Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Benzodiazepines (Bzs) are well-established, clinically useful anxiolytics. Some effects that are characteristic of these drugs, particularly sedation and muscle relaxation, have been considered unwanted side effects. Identification of the Bz modulatory site on the -aminobutyric acid type A (GABA_A) complex facilitated a search for compounds that relieve anxiety without such side effects. Development of Bz partial agonists has been one route toward this goal (Costa and Guidotti, 1996).

U-78875 [3-(5-cyclopropyl-1,2,4-oxadiazol-3-yl)-5-(1-methylethyl)imidazol(1,5-a)quinoxalin-4(5H)-o-ne] is one of a series of imidazoquinoxaline derivatives. The generic name assigned to U-78875 is panadiplon. These compounds bind Bz sites on GABA_A receptors with high affinity but vary in efficacy across in vitro functional assays. Based on in vitro data, Petke et al. (1992) concluded that U-78875 is an antagonist at the Bz site. Unlike diazepam (and unlike Bzs generally), U-78875 had little effect on GABA-induced ³⁶Cl uptake in rat cerebrocortical synaptoneurosomes and only minimally enhanced t-butylbicyclophosphorothionate binding at GABA_A receptors. Instead, U-78875 antagonized diazepam in such preparations more potently than the Bz antagonist flumazenil/RO 15-1788 (Petke et al., 1992).

In behavioral studies, however, U-78875 had effects similar to those of Bz agonists in procedures that traditionally predict anxiolytic efficacy; the potency was similar to that of diazepam (Tang et al., 1991). U-78875 was less effective than diazepam, though in behavioral assays that predict sedation (e.g., it produced only a small decrease in time on a rotarod). Like the in vitro profile, U-78875 dose dependently antagonized effects of Bzs in behavioral and other in vivo pharmacological assays (Tang et al., 1991, 1993).

Drug discrimination procedures are used to classify subjective drug effects as similar or dissimilar to those of the particular drug used to train the discrimination. The generalization profiles obtained in drug discrimination testing usually are related closely to the pharmacological specificity of the training drug. Tang and Franklin (1991) found unusual discriminative effects for U-78875 in that it produced full generalization in a group of rats trained to discriminate 1.0 mg/kg diazepam but not in a group trained to discriminate 10 mg/kg diazepam. U-78875 also potently antagonized the 10-mg/kg diazepam dose (no interactions with 1.0 mg/kg were reported). These results were unique because previous studies in which diazepam training dose had been manipulated (albeit not to the high dose they used) had not found qualitative differences in generalization to anxiolytics or sedatives (Shannon and Herling, 1983).

The present investigation studied U-78875 in animals trained to discriminate either a full agonist Bz, lorazepam or diazepam, or a barbiturate, pentobarbital. Lorazepam, diazepam, and pentobarbital all enhance GABA through binding sites on the GABA_A receptor complex. Animals trained to discriminate pentobarbital and most Bzs typically have shown indistinguishable generalization profiles in that a wide range of sedative-anxiolytics, including barbiturates and neuroactive steroids, occasion drug-appropriate responding (Ator and Griffiths, 1989a,b; Griffiths et al., 1992; Ator et al., 1993). However, this has not been true of animals trained to discriminate lorazepam. Baboons and rats trained to discriminate lorazepam have generalized reliably to full agonists at the Bz site, whether or not the test compounds are Bz-subtype selective, but they have not generalized reliably to other compounds that enhance GABA (e.g., barbiturates and neuroactive steroids; for review see Ator and Griffiths, 1997). Direct comparisons in our laboratory have differentiated the lorazepam from the diazepam training condition (Ator and Griffiths, 1989a; Ator et al., 1993).

Because partial agonists are, by definition, compounds that produce less than the maximal effect of an agonist in a particular assay, we predicted that U-78875 would not produce full generalization in the lorazepam training condition. Whether the effects of U-78875, being a partial Bz agonist, would be differentiated from agonist Bzs by not occasioning full drug-appropriate responding in animals trained to discriminate pentobarbital or diazepam was of particular interest. Also, partial agonists classically have been characterized by their ability to antagonize (i.e., reduce the effects of) a full agonist at the same binding site. U-78875 was studied in combination with the training drugs to determine whether U-78875 would interact differentially with the barbiturate and the Bzs. Given its binding profile, we predicted that U-78875 likely would antagonize both lorazepam and diazepam but not pentobarbital. U-78875 was studied under comparable training conditions in both rats and baboons to investigate cross-species generality after initial results in baboons showed that the ability of U-78875 to antagonize lorazepam was dependent upon whether U-78875 shared discriminative effects with lorazepam.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Rats

Subjects. Adult male Long-Evans Hooded rats (n = 34, Rattus norvegicus; Blue Spruce Farms stock from Harlan-Sprague-Dawley, Altamount, NY) were housed individually in plastic cages with continuous access to water. Most (n = 22) had served in drug discrimination studies before and had been tested with other GABAergic sedative/anxiolytic drugs; 12 rats were experimentally naive. Individualized rations of commercial rat diet were provided at approximately the same time each day, which was about 30 min after the experimental session. The rats were weighed before each experimental session, and their weights were maintained at 330 g ± 10 g (cf., Ator, 1991). Lights in the colony room were on a 12-h cycle (lights on at 6 AM).

Apparatus. Six identical, custom-made experimental chambers (27.7 × 30.3 × 53.2 cm high) were used (see Fig. 4 in Ator, 1991). In each, two stainless steel response levers (Gerbrands Corp., Arlington, MA) were mounted 13 cm apart and 5 cm above the floor. A 28-VDC cue light with a translucent, colored cap was centered 6 cm over each lever. Although cap colors differed across chambers, they were the same within chambers. A cast iron food cup, into which an electromechanical pellet dispenser delivered 45-mg food pellets ("precision" pellets; P. J. Noyes Co., Lancaster, NH), was centered on the wall opposite the levers, 4 cm above the floor. The chamber was enclosed in a custom-made sound-attenuating chamber. It was equipped with a ventilation fan and a speaker through which 18 dB white noise (for masking extraneous sounds while the rats were in the chambers) was delivered. Chambers were interfaced to computers; experimental conditions were programmed using state notation language (MedPC). An event recorder recorded lever operations and pellet deliveries when the cue lights were illuminated.

Training Session Procedures. Most of the rats had been trained before the present study to discriminate a training drug dose (D) from the no-drug (ND) condition using methods previous described (Ator, 1990). The D conditions were: 10 mg/kg pentobarbital given i.p. 15 min before the session, 1.0 mg/kg diazepam given i.p. 30 min before the session, and 1.0 mg/kg lorazepam given i.p. 60 min before the session. There were a total of 6 rats trained to discriminate pentobarbital, 12 trained to discriminate diazepam, and 16 trained to discriminate lorazepam. Of those, 6 diazepam-trained rats and 6 lorazepam-trained rats had no testing history with drugs other than the training drug when they were tested with U-78875.

Baboons

Adult male baboons [9 Papio anubis and 1 Papio cynocephalus (baboon GI), Primate Imports, New York, NY] were individually caged and had continuous access to water. The baboons had served in studies of i.v. drug self-administration before being trained in drug discrimination. For baboon AR, U-78875 was the first drug tested other than the training drug; for baboon SE it was the second. The other baboons had 6- to 9-year histories of tests with other drugs (Ator and Griffiths, 1997). Individualized rations of commercial monkey diet, a multivitamin, and a piece of fruit or a carrot were provided at about the same time each day, which was about 75 min after the experimental session. The baboons were weighed every 2 weeks under ketamine hydrochloride anesthesia (preceded by atropine sulfate to control salivation). Food rations were chosen to keep weights stable or slowly increasing (Ator, 1991). Body weights were: 40 kg for baboon AR, 24 to 30 kg for CE, 29 to 32 kg for JA, 29 to 35 kg for LO; 30 to 36 kg for ML, 27 to 38 kg for MS, 28 to 34 kg for RA and RF, and 26 to 27 kg for SE. Ranges were wider for some baboons because data were collected over longer periods of time. Lights were on a 12-h cycle (lights on at 6 AM; there also was natural light through laboratory windows).

Apparatus. A custom-made intelligence panel formed the rear wall of the stainless steel primate cage in which the baboon was housed (see Fig. 7 in Ator, 1991). The panel contained either two Lindsley operanda (Gerbrands Corp., Arlington, MA) or two custom-made stainless steel levers. The levers were approximately 15 cm apart and a cue light like those described above was mounted over each. A recessed food hopper, into which an electromechanical feeder delivered 1-g banana-flavored pellets (either P.J. Noyes, Lancaster, NH or BIO-Serv, Inc., Frenchtown, NJ, depending upon the baboon) was above and to the right of the two levers. The hopper was illuminated for the duration of feeder operation (100 ms). White noise and tones could be delivered through a speaker mounted on the back of the intelligence panel. A 5 × 5-cm white Plexiglas panel, which could be transilluminated, was mounted in the upper right quadrant of the panel. A solenoid-operated drinkometer (Kandota Instruments, Sauk Center, MN), used for oral drug delivery, was mounted 21 cm above the levers. The baboon had easy access to the levers, drinkometer, and food cup when the baboon sat on the cage bench facing the panel. The drinking water spout and the hopper into which the daily rations of food were delivered were at the other end of the bench. Moderate visual and sound attenuation was provided by an enclosure chamber. Baboons AR and SE were studied after the lab was renovated, however; so there was no sound attenuation and they could see baboons opposite them. Chambers were interfaced to computers as described for the rats. A cumulative recorder recorded lever operations and pellet deliveries when the cue lights were illuminated.

Training Session Procedures. Before the present study, the baboons were trained to discriminate D from ND using procedures like those used for the rats (Ator, 1990). The D conditions were 10 mg/kg pentobarbital given p.o. 45 min before the session for four baboons and 1.8 mg/kg lorazepam given p.o. 60 min before the session for six baboons. Training session procedures and session duration were like those for the rats with the following additions: white noise and illumination of the translucent panel began with the presession time-out and continued until the end of the session. When the session began, the lights over the levers were illuminated and a 3-s tone sounded. For the baboons, a response on either lever when the lights were on produced a 0.1-s tone. The time-out that began with pellet delivery was 6 s. The consecutive response requirement was 20 for 5 baboons (CE, GI, JA, ML, and SE). For the other five baboons the response requirements were: MS, 15; AR, 30; LO and RF, 35; and RA, 40). When the latter baboons were being trained (before the present study), they did not reliably show a criterion level performance at 20. Each baboon's own pattern of switching levers was studied to determine how to adjust the response requirement. When a value worked well for maintaining criterion performance, it was retained.

Training Criteria

For both species, tests were conducted if the following criteria were met in two consecutive training sessions (in either the D, ND order or the ND, D order): first, the required number of consecutive responses first must have been made on the correct lever without that sequence yet having occurred on the incorrect lever, and second, 95% of total responses must have been on the correct lever. If performance failed to meet both criteria in any one training session, then four consecutive training sessions (either ND, D, ND, D or D, ND, D, ND) in which both criteria were met (defined then as criterion level performance) were required before the next test. Response rate was not considered in determining whether performance was at criterion level; but if rate was lower than other ND rates, training continued until rate was stable.

Test Procedures

Test sessions were identical with training sessions for each species except that meeting the response requirement on either lever produced a food pellet. The order of D and ND sessions between tests was counterbalanced so that test sessions were preceded equally often by D and ND training sessions. To determine whether performance was under control by the training drug per se (and not the dosing procedure), tests with D and its vehicle were given at the beginning of the study and again before study of drug combinations began. The fact that criterion level performance was shown reliably in tests preceded by vehicle for both rats and baboons confirmed that performance was not based on whether the i.p. or p.o. dosing procedure, respectively, had occurred. If good control was not demonstrated in consecutive D and vehicle tests, which rarely occurred, four D and ND training sessions were conducted as described above, and D and vehicle both were retested.

A single-subject design was used in which each animal served as his own control and replications across animals were used to determine generality (Sidman, 1960). Full dose-effect curves generally were determined for each animal except as noted below. Study of drug combinations was planned to be able to determine for each animal individually whether the agonist curve would be shifted to the right or left by U-78875, but illness in rats resulted in inability to complete such curves for some animals. Each dose or combination generally was administered once. For better characterization of discriminative thresholds in individual animals, doses sometimes were repeated; those exceptions are noted under Results.

Rats. U-78875 was given 30 min before the test session and the presession time-out corresponded to the pretreatment time. For U-78875 in combination with a training drug, each drug was given at its usual pretreatment time, and the rat was placed in the chamber after the second injection.

Baboons. U-78875 was given 60 min before the test session, and the presession time-out corresponded to the pretreatment time. Training sessions were not conducted for 2 or 3 days after doses greater than 3.2 mg/kg U-78875 were given because we suspected carryover of drug effects. Full dose-effect curves were determined for each baboon. For baboons that generalized to U-78875, the Bz antagonist flumazenil was given either p.o. or i.m. (into a thigh muscle) just before the U-78875 to determine whether it would block U-78875's discriminative effects. The drug time course was studied on most days on which U-78875 or vehicle was given, because previous studies showed orderly dose-dependent effects under this procedure (Ator and Griffiths, 1985; Ator, 1990; Sannerud et al., 1992). After the initial 20-min test session an hour after dosing, 10-min test sessions occurred 3, 5, 7, 9, 11, 13, and 15 h after dosing, each preceded by a 5-min time-out. The ration of monkey chow was omitted on those days.

Drugs

U-78875 was donated by The Upjohn Co. (Kalamazoo, MI). Pentobarbital sodium was purchased from Sigma Chemical Co. (St. Louis, MO), and doses are expressed as the salt. Diazepam and flumazenil were donated by Hoffmann-LaRoche, Inc. (Nutley, NJ). Lorazepam was donated by Wyeth-Ayerst Laboratories (Princeton, NJ).

Rats. Drugs were administered i.p. in a volume of 1 ml/kg, except that 10 and 18 mg/kg U-78875 were in a volume of 2 ml/kg due to solubility limitations. Diazepam and lorazepam were prepared as stock solutions that did not contain water and were maintained for up to 30 days (diazepam in 80:20 propylene glycol/95% v/v ethanol; lorazepam in 80:20 propylene glycol/polyethylene glycol 400). These stocks were diluted for injection by 50% with sterile water for diazepam and 0.9% saline for lorazepam, and were maintained for up to 7 days. Pentobarbital was dissolved in 0.9% saline. U-78875 was stirred and then sonicated for approximately 20 min in a 1% carboxymethylcellulose solution. Both drugs were prepared within the hour before injection.

Baboons. Drugs were administered p.o. in 60 ml of a 1 g/liter BIO-Serv Agent K matrix that was flavored with orange-drink powder, except that flumazenil was given i.m. in the same vehicle described above for diazepam. For all drugs given orally, the dose was prepared in an electric blender before the session (the matrix itself was maintained under refrigeration for up to 3 days). A dose was delivered through the drinkometer spout, followed immediately by 40 ml of vehicle to flush the line (which also may have reinforced the baboon's consuming the dose). For p.o. vehicle tests, quinine sulfate (0.32 mg/ml) was included as a taste control.

Data Analysis

The primary dependent variables were percentage of total test session responses that occurred on the D lever and the rate of responding. Consistent with the drug discrimination literature, a conservative criterion (given the percentage used in the training criterion) was adopted for judging that test drug effects were not qualitatively different from D: Full generalization was concluded if percentage of D-lever responding was 80%, as long as the response requirement had been completed at least once (i.e., obtaining at least one pellet defined having made a lever choice). Conversely, 20% D-lever responding was not considered to be qualitatively different from ND. Given the 95% accuracy required in the criterion level training sessions, the 80% criterion can be considered significantly different from chance in a two-choice discrimination (Sidman, 1980). Intermediate percentages (21-79% D-lever responding) were interpreted as partial generalization akin to a psychophysical threshold for detection of a drug effect like the training stimulus (Ator, 1990). A log₁₀ scale was used on the abscissae for the dose-effect curves. An ED₅₀ was determined by interpolation (Barry, 1974) to be the nearest quarter of a log₁₀-unit dose at which the generalization gradient reached or crossed 50%. An ED₈₀ was determined in the same manner with respect to the dose at which the generalization gradient reached or crossed 80% (the criterion for sharing discriminative effects with the training drug). The rate of responding on both levers combined, excluding time-outs, was converted to percentages of the mean rate in ND control sessions (i.e., the ND training session that most closely preceded each test). Significant differences (p < .05) between ED₅₀s (and ED₈₀s) for training drug generalization gradients in the absence of U-78875 compared with gradients generated in the presence of U-78875 were determined via two-tailed t test for paired data.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Rats

Control Performance. Criterion level performance reliably occurred in training sessions, regardless of training drug. In test sessions with the training drug dose, the percentage of D-appropriate responding was 98 to 100%. In test sessions with vehicle, the percentage of D-appropriate responding was 0 to 2%, despite the fact that vehicle injections did not routinely precede ND training sessions. For most of the 18 rats in the pentobarbital and diazepam training groups, the mean response rates in D and/or ND training sessions were 1.5 to 2.5 responses/s, and the range of rates in D sessions overlapped the range of rates in ND sessions. For the five rats for which response rate in one type of session was reliably higher than the other, four responded faster in D than in ND sessions. For most (n = 11) of the 16 lorazepam-trained rats, however, response rates were lower in D than in ND sessions (for most rats, mean D rate was 1.0 response/s and mean ND rate was 1.5 responses/s).

U-78875 Generalization. The mean U-78875 generalization gradients for the pentobarbital- and diazepam-trained rats generally were a monotonically increasing function of dose (Fig. 1). The ED₅₀ for U-78875 was 1.8 mg/kg for the pentobarbital-trained rats and 0.056 to 0.1 mg/kg for the two groups of diazepam-trained rats. The ED₈₀ was 10 mg/kg for the pentobarbital-trained rats and 0.18 to 0.32 mg/kg for the diazepam-trained rats. Thus, U-78875 was at least 10 times more potent in diazepam- than pentobarbital-trained rats. On an individual-subject level, all but 1 of the 6 pentobarbital-trained rats generalized to U-78875, and all 12 diazepam-trained rats did so.

View larger version (25K): [in this window] [in a new window]   Fig. 1.   Generalization to U-78875 in rats trained to discriminate pentobarbital (n = 6), diazepam (n = 12), or lorazepam (n = 10) from no-drug (ND) condition. Upper panels, percentages of responding on lever paired with training drug in test sessions preceded by injections of U-78875 or its vehicle (V). Lower panels, rates of lever pressing on both levers combined in those same test sessions as percentages of mean response rate in control ND training sessions. For lorazepam and diazepam training conditions, triangles represent data from groups of rats that did not have a history of being tested with other drugs (in these rats, U-78875 injections were preceded by injections of training drug V in preparation for interaction studies). Each open symbol generally represents a single test session for a single rat. Open triangles for diazepam 0.01 to 0.1 doses are means of two or three determinations for three of six rats. Closed symbols give the group mean. Mean control ND responses/s (+ range) were: pentobarbital, 1.4 (0.5-2.1); diazepam circles, 2.9 (1.8-4.5); diazepam triangles, 1.6 (0.9-2.0); lorazepam circles, 1.4 (0.5-2.5); lorazepam triangles, 1.7 (0.6-3.5). Six other lorazepam-trained rats tested only with 10 mg/kg are not shown (see text): those percentages were 0 for three rats and 99%, 100% and 53% for the others (mean = 42%).

Interactions with Pentobarbital. In the pentobarbital-trained rats, pentobarbital itself produced a generalization gradient with an ED₅₀ of 5.6 mg/kg. U-78875 shifted the pentobarbital generalization gradient to the left (Fig. 2). The pentobarbital ED₅₀ decreased to <1.0 mg/kg in the presence of both U-78875 0.1 mg/kg [t(5) = 3.1, p = .027] and .32 mg/kg [t(5) = 15.3, p = .0001]. A lower U-78875 dose (0.032 mg/kg) had no effect in combination with 1.0 mg/kg pentobarbital (Fig. 2). Unfortunately, 0.032 mg/kg U-78875 was not tested in combination with 3.2 mg/kg pentobarbital to complete that curve. At an individual level, the percentage of pentobarbital-appropriate responding was no greater than 3% for any rat at 1.0 or 3.2 mg/kg pentobarbital nor at U-78875 doses of 0.32 mg/kg or less. Yet all six rats showed the superadditive effect of responding 100% on the pentobarbital-appropriate lever in one or both combinations with 0.32 mg/kg U-78875, and four of the six rats did so in one or both combinations with 0.1 mg/kg U-78875.

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Interactions of U-78875 with pentobarbital, diazepam, and lorazepam in rats trained to discriminate those drugs from no-drug (ND) condition. Upper panels, group mean percentages of responding on training drug lever after an injection of U-78875 or vehicle (V) followed by an injection of training drug or its V; lower panels, response rates as in Fig. 1. Each rat was tested with a particular combination once, except that for diazepam group, about half the tests with U-78875 0.01 or 0.032 mg/kg were repeated up to three times. Data for U-78875 plus the training drug V are the same as those in Fig. 1. Points with pentobarbital represent six rats. Points with diazepam generally include 4 to 11 rats, and points with lorazepam generally include 4 to 9 rats. For lorazepam, the points connected by dashed lines are for six rats tested only with 3.2 mg/kg lorazepam, 10.0 mg/kg U-78875, and the combination. Mean control ND responses/s (range) for each training group were: pentobarbital 1.3 (0.4-2.1); diazepam 2.0 (0.9-3.6); lorazepam 1.7 (0.6-3.5).

Interactions with Diazepam. In the diazepam-trained rats, diazepam itself produced a generalization gradient with an ED₅₀ of 0.78 mg/kg. Because the U-78875 generalization gradient for diazepam-trained rats (Fig. 1) was broad and the diazepam gradient itself (Fig. 2) was broad, only doses of both drugs that were much lower than their ED₅₀s for discriminative effects could be used to study potentiation of diazepam's discriminative stimulus effects. When U-78875 doses that did not occasion D-lever responding in most rats were given in combination with diazepam doses that did not occasion >20% D-lever responding in most rats (0.1 mg/kg), no effect greater than that produced by each dose alone was found (Fig. 2). Intermediate doses of U-78875 (0.1 and 0.32 mg/kg), which generally produced intermediate to high percentages of diazepam-appropriate responding by themselves, also did so in combination with an intermediate diazepam dose (Fig. 2). Thus, potentiation of diazepam's discriminative effects was not found. Alternatively, there was the possibility that U-78875, as a partial Bz-site agonist, would antagonize diazepam. In fact, the intermediate U-78875 dose, 0.1 mg/kg, did decrease diazepam-appropriate responding at 1.0 mg/kg from 100% to 70%. Higher U-78875 doses produced no effect different from when they were given alone (Fig. 2).

Interactions with Lorazepam. In the lorazepam-trained rats, lorazepam itself produced a generalization gradient with an ED₅₀ of 0.18 mg/kg (Fig. 2). No potentiation of lorazepam by U-78875 was found. Instead, U-78875 shifted the lorazepam gradient to the right, antagonizing the discriminative stimulus effects of lorazepam. Surmountability of the U-78875 antagonism by lorazepam was demonstrated in the dose-effect curves for 0.1 and 3.2 mg/kg U-78875 (Fig. 2). Due to personnel changes in the lab, we were not able to study surmountability of the 10 mg/kg U-78875 antagonism by 3.2 mg/kg lorazepam in that same group of rats. This interaction (as well as each dose in combination with the vehicle of the other drug) was studied in another six lorazepam-trained rats, but surmountability was not achieved (Fig. 2). For the U-78875/lorazepam dose combinations at which group curves did show surmountability (0.1 and 3.2 mg/kg U-78875), t tests were not significant at p < .05. For combinations with U-78875 0.32 and 1.0 mg/kg, evaluation of ED₅₀s would have been compromised, because the group curves did not reach the maximum for concluding generalization.

View larger version (37K): [in this window] [in a new window]   Fig. 3.   Interactions of U-78875 with lorazepam as a function of whether or not the lorazepam-trained rat generalized to U-78875. Data for 10 lorazepam-trained rats that made <80% total responses on lorazepam-appropriate lever after all doses of U-78875 are shown on left, and data for 5 rats that made >80% of total responses on drug lever after at least one dose of U-78875 are on right. Points connected by solid lines generally include four to six rats for curves on left and two or three rats for curves on right. Points connected by dashed lines are data from rats tested only with 3.2 mg/kg lorazepam, 10.0 mg/kg U-78875, and their combination (see text). Mean control ND responses/s (range) for group on left was 1.6 (0.7-2.5) and 2.0 (0.6-3.5) for group on the right.

Effects of U-78875 on Response Rates. U-78875 substantially increased rates of responding for some rats, and there were some decreases to 50% of the control rate or, rarely, below (Fig. 1, bottom panels). On the average, however, U-78875 did not affect response rates. In the interaction studies, U-78875 seemed able to antagonize the effects the training drugs had on rate (individual data not shown; Figs. 2 and 3, bottom panels).

Baboons

Control Performance. Criterion level performance occurred reliably in training sessions. In test sessions with the training drug dose, D-lever responding was 94 to 100%. In the test sessions with vehicle, D-lever responding was 0%, despite the fact that ND sessions were not preceded by the oral dosing procedure. For the pentobarbital-trained baboons, mean response rates in control ND training sessions were 1.2 to 2.2 responses/s; for the lorazepam-trained baboons those rates were 0.8 to 3.2 responses/s. Response rates in D training sessions tended to be within the same range as ND rates for both pentobarbital- and lorazepam-trained baboons, except that lorazepam-trained baboon LO responded about twice as fast in D as ND training sessions.

U-78875 Generalization and Flumazenil Antagonism. As in the lorazepam-trained rats, U-78875 produced dose-dependent generalization in a minority (n = 2) of the six lorazepam-trained baboons (Fig. 4). Unlike the pentobarbital-trained rats, generalization to U-78875 occurred in a minority (n = 1) of the four pentobarbital-trained baboons (Fig. 4). The ED₅₀s for the individual baboons that generalized to U-78875 ranged from 0.56 to 5.6 mg/kg. One baboon (SE) showed partial generalization at 10 mg/kg, but could not be induced to consume the full amount of higher doses. For the other pentobarbital- and lorazepam-trained baboons, D-lever responding was 0% even at doses as high as 32 and 56 mg/kg. The highest doses tested in the baboons represented from 0.5 to 2.25 total grams of U-78875 (see body weights above). Higher doses were not given because of practical considerations in administering them. Response rates were not decreased by U-78875 compared with vehicle (Fig. 4). Response rates after U-78875 were higher than after vehicle for two baboons that generalized, but otherwise were unaffected, even by very high doses.

View larger version (31K): [in this window] [in a new window]   Fig. 4.   Generalization to U-78875 in baboons trained to discriminate pentobarbital (n = 4) or lorazepam (n = 6) from the no-drug (ND) condition. Upper panels, percentages of responding on lever paired with training drug after U-78875 or its vehicle (V); lower panels, response rates as in Fig. 1. Each symbol represents a single test session for a single baboon. Mean control ND responses/s for each baboon were: LO, 0.8; GI, 1.2; SE, 1.4; MS, 1.7; AR, CE, and JA, 2.0; RF, 2.6; ML, 2.8; RA, 3.2.

Time Course of U-78875 Discriminative Stimulus Effects. Time course was studied with each dose of U-78875 in all baboons except baboon SE. Baboons that did not generalize to U-78875 in the test session 1 h after dosing did not generalize in any subsequent tests in the 15 h after the dose had been given (data not shown). The three baboons that did generalize to U-78875 60 min after dosing showed an offset of discriminative stimulus effects in a generally time-dependent fashion (Fig. 5). However, at one dose for all three baboons, percentage of D-lever responding went from >50% in one session to zero in the next session, and increased to 50% or more again in the next test session. This would be unremarkable except that it occurred for all three baboons and may reflect some aspect of metabolism of this compound. Interestingly, the duration of discriminative stimulus effects was not an increasing function of dose (i.e., higher doses showed shorter durations of effect than did lower doses for both LO and JA; see Fig. 5).

View larger version (18K): [in this window] [in a new window]   Fig. 5.   Time course of discriminative stimulus effects of U-78875 compared with vehicle (V) in two lorazepam- and one pentobarbital-trained baboons in which U-78875 occasioned drug-appropriate responding. Data are for 20-min tests 1 h after dosing and consecutive 10-min test sessions 3 to 15 h after dosing; each point represents a single session. Data are missing for baboon JA for some test sessions >7 h after dosing because he did not respond.

Interactions with Pentobarbital. Pentobarbital itself had an ED₅₀ of 3.2 to 7.8 mg/kg across baboons (Fig. 6). As in the rats, doses of U-78875 that did not occasion D-lever responding when given alone shifted the pentobarbital gradient to the left, potentiating the discriminative stimulus effects of pentobarbital. U-78875, 0.32 mg/kg, combined with pentobarbital, reduced the pentobarbital ED₅₀ to 1.8 or 0.56 mg/kg in all three baboons. Even greater shifts were produced by higher U-78875 doses in baboons CE and SE. As would have been expected for baboon JA, higher doses of U-78875, which on their own occasioned between 40 and 80% D-lever responding, also shifted the pentobarbital gradient to the left (data not shown).

View larger version (22K): [in this window] [in a new window]   Fig. 6.   Potentiation of pentobarbital by U-78875 in pentobarbital-trained baboons. Upper panels, percentages of responding on pentobarbital-paired lever after combinations of pentobarbital or its vehicle (V) with doses of U-78875 that did not occasion >20% pentobarbital-lever responding when given alone; lower panels, response rates as in Fig. 1. Each point represents a single test. However, each combination of pentobarbital +V represents two to four tests for baboon JA; for SE, those of pentobarbital >1.8 mg/kg + V and pentobarbital V + 3.2 mg/kg U-78875 represent two tests. See text for combinations of 10 mg/kg pentobarbital with higher U-78875 doses. Mean control ND response rate was 1.7 responses/s for each baboon.

Interactions with Lorazepam. Baboons ML and JA both generalized to U-78875. U-78875 doses that did not occasion D-lever responding when given alone potentiated lorazepam, shifting lorazepam ED₅₀s from 0.18 mg/kg (ML) or 0.32 mg/kg (JA) to <0.1 mg/kg (Fig. 7). Furthermore, no antagonism of lorazepam's discriminative effect was produced in these baboons by giving U-78875 in combination with lorazepam (Fig. 7), nor was there antagonism by 3.2 mg/kg U-78875, tested in combination with 1.0 mg/kg lorazepam in baboon JA (data not shown). Potentiation was not demonstrated in the third baboon (LO) that generalized to U-78875, but U-78875 (0.32 and 1.0 mg/kg) did not antagonize 1.0 or 1.8 mg/kg lorazepam for baboon LO either.

View larger version (23K): [in this window] [in a new window]   Fig. 7.   Potentiation of lorazepam by U-78875 in one lorazepam-trained baboon (ML) and one pentobarbital-trained baboon (JA) that generalized to U-78875. Upper panels, percentage of total responding on lever paired with training drug after combinations of lorazepam with vehicle (V) and with doses of U-78875 that did not occasion >20% drug-lever responding when given alone; lower panels, response rates as in Fig. 1. Mean control ND rate was 2.8 responses/s for baboon ML and 1.2 responses/s for JA.

View larger version (16K): [in this window] [in a new window]   Fig. 8.   Antagonism of discriminative stimulus effects of 1.8 or 3.2 mg/kg lorazepam by U-78875 in three lorazepam-trained baboons (MS, RF, and RA) and two pentobarbital-trained baboons (CE and SE) that had not generalized to U-78875 given alone. Upper panels, percentage of total test session responding on training drug-appropriate lever after a lorazepam (LRZ) dose given alone and that dose in combination with U-78875; lower panels, response rates as in Fig. 1. Mean control ND response rates (responses/s) were: MS and SE, 1.7; RA, 3.2; RF, 2.6; and CE, 2.2. Each point indicates results of a single test session, but some for CE and SE are means of two tests, where vertical bars indicate range. See text for interactions of U-78875 with lower lorazepam doses.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

U-78875 produced differential generalization across training conditions and across species. Rats trained to discriminate diazepam or pentobarbital generalized to U-78875, and U-78875 was at least 10 times more potent in diazepam- than in pentobarbital-trained rats. Yet, most rats trained to discriminate lorazepam did not generalize to U-78875, even though the tested doses were more than one and two log₁₀ units higher than the ED₅₀s for the rats trained to discriminate pentobarbital and diazepam, respectively. Differential generalization to U-78875 in diazepam- versus lorazepam-trained rats was not a function of testing history with other compounds: The results were the same for groups of rats newly trained to discriminate diazepam and lorazepam. The failure of lorazepam-trained rats to generalize to a test compound while diazepam- and pentobarbital-trained rats do has also occurred in previous studies that tested barbiturates, neuroactive steroids, and the Bz partial agonist bretazenil (Ator and Griffiths, 1989a,b; Ator et al., 1993, 1995). Rats trained under these training conditions have generalized to compounds that have been characterized as showing full efficacy at the Bz binding site (Ator and Griffiths, 1986; 1989a,b).

The lack of generalization to U-78875 by lorazepam-trained rats was replicated in lorazepam-trained baboons, but full generalization to U-78875 by pentobarbital-trained rats was not replicated in pentobarbital-trained baboons. Although these results could lead one to question the bioavailability of U-78875 in baboons, the interactions of U-78875 with lorazepam and pentobarbital confirmed that oral U-78875 is behaviorally active in baboons. As for rats trained to discriminate lorazepam, baboons trained to discriminate lorazepam have generalized reliably only to compounds otherwise characterized as full agonists in modulating GABA (reviewed in Ator and Griffiths, 1997). Thus, the present results with U-78875 in lorazepam-trained baboons are consistent with that generalization profile.

On the other hand, pentobarbital-trained animals, regardless of species, have generalized to a wide range of sedative/anxiolytic drugs, including Bz-site ligands and compounds not known to be GABAergic (Ator and Griffiths, 1989a; Griffiths et al., 1992). Lack of full generalization to a Bz-site ligand in pentobarbital-trained baboons was unique in our experience with GABAergic sedative-anxiolytics. Recently, Rowlett and Woolverton (1998) reported a comparable finding for bretazenil in rhesus monkeys trained to discriminate pentobarbital under a shock-avoidance procedure. They concluded that lower efficacy at the Bz site may be the variable that differentiated the discriminative effects of bretazenil from barbiturates. However, we have found that pentobarbital-trained rats did generalize to bretazenil (Ator et al., 1995). It may be that lower efficacy at the Bz site is more likely to be functionally relevant for stimulus effects in primates than in rats.

U-78875 was similarly effective in both baboons and rats in potentiating the pentobarbital discriminative stimulus. U-78875 and pentobarbital doses that alone did not have discriminative effects, together occasioned 50 to 100% pentobarbital-appropriate responding. This behavioral result is consistent with pentobarbital and U-78875 both enhancing GABA through different allosteric sites on the GABA_A receptor complex, and could reflect positive cooperativity (Leff, 1987), or it may be that pentobarbital enhanced U-78875 binding. Pentobarbital has been reported to enhance the binding of some Bzs and flumazenil (Skolnick et al., 1981; Miller et al., 1988; Carlson et al., 1992). Thus, despite the minimal potentiation of GABA by U-78875 found by Petke et al. (1992), occupation of the Bz site by U-78875 concomitant with occupation of the barbiturate site by pentobarbital can have functional relevance. Given the strong potentiation of pentobarbital's discriminative stimulus effects, it is worth noting that U-78875 did not also potentiate rate-decreasing or toxic effects of pentobarbital, even at high dose combinations, which may be favorable for the clinical usefulness of such compounds.

Explaining the interaction of the two Bzs with U-78875 seems more complex than could have been predicted, given that diazepam and lorazepam both have well-established in vivo and in vitro profiles as full agonists and that U-78875's profile, as described in the Introduction, seems consistent with that of a partial Bz agonist. Classic receptor theory would predict that a partial agonist would produce less than the maximal effect at a binding site and that when combined with a full agonist would reduce the effect of the full agonist alone (Kenakin, 1993). The group data for lorazepam-trained rats and baboons is consistent with both predictions. It is at the individual subject level that the behavioral data do not always match the predictions from receptor theory. For some animals, U-78875 occasioned lorazepam-appropriate responding and thus functioned as an agonist. In those animals, U-78875 did not act as an antagonist; and in the baboons, it produced a superadditive effect in combination with lorazepam. The best predictor of whether U-78875 would act as an antagonist in combination with lorazepam for a particular animal was not the group results but whether or not it had acted as a full agonist when given alone to that animal. If the results from the lorazepam-trained rats that generalized to U-78875 and the results from the lorazepam-trained baboons that generalized to U-78875 were collected in isolation from each other, one might be inclined to dismiss them. The fact that the data from these animals support each other suggests that the phenomenon has biological generality. Importantly, after the study of U-78875 began, we began study of the putative Bz partial agonist bretazenil. As with U-78875, bretazenil potentiated the discriminative stimulus effects of lorazepam in baboons that generalized to it and antagonized lorazepam in baboons that did not (Ator et al., 1995).

U-78875 neither potentiated nor antagonized diazepam. These results are only surprising in the context that diazepam has been almost uniformly considered as a prototypic full Bz agonist, including in drug discrimination studies (cf. Shannon and Herling, 1983). The generalization from diazepam to U-78875 in the present study taken together with the interaction data suggests similar efficacy of diazepam and U-78875 in diazepam-trained rats, which suggests that the 1.0 mg/kg diazepam training stimulus was a low efficacy Bz stimulus. The present results and this interpretation support those of Tang and Franklin (1991) for rats trained to discriminate 1.0 mg/kg diazepam compared with rats trained to discriminate 10 mg/kg diazepam. The former training group showed full generalization to U-78875 but the latter did not. Tang and Franklin (1991) did not report interactions of U-78875 with diazepam in rats trained to discriminate 1.0 mg/kg diazepam. They did show, however, that U-78875 antagonized diazepam in rats trained to discriminate 10 mg/kg diazepam. Thus, similar to the data for lorazepam-trained animals discussed above, the best predictor of whether U-78875 serves as a diazepam antagonist in these studies taken together is whether or not it occasioned diazepam-appropriate responding.

Because Tang and Franklin (1991) found differential generalization to U-78875 as a function of diazepam training dose, the possibility that the selectivity of the lorazepam training condition might also be dose dependent must be considered. That 1.0 mg/kg lorazepam might be a higher efficacy training stimulus than diazepam is suggested by the fact that response rates in D training sessions for rats were generally lower than rates in ND sessions for lorazepam-trained rats, but not for pentobarbital- and diazepam-trained rats. However, response rates were not lower in lorazepam training sessions compared with ND sessions for the baboons. Another piece of data that suggests, but does not prove, that the 1.0 mg/kg lorazepam and the 1.0 mg/kg diazepam training conditions have similar efficacy is that the ED₅₀s for diazepam in rats trained to discriminate 1.0 mg/kg lorazepam and for lorazepam in rats trained to discriminate 1.0 mg/kg diazepam were very similar (1.0 mg/kg and 0.56 mg/kg, respectively; Ator and Griffiths, 1989b).

U-78875 is one in a series of novel quinoxalinones screened for its therapeutic potential as a partial agonist (Tang et al., 1991). Other compounds in this series bound the Bz site and showed in vitro agonistic activity predictive of usefulness as an anxiolytic but also interacted in a concentration-dependent manner with a second, low-affinity, flumazenil-insensitive, site on GABA_A receptors (Im et al., 1995, 1996). Potentiation of GABA-mediated chloride current was reversed as the concentration increased. The possibility of such an interaction was not reported for U-78875; however, its complex behavioral profile may reflect more complex receptor interactions, including concentration-dependent activity at multiple GABA_A subtypes, than were initially presumed. Although the discriminative effect of U-78875 in baboons was antagonized by flumazenil, which is consistent with flumazenil antagonism of U-78875's effect on electroencephalogram recordings in rats (Tang et al., 1991), orderly dose-effect functions for flumazenil in combination with U-78875 across baboons did not emerge and the effort was abandoned. Recent reports of unusual patterns of flumazenil interaction with later compounds in this series (Im et al., 1998) further suggest a possible differential interaction of U-78875 itself at multiple subtypes not investigated in the earlier work.

In the present study, the interactions with diazepam and lorazepam in animals trained to discriminate those drugs showed that the best predictor of whether U-78875 would antagonize the discriminative effects in a particular animal was whether it occasioned full drug-appropriate responding or not. Within-subject duality in function of U-78875 was further evidenced in pentobarbital-trained baboons that did not generalize to U-78875. In those animals, U-78875 potentiated pentobarbital but antagonized lorazepam. Receptor theory makes clear that apparent efficacy can vary as a function of the preparation, and that agonists might have multiple intrinsic efficacies (Kenakin, 1993; Clarke and Bond, 1998). We are reminded that "full agonist" and "partial agonist" are terms that derive from an observed drug effect under a particular set of conditions. Just as it is axiomatic in behavioral psychology to presume that stimuli have immutable functional properties (e.g., as reinforcers or punishers; Morse and Kelleher, 1977), the behavioral pharmacology of putative partial Bz agonists reminds us that multiple factors determine whether these compounds function as agonists or antagonists in any given context.