Opioids, Reward and Addiction: An Encounter of Biology, Psychology, and Medicine

A. Early History

Opium is the dried milky juice of the unripe seed capsule of the poppy, the Papaver somniferum. The word opium is derived from “opos”, the Greek word for juice. The first reference to this juice was by Theophrastus (300 B.C.), mentioning it mekonion. The medicinal and nonmedicinal use of opium by the ancient Greeks and Romans is not well documented, but it is generally believed that they were aware of the euphoric and narcotic (from the Greek word for stupor) properties of opium. They probably also knew that it could be applied for pain relief and dysentery. There are suggestions that the opium poppy was cultivated in Persia back to the end of the third millennium B.C. Arabic physicians used opium quite often and Arabic traders brought opium from the eight century A.D. on, first to the East, to India and China, and later to Europe. The Mohammedan prohibition of wine and the banning of tobacco smoking in China may have favored the spread of opium. With the “worldwide” availability of opium, the phenomenon addiction raised its head. An attempt to forbid the import of opium into China by the authorities, led to the so-called “Opium War” between England and China, with the result that opium trade was permitted (Macht, 1915).

Medicinal use of opium was stimulated by the famous physician Paracelsus at the end of the middle ages by the introduction of tincture of opium or laudanum. The name laudanum is probably derived from the Latin “laudandum”, which means something to be praised. Several preparations of laudanum were made, all of which contained more or less opium and many other ingredients. Laudanum and other preparations of opium (e.g., extracts of opium and pilulae opii) were widely used for a number of indications. In the beginning of the 19th century, the pharmacist Sertürner isolated an important active principle of opium, the alkaloid morphine (Sertürner, 1806). Morphine was named after the Greek god of dreams, Morpheus. During the nineteenth century many other alkaloids were isolated from opium, some of them with a comparable, but weaker action than morphine and others with a different pharmacological profile. From the mid-nineteenth century on, morphine was parenterally administered as premedication for surgical procedures and for postoperative and chronic pain.

Morphine appeared to be as addictive as opium. This stimulated research to develop nonaddictive opiates, substances with the beneficial therapeutic actions of morphine but lacking its addictive potential. In 1898, heroin was introduced as the ideal nonaddictive substitute for morphine. It lasted quite a long time before it became clear that heroin has a higher addictive potential than morphine. Several claims for nonaddictive opiates followed, but to date, none of these claims have been substantiated. During the 20th century a number of drugs were synthesized with a morphine-like action, but with a structure somewhat different from that of morphine. Examples are meperidine (1939) and methadone (1946). Structure-activity studies with the morphine molecule as starting point resulted in the synthesis of nalorphine, a mixed agonist-antagonist: the drug reverses the typical actions of morphine and it precipitates the abstinence syndrome in opiate addicts, but it also has analgesic properties. Additional research led to the discovery of pure opiate antagonists such as naloxone.

B. Opioid Receptors and Endogenous Opioids

The structural similarities between all substances with an opiate-like action and the discovery of opiate agonists, mixed agonist-antagonists and antagonists, generated the concept of opiate receptors. Goldstein et al. (1971) used radiolabeled levorphanol to discover opiate-binding sites in subcellular fractions of mouse brain. When radioligands with high specific activity became available, stereospecific opiate-binding sites in the central nervous system were demonstrated (Pert and Snyder, 1973; Simon et al., 1973; Terenius, 1973). The finding of opiate-binding sites and the fact that opiate antagonists exerted some intrinsic activity in opiate naive subjects and could diminish nondrug-induced analgesia stimulated thoughts about endogenous compounds with opiate-like action (Lasagna, 1965; Jacob et al., 1974; Akil et al., 1976; Buchsbaum et al., 1977).

In this review, the term opioid will be used for all substances with an opiate-agonistic action. Endogenous and exogenous opioids can be distinguished, depending on whether the substances are normally present in the body or not. The first indication for endogenous opioids came from studies showing that brain extracts contain opioid-like activity (Terenius and Wahlström, 1974; Kosterlitz and Waterfield, 1975). Further investigations led to the isolation and characterization of the enkephalins (from the Greek “in the head”), the first discovered endogenous opioids (Hughes et al. 1975). There appeared to be two pentapeptides, Met- and Leu-enkephalin. The structure of Met-enkephalin was also present as the N-terminal part of the earlier isolated C fragment, part of the fat-mobilizing pituitary hormone β-lipotropin (Bradbury et al., 1976). The C fragment, later termed β-endorphin (from endogenous morphine), and the enkephalins were shown to induce similar actions as morphine in a number of in vitro and in vivo test procedures. Repeated administration of β-endorphin led to tolerance to its analgesic action and to morphine-like withdrawal symptoms upon a challenge with naloxone (Van Ree et al., 1976; Wei and Loh, 1976). Furthermore, β-endorphin and the enkephalins were self-administered by laboratory animals, indicating the rewarding properties and addictive potential of these substances (Belluzzi and Stein, 1977; Van Ree et al., 1979;Goeders et al., 1984). Thus, the endogenous opioids may share all its typical opioid-like actions with morphine, both after acute and chronic administration.

After the discovery of another class of endogenous opioids, the dynorphins, (dyn… . from Greek dynamis = power) (Goldstein et al., 1979, 1981), it appeared that most endogenous opioids are generated by enzymatic processing from three precursor molecules, pro-opiomelanocortin (POMC),²proenkephalin (ProEnk), and prodynorphin (ProDyn) (Nakanishi et al., 1979; Kakidani et al., 1982;Noda et al., 1982). Each of these precursors has an unique anatomical distribution throughout the central nervous system (CNS) and in peripheral organs (Akil et al., 1984; Khachaturian et al., 1985). The anterior and neurointermediate lobes of the pituitary gland are major sites of POMC biosynthesis. In the brain, there are two distinct nuclei that contain POMC neurons: the arcuate nucleus of the hypothalamus and the nucleus tractus solitarius. Widespread projections from these neurons are present throughout the brain. From POMC the opioid β-endorphin is generated, but also α- and γ-endorphin and several nonopioid peptides, e.g., adrenocorticotropin and β- and γ-melanocyte-stimulating hormones. ProEnk-containing neurons are widely distributed throughout the brain and consist of both local circuits and long projection neurons. ProEnk is the source of Leu- and Met-enkephalin and several extended forms of these pentapeptides. ProDyn-containing cell bodies have a characteristic widespread distribution throughout the CNS. ProDyn-containing neurons have both short and long projection pathways and can generate several opioid peptides, including α- and β-neoendorphin, dynorphin A, and dynorphin B.

Martin et al. (1976) first postulated the existence of multiple types of opioid receptors. Based on their behavioral and neurophysiological findings in the chronic spinal dog, they distinguished between the μ type (for morphine, which induces analgesia, hypothermia, and meiosis among others), the κ-type (for ketocyclazocine, which induces depression of flexor reflexes and sedation among others), and ς-type (for SKF10,047 orN-allylnormetazocine, which induces tachycardia, delirium, and increased respiration among others). Later, a fourth type of opioid receptor, named δ (for vas deferens) was identified (Lord et al., 1977). Additional research revealed that the ς-type receptor is nonopioid in nature, leaving three main type of opioid receptors, μ, δ, and κ (Mannalack et al., 1986). These receptors, belonging to the family of seven transmembrane G protein-coupled receptors, have been cloned using molecular biological techniques (Evans et al., 1992;Kieffer et al., 1992; Reisine and Bell, 1993; Uhl et al., 1994; Knapp et al., 1995). Apart from occurring as separate molecules, brain μ- and δ-opioid receptors have also been suggested to function as a μ-δ receptor complex (for review, see Rothman et al., 1993). In slices of rat neostriatum, activation of this complex, which displays an affinity profile for opioid ligands different from nonassociated μ- and δ-opioid receptors, has been shown to inhibit dopamine (DA) D1-receptor-stimulated adenylate cyclase activity (Schoffelmeer et al., 1992, 1993).

Interestingly, there seems to be some preference for the different endogenous opioid ligands for the different receptors: β-endorphin for μ, enkephalins for δ, and dynorphins for κ. Subtypes of these receptors have been proposed (μ₁, μ₂; δ₁, δ₂; κ1, κ2, κ3) (Dhawan et al., 1996) and some evidence is available for some other receptor types [e.g., the ε receptor which was labeled as β-endorphin specific (Wüster et al., 1979; Narita and Tseng, 1998)]. The International Union of Pharmacology subcommittee on opioid receptors has proposed another terminology to distinguish the opioid receptors: OP1, OP2, and OP3 for the δ, κ, and μ receptor, respectively (Dhawan et al., 1996) (Table1). Another opioid-like receptor has been cloned, termed the ORL-1 opioid receptor (Fukuda et al., 1994;Mollereau et al., 1994; Lachowitz et al., 1995). In addition, some novel endogenous opioids have been isolated, termed orphanin FQ which seems to be an endogenous ligand for ORL-1 and endomorphin-1 and endomorphin-2 which have been proposed to represent a highly specific endogenous ligands for the μ receptor (Meunier et al., 1995,Reinscheid et al., 1995; Zadina et al., 1997). Since the discovery of orphanin FQ/nociceptin released the ORL-1-opioid-like receptor of its orphan status, a novel nomenclature of this receptor and its endogenous ligand has been proposed. By analogy to the known opioid receptors (μ, δ, and κ) the new name for ORL-1 would be o (omicron), after its endogenous ligand (orphanin). Metonymorphin, xenorphin, or endomicron were proposed as possible new names for orphanin FQ/nociceptin (Henderson and McKnight, 1997). It should, however, be noted that, if orphanin FQ/nociceptin were given a new name, then the possible new name for ORL-1 would change as well. We suggest the use of the combination xenorphin/ξ receptor and consequently XOR and OP4 for the molecular biology and International Union of Pharmacology recommendation nomenclature, respectively (Table 1).

C. Addiction

Opioids are drugs used for pain relief, against dysentery, and for a number of other therapeutic indications. During repeated treatment, tolerance to certain effects of opioids develops, e.g., to their analgesic action, which could result in discontinuation of the treatment, either or not after an increase of the daily dose. Another phenomenon occurring upon repeated treatment is the induction of physical dependence, characterized by withdrawal symptoms after discontinuation of drug treatment. Pathognostic for withdrawal symptoms is that they are suppressed by administration of the drug. Thus, the presence or the expectation of withdrawal symptoms could be an important incentive for restart or continuation of drug use. Although this does not seem to be a major problem in clinical practice, withdrawal symptoms have dominated postulates about the underlying mechanisms of addictive behavior for a long time.

It was generally believed that addicts will initiate their drug-taking habit because of the inherent euphoric action of opioids and will continue their habit to prevent the occurrence of withdrawal symptoms. Therefore, most addiction research was directed at the underlying mechanisms of physical dependence and related withdrawal symptoms. In this framework, drug-taking behavior has been conceptualized in the context of drive reduction (Hull, 1943). There emerged, however, some problems with this concept. The relapse rate in opioid addiction is high, also when the withdrawal symptoms have already disappeared for a long time. Moreover, physical dependence also develops in patients treated with opioids, for example, pain relief, but the percentage of these patients that initiates addictive behavior is quite low. Furthermore, physical dependence hardly develops with some other drugs with high-addictive potential such as cocaine. These observations stimulated research to delineate other factors that could explain the development and maintenance of opioid addiction and drug addiction in general. Among these are the reinforcing properties of drugs, drug-induced craving, and the concept of psychic dependence. During the last decades, several consensus meetings have been organized to provide workable terminology and concepts. However, in the literature of today, the terms addiction, dependence, and drug abuse are still used interchangeably.

Drug abuse may refer to “the use, usually by self-administration, of any drug in a manner that deviates from the approved medical or social patterns within a given culture” (Jaffe, 1990). Drug dependence may be a syndrome manifested by a behavioral pattern in which the use of a given psychoactive drug or class of drugs is given much higher priority than other behaviors that once had higher value. In its extreme form drug dependence is associated with the need for continued drug exposure (compulsive drug use), and it exhibits the characteristic of a chronic relapsing disorder (Edwards et al., 1981). Addiction can be regarded as a severe degree of drug dependence that is an extreme on the continuum of involvement with drug use (Jaffe, 1990). The system of diagnosis for mental disorders published in DSM-IV by the American Psychiatric Association (1994) uses the term substance dependence instead of addiction for the overall behavioral syndrome. Substance dependence is defined as “a cluster of symptoms indicating that the individual continuing use of the substance despite significant substance-related problems”. Withdrawal symptoms and tolerance can be present but are not a conditio sine qua non for the diagnosis substance dependence. Substance abuse, a less severe diagnosis, involves a pattern of adverse consequences from repeated use that does not meet criteria for substance dependence (O’Brien, 1996).

The need for continued drug use in drug dependence and addiction is basically of a psychic nature. Psychic dependence has been defined by “a condition in which a drug produces a feeling of satisfaction and a psychic drive that requires periodic or continuous administration of the drug to produce pleasure or to avoid discomforts” (Eddy et al., 1965). Besides development of psychic dependence, physical dependence [“an adaptive state that manifests itself by intense physical disturbances when the administration of the drug is suspended” (Eddy et al., 1965)] can contribute to compulsive drug use but it is not necessary for continued use. Although the nature of psychic and physical dependence is different, both are considered a priori to result from adaptive changes of neural systems in the brain in response to repeated drug use and/or exposure.

With regard to use of drugs, there exists a continuum from no drug use via controlled use to an actual dependence on the drug. The transition from controlled use to dependence may be referred to as initiation of drug dependence. It has been suggested that initially the use of a particular drug is related to its ability to produce effects of well being and euphoria. Environmental variables and/or individual characteristics contribute to whether or not an individual becomes dependent on the drug. At this point a basic emotional feature may have been altered by repeated drug use, which in turn is responsible for the need to experience the effect of the drug again and again. This need is basically of a psychic nature, but it can contain physical elements such as physical dependence. Once a person has become dependent on a drug, discontinuation of drug use is difficult. Even after a prolonged period of abstinence, addicts can relapse into their former habit of drug dependence. A factor that may be important for relapse is craving, a (intense) desire to re-experience the effects of the drug (Rankin et al., 1979; Markou et al., 1993). Drug craving can be conceptualized as the incentive motivation to self-administer a previously consumed drug. This craving may be present during continuous use of the drug and long after abstinence, and may develop on basis of incentive sensitization mechanisms in which associative learning plays a role (Bolles, 1975;Stewart et al., 1984; Robinson and Berridge, 1993). Besides craving, other factors may contribute to relapse, which is the major target for treatment programs of drug addiction.

A. Self-Administration

Drug self-administration is the most widely used model for the experimental analysis of drug addiction and is based on the concepts of operant conditioning. The administration of a drug of abuse is made contingent upon a behavioral response of the animal. This response may consist of alleyway running, arm choice in Y-maze and drinking of flavored solutions, yet most studies use lever-pressing as the behavioral act. An increase in the frequency of the response provides evidence that the drug is self-administered, and thus serves as a positive reinforcer.

In 1940, Spragg (1940) first suggested that drugs could function as positive reinforcers. His suggestion was based on experiments with chimpanzees, which were made physically dependent on morphine by daily treatment with morphine for several months. Then the animals could learn to select one of two boxes concealing a syringe filled with a morphine solution, which would subsequently be administered to the animal by the experimenter. The monkeys opened the box containing the morphine syringe more often than the other box that contained food.

Self-injection by animals was first reported by Headlee et al. (1955), who demonstrated that morphine was injected i.p. by physically dependent rats. In the early 1960s, several investigators developed techniques for i.v. self-administration in rats and monkeys (Weeks, 1962; Thompson and Schuster, 1964). Typically, an animal is surgically prepared with a chronic, indwelling i.v. catheter, which is guided s.c. to the arm, back, or head. Depending on whether primates or rodents are tested, restrainment in the test cages is used. Whereas monkeys are usually restrained by a primate chair or harness and arm arrangement, rats are allowed to move about freely in the test cage. The i.v. catheter is connected with an automatic infusion pump. Intravenous drug injections are made contingent upon a certain behavioral response under specified schedules of reinforcement.

Initial research with the i.v. self-administration method demonstrated that both opioid-dependent and opioid-naive animals would press a lever to receive injections with morphine (Weeks, 1962;Thompson and Schuster, 1964; Deneau et al., 1969). It became clear that besides morphine a wide variety of psychoactive drugs from different pharmacological classes could serve as positive reinforcers in animals. These drugs include psychomotor stimulants, such as amphetamine and cocaine (Pickens and Harris, 1968; Pickens and Thompson, 1968; Van Ree et al., 1978), dissociative anesthetics, such as barbiturates and benzodiazepines (Davis et al., 1968; Pilotto et al., 1984), ethanol (Smith and Davis, 1974), Δ⁹-tetrahydrocannabinol and the cannabinoid receptor agonist WIN 55,212 (Van Ree et al., 1978; Takahashi and Singer, 1979; Martellotta et al., 1998), phencyclidine (Balster and Woolverton, 1980), and nicotine (Lang et al., 1977; Goldberg and Spealman, 1982). In general, drugs that are self-administered by animals are abused to some extent by humans although there are exceptions. For example, rats will readily self-administer apomorphine (Baxter et al., 1974; Colpaert et al., 1976), whereas humans will not become dependent on this drug because of its nausea-promoting effects. Conversely, drugs that fail to initiate or maintain self-administration behavior in animals have no or little abuse potential in humans. It should, however, be noted that not all drugs are equally powerful as positive reinforcers in animals. For instance, nicotine is self-administered under a narrower unit dose range than opioids and cocaine. Nonetheless, the drug self-administration model can serve as a useful model for the prediction of the abuse potential of drugs in humans (Thompson and Young, 1978; Van Ree et al., 1978; Van Ree, 1979;Collins et al., 1984).

Intravenous self-administration in rats and monkeys is the most frequently used to assess the reinforcing effects of drugs. However, other models using other species (e.g., dogs, cats, mice, or pigeons) or other routes of administration (e.g., intragastric, oral, inhalation, i.c.v., or intracerebral) have been developed (e.g., Smith et al., 1976; Jones and Prada, 1977; Carroll and Meisch, 1978; Van Ree and Niesink, 1978; Van Ree et al., 1979; Kilbey and Ellinwood, 1980;Van Ree and De Wied, 1980; Bozarth and Wise, 1981b; Criswell, 1982;France et al., 1991; Mattox and Carroll, 1996).

Although the positive reinforcing effects of a drug are the most important stimuli in self-administration behavior, other factors may contribute significantly to operant behavior and thus self-administration behavior. These factors include, among others, conditioned or secondary reinforcement and negative reinforcement. Distinctive, neutral environmental stimuli that are repeatedly associated with the primary reinforcing effects of a drug, can acquire (secondary) reinforcing properties through classical conditioning (Davis and Smith, 1976; Beninger, 1983; Stewart et al., 1984). These stimuli are then called conditioned or secondary reinforcers. Although the primary reinforcing effects of the drug mainly determine the initiation of selfadministration behavior, the conditioned or secondary reinforcers maintain this behavior over time, even in the absence of the primary reinforcer. For example, a red light switched on when a monkey presses a lever to obtain a morphine injection subsequently supports lever-pressing when morphine is temporarily not available (Schuster and Woods, 1968). The effects of conditioned reinforcers diminish over time when the drug injection is no longer available. In animals made physically dependent on drugs, an additional factor influencing self-administration behavior is exerted by negative reinforcement, i.e., the animals will continue to self-administer a drug to alleviate or overcome the presumably aversive (negative) state of withdrawal (Solomon, 1980; Koob et al., 1989a).

B. Intracranial Electrical Self-Stimulation

Intracranial electrical self-stimulation (ICSS) is widely used to explore the involvement of particular brain circuits in reward. Typically, when an animal is equipped with an electrode placed in a “positive” brain area and given the opportunity to perform a behavioral response, e.g., pressing a lever, that is followed by a short-pulse train of electrical current via the electrode, the animal will initiate and maintain responding. Thus, the stimulation serves as an operant reinforcer (Skinner, 1938). The phenomenon of ICSS has been described initially by Olds and Milner (1954), who observed this behavioral pattern in rats equipped with electrodes in the septal area of the brain. ICSS was suggested to be linked to brain circuits implicated in natural incentives such as food and sexual contact (Olds and Milner, 1954; Trowill et al., 1969; Mogenson and Wu, 1982). However, it appeared that a variety of brain structures, related and not related to natural incentives, could support ICSS (Olds et al., 1971; Wise, 1996). Although ICSS resembles other types of reward, it has some unique properties. In most stimulated sites, the rewards are strong and immediately present during stimulation and it lasts not much longer than the stimulus itself. The brain structures in which ICSS can be elicited have been designated as reward or pleasure centers. Whether these various brain structures belong to a single system or to multiple reward circuits operating in parallel is still a matter of debate.

In general, drugs of abuse facilitate ICSS in that the frequency current-response function is shifted leftward in a parallel manner and/or the threshold for eliciting ICSS is decreased. Such findings have been documented for morphine and heroin (Esposito and Kornetsky, 1977; Van Wolfswinkel and Van Ree, 1985b; Hubner and Kornetsky, 1992;Bauco et al., 1993), amphetamines (Gallistel and Karras, 1984;Schaefer and Michael, 1988b), cocaine (Bain and Kornetsky, 1987; Frank et al., 1988; Van Wolfswinkel et al., 1988; Bauco and Wise, 1997), nicotine (Huston-Lyons et al., 1992; Bauco and Wise, 1994; Ivanova and Greenshaw, 1997; Wise et al., 1998), phencyclidine (Kornetsky and Esposito, 1979; Carlezon and Wise, 1993b), and Δ⁹-tetrahydrocannabinol (Gardner et al., 1988,1989; Lepore et al., 1996). With respect to ethanol, the data so far are not fully consistent (De Witte and Bada, 1983; Schaefer and Michael, 1987; Bain and Kornetsky, 1989, Moolten and Kornetsky, 1990;Lewis and June, 1994). It seems that facilitation of ICSS is an effect that drugs of abuse have in common, despite the differential pharmacological characteristics of these drugs. Thus, facilitation of ICSS may be relevant for the dependence-creating properties of drugs and worthwhile to analyze in detail to understand the basic mechanisms of drug dependence.

Concerning the neurobiology of ICSS, catecholamines and especially DA have been implicated as important neurotransmitters in the reward circuit (Crow, 1972; German and Bowden, 1974; Wise, 1978). Evidence that DA is involved in ICSS stems from anatomical studies (Corbett and Wise, 1980), lesion experiments (Fibiger et al., 1987), pharmacological manipulations (Zarevics and Setler, 1979; Wise and Rompré, 1989), and neurochemical studies (Nakahara et al., 1992; Fiorino et al., 1993;Di Chiara, 1995). It has been suggested that in particular the mesocorticolimbic DA system is important for ICSS.

C. Conditioned Place Preference

Affective (rewarding or punishing) stimuli can evoke approach or avoidance behavior, respectively (Schneirla, 1959). When these stimuli are paired with a neutral environment, these neutral environmental stimuli can gain the capacity of evoking similar approach or avoidance behavior as the affective stimulus (Pavlov, 1927). Affective effects of drugs can thus be assessed by giving the animal the possibility to express attraction or aversion to environments paired with effects of drugs. This principle is the basis of the place-conditioning method. In place-conditioning, individuals are exposed to distinctive neutral environmental cues as conditioning stimulus, after being treated with a drug, the effects of which will then act as an unconditioned stimulus (Hoffman, 1989; Schechter and Calcagnetti, 1993; Bardo et al., 1995). A test apparatus, consisting of (at least) two compartments with distinct visual, olfactory, or tactile cues is used, so that an animal will be able to distinguish between these compartments. In alternating conditioning sessions, the animal is confined to one compartment after being injected with a drug. In a subsequent session, the animal is injected with placebo and placed in the other compartment. This procedure is repeated several times [although significant place-conditioning can be achieved with a single conditioning session (Mucha et al., 1982; Bardo and Neisewander, 1986)] so that the animal learns to associate the cues of one distinct compartment of the test apparatus with the effects of the test drug. On a day after these conditioning sessions, the animal is placed in the test apparatus, without any confinements. The animal will have the opportunity to move freely around the different compartments, and the relative amount of time spent in these compartments is measured. When the animal spends significantly more time in the drug-paired part of the apparatus, it is said to display conditioned place preference. Likewise, a smaller amount of time spent in the drug-paired environment reflects conditioned place aversion.

With respect to conditioned place preference, there are some experimental variables that need to be taken account of. First, the structure of the place-conditioning apparatus may be such that an animal has an initial preference for one or the other compartment. If, for example, an apparatus consisting of a black and a white compartment is used, rats usually have a natural preference for the black (i.e., darker) side of the apparatus. Such preference may be revealed by subjecting the animals to a pretest before conditioning commences. In this case, termed a biased design, the conditioning drug of which a preference is expected is mostly paired with the least preferred compartment of the apparatus. Place-conditioning is then expressed as the amount of time spent in the drug-paired compartment of the test apparatus minus the amount of time spent in that environment during the pretest. A criticism on the biased design is that approach behavior might not be due to positive motivational but to anxiolytic properties of the drug, decreasing the animals’ anxiety in a nonpreferred environment. However, the finding that place preference even in an biased design is consistently found with psychostimulant drugs such as amphetamine and cocaine, which do have anxiogenic properties, is hard to reconcile with such reasoning. An alternative experimental design is to pair the conditioning drug with one side of the apparatus in half of the animals and with the other side in the other half of the animals (counterbalanced design). The most elegant approach is to manipulate the different environments in such a way using, for example, different odors that animals no longer prefer one side of the apparatus over the other (unbiased design), and to subsequently condition the animals in a counterbalanced fashion. A second factor is that the effects of a conditioning drug might interfere with exploration of the drug-paired environment. Since environmental novelty can elicit both approach and avoidance behavior, this might influence the expression of conditioned place preference or aversion. To circumvent this, a placeconditioning apparatus consisting of three compartments can be used. Two parts of this apparatus will then be paired with placebo or drug, respectively, whereas the third compartment is completely novel to the animal on the test day. The presence of a novel compartment might then obviate for any influences of exploratory behavior on preference for the placebo- or drug-paired environment, supposedly overruling the relative novelty of one of the two conditioning compartments. Environmental novelty has been shown to elicit conditioned place preference (Bardo et al., 1989) and the capacity of novelty to induce place preference has also been compared with drug-induced place preference. It was shown that rats consistently preferred environments paired with morphine, amphetamine, or apomorphine over novel environments, which were in turn preferred over familiar, saline-paired environments (Parker, 1992).

Place-conditioning studies have been performed with a wide variety of psychoactive drugs as well as with nonpharmacological stimuli. For example, conditioned place preference has consistently been observed using opioids such as morphine, fentanyl, heroin, and β-endorphin, psychostimulants, e.g., amphetamine, methylphenidate, cocaine, 3,4-methylenedioxymethamphetamine (“ecstasy”), and nicotine, as well as with benzodiazepines such as diazepam. In addition, nonpharmacological cues such as social and sexual interaction, environmental novelty, and sucrose drinking elicit conditioned place preference. Place aversion has been reported for opioid antagonists such as naloxone, for lithium chloride, and for aggressive attacks, ionizing radiation, and footshock. Mixed results have been reported for apomorphine, caffeine, ethanol, and phencyclidine (for reviews, see Hoffman, 1989; Schechter and Calcagnetti, 1993).

The place-conditioning paradigm has some interesting properties in comparison to other models generally used in addiction research. First, next to positive motivational, also aversive properties of drugs can be assessed. This is also possible with the self-administration method. In that case, the number of self-injections with the drug with aversive effects will be lower than the number of placebo self-injections. To observe these aversive effects, however, relatively high levels of responding for placebo will be required, whereas in the place-conditioning method there is no such restriction. Second, the place-conditioning paradigm is not an operant task, but rather a classical (Pavlovian) conditioning paradigm. During conditioning, the drug will be delivered, irrespective of the behavior of the animal. What is learned is therefore not response conditioning but stimulus conditioning. However, what is measured during testing is not merely the result of classical conditioning because the approach behavior that is measured is not necessarily part of the primary (unconditioned) response of the animal to the drug. Third, the behavior of the animal is measured in the drug-free state. Any drug effects that might interfere with behavioral performance during testing is avoided. However, testing animals in a drug-free state is also a disadvantage. Since conditioning and testing are conducted in two different interoceptive states, state-dependent learning might occur, causing a risk of false negatives. Finally, since significant place-conditioning can be obtained with a small number of conditioning trials, the duration of the experiments is relatively short.

The most serious disadvantage of the place-conditioning technique is that the interpretation of data is difficult. As most if not all substances that are self-administered by humans and laboratory animals also have the property to induce conditioned place preference, it is suggested that their positive affective properties play a significant role in the development of place preference. It is, however, hard to interpret what the animal is exactly expressing with its approach or avoidance behavior. The most likely explanation is that conditioned place preference reflects the desire to experience the effects of the drug. However, since in most cases only time spent in a certain compartment is scored, such a conclusion cannot definitely be drawn.

A. Intravenous Opioid Self-Administration

The first studies on i.v. opioid self-administration were performed in the 1960s. Different groups demonstrated that rhesus monkeys and rats would learn to press a lever to receive i.v. infusions with morphine (Weeks, 1962; Thompson and Schuster, 1964; Deneau et al., 1969). These experiments showed that morphine served as a positive reinforcer of self-administration behavior in animals made physically dependent on opioids as well as in animals which were naive to the drug (nondependent). These initial experiments have been replicated and extended by many laboratories over the last decades. Reviewing the studies evaluating the reinforcing properties of opioids in the self-administration paradigm revealed several procedures that could be classified under different headings. In the present overview, we divide them into the “acute” method and the “substitution” method.

In the first method, the acute method, the animal is allowed access to the test drug without previous experience with this drug or any standard drug. Different procedures can be applied. For example, the animal is given initial access to vehicle (usually saline) for a few days to obtain control rates of responding before being offered the test drug. Alternatively, the animal is initially trained to lever-press for a nondrug reinforcer (usually food pellets) on a continuous reinforcement schedule to familiarize the animal with lever-press responding. In a third procedure an animal without any previous experience with the behavioral requirements for the delivery of the drug is offered access to the test drug. During the period of actual access each lever-press results in an infusion. The acute method is useful in assessing whether an animal will initiate self-administration of the test compound, whether responding for the compound will change over time, and to assess dose-response curves of the test compound.

In the substitution method, drug self-administration is first established with a standard compound known to be reinforcing. In short, during daily experimental sessions, animals are trained to respond for an i.v. delivery of a standard compound. This compound is referred to as baseline drug and is known to produce reliable self-administration rates over sessions. In substitution studies evaluating the reinforcing properties of opioids, codeine, a pure opioid agonist, is mostly used as baseline drug, although some studies use cocaine as such. After responding becomes stable, a dose of a test compound or vehicle (usually saline) is substituted for the baseline drug for one or several sessions. If saline is substituted, responding tends to decline to relatively low rates (negative control). On the other hand, when a test compound is substituted, the compound may maintain responding at some level above that of saline. If this occurs, the drug is classified as a positive reinforcer. Using this method dose-response curves of test drugs can be generated and relative reinforcement potencies of several drugs can be determined.

The results of the opioid self-administration studies with these methods have been reviewed (e.g., Johanson and Balster, 1978; Griffiths and Balster, 1979; Woolverton and Schuster, 1983; Collins et al., 1984;Young et al., 1984; Balster and Lukas, 1985; Vaupel et al., 1986; Woods and Winger, 1987). In the next paragraphs, we summarize the early findings and refer to previous reviews for more detailed discussion.

Johanson and Balster (1978) summarized data generated using the substitution method in monkeys to assess the reinforcing properties of several opioid drugs. They reported that, in general, all tested pure opioid agonists are readily self-administered under a variety of experimental protocols. These agonists include the opioid agonistsl-α-acetylmethadol (LAAM), alfentanil, codeine, dihydroetorphine, etonitazine, etorphine, fentanyl, heroin, hydromorphone, levorphanol, methadone, meperidine, morphine, and propoxyphene (Ternes et al., 1985; Bertalmio and Woods, 1989; Beardsley and Harris, 1997). An exception was the opioid agonist tilidine, which acted as a positive reinforcer in one laboratory but not in another laboratory. The mixed opioid agonist-antagonists buprenorphine, butorphanol, α-etazocine, nalbuphine, pentazocine, and propiram tested in different laboratories also served as positive reinforcers in the substitution procedure. In contrast, drugs like cyclazocine, ketocyclazocine, levallorphan, nalorphine, naloxone, and naltrexone were not self-administered by monkeys (Slifer and Balster, 1983). Self-administration of opioids appeared to be stereoselective in the sense that the d-isomer dextrorphan was not self-administered, whereas the l-isomer levorphanol was (Winter, 1975). In addition i.v. self-injections have been observed with the enkephalin analog FK-33-824 in monkeys physically dependent on morphine (Mello and Mendelson, 1978).

Comparing the positive reinforcing properties of opioids with their discriminative and antagonistic effects (reversal of behavioral effects of opioid agonists and direct rate-decreasing action), a distinction can be made in three groups, i.e., opioids with morphine-like properties (μ-type opioid), opioids represented by ethylketazocine (κ-type opioid), and opioid antagonists (Woods et al., 1982;Woolverton and Schuster, 1983; Young et al., 1984; Woods and Winger, 1987). In general, preferential μ-type opioid agonists (such as alfentanil, codeine, etorphine, fentanyl, heroin, levorphanol, meperidine, methadone, and morphine) and the mixed μ-type opioid agonist-antagonists (such as buprenorphine, butorphanol, nalbuphine, pentazocine, profadol, and propiram) that exert morphine-like effects in other systems readily maintain i.v. self-administration in rhesus monkeys. In contrast, the ethylketazocine-like opioid agonists, such as bremazocine, ethylketazocine, ketacyclazocine, ketazocine, and other benzomorphan ligands, in general do not maintain self-administration behavior. Also, the mixed opioid agonist-antagonists, like cyclazocine, nalorphine, and oxilorphan, which share agonistic actions primarily with the ethylketazocine-like opioid agonists, are not self-administered by monkeys. In fact, they act as negative reinforcers in the self-administration procedure. The third group, the opioid antagonists such as naloxone and naltrexone, fails to maintain responding and in higher doses serve as negative reinforcers.

Griffiths and Balster (1979) investigated whether the self-administration paradigm could be useful as a tool in the assessment of the subjective effects of opioids in humans. For that the results of drug self-administration of the above-mentioned opioid drugs by monkeys using the substitution procedure were compared with clinical evaluations of morphine-like signs, symptoms, and subjective effects of a single dose of these opioid drugs in humans (Jasinski, 1977). Of the 33 drugs examined, 4 drugs did not produce similar results in humans and monkeys. Of these four drugs, three (i.e., dextromethorphan, butorphanol, and nalbuphine) were self-administered by monkeys but induced no or equivocal morphine-like effects in humans. One drug, tilidine, produced clear morphine-like subjective effects in humans but did not readily maintain self-administration in monkeys. The opioid antagonists were not self-administered by monkeys nor produced morphine-like effects in humans. The concordance between the human and animal results in this study validates the use of the self-administration paradigm to subjective effects of opioids in humans.

Although the monkey has been the preferred species for testing the abuse potential of new drugs, the increasing costs of this experimental animal led Collins et al. (1984) to develop a rapid screening test for the reinforcing actions of, among others, several opioid drugs in a less expensive animal, the rat. Using the acute self-administration method with drug-naive rats, they showed that the self-administration results with opioid drugs in rats paralleled those obtained with monkeys. Moreover, using drug-naive animals, these results demonstrated that opioid drugs, besides maintaining self-administration behavior, also readily initiate selfadministration behavior. In rats, self-administration has been shown for μ-type opioid agonists 6-acetylmorphine, codeine, dihydroetorphine, dihydromorphine, etonitazene, fentanyl, heroin, meperidine, methadone, morphine, and propoxyphene; the mixed μ-type opioid agonist-antagonists butorphanol, nalbuphine, nalorphine, and pentazocine; and the κ-type opioids ethylketocyclazocine and ketocyclazocine (e.g., Weeks and Collins, 1964, 1979; Collins and Weeks, 1965; Smith et al., 1976; Van Ree et al., 1978; Collins et al., 1984; Young and Khazan, 1984; Dai et al., 1989; Martin et al., 1997). Furthermore, [d-Ala²]-Met-enkephalin, dynorphin-(1–13), and [d-Ala²]-dynorphin-(1–11) were i.v. self-administred in rats physically dependent on morphine and pretrained to self-inject morphine (Tortella and Moreton, 1980; Khazan et al., 1983). The μ-type antagonists cyclazocine and naloxone were not self-administered (Collins et al., 1984).

In more recent studies another method, the reinstatement model, has been used to assess the effects of opioids on opioid-seeking behavior. Typically, animals are trained to i.v. self-administer opioids. After reliable self-administration, extinction sessions are given during which saline is substituted for the opioid. After termination of responding under extinction conditions, a priming injection with a test drug is given and lever-pressing is assessed. Stewart and coworkers (De Wit and Stewart, 1983; Stewart, 1983) found that priming injections of 50 to 200 μg/kg heroin effectively reinstated heroin-seeking behavior in rats. Priming infusion of pharmacologically related drugs also reinstated responding on the lever associated with heroin infusions. Under these conditions, injections with nalorphine (10 mg/kg) however did not reinstate lever-press behavior and naltrexone (2 mg/kg) suppressed responding below the levels seen after saline injections (Stewart and Wise, 1992). Using this reinstatement procedure, the same research group replicated their findings, in that reexposure to heroin after abstinence reinstated heroin-seeking behavior, whereas an injection of naloxone did not (Shaham and Stewart, 1996; Shaham et al., 1996, 1997). Moreover, they found that brief exposure to footshock stress or corticotropin-releasing hormone reinstated heroin-seeking behavior, thereby mimicking the effect of a noncontingent priming infusion of heroin, although differences may be present with regard to the neurobiology of the various reinstatement stimuli (Shaham and Stewart, 1994, 1995; Shaham et al., 1996, 1997). By contrast, the aversive state of opioid withdrawal, induced by an injection with naltrexone, did not reinstate drug-seeking behavior. Other abused drugs like amphetamine and cocaine are also capable to reinstate drug-seeking behavior in animals, previously trained to self-administer heroin (De Wit and Stewart, 1983; De Vries et al., 1998). It has been argued that the reinstatement model is relevant for assessing opioid-induced relapse.

In general, it seems that both the pure μ-opioid agonists (such as morphine, methadone, codeine, and heroin) and the mixed μ-opioid agonist-antagonists (such as butorphanol, nalbuphine, and buprenorphine) serve as positive reinforcers in the different i.v. self-administration paradigms in monkeys and rats. On the other hand, κ-opioid agonists (such as ethylketazocine, ketazocine, and benzomorphan ligands) and opioid antagonists (such as levallorphan, naloxone, and naltrexone) do not maintain self-administration behavior or even maintain responses that postpone or terminate their injection (negative reinforcers). In the rat, however, self-administration of (ethyl)ketazocine and nalorphine has been reported.

B. Variables Interfering with Opioid
Self-Administration

Drug-taking behavior in general, and opioid selfadministration in particular, is controlled by a number of variables. The most commonly discussed variables concern those that can readily be manipulated, e.g., the dose of the drug administered, the route of administration, the schedules of reinforcement, and the presence or absence of physical dependence, tolerance, and sensitization. In addition, a number of predisposing variables (e.g., preexposure to opioids during gestation, environmental factors, and genetic factors) can affect drugtaking behavior. The contribution of these factors will be discussed briefly. The discussion will focus on studies with rhesus monkeys and rats since the reinforcing properties of opioids have been studied most extensively in these species.

1. Dose of the Drug.

The dose of the drug per injection (“unit dose”) is an important and critical factor in drug-taking behavior. A linear function between the log dose of drug delivered per injection and the amount of drug taken (Fig.1) has been shown to exist for various opioids in different species of animals (Weeks and Collins, 1964, 1979;Smith et al., 1976; Harrigan and Downs, 1978a; Van Ree et al., 1978;Dai et al., 1989). Moreover, studying heroin self-administration, it was demonstrated that the linear relationship between the log unit dose and drug intake is present during initial exposure to heroin and in animals physically dependent on heroin (Van Ree et al., 1978; Dai et al., 1989).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Self-administration of morphine sulfate in rats. Presented are the mean number of self-infusions (A) and the mean amount of drug intake (B) by either the i.v. (●) or the intragastric route (○) versus the unit dose of morphine sulfate per infusion. Data fromSmith et al. (1976).

An intermediate dose range is optimal for maintaining self-administration of morphine, whereas lower and higher doses do not favor it. Data from a wide range of unit doses for drug-naive rats taking i.v. morphine on a continuous schedule of reinforcement have been reported (Smith et al., 1976; Weeks and Collins, 1979). On the fifth or sixth day of self-administration, a medium dose of around 30 μg/kg morphine per infusion maintained a maximum number of i.v. deliveries. When lower doses of morphine were tested, a progressive decrease in the rate of self-administration was found, probably due to a decrease in reinforcing efficacy of the drug. Similarly, with higher doses of morphine [100–10,000 μg/kg/infusion (inf)] a decrease in rate of responding for drug infusions was observed. Similar data were obtained in rats that were exposed to graded unit doses of heroin (Dai et al., 1989). In general, data from both rats and monkeys revealed that the relationship between log unit dose of a drug and the self-administration rate seems to be an inverted U-shaped curve (Harrigan and Downs, 1978a; Young et al., 1981; Collins et al., 1984;Balster and Lukas, 1985; Lukas et al., 1986; Vaupel et al., 1986;Martin et al., 1997). The decrease in responding with higher doses of the drug might be explained by a cumulative effect of increased reinforcement per infusion, which diminishes the drive of the animal to respond for morphine, and by catalepsy, gnawing, sedation, and satiation. Taken into account the linear function between the log unit dose and drug intake, it seems that the amount of drug intake is a more informative index for the reinforcing efficacy of opioids and of other abused drugs than the number of self-injections (Van Ree et al., 1978). The rate of opioid intake seems not only dependent on the unit dose available for self-administration, but upon the training dose as well.Martin et al. (1998) demonstrated that doses of heroin lower than 5.4 μg/inf maintained higher rates of drug intake in rats trained with 5.4 μg/inf as compared to rats trained with 18 μg/inf heroin, whereas doses higher than 5.4 μg/inf maintained similar rates of heroin intake in both groups of animals.

The various opioids differ markedly in their potencies, i.e., in the unit doses that maintain maximum rates of responding under similar conditions. Comparison of the log dose-response curves of heroin, morphine, methadone, and LAAM in rhesus monkeys showed that all opioids maintained comparable peak self-administration rates, but at different unit doses (Harrigan and Downs, 1978a). The rank order of the relative potency was heroin > LAAM > morphine > methadone (1, 4, 10, and 16 μg/kg/inf, respectively). Systematic studies with several opioid agonists and mixed opioid agonist-antagonists in different species reveal that under identical access conditions the relative potencies of the different opioid drugs can vary by more than 10,000-fold when taking the peak of the inverted U-shaped dose-response (self-administration rate) as assessment (Young et al., 1981; Collins et al., 1984; Balster and Lukas, 1985; Lukas et al., 1986; Vaupel et al., 1986; Martin et al., 1997).

Another method to estimate the relative potencies of different opioids is to compare their ED₅₀ values (i.e., the unit dose of a drug that initiates and maintains reliable self-administration behavior above saline level in 50% of the animals). By offering various unit dose levels of morphine, fentanyl, and heroin, a substantial portion of the animals readily initiated self-administration behavior according to a linear dose-response curve (Van Ree et al., 1978). The calculated relative ED₅₀ values were 2.5 μg/kg/inf for fentanyl, 50 μg/kg/inf for heroin, and 650 μg/kg/inf for morphine. Interestingly, comparison of the reinforcing and analgesic properties of these opioids revealed an accurate similarity between the relative ED₅₀ values for self-administration behavior and the relative potencies of these drugs for analgesia.

The dose of a drug is also an important determinant of the temporal distribution of opioid self-injections (Van Ree et al., 1978; Weeks and Collins, 1979). Drug-naive rats offered 10 mg/kg/inf morphine under a continuous reinforcement schedule seldom took more than one injection at the time. When the dose of the drug is reduced to 3.2 mg/kg/inf, double and an occasional triple injections were seen. With much smaller doses (32 μg/kg/inf), rats usually take morphine in series of closely spaced injections (up to 50 injections), and then a pause with only sporadic injections until the next series (Weeks and Collins, 1979). Looking at the interinfusion intervals, a higher percentage of relatively low interinfusion intervals (1–10 min) was found when a low dose of morphine, heroin, or fentanyl was offered, whereas longer intervals (30–60 min) between infusions occurred more frequently when the unit dose per injection was higher (Van Ree et al., 1978).

In conclusion, the unit dose of a drug delivered is one of the main factors which determines the ultimate level of drug intake during self-administration. The amount of drug taken can serve as a useful index of the reinforcing efficacy for the reinforcer (i.e., drug injection).

C. Endogenous Opioids and Opioid Drugs of Abuse

A. Effects of Opioids

The first report about the effect of morphine on ICSS was fromOlds and Travis (1960). The self-stimulation behavior was studied over a range of stimulus intensities in animals with electrodes implanted in the lateral hypothalamus (LH), septal area, or VTA. Although it was found that morphine (7.0 mg/kg i.p.) caused a significant decrease in the response rate in most of the animals, some facilitation of the rate was seen as well. There seemed to be some site specificity of the effects of morphine. Self-stimulation from the VTA was more facilitated by morphine than that from the septal preparations. Conversely, there were more inhibitory effects in the septal than in the tegmental preparations. In most cases, morphine decreased the response rate in animals with electrodes in the hypothalamus.

It lasted about a decade before the next report on this issue was published, probably because of the interest in the procedure of self-administration of morphine and other drugs of abuse as a model to investigate reinforcing properties (Weeks, 1962; Deneau et al., 1969;Van Ree, 1979). In 1972, Adams et al. reported that morphine (10 mg/kg s.c.), decreased self-stimulation behavior in rats with electrodes in the medial forebrain bundle (MFB) during the first 2 h after drug administration. However, thereafter a facilitation of the response rate was observed. Morphine was administered for 5 consecutive days. By day 3, there appeared to be complete tolerance to the inhibitory effect on the response rate along with no tolerance to the facilitating action of morphine. These findings of decreasing and facilitating effects of morphine after acute and repeated administration of the drug has been further analyzed in a number of studies.

Morphine can stimulate and depress motor performance depending upon several variables such as the dose, time between injecting and testing, and presence or absence of tolerance. Since in most ICSS studies a motor response is the measured variable, the effects of morphine on motor performance may interfere with the drug-induced changes in reinforcement, as attempted to measure with ICSS, and may hamper the interpretation of the observed effects. A way to circumvent problems associated with performance changes in rewarded behavior is the determination of the threshold for that behavior. Such a threshold method, usually associated with a low rate of motor performance, may measure reward-induced changes in behavior more accurately and physiologically than methods that are highly dependent on motor performance. Several methods to determine a threshold in operant behavior have been designed (Stein and Ray, 1960; Franklin, 1978;Schaefer and Holtzman, 1979; Ettenberg, 1980). Esposito and Kornetsky (1977) observed a threshold decrease after the administration of morphine in a rate-insensitive “double staircase” psychophysical method to determine threshold of ICSS in rats with electrodes in the MFB. They tested the effect of 1 to 10 mg/kg s.c. morphine on threshold repeatedly during 2 to 4 weeks and found no tolerance for this effect of morphine. Although they administered 8 to 10 mg/kg daily after the test sessions, the lower dose given the next day before the test remained effective on threshold. Van Wolfswinkel and Van Ree (1985b)compared the effect of graded doses of morphine (0.3–5 mg/kg s.c.) using three different procedures to measure the threshold of ICSS in rats with electrodes in the VTA. The procedures were 1) determination of response rate, i.e., the number of responses, to high and threshold currents; 2) measuring threshold current when the response rate was kept low and relatively constant; and 3) determination of “behavioral” threshold using a two-lever procedure in which a response on one lever resulted in a reset of the decreasing current to a high-current contingent on a response to the other lever (see also,Stein and Ray, 1960). Different groups of animals were used for the three procedures and five doses of morphine were administered in increasing dose, spaced at least for 2 days. As compared to placebo treatment the previous day, morphine induced a slight decrease (low doses) and increase (high dose) of the threshold current in the response rate procedure, no effect in the constant response rate procedure, and a dose-related decrease of the threshold current in the behavioral threshold procedure. During this latter procedure, no change in response rate was observed after morphine treatment. A similar effect of morphine in the behavioral threshold procedure was observed in rats used before for the two other procedures. The behavioral threshold procedure, in which the rat can select its own threshold current, is theoretically the most insensitive to nonreward-related motor performance effects. Response rate is not used for calculation of the threshold and was not affected by morphine treatment (see also,Zarevics and Setler, 1979). In subsequent experiments, using the same behavioral threshold procedure, no tolerance to the morphine-induced decrease in threshold was observed when morphine (5 mg/kg s.c.) was administered for 15 days before ICSS testing (Van Wolfswinkel et al., 1985). In this experiment, a decrease in response rate was found, but only during the first 2 days of morphine treatment. Thus, enhanced brain reinforcement can be observed after acute and chronic treatment with morphine when a response rate-insensitive procedure is used to measure ICSS behavior. This conclusion corroborates with other experiments using threshold determinations of ICSS (Esposito and Kornetsky, 1977, 1978; Esposito et al., 1979; Nazzaro et al., 1981;Kornetsky and Bain, 1983).

When the response rate is measured as a dependent variable for determining ICSS, the effect of morphine depends on the drug dose, the time between treatment and testing, and whether or not the animals are drug-naive. In general, systemically administered low doses of morphine (<3 mg/kg) can increase, whereas larger doses disrupt responding in the period shortly after administration (Adams et al., 1972; Glick and Rapaport, 1974; Wauquier and Niemegeers, 1976; Schaefer and Holtzman, 1977; Weibel and Wolf, 1979; Van Wolfswinkel and Van Ree, 1985b). This disruption is followed by an increase in ICSS responding 2 to 6 h after morphine treatment (Adams et al., 1972; Lorens and Mitchell, 1973; Lorens, 1976). This delayed facilitation is enhanced and present earlier after repeated injection of morphine (Adams et al., 1972;Kelley and Reid, 1977; Schaefer and Holtzman, 1977). Using a procedure in which the stimulus frequency of the electrical current is varied yielding a response rate-frequency function that resembles the traditional pharmacological doseresponse curve, morphine induced a leftward shift of the response rate-frequency function, indicating facilitation of ICSS (Rompré and Wise, 1989; Bauco et al., 1993;Carlezon and Wise, 1993a; Wise, 1996). In tolerant animals, an enhancement of ICSS has consistently been found (Lorens and Mitchell, 1973; Bush et al., 1976; Weibel and Wolf, 1979; Van Wolfswinkel et al., 1985).

The facilitation of ICSS by morphine is mimicked by other opioids administered systemically, as shown by experiments in which heroin, 6-acetylmorphine, methadone, levorphanol, or pentazocine was tested (Kornetsky et al., 1979; Weibel and Wolf, 1979; Bozarth et al., 1980;Stutz et al., 1980; Gerber et al., 1981; Preshaw et al., 1982; Schenk and Nawiesniak, 1985; Hubner and Kornetsky, 1992). The facilitation of ICSS appeared to be stereoselective in that dextrorphan did not enhance ICSS, and opioid antagonists blocked the effect (Weibel and Wolf, 1979). Thus, opioid receptors are probably involved in the opioid-induced facilitation of ICSS.

From the data, it can be concluded that morphine and other opioids can facilitate ICSS reward and that no tolerance developed for the facilitating effect of morphine. Depending on the procedure used, an initial depression of behavior is present in morphine-naive animals, but tolerance to this probably nonreward-related effect develops upon repeated drug administration.

In experiments in which systemically administered morphine facilitated ICSS, the electrodes were in general implanted in the MFB/LH area or in the VTA. Although a direct comparison between these areas with respect to morphine action has not been performed, the obtained data are quite comparable: doses of morphine around 1 mg/kg and higher facilitated ICSS. In some studies other brain sites have been studied. When the electrodes were implanted in the medial prefrontal cortex, locus ceruleus, dorsal raphe nucleus, or mesencephalic central gray, similar effects of morphine were found (Lorens, 1976; Liebman and Segal, 1977;Esposito et al., 1979; Jackler et al., 1979; Nelson et al., 1981;Schenk et al., 1981). But a facilitation of ICSS by morphine was not present when the electrodes were implanted in the substantia nigra or in the medial part of the anterior prefrontal cortex (Nazzaro et al., 1981; Corbett, 1992). Thus, not all sites from which ICSS behavior can be elicited seem to be influenced by systemically administered morphine.

The facilitating effect of systemically administered opioids was mimicked when relatively low doses of morphine or levorphanol, but not dextrorphan, were injected directly into the brain ventricle, implicating central opioid receptors in this opioid action (Weibel and Wolf, 1979; Shaw et al., 1984). A number of studies have addressed the site of action of morphine and other opioids in facilitating ICSS behavior. Morphine (1 μg) injected bilaterally into the ventral tegmental/substantia nigra area but not in the NAC or the striatum facilitated ICSS behavior elicited from electrodes placed in the MFB (Broekkamp and Van Rossum, 1975; Broekkamp et al., 1976; Broekkamp and Phillips, 1979). Morphine was effective at a dose of 200 ng, but not of 50 ng, and the drug effect was blocked by systemically administered naloxone. The effect of morphine was mimicked by injection of [d-Ala²]-Met-enkephalinamide into the same area. A dose-dependent decrease in the frequency threshold for ICSS from the MFB was found after injecting morphine into the VTA (Rompré and Wise, 1989; Bauco et al., 1993). Neither sensitization nor tolerance was observed following repeated morphine injection (Bauco et al., 1993).

Selective μ-, δ-, and κ-opioid receptor ligands have been injected into the VTA in rats with electrodes in the MFB. The effects on ICSS were, however, not consistent and both facilitating effects and no effects have been reported (Jenck et al., 1987; Heidbreder et al., 1992; Singh et al., 1994). Other studies have shown stimulating effects of the μ-opioid receptor ligand DAMGO injected into the lateral accumbens core or the caudal ventral pallidum and of the δ-specific ligand DPDPE injected into the caudal ventral pallidum or ventromedial striatum (Johnson et al., 1993, 1995a; Johnson and Stellar, 1994). ICSS behavior elicited from electrodes in the NAC was facilitated by morphine injected in a dose of 50 ng or higher into the VTA using the behavioral threshold method (Van Wolfswinkel and Van Ree, 1985a). Interestingly, morphine injected into the NAC did not affect ICSS elicited from the VTA (Van Wolfswinkel and Van Ree, 1985a). Injection of specific μ-, δ-, or κ-opioid receptor ligands, DAMGO, [d-Ala²,d-Met⁵]-enkephalin, and dynorphin B, respectively, into the VTA facilitated ICSS from this area, whereas the same ligands were ineffective when injected into the MFB (Singh et al., 1994). A decrease of threshold for ICSS from the VTA was found when the μ-opioid receptor agonist DAMGO or the δ-opioid receptor agonist DPDPE was injected into the NAC. This effect was blocked by peripheral administration of the δ-antagonist naltrindole (Duvauchelle et al., 1996; Duvauchelle et al., 1997). Finally, morphine injected into the medial prefrontal cortex did not modify ICSS from this area (Shaw et al., 1984). Taken together, the data collected so far provide evidence that the VTA is a sensitive site for morphine and other opioids in facilitating ICSS reinforcement, although this may not be the only brain site.

With respect to the neurochemical systems involved in opioid-induced facilitation of ICSS, little information is available with the exception of the endogenous opioids and DA systems (see VII. Brain DA and Opioid Drugs of Abuse). Blockade of NMDA receptors by MK-801 potentiated the morphine-induced facilitation of ICSS (Carlezon and Wise, 1993a).

B. Endogenous Opioids

A useful approach to investigate the role of endogenous opioids in certain behaviors is to analyze the effects of opioid antagonists on that behavior. A number of studies have been performed dealing with opioid antagonists and ICSS. The first reports were on the opioid antagonist naloxone and ICSS for electrodes implanted in the MFB/LH area (Wauquir et al., 1974; Holtzman, 1976; Goldstein and Malick, 1977;Van der Kooy et al., 1977). In general, no marked effects of naloxone were found. In contrast, a large decrease of ICSS behavior by naloxone was reported when the electrodes were implanted in the central gray area of the midbrain (Belluzzi and Stein, 1977). After these initial reports, several studies have addressed the reason for these equivocal results. The data obtained have extensively been discussed in a review by Schaefer (1988).

It appears that the effect of opioid antagonists depends on the site of the stimulation electrode. In addition to the central gray area, decreases after naloxone have been reported when the electrodes were implanted in the locus ceruleus, substantia nigra, septal area, paratenial nucleus of the thalamus, dentate gyrus, NAC, medial entorhinal cortex, or amygdala (for references, see Schaefer, 1988;Trujillo et al., 1989a). The effects were, however, not always consistently found among the different laboratories and the effective dose of naloxone varied between less than 1 mg/kg to 10 mg/kg. It seems that the MFB/LH area is the least reliable site for the effects of opioid antagonists. Other factors influencing the effect of naloxone on ICSS are the amount of effort required of the animal but not the response difficulty (Trujillo et al., 1989c) and the schedule of reinforcement. The studies cited above used the continuous reinforcement schedule. When FR schedules were used, opioid antagonists produced marked dose-dependent decreases in the rate of lever-pressing (Schaefer and Michael, 1981, 1985; West et al., 1983). Also using this procedure, it appeared that the central gray area was a much more sensitive site for opioid antagonists than the MFB/LH area. Furthermore, the effectiveness of naloxone increased when the FR requirement was raised. The effect of naloxone persisted for about 2 h, which is consistent with the duration of action of this drug. Thus, intermittent reinforcement schedules can reliably disclose the effects of naloxone (Franklin and Robertson, 1982).

Another interesting observation is that the effect of opioid antagonists seem to be stronger in longer than in shorter test sessions of ICSS. This could indicate that the antagonists block the reinforcing value of ICSS, resulting in an extinction-like pattern of responding (Katz, 1981; Trujillo et al., 1989b). Only a few studies have been performed using rate-independent procedures of ICSS. Comparing these procedures, measuring the threshold for ICSS in rats with electrodes in the VTA, it was consistently found that a rather high dose of naloxone (10 mg/kg s.c.) raised the threshold for ICSS (Van Wolfswinkel and Van Ree, 1985b). Using the behavioral threshold procedure and testing and treating the animals repeatedly with naloxone, it appeared that the naloxone-induced increase of threshold became more pronounced during the 3 weeks of the experiment (Van Wolfswinkel et al., 1985). Interestingly, this effect persisted for at least 3 days after discontinuation of naloxone treatment. It was concluded that blockade of opioid receptors may induce long-term changes in the setpoint of ICSS. Accordingly, continuous s.c. administration of naloxone shifted ventral tegmental ICSS rate-frequency curves to the right, without suppressing behavioral performance (Hawkins and Stein, 1991). Since the acute effect of naloxone on the threshold was lower in animals more experienced with ICSS behavior, a study was performed on acquisition of the behavioral threshold procedure (Van Wolfswinkel and Van Ree, 1985c). It was found that this acquisition was disrupted by repeated treatment with naloxone, whereas the acquisition of a comparable food reinforced behavior was not affected by naloxone treatment. It has been argued that these data are consistent with those obtained with the FR schedule of reinforcement (Schaefer, 1988).

Most studies have used the antagonist naloxone. But similar effects have been reported with naltrexone and diprenorphine (Schaefer and Michael, 1981, 1985, 1988a). With respect to diprenorphine, opposite effects, i.e., an increase of responding for ICSS, have been found as well (Pollerberg et al., 1983; LaGasse et al., 1987). The effects of opioid antagonists are likely to be produced in the CNS, since both naloxone methobromide and naltrexone methobromide, compounds that rarely cross the blood-brain barrier, were without effect after systemic administration (Schaefer and Michael, 1985; Trujillo et al., 1989a). Some studies have been performed with mixed opioid agonist-antagonists. Decreased responding for ICSS has been reported for cyclazocine, nalorphine, and pentazocine (Holtzman, 1976; Schaefer, 1988). However, in another study, a lowering of the threshold for ICSS after nalbuphine or pentazocine was observed and no changes in threshold after cyclazocine or nalorphine (Kornetsky et al., 1979). Intracerebroventricular administration of high doses of the δ-opioid antagonist naltrindole, but not the μ-antagonist d-Tic-Cys-Tyr-d-Trp-Arg-Thr-Pen-Thr-NH₂ and the κ-antagonist nor-BNI raised the ICSS threshold (Carr and Papadouka, 1994; Carr, 1996).

Chronic food restriction facilitated ICSS from the LH, as has been shown for drug self-administration (Carr and Wolinsky, 1993; Carr, 1996). This facilitatory effect was blocked by i.c.v. naltrexone, the μ antagonist TCTAP, and the κ antagonist nor-BNI, suggesting that μ-, and κ-opioid receptors are involved in this facilitation (Carr and Papadouka, 1994; Carr, 1996).

Strains of rats, selectively bred for high versus low rate of lateral hypothalamic ICSS were analyzed for their density of μ-opioid receptors in discrete brain areas using the ligand [³H]DAMGO and in vitro autoradiography. The high-rate animals showed a higher and lower density of μ-opioid receptors in the ventral hippocampus and NAC, respectively, as compared to the low-rate animals (Gross-Isseroff et al., 1992). In addition, there is some evidence that endogenous opioids are released during self-stimulation of the VTA, as measured by the in vivo receptor occupancy procedure (Stein, 1993).

In conclusion, there seems to be evidence that endogenous opioid systems are involved in ICSS. The data collected so far point to a modulatory role rather than that reward from ICSS is mediated by endogenous opioids. More studies are needed, in particular after chronic blockade of endogenous opioids, to delineate more precisely the significance of endogenous opioids for ICSS.

A. Opioid Place Preference

Beach (1957) was the first to report that morphine elicits conditioned place preference. In that study it was shown that in rats made physically dependent upon morphine, administration of morphine during extensive training (12–22 days of conditioning sessions, using training doses of 5–20 mg/kg morphine, injected either s.c. or i.p.) resulted in preferences for the previously nonpreferred side of a test box. During preference testing, morphine was still administered to the animals. Interestingly, a place preference was observed both when, according to the conditioning schedule, the animals were expecting an injection with morphine in the conditioned compartment (“needing morphine”) or when they had been injected with morphine 10 min to 4 h before the test session (“sated for morphine”). These results suggested that both relief from morphine withdrawal and morphine’s positive affective properties could contribute to the establishment of conditioned place preference. The place preference induced by morphine withdrawal relief appeared to persist for 3 weeks. These findings were replicated in a later study investigating the involvement of monoamines in withdrawal relief-induced conditioned place preference (Schwartz and Marchok, 1974). In the late 1970s, morphineinduced conditioned place preference was first reported in animals not previously made physically dependent on morphine (Rossi and Reid, 1976; Katz and Gormezano, 1979). It was observed that when rats were conditioned at times when morphine (10 mg/kg s.c.) was expected to facilitate ICSS (1–4.5 h, but not 7 h postinjection), conditioned place preference was induced (Rossi and Reid, 1976). Another study showed that as few as three conditioning sessions with morphine or an enkephalin analog, administered i.c.v., were sufficient to produce place preference (Katz and Gormezano, 1979).

Using morphine and naloxone as conditioning drugs, Mucha and colleagues (Mucha et al., 1982; Mucha and Iversen, 1984; Mucha and Herz, 1986) have systematically investigated several methodological variables that can influence opioid-induced place-conditioning, e.g., dose of drug, route of administration, trial duration, number of conditioning trials, and stereospecificity of the opioids. Using four conditioning trials and i.v. administration of morphine, significant place preference was found with doses ranging from 0.08 to 10 mg/kg. Trial duration of 10 to 90 min induced similar levels of place preference. It appeared that one conditioning trial with 4 mg/kg morphine was sufficient to induce place preference. When morphine was administered s.c., place preference was found with 0.2 to 5 mg/kg, whereas 0.04 mg/kg was ineffective. Naloxone induced place aversion, in doses ranging from 0.02 to 2 mg/kg, and 0.1 to 45 mg/kg, when administered s.c. or i.p., respectively. Upon s.c. administration, three trials with morphine (1 mg/kg) or naloxone (0.5 mg/kg) were necessary to induce a significant place preference or aversion, respectively (Mucha et al., 1982; Mucha and Iversen, 1984). The development of place preference induced by morphine (0.5–2 mg/kg i.v.) was inhibited by naloxone (2 mg/kg i.p.) (Mucha et al., 1982). Stereospecificity of opioid-induced place-conditioning was demonstrated using levorphanol, which, in contrast to its inactive stereoisomer dextrorphan, induced place preference (Mucha et al., 1982; Mucha and Herz, 1986). In addition, although conditioning with (−)-morphine and (−)-naloxone caused place preference and aversion, respectively, (+)-morphine and (+)-naloxone were ineffective (Mucha et al., 1982).

In a follow-up study, the influence of environmental novelty and interoceptive states on morphine-induced place preference was investigated (Mucha and Iversen, 1984). It appeared that animals conditioned with morphine (4 trials, 1 mg/kg s.c.) and tested after injection of saline or morphine (1 mg/kg s.c.) displayed nearly identical levels of place preference. This seems to rule out any effects of state-dependent learning on the expression of morphine-induced place preference. Morphine-induced place-conditioning was also performed in a three-compartment apparatus, with one compartment being completely novel to the animals on the test day. Here, a clear preference for the morphine-paired side, over both the novel and the familiar saline-paired part of the apparatus, was observed. In this experiment, the animals spent more time (albeit not statistically significant) in the novel compartment, as compared to the saline-paired environment. In a subsequent experiment, rats were placed four times in one side of the two-compartment apparatus without any injections, and no preference for the novel or familiar side was found. However, when conditioning was performed with morphine, without intervening saline trials, a clear preference of the morphine-paired over the novel environment was found on the test day. These findings seem to exclude any major influence of exploration or environmental novelty on morphine-induced conditioned place preference. In this study, it was also shown that testing at two postconditioning intervals (1 day or 1 month) did not affect the strength of morphine-induced placeconditioning (Mucha and Iversen, 1984).

The studies described above clearly demonstrate that morphine reliably induces conditioned place preference and naloxone place aversion (Mucha et al., 1982; Mucha and Iversen, 1984; Mucha and Herz, 1986). These effects are most likely mediated through opioid receptors, with only marginal interference of state-dependent learning and environmental novelty or familiarity.

B. Variables Interfering with Opioid Place Preference

A meta-analysis has been conducted on conditioned place preference studies with morphine and heroin (as well as cocaine and amphetamine) in rats, published between 1979 and 1992 (Bardo et al., 1995). The influence of a variety of experimental factors was analyzed. These included drug dose, route of administration, number of conditioning trials, trial duration, test duration, drug compartment, number of apparatus compartments, and use of intervening saline trials or a preconditioning test, as well as sex, strain, and housing conditions of the animals. The data revealed no consistent influence of the sex of the animals used. When the data were analyzed for strain effects, Sprague-Dawley and Wistar rats were found to be more sensitive to the place-conditioning effects of morphine and heroin than other strains. With regard to social circumstances, individual housing at the time of the experiment appeared to enhance the sensitivity for heroin-induced place-conditioning. It has, however, also been described that isolation rearing made rats less sensitive to the ability of morphine and heroin to establish conditioned place preference (Schenk et al., 1983, 1985; Wongwitdecha and Marsden, 1996). Indeed, it was recently shown that rats reared in enriched environments were more sensitive to morphine-induced place-conditioning than animals reared under impoverished circumstances, which included social isolation (Bardo et al., 1997). With respect to social status, dominant animals seem to be more sensitive to the place-conditioning effects of morphine than their submissive counterparts (Coventry et al., 1997). There is, as yet, only one report on the effects on prenatal morphine treatment on place-conditioning. In that study, rats prenatally exposed to morphine appeared to be more sensitive to the place preference-inducing effects of morphine (Gagin et al., 1997). With respect to the effects of stress, it appeared that uncontrollable, but not escapable, footshock stress potentiated the effects of morphine on place-conditioning (Will et al., 1998).

Dose dependence was found for both morphine, with doses above 1 mg/kg generally producing place preference, and heroin, with doses above 0.3 mg/kg generally producing place preference. Although for heroin a relationship between drug dose and effect magnitude was found, such a relationship was much less clear for morphine. Regarding the route of administration, it was found that for morphine, s.c., as compared to i.p. and i.v., administration was slightly more effective. In the case of heroin, i.p. administration appeared to be much more effective than s.c. administration in inducing conditioned place preference. Regarding the number of conditioning trials, no clear picture emerged. This was due to the fact that in the case of morphine, a single i.v. administration had been reported to be effective in inducing conditioned place preference (Mucha et al., 1982; Bardo and Neisewander, 1986), whereas for heroin, only reports using three or four trials were available for analysis. In general, a longer conditioning trial duration (45 min or more for morphine, 25 min or more for heroin) appeared to be more effective, although in this case the analysis was quite complicated. For example, upon i.v. morphine administration, 10- or 90-min conditioning trials did not induce different levels of place preference (Mucha et al., 1982), whereas in the case of heroin, an inverted U-shaped relationship appears to exist: 10- or 100-min trials being less effective than 30-min trials (Bozarth, 1987a). With regard to experimental design, it also appeared that the test duration (10–30 min) had no marked influence on the expression of conditioned place preference. Counterbalanced administration of the conditioning drugs (as opposed to administering the drugs in the least preferred compartment in a biased design, or in the white compartment of a black-white apparatus) enhanced the strength of the place-conditioning, as did the use of saline trials in the nondrug-paired environment between conditioning trials (e.g., Scoles and Siegel, 1986). A preconditioning test, used to determine initial preference of the animal for a certain compartment of the test apparatus, negatively influenced morphineinduced place-conditioning. Finally, in the case of heroin, the use of a three-compartment apparatus yielded a more sensitive design than a two-compartment apparatus (Bardo et al., 1995).

A. Psychostimulants

In a human study with cocaine abusers, it was found that chronic treatment with the opioid antagonist naltrexone reduced euphoria and the “crash” from an i.v. cocaine injection (Kosten et al., 1992, but see Walsh et al., 1996). This finding suggests that the endogenous opioid system may be involved in certain aspects of cocaine addiction. Results from animal studies wherein the effect of opioid blockade on cocaine self-administration was studied generally seem to confirm such an involvement (for review, see Mello and Negus, 1996). In rats trained to i.v. self-administer cocaine, systemic pretreatment with the opioid antagonists naloxone or naltrexone dose-dependently (0.1–10 mg/kg s.c.) decreased cocaine self-administration, supposedly by a decrease of the reinforcing effects of cocaine (Corrigall and Coen, 1991). Similarly, daily treatment with naltrexone (0.32–3.2 mg/kg i.v.) decreased cocaine’s reinforcing properties in monkeys (Mello et al., 1990). In another study, naltrexone was found to increase cocaine self-administration in trained rats, but only under certain conditions of food supply (Carroll et al., 1986a). The authors suggested that naltrexone either increased the reinforcing effects of cocaine, resulting in a higher cocaine intake, or decreased the reinforcing effects in which case more responding is needed to produce the same drug effect. Nonetheless, the study supports the existence of an effect of opioid blockade on cocaine self-administration. In contrast to the above-mentioned findings, other studies did not find a significant effect of naltrexone treatment on the rate or pattern of cocaine intake in rats and metamphetamine intake in rhesus monkeys (Harrigan and Downs, 1978b; Ettenberg et al., 1982;Hemby et al., 1996). Moreover, treatment with the pure opioid antagonist quadazocine failed to affect cocaine-reinforced responding in rhesus monkeys (Winger et al., 1992).

Beside pure opioid antagonists, mixed opioid agonist-antagonists are also able to antagonize the reinforcing effects of cocaine during self-administration. Daily and intermittant buprenorphine treatment (0.1–0.7 mg/kg i.v.) significantly suppressed cocaine self-administration by rhesus monkeys, even more potently than the pure opioid antagonist naltrexone (Mello et al., 1989, 1990, 1992, 1993b;Winger et al., 1992; Lukas et al., 1995). Similar effects of buprenorphine have been found in rhesus monkeys self-administering smoked cocaine and in rats and mice i.v. self-administering cocaine (Carroll et al., 1992; Comer et al., 1996; A. V. Kuzmin, M. A. F. M. Gerrits, E. E. Zvartau, J. N. Van Ree, unpublished data). Although the exact mechanisms by which buprenorphine reduces cocaine self-administration are unknown, it has been suggested that the μ agonistic properties of buprenorphine are important for its interactions with cocaine. When buprenorphine and naltrexone were administered simultaneously, naltrexone significantly attenuated buprenorphine’s suppressive effects on cocaine self-administration, probably through antagonism of the μ agonist component of buprenorphine (Mello et al., 1993c; A. V. Kuzmin, M. A. F. M. Gerrits, E. E. Zvartau, J. N. Van Ree, unpublished data). In addition, buprenorphine has κ antagonist effects, which might contribute to its suppressive effects on cocaine self-administration (Brown et al., 1991). Other mixed opioid agonists-antagonists, such as nalbuphine and butorphanol, also reduced cocaine self-administration in rhesus monkeys, but this effect was not selective since food self-administration also decreased in a dose-dependent manner (Winger et al., 1992; Mello et al., 1993a). In mice, treatment with butorphanol and nalbuphine decreased initiation of i.v. cocaine self-administration (A. V. Kuzmin, M. A. F. M. Gerrits, E. E. Zvartau, J. N. Van Ree, unpublished data). When tested against a scale of different cocaine unit doses, butorphanol produced a rightward shift in the unit dose-response curve for cocaine, indicating a decrease of the reinforcing effects of cocaine, whereas nalbuphine shifted the dose-response curve to the left. Coadministration of naloxone did not influence the effects of butorphanol, suggesting the involvement of κ receptors in this effect.

During the initiation phase of cocaine self-administration (i.e., in drug-naive animals), treatment with naltrexone (1 mg/kg) decreased cocaine intake in rats, presumably by an attenuation of the reinforcing effects of cocaine (De Vry et al., 1989a). This suppressive effect was found when a threshold dose of cocaine (0.16 mg/kg/inf) was used, but not when a higher cocaine unit dose was available. In fact, naltrexone caused a rightward shift in the dose-response curve for cocaine, indicating that cocaine is less reinforcing after opioid blockade. A similar shift in dose-response curve for cocaine has been observed in mice treated with naloxone (0.01–1.0 mg/kg) (Kuzmin et al., 1997a). That naltrexone treatment was effective within a critical cocaine unit dose range is supported by the finding that naltrexone decreased cocaine self-administration at unit doses of 0.1 and 0.3 mg/kg/inf, but not at 1.0 mg/kg/inf (Corrigall and Coen, 1991). The proposed involvement of opioid systems in the reinforcing effects of cocaine is also supported by the observation that chronic treatment with naltrexone (10 mg/kg/day for 12 days) followed by a naltrexonefree interval facilitated the initiation of cocaine self-administration, probably by enhancing the reinforcing effects of cocaine (Ramsey and Van Ree, 1990). Suppression of cocaine intake after i.c.v. administration of naltrexone suggests that naltrexone exerts its effect on initiation of cocaine self-administration through an action on the CNS (Ramsey and Van Ree, 1991). With regard to the local opioid systems in the brain, treatment with naltrexone (1 μg/site) in the VTA, but not in the NAC, caudate putamen, central amygdala, or medial prefrontal cortex, attenuated cocaine self-administration behavior (Ramsey et al., 1999). Thus, opioid systems in the VTA may be implicated in modulating initiation of cocaine self-administration.

The above-mentioned effects on cocaine selfadministration have been found after blockade with nonselective μ-opioid antagonists such as naloxone and naltrexone. Recently, the effects of more selective opioid ligands on the reinforcing effects of cocaine were studied. After a number of days of stable cocaine intake by rats, acute blockade of the δ-opioid receptor by naltrindole (3–10 mg/kg) reduced the self-administration of cocaine (0.4 mg/kg/inf) (Reid et al., 1995). In another study, naltrindole (0.03–10 mg/kg) failed to affect cocaine self-administration at unit doses of 0.25 and 1.0 mg/kg/inf in trained rats (De Vries et al., 1995). In rhesus monkeys trained to self-administer cocaine, treatment with the δ-opioid antagonist naltrindole (0.1–3.2 mg/kg) for 10 consecutive days decreased cocaine intake, although in some monkeys the response rate for cocaine recovered to baseline levels during the last days of naltrindole treatment (Negus et al., 1995). The reduction of cocaine intake seems to be dependent on the dose of cocaine offered, since naltrindole effectively decreased the self-administration of 0.01 mg/kg/inf cocaine, but was ineffective or less effective in decreasing the self-administration of either higher or lower unit doses of cocaine. Thus, naltrindole may modulate the reinforcing effects of cocaine, probably by blocking δ-opioid receptors. The involvement of κ-opioid receptors in cocaine reinforcement has also been demonstrated. Treatment with the selective κ-opioid agonists U50,488H and spiradoline dose-dependently decreased cocaine selfadministration in rats (Glick et al., 1995, Kuzmin et al., 1997b). Although the κ antagonist nor-BNI had no effect on cocaine self-administration, it fully antagonized the effect of U50,488H (Glick et al., 1995). In monkeys, treatment with the κ agonists ethylketocyclazocine and U50,488H dose-dependently decreased cocaine self-administration of unit doses at the peak of the cocaine dose-effect curve (Negus et al., 1997). The effect of the κ agonist was blocked by the κ antagonist nor-BNI and naloxone and was accompanied by some undesirable effects. Interestingly, Kuzmin et al. (1997b) showed that treatment with U50,488H induced proper self-administration behavior with lower subthreshold unit doses of cocaine, doses that did not initiate self-administration under control conditions. In fact, the dose-response curve for cocaine reinforcement was shifted to the left, which may explain the observation of the decreased self-administration of cocaine by U50,488H when higher unit doses of cocaine were used (Glick et al., 1995, Kuzmin et al., 1997b). This latter finding suggests that activation of the κ-opioid systems increases the sensitivity for cocaine’s reinforcing effects. Accordingly, the κ-antagonist nor-BNI shifted the dose-response curve for cocaine reinforcement to the right. (Kuzmin et al., 1998). Thus, it seems that blockade of the μ-, δ-, or κ-opioid receptors, may make the animals less sensitive for cocaine reinforcement, whereas activation of κ-opioid receptors may result in the opposite. The results obtained so far stress the necessity of complete dose-response studies with respect to cocaine reinforcement before definitive conclusions are drawn.

Some internal and external factors affecting ICSS may use endogenous opioid systems. The potentiating effect of chronic food restriction on ICSS was reversed by naloxone and the μ antagonist TCTAP and the κ antagonist nor-BNI (Carr and Simon, 1984; Carr and Wolinsky, 1993; Carr and Papadouka, 1994). Also, the decrease in response rate of ICSS induced by inescapable footshock was antagonized by naloxone (Kamata et al., 1986). As already described before, drugs of abuse in general facilitate ICSS. The threshold-lowering effect induced by cocaine, amphetamine, or phencyclidine was reversed by naloxone treatment (Esposito et al., 1980; Kornetsky et al., 1981a,b.; Bain and Kornetsky, 1987; Van Wolfswinkel et al., 1988). Similar effects have been reported with respect to increase in response rate by amphetamine (Holtzman, 1974). Also, under the condition of a fixed interval schedule of reinforcement, naloxone attenuated the increase in response rate induced by amphetamine but not that by phencyclidine (Schaefer and Michael, 1990). The cocaine-induced facilitation of ICSS from the medial forebrain bundle was blocked by systemic treatment with the δ-opioid antagonist naltrindole (Reid et al., 1993). Thus, endogenous opioids may play a mediatory role for the effects of some drugs of abuse on ICSS.

Although the place preference induced by opioid drugs has amply been demonstrated to be mediated through opioid receptors, this has been shown for other drugs as well. In the case of cocaine, the development of conditioned place preference was prevented by coadministration of low doses of naloxone or naltrexone (Houdi et al., 1989; Suzuki et al., 1992b; Gerrits et al., 1995; Kim et al., 1997; Kuzmin et al., 1997a). Methadone appeared to enhance cocaine’s place-conditioning effects, whereas the opioid mixed agonist-antagonist buprenorphine has been reported to both block and enhance cocaineinduced place preference conditioning (Brown et al., 1991; Kosten et al., 1991; Bilsky et al., 1992; Suzuki et al., 1992b). Apart from blocking the development, naloxone also blocked the expression of cocaine-induced conditioned place preference (Gerrits et al., 1995). The lowest effective doses of naloxone were 0.032 and 0.1 mg/kg s.c. for expression and development of cocaine-induced place preference, respectively (Gerrits et al., 1995).

δ-Opioid receptors have also been suggested to be involved in cocaine-induced place-conditioning, as naltrindole (nonspecific δ antagonist), naltriben (δ₂ antagonist), but not BNTX (δ₁ antagonist) appeared to block the development of cocaine-induced place preference (Menkens et al., 1992;Suzuki et al., 1994a). However, a detailed study on the influence of naltrindole reported no effect of naltrindole on cocaine-induced place-conditioning (De Vries et al., 1995). Recently, i.c.v. administration of an antisense oligodeoxynucleotide to δ receptors was found to inhibit cocaine-induced place preference (Suzuki et al., 1997b). The κ agonist U50,488H could also block cocaine-induced place preference (Suzuki et al., 1992b; Crawford et al., 1995). In addition to attenuating cocaine-induced place preference, naloxone, naltrindole, and naltriben also blocked amphetamine-induced place preference (Trujillo et al., 1991; Suzuki et al., 1994a). Similar to morphine pretreatment (see V. Tolerance, Physical Dependence, and Sensitization), pretreatment with cocaine has been shown to induce sensitization to the place-conditioning effects of cocaine and morphine in rats. This sensitization induced by preexposure to cocaine could be attenuated by coadministration of the δ antagonist naltrindole and the κ agonist U50,488H or U69,593 (Shippenberg and Heidbreder, 1995a,b; Shippenberg et al., 1996b, 1998; Shippenberg and Rea, 1997). Remarkably, the sensitization of the place-conditioning effects of morphine and cocaine induced by morphine pretreatment could not be attenuated by U69,593 coadministration (Shippenberg et al., 1998).

Thus, stimulation of opioid receptors seems to be involved in psychostimulant-induced place-conditioning. The involvement of δ-opioid receptors is not undebated, but the low doses of naloxone and naltrexone sufficient to inhibit cocaine- and amphetamine-induced place-conditioning suggest a prominent role for μ receptors. In addition, δ- and κ-opioid receptors seem to be involved in the cocaine-induced sensitization to the place-conditioning effects of cocaine.

Studying the neostriatum of human subjects with a history of cocaine dependence, it was found that cocaine dependence was linked with selective alterations in striatal opioid mRNA expression and opioid receptor binding (Hurd and Herkelham, 1993). Reductions in the levels of enkephalin mRNA and μ-opioid receptor binding were found in the striatum concomitant with elevations in levels of dynorphin mRNA and κ-opioid receptor binding. Using positron emission tomography, it was found that μ-opioid receptor binding was increased in several brain regions of cocaine addicts (Zubieta et al., 1996). The change in binding was positively correlated with the degree of cocaine craving the addicts experienced. In cocaine overdose addicts, an increase in κ₂-opioid receptors was found in the NAC and amygdala (Staley et al., 1997). Although the functional relationship between these alterations and cocaine dependence is not clear, this finding provides neurochemical evidence for an involvement of endogenous opioid systems in cocaine dependence.

A number of studies have investigated the effect of administration of psychostimulants in rats on the levels of endogenous opioids, expression of opioid mRNA and opioid receptors in the brain (for review, see Trujillo et al., 1993). Investigations on the action of psychostimulant drugs on the β-endorphin system are limited. It has been reported that acute and chronic treatment with cocaine induced an increase in levels of βE-IR in plasma and pituitary (Moldow and Fischman, 1987; Forman and Estilow, 1988). Furthermore, chronic treatment with cocaine induced an increased release of β-endorphin from the pituitary in vitro (Forman and Estilow, 1988). No effect of chronic treatment with cocaine and amphetamine on the levels of βE-IR in the hypothalamus were found (Harsing et al., 1982; Agarwal et al., 1985; Forman and Estilow, 1988).

Reports on the enkephalin system are consistent in demonstrating a lack of effect of acute and chronic treatment with psychostimulants on enkephalin immunoreactivity in the striatum and in other brain structures such as cortex, hippocampus, hypothalamus, and brain stem (Harsing et al., 1982; Sivam, 1989; Li et al., 1990; Trujillo et al., 1990). In contrast to a lack of effect on levels of enkephalin peptides in the brain, acute cocaine and metamphetamine increased the expression of enkephalin mRNA in the striatum and amygdala (Bannon et al., 1989;Cohen et al., 1991; Hurd et al., 1992; Wang and McGinty, 1995). Subchronic or chronic treatment did not, however, affect the enkephalin mRNA expression (Sivam, 1989; Branch et al., 1992; Daunais and McGinty, 1994). The correlation between changes in expression of enkephalin mRNA and levels of enkephalin peptides needs, however, to be elucidated, including its relevance for cocaine reinforcement and dependence.

The effect of acute administration of psychostimulants, such as cocaine, amphetamine, and metamphetamine, on the dynorphin system has been examined extensively, and the results are equivocal. The investigations have primarily been focused on the striatonigral dynorphin system. Summarizing, acute administration of psychostimulants increased, decreased, or had no effect on dynorphin immunoreactivity levels in the striatum and/or substantia nigra (Peterson and Robertson, 1984; Hanson et al., 1988; Li et al., 1988; Sivam, 1989; Trujillo et al., 1990; Johnson et al., 1991; Singh et al., 1991). Moreover, acute treatment increased or did not affect the expression of dynorphin mRNA in the striatum and NAC (Hurd et al., 1992; Daunais and McGinty, 1994;Wang and McGinty, 1995). More consistent effects were found after subchronic or repeated treatment with psychostimulants; increased striatal levels of dynorphin immunoreactivity and dynorphin mRNA levels have been reported (Peterson and Robertson, 1984, Li et al., 1986,1988; Hanson et al., 1987, 1988, 1989; Sivam, 1989; Trujillo and Akil, 1989, 1990; Smiley et al., 1990; Trujillo et al., 1990; Gerfen et al., 1991; Hurd et al., 1992; Steiner and Gerfen, 1993; Daunais and McGinty, 1994; Smith and McGinty, 1994). In addition, increased dynorphin levels were demonstrated in the substantia nigra, NAC, but not hippocampus (Sivam, 1989; Smiley et al., 1990; Trujillo et al., 1990).

Studies on animals self-administering cocaine shed some more light on the potential involvement of β-endorphin in processes underlying cocaine dependence. Sweep et al. (1988, 1989) found marked decreases in βE-IR levels in the anterior part of the limbic system (i.e., NAC, septum, hippocampus, and rostral striatum) in animals self-administering cocaine. Furthermore, using an in vivo autoradiographic technique, a decrease in opioid receptor occupancy was found in restricted subcortical brain regions of animals self-administering cocaine, including limbic areas (i.e., lateral septum, ventral pallidum, nucleus stria terminalis, and amygdala) and some regions of the hypothalamus and thalamus (Gerrits et al., 1999). The decrease in opioid receptor occupancy is probably due to a release of endogenous opioids in these particular brain regions. Interestingly, both of these changes (i.e., decreased βE-IR content and increased endogenous opioid release) were present just before a scheduled next cocaine self-administration session would have taken place; thus, when the desire or need for the drug is assumed to be high. This might suggest an involvement of endogenous opioids, and possibly β-endorphin, in the processes underlying the need for cocaine. When the same methodologies were used in rats that just had completed their daily cocaine self-administration session, it appeared that βE-IR levels in the brain were hardly changed and that a decrease of opioid receptor occupancy was present in many brain areas, including the NAC (Sweep et al., 1988; Gerrits et al., 1999). Daunais et al. (1993), investigating the effect of cocaine self-administration on the expression of dynorphin mRNA, found an increased expression in the patch-like areas of the dorsal, but not in those of the ventral, striatum. Because repeated high doses of cocaine given for 6 to 7 days were necessary to induce this effect, the authors concluded that the increased dynorphin mRNA expression does not underlie the acute reinforcing effects of cocaine but is more associated with long-term adaptation and sensitization.

Besides regulating the level of endogenous opioids and the expression of opioid mRNA in the brain, studies have demonstrated that psychostimulant drugs also regulate the density of opioid receptors in the brain. Chronic cocaine exposure reduced opioid receptor density, as labeled by [³H]naloxone, in the hippocampus and striatum and in the basolateral nucleus of the amygdala, VTA, substantia nigra, and dorsal raphe nuclei (Ishizuka et al., 1988;Hammer, 1989). An increase in opioid receptor density was found in the NAC, ventral pallidum, and lateral hypothalamus. In light of the selective opioid receptors in the brain, studies were performed with more selective radioligands. In short, chronic treatment of rats with cocaine caused an up-regulation of μ-opioid receptors in the cingulate cortex, NAC, rostral caudate putamen, and basolateral amygdala; an up-regulation of κ-opioid receptors in the cingulate cortex, rostral caudate putamen, olfactory tubercle, and VTA; and no change in δ-opioid receptors in any of the brain regions examined (Unterwald et al., 1992, 1994). Itzhak (1993), studying the effect of chronic cocaine treatment on opioid receptor densities in guinea pigs, found a significant down-regulation of μ-opioid receptors in frontal cortex, amygdala, thalamus, and hippocampus; an alteration in the expression of κ-opioid receptors in the cerebellum; and no significant changes in δ-opioid receptor expression. Furthermore, it was shown that “binge” pattern of cocaine administration led to significant decreases in the level of κ-opioid receptor mRNA in the substantia nigra but not in the caudate putamen (Spangler et al., 1997). Finally, to determine the functional consequences of chronic cocaine on opioid receptors, Unterwald et al. (1993) measured changes in adenyl cyclase activity. They found that chronic cocaine administration resulted in a selective impairment of δ-opioid receptor-mediated function in the caudate putamen and NAC.

In conclusion, a role for μ-opioid receptors and βendorphin in cocaine dependence seems likely. This role may be at the level of modulating the reinforcing action of cocaine and the motivational state induced by repeated cocaine exposure. Concerning the κ- and δ-opioid receptors and the dynorphin and enkephalin systems, the data so far do not allow definitive conclusions. In particular, it is not clear how to link effects of passive administration of cocaine with the addiction, e.g., self-administration process.

B. Ethanol

In 1970, a biochemical link was proposed between ethanol and opioid systems based on the finding that condensation of the ethanol metabolite acetaldehyde and biogenic amines produced tetrahydroisoquinolines (TIQs). These TIQs seemed to have opioid-like effects (Davish and Walsh, 1970; Cohen and Collins, 1970; Fertel et al., 1980). Long-term ethanol self-administration induced the formation of TIQs in the brain of rats (Collins et al., 1990; Haber et al., 1996). Furthermore, TIQs induced excessive alcohol drinking, an effect that was modulated by morphine and naloxone (Critcher et al., 1983). Evidence for an involvement of the endogenous opioid systems in ethanol reinforcement and addiction is provided by studies with opioid antagonists and agonists (for reviews, see Herz, 1997; Spanagel and Zieglsgängsberger, 1997).

Opioid antagonists, such as naltrexone and naloxone, decrease ethanol self-administration in rodents and monkeys under a variety of different experimental conditions. Although opioid blockade by antagonists blocked intragastric (Sinden et al., 1983) and i.v. (Altshuler et al., 1980; Martin et al., 1983) self-administration of ethanol, the majority of studies on the involvement of endogenous opioids in ethanol reinforcement have used an oral self-administration paradigm. Altshuler et al. (1980), using rhesus monkeys experienced with alcohol intake, found that chronic treatment with naltrexone (1–3 mg/kg i.m.) on a daily basis for 15 days dosedependently decreased i.v. ethanol administration by as much as 50%. A later study with alcohol-drinking rhesus monkeys supported this finding in that the total oral ethanol intake was reduced by acute treatment with naltrexone in a graded dose-dependent manner (0.02–1.5 mg/kg i.m.). The consumption of drinking water was much less affected by naltrexone (Kornet et al., 1991). Using different variants of ethanol-water choice procedures, treatment with low doses of naloxone (0.1–1 mg/kg) selectively and dose-dependently decreased the preference for ethanol in rats (De Witte, 1984; Sandi et al., 1988; Schwartz-Stevens et al., 1992). Evidence has been presented that the amount of ethanol intake could be decreased without altering the water intake. Furthermore, the decrease in ethanol intake was found to be independent of palatability of the presented alcohol (i.e., alcohol mixed with saccharin or quinine). A lack of blockade of ethanol intake by the peripherally acting opioid antagonist methylnaltrexone indicates a central site of action of opioid blockade of ethanol intake (Linseman, 1989). A number of other reports confirmed the decreasing effect of opioid antagonists on ethanol intake (e.g., Marfaing-Jallat et al., 1983; Reid and Hunter, 1984; Samson and Doyle, 1985; Hubbell et al., 1986, 1991; Mason et al., 1993; Froehlich, 1995; Ulm et al., 1995; Davidson and Amit, 1996).

The effect of opioid blockade was found in nondeprived animals and during conditions of continuous and concurrent supply, but has also been investigated in alcohol-abstinence studies. In rhesus monkeys who had about 1 year of experience with alcohol drinking, short and longer periods of imposed interruptions of alcohol supply (up to 7 days) led to a temporary increase in ethanol intake (“catch-up” phenomenon) and a subsequent relapse in the preinterruption drinking habit (Kornet et al., 1990). Blockade of opioid receptors with naltrexone after 2 days of imposed abstinence dosedependently reduced the abstinence-induced increase in ethanol intake after renewed presentation of ethanol (Kornet et al., 1991). Interestingly, a lower dose of naltrexone (0.17 mg/kg i.m.) was effective in reducing ethanol intake after imposed abstinence as compared to during continuous supply of alcohol, suggesting a role for endogenous opioids in the catch-up phenomenon (Kornet et al., 1991). Reid et al. (1991) used a different regimen of imposed abstinence in rats involving 22 h of deprivation of fluids followed by 2 h of access to water and a sweetened ethanol solution. Treatment with naloxone (4 mg/kg) 30 min before a day’s opportunity to take fluids decreased the intake of ethanol, whereas an injection with naloxone 4 h before alcohol supply increased ethanol intake. Taken together, the results from the studies with opioid blockade suggest an involvement of endogenous opioids in ethanol reinforcement.

The effect of more selective opioid receptor antagonists on ethanol intake has mostly been studied in selectively bred strains of high alcohol-drinking or -preferring rats (for review, see Froehlich, 1995). In these rat strains [e.g., high alcohol drinking (HAD) and the alko alcohol (AA)], treatment with the nonselective opioid antagonists naloxone and naltrexone dose-dependently decreased voluntary oral ethanol intake (Pulvirenti and Kastin, 1988; Froehlich et al., 1990;Hyytiä and Sinclair, 1993; Myers and Lankford, 1996). The selective μ-opioid antagonist CTOP, administered i.c.v., significantly decreased ethanol intake in AA rats. In the same rat strain, selective blockade of the δ-opioid receptor with ICI 174,864 or naltrindole had no effect on alcohol drinking (Hyytiä, 1993;Honkanen et al., 1996). Contrary to these findings, ICI 174,864 and naltrindole significantly decreased ethanol consumption as efficiently as naloxone in the high-drinking HAD rats (Froehlich et al., 1991;Froehlich, 1995). ICI 174,864 and naltrindole also suppressed ethanol intake in another rat strain with high alcohol preference (P-line). This effect of naltrindole was, however, not specific for ethanol, as evidenced by the fact that naltrindole reduced intake of saccharin solutions with and without ethanol (Krishnan-Sarin et al., 1995a). Furthermore, naltriben, an antagonist selective for the δ₂-opioid receptor, suppressed ethanol intake in rats of the P-line (Krishnan-Sarin et al., 1995b). This effect appeared to be specific for ethanol and independent of alcohol palatability. The involvement of μ- and δ-opioid receptors in ethanol reinforcement has also been investigated in the alcohol-preferring C57BL/6 mice. Naltrexone reduced ethanol intake in these mice, but the effect waned at increasing doses of naltrexone. Furthermore, chronic naltrexone treatment stimulated ethanol intake (Phillips et al., 1997). Administration of naltrindole decreased ethanol intake, but blockade of the μ-opioid receptor with β-funaltrexamine in these mice had no effect on alcohol consumption (Dzung et al., 1993). These data may suggest that in the AA rats the μ-opioid receptor is important in mediating ethanol reinforcement and in the HAD and P-line rats and C57BL/6 mice, the δ-opioid receptor. The effects of naltrexone and selective opioid receptor antagonists on ethanol consumption have also been studied in the Wistar rats (Stromberg et al., 1998). Naltrexone and the μselective opioid antagonist β-funaltrexamine significantly decreased the intake of an ethanol solution using a limited access procedure. Blockade of the δ-opioid receptor with naltrindole failed to significantly reduce ethanol consumption. These data suggest that in the outbred rat the μ-opioid receptor rather than the δ-opioid receptor is involved in ethanol reinforcement.

Bremazocine was found to potently suppress ethanol drinking in rats in a free-choice unlimited access model. The effect of bremazocine, which combines agonism at κ receptors and the μ-δ receptor complex, with antagonistic actions at μ receptors (Heijna et al., 1989; Schoffelmeer et al., 1992) was not secondary to effects on motor activity or fluid intake. In addition, bremazocine did not affect intake of a highly preferred sucrose solution. Both naltrexone and the κ agonist U50,488H had only modest and transient suppressant effects on ethanol drinking, suggesting a role of the μ-δ receptor complex in the effect of bremazocine (Nestby et al., 1999). Buprenorphine, was found to reduce i.v. ethanol self-injection in rats and oral ethanol intake in rhesus monkeys (Martin et al., 1983; Carroll et al., 1992). In the latter study, buprenorphine, however, also attenuated saccharin-maintained responding.

The involvement of endogenous opioid systems in ethanol consumption has also been studied using opioid agonists. However, the results with opioid agonists are less consistent than those with opioid antagonists. In general, low doses of morphine stimulated ethanol intake in animals (Reid and Hunter, 1984; Hubbell et al., 1986, 1987, 1993; Reid et al., 1991), whereas moderate to high doses of morphine have been reported to suppress alcohol consumption (Sinclair et al., 1973; Sinclair, 1974; Ho et al., 1976; Czirr et al., 1987; Linseman, 1989; Volpicelli et al., 1991; Schwartz-Stevens et al., 1992). An increase in ethanol intake after i.c.v. administration of morphine indicated that the low-dose effect of morphine is centrally located (Linseman and Harding, 1990). Some unclarity about the low-morphine dose effect exists since other studies showed that low doses of morphine hardly affected or decreased ethanol intake in rats and monkeys, respectively (Kornet et al., 1992b;Schwartz-Stevens et al., 1992).

Measuring the preference for alcohol, Volpicelli et al. (1991)demonstrated that morphine lowered alcohol preference. The suppression of alcohol preference was related to the dose in that a high dose of morphine suppressed alcohol preference more than a low dose of morphine. Similar impairments in the acquisition of alcohol preference were reported after systemic administration of endogenous opioids such as β-endorphin, Leu-enkephalin, and a synthetic analog of Met-enkephalin (Sandi et al., 1989, 1990a,b). Using a daily abstinence regimen of 22 h of deprivation and 2 h of access to fluids, a low dose of morphine administered 30 min before the daily opportunity to drink fluids increased the intake of ethanol, whereas morphine administered 4 h before renewed alcohol supply decreased alcohol drinking (Reid et al., 1991). Furthermore, it has been shown that morphine enhanced ethanol place preference (Marglin et al., 1988). The development of ethanol-induced place aversion in rats was enhanced by naloxone, whereas naloxone did not influence the expression of ethanol-induced place aversion (Bormann and Cunningham, 1997).

Based on, among others, the above-mentioned preclinical studies wherein opioid antagonists reliably reduce alcohol consumption under a variety of circumstances, clinical studies have been undertaken to assess the effect of naltrexone treatment in alcoholics. During a 12-week, double-blind, placebo-controlled trial, alcoholdependent patients were treated with naltrexone-hydrochloride (50 mg/day) in adjunct to psychosocial treatment following alcohol detoxification. Subjects taking naltrexone reported significantly less alcohol craving. The number of days in which alcohol was consumed was significantly decreased by naltrexone and relapse was reduced. Of the placebo-treated patients, 95% relapsed after they drank alcohol again, whereas only 50% of the naltrexone-treated patients exposed to alcohol relapsed (Volpicelli et al., 1992). Additionally, a majority of the naltrexone-treated patients reported that the “high” produced by alcohol was significantly less than usual (Volpicelli et al., 1995c). These findings were replicated and extended by O’Malley et al. (1992,1996) who, in addition, found that the reducing effects of naltrexone on alcohol drinking, craving, and relapse interacted with the type of supportive therapy the patients received. Naltrexone has recently received approval for the treatment of relapse in alcohol dependence, and thereby may offer a new treatment regimen in combination with psychosocial therapy to reduce relapse following alcohol detoxification (O’Malley, 1995; Swift, 1995; Volpicelli et al., 1995a,b). Another opioid antagonist, nalmefene, also has been reported to reduce alcohol consumption and to prevent relapse (Mason et al., 1994). Furthermore, naltrexone increased the latency to drink alcohol in social drinkers (Davidson et al., 1996, but see Doty and De Wit, 1995). The interaction between naltrexone and the subjective alcohol response seemed to depend on the degree of being at risk for alcoholism (King et al., 1997).

A possible involvement of endogenous opioids in alcohol addiction is supported by an early study demonstrating a 3-fold lower level of β-endorphin in the cerebrospinal fluid of abstinent, chronic alcohol addicts as compared with controls (Genazzani et al., 1982). Over the years, the interaction between ethanol and the activity of endogenous opioid systems and its possible implication for ethanol reinforcement and dependence have been studied in animals, but the results of these studies have not yielded an unified theory about the role of endogenous opioids in ethanol dependence (for overview and details, seeGianoulakis, 1989; Froelich and Li, 1993; Gianoulakis, 1993; Trujillo et al., 1993; Tabakoff et al., 1996).

Studies examining the effect of ethanol on the β-endorphin system have shown a variety of effects on brain β-endorphin. Acute treatment with ethanol increased, decreased, or had no effect on β-endorphin content in the pituitary and hypothalamus (Schultz et al., 1980;Seizinger et al., 1983; Wilkinson et al., 1986; Patel and Pohorecky, 1989; Przewlocka et al., 1990). An increased in vivo release of β-endorphin from the pituitary and hypothalamus after acute ethanol has been demonstrated (Gianoulakis and Barcomb, 1987). With regard to chronic ethanol administration, also decreases, increases, or no effects on the β-endorphin levels were found in the pituitary and hypothalamus (Schultz et al., 1980; Cheng and Tseng, 1982; Seizinger et al., 1983). Similar discrepancies have been found for the effects of ethanol exposure on β-endorphin content in other brain regions (Schultz et al., 1980; Seizinger et al., 1983; Wilkinson et al., 1986;Przewlocka et al., 1990). Such inconsistencies were also found for the effects of ethanol exposure on brain content of enkephalin and dynorphin peptides and on the expression of opioid receptors in the brain (see Gianoulakis, 1989, 1993; Froehlich and Li, 1993; Trujillo et al., 1993). The discrepancy in ethanolinduced effects are in part due to differences in procedural variables (i.e., animal species examined, dose, route and duration of ethanol administration, areas of the brain examined, and whether ethanol-induced changes in opioid peptides were examined during or after ethanol administration). A different approach to study the link between endogenous opioids and ethanol reinforcement is the examination of strains of mice with different genetic propensities to drink alcohol. For example,Gianoulakis and Gupta (1986) demonstrated that the hypothalamic β-endorphin level was about 25% lower in the alcohol-nonpreferring DBA/2 mice as compared with the alcohol-preferring C57BL/6 mice. Moreover, the β-endorphin content in the hypothalamus decreased in the C57BL/6 mice in response to acute injection of ethanol but not in the DBA/2 mice. Additional studies showed that in C57BL/6 mice ethanol induced an enhanced in vitro release of hypothalamic β-endorphin, lower levels of β-endorphin in the NAC under basal conditions, and an increase in hypothalamic content of POMC-mRNA after 3 weeks of ethanol consumption (De Waele and Gianoulakis, 1993, 1994). Comparing alcohol-preferring AA and alcohol-avoiding ANA lines of rats, differences in the density of both μ- and δ-opioid receptors in distinct brain regions and in the dynorphin and enkephalin levels in the NAC were found (Nylander et al., 1994; De Waele et al., 1995). For a detailed discussion of genetically determined differences in the opioid system, we refer to the reviews of Gianoukalis and coworkers (Gianoukalis and De Waele, 1994; Gianoukalis et al., 1996). Subjects at high risk for alcoholism showed an increase in plasma levels of βE-IR upon administration of moderate doses of ethanol, whereas subjects at low risk did not respond in this way (NGianoukalis et al., 1996). In monkeys, differential effect on plasma β-endorphin levels in relation to the increase in alcohol consumption during initiation of alcohol self-administration has been reported (Kornet et al., 1992a). Alcoholism was accompanied by increased [³H]naloxone binding in several brain regions, particularly the frontal cortex (Ritchie and Noble, 1996). Thus, the genetic makeup of the endogenous opioid systems as well as the interaction between alcohol and these systems may contribute to the development of alcoholism.

There seems to be agreement on the assumption that ethanol drinking or administration stimulates the activity of the endogenous opioid systems that serves to reinforce further alcohol drinking and, in time, leads to the development of ethanol dependence. Two theories have been developed that focus on basal endogenous opioid activity as a “predisposing factor” for alcohol drinking and abuse. One theory, the “opioid deficit” or “opioid compensation” hypothesis predicts that a deficiency in endogenous opioids leads to alcohol-craving and increased alcohol-drinking (Blum, 1983; Erickson, 1990; Ulm et al., 1995). This theory assumes that because ethanol stimulates activity within the opioid system, ethanol is consumed to compensate for low basal levels of endogenous opioids. The other theory, the “opioid surfeit hypothesis” assumes that an excess of opioid activity leads to alcohol-craving and increased alcohol intake which is then reinforced by further ethanolinduced increases in opioid activity that culminates in ethanol dependence (Hunter et al., 1984; Reid et al., 1991). Little experimental evidence exists to substantiate either theory, but most findings so far can be best explained in the context of the opioid compensation hypothesis. That is, during conditions of excess opioid receptor activity, e.g., after treatment with morphine, alcohol consumption decreases. In contrast, during conditions with a deficiency in opioid activity, consumption of alcohol increases. A condition with a deficiency in opioid activity could be imposed abstinence. Short and longer periods of imposed interruptions of ethanol supply lead to a temporary increase in ethanol intake and a subsequent relapse in preinterruption drinking habit (Kornet et al., 1990). Blockade of opioid receptors with low doses of naltrexone reduced the abstinence-induced increase in ethanol intake, suggesting that craving and relapse are opioid-mediated (Kornet et al., 1991). A recent observation using an in vivo opioid receptor occupancy technique, showing that endogenous opioids were released in some limbic brain regions, i.e., the amygdala, hippocampus, ventral pallidum, nucleus stria terminalis, when the desire for ethanol is high in contrast to when the desire is probably low, agrees well with this hypothesis (Gerrits et al., 1999). In addition, it corroborates with recent findings in human alcoholics demonstrating that craving and relapse are attenuated after treatment with the opioid antagonist naltrexone (O’Malley et al., 1992; Volpicelli et al., 1992). Furthermore, some clinical evidence is available for an inverse relationship between alcohol and opiate use in heroin addicts (see Ulm et al., 1995).