in the rat. Neurotransmitters involved in these neural inputs to the SF0 are largely unknown, though in the study in vitro SF0 neurons responded to carbachol, an- giotensin II, and serotonin (2).

The region surrounding the anteroventral third ventricle (AV3V) receives fibers from and projects to the SF0 (5,12,18,19). These mutual connections between the SF0 and the AV3V are

considered to be important for the neurohumoral integration, particularly water homeostasis (8,24). The purpose of this paper, therefore, is to clarify the relationships between the SF0 and the AV3V using electrophysiological techniques. We studied changes in the SF0 neuron activity in response to stimulations of various parts of the AV3V; namely, the POMn, the organum vasculosum of the lamina terminalis (OVLT), and the fornix. The ventral (vSF0) and dorsal (dSF0) stalks of the SF0 were also stimulated, because the former contains most nerve fibers to and from the SF0 neurons, while the latter was used for comparison. In addition, involvement of GABA and the excit- atory amino acids as possible transmitters for the inhibitory and excitatory responses of the SF0 neurons, respectively, were ex- amined by local applications of the chemicals.

METHOD

Surgical Procedures

Experiments were carried out in 16 cats of either sex, weighing 2.7-3.6 kg. Animals were anesthetized with nembutal (35 mg/ kg. IP), the supplemental doses (5 mg/kg, IV) were given as needed . A trachea cannula was inserted, a radial vein and a femoral artery were cannulated for fluid administration, and blood pressure and heart rate recordings, respectively. A copper-

’ Requests for reprints should be addressed to Dr. K. Koizumi, Department of Physiology, Box 3 I, State University of New York Health Science Center at Brooklyn, 450 Clarkson Avenue, Brooklyn, NY 11203.

581

582

constantan thermocouple was inserted into the trachea cannula to monitor respiration. Body temperature was maintained be- tween 36 and 38°C by a heating pad.

The SF0 and AV3V regions were exposed by hemispherec- tomy. After extended craniotomy of the left hemisphere, the corpus callosum and the thalamus were cut midsagittally, and the rostra1 forebrain was cut parasagittally along the lateral ven- tricle. The left hemisphere was cut transversely at the caudal edge of the third ventricle and was removed from the animal. Cerebral arteries were ligated at the cranial floor. The cut surface was covered with thin slices of gelatin-sponge to prevent any gradual bleeding. The residual anterior hypothalamic area and the lateral septum of the left hemisphere was slightly pulled ros- tro-laterally using a thin stainless steel plate to view clearly the SF0 and structures in the preoptic recess of the third ventricle (Fig. I). Care was taken not to damage the SF0 and regions around the anterior commissure. Cerebrospinal fluid sponta- neously flowed from the choroid plexus and covered exposed surface of the SF0 and AV3V regions.

Electricul Stimulution

Concentric bipolar stainless steel electrodes (0. I mm-tip diameter; 0.25 mm-outside diameter; 0.25 mm-distance from tip to concentric ring) were placed on two of the following peri- ventricular regions: the dSF0 and the vSF0 that were 0.8 mm from the center of the SFO, dorsal and ventral parts of the POMn, the fornix at the level of dorsal POMn, and the OVLT. The electrodes were placed diagonally from the ventricular side under direct visual control through a dissecting microscope. Only tips of the electrodes were inserted into the tissue to confine the current spread and to minimize the damage in structures; outer rings of the electrodes were left in the cerebrospinal fluid at the wall of the ventricle. Stimuli were square pulses with a duration of 0.5 ms and a strength of 10-2 10 PA. Single pulse or a brief train of three pulses at 180 Hz was employed

OSAKA, YAMASHITA AND KOIZUMI

Recording

Glass microelectrodes filled with 0.5 M sodium acetate so- lution containing 2% Pontamine Sky Blue (Chicago Sky Blue, Sigma, St. Louis, MO) were inserted in the SF0 under visual control. Recordings were made from the surface to 0.4 mm depth and within the middle part of the SF0 to avoid recording from the septo-fimbrial nucleus or the hippocampal commissure, be- cause the thickness of the SF0 was 0.4 mm as judged by re- cording of neuronal activity and by histology with dye marking of recorded sites.

Recording sites were easily identified under the dissecting microscope, but the position was occasionally verified by dye marking. At the end of experiments Pontamine Sky Blue dye in a recording electrode was deposited by the current application (5 FA, 5 min). The brain was then fixed in 10% formalin; frozen sagittal sections were made and stained by neutral red.

Drug Application by Multiburrel Pipette

A multibarrel glass micropipette was used to apply drugs to SF0 neurons. Tip diameter of multibarrel pipette was 2-5 pm, and a recording microelectrode was glued to the pipette so as to protrude beyond the pipette tip by approximately 30 pm. Pres- surized nitrogen was used to eject drugs from the multibarrel pipette. Nitrogen gas from a regulated pressure source was fed to each barrel ofthe pipette via a three-way solenoid valve (Gen- eral Valve Corp.) and an eight-way valved connector (Omnifit Ltd.). The pressure ranged from 30 to 820 kPa (4.4 to 120 psi). The threshold pressure for ejection and the amount of ejected solution at a certain level of pressure was measured for each barrel in mineral oil under microscopic observation before and after use. Ejecta were usually in the similar range for different barrels of one pipette. but varied from one pipette to another. Typically, a 20-50 pm diameter (4-65 pl) droplet was formed at the tip of a micropipette during application of pressure for IO-20 s.

-. -- _ -- __.

tiy~oiholomus

Inhibition Excitation Antidromic

(X) 40 30 20 10 0 10 20 30 30 20 10 0

OVLT n=84

FIG. I Percent incidence of inhibition (solid bar), excitation (horizontal crosshatch), and antidromic activation (diagonal hatch) of SF0 neurons to stimulation of the rostra1 periventricular regions. Some neurons are classified into several categories of responses, since they showed various responses by varying the intensity of stimulation. The diagram shows the exposed periventricular surface of the right hemisphere. SFO; subfornical organ, dSF0; dorsal stalk of the SFO, vSF0; ventral stalk of the SFO, dPOMn; dorsal part of the median preoptic nucleus, vPOMn; ventral part of the median preoptic nucleus. AC: anterior commissure, OVLT: organum vasculosum of the

lamina terminalis.

INHIBITORY INPUTS TO THE SUBFORNICAL ORGAN 583

The following drugs were used: GABA; bicuculline methio- dide, an antagonist for GABA; monosodium glutamate; kynu- renic acid, a nonspecific antagonist for excitatory amino acids; and vehicle, artificial cerebrospinal fluid (ACSF). The constit- uents of ACSF were (mmol/kg): NaCl 156; KCI 1.6; CaC& 1.26; MgS04 1.16; NaHC03 8.4; KI-12P04 1.25; glucose 3.3. Drugs were applied at concentration of 1 mmol/kg.

Analysis

Extracellularly recorded action potentials of single neurons were conventionally amplified, displayed on a storage oscillo- scope, and led to a ratemeter. The criteria for antidromically evoked action potentials were:

1. a fixed latency of responses, 2. the ability to respond to stimuli at 100 Hz or higher in one-

to-one relation, 3. collision of an antidromic action potential with a sponta-

neously occurring spike, 4. separation of initial segment and somadendritic spikes at a

high frequency stimulation.

A peristimulus time histogram, the composite of 100-300 sweeps, was constructed from recordings from each neuron; the sweep time was 204.8 ms with 5 12 bins. Responses were judged to be excited or inhibited when more than 30% increase or de- crease in firing rates over control level occurred after stimulation or drug application. Latencies and durations of responses were expressed as mean * SE and compared statistically by use of Mann-Whitney U-test or analysis of variance and post hoc Newman-Keuls’ test.

RESULTS

Analyses were made on a total of 328 SF0 neurons, which showed spontaneous and stable activities for periods long enough to observe the effects of electrical stimulation of at least one periventricular site (dSF0, n = 72; vSF0, n = 90; POMn, n = 112; the fornix, n = 109; OVLT, n = 84). Since responses of SF0 neurons to stimulation of the dorsal and ventral parts of POMn were similar, these data were pooled. Basal firing rates of SF0 neurons ranged from 0.3 to 20 Hz; 4 to 7 Hz being the most frequent.

Inhibition

The spontaneous activity of SF0 neurons was most often inhibited by stimulation of various sites (Fig. 2). Figure 1 shows the summary of neuron responses. Percentages of orthodromi- tally inhibited and excited as well as antidromically activated cells are indicated. The inhibition was more often observed than the excitation by stimulations of most sites examined. Percent- ages of inhibited neurons (27-33%) by the stimuli applied to all sites, except the OVLT (13%) were similar. The onset latencies of inhibition were: the dSF0 stimulation, 13 + 3 ms; vSF0, 7 f 1 ms; POMn, 22 + 2 ms; OVLT, 2 1 + 2 ms and fomix, 19 ? 2 ms. The duration of inhibition was in the range of 10 to 300 ms. The inhibitory effect was more clearly detected with stimuli of three consecutive pulses than a single pulse, although the latencies of the inhibitory responses did not differ significantly (Fig. 2).

A small proportion of the inhibitory responses (20 out of total 126) was accompanied by antidromic excitation, though its threshold was much higher than that for the inhibitory re- sponse. In an example illustrated in Fig. 2, the threshold for antidromic excitation of an SF0 neuron was 21 PA, while the

vSF0 stem

t 68pA (threshold 21pA)

r

I single, 13pA, 200 sweeps

+ 3-pulses, 13pA, 120 sweeps

- - 100 ms IO .ns

FIG. 2. Inhibitory responses of an SF0 neuron to stimulation of the vSF0. Top: oscilloscope tracings of antidromic spike (five superimposed sweeps). Middle and bottom: inhibitory responses to single and three- pulse stimulations (subthreshold strength to antidromic activation). On the right, 10 times expanded tracings show that the latencies of the in- hibition in both cases are similar.

neuron was inhibited by a stimulus of 13 PA. Figure 3B and C show another example of inhibition by a stimulus subthreshold to antidromic excitation. Moreover, in another two neurons, though no clear inhibition occurred after antidromic potentials, threee-pulse stimulations at subthreshold to antidromic excita- tion inhibited the neurons.

Inhibition by GABA and excitation by bicuculline. Responses of SF0 neurons to application of GABA were examined on 24 neurons. Of these, 16 neurons were inhibited and the remaining 8 were unresponsive. At low ejection pressure of GABA, the firing rate gradually decreased to the minimum within approx- imately 10 s, while the firing ceased when the ejection pressure was high. All the 16 neurons tested stopped firing during the application of GABA, as an example shown in Fig. 4. After the end of the application, the firing rate recovered to the pretreat- ment level following a transient weak excitatory period.

Application of bicuculline excited 11 neurons (Fig. 4) in- hibited 4 and unaffected 11 among 26 neurons tested. The ex- citation was an alternation of burst-like activities for 2-3 s fol- lowed by a similar period of low discharge rate, but the mean firing rate increased from 4.1 k 0.6 Hz (mean t SEM) to 9.1 + 1.5 Hz during the application of bicuculline.

Blocking action of bicuculline on the inhibitory response in- duced b_v electrical stimulation. Effects of bicuculline on the in- hibitory responses induced by stimulation of the vSF0 (n = 6) or the dSF0 (n = 2) were examined. Bicuculline attenuated the

584 OSAKA. YAMASHITA AND KOIZUMI

dSF0 stem

1 87 /LA (Thresnold)

ImV

L--1 5ms

1 27/~A

li

0

t 15,uA

/ 14

ron, but at 110 HA. which was subthreshold to antidromic ex- citation, stimulation of the same site evoked spikes with slightly fluctuating latencies of 2 1 ms. Moreover, summation of two or three consecutive stimuli was necessary to cause this excitatory response (Fig. 6B and C). The excitatory responses having fixed latencies were observed in four cells to stimulation of the OVLT (two cells) and the dSF0 (two cells).

In six antidromically activated SF0 cells, a weak stimulus subthreshold to antidromic activation produced the excitatory response (Figs. 3 and 6). In Fig. 3, the stimulation at only the lowest intensity ( 1 1 FA) excited, while stronger ( 15 and 27 PA) stimulation inhibited the cell, but even this latter intensity was far below the threshold for antidromic excitation (87 hA). These observations suggest that the stimuli were applied to sites where both excitatory and inhibitory afferents to the SF0 as well as efferent fibers from the SF0 also exist. The excitatory response of SF0 neurons might be suppressed by the concurrent inhibi- tion.

Excitation !I)! gluturnutc~ and inhibition by kynurenic acid.

Application of glutamate excited 29 of 4 1 SF0 neurons tested. At the high ejection pressure, glutamate caused high-frequency firing with rapidly decreasing amplitude of action potentials for a few seconds, suggesting depolarization of the cells. After ap- plication of glutamate, neurons were not spontaneously active for a few seconds but soon recovered to their basal firing level (Fig. 4). Kynurenic acid, a nonspecific antagonist for excitatory

1

50ms amino acids, was applied to 18 neurons. Of these, 12 neurons were unresponsive, while the remaining 6 neurons were inhibited.

FIG. 3. Responses of an SF0 neuron obtained by varying intensity of In two of six inhibited neurons spontaneous discharges ceased the dSF0 stimulation. A: oscilloscope tracings of antidromic action po- during application of kynurenic acid (Fig. 4), while in the other tentials (six sweeps; evoked at the threshold intensity, 87 rA). B to D: Peristimulus time histograms of 100 sweeps when stimulus intensity was

four neurons their firing rate decreased to 50% of their sponta-

reduced. B and C: Inhibitory responses at 27 and 15 PA intensities. neous firing level even when the maximum ejection pressure

respectively. D: Lowering the stimulus intensity further to I I PA altered was used.

the response from inhibition to excitation. Fuihm (f k~wurmic ucid to block the mcitutory responsr

cwkcd by clectricul stimuiution. Effects of kynurenic acid were examined on the excitatory responses after stimulation of the

inhibitory responses in seven neurons (Fig. 5), and did not affect one response. Since bicuculline itself excited only one neuron of these eight neurons, it is unlikely that effects of electrical stim- ulation were masked by the excitation induced by bicuculline. Recovery of the inhibitory response to electrical stimulation fol- lowing bicuculline was tested in two neurons. In both cases, the inhibition could be evoked within 1 min after application of bicuculline.

Excitation

A proportion of the SF0 neurons that were excited by peri- ventricular structures tested in this experiment was smaller than that inhibited (Fig. 1). The onset latencies of excitation were: for the dSF0 stimulation, 17 -t 2 ms; vSF0, 14 + 2 ms; POMn, 20 + 1 ms; OVLT, 32 f 4 ms and fornix, 22 f 2 ms. The latencies of excitation to stimulation of the vSF0 and the fomix were significantly longer than those of inhibition (p < 0.01 in both cases).

Most excitation had long (up to 100 ms) periods of increased firing probability after stimulation, suggesting polysynaptic pathways. In a few instances, however, evoked spikes after con- stant latency were observed. Such evoked response could not follow every stimulating pulse in one-to-one manner and showed no collision cancellation with a spontaneously occurring action potential. Such neurons were probably excited orthodromically rather than antidromically, through monosynaptic pathways. Figure 6 shows that the OVLT stimulation at 2 10 PA evoked antidromic potentials with the latency of 20 ms in an SF0 neu-

vSF0 (n = 7) and the dSF0 (n = 4). The application of kynurenic acid did not alter any of the excitatory responses. Although ap- plication of kynurenic acid alone stopped firing of one neuron, electrical stimulation still evoked the excitatory response.

Altogether 103 cells were antidromically activated by stim- ulation of every region studied (Fig. l), showing direct projections from the SFO. Figs. 2. 3, and 7 show oscilloscope tracings of antidromic action potentials of SF0 neurons. Stimulation of the vSF0 and the POMn effectively evoked antidromic spikes in 32% and 34% of total neurons tested, respectively, while the OVLT evoked antidromic spikes in 23% of neurons. Since the vSF0 consists mainly of fibers. the results suggest that the SF0 efferents pass through the vSF0 and terminate at the AV3V. The fornix stimulation evoked antidromic spikes in only 5 out of IO9 cells. On the other hand, 17% of SF0 neurons projected dorsally. The antidromically excited neurons were present throughout the middle portion of the SF0 from the surface to 0.4 mm depth and no relationship was found between stimu- lation sites and locations of these neurons.

The threshold currents required to evoke the antidromic ac- tion potentials by stimulation of the vSF0 ranged from 13 to 110 HA (mean + SE, 43 f 6 PA), which were similar to those required for the dSF0 stimulation (46 f 8 PA). Threshold values were higher for the POMn ( 158 + 20 PA), the OVLT ( 105 rt 11 PA) and the fornix (197 f 29 PA). The latencies of antidromic responses were: for the dSF0 stimulation, 4.4 f 0.7 ms; vSF0,

INHIBITORY INPUTS TO THE SUBFORNICAL ORGAN

C& G;A GABA Biczline KynFic acid ; - 60 60 80 60 60 40

1 min

FIG. 4. A ratemeter record showing responses of an SF0 neuron to locally applied sodium glutamate solution (Glu), GABA, bicuculline, and kynurenic acid. Horizontal bars indicate the periods of pressure application of solutions. Numerals under bars show ejection pressures in kPa.

4.1 210.3 ms; POMn, 15 + 1 ms; OVLT, 17 f: 1 ms and fornix 6.5 Ifr 0.7 ms. The values were significantly @ < 0.05) shorter than those of orthodromic inhibition and excitation, except the case of the OVLT stimulation. However, the majority (94%) of antidromic potentials were conducted at 0.1-0.5 m/s, suggesting the similar size of axons of SF0 neurons.

Three SF0 neurons were antidromically activated by stim- ulations of two different sites. These pairs of stimulation sites were fomix-ventral POMn, fomix-OVLT, and dorsal POMn- ventral POMn. An SF0 neuron projecting to both the OVLT and the fornix was examined by the collision test (Fig. 7). The result suggests that the SF0 neuron was sending an axon to the OVLT by way of the fornix, or that the axon bifurcated at point close to the fornix. We could not perform detailed collision test for the other two neurons that displayed antidromic activation from two sites. No data suggested, however, that bifurcation of axons occurred at a point close to the cell body.

Before

Bicuculline

I 1 rnv 111’

. .

I 50 FIG. 5. Oscilloscope tracings of an SF0 neuron showing the effect of application of bicuculline on the inhibition induced by stimulation of the vSF0. The tracings (100 superimposed sweeps) were obtained before (top), during (middle), and 1 min after (bottom) application of bicuculline. Application of bicuculline did not affect basal firing rate of the neuron but attenuated the inhibitory input from the vSF0. Ejection pressure was 390 kPa.

585

DISCUSSION

The present results provide electrophysiological evidence for bidirectional connections between the SF0 and the AV3V in the cat. Anatomical studies have shown direct projections from the POMn and OVLT to the SF0 (12,13,18). Orthodromically induced inhibitory and excitatory responses of SF0 neurons were

OVLT stim. 210 /LA

-lOms HO/LA

FIG. 6. Antidromically (A) and orthodromically evoked (B and C) action potentials of an SF0 neuron following stimulation of the OVLT. A la- tency of antidromic potentials was 20 ms (A), while at lower intensity of stimulation (subthreshold for antidromic excitation) an orthodromic activation of the neuron occurred with a latency of 2 I ms (B). Summation of two or three consecutive stimulus pulses was necessary to produce orthodromic action potentials whose latencies slightly fluctuated. The time scale is different in C.

586 OSAKA. YAMASHITA AND KOIZUMI

Fornix

I I

FIG. 7. Oscilloscope tracings of responses of an SF0 neuron that was activated antidromic~ly by stimulation of both the fomix and the OVLT. A: a constant latency (6.5 ms) response to two suprath~shold stimuli to the fornix. B and C: Responses to sequential stimuli given to the OVLT (thick arrow) and the fomix (thin arrow) at different intervals. Antidromic action potentials were evoked by each stimulus. In C, the action potential evoked by the OVLT stimulation is unclear by a stimulus artifact but is present as indicated by an arrow head. Collision did not occur even though the fomix stimulation was applied before recording of an impulse evoked by the OVLT stimulation. D: Cancellation of an antidromic spike evoked by the fornix stimulation, due to its collision with that evoked by the OVLT stimulation. E: Two antidromic action potentials (latency. I6 ms) evoked by the OVLT stimulation. F and G: Stimulation of the fornix preceded that of the OVLT at varied intervals. Collision of the 2nd spike occurred in G. Intervals of the two consecutive stimuli were: 6.5 ms (A), 17.5 ms (B), 15.5 ms (C). 13.5 ms (D). 5.5 ms (E), 20.5 ms (F) and 13.5 ms (Cf.

found more frequently than antidromic activation. In a few cases, excitation caused presumably through monosynaptic connection was also found. However, it is difficult to draw a definite con- clusion from the extracellular recordings whether orthodromic responses are evoked monosynaptically or polysynaptically. es- pecially for inhibitory response. Stimulations of the dSF0 and the vSF0, which situated very close to the SF0 (approximately 0.8 mm away) and were mainly consisted of fibers but not cell bodies, produced orthodromic responses in many SF0 neurons with latencies longer than those of antidromic action potentials. This may suggest that interneuronal connections are present within the SFO. Intracellular study on SF0 neurons in vitro showed multiple EPSPs and IPSPs with marked differences in

latencies evoked by stimulation of the fornix, suggesting the ex- istence of interneurons in the SF0 (3).

Based on the silent period detected after antidromic activa- tion, Buranarugsa and Hubbard suggested an inhibitory inter- neuron that forms a recurrent inhibitory circuit in the SF0 (2,3). The inhibitory responses observed in the present study, therefore, may result from activation of recurrent collaterals by the anti- dromically evoked spikes. However, we consider the inhibition was more likely to be due to a postsynapti~ inhibitor process o~hodromically activated by stimulating afferent fibers to the SFO. The reasons are:

Of 126 neurons that were inhibited by the stimulation of the AV3V. the majority (106 neurons) were not accompanied by antidromic excitation. If the recurrent collaterals of SF0 neurons mediated the greater part of the inhibition, most of the inhibition would have been accompanied by antidromic excitation. Of 20 neurons that were inhibited as well as antidromically excited by the same stimulus, the majority ( I5 neurons) were inhibited at the stimulus strength far below threshold for an- tidromic activation. It is unlikely that such subthreshold stimulation inhibited SF0 neurons through the recurrent collaterals. If the recurrent collaterals mediate the most of inhibition. the incidence of inhibition should be correlated with that of antidromic excitation. However, the incidence ofantidromic excitation by stimulating different regions of the AV3V varied from 5% for fornix stimulation to 34% for POMn stimulation. while the incidence of inhibition evoked by stimulations of dSF0, vSF0. POMn and fornix was similar, approximately 30% of SF0 neurons (Fig. 1).

The neurotransmitter involved in the inhibitory response in the SF0 is probably GABA. The reasons are:

1. Application of GABA inhibited two-thirds of SF0 neurons tested:

2. Application of bicuculline, an antagonist of GABA, excited nearly one-half of SF0 neurons. suggesting the tonic release of GABA to the recorded neurons:

3. Most inhibitory responses induced by electrical stimulation of the dSF0 and the vSF0 were blocked or reduced by ap- plication of bicuculline.

Transmitters other than GABA also may be involved in the inhibitory responses in the SFO. Although we did not test any other possible inhibitory substances, LHRH and somatostatin may mediate inhibitory action from other periventricular regions. These peptide-positive fibers have been found in the SF0 (14). while cell bodies of the peptides are found in the AV3V and in some neurons projecting dorsally over the anterior commissure. ascending along the wall of the ventricle (13). Microinjection of LHRH and somatostatin into the third ventricle suppressed water intake in dehydrated rats (25). LHRH and somatostatin are known to act as CNS depressants (2 I).

Application of kynurenic acid did not affect the excitatory responses induced by electrical stimulation ofthe dSF0 and the YSFO in all I I neurons tested. Therefore, it is unlikely that ex- citatory amino acids mediate excitation in response to stimu- lation of the AV3V regions. However, excitatory amino acids may be neurotransmitter(s) of the intemeurons in the SFO, since application of glutamate excited most SF0 neurons and appli- cation of kynurenic acid inhibited one-third of SF0 neurons tested.

Excitatory responses observed in the present study may be mediated by other transmitters such as acetylcholine, serotonin,

INHIBITORY INPUTS TO THE SUBFORNICAL ORGAN 587

and angiotensin II. These substances are suggested as possible excitatory neurotransmitters in the SF0 (2,15,17). Our study (Osaka and Kawano, unpublished observation) showed, however, that stimulation of the median and dorsal raphe nuclei in the cat produced no excitation, rather inhibition of a few SF0 neu- rons. This suggests a minor role of the raphe nuclei and serotonin in the cat SFO.

Our results indicate that ventrally directed massive efferents run from the SF0 to the AV3V regions, particularly to the POMn, and a few efferents run to the fornix. This observation accords well with neuroanatomical studies in the rat (5,18,19,22). The efferent fibers from the SF0 seemed to form a homogenous population, judging from the uniformity in their conduction velocities (0.1-0.5 m/s), which are similar to the value reported for the projection from the SF0 to the POMn in the rat (9). However, the conduction velocities of these efferents found in the present study are slower than those projecting to the para-

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

ventricular nucleus (PVN) in the cat (26) and rat (9). The dif- ferent conduction velocities of axons of SF0 neurons projecting to the AV3V and the PVN may reflect functional differentiation in the efferent neurons of the SFO. Lind and Johnson (16) im- plicated functional separation of SF0 efferents by the knife-cut experiments. After cutting between the SF0 and the POMn, they found attenuated angiotensin B-induced drinking, but not pressor response that was presumably mediated by the projec- tions to the PVN (7).

In conclusion, this study has shown abundant inhibitory as well as excitatory inputs to the SF0 from the AV3V regions. The SF0 is not simply detecting blood-borne substances but also integrating information reaching from the AV3V through neuronal pathways. GABAergic inputs from the AV3V may control the excitability of SF0 neurons to blood-borne sub- stances, though functional role of these neural inputs to the SF0 must be clarified in the future.

REFERENCES

Akert, K.; Pfenninger, K.; Sandri, C. The fine structure of synapses in the subfornical organ of the cat. Z. ZeIIforsch. 81:537-556; 1967. Buranarugsa, P.; Hubbard, J. I. The neuronal organization of the rat subfornical organ in vitro and a test of the osmo- and morphine-receptor hypotheses. J. Physiol. (Lond.) 291: 101-I 16; 1979. Buranarugsa, P.; Hubbard, J. I. Intracellular recording from neurones of the rat subfornical organ in vitro. J. Physiol. (Lond.) 294:23-32; 1979. Buranarugsa, P.; Hubbard, J. I. Excitatory effects of atria1 natriuretic peptide on rat subfornical organ neurons in vitro. Brain Res. Bull. 201627-63 1; 1988. Camacho, A.; Phillips, M. I. Horseradish peroxidase study in rat of the neural connections of the organum vasculosum of the lamina terminalis. Neurosci. Lett. 25:201-204; 198 I. Dellmann, H. D.; Simpson, J. B. The subfornical organ. Int. Rev. Cytol. 58:333-421; 1979. Ferguson, A. V.; Renaud, L. P. Hypothalamic paraventricular nu- cleus lesions decrease pressor responses to subfornical organ stim- ulation. Brain Res. 305:361-364; 1984. Gross, P. M. The subfomical organ as a model of neurohumoral integration. Brain Res. Bull. 15:65-70; 1985. Gutman, M. B.; Ciriello, J.; Mogenson, G. J. Electrophysiological identification of forebrain connections of the subfomical organ. Brain Res. 382:119-128; 1986. Gutman, M. B.; Ciriello, J.; Mogenson, G. J. Effects of plasma an- giotensin II and hypernatremia on subfornical organ neurons. Am. J. Physiol. 254:R746-R754; 1988. Hattori, Y.; Kasai, M.; Uesugi, S.; Kawata, M.; Yamashita, H. Atria1 natriuretic polypeptide depresses angiotensin II induced excitation of neurons in the rat subfornical organ in vitro. Brain Res. 443:355- 359; 1988. Hernesniemi, J.; Kawana, E.; Bruppacher, H.; Sandri, C. Afferent connections of the subfornical organ and of the supraoptic crest. Acta Anat. 81:321-336; 1972. Knigge, K. M.; Bennett-Clarke, C.; Burchanowski, B.; Joseph, S. A.; Romagnano, M. A.; Stemberger, L. A. Relationship of some releasing-hormone-producing neuron systems to the ventricles of

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

the brain. In: Motta, M., ed. The endocrine functions of the brain. New York: Raven Press; 1980: 195-206. Krisch, B.; Leonhardt, H. Luliberin and somatostatin fiber-terminals in the subfornical organ of the rat. Cell Tissue Res. 2 10:33-45; 1980. Lind, R. W. Bi-directional, chemically specified neural connections between the subfomical organ and the midbrain raphe system. Brain Res. 384:250-261; 1986. Lind, R. W.; Johnson, A. K. Subfornical organ-median preoptic connections and drinking and pressor responses to angiotensin II. J. Neurosci. 2:1043-1051; 1982. Lind, R. W.; Swanson, L. W.; Ganten, D. Angiotensin II immu- noreactivity in the neural afferents and efferents of the subfornical organ of the rat. Brain Res. 32 1:209-2 15; 1984. Lind, R. W.; van Hoesen, G. W.; Johnson, A. K. An HRP study of the connections of the subfornical organ of the rat. J. Comp. Neurol. 2 10:265-277; 1982. Miselis, R. R. The efferent projections of the subfornical organ of the rat: A circumventricular organ within a neural network subserving water balance, Brain Res. 230: l-23; 198 1. Phillips, M. I.; Felix, D. Specific angiotensin II receptive neurons in the cat subfornical organ. Brain Res. 109:531-540; 1976. Renaud. L. P.: Martin, J. B.; Brazeau, P. Depressant action ofTRH, LH-RH and somatostatin on activity of central neurones. Nature 2551233-235; 1915. Saper, C. B.; Levisohn, D. Afferent connections of the median preoptic nucleus in the rat: Anatomical evidence for a cardiovascular integrative mechanism in the anteroventral third ventricular (AV3V) region. Brain Res. 288:2 l-3 1; 1983. Sibbald, J. R.; Hubbard, J. I.; Sirett, N. E. Responses from osmo- sensitive neurons of the rat subfornical organ in vitro. Brain Res. 461:205-214: 1988. Simpson, J. B. The circumventricular organs and the central actions of angiotensin. Neuroendocrinology 32:248-256; I98 1. Vijayan, E.; McCann, S. M. Suppression of feeding and drinking activity in rats following intraventricular injection of thyrotropin releasing hormone (TRH). Endocrinology 100: 1727- 1730; 1977. Yamashita, H.; Osaka, T.; Kannan, H. Effects of electrical and chemical stimulation of the paraventricular nucleus on neurons in the subfornical organ of cats. Brain Res. 323: 176-180; 1984.