Opposite regulation of conditioned place preference and intravenous drug self-administration in rodent models: motivational and non-motivational examples

Abstract

Although developed from a common antecedent, conditioned place preference (CPP) and intravenous drug self-administration (SA) represent different behavioral paradigms, each with strong face validity. The field has treated results from these studies largely interchangeably; however, there is considerable evidence of opposite modulation of CPP vs. SA. This review outlines four manipulations that differentially affect CPP and SA based on alterations of motivation. These examples are contrasted with one example of differential CPP and SA results that can be explained by simple parallel shifts in dose-response functions. The final two examples have yet to be classified as motivation-based or parallel shifts. Important aspects, including motivation, volitional control of drug administration, reward, and the role of cues are discussed. One major conclusion of this paper is that explanations for apparent discrepancies between CPP and SA require full dose effect functions and assessment of PR breakpoints. Overall, this manuscript offers a more nuanced insight into how CPP and SA can be used to study different aspects of substance use disorders.

Introduction

Although no animal model can recapitulate all aspects of drug abuse in humans, conditioned place preference (CPP) and intravenous drug self-administration (SA) procedures can be used to model many aspects of drug reward and reinforcement. With CPP, when human and non-human animals receive experimenter-delivered drug in one context and placebo in a different context, they show a preference for the drug-paired context when tested in a drug-free state. Since the drug delivery is not operant-based, the preference is referred to as drug “reward”, which contrasts it with “reinforcement” as used to define the response-contingent delivery of drug with self-administration, which has hard-to-define components of both incentive motivation (response initiation and strengthening) and reward (response-independent positive hedonic state). A number of comprehensive reviews are available that discuss the procedures, strengths and limitations of CPP (Bardo, Rowlett et al. 1995, Tzschentke 1998, Prus, James et al. 2009) and SA (Balster and Lukas 1985, Katz 1989, Panlilio and Goldberg 2007, O’Connor, Chapman et al. 2011), and much evidence from those reviews illustrates how results obtained with either CPP or SA are often consonant. Perhaps most compelling, with few exceptions, drugs that induce CPP are self-administered and vice-versa (Bardo and Bevins 2000). Additionally, several manipulations produce coordinated regulation of CPP and SA, such as lesions of the nucleus accumbens (NAc) that decrease both amphetamine CPP and SA (Lyness, Friedle et al. 1979, Spyraki, Fibiger et al. 1982), as well as similar disruption by dopamine transporter manipulations (Medvedev, Gainetdinov et al. 2005, Chen, Tilley et al. 2006, Thomsen, Hall et al. 2009). Moreover, different genetically inbred rodent strains such as abuse susceptible C57 vs. resistant DBA mice (Nocjar, Middaugh et al. 1999) or abuse susceptible Lewis rats vs. resistant Fischer 344 rats (Kosten, Miserendino et al. 1994, Horan, Smith et al. 1997, Brower, Fu et al. 2002, Philibin, Vann et al. 2005, Freeman, Kearns et al. 2009, Picetti, Ho et al. 2010) show consonance between CPP and SA, as do sex differences (Lynch and Carroll 1999, Cicero, Ennis et al. 2000, Carroll, Morgan et al. 2002, Russo, Jenab et al. 2003, Zakharova, Wade et al. 2009). Nonetheless, these are not redundant measures of abuse liability, and it has been shown that individual differences in CPP are not predictive of individual differences SA (Bardo, Valone et al. 1999).

In the current review, we attempt to avoid redundancy with the earlier extensive literature showing consistencies in CPP and SA results. Instead, the purpose is to describe two substantive areas of research in which there exists a major discrepancy between the results obtained with CPP and SA. These examples should be considered representative examples found in our and others’ research, but this list will not be exhaustive, as future research adds new examples. The first classification of examples relates to the dissociation of interoceptive reward cues (CPP) from motivation (SA). Of the four examples that will be discussed, one stems from environmental factors (enriched housing), two from genetic/molecular manipulations in vivo (CREB inhibition or G9a overexpression), and one from a pharmacological manipulation (clozapine treatment). The second classification of examples for the apparent discrepancy between CPP and SA may be attributed to simple parallel shifts in dose-response functions. Each of these areas of discrepancy will be outlined and then we will discuss how information gathered from these substantive areas may shed light on how CPP and SA differ in measuring drug abuse liability. Finally, some thoughts on the uses of each preparation in future research will be delineated.

It should be noted that every study has limitations, usually stemming from pragmatic compromises. Much of the information on which these conclusions are drawn is incomplete. For example, in investigating the effect of some manipulation of sensitivity to cocaine CPP, one has to add at least two groups of animals for each cocaine dose. Thus, much of the CPP data discussed below is drawn from one or two doses (Green, Alibhai et al. 2010, Rae, Zanos et al. 2018). Similarly, some manipulations such as herpes-based viral vectors only last for 4 days, meaning that reasonably only two doses can be assessed for PR schedules (Larson, Graham et al. 2011). These considerations add caveats to the conclusions drawn from the results; however, in aggregate, enough data has been produced for us to draw reasonable conclusions from the results discussed here.

Examples of discordant CPP vs. SA results based on motivation

Environmental manipulation has opposite effects on CPP and SA

Environmental enrichment increases CPP but decreases SA

The environmental enrichment paradigm is a developmental model of differential rearing in which adolescent rats are raised in one of two different environments: (1) an enriched condition (EC) consisting of group housing (8 – 12 per cage) in a large cage with multiple hard plastic objects replaced and rearranged daily for optimal novelty; or (2) an isolated condition (IC) where rats are housed individually with negligible social contact, exercise, or novelty. In many cases, a social condition (SC) consisting of either pair- or group-housing without novel objects is included for comparison in order to determine if the social peers and/or the novel objects contribute to any differences observed between EC and IC groups (Green, Cain et al. 2003, Gipson, Beckmann et al. 2011, Hofford, Darna et al. 2014, Hofford, Chow et al. 2017). These differential environments represent a non-drug, non-surgical, non-genetic manipulation of individual differences in response to drugs of abuse assessed in adulthood.

Initial studies using CPP indicated that environmental enrichment may increase vulnerability for addiction-related behaviors. These CPP studies showed that EC rats are more sensitive to the effects of amphetamine and cocaine than IC rats (Bowling and Bardo 1994, Green, Alibhai et al. 2010). In the study by Bowling and Bardo (1994), EC rats, but not IC rats, showed CPP with a low dose of amphetamine (0.5 mg/kg). At a higher dose (2.0 mg/kg), both EC and IC rats showed CPP, but the effect in EC rats was greater across repeated tests. The cocaine study used only saline and 1 dose of cocaine (10 mg/kg) but the results were congruent with the amphetamine results. In parallel with these CPP results, EC rats also display greater sensitivity to stimulant-induced locomotor activity (Bowling, Rowlett et al. 1993, Bowling and Bardo 1994), suggesting at least initially that enrichment may enhance addiction vulnerability to stimulant drugs. Interestingly, mice may not show the same pattern, as EC mice are less sensitive in CPP procedures than group-housed mice (Solinas, Thiriet et al. 2009); however, this study did not include an IC group for comparison.

Based on results obtained with stimulant CPP in rats, it was hypothesized that EC rats would self-administer stimulants more readily than IC rats, and thus would be more vulnerable to addiction. Contrary to this hypothesis, however, the overwhelming evidence indicates that environmental enrichment protects against drug abuse vulnerability as measured by SA of stimulants such as amphetamine, cocaine, nicotine and methylphenidate. For example, EC rats self-administer less amphetamine and cocaine than IC rats on a fixed ratio (FR) schedule, particularly at low unit doses during acquisition, maintenance responding, extinction, and reinstatement procedures (Bardo, Klebaur et al. 2001, Green, Gehrke et al. 2002, Green, Alibhai et al. 2010). These studies employed full dose response ranges for amphetamine and cocaine. These studies were replicated several times and expanded to include EC protection in re-acquisition, escalation, cue responding, incubation of craving, and a conflict model of stimulant SA (Stairs, Klein et al. 2006, Chauvet, Lardeux et al. 2009, Gipson, Beckmann et al. 2011, Stairs, Prendergast et al. 2011, Alvers, Marusich et al. 2012, Chauvet, Goldberg et al. 2012, Hofford, Darna et al. 2014, Hofford, Beckmann et al. 2016, Gauthier, Lin et al. 2017, Stairs, Ewin et al. 2017, Ewing and Ranaldi 2018). Moreover, enrichment has been shown to reduce risk in various prevention, treatment, and behavioral economics models alike (Thiel, Pentkowski et al. 2010, Ranaldi, Kest et al. 2011, Thiel, Engelhardt et al. 2011, Thiel, Painter et al. 2012, Yates, Bardo et al. 2017). When assessed using a progressive ratio (PR) schedule, environmental enrichment sharpens the amphetamine dose response curve, with EC rats showing reduced responding at a low unit doses, but a more rapid rise (i.e. slopes show an Environment X Dose interaction) in responding with increasing unit doses compared to IC rats (Green, Gehrke et al. 2002). Importantly, however, the maximal breakpoint for responding on a PR schedule does not differ between EC and IC rats. Taken together, these results illustrate that, in contrast to CPP, environmental enrichment reduces stimulant addiction vulnerability, at least at low unit doses.

The apparent paradox indicating that enrichment increases CPP, while reducing SA, does not appear to be specific to stimulant drugs. With ethanol, for example, environment enrichment increases ethanol CPP in mice (Rae, Zanos et al. 2018), whereas voluntary oral ethanol SA is reduced (Holgate, Garcia et al. 2017, Rodriguez-Ortega, de la Fuente et al. 2018). Conversely, voluntary oral ethanol SA is reduced by enrichment in alcohol-preferring (P) rats (Deehan, Palmatier et al. 2011), although not in outbred rats (Rockman, Gibson et al. 1989, Berardo, Fabio et al. 2016). Nonetheless, when outbred rats are required to perform an operant response for ethanol, environmental enrichment reduces intake (Deehan, Cain et al. 2007). Thus, similar to stimulants, environmental enrichment increases ethanol CPP, while decreasing ethanol SA, especially when ethanol access is response-contingent.

Results obtained with opioids indicate a further dissociation between CPP and SA following environmental enrichment, although the results are more mixed compared to stimulants and ethanol. Similar to the effects obtained with stimulants and ethanol, enrichment increases CPP produced by low-efficacy μ opioid agonists such as buprenorphine, butorphanol and nalbuphine (Smith, Chisholm et al. 2005); however, negative or contrary results have been obtained with high-efficacy μ agonists such as morphine or heroin (Smith, Chisholm et al. 2005, El Rawas, Thiriet et al. 2009). With SA, however, limited results uniformly show a decrease in opioid SA following environmental enrichment. That is, compared to IC rats, EC rats showed less SA assessed by either voluntary oral morphine consumption (Alexander, Coambs et al. 1978) or response-contingent intake of remifentanil (Hofford, Chow et al. 2017);Thus, taken altogether, these results from studies examining the effects of environmental enrichment illustrate a clear discrepancy between the results obtained with CPP and SA.

Genetic/molecular manipulations having opposite effects on CPP and SA

CREB blockade in NAc shell increases cocaine CPP but decreases FR- and PR-SA

The basic-region leucine zipper (BZIP) transcription factor cAMP response element binding protein (CREB) is a critical master switch for addiction-related behavior (Carlezon, Duman et al. 2005). This transcription factor induces expression of many immediate early genes induced by drugs of abuse, including cFOS, FOSB, EGR1 (zif-268), ATF3, and others (Alibhai, Green et al. 2007, Green, Alibhai et al. 2008, Renthal, Carle et al. 2008, Duclot and Kabbaj 2017). Increased CREB activity in vivo can be achieved via viral vectors to overexpress wild-type CREB protein or a constitutively active CREB protein (Dong, Green et al. 2006, Larson, Graham et al. 2011). In contrast, CREB activity can be suppressed by knocking down CREB expression with an shRNA-expressing vector, a vector overexpressing a dominant-negative mutant CREB (mCREB), or by overexpressing the wild-type CREB inhibitor inducible cAMP early repressor named ICER (Carlezon, Thome et al. 1998, Dong, Green et al. 2006, Green, Alibhai et al. 2006, Green, Alibhai et al. 2010, Larson, Graham et al. 2011). CREB exerts disparate effects on behavior depending on which brain region and cell types are activated (Carlezon, Duman et al. 2005), but in the nucleus accumbens (NAc) shell, CREB activation decreases cocaine CPP and increases cocaine PR-SA and low-dose FR-SA (Carlezon, Thome et al. 1998, Larson, Graham et al. 2011). Conversely, inhibition of CREB in NAc shell produces an increase in cocaine CPP with a decrease in cocaine FR-SA, particularly at low unit doses (Carlezon, Thome et al. 1998, Larson, Graham et al. 2011). These were tested across 3 doses of cocaine for CPP, 4 doses for FR-SA and 2 doses for PR-SA responding. Given the strength of the evidence, particularly in bidirectional modulation of CREB activity, the most reasonable conclusion to draw is that CREB activation in NAc shifts the dose-response for CPP and FR-SA to the right, but at the same time increases motivation as measured by PR-SA.

Interestingly, the low CREB phenotype is similar to the phenotype produced by environmental enrichment (Green, Alibhai et al. 2010). Also similar to environmental enrichment, decreased CREB function in medium spiny neurons of NAc shell decreases neuronal excitability (Dong, Green et al. 2006, Scala, Nenov et al. 2018). Not surprisingly, environmentally enriched rats exhibit decreased active (i.e. phosphorylated) CREB in NAc, thus leading to the hypothesis that CREB inhibition in accumbal medium spiny neurons is a molecular mechanism mediating the environmental enrichment phenotype characterized by a paradoxical increase in drug CPP and decrease in SA (Green, Alibhai et al. 2010). Again, mice seem to be different from rats, as environmental enrichment does not affect pCREB in EC vs. social-housed mice (Nader, Chauvet et al. 2012); however, this study did not include an IC group for comparison.

G9a overexpression decreases CPP but increases FR- and PR-SA

Substance use disorders are no doubt a function of neuroplasticity, and histone acetylation has been implicated in this plasticity (Kalda, Heidmets et al. 2007, Renthal, Maze et al. 2007). More recently, histone methylation has also received attention. The histone lysine methyltransferase G9a (gene name ehmt2) is among the top 3% of highest expressing transcripts in NAc in rats (Zhang, Kong et al. 2016) and its transcript is decreased by repeated experimenter-delivered cocaine exposure (Maze, Covington et al. 2010). Overexpression of G9a in NAc decreases cocaine CPP in mice, an effect that seems to be associated with dopaminergic D2-expressing medium spiny neurons (Maze, Covington et al. 2010, Maze, Chaudhury et al. 2014); CPP was tested at a single cocaine dose. In contrast, using 4 doses for FR-SA and 2 doses for PR-SA, overexpression of G9a in NAc in rats increased cocaine FR-SA, particularly at low unit doses, while also increasing PR-SA at high unit doses (Anderson, Larson et al. 2018), again corroborating a genetically-based divergence between CPP and SA outcomes. Despite the limited dose range for the CPP experiment, the most reasonable conclusion is that G9a overexpression decreases cocaine sensitivity (shifting the dose-response functions to the right for CPP and FR-SA), yet also increases motivation for the drug as measured by PR-SA.

Pharmacological manipulation having opposite effects on CPP and SA

Clozapine administration decreases cocaine CPP but increases PR-SA

In addition to environmental and genetic manipulations, there is also evidence of a dissociation between the effects of CPP and SA following acute treatment with clozapine, particularly when examining cocaine reward. In contrast to typical neuroleptics like haloperidol that act through dopaminergic D₂ receptors, the atypical neuroleptic clozapine has weak affinity for the D₂ receptor in favor of serotonergic receptor antagonism, primarily the 5-HT_2A subtype (Meltzer, Massey et al. 2012), as well as interaction with the GABA_B receptor (Wu, Blichowski et al. 2011), thus producing a behavioral profile different from haloperidol. In rats, while clozapine pretreatment blocks the development of cocaine CPP (Kosten and Nestler 1994), it increases PR-SA for cocaine at moderate clozapine doses (Loh, Fitch et al. 1992). In contrast to PR-SA, however, clozapine decreases cocaine FR-SA in rats, with no compensatory increase in intake that is typically observed with D₂ antagonists such as haloperidol (Roberts and Vickers 1984). The CPP study tested 1 dose of clozapine against 1 dose of cocaine and the SA studies each tested 3 doses of clozapine against a single cocaine dose. Monkeys, however, do show an increase in FR responding with clozapine, suggesting that it is possible that the rat study did not employ clozapine doses low enough to increase responding (Vanover, Piercey et al. 1993). Regardless of FR results, the fact that clozapine decreases CPP yet increases PR-SA alone meets the definition of incongruous CPP vs. SA.

The paradoxical results obtained between cocaine CPP and SA following clozapine treatment are not replicated entirely with other drugs of abuse, although some results corroborate the paradox. For example, similar to cocaine CPP, clozapine blocks morphine CPP in male mice (Manzanedo, Aguilar et al. 2001). Unlike haloperidol, however, clozapine does not block nicotine CPP in rats (Brown, Kirby et al. 2018), although it does attenuate nicotine SA measured with either voluntary oral intake using a 2-bottle choice (Kameda, Dadmarz et al. 2000) or operant-based FR or PR schedules, as well as cue- and drug-induced reinstatement of nicotine seeking (Abela, Li et al. 2018). Interestingly, this latter study also showed that clozapine enhances PR responding for food at an intermediate dose (2.5 mg/kg) (Abela, Li et al. 2018) and another study showed that clozapine increases shock-suppressed responding for food in squirrel monkeys (Bergman and Spealman 1986), results that are consistent with the body weight gain that typifies use of this atypical antipsychotic (De Berardis, Rapini et al. 2018). Thus, while the paradoxical effects on CPP and SA are not as consistent as described earlier for environmental and genetic manipulations, some evidence indicates that clozapine blocks drug reward measured by CPP, while enhancing drug and non-drug SA.

Insights into possible factors contributing to differential results

A common antecedent

Despite fundamental differences between CPP and SA procedures discussed below, both behavioral paradigms can be traced back to a common antecedent published by Shirley Spragg in 1940 (Spragg 1940). Spragg was interested in studying morphine addiction using non-human (and thus non-verbal) primates. In these studies, Spragg utilized intramuscular morphine injections to induce physical dependence in chimpanzees. Thus, these experiments were testing negative reinforcement--what the chimpanzees were willing to do to avoid the withdrawal symptoms of morphine abstinence. Spragg first noticed that subjects would congregate in the room where the morphine injections were administered, specifically when they were in a state of withdrawal, a clear preference for a place associated with (i.e. conditioned to) morphine. Further, the chimpanzees would coax, cajole, and pull Spragg into the injection room, a type of operant response made by the subject in an attempt to receive an experimenter-delivered morphine injection.

To increase the rigor of his observations, Spragg then devised a choice procedure for the chimpanzees. He built a locked black box that contained a food reward (usually a banana) and a white box that contained a syringe of morphine. The subject was allowed to choose either a triangular-shaped black key that would fit the triangular keyhole of the black box or a round-shaped white key to open the white box. When the subject chose the white box, Spragg would administer the morphine intramuscularly. Not surprisingly, the subject would choose the white key and open the syringe-containing box when in morphine withdrawal, but would choose the black key for a banana when motivated by hunger rather than drug withdrawal. While this choice procedure does not conform exactly to either current CPP or SA procedures, it is an interesting historical illustration that integrates both of these models. Specifically, like CPP, this procedure produces a preference for an environment previously associated with drug and, like SA, it includes a manipulandum for making an operant response to receive a drug injection. Since the time of Spragg’s work in chimpanzees, most preclinical studies have tended to rely on either CPP or SA exclusively, although at least one other model has been developed to integrate CPP and SA in rodents (Ettenberg 2009).

While CPP is typically interpreted to represent a Pavlovian conditioning paradigm, whereas SA is an operant conditioning paradigm, it is important to remember that this distinction is not absolute. In particular, CPP has an operant-like component (i.e. moving from one compartment to another) that can be interpreted to reflect behavior controlled by the secondary reinforcing effect of the drug-paired context. Conversely, SA also has Pavlovian components that control behavior, including components associated with both discrete stimuli (cue illumination, lever press sound, etc.) and contextual stimuli (olfactory, visual and tactile cues associated with the apparatus).

Differences between CPP and SA: parallels with human psychopharmacology

In humans, drugs of abuse usually produce subjective feelings (i.e. interoceptive cues) of “well-being”, “euphoria”, or “high” (i.e. reward). Additionally, drugs of abuse can increase observable behavior such as drug taking (i.e. reinforcement) that can become compulsive in some individuals (i.e. substance use disorders). Early human psychopharmacology research found that the subjective effects of drugs do not always correlate with human SA in laboratory settings. For example, one study found that low cocaine doses (4 and 8 mg i.v.) and procaine (48 mg i.v.) do not produce any subjective reports of “high”, yet they are reliably self-administered by those same subjects in a choice procedure (Fischman 1989). Additionally, desipramine therapy is able to decrease subjective self-reports from cocaine without changing cocaine-taking behavior (Fischman 1989). Similar effects are seen with morphine administration (Lamb, Preston et al. 1991). Conversely, naltrexone decreases methamphetamine drug taking without affecting subjective reports (Marks, Lile et al. 2016). Indeed, one study found only modest correlations in amphetamine taking and subjective reports among only 6 of 17 subjective report items (Bolin, Reynolds et al. 2013). These studies demonstrate that self-reported subjective interoceptive cues can be distinguished from drug-taking behavior. Indeed, individuals experimenting with drugs of abuse can achieve robust interoceptive reward cues without developing the craving and compulsive drug taking associated with substance use disorders.

What do animal models of CPP and SA really measure?

Our hypothesis is that discordance in CPP vs. SA in the cases described above is a function of dissociating interoceptive reward cues of a drug measured by CPP from motivation for the drug measured by SA. A tempting response to the question “What do CPP and SA really measure?” would be to say that CPP procedures measure Pavlovian classical conditioning aspects of substance use disorders, whereas SA measures operant conditioning aspects. However, this statement is an oversimplification based on merely procedural differences in the arrangement of the drug cue in relation to a either a contextual stimulus or a response (i.e., S-S vs S-R learning) (Mackintosh 1974). These procedural differences become blurred when one considers that CPP involves an experimental animal moving from a neutral compartment to one previously paired with a drug, thus reflecting a type of operant response (i.e. the presence of a discrete manipulandum is not necessary to demonstrate that an approach behavior can be reinforced). Further, with SA, animals do not self-administer drugs at mid to low unit doses without Pavlovian cues such as lights and/or tones that are paired with each drug infusion (Schenk and Partridge 2001). With nicotine SA, for example, the absence of a discrete cue signaling each drug infusion leads to negligible rates of intake even at high doses (Caggiula, Donny et al. 2001, Donny, Chaudhri et al. 2003).

Strict behaviorist interpretations of CPP and SA procedures avoid using terms like “reward”, “craving”, and “euphoria”, instead restricting discussion to the measurable term response-contingent “reinforcement” for SA exclusively. In contrast, the term “reward” is more often applied to CPP exclusively, which presumably denotes approach to one context due its Pavlovian association with the positive interoceptive state induced by the drug. The simple truth of the matter is that the current scientific lexicon that includes words like “motivation”, “reward”, “reinforcement”, “craving”, etc. was developed long ago from lay language and does not superimpose well on the neural systems that underlie either CPP or SA. Thus, researchers in the field are constrained to a few limited options: 1) adhere to a strict behaviorist approach describing only quantifiable operant responses or time spent in a chamber without inferring more complicated constructs; (2) redefine current lay terms to match the neural systems; or (3) develop new terms that better fit the neural systems.

The role of motivation

The term “motivation” has a variety of definitions and even more connotations. In an effort to avoid splitting hairs, we refer to motivation simply by the Oxford Dictionary’s “willingness to do something”. This definition implies that the “something” is to emit behavior in order to achieve an outcome. In animal models, motivation is measured most easily by operant responding. The number of operant responses a subject is willing to exert is a measure of the motivation for that stimulus, the basis of the PR, extinction, and reinstatement procedures. A behavioral economic analysis also can be used to graphically formalize the relationship between the price of the commodity (drug) and consumption of the commodity at a given price. Procedures where responses are either reinforced or not under progressively leaner schedules or higher prices are direct measures of how much effort the subject is willing to “pay” for the drug, which is analogous to questionnaires in human studies asking how much money the subject would be willing to pay for the drug.

In contrast to the non- or partially-reinforced procedures above, the linear descending limb of the FR operant schedule for SA demonstrates how the motivation for the drug infusion decreases as the subject attains optimal/maximal dosing, a form of satiety (Tsibulsky and Norman 1999). As the drug is cleared from the body, drug levels fall below the optimal/maximal level and the subject regains motivation and responds to regain optimal/maximal drug levels. This effect is analogous to food satiety, where sated subjects lose motivation to respond for food until hunger motivation returns. Thus, maintenance responding at supra-threshold doses, including escalation procedures, is a measure of the optimal/maximal dose level for satiety. By this framework, the ascending limb of the maintenance responding FR dose-response function would be the threshold dose for motivated responding and a rightward shift in the ascending limb would represent an increase in motivational threshold.

As mentioned above, CPP testing has some form of an operant response (moving to the drug-paired side) that could be construed as motivation. However, in contrast to SA, entry into the drug-paired compartment is qualitatively different from a typical lever press or nose poke measured in SA. In contrast to SA, the dependent measure of CPP is nearly always duration in the paired compartment, which is a graded measure that is not a discrete operant behavior, and thus not a measure of motivation. In CPP, the subject explores all chambers in the apparatus repeatedly, and early work has shownthat CPP reflects an increase in the duration spent in the drug-paired compartment and not an increase in the number of entries into the drug-paired compartment (Bardo, Miller et al. 1984). As an extreme example of duration not being a measure of motivation, a subject falling asleep in the drug-paired chamber of a CPP test would score a robust place preference but an identical subject falling asleep in a SA procedure would score as having very little motivation for the drug. By this conceptualization, CPP is not a measure of motivation because duration on the drug-paired side has no cost (the subject is always somewhere). Thus, CPP is a more pure measure of reward rather than motivation.

There is significant evidence that motivation plays a role in differential results between CPP and SA based on the environmental, genetic and pharmacological studies cited earlier (see Table 1). Environmental enrichment, the most robust example of discordance between CPP and SA results, produces differential effects on sucrose pellet SA based on hunger motivation. When EC and IC rats have ad libitum free access to sucrose pellets (in the home cage) there are no differences in intake; however, differences between EC and IC rats emerge when an operant response contingency is introduced. That is, at 100% free-feed body weight (i.e. no hunger motivation) EC rats respond for fewer sucrose pellets than IC rats, but at 85% free-feed body weight EC rats consistently outperform IC rats (Green, Alibhai et al. 2010). Similarly, EC rats respond less for low doses of amphetamine under a PR schedule than IC rats, yet they respond more than IC rats at a higher more motivating dose (Green, Gehrke et al. 2002). Hunger motivation also increases cocaine SA (Carroll, Lac et al. 1986). Thus, motivational factors play a more important role in SA than CPP.

Table 1:

Studies producing discordant effects on SA vs. CPP due to alterations in motivation (exceptions discussed in accompanying text).

Manipulation	Drug	Species	CPP	Self-admin	Refs
Environmental Manipulation
Environmental enrichment	Cocaine or amphetamine	Rats	Increased induction	Decreased acquisition, maintenance (low dose FR and PR), extinction, reinstatement, re-acquisition, escalation, conflict, cue responding, incubation of craving	(Bowling, Rowlett et al. 1993, Bowling and Bardo 1994, Bardo, Klebaur et al. 2001, Green, Gehrke et al. 2002, Stairs, Klein et al. 2006, Chauvet, Lardeux et al. 2009, Green, Alibhai et al. 2010, Gipson, Beckmann et al. 2011, Stairs, Prendergast et al. 2011, Alvers, Marusich et al. 2012, Chauvet, Goldberg et al. 2012, Hofford, Darna et al. 2014, Hofford, Beckmann et al. 2016, Gauthier, Lin et al. 2017, Stairs, Ewin et al. 2017, Ewing and Ranaldi 2018)
	Ethanol	Rats Mice	Increased CPP	Decreased free access oral intake and operant-based oral intake	(Deehan, Cain et al. 2007, Holgate, Garcia et al. 2017, Rae, Zanos et al. 2018, Rodriguez-Ortega, de la Fuente et al. 2018)
Genetic Manipulations
Decreased CREB in NAc shell	Cocaine	Rats	Increased induction	Decreased maintenance responding (PR and low dose FR)	(Carlezon, Thome et al. 1998, Larson, Graham et al. 2011)
G9a overexpression in NAc shell	Cocaine	Rats	Decreased induction	Increased maintenance responding (PR and low dose FR)	(Maze, Covington et al. 2010, Maze, Chaudhury et al. 2014, Anderson, Larson et al. 2018)
Pharmacological Manipulation
Clozapine administration (10 mg/kg)	Cocaine Morphine Amphetamine	Rats	Decreased induction	Increased maintenance (PR) Increased oral consumption	(Loh, Fitch et al. 1992, Kosten and Nestler 1994, Ufer, Dadmarz et al. 1999)

Non-motivational discrepancies between CPP and SA: parallel shifts in monophasic vs. biphasic dose response relationships

Volitional control and the operant response

The second classification of apparent discrepancies between CPP and SA can be accounted for by simple parallel shifts in dose response functions rather than motivational differences. While dose response curves derived with CPP typically do not include an extensive number of doses because each dose requires a separate group of mice or rats (i.e., between-subject dose effect curves), a meta-analysis suggests a monotonic function (Bardo, Rowlett et al. 1995); see Figure 1 left. In contrast, FR dose-response functions become biphasic inverted U-shaped curves, with a descending limb where subjects decrease responding for higher doses due to satiety (Figure 1 middle). That is, higher doses reach or maintain the satiety level with fewer self-infusions than low doses, producing fewer responses/infusions at higher unit doses to maintain the same satiety level. In contrast to FR-SA, the dose response curve for PR-SA is monotonic, similar to CPP. Partial pharmacological blockade (often competitive antagonism such as the case of haloperidol) of the drug of abuse will decrease the “apparent” dose of the drug of abuse and shift dose-response functions to the right in all three models (CPP, FR- and PR--SA) in a similar parallel fashion (Figure 1). This will decrease CPP and PR-SA, as the lower apparent dose will produce less reward and motivation. This will also lead to extinction for low unit doses of the drug in FR-SA. However, at higher unit doses, the drug of abuse is still able to reach the threshold for reinforcement and will be self-administered at an even greater rate to compensate for the partial blockade (full blockade will flat-line the dose-response function of all three paradigms).

Figure 1:

Parallel rightward shifts in dose-response function

Fundamentals of psychopharmacology emphasize the importance of testing manipulations across the full range of doses from control (usually saline) to subthreshold doses, to moderate doses, and finally to high doses bordering on toxic levels. Addiction models represent a useful demonstration of the importance in psychopharmacology of determining full dose response functions, as a decrease in FR-SA of a single dose could represent a supra-threshold leftward shift (agonism), a rightward shift rendering the dose subthreshold (antagonism), or a flattening of the function (non-specific rate suppressant effects). Although most published literature does not test the pretreatment of compounds across the full dose range of SA, we can draw inferences based on the available evidence.

Additionally, properly-designed psychopharmacology experiments should consider potency (left-right shifts in dose response) and effectiveness (up/down shifts), again highlighting the need for full dose-response functions. Unfortunately, CPP, FR and PR procedures are not ideal preparations for assessing effectiveness, since high drug doses can induce secondary effects that limit the detection of reward or reinforcer effectiveness. If CPP studies produced a plateau at high doses, effectiveness would be easily assessed. However, CPP experiments typically produce a monotonic rising function that is interrupted only by side effects such as stereotypy, lethargy, or toxicity during conditioning at very high doses that limit reward and/or learning (Brown, Kirby et al. 2018). Additionally, by nature, rats will repeatedly investigate the whole of the apparatus on a test trial, meaning that there is an upper limit (i.e. ceiling) on preference not determined by interoceptive reward. Similarly, for FR schedules, intake is limited by satiety brain levels (Tsibulsky and Norman 1999), although one could argue that this in itself is a form of effectiveness. While PR schedules can produce a clear plateau at high doses (Green, Gehrke et al. 2002), that plateau is likely determined to a large degree by fatigue. For example, in one widely used PR methodology for rats (Richardson and Roberts 1996), the 26^th infusion requires 1102 responses, which cumulates with 4,835 previous responses for a total of 5937 responses. The 29^th infusion requires 2012 responses for a total of 10,943 responses. This strenuous schedule forces a plateau that cannot be easily overcome due to fatigue, thus limiting sensitivity to observe potential treatment differences at high breakpoints. Thus, although upward/downward shifts are important, these particular procedures have substantial caveats with respect to measuring effectiveness. This does muddy interpretations, as a left/right shift in the dose-response function for CPP often looks identical to an up/down shift.

Haloperidol produces a rightward shift in dose-response functions

It has been known for a long time that cocaine’s effects are due in large part to indirect dopamine D2 receptor activation via inhibition of the plasmalemma dopamine transporter (DAT). Similar to the previous examples of discrepant CPP and SA results (see Table 1), haloperidol decreases cocaine CPP (Spyraki, Nomikos et al. 1987), but increases FR cocaine SA (Roberts, Dalton et al. 1987). Importantly, however, haloperidol decreases PR cocaine SA (Roberts, Loh et al. 1989), thus indicating that regardless of the procedure (CPP, FR-SA or PR-SA), all three results are indicative of a rightward shift in the dose response function. The CPP study tested 2 doses of cocaine while the FR-SA and PR-SA tested 5 doses and 2 doses of haloperidol, respectively, against a single dose of cocaine. Rightward shifts in the cocaine dose response function typically only shift the peak FR responding by one dose before the dopaminergic antagonism is unable to be surmounted and non-specific rate suppressant effects flatten the dose response curve. Inversely to D2 antagonism, indirect dopaminergic agonists GBR12909, WIN35428, and indatraline cause a parallel leftward shift in the dose response function (Schenk 2002).Taken as a whole, the seeming incongruity of decreases in CPP with increases in FR-SA can be accounted for in this case by volitional administration of cocaine where animals alter response rates to compensate for the haloperidol pretreatment and maintain satiety-level dosing, shifting the dose-response function to the right.

Yet to be determined discrepancies between CPP and SA

The examples above are classified as either motivational discrepancies in CPP and SA or as parallel shifts in the dose response functions. The key factor in determining the classification is PR responding, as this procedure directly measures motivation. If PR is concordant with FR responding, the manipulation would be classified as motivation-based, but if the PR is concordant with CPP (and opposite to FR), the manipulation would be a simple parallel shift in dose-response functions. Based on this analysis, we would conclude that motivational discrepancies may account for the CPP vs SA differences noted in Table 1, whereas a shift in the dose-response curve may account for the CPP vs SA differences noted with haloperidol.

Table 2:

Manipulation producing simple shifts in dose-response functions and manipulations yet to be determined.

Manipulation	Drug	Species	CPP	Self-admin	Refs
Manipulation producing shifts in dose-response functions
Haloperidol administration	Cocaine	Rats	Right-shifted	PR and FR right-shifted	(Roberts, Dalton et al. 1987, Spyraki, Nomikos et al. 1987, Roberts, Loh et al. 1989)
Yet to be determined manipulations
Social defeat	Cocaine	Mice	Increased, slower extinction, enhanced reinstatement	Delays acquisition of FR-SA	(Reguilon, Montagud-Romero et al. 2017, Rodriguez-Arias, Montagud-Romero et al. 2017, Ferrer-Perez, Reguilon et al. 2018)
Naltrexone administration	Cocaine	Rats	Decreased	Increases FR-SA	(Carroll, Lac et al. 1986, Suzuki, Shiozaki et al. 1992, Sala, Braida et al. 1995)

Beyond these classifications, the following examples lack PR data, and thus they cannot yet be classified.

Social Defeat increases cocaine CPP yet delays acquisition of cocaine FR-SA

Similar to the discordant findings observed between CPP and SA following environmental enrichment, a few studies have also observed that exposure to adolescent social defeat in mice increases CPP, while decreasing SA when tested in adulthood. In this model, group-housed male mice are individually exposed to an aggressive resident mouse that is older and has been individually caged in order to enhance aggression. Additional exposure to threatening behavior from the aggressive resident occurs with the intruder being protected from injury by a partition within the resident cage. In one study (Rodriguez-Arias, Montagud-Romero et al. 2017), male OF1 mice were exposed to repeated social defeat from 27–36 days of age and then were tested for cocaine CPP or SA beginning at 58 days of age. Defeated mice displayed CPP at a dose that is subthreshold for control mice (1 mg/kg), suggesting enhanced reward measured by CPP. In contrast, defeated mice showed delayed acquisition of cocaine FR-SA, suggesting attenuated reinforcement measured by SA. The CPP was tested at 2 different cocaine doses and the FR-SA at a single dose. The CPP results have been bolstered by additional studies showing that social defeat increases the number of trials required to extinguish CPP and enhances reinstatement of cocaine CPP (Reguilon, Montagud-Romero et al. 2017, Ferrer-Perez, Reguilon et al. 2018). While social defeat impairs the blood-brain barrier (Rodriguez-Arias, Montagud-Romero et al. 2017), changes in cocaine pharmacokinetics do not likely play a role in these behavioral changes because cocaine has high brain penetrance. Instead, the differential effects of social defeat on CPP and SA more likely result from a fundamental difference in these outcome measures. Thus, the reasonable conclusion given the available data is that social defeat does indeed produce disparate effects on CPP vs. FR-SA, but no conclusions can be reached yet regarding motivation.

While results from mice reveal a differential effect of social defeat on CPP and SA, it is not clear if this difference generalizes to rats. Rats that are socially defeated in adolescence and tested in adulthood show enhanced amphetamine CPP (Burke, Watt et al. 2011), as well as enhanced acquisition of cocaine SA, PR responding, and binge responding (Burke and Miczek 2015), suggesting no discrepancy in CPP and SA. Interestingly, however, the defeat-induced increase in cocaine SA is moderated by housing condition. That is, this effect is obtained in paired-housed rats, but not in single-housed rats (Burke and Miczek 2015). It should be noted that the CPP was performed with amphetamine and the SA with cocaine, but generally these two drugs act in a similar (yet not identical) manner. Thus, it is unclear if the differences are due to species, housing, experimental details (e.g. amphetamine vs. cocaine), or some as yet unidentified factor. Regardless, the apples-to-apples comparison in mice above shows opposite modulation of drug-induced CPP and SA by adolescent social defeat.

Naltrexone decreases cocaine CPP and increases cocaine FR-SA

The second example of indeterminate discrepancy comes from administration of the opioid receptor antagonist naltrexone, which shifts the dose-response functions to the right for both CPP and FR-SA (Carroll, Lac et al. 1986, Suzuki, Shiozaki et al. 1992, Sala, Braida et al. 1995). The two CPP studies tested 1 dose of naltrexone against 1 dose of cocaine, while the FR-SA study tested 3 doses of naloxone against 1 dose of cocaine. The two CPP studies each tested a different cocaine dose (4 mg/kg and 10 mg/kg), providing some dose-response information. The fact that cocaine FR-SA increased responding rules out any non-specific rate-suppressant effects and thus the reasonable conclusion is that naltrexone shifted both curves to the right.

Interestingly, naltrexone is a mu opioid antagonist and thus does not directly block monoamine neurotransmission. These results are possibly a function of the antagonism of mu opioid signaling from cocaine-induced release of endorphins in the nucleus accumbens (Olive, Koenig et al. 2001). Regardless, the satiation of FR cocaine SA does not provide a strong assessment of motivation, so at this point it is impossible to classify this example as a function of motivation (as with the first set of examples) or merely a parallel shift in dose response functions (as with the second classification).

Concluding Remarks

To first summarize the information above, we presented 7 examples of manipulations that show opposite responses in CPP and FR-SA procedures. In four of these examples (enrichment, CREB, G9a and clozapine), PR-SA responding was concordant with FR-SA suggesting a motivational effect. Another example (haloperidol) simply shifted dose-response functions to the right in all cases, interpreted as a non-motivational effect. Two final examples (social defeat, naltrexone) lack PR-SA data, and as such, cannot be classified as motivational or parallel shifts.

Our conceptualization of substance use disorders is that reinforcement as measured by SA is a reflection of interoceptive reward modified by motivation, whereas CPP is a purer measure of interoceptive reward without motivational aspects. Substance use disorders are a function of hyper-motivation regardless of subjective reward. This may sound reminiscent of Robinson and Berridge’s Incentive Sensitization theory where “wanting” for drugs becomes sensitized while “liking” either has no change or even decreases with continued use (Robinson and Berridge 1993, Berridge and Robinson 2016). It would be tempting to say that reward as measured by CPP represents “liking” whereas the motivation aspect of SA represents “wanting”; however, the Incentive Sensitization theory is clear that “liking” is not dopamine-dependent and does not show sensitization with repeated drug use. CPP, however, is clearly dopamine-dependent and does sensitize (Spyraki, Nomikos et al. 1987, Shippenberg and Heidbreder 1995). Thus, by our conceptualization, both CPP and SA could fit under the umbrella of “wanting”. However, if one insists that “wanting” is synonymous with motivation, then CPP would represent neither “wanting” nor “liking”.

This review provides multiple conclusions regarding CPP and SA procedures. First, despite a common antecedent, CPP and SA are not isomorphic measures of addiction-related behavior. Second, dissociations of CPP vs. PR-SA vs. FR-SA may be analogous to dissociations in human studies of subjective self-reports of reward vs. subjective indices of value/craving vs. drug taking behavior, respectively. Third, incongruous results between CPP and SA in at least four cases are most likely due to motivation; in contrast, some apparent CPP vs. SA dissociations are merely a function of parallel shifts in dose response functions. Lastly, it is concluded that full dose response functions across multiple procedures (CPP, PR-SA, and FR-SA) are necessary to determine whether discordant results between CPP and SA are due to motivational effects or merely parallel shifts in dose response functions.

There are many problematic issues in interpreting data from these procedures, including lack of full dose-response data (particularly for CPP studies), nonspecific drug effects at high doses (lethargy, amnesia, nausea, stereotypy, etc.), and inconsistencies in procedures. At the very least, we feel that there are enough instances of incongruity between CPP and SA to confidently state that these are not isomorphic models of addiction, despite having a common antecedent. We first presented four incongruities that cannot be explained by consistent parallel shifts in dose response relationships. Regarding the underlying factors for these incongruities, we focused mainly on motivation, as SA requires a voluntary active response that has a small but significant associated cost that can be subject to motivation (e.g. PR-SA). Motivation-based procedures of SA such as extinction, reinstatement, and PR procedures are key for determining motivational aspects of the incongruities vs. mere parallel shifts in dose-response functions.

In addition to the four motivation-based examples of incongruity between CPP and SA, we presented one representative example (haloperidol) of what initially seemed like incongruous results between CPP and SA, but that is easily explained by simple antagonist-driven parallel rightward shifts in dose response functions rather than differences in motivation. This example underscores the importance of testing full dose response functions rather than a single dose. Lastly, we presented two examples of discordant CPP and FR-SA results where PR (or SA extinction/reinstatement) data are not available, illustrating that the importance of PR-SA, SA extinction, and/or reinstatement procedures for interpreting the discordance. If the PR results were to mirror FR-SA results, then the discrepant findings could be attributed to motivational differences. In contrast, if the results were to mirror CPP, then the results are most likely due to parallel dose response shifts. The question then arises: are FR procedures of limited value? Quite the contrary. FR-SA procedures measure reinforcement threshold (i.e. ascending limb) and drug satiety level (i.e. the descending limb), both important measures of addiction-related behavior. In addition, FR schedules, whether they are simple single schedules or FR choice schedules (e.g., food vs. drug), also model real-world drug-taking better than PR schedules.

Finally, no single animal paradigm can begin to model all the complexities of human substance use disorders as a whole, but rather can only model circumscribed aspects (endophenotypes) of substance use disorders. Thus, multiple behavioral models are necessary to understand the various important aspects of substance use disorders (craving, impulsivity, stress, habit, frustration, etc.). It is incumbent upon us to better understand what information each model can (or cannot) provide. Although not all is known of the differences between CPP and SA models, the evidence is clear that these are not isomorphic measures of addiction-related behavior, as a number of manipulations discussed above show opposite modulation.

Highlights.

• Conditioned place preference (CPP) and intravenous drug self-administration (SA) derive from a common antecedent

• Although most genetic and pharmacological manipulations produce concordant effects on CPP and SA, several manipulations produce discordant effects in these paradigms

• Examples of discordant effects include environmental enrichment, CREB modulation in the nucleus accumbens, G9a modulation in the accumbens, clozapine administration, haloperidol administration, social defeat stress, and naltrexone administration

• Discordant examples can be classified as driven by motivation or as simple parallel shifts in dose-response functions