Psychopharmacology (1988) 94:371-378 Psychopharmacology © Springer-Verlag 1988

Effects of RO 15-1788 on a running respon rewarded on continuous or partial reinforcement schedules

Marjorie Hawki~*, John Sinden**, Ian Martin***, and Jeffrey A. Gray** MRC Brain, Behaviour and Psychiatry Research Group, Department of Psychology, Institute of Psychiatry, London, UK

Abstract. Two experiments were run in which rats were rewarded with food for running in a straight alley at one trial a day, followed by extinction of the running response. During acquisition of the response, reward was delivered either on a continuous reinforcement (CRF) or on a quasi- random 50% partial reinforcement (PRF) schedule. The groups given PRF were more resistant to extinction than those given CRY, the well-known partial reinforcement ex- tinction effect. In Experiment 1 different groups of rats were injected during acquisition only with 1, 5 or 10 mg/kg of the benzodiazepine antagonist, RO 15-1788, or with pla- cebo. In Experiment 2, 5 mg/kg RO 15-1788 or placebo were administered in a full cross-over design during acquisi- tion, extinction or both. At the end of Experiment 2 only [3H]-flunitrazepam binding was measured in either the pres- ence or absence of added y-aminobutyrate (GABA) in ho- mogenates of hippocampi dissected from the animals that had received behavioural training. The drug affected run- ning speeds during both acquisition and extinction in differ- ent ways depending upon the schedule of reinforcement (CRY or PRF) and also gave rise to enhanced GABA stim- ulation of [3H]-flunitrazepam binding. The results are dis- cussed in relation to the hypothesis that the neurochemical pathways by which reinforcement schedules modify behav- iour include a step influenced by benzodiazepine receptors.

Key words: RO 15~1788 - Benzodiazepine receptors- Par- tial reinforcement extinction effect

The partaal reinforcement extinction effect (PREE) is one of the most robust phenomena in animal learning. It con- sists in the fact that animals trained on a partial reinforce- ment (PRY) schedule, i.e. one in which rewarded and non- rewarded trials are randomly interspersed with one another, are more resistant to extinction than animals trained on a continuous reinforcement (CRF) schedule, i.e. with all trials rewarded. Provided trials are spaced widely apart

Present addresses * Department of Pharmacology, The University of Illinois, College of Medicine at Chicago, 83 South Wolcott Avenue, Box 6998, Chicago, IL 60680, USA ** Department of Psychology, Institute of Psychiatry, De Cre- spigny Park, Denmark Hill, London SE5 8AF, UK *** MRC Molecular Neurobiology Unit, University of Cambridge Medical School, Hills Road, Cambridge CB2 2QH, UK

Offprint requests to: J.A. Gray

(24 h), the PREE is abolished if an anxiolytic dose of the benzodiazepine chlordiazepoxide is administered daily dur- ing training (Feldon and Gray, 1981a, b; McNaughton 1984). This effect is not an instance of state dependency (Overton 1966), since blockade of the PREE occurs whether the animal is tested under the drug or placebo. Abolition of the PREE by chlordiazepoxide is due to changes in the PRF-trained animals, who behave under the drug like CRF-trained animals with or without the drug. Thus, in the one-trial-a-day paradigm, the benzodiazepine blocks the behavioural effects of the nonrewarded trials of the PRF schedule.

The benzodiazepines are thought to produce their overt effects through interaction with specific binding sites which are located on a subpopulation of the receptors for the inhibitory amino acid transmitter, gamma-aminobutyric acid (GABA). These receptors control chloride ion channel gating in neuronal membranes in a complex manner; they are responsive to a variety of drugs which modulate this gating phenomenon through accessory binding sites on the receptor complex. The binding site for the benzodiazepines is but one of these accessory sites (Haefely et al. 1985). Radioligand binding studies allow the biochemical charac- terisation of this receptor complex, not only with regard to individual accessory sites themselves, but also with re- gard to the ways in which these sites are linked to the total receptor complex. One such linkage is revealed by the ability of GABA to increase the affinity of the benzodia- zepine binding site for its agonistic ligands. Subtle changes in these interactions may well be associated with small mod- ifications in the behaviour of animals treated with com- pounds of this class.

The behavioural effects of chlordiazepoxide outlined above raise the possibility that the PREE is mediated by neural mechanisms that involve the benzodiazepine recep- tor and its putative endogenous ligand (Haefely 1984). As an example of such a mechanism, consider the noradrener- gic neurons, with cell-bodies in the locus coeruleus, whose axons travel in the dorsal bundle to innervate, among other sites, the hippocampal formation. Destruction of either the dorsal bundle (Owen et al. 1982) or the hippocampus (Raw- lins et al. 1980) has been reported to abolish the PREE. Furthermore, training on CRY and PRF schedules, respec- tively, gives rise to differing levels of activity of tyrosine hydroxylase (the rate-limiting enzyme in the synthesis of noradrenaline) measured post-mortem in the rat hippocam- pus (Boarder et al. 1979), suggesting that the PREE may be mediated by changes in noradrenaline synthesis and re-

372

lease as a function of reinforcement schedule. Such changes may in turn depend on the action of endogenous agonists or inverse agonists (Haefely et al. 1985) at benzodiazepine binding sites associated with the GABA receptors that have been described both on the oell-bodies of noradrenergic neurons in the locus coeruleus (Guyenet and Aghajanian 1979) and on their terminals in the hippoeampus (Fung and Fillenz, cited in Gray et al. 1984). The blockade of the PREE by exogenous benzodiazepines might then be mediated by an action on these same receptors (Gray et al. 1978). Consistent with this hypothesis, electrophysiologi- cal evidence suggests that benzodiazepines, in low anxio- lytic doses, rather specifically reduce the noradrenergic input to the septohippoeampal system (Gray et al. 1975; McNaughton et al. 1977, Quintero et al 1985).

If the above arguments are correct, one may be able to affect the PREE by blocking the action of the putative endogenous benzodiazepine-receptor ligand by administra- tion of the benzodiazepine antagonist RO 15-1788 (fluma- zepil; Hunkeler et al. 1981). This imidazodiazepine anta- gonises the effects of both agonists and inverse agonists on central benzodiazepine receptors, and it also possesses intrinsic activity in a variety of behavioural, neurological, electrophysiologieal and biochemical tests in both animals and Man (File and Pellow 1986). Various possible mecha- nisms of action for the intrinsic activity of RO 15-1788 have been suggested, including benzodiazepine receptor ac- tivation and antagonism of the endogenous benzodiazepine receptor ligand. I f RO 15-1788 acts in either of these ways, it might be expected to affect the PREE.

We report here two experiments in the straight alley in which we investigated this possibility. In the first, we administered three doses (1, 5 and 10 mg/kg) of RO 15-1788 during acquisition only. In the second, we administered the 5 mg/kg dose in a full cross-over design, i.e. drug or placebo during acquisition combined orthogonally with drug or pla- cebo during extinction. In addition, in order to assess possi- ble changes in the GABA/benzodiazepine receptor complex wrought by the combined behavioural and pharmacological treatment, at the end of Experiment 2 we measured tritiated flunitrazepam binding, in both the presence and the absence of exogenous GABA, to homogenates of brain tissue pre- pared from the animals that had undergone behavioural testing. For this purpose we used samples from the hippo- eampal formation, since (as noted above) destruction of this region has been shown to abolish the PREE (Rawlins et al. 1980).

Methods

Experiment 1

Subjects. Seventy-two male Sprague-Dawley rats, experi- mentally naive and with a mean weight of 364 g, were ran- domly assigned to eight equal groups of nine rats each, composing a 4 x 2 factorial design made up of drug treat- ment (vehicle, or 1, 5 or 10 mg/kg RO 15-1788) and rein- forcement (CRF of PRF). They were caged in groups of four, one from each experimental condition going into two adjacent cages. During the 10 days preceding the start of training, all rats were gradually introduced to a 23-h food- deprivation schedule; water was freely available in the home cage at all times. Once behavioural training commenced,

the daily 1-h feeding period was scheduled to occur at least 1 h after the last rat was tested.

Apparatus. The aluminium straight alley was 163.7 cm long, 31.8 cm high, and divided by doors into a 20.4 x 20.8 cm startbox, a 122.4 x 15.6 cm run section, and a 20.7 x 20.5 cm goalbox. Start time (to the nearest centisec) was measured from the solenoid-operated opening of the startbox door till the animal crossed a visible-light photobeam located 12.5 cm from the startbox door; run time, from here to a second photobeam 94,5 cm further on; and goal time, from here to a third photobeam located just in front of the food-cup attached to the rear wall of the goalbox. The startbox door was opened 10 s after the rat had been placed in the startbox: the goalbox door was gently lowered after it had entered the goalbox.

Behaviouralprocedures. There were four consecutive stages. Handling lasted 14 days (groups of four for 10 rain/day). Pretraining lasted 3 days: on the 1st, groups of four were placed in the runway for 20 rain free exploration with ad fib food pellets in the goalbox; on the 2rid, the rats received the same treatment for 10 rain in groups of two; and on the 3rd each rat was run for two rewarded trials with the doors in operation. Acquisition then lasted 16 days at one trial per day. The reward consisted of 20 45-mg pellets placed before the trial in the food-cup. The CRF groups were rewarded on every trial; the PRF groups, on the sched- ule (R=reward, N=nonreward) R, N, R, N, R, N, R, N, R, N, N, R, R, N, N, R. On rewarded trials the rat was returned to its home cage as soon as it had consumed the reward; on nonrewarded trials, goalbox confinement time was 30 s. Extinction followed on the next day after acquisition and lasted 10 days at one trial a day, each identi- cal to the nonrewarded trials during acquisition. I f an ani- mai failed to reach the goalbox within 100 s it was removed from the alley. After two consecutive such "criterion" trials the animal was no longer run and given the notional score of 100 s for all sections of the alley on all remaining trials.

Drugs. During pretraining each rat was injected intraperito- neally (IP) with 1 ml/kg distilled water 10 rain before being placed in the alley. During acquisition each rat received an IP injection 10-15 min before the trial of either 1 ml/kg water plus Tween 40 (2 drops in 10 ml) or this vehicle con- taining 1, 5 or 10 mg/ml RO 15-1788 depending upon allo- cation to drug treatment. During extinction all rats were similarly injected, but only with the vehicle.

Statistics. All times were converted to speeds (one/s) and submitted to analyses of variance separately for start, run and goal sections and separately for acquisition and extinc- tion. The analyses were done with the Genstat program (Rothamsted Experimental Station), which permits com- parison of trends by the method of orthogonal polynomials. Reported t-tests are based on the appropriate error term from the analyses of variance.

Experiment 2

The eight groups of nine rats each (mean weight = 270 g) were randomly chosen so as to compose a 2 × 2 × 2 factorial design contrasting reinforcement schedule (CRF versus PRF), acquisition drug (5 mg/kg RO 15-1788 versus pla-

cebo) and extinction drug (5 mg/kg RO 15-1788 versus pla- cebo). All procedures were as in Experiment 1, except that the duration of acquisition was reduced to 10 days (one trial/day) so as not overly to prolong the period of drug administration in the animals given RO 15-1788 during both acquisition and extinction. The PRF schedule was: R, N, R, N, N, R, R, N, N, R. Extinction lasted 15 days, but because of the reduced variance the behavioural data from the final 5 days were not submitted to statistical analy- sis.

Biochemicalprocedures. The rats were killed by cervical dis- location on the day after the end of extinction. The brains were removed and the hippocampus dissected out on ice. The two hippocampi from each animal were pooled and homogenised 1:20 (w/v) in ice-cold 0.1 M Tris-citrate buffer pH 7.1, using a Potter-Elvejhem homogeniser. The homog- enate was centrifuged at 16000 g for 20 min at 4°C and the pellet was washed a further four times by re-homoge- nisation and centrifugation in the same volume of fresh ice-cold buffer. The membrane pellet was resuspended in the same volume of fresh buffer and dispersed using a Poly- tron homogeniser at setting 5 for 5 s immediately before use in the binding experiments. Aliquots of the crude mem- brane suspension (equivalent to about 75 gg protein) were incubated in a fmal volume of 1.0 ml with 0.5 nM [SH]- flunitrazepam (S. A. 79 Ci/nmol, Amersham International) for 60 min at 4 ° C in the presence and absence of 100 laM GABA and 150 mM sodium chloride. Non-specific binding was defined as that not displaceable by 3 laM clonazepam. At the end of the incubation period bound and free ligand were separated by the addition of 4 ml ice cold buffer and immediate vacuum filtration through Whatman GF/B filters; this was followed by two further 5 ml washes with the same buffer. Aquasol (3 ml, N. E. N.) was then added to the filters with subsequent counting by conventional liq- uid scintillation methods. Both total and non-specific bind- ing were determined in triplicate on each tissue sample. Analyses of variance were carried out separately on the results of the binding assays with and without added GABA, and also on the change in [3H]-flunitrazepam bind- ing induced by GABA (expressed as a percentage of the value obtained for each sample in the absence of GABA).

Results

Effect of RO 1~1788 on acquisition

Experiment 1. The vehicle-treated PRF group gained speed more slowly over days than the CRF group in both the run and goal sections. This effect was unchanged by 1 mg/ kg RO 15-1788, and only marginally altered by the 10 mg/ kg dose. However, the 5 mg/kg dose reversed the effect of reinforcement schedule, so that PRF-trained animals now gained speed more quickly over days than the CRF group.

In the run section the 5 mg/kg dose of RO 15-1788 reduced the rate of gain of speed over days in the CRF condition only. This effect reflected in particular an increase in speeds early in acquisition in the treated group. Neither the groups treated with 1 mg/kg nor those treated with 10 mg/kg differed from controls. Statistically, this pattern emerged as a significant linear component in the Drug x

2 .0

I . S

1 .0

0 ,S

Exp. 1

373

V e h i c l e CRF S =g/kg CRF S mg/kg PFIF

5 18 15

Oeys

2 . 0

1.S

1 .0

0 .S

Exp. 2

~ ~ ~ Whlcle CWF VehJcle PF~

~_--~Z 5 mg/kg CRF

~ ~ 5 mg/l<g PRF

i i i 1 i i i i i

q G 9

D e y s

Fig. l. Goal speeds during acquisition in Experiments 1 (top) and 2 (bottom) as a function of reinforcement schedule (CRF, continu- ous reinforcement or PRF, partial reinforcement) and administra- tion of vehicle or 5 mg/kg RO 1~1788

Days x Schedule interaction [F(3, 960)=7.97, P<0.001]. Examination of the coefficients of linear regression of run speeds upon days showed that CRF-trained animals gained speed significantly (P<0.01) faster than PRF-trained ani- mals in all drug conditions except 5 mg/kg RO 15-1788, in which PRF-trained animals gained speed significantly faster than the CRF group [t(960)=2.36, p<0.05]. This effect was due exclusively to a reduction in the slope of linear regression in the 5 mg/kg RO 15-1788 CRF condi- tion; this differed significantly from the slope observed in the vehicle-treated CRF controls [t(960)= 4.15, P < 0.001].

In the goal section the 5 mg/kg dose of RO 15-1788 again selectively decreased the slope of linear regression of speeds upon days in the CRF condition (Fig. 1, upper panel): linear component of the Drug x Days x Schedule in- teraction [F(3, 960)=7.54, P<0.001]; difference in linear regression coefficients between vehicle and 5 mg/kg groups [t(960)=2.88, P<0.01]. But, in addition, 5 mg/kg of the drug also increased the slope of linear regression in the PRF condition to an exactly equal degree (t=2.88). The latter effect was also produced by the 10 mg/kg dose of RO 15-1788 (t=2.40, P<0.05). In consequence of these changes, the rate of gain of speed was signifieantly ( P < 0.01) greater in the CRF than the PRF condition in the

374

.6

"io

o. co

¢-

t3 o

/ - / / , / / / / / /

/// /// /// /// /// /// ///

//i /// /// ///

i i I

1 S I@

mg / k~

Fig. 2. Effects of RO 15-1788 (i, 5 or 10 m~kg) given dung ac- quisition on mean run speeds during extinction in Experiment 1. Bar marker: 2 standard errors

0 , 8

0 ° 6 ell

~ B . 4

~ 0 . 2 + Vehicle

[ ] S mg/kg

Days

Fig. 3. Effects of 5 mg/kg RO 15-1788 given during acquisition on daily run speeds during extinction in Experiment 2

vehicle and 1 mg/kg groups, and still so (but only at P < 0.05) in the 10mg/kg groups; however, in the 5 mg/kg groups the PRF trained animals gained speeds over days significantly faster than the CRF animals (t=2.07, P < 0.05).

Experiment 2. The results obtained with the 5 mg/kg dose of RO 15-1788 in Experiment 2 partially confirmed those obtained in Experiment 1. The drug again reduced the rate of gain of speed over days in the run and goal sections. However, this effect differed in three respects from the pat- tern seen in Experiment 1. First, it did not interact to a statistically significant degree with Schedule, though inspec- tion of the relevant means (Fig. 1, lower panel) suggests that the influence or the drug was greater in the CRF condi- tion, as in Experiment 1. Second, the effect was due to re- duced speeds late in acquisition in the drugged groups, rather than increased speeds early in acquisition as in Ex- periment 1. Finally, there was no indication of the addition- al effect, seen in Experiment 1, of an increased rate of gain of goal speeds in PRF-trained animals ; however, this effect did not emerge in Experiment 1 until Day 9 of training (Fig. 1, upper panel), and training in Experiment 2 termin- ated on Day 10.

Statistically, these effects were confirmed by significant linear components in the Drug × Days interactions [F(1, 585)=6.78 and 13.29, p<0.01 and 0.001], for the run and goal sections, respectively. Asymptotically, the drug-treated rats were significantly slower than controls on the last two

days of acquisition, but only in the goal section. In addition, reinforcement schedule affected speeds (PRF slower), as shown by the significant interactions between Days and Schedule [F(9, 585)=4.18 and 2.21, P<0.001 and 0.05], for the run and goal sections, respectively. These effects of Schedule did not interact with Drug.

Effects upon extinction of RO 15-1788 given during acquisition

Experiment 1. There was a PREE (i.e. increased resistance to extinction of PRF-trained rats relative to CRF-trained animals), and this was unaffected by acquisition drug treat- ment. However, this treatment did affect overall extinction speeds in an inversely dose-dependent manner, the animals which had received 1 mg/kg RO 15-1788 being slower in both the start and run sections (Fig. 2). This effect may reflect the influence of change of drug state between acquisi- tion and extinction.

Statistically, the PREE was demonstrated by significant interactions between Schedule and Days IF(9, 648)=2.74 and 1.93, P<0.01 and 0.05, for run and goal, respectively]. The overall decrease of extinction speeds caused by RO 15-1788 was supported by main effects of Acquisition Drug [F(3, 64)=2.96 and 3.70, P<0.05], for the start and run sections, respectively. Application of t-tests showed that, in the run section (Fig. 2), the 1 mg/kg group were slower than the vehicle and 10 mg/kg groups [t(64) = 1.95 and 2.07, respectively, P=0.05]. The same pattern was seen in the start section, but only the 1 and 10 mg/kg groups differed significantly (t = 2.07, P = 0.05). These effects were not due to asymptotic differences at the end of acquisition, nor was there statistically reliable evidence that they changed during the course of extinction.

Experiment 2. The rate of extinction was changed by prior drug treatment in all three alley sections in essentially the same manner. As illustrated for the run section in Fig. 3, animals given 5 mg/kg RO 15-1788 during acquisition com- menced extinction with slower speeds but extinguished more slowly than controls, the two resulting curves converg- ing mid-way through extinction. These effects were fully significant in every alley section, as shown, e.g. for the data shown in Fig. 3 by both the main effect of Acquisition Drug [F(1, 64) = 5.60, P < 0.05], and the interaction between Drug and Days [F(9, 576)=2.53, P<0.01]. Only in the goal sec- tion is it likely that these effects reflect a continuation of differences already present at the end of acquisition.

Effects of RO 15-1788 given during extinction

The drug was administered during extinction only in Exper- iment 2. As shown in Fig. 4, the effects of this treatment varied as a function of (1) reinforcement schedule during training and (2) alley section. In the goal section, where the CRF and PRF groups differed within the vehicle condi- tion, there was no drug effect; whereas in the start section, where the CRF and PRF groups did not differ within the vehicle condition, the drug induced a difference (PRF- trained animals faster than CRF) of the kind observed in the vehicle groups in the goal section.

The drug effect in the start section was supported sta- tistically by the interaction between Extinction Drug and Schedule [F(1, 64)= 5.13, P < 0.05]. The t-tests showed that,

03

1 .5

1 .8

8 .5

START

R U N

CRF

5

PRF

,'%q

5

03

8 .6

8 .4

8 ,2

CRF

5

PRF

5

1 .8

8.B

8 .6

@.4

8 . 2

G O A L CRF

5 @

mg / kg

PRF

\NN N\'1 \ \ ' q

\ \ N \ \ N NNN \ \ N \ \ x t \ \ N

5

Fig. 4. Effects of 5 mg/kg RO 15-1788 given during extinction on mean start (above), run (middle) and goal (below) speeds during extinction after training on continuous (CRF) or partial (PR_F) reinforcement schedules in Experiment 3. Bar marker:2 standard errors

fll 0A

¢

375

2 . 0 -

1 . 5

1 . 8

g . 5

V e h i c l e in A c q u i s i t i o n

i i i i t ~ i i i i

3 6 9

Days

RO 5 mg/kg ,n A c q u i s i t i o n

2 . 8 (9 (D V e h , c l e CRF

~" - - "Q) V e h i c l e PRF

[] [] RO 5 mo CRF 1.5

E ] - - - ~ ] RO 5 mg PRF

, . 8

\%. 8.5 ~ 0.\ "J~, l;~ "'~_

i i i i i i t i L L

3 6 9

D a y s

Fig. 5. Goal speeds during extinction in Experiment 2 as a function of acquisition drug treatment (above, vehicle; below, 5 mg/kg RO 15-1788), extinction drug treatment (key) and acquisition rein- forcement schedule (key: CRF, continuous reinforcement; PRF, partial reinforcement)

whereas the vehicle animals did not differ as a function of reinforcement schedule, there was a significant difference between the drug-treated CRF and PRF groups [t(64)= 2.40, P<0.05] . This effect was uncomplicated by further interaction with Acquisition Drug (F< 1). A similar effect (Fig. 4) was present in the run section, but well short of significance; and absent altogether in the goal section, in which the difference between CRF and PRF conditions was significant within the vehicle-treated animals.

In addition to these schedule-dependent effects, Extinc- tion Drug affected start speeds in interaction with Days [F(9, 576)= 1.92, P<0.05] , in a manner that was indepen- dent of reinforcement schedule. This effect was in part due to the drug's decreasing speed on the first 2 days of extinc- tion and in part to its increasing speed later in extinction. This pattern gave rise also to a significant linear component in the Extinction Drug x Days interaction [F(1,

576)= 12.86, P < 0.001], reflecting the slower rate of extinc- tion in the animals drugged during extinction.

Interaction between acquisition and extinction drug treatments

There was evidence of the PREE in all three alley sections in Experiment 2. The PREE was unaffected by either acqui- sition or extinction drug treatment in the start or run sec- tions. However, in the goal sections (Fig. 5) the PREE was modified (although remaining substantial) by both acquisi- tion and extinction drug treatments separately but non- additively. Each of these treatments retarded the rate of extinction of the CRF-trained animals without affecting PRF animals. These drug-induced changes in the rate of extinction, however, largely reflect the fact that all three drug-treated CRF groups (i.e. those given RO 15-1788 only

376

Table l. Linear regression coefficients (C) for goal speeds on days during extinction in Experiment 2 as a function of drug treatment during acquisition (Acq) and extanction (Ext) and reinforcement schedule during acquisitmn. Values of t (with 576 dO in the right- hand cohmm are for comparisons of continuous (CRF) and partial (PRF) reinforcement groups m the same row; values of t in the other two columns are for comparisons of groups treated with 5mg/kg RO 15-1788 (RO) against the vehicle-vehicle control group in the same column

Acq Ext CRF PRF t Drug Drug

Vehicle Vehicle C 0.169 0.089 3.84** C 0.109 0.101 0.41 Vehicle RO t 0.286* 0.57 C 0.091 0.095 0.16 RO Vetucle t 3.72** 0.28 C 0.095 0.067 1.36 RO RO t 3.53 ** 1.06

*P<O.OI; **P<O.O01. All coefficients are negative (minus signs omitted

in acquisition, only in extinction, or during both phases of the experiment) commenced extinction with slower speeds than the vehicle-treated controls (Fig. 5).

The PREE emerged statistically in the start section as a linear component in the Days x Schedule interaction [F(1, 576) = 8.56]; and in the run and goal sections as the overall interaction between these variables [F(9, 576)=2.53 and 3.03, respectively; P<0.01 in each case]. In the goal section only there was a si~ificant linear component in the four- way interaction between Schedule, Days, Acquisition Drug and Extinction Drug [F(1,576)=6.51, P<0.01]. As shown in Fig. 5, this interaction arose because, given during either training or extinction, RO 15~1788 reduced the rate of ex- tinction in the CRF condition only, and these two effects were not additive. Comparison of the linear regression coef- ficients by t-tests (Table 1) confirmed this interpretation of the interaction. Only among the vehicle-treated groups was the slope of linear regression steeper in the CRF than the PRF condition; within the CRF condition, all three drug groups (drug during acquisition, drug during extinction, and drug during both acquisition and extinction) had signif- icantly shallower slopes of linear regression than the con- trols; and, within the PRF condition, none of the drug- treated groups differed significantly from the controls. Thus, as measured by the linear regression coefficients, ac- quisition and extinction drug treatments each separately and non-additively reduced the PREE by retarding the rate of extinction of the CRF-trained animals. Note, however, that the PREE remained substantial within all three drug conditions. Not only did the overall interaction between Days and Schedule fail to enter into interaction with either Acquisition Drug or Extinction Drug, but in addition there was a large quadratic component of this interaction [F(1, 576) = 9.27, P < 0.01], which also persisted uninfluenced by drug treatment (Fig. 5).

Biochemical results

There were no effects of Schedule, Acquisition Drug or Extinction Drug on the raw data obtained in the [3H]-fluni- trazepam bindings assays in Experiment 2, whether these

were carried out in the presence or the absence of GABA. However, the analysis of percentage change in binding in- duced by GABA revealed a significant effect of Acquisition Drug [F(1, 61)= 8.27, P < 0.01]. GABA increased [3H]-flun- itrazepam binding by 74% in the samples prepared from animals given placebo during training, but by 82% in those given RO 15-1788 (standard error, 2.72%). This effect was uninfluenced by drug treatment during extinction.

Discussion

The experiments reported here are the ftrst, to our knowl- edge, to have investigated the influence on the PREE of the benzodiazepine antagonist, RO 15-1788. There are no previous data, therefore, against which to judge the reliabili- ty of our results. Furthermore, although for the 5 mg/kg dose of RO 1 ~1788 given during acquisition (the only con- dition in common between the two experiments reported here) there were similar patterns of results in Experiments 1 and 2, there were also a number of differences, not all of which are likely to be due to the main procedural difference between the two experiments, viz, length of training (see the Results section on the effect of RO 15-1788 on acquisi- tion). Thus our results must be treated with caution. Not- withstanding this caveat, it is clear that the benzodiazepine antagonist, given over the dose range 1-10 mg/kg during acquisition, does not produce a major disruption in the PREE, as does the benzodiazepine agonist, chlordiazepox- ide (Feldon and Gray 1981a, b; McNaughton 1984). It is unlikely, therefore, that, assuming the existence of an endogenous ligand for the benzodiazepine receptor, an ac- tion of this ligand plays a critical role in the production of the PREE along the fines speculatively discussed in the Introduction.

Our results nonetheless show subtle but substantial ef- fects of RO 15-1788 upon food-rewarded alley-running. As already noted, these effects differ from those reported in the same task with the anxiolytic benzodiazepine, chlor- diazepoxide (Feldon and Gray 1981 a, b); nor are they sys- tematically opposite in sign to the latter, as might be ex- pected if the compound had anxiogenic-like effects (File and Pellow 1986). However, the pattern of change seen with RO 15-1788 resembles that produced by chlordiaz- epoxide in one important respect: several of the effects we observed were schedule dependent, that is, they differed depending upon the reinforcement schedule, continuous or partial, used in training.

Effects of this kind were observed both during acquisi- tion, when the different reinforcement schedules were actu- ally in force, and during extinction, when both CRF- and PRF-trained animals were treated identically. Thus, during acquisition in Experiment 1, 5mg/kg RO 15-1788 de- creased the rate of gain of both run and goal speeds in the CRF condition only; while both 5 and 10 mg/kg of the drug increased the rate of gain of goal speeds in the PRF condition only. One consequence of these changes was that the difference between the CRF and PRF groups, pres- ent in the controls, was reduced in the drug-treated animals. A second schedule-dependent effect which similarly reduced the differences between CRF- and PRF-trained animals was observed in the goal section during extinction in Experi- ment 2: RO 15-1788, whether given during acquisition or extinction, reduced the rate of loss of speed over days in the CRF condition only (Fig. 5). This pattern of change

377

could not have been due to state dependency, i.e. limitations of the expression of a learnt response to the drug state in which it was learned (Overton 1966), since the animals both trained and extinguished under the drug behaved identically to those drugged only in training or only in ex- tinction. Although different in detail from the effects of chlordiazepoxide (Feldon and Gray 1981 a, b), these change have in common with those seen after administration of the benzodiazepine that they reduce the difference between CRF- and PRF-trained animals. However, a third schedule- dependent effect of RO 15-1788, in contrast to the first two, enhanced the difference between CRF- and PRF- trained animals. This effect (Fig. 4) was observed in the start section in Experiment 2, in which administration of the drug during extinction produced a difference (PRF speeds faster) that was absent in the vehicle controls. Since the same kind of CRF-PRF difference was present in the controls in the goal, where it was unaffected by the drug, it seems likely that in the start section RO 15-1788 un- masked an effect of reinforcement schedule that was latent in the controls.

Although these schedule-dependent effects of RO 15-1788 do not appear to involve state dependency, some of our other observations may reflect this phenomenon. Thus, in both experiments extinction speeds were slower in animals that had been drugged during training only (Figs. 2, 3). This effect was not secondary to drug-induced differences present at the end of acquisition. Thus it was perhaps due to the change in drug state (from RO 15-1788 to placebo) between acquisition and extinction. The con- verse change, from placebo during acquisition to drug in extinction, also gave rise to slowed exit from the start box on the first two days of extinction in Experiment 2. Assum- ing that these effects indeed constitute state dependency, the results of Experiment 1 are somewhat surprising in showing an inverse dependence of the phenomenon upon dose (Fig. 2). Since in the same experiment, schedule-depen- dent effects emerged only at the two higher doses (5 and 10 mg/kg), it is unlikely that the state-dependent and sched- ule-dependent effects of RO 1 5-1788 are mediated by iden- tical mechanisms.

The most intriguing of the behavioural effects of the benzodiazepine antagonist is the enhanced influence of rein- forcement schedule seen when the drug was injected during extinction in Experiment 2 (Fig. 4). As noted above, the pattern of results obtained in the different alley sections suggests that the drug unmasked a schedule effect that was latent in the controls. Such an effect might conceivably be due to changes in the benzodiazepine receptor or in an endogenous ligand for this receptor, if one exists. How- ever, we found no effects of either Reinforcement Schedule or Extinction Drug on the binding of tritiated flunitraze- pam.

It should be noted that the design of our experiments does not permit the inference that our results necessarily depend upon the presence of the drug at the time that the animal is exposed to a particular training experience. To make such an inference it would have been necessary also to include groups receiving the same drug regime, but in- jected, e.g. after the daily training trial (Salmon and Gray 1985). Similarly, drug effects that were observed late in training or during extinction cannot unequivocably be at- tributed to an effect of the drug on the particular learning or behavioural state that holds at the time the drug effect

is observed; nor can they unequivocably be attributed to the duration of drug treatment up to that point; for these two variables length of training and length of drug treat- ment - were confounded in our experimental design (as in most similar studies). Deconfounding of these variables would require groups drugged for varying lengths of time before training commences.

The one significant effect disclosed by the biochemical assays consisted of enhanced GABA stimulation of [3H]- flunitrazepam binding in samples prepared from animals given 5 mg/kg RO 15-1788 during acquisition. Since this effect emerged only in the analysis of the percentage change measure, and not in the raw results of the assays carried out with or without GABA, respectively, it should be treated with caution. Available evidence suggests that GABA facilitation of [3H]-flunitrazepam binding provides information about the coupling between the benzodiazepine and GABA components of the overall receptor complex. It has been shown, for example, that chronic treatment with diazepam results in a decrease in the maximal facilita- tion of flunitrazepam binding caused by GABA with no change in the binding of flunitrazepam itself (the opposite of the effect we found after RO 15-1788 given during acqui- sition), together with a decreased post-synapfic sensitivity to GABA in vivo (Gallager et al. 1984). Assuming that the biochemical changes we observed are reliable, it is surpris- ing that an assay conducted at the end of the combined period of training and extinction should pick up an effect of the drug given during training, but no effect of the drug given during extinction; and also surprising that the latter treatment should not have modified the long-term biochem- ical effects of the former. In further experiments we have been unable to modify [3H]-flunitrazepam binding, or the effect on such binding of added GABA, by either 10 or 20 days of continuous administration of 5 mg/kg RO 15-1788 without behavioural treatment. Thus the effect of the benzodiazepine antagonist on GABA stimulation of [3H]-flunitrazepam binding observed in the present experi- ment may depend critically on certain aspects of the ani- mal's behavioural experience. There were a number of ef- fects of Acquisition Drug upon behaviour during both ac- quisition and extinction (see Results); but, without further data, it is not possible to relate any particular one of them to our biochemical observations.

These biochemical observations suggest that the changes in the GABA/benzodiazepine receptor complex wrought by the combination of behavioural and pharmacological treat- ments employed in our experiments are quite subtle, as are the behavioural changes that we also observed. Thas is as one would expect in dealing with behavioural processes as complex as those that undoubtedly underlie the PREE. But an understanding of these processes, and their biochemical basis, is an important objective. The PREE is in many ways an analogue of the type of behaviour therapy which is effec- tive in eliminating phobic symptoms in patients suffering from neurotic disorders (Gray et al. 1982). Many such pa- tients simultaneously receive behaviour therapy and anxio- lyric drug treatment. Interactions between these two forms of therapy may therefore have a serious influence on the outcome of treatment (Sartory 1983). Rational choice of therapy requires a thorough understanding of such interac- tions at the neurochemical level.

Overall, as in a number of other recent reports (File and Pellow 1986), the pattern of results revealed by these

378

experiments, the first in which a benzodiazepine antagonis t has been examined in the par t ia l reinforcement paradigm, is too complex to permit easy generalisation. However, the interactions we have observed between RO 15-1788 and reinforcement schedule, coupled with our previous reports that chlordiazepoxide given during training in this para- digm abolishes the PREE (Feldon and Gray 1981a, b), suggest that the neurochemical pa thways by which rein- forcement schedules modify behaviour involve a step or steps that are influenced by benzodiazepine receptors.

References

Boarder MR, Feldon J, Gray JA, Fillenz M. (1979) Effect of run- way training on rat brain tyrosine hydroxylase: differential ef- fect of continuous and partial reinforcement schedules. Neu- rosci Lett 15:211-215

Feldon J, Gray JA (1981 a) The partial reinforcement extinction effect after treatment with chlordiazepoxide. Psychopharma- cology 73 : 269-275

Feldon J, Gray JA (1981 b) The partial reinforcement extinction effect: influence of chlordiazepoxide in septal-lesioned rats. Psy- chopharmacology 74: 280-289

File SE, Pellow S (1986) Intrinsic actions of the benzodiazepine receptor antagonist RO 15-1788. Psychopharmacology 88:1-11

Gallager DW, Lakoski JM, Gousalves SF, Rauch SL (1984) Chronic benzodiazepine treatment decreases postsynaptic GABA sensitivity. Nature 308:74-77

Gray JA, McNaughton N, James DTD, Kelly PH (1975) Effect of minor tranqui/lisers on hippocampal theta rhythm mimicked by depletion of forebrain noradrenaline. Nature 258:424-425

Gray JA, Feldon J, Rawlins JNP, Owen S, McNaughton N (1978) The role of the septo-hippocampal system and its noradrenergic afferents in behavioural responses to nonreward. In: Elliott K, Whelan J (eds) Functions of the septo-hippocampal system. Ciba Foundation Symposium No. 58 (New Series), Elsevier, Amsterdam, pp 275-300

Gray JA, McNaughton N, Holt L, Tsaltas E, Feldon J, Shemer A (1982) The effects of anti-anxiety drugs on tolerance for stress.

In: Levy A, Spiegelstein MY (eds) Behavioural models and the analysis of drug action. Elsevier, Amsterdam, pp 175-194

Gray JA, Quintero A, Mellanby J, Buckland C, Fillenz M, Fung SC (1984) Some biochemical, behavioural and electrophysiolog- ical tests of the GABA hypothesis of anti-anxiety drug action. In: Bowery NG (ed) Actions and interactions of GABA and benzodiazepines. Raven Press, New York, pp 239-262

Guyenet PG, Aghajanian GK (1979) ACh, substance P and met- enkephalin in the locus coeruleus: pharmacologdcal evidence for independent sites of action. Eur J Pharmacol 53:319-328

Haefely W (1984) Actions and interactions of benzodiazepine ago- nists at GABAergic synapses. In: Bowery NG (ed) Actions and interactions of GABA and benzodiazepines. Raven Press, New York, pp 263-285

Haefely W, Kyburz E, Gerecke M, M6hler H (1985) Recent ad- vances in the molecular biology of benzodiazepine receptors and in the structure-activity relationships of their agouists and antagonists. In: Testa B (ed) Advances in drug research. Aca- demic Press, London, pp 165-322

Hunkeler M, Mohler H, Pierie L, Polc C, Bonnetti EP, Cumin R, Sehaffner R, Haefely W (1981) Selective antagonists of ben- zodiaTepines. Nature 290:514

McNaughton N (1984) Effects of anxiolytic drugs on the partial reinforcement effect in runway and Skinner box. Q J Exp Psy- chol 36 B:319-330

McNaughton N, James DTD, Stewart J, Gray JA, Valero I and Drewnowski A (1977) Septal driving of hippocampal theta rhythm as a function of frequency in the male rat: effects of drugs. Neuroscience 2:1019-1027

Overton DA (1966) State-dependent learning produced by depres- sant and atropine-like drugs. Psychopharmacologia 10:6-31

Owen S, Boarder M, Gray JA and Fillenz M (1982) Acquisition and extinction of continuously and partially reinforced running in rats with lesions of the dorsal noradrenergic bundle. Behav Brain Res 5:11-41

Quintero S, Mellanby J, Thompson MR, Nordeen H, Nutt D, McNaughton N, Gray JA (1985) Septal driving of hippocampal theta rhythm: role of 7-aminobutyrate-benzodiazepine receptor complex in mediating effects of anxiolyties. Neuroscience 16:875~84

Received October 10, 1986 / Final version September 23, 1987