1. Novelty Seeking

One of the most critical individual difference factors predicting drug use among humans is novelty seeking or sensation seeking (Kosten et al., 1994; Zuckerman, 1994; Wills et al., 1998; Ball, 2004). Zuckerman (1994) defines sensation seeking as a trait defined by the seeking of varied, novel, complex, and intense sensations and experiences and the willingness to take physical, social, legal, and financial risks for the sake of such experiences. Although drug use can increase sensation seeking (Ersche et al., 2010), longitudinal results also indicate a direct path leading from sensation seeking to initiation (Horvath et al., 2004). Adolescent sensation seekers are at increased risk for use of various drugs, including alcohol, tobacco, and marijuana (Martin et al., 2002, 2004; Sargent et al., 2010). Young adult high sensation seekers also are more sensitive to the effects of drugs (self-administration, positive subjective effects) than are low sensation seekers under controlled laboratory conditions (Stoops et al., 2007; Perkins et al., 2008; Fillmore et al., 2009; Kelly et al., 2009), suggesting that this facet is present before problem use. Neurobiological studies indicate that the underlying neurocircuitry associated with sensation seeking involves, at least in part, the nucleus accumbens. Human neuroimaging studies indicate that amphetamine produces the greatest dopamine release in the nucleus accumbens among high sensation seekers (Leyton et al., 2002), although this effect may be gender specific (Riccardi et al., 2006). When viewing highly arousing images, high sensation seekers show enhanced activation in regions involved in emotional induction and reward, as well as reduced activation of regions involved in emotional regulation (Joseph et al., 2009). Thus, brain systems altered by exposure to salient stimuli share a common link with drug reward, which may mediate the association between sensation seeking and drug use.

2. Impulsivity

Another individual difference factor predicting drug use among humans is impulsivity, defined broadly as the tendency to engage in premature, inappropriate, or maladaptive behavior without foresight (e.g., Dalley et al., 2011). Although impulsivity is a broadly defined biologically based trait that appears in most major theories of personality, it can be parsed into different facets (Whiteside and Lynam, 2001). Impulsivity reflects the balance of two independent behavioral processes: 1) approach or activation associated with reinforcement and 2) inhibition associated with punishment, with impulsive individuals exhibiting hyperactivation or hypoinhibition (e.g., Bechara, 2005). Neuroimaging studies have confirmed the interplay between reinforcement pathways (ventral striatum/nucleus accumbens and orbitofrontal, dorsal, and lateral regions of the prefrontal cortex) and inhibitory pathways (amygdala, insula, anterior cingulate, right inferior frontal gyrus, subthalamic nucleus, and supplementary motor areas of the prefrontal cortex) associated with impulsivity and drug abuse vulnerability (Nigg et al., 2006; Aragues et al., 2011; Dalley et al., 2011; Hommer et al., 2011). Impulsive individuals, determined using either personality- or performance-based criteria, initiate drug use at earlier ages, escalate to heavy use, and transition to abuse and dependence more quickly and are less likely to remain abstinent after treatment compared with nonimpulsive individuals (de Wit, 2009; Dick et al., 2010; Dalley et al., 2011). Longitudinal studies of early childhood temperament indicate that the association between impulsivity and vulnerability to drug abuse is present at an early age (Zucker et al., 2008; Chartier et al., 2010). Thus, individuals who are high in impulsivity are likely to engage in a variety of risky behaviors, including drug use, and are more sensitive to the reinforcing and other pharmacodynamic effects of drugs after initial use, thereby making them more inclined to continue and escalate drug use.

1. Social Experiences during Development

Environmental experiences during early development impact the brain directly or modify the genetic developmental processes through gene-environment interactions (Zucker et al., 2008). Low prosocial family interaction, for example, as well as peer deviance and drug use, are risk factors for the initiation of alcohol use during childhood (Oxford et al., 2001; Rose et al., 2001). Other social context factors influencing alcohol use by children include alcohol use by parents and exposure to alcohol through the media. Peer influence on alcohol and drug use becomes more salient among older adolescents (ages 11–13). Peers influence use through three pathways: 1) direct modeling of use; 2) self-sustaining affiliation with like-minded peers; and 3) overestimation of peer drug use (Windle et al., 2008; Chartier et al., 2010). Adolescents reporting high rates of drug use also report lower levels of social support from family and peers, whereas adolescents with heavy alcohol use actually report higher social support from friends. Recent studies link social environment with drug abuse vulnerability through underlying neurobiological mechanisms in that dopamine D2/3 receptor binding efficacy, which is associated with the reinforcing effects of drugs, is correlated with both social status and perceived level of social support and inversely correlated with social detachment (e.g., Schneier et al., 2000; Martinez et al., 2010).

2. Social Context during Drug Use

Drug use is a learned behavior driven in large part by the functional effects of the drug. As mentioned earlier, the functional effect of a drug is not intrinsic to the pharmacology of the drug, but rather it reflects an interaction between pharmacology and the context in which drug is available (Hughes et al., 1988). The social environment can play a prominent role in modulating the functional effects of abused drugs. One conceptual model for social context as a determinant of the functional effects of drugs is based on the manner in which drug use sets the occasion for access to social opportunities (Falk, 1983). This model postulates that drug use becomes a requirement for affiliation with selective groups, and intoxication becomes a “time-out” from social rules, which permits individuals to engage in behaviors not otherwise acceptable. Subsequent studies provide support for this model. For example, clinical laboratory studies show that stimulant and sedative drugs increase verbal behavior and social interaction (Stitzer et al., 1981; Higgins and Stitzer, 1986; Ward et al., 1997), suggesting that the reinforcing effects of a drug are increased when there is a pharmacologically induced enhancement of social interaction (i.e., the cocktail-party phenomenon). This possibility has been examined by testing the effects of amphetamine, secobarbital, and marijuana on talking behavior and preference for social interaction among healthy volunteers (Heishman and Stitzer, 1989). Relative to placebo, both amphetamine and secobarbital increased rates of talking, as well as preference for social interaction (i.e., increased the reinforcing efficacy of social interaction). In comparison, marijuana had little effect on talking rate or preference for social interaction.

Drug-taking behavior can also be altered by the social consequences of drug use. In an early series of residential laboratory studies, Bigelow and colleagues demonstrated that time-out from social interaction resulted in decreased alcohol self-administration among heavy alcohol users (Griffiths et al., 1974, 1977). These early studies helped to establish the importance of social access as a primary motive for alcohol and drug use (Cooper, 1994). The possible role of negative reinforcement (e.g., drug modulation of negative socially induced mood states such as social anxiety) also is a likely determinant of drug abuse vulnerability (DeMartini and Carey, 2011). Despite these behavioral results, little is known currently about the neural factors that explain why social interaction alters the behavioral effects of abused drugs.

1. Response to Novelty

The most reliable individual difference predicting the psychostimulant and reinforcing effect of drugs is the novelty “responder” test characterized initially by Piazza et al. (1989) and subsequently examined across many laboratories (Kabbaj, 2006). Rats are classified as high (HRs) or low responders (LRs) based on the amount of ambulatory activity recorded in an inescapable novel environment. The test is relatively brief, lasting 60 min or less, and rats are categorized as HR or LR based on a median split analysis. Since the initial report by Piazza et al. (1989), several studies have confirmed that HR rats show increased amphetamine-induced activity (Bevins et al., 1997) and amphetamine self-administration (Piazza et al., 1990, 2000; Pierre and Vezina, 1997; Klebaur et al., 2001a; Cain et al., 2005, 2006, 2008) compared with LR rats. The differences between HR and LR rats in amphetamine self-administration are similar for both males and females (Klebaur et al., 2001a), and these individual differences generalize to other stimulant drugs, including cocaine (Mantsch et al., 2001; Sell et al., 2005; Kabbaj, 2006; Belin et al., 2008, 2011; Walker et al., 2009) and methamphetamine (Bevins and Peterson, 2004; Gancarz et al., 2011). Nicotine self-administration also is greater in HR rats than in LR rats (Suto et al., 2001), although these individual differences may not occur in nicotine-induced hyperactivity (Coolon and Cain, 2009). Although individual differences have been reported for maintenance across various unit doses in cocaine self-administration (Piazza et al., 2000), they are most influential for acquisition at low unit doses (0.25 mg/kg per infusion; Mantsch et al., 2001), suggesting that HR/LR differences may be most closely associated with sensitivity to drug reinforcement. The HR/LR difference probably is due, at least in part, to genetic factors because inbred lines possessing the HR and LR phenotypes also display differences in the psychostimulant and reinforcing effects of stimulants (Gingras and Cools, 1997; Davis et al., 2008; Turner et al., 2008; Cummings et al., 2011).

A recently developed variation of the Piazza responder test was developed by Zahniser and colleagues (Gulley et al., 2003). In this test, rats are categorized as HR or LR based on their locomotor response to acute cocaine. This individual difference is linked to dopamine transporter (DAT) function (Sabeti et al., 2003; Briegleb et al., 2004) and cocaine CPP (Allen et al., 2007). However, in contrast to the HR/LR test, high cocaine responders do not differ from low cocaine responders in acquisition of cocaine self-administration (Mandt et al., 2008, 2012).

One apparent exception to the general finding that HR rats are more sensitive than LR rats to the locomotor and reinforcing effects of stimulants has been reported for methylphenidate. In the only study to date, Wooters et al. (2006) reported that male and female HR/LR rats do not differ in locomotor activity after acute methylphenidate during either adolescence or young adulthood. Although this conclusion from a single study requires confirmatory work, methylphenidate differs from cocaine and amphetamine by its greater potency at the dopamine and norepinephrine transporters relative to the 5-HT transporter (Han and Gu, 2006). This suggests the possibility that differences in 5-HT transporter function may play a role in the difference between HRs and LRs. Consistent with this possibility, tissue concentration of 5-HT is reduced in HR prefrontal cortex (Thiel et al., 1999). Further, using the conditioned place preference (CPP) paradigm in which rats are allowed to choose between two contexts paired previously with either drug or saline (Rossi and Reid, 1976), a 5-HT_2C antagonist differentially affects cocaine reward in HR and LR rats (Capriles et al., 2012) .

Differences between HR and LR rats also occur among drugs outside the stimulant class. With opiates, HR rats show enhanced morphine locomotion and self-administration (Deroche et al., 1993; Kalinichev et al., 2004). These effects are modulated by social context, as the influence of social crowding on morphine hyperactivity is evident in HR rats but not LR rats (Xigeng et al., 2004). The locomotor effects of cannabinoids and alcohol also are enhanced in HR rats compared with LR rats (Galanopoulos et al., 2011). With alcohol, oral intake is enhanced in HR rats using an FR3 schedule of reinforcement (Nadal et al., 2002) but not using either a free-access two-bottle test or an FR1 schedule of reinforcement (Bisaga and Kostowski, 1993; Gingras and Cools, 1995; Bienkowski et al., 2001; Hayton et al., 2012), suggesting that HR rats are more motivated than are LR rats to earn alcohol under high-effort schedules. Other than alcohol, however, little is known about HR and LR differences in response to anxiolytic drugs. This is a notable gap in information because individual differences in anxiety (or novelty seeking) based on the elevated plus maze predict cocaine self-administration, with low-anxiety rats showing enhanced responding for cocaine (Bush and Vaccarino, 2007). Nonetheless, because both low-anxious and HR rats show an increased propensity to self-administer cocaine, these findings appear to rule out the possibility that the elevated activity in HR rats simply reflects enhanced anxiety in the inescapable novel environment.

In addition to abused drugs, individual differences in the novelty responder test are observed with nondrug reinforcers such as palatable food. Operant responding using sucrose pellet reinforcers is greater in HR rats than in LR rats (Dellu et al., 1996; Klebaur et al., 2001a; Cain et al., 2006), although this effect may not generalize to less palatable reinforcers such as a standard food diet (Gulley, 2007). The ability of individual differences to predict responding for both drug and palatable nondrug reinforcers opens the possibility that individual differences exist in learning generally, rather than drug sensitivity specifically (Mitchell et al., 2005). Alternatively, because both drug and nondrug reinforcers involve overlapping neurocircuitry (Kelley and Berridge, 2002), these individual differences may be associated with common neural systems. Although there is little information to address this issue directly, Xu et al. (2001) found that individual differences in performance in either the Morris water or Y maze are not associated with the magnitude of morphine CPP. Thus, individual differences in general learning do not account readily for HR and LR differences in behaviors reinforced by drug or palatable food.

In addition to food reinforcement, HR rats show greater responding for novel visual stimulation. For example, maintenance of amphetamine self-administration is disrupted by novel stimuli, and this disruption is greater in HR rats than it is in LR rats (Cain et al., 2004). HR rats also show more responding than LR rats to earn novel visual stimuli (cue light illumination; Gancarz et al., 2011). These latter results are important because drug self-administration studies often use cue light illumination to signal the drug infusion or a time-out period (no drug availability) after the infusion. The enhanced responding for visual stimuli observed in HR rats also may have implications for cue-elicited effects associated with drug, such as conditioned reinforcement and cue-induced relapse. Thus, it is important to determine whether the differences between HR and LR rats reflect primary reinforcement associated with the drug or the cue.

One somewhat puzzling feature of the novelty responder test is that it does not reliably predict drug-induced CPP. With cocaine CPP, HR rats have been reported to be less sensitive or equally sensitive compared with LR rats (Erb and Parker, 1994; Dellu et al., 1996; Kosten and Miserendino, 1998; Shimosato and Watanabe, 2003; Mathews et al., 2010). The lack of difference between HR and LR rats may be specific to cocaine, however, as HR rats are more sensitive to amphetamine and morphine CPP (Zheng et al., 2003, 2004; Pelloux et al., 2004). Thus, individual differences may predict only the direct primary reinforcing effect of cocaine as measured by self-administration but not the conditioned rewarding effect as measured by CPP (Bardo and Bevins, 2000).

Regarding the neural mechanisms mediating HR and LR differences (Fig. 1), a general view is that an ascending mesolimbic dopamine projection emanating from the midbrain ventral tegmental area to the nucleus accumbens via the medial forebrain bundle represents one important component of the neural circuitry, thus overlapping with drug reward-relevant circuitry (Wise and Rompre, 1989; Bardo et al., 1996; Berridge and Robinson, 1998; Kelley and Berridge, 2002). Although experimenter-delivered electrical stimulation of the medial forebrain bundle produces similar reward in HR and LR rats (Antoniou et al., 2004), basal firing of midbrain dopamine neurons is enhanced in HR rats, perhaps as a result of subsensitivity to impulse-regulating autoreceptors (Marinelli and White, 2000). HR rats also show more persistent mesoaccumbal impulse flow after withdrawal from cocaine self-administration (McCutcheon et al., 2009). Several downstream cellular changes also have been identified, including greater extracellular dopamine in nucleus accumbens and striatum (Piazza et al., 1991b; Hooks et al., 1992; Thiel et al., 1999), greater velocity of dopamine uptake in nucleus accumbens (Chefer et al., 2003), and greater mRNA levels for tyrosine hydroxylase and dopamine D1 receptors (Saigusa et al., 1999). Additionally, HR rats have a reduced density of accumbal dopamine D2 receptors (Hooks et al., 1994b), suggesting a decreased number of release-regulating autoreceptors or a compensatory downregulation of postsynaptic receptors in response to increased presynaptic dopamine release.

Individual differences in reward neurocircuitry also extend to the mesocortical dopamine system. HR rats have decreased dopamine transport function in the medial prefrontal cortex (Kosten et al., 2005b), which presumably results in increased basal extracellular dopamine, similar to the results obtained in nucleus accumbens. A role for the dopamine transporter also is implicated by HR rats showing greater locomotor activity after the DAT inhibitor GBR12909 compared with LR rats (Hooks et al., 1994a). The immediate early gene Narp is enhanced in prefrontal cortex of HR rats (Lu et al., 2002). Since prefrontal cortical regions are important for behavioral inhibition (Fuster, 2008), these cellular differences may be involved in the disinhibited responding observed in HR rats using a differential reinforcement of low rate (DRL) schedule (Stoffel and Cunningham, 2008). In a DRL schedule, reinforcement is contingent on the occurrence of a response separated from the preceding response by a fixed time (e.g., 10 s), thus providing a measure of inhibitory control during timed behavior (Hodos et al., 1962).

The prefrontal cortex interconnects with amygdala and hippocampal circuitry involved in affective disorders, fear conditioning, and output modulation of the hypothalamic-pituitary-adrenal (HPA) axis (Ressler and Mayberg, 2007; Rodrigues et al., 2009). Activation of the HPA axis involves triggering corticotrophin-releasing hormone (CRH) from the hypothalamus, which elicits release of adrenocorticotropic hormone (ACTH) from the pituitary and subsequently elevates corticosterone from the adrenals (Sapolsky et al., 2000). The neurocircuitry involving hippocampal-prefrontal-amygdala interconnections with the HPA axis may play a role in individual HR and LR differences. Whereas basal levels of corticosterone do not differ between HR and LR rats (Piazza et al., 1991a; Mantsch et al., 2001), HR rats have elevated corticosterone in response to novelty and after repeated cocaine relative to LR rats (Piazza et al., 1990; Dellu et al., 1996; Lucas et al., 1997). These results implicate a differential HPA sensitivity between HR and LR animals.

Blocking different components of the HPA axis can reduce the differences between HR and LR rats. For example, adrenalectomy reduces HR/LR differences in morphine-induced hyperactivity (Deroche et al., 1993). With regard to the hippocampal-prefrontal-amygdala circuitry, inactivation of the central nucleus of the amygdala with the GABA_A agonist muscimol also reduces the individual differences in amphetamine self-administration and locomotor sensitization (Cain et al., 2008, 2009). Whereas direct blockade of hippocampal or prefrontal activity has not been examined directly in HR and LR rats, affinity of hippocampal type I and II corticosteroid receptors is decreased in HR rats (Maccari et al., 1991), likely because of a compensatory response to excessive levels of corticosterone. HR rats also have increased fibroblast growth factor expression in hippocampus compared with LR rats (Turner et al., 2008), although this difference is negated by repeated cocaine. This last study indicates that individual differences interact with cocaine to regulate gene expression in stress-relevant brain systems.

Since learning may be enhanced when glucocorticoids are elevated to some optimal level (de Quervain et al., 2009), novelty- or drug-induced elevations in corticosterone in HR rats may increase drug self-administration as a result of enhanced learning of the operant response. However, as indicated previously, HR rats do not show superior performance over LR rats when reinforced with a standard food diet (Gulley, 2007) or when tested in nonoperant tasks (Xu et al., 2001), suggesting that the increase in drug self-administration in HR rats does not simply reflect improved learning. Instead, corticosterone may have direct reinforcing effects alone (Piazza et al., 1991a, 1993), as well as potentiating the reinforcing effects of stimulant drugs (Piazza et al., 1991a; Goeders, 2002). The ability of corticosterone to potentiate drug reinforcement likely reflects an interaction with the mesolimbic dopamine system (Gilad et al., 1987; Piazza and Le Moal, 1997).

In summary, individual differences in activity in an inescapable novel environment is a reliable predictor of the stimulant and reinforcing effects of abuse drugs, with HR rats showing greater sensitivity than LR rats. This relation exists across each major drug class. HR rats also show greater sensitivity to the reinforcing effects of palatable food (sucrose) and novel visual stimuli. At least two overlapping neural systems play a role in the behavioral response to drugs in HR rats compared with LR rats: 1) the reward-relevant mesolimbic dopamine system and 2) the stress-relevant HPA system regulated by a hippocampal-prefrontal-amygdala loop.

2. Novelty Seeking

Although the novelty responder test used to define HR and LR animals has sometimes been referred to as a measure of “novelty seeking,” this is inappropriate because it is unclear whether activity in an inescapable novel environment reflects either exploratory or escape behavior. Instead, novelty seeking may be better measured as a preference for a novel context compared with a familiar context, a measure that is not correlated reliably with activity in the novelty responder test (Pelloux et al., 2004). Thus, novelty seeking should be defined by giving animals a choice to either approach or avoid novelty.

The most common method used to measure individual differences in novelty seeking in rats is the novelty place preference test. Rats are first habituated to one distinct context of a CPP apparatus by being placed in a side compartment for one or more sessions, typically lasting 30 min each. Rats then are given free-choice access to the entire apparatus and are categorized as either high or low novelty seekers based on their preference for the novel compartment relative to the familiar compartment (Hughes, 1968; Parker, 1992; Bardo et al., 1993). A variation of this method is to allow rats a choice between novel or familiar objects. This can be accomplished using either two objects (one novel and one familiar; Ennaceur and Aggleton, 1997) or multiple objects (one novel object and several familiar objects; Nicholls et al., 1992). Regardless of whether a two-object or a multiple-object test is used, it is critical that rats are habituated to the context in which the objects are presented. Thus, the only novel feature on the test day is a single object.

In mice, a holeboard test also can be used to measure individual differences in novelty seeking. In this test, mice are placed in an open-field apparatus that has a series of holes in the floor. The number of head dips is used to measure novelty seeking (File and Wardill, 1975; Kliethermes and Crabbe, 2006). Although this test has some features reminiscent of the novelty responder test (i.e., it is inescapable), it is interpreted typically as a measure of novelty seeking, and thus it is included here.

Operant conditioning procedures also are useful for measuring individual differences in novelty seeking. Some classic work has shown that primates and rodents will emit an operant response to obtain access to novel stimuli (Marx et al., 1955; Fiske and Maddi, 1961). More recently, operant conditioning measures of novelty seeking have been used to predict individual differences in sucrose self-administration in mice (Olsen and Winder, 2010). This measure of novelty seeking overlaps to some extent with the novelty responder test, as HR rats respond more for a novel visual reinforcer compared with LR rats (Gancarz et al., 2011).

Compared with the novelty responder test of Piazza et al. (1989), the predictive relation between individual differences in novelty seeking and drug reward is weaker (Table 1). Some reports show that high novelty seekers are more sensitive than low novelty seekers to the locomotor and reinforcing effects of amphetamine (Bevins et al., 1997; Cain et al., 2005, 2006) and oral nicotine self-administration (Abreu-Villaca et al., 2006). However, other reports show no significant relation between novelty seeking and stimulant activity induced by amphetamine or methylphenidate (Klebaur and Bardo, 1999; Wooters et al., 2006) or self-administration of amphetamine (Klebaur et al., 2001a; Marusich et al., 2011), and the effects with oral alcohol self-administration are mixed (Bienkowski et al., 2001; Johansson and Hansen, 2002). The effects with CPP are also mixed, as individual differences in novelty seeking predict amphetamine CPP in some studies (Klebaur and Bardo, 1999; Robinet et al., 1998), but not in all studies (Pelloux et al., 2004). Given the range of methodological differences among studies, it is difficult to ascertain the critical variables that may explain these mixed findings. However, most of these studies examined only acquisition and/or maintenance of behavior. In contrast, recent evidence indicates that novelty seeking may be a better predictor of compulsive self-administration (persistence of responding in the absence of drug) compared with acquisition of self-administration (Belin et al., 2011). “Compulsivity” is defined operationally by various procedures, including 1) persistent responding when a cue signaling the drug is no longer available; 2) high progressive ratio (PR) breakpoints, determined by increasing the response requirement after each reinforcer delivery until there is a cessation of responding; and 3) resistance to punishment (Belin et al., 2011). To the extent that compulsivity models dysfunctional drug abuse, these results suggest that novelty seeking may be useful for identifying individuals at greatest risk.

Genetic models of novelty seeking yield mixed effects on drug self-administration. In one study, mice bred selectively for novelty seeking using the holeboard test showed no differences in response to amphetamine activity, CPP, or oral self-administration of methamphetamine or alcohol (Kliethermes et al., 2007). In contrast, another study using inbred rat strains showed a relation between novelty seeking using the place preference test and i.v. amphetamine self-administration (Meyer et al., 2010); no strain-dependent differences were observed using the novelty responder test. These discrepant findings do not likely relate to the different routes of administration, as novelty seeking in the place preference test also predicts oral amphetamine self-administration in rats (Pelloux et al., 2004). Instead, the discrepant results more likely reflect a difference in species (mouse versus rat) and/or novelty test (holeboard versus place preference) used between studies.

Further insight regarding the differences between the novelty responder and novelty seeking tests were revealed by Cain et al. (2005). In that report, a large number of rats (n =165) were tested in both the novelty responder and novelty place preference tests before being trained in amphetamine self-administration across different FR schedules of reinforcement; the large sample size afforded the opportunity to apply analytic techniques normally reserved for large-sample human studies. A hierarchical regression analysis of these preclinical results indicated that the novelty responder test was a better predictor than the novelty seeking test for acquisition of amphetamine self-administration. However, when each variable was entered sequentially into the regression, novelty seeking significantly improved the predictive power of the novelty responder test. These latter results indicate that novelty seeking is not redundant with the novelty responder test and that individual differences in novelty seeking contribute to drug self-administration.

Individual differences in novelty seeking also play a role in the relation between reward cues and drug self-administration. Rats showing the greatest approach to food-associated cues (“sign trackers”) show the greatest novelty seeking (Beckmann et al., 2011) and cocaine self-administration (Beckmann et al., 2011; Saunders and Robinson, 2011). In contrast, no relation is evident between sign tracking and the HR/LR test, suggesting that novelty seeking may mediate uniquely the relation between approach to reward-related cues and cocaine reinforcement.

From a neurobehavioral perspective, one reason why novelty seeking may be a weaker predictor of drug self-administration than the HR/LR novelty responder test is that it is mediated by only a portion of the circuitry outlined in Fig. 1. Specifically, novelty seeking may differ from the novelty responder test because it involves primarily the mesolimbic dopamine system (Bardo et al., 1996), rather than the stress axis that modulates drug reward (Piazza and Le Moal, 1997). Consistent with this, inescapable novelty, but not free-choice novelty, elevates levels of corticosterone (Misslin et al., 1982).

Considerable evidence indicates that free-choice approach to novelty activates directly the reward-relevant mesocorticolimbic dopamine circuitry. Novelty place preference is blocked by dopamine antagonists (Misslin et al., 1984; Bardo et al., 1993) and by depleting dopamine levels in nucleus accumbens and forebrain with the neurotoxin 6-hydroxydopamine (Pierce et al., 1990). When rats enter a novel compartment, there is a rapid, transient surge in extracellular accumbal dopamine measured by in vivo voltammetry (Rebec et al., 1997a, 1997b), as well as a novelty-induced response in accumbal single-unit electrophysiological activity (Wood and Rebec, 2004). Thus, high novelty seekers may show enhanced drug reward as a result of greater activation of mesocorticolimbic dopamine systems.

Monoamines other than dopamine also play a role in novelty seeking. For example, accumulating evidence indicates that norepinephrine activity in the hippocampus is involved in novelty signal detection (Knight, 1996). Although no direct neural connections have been found between the ventral tegmental area and the hippocampus, the hippocampus sends projections to the medial prefrontal cortex, nucleus accumbens, amygdala, and septal area, which relay input to the ventral tegmental area (Floresco and Grace, 2003; Lisman and Grace, 2005; Luo et al., 2011). Similar to the novelty-induced increase in mesocorticolimbic dopamine, exposure to novel environmental stimuli increases the concentration of extracellular norepinephrine in the frontal cortex assessed by in vivo microdialysis (Feenstra et al., 2000), which presumably reflects an increased impulse flow of neurons in the locus coeruleus. In any case, these results indicate that the ventral tegmental area acquires information about stimulus novelty via circuitry involving the prefrontal cortex and hippocampus, sites for memory storage and novelty detection.

Both 5-HT and glutamate systems are also implicated in novelty seeking. The 5-HT1A agonist 8-OHDPAT produces a dose-dependent decrease in novel object exploration in rats, whereas the 5-HT1A antagonist WAY-100635 produces an increase in novel object exploration (Carey et al., 2008). It is important to note that these 5-HT-mediated effects are obtained without any change in locomotor activity, thus demonstrating that the effects are specific for approach to novelty rather than general exploratory behavior. Similarly, a role for glutamate has been illustrated in rats conditioned to approach an environmental context paired previously with a novel object (Bevins and Bardo, 1999). The noncompetitive NMDA antagonist MK-801 given during the conditioning phase blocks this effect, thus implicating a role of NMDA receptors in learning produced by stimulus novelty. Given the importance of glutamate-dopamine interactions in drug reinforcement and addiction (Kalivas, 2009), a more complete investigation of glutamatergic systems mediating novelty seeking is warranted.

In summary, although individual differences in novelty seeking measured in a free-choice preference test are not as strongly predictive of drug self-administration as individual differences in inescapable novelty (HR/LR test), they predict drug self-administration using a large sample size and may be especially predictive of compulsive drug self-administration. However, based on a lack of information, it is unclear whether these conclusions generalize beyond the stimulant class. In any case, in contrast to the two overlapping neural systems (reward and stress) involved in inescapable novelty, the reward-relevant mesolimbic dopamine system is associated primarily with individual differences in novelty seeking.

3. Impulsivity

Impulsivity is a broad psychologic construct that appears in virtually all major theories of personality (Whiteside and Lynam, 2001), being incorporated into various psychiatric diagnoses such as anxiety and bipolar mood disorders, as well as conduct disorder, attention deficit hyperactivity disorder, and substance use disorders (American Psychiatric Association, 2000). With preclinical models, although the term disinhibition may be more appropriate because it connotes a task-specific deficit or loss of an active neurobehavioral process (inhibition), we use the more commonly used broad term impulsivity. Individual differences among laboratory animals in impulsivity are measured by a host of behavioral tasks, including delay discounting, fixed consecutive number, five-choice serial reaction time task (5-CSRTT), go/no-go, DRL, and stop-signal reaction time. A comprehensive coverage of these tasks is provided in several excellent reviews (Evenden, 1999; Winstanley et al., 2010; Dalley et al., 2011). It is important to note that there is little relation in performance among these various tasks (Anker et al., 2009; Marusich et al., 2011), suggesting that each task measures a different facet of impulsivity. One nomenclature is to parse impulsivity tasks into three broad categories: 1) impulsive choice, which is primarily decision making when choosing between a small immediate reward and a larger delayed reward; 2) impulsive action, which is primarily motoric; and 3) impulsive reflection, which is premature responding before adequate sensory processing (Dalley et al., 2011). Regardless of the category, individual differences in impulsivity do not correlate with individual differences in either the novelty responder test or novelty seeking (Bardo et al., 2006; Marusich et al., 2011; Molander et al., 2011), thus implicating dissociable neurobehavioral systems.

Among the various tasks, those categorized as measures of impulsive choice using a delay of reward are linked most closely to drug self-administration (Table 2). Perhaps the best example is the delay discounting task, which allows animals to choose between an immediate small reward and a delayed large reward, with impulsivity being defined as a preference for the small immediate reward (Rodriguez and Logue, 1988). When screened initially on delay discounting, rats that are high in impulsivity show faster acquisition, escalation, and reinstatement of self-administration with cocaine (Perry et al., 2005, 2008; Anker et al., 2009), methylphenidate (Marusich and Bardo, 2009), nicotine (Diergaarde et al., 2008, 2012), morphine (Garcia-Lecumberri et al., 2011), and alcohol (Poulos et al., 1995). The predictive effect of delay discounting generalizes to amphetamine CPP (Yates et al., 2012), indicating that lever pressing is not a prerequisite to show a relation between individual differences in impulsive choice and drug reward. In contrast, this last study did not find any relation between impulsive choice and amphetamine-stimulated activity. Similarly, rats that are high or low in impulsive choice do not differ in cocaine-induced hyperactivity (Perry et al., 2005), and mice that are high in impulsivity are less sensitive to the hyperactivity produced by acute alcohol (Mitchell et al., 2006). Thus, impulsive choice specifically predicts the reinforcing effect of abused drugs rather than a nonspecific alteration in ongoing behavior.

In addition to delay discounting, individual differences in 5-CSRTT performance predict stimulant self-administration (Dalley et al., 2007; Diergaarde et al., 2008). 5-CSRTT involves detection of five visual targets to earn food, with premature responding to a target being punished by a time-out period (food omission), thus defining impulsivity (Robbins, 2002). In contrast to delay discounting, 5-CSRTT is a better predictor of compulsive drug intake than regulated intake (Dalley et al., 2011). For example, with 5-CSRTT performance, rats that are high in impulsivity display greater cocaine seeking than rats that are low in impulsivity, even when responding is punished (Belin et al., 2008; Economidou et al., 2009). This conclusion may not generalize beyond stimulants, however, because individual differences in 5-CSRTT performance do not predict heroin self-administration (McNamara et al., 2010; Schippers et al., 2012).

Lesion studies show that several brain regions are associated with individual differences in impulsivity, although the role of these regions varies across tasks. In general, dopamine-mediated sensorimotor systems involving the nucleus accumbens core and neostriatum are implicated most strongly across various facets of impulsivity. For example, impulsive choice measured by delay of reward, impulsive action measured by stop-signal reaction time, and impulsive reflection measured by 5-CSRTT and DRL are each increased with damage to nucleus accumbens core or dorsal striatum but not the nucleus accumbens shell (Cardinal et al., 2001; Eagle and Robbins, 2003; Pothuizen et al., 2005). The neostriatal brain system is also implicated in habit formation after repeated drug exposure (Belin et al., 2009; Goldstein et al., 2009; Balleine and O’Doherty, 2010). However, the task-dependent facets of impulsivity also are dissociable based on various lesion studies showing that 1) impulsive choice is increased by damage to medial prefrontal cortex and basolateral amygdala (Weissenborn et al., 1997; Winstanley et al., 2004; Gill et al., 2010); 2) impulsive action is increased by damage to the orbitofrontal cortex and subthalamic nucleus (Eagle et al., 2008); and 3) impulsive reflection is increased by damage to anterior cingulate cortex, infralimbic cortex, nucleus basalis magnocellularis, and hippocampus (Muir et al., 1996; Bannerman et al., 1999; Chudasama et al., 2003; Harati et al., 2008). Although some discrepancies can be found in the literature, this evidence supports the general conclusion that dopamine-rich accumbal core and dorsal striatum systems involved in different forms of behavioral activation are modulated by multiple top-down cortical inputs (Dalley et al., 2011).

It is uncommon to find lesion sites that decrease impulsivity. In one exception, damage to the orbitofrontal cortex decreases impulsivity in a delay discounting task, although this effect may depend on individual differences in baseline impulsivity (Zeeb et al., 2010). This finding contrasts with other results showing that an orbitofrontal cortex lesion increases impulsivity in delay discounting (Mobini et al., 2002). However, orbitofrontal cortex is not a homogeneous structure, and the extent of the lesion site into medial and lateral boundaries may provide an explanation for these discrepant reports (Mar et al., 2011).

Given the lesion results showing roles of nucleus accumbens core and striatal terminal regions, it is not surprising that dopamine is implicated in impulsivity. Positron emission tomography (PET) scans reveal that rats showing high impulsivity on 5-CSRTT have greater [¹⁸F]fallypride binding in nucleus accumbens than rats showing low impulsivity, indicating that high impulsivity is associated with greater D2-like dopamine receptor availability (Dalley et al., 2007). Dopamine release may also be altered, as electrically evoked dopamine release is attenuated in accumbal tissue slices obtained from rats that are high in impulsivity based on either delay of reward or 5-CSRTT performance (Diergaarde et al., 2008). Dopamine in the medial prefrontal cortex is also involved in impulsive choice, as electrically evoked dopamine release is attenuated in this region in rats that are high in impulsivity based on a delay of reward task (Diergaarde et al., 2008).

In addition to mesocorticolimbic dopamine, in vivo microdialysis results reveal an increase in 5-HT in orbitofrontal cortex in rats performing the delay discounting task relative to yoked controls (Winstanley et al., 2006). These results are generally consistent with work from nonhuman primates showing an inverse relationship between impulsivity and 5-HT metabolites in cerebrospinal fluid (Westergaard et al., 2003). In rats, 5-HT2A antagonists and 5-HT2C agonists also attenuate impulsivity measured by either delay discounting or 5-CSRTT tasks (Paterson et al., 2012; Homberg, 2012). More work is needed to identify the specific brain regions involved in impulsivity modulated by 5-HT systems, as well as other neurotransmitter systems such as glutamate and the endocannabinoids (Pattij and Vanderschuren, 2008).

In summary, impulsivity is a broad term that has multiple facets that can be measured by a host of behavioral tests. Among the various tests, individual differences in impulsive choice measured by delay discounting are perhaps the most reliable predictor of self-administration of stimulants, opiates, and alcohol. This relation also occurs with stimulant CPP but not with hyperactivity. Brain microinjection, lesion, and microdialysis studies have revealed intricate dopamine and 5-HT neural systems involved in various facets of impulsivity. Among the critical brain regions, impulsive choice involves the nucleus accumbens core, neostriatum, medial prefrontal cortex, and basolateral amygdala.

4. Other Individual Differences

Individual differences in consumption of highly palatable tastes (e.g., saccharin/sucrose) also predict various responses to abused drugs (Carroll et al., 2008). High sucrose consumers are more sensitive than low consumers to the locomotor stimulant effect of amphetamine (Sills and Vaccarino, 1994). In contrast, high consumers are less sensitive to the locomotor stimulant effect of morphine (Sills and Vaccarino, 1998). These latter results may be explained by the biphasic effect of morphine, which is characterized by an initial depression, followed by rebound hyperactivity (Vasko and Domino, 1978). Perhaps high sucrose consumers are more sensitive to the depressant phase, as opposed to being less sensitive to the hyperactive phase. In any case, high sweet consumers also self-administer more cocaine and alcohol compared with low consumers (Bell et al., 1994; Gosnell, 2000; Carroll et al., 2002) and are more impulsive on a delay discounting task using food as a reinforcer (Perry et al., 2007). Thus, overlapping neurobehavioral mechanisms exist in preference for palatable tastes, drug intake, and impulsive choice.

Finally, individual differences in wheel-running activity also are related to stimulant self-administration in rats. High wheel runners self-administer more cocaine than low wheel runners (Larson and Carroll, 2005), although this difference does not generalize to cocaine-induced locomotion. As with sucrose/saccharin preference, however, the neural mechanisms linking wheel-running behavior to drug reinforcement are largely unknown.

1. Maternal Separation

One animal model of early life social stress is the maternal separation (MS) model (Hofer, 1970; Smotherman et al., 1977). In the rodent version of this model, neonatal pups are separated from their dam for a period of time before weaning. Studies of rodent maternal separation have used a variety of experimental parameters that affect outcome, including the duration and number of separations from the dam, the preweanling period when the separations occur, whether pups are separated as a litter or individually, litter composition, and the age when testing occurs later in life. Maternal separation has been described using different terminology across laboratories, as reviewed previously (Moffett et al., 2007); this review uses the MS notation followed by a number that corresponds to the minutes during which the separation occurs unless otherwise specified (e.g., MS360 denotes maternal separation for a 360-min period). Studies using this model also differ in terms of the control groups used, with many including animal facility-reared controls (AFR) that are left undisturbed except for custodial cage changes, nonhandled controls (NH) that are not handled or in some cases are handled by the experimenters only for cage changes, or MS0 that are briefly handled and returned to the dam each time other MS groups are separated from the dam for a specified period. For comparison across studies, these group notations will be used, even though in some cases they are not the same or may even conflict with group notation used in the references cited.

Effects of maternal separation on drug abuse later in life occur across different pharmacological classes (Table 3). In general, both increases and decreases in sensitivity to drug reward occur depending on separation length, with separations of ≥60 min generally increasing sensitivity, and brief separations of ≤15 min (handling controls) decreasing sensitivity. Interestingly, minimal separation that occurs in AFR and NH conditions is sometimes less protective against drug abuse-related behaviors than MS15. One explanation for the less-than-optimal protection in AFR and NH groups is that having the dam and pups in the same vicinity continuously may be stressful because dams in a natural setting typically leave the nest completely for brief periods (Moffett et al., 2007). Another explanation is that brief (AFR or NH groups) and longer (MS15 group) separations differentially alter maternal behaviors toward the pups once they are returned (Marmendal et al., 2004; Francis and Kuhar, 2008; Der-Avakian and Markou, 2010).

The effects of long periods of maternal separation (≥60 min) on locomotor activity produced by acute stimulant administration later in life have been mixed. Most studies show an increase (Zimmerberg and Shartrand, 1992; Kehoe et al., 1996, 1998; Pryce et al., 2001; Brake et al., 2004; Marin and Planeta, 2004; Kikusui et al., 2005) or no effect (Lehmann et al., 1998; Li et al., 2003; Marmendal et al., 2004; Kosten et al., 2005a,c; Hensleigh et al., 2011) compared with stimulant effects in controls, whereas others show a decrease (Matthews et al., 1996a; Moffett et al., 2006). Factors that likely account for these differences are drug history before acute stimulant administration, doses of stimulant tested, and sex. Effects of repeated stimulant administration are also mixed, although most studies have found that maternal separation either decreases (Matthews et al., 1996b; Li et al., 2003) or has no effect (Weiss et al., 2001; Planeta and Marin, 2002; Brake et al., 2004; Muhammad and Kolb, 2011) on locomotor sensitization, whereas one study found enhanced sensitization (Kikusui et al., 2005). The age at which animals are tested for sensitization may contribute to differences observed across these studies as Weiss et al. (2001) and Kikusui et al. (2005) tested animals during adolescence, and in fact, the latter study found that the sensitized response did not persist into adulthood in female mice, although it did persist in male mice. Another complicating factor is that a previous history of saline injections can sensitize rats to acute amphetamine (Brake et al., 2004), potentially obscuring drug-sensitization effects.

Prolonged maternal separation increases stimulant reward, although not all studies support this conclusion (Faure et al., 2009), and the effect appears more reliable in males than in females. In males, maternal separation increases intake of oral amphetamine (Vazquez et al. (2006) and i.v. cocaine (Kosten et al., 2000, 2004; Zhang et al., 2005; Moffett et al., 2006), although one study using a 24-h discrete trial reinforcement schedule found no effect of maternal separation in cocaine self-administration in either male or female rats (Lynch et al., 2005). In females, increases in cocaine self-administration are obtained (Matthews et al., 1999; Kosten et al., 2006; Moffett et al., 2006); however, decreases in intake also have been reported (Kosten et al., 2004; Matthews et al., 1999). This sex dependency is corroborated by other evidence showing that maternal separation increases amphetamine reward as measured by intracranial brain stimulation threshold in male rats (Der-Avakian and Markou, 2010) but not in female rats (Matthews and Robbins, 2003). In maternally separated females, the D2 antagonist raclopride increases brain stimulation reward threshold (Matthews and Robbins, 2003), suggesting that females have enhanced anhedonia in adulthood after prolonged maternal separation. The reason for the discrepancies across studies using female rats is unclear, but it likely relates to hormonal fluctuations during the estrous cycle, which influence stimulant self-administration and stimulant seeking in female rats (Roth et al., 2004; Fuchs et al., 2005; Kippin et al., 2005; Jackson et al., 2006; Feltenstein and See, 2007).

There is some support for a protective effect of brief maternal separation on stimulant reward. Campbell and Spear (1999) found that brief neonatal isolation attenuates amphetamine CPP during early adulthood. With cocaine self-administration, MS15 male and female rats fail to acquire cocaine self-administration with a unit dose (0.0625 mg/kg per infusion) that supports self-administration in controls and MS180 rats (Moffett et al., 2006). Further, a protective effect of brief maternal separation against cocaine self-administration is obtained in MS10 males receiving the separation during the first postnatal week, whereas MS10 females separated briefly during the second, but not the first, postnatal week self-administer more cocaine (Flagel et al., 2003). Thus, sex- and time-dependent differences exist in the protective effect of brief maternal separation.

Other models of neonatal experience have examined vulnerability to stimulant behavioral effects. Neonatal sibling deprivation during the first 2–3 weeks of life attenuates the locomotor and rewarding effects of cocaine in male rats relative to controls reared in a litter of four males and four females (Li et al., 2008). Although no differences were reported in females, only one conditioning regimen was examined, and thus it is possible that sensitivity of females to stimulants may be affected under different conditioning parameters. In any case, this sex-dependent effect is interesting in light of other results showing increased vulnerability to cocaine self-administration in male rats, but not female rats, subjected to neonatal stress from cross-fostering (Vathy et al., 2007). Increases in locomotor activity to amphetamine also occur in male rats artificially reared in isolation with low levels of maternal-like stimulation, but not with high levels of stimulation (Lovic et al., 2006). These findings corroborate the general conclusion that males are more susceptible than females to stimulant effects after the loss of normal social interactions during the neonatal period.

Prolonged maternal separation also enhances the abuse-related effects of opiates. Locomotor sensitization to repeated morphine is enhanced by maternal separation in rats (Kalinichev et al., 2002). Maternal separation also increases morphine CPP (Vazquez et al., 2005b, 2007; Michaels and Holtzman, 2008). Morphine CPP is enhanced in a model comparing artificial rearing to mother rearing in male rats (Lomanowska et al., 2006), and neonatally isolated rats consume more morphine than controls when given access via a two-bottle choice between morphine and water (Vazquez et al., 2005b). Interestingly, conditioned place aversion produced by a κ-receptor agonist, spiradoline, is attenuated by maternal separation in males but not in females (Michaels and Holtzman, 2008). These results suggest that, at least in males, early life stress may enhance opiate abuse vulnerability by increasing reward sensitivity, as well as by decreasing the aversive effect produced at κ-opiate receptors.

Surprisingly, there is little information about the effect of maternal separation on opiate self-administration. Using brain stimulation reward, neither morphine nor naltrexone alters reward threshold after maternal separation (Michaels et al., 2007). However, basal stimulation response rates are attenuated by maternal separation (Michaels et al., 2007), suggesting that maternal separation may produce anhedonia during adulthood, thus exacerbating drug self-administration.

Maternal separation alters alcohol reward as measured by voluntary alcohol consumption later in life, although the precise conditions required for this outcome are unclear. In general, studies reveal a U-shaped time-dependent effect in which MS60-360 and control (NH or AFR) groups show greater alcohol intake relative to MS15 in both rats (Hilakivi-Clarke et al., 1991; Ploj et al., 2003a; Roman et al., 2003; Jaworski et al., 2005) and mice (Cruz et al., 2008). However, Huot et al. (2001) failed to observe a difference between male MS15 and AFR controls, although MS180 males did exhibit enhanced alcohol intake relative to AFR controls. These last findings suggest that the detrimental effects of prolonged maternal separation are greater than the protective effects of brief separation.

A number of studies failed to observe significant effects of maternal separation on alcohol intake in rodents (Marmendal et al., 2004; Roman et al., 2004, 2005; Gustafsson et al., 2005, 2007; Gustafsson and Nylander, 2006; Marmendal et al., 2006; Advani et al., 2007; Daoura and Nylander, 2011; Oreland et al., 2011). Although the reason for the discrepancy is not clear, because of the myriad of procedural differences across studies, several tentative conclusions may be drawn. First, the time of test is important, as rats are more sensitive to maternal separation effects on alcohol intake when tested in adulthood rather than adolescence (Daoura et al., 2011). Since baseline alcohol intake is greater in adolescent rats than in adults, this difference may mask maternal separation effects in adolescents. Second, sex differences likely play a role because two studies examining female rats failed to observe effects of prolonged maternal separation on alcohol consumption (Roman et al., 2004; Gustafsson et al., 2005), although a protective effect of brief maternal separation was observed in females (Hilakivi-Clarke et al., 1991). In a study of alcohol-preferring rats, no effect of maternal separation was observed in females (Roman and Nylander, 2005). Third, alcohol concentration is important because preference varies for different concentrations across different maternal separation conditions in male rats, with MS15 rats preferring 5% alcohol and MS360 rats preferring 20% alcohol (Gustafsson et al., 2007).

Altered vulnerability to alcohol use later in life occurs in nonhuman primates using a model comparing peer-reared to mother-reared rhesus monkeys. These studies report an increase in alcohol consumption in peer-reared monkeys (Barr et al., 2004b; Higley et al., 1991, 1996; Fahlke et al., 2000; Zhang and Kosten, 2007; Newman et al., 2009). Although this effect has been observed in both male and female monkeys, there are likely sex differences in the genes that are involved. Female peer-reared monkeys with the long/short allele for the 5-HT transporter gene promoter exhibit higher alcohol intake than peer-reared females with the long/long allele, as well as higher alcohol intake than maternal-reared females regardless of genotype (Hall and Degenhardt, 2009). This finding suggests that neither peer-rearing nor genotype alone influences alcohol intake, but both are risk factors that interact such that the environmental peer-rearing factor increases vulnerability in individuals with the long/short genotype. Similarly, vulnerability to cocaine abuse is associated with a genotype x maternal rearing interaction in mice (van der Veen et al., 2008). This study showed that cross fostering two different strains of mice, C57BL/6J and DBA/2J, with two other strains of mice that exhibit high (C3H/HeN strain) versus low (AKR stain) maternal licking resulted in a maternal influence on the acquisition of cocaine self-administration. The latter mice fostered by AKR dams self-administered less cocaine relative to those fostered by C3H/HeN dams. Again, these findings suggest that genotype makes a difference only under certain environmental rearing conditions.

As with other abuse-related psychosocial differences, dopamine systems play a key role in the effects of maternal separation (Meaney et al., 2002). Prolonged maternal separation increases extracellular dopamine levels in striatum after both K⁺ and amphetamine challenge (Hall et al., 1999), as well as increasing total tissue levels of dopamine (Matthews et al., 2001; Brake et al., 2004). Maternal separation also reduces DAT levels in striatum (Meaney et al., 2002), which may contribute to enhanced dopamine responses. In contrast, brief maternal separation decreases tail pinch-induced dopamine release compared with controls; dopamine D3 receptor binding and mRNA are also decreased compared with controls (Brake et al., 2004).

5-HT systems also are linked to maternal separation-induced changes in sensitivity to stress and abused drugs (Miczek et al., 2008; Valentino et al., 2010). In rats, maternal separation decreases 5-HT functioning in amygdala, including reduced 5-HT levels, 5-HT1A autoreceptors, and 5-HT transporters (Vicentic et al., 2006; Oreland et al., 2009). In monkeys, peer-reared adults have decreased 5-HT transporters measured by PET in amygdala and related structures compared with maternal-reared adults (Ichise et al., 2006). In addition, peer-reared monkeys have decreased cerebrospinal fluid (CSF) levels of the 5-HT metabolite 5-hydroxyindoleacetic acid (5-HIAA), which is negatively correlated with alcohol consumption (Higley et al., 1996; Ichise et al., 2006). The decrease in limbic 5-HT function after maternal separation may disinhibit dopamine reward processes (De Deurwaerdere et al., 2004; O’Dell and Parsons, 2004).

Since maternal separation is a powerful stressor in neonates, it is not surprising that the HPA axis is implicated as a key neurobiological factor associated with the effects of maternal separation and abuse-related behaviors. Maternal separation alters the HPA axis, with reports of either hypersensitivity marked by elevated basal and stress-induced blood corticosterone measured later in life (Higley et al., 1991; Plotsky and Meaney, 1993; Vazquez et al., 2005a) or hyposensitivity marked by lower basal and stress-induced blood corticosterone levels (Greisen et al., 2005; Kim et al., 2005; Roman et al., 2006); however, these latter studies assayed blood after testing animals in different behavioral tasks such as the elevated plus maze. Despite these discrepant findings, prolonged maternal separation blunts the negative feedback of glucocorticoids on the HPA axis (Ladd et al., 2004) and produces a constellation of behavioral changes reflecting increased anxiety and depressive symptoms (Zhang et al., 2004; Faturi et al., 2010). Even more subtle early life stress, such as being reared by a mother who engages in low levels of licking and grooming her offspring, engenders hypersensitivity and slower negative feedback of the HPA axis (Liu et al., 1997), thus attesting to the long-lasting influence of early life mother-infant interactions.

In relation to drug abuse–related behaviors, early life levels of corticosterone are positively correlated with alcohol intake later in life. Peer-reared monkeys have higher cortisol levels than mother-reared monkeys, regardless of sex (Fahlke et al., 2000), and consume more alcohol than mother-reared monkeys (Higley et al., 1991). In addition to cortisol, peer-reared monkeys have elevated blood levels of ACTH and CSF levels of the norepinephrine metabolite 3-methoxy-4-hydroxyphenylglycol (Higley et al., 1991). The effects of maternal separation on the HPA axis are sex dependent, as the alcohol-induced elevation in ACTH is greater in peer-reared female monkeys compared with peer-reared males (Barr et al., 2004a).

As another component of the stress system, μ-opioids are implicated in the effects of maternal separation on drug abuse-related behaviors. When separated from the dam, rat pups exhibit increased distress calling, and on return they exhibit increased approach to the dam; these behaviors are decreased and increased, respectively, by μ-opioid agonists and antagonists (Carden et al., 1991; Agmo et al., 1997). μ-Opiate receptor knock-out mice fail to display preference for mother-associated cues or vocalizations in response to separation from their mothers (Moles et al., 2004). Food-deprivation stress during adulthood in MS180 female rats is associated with decreased cocaine amphetamine-related transcript (CART) and pro-opiomelanocortin (Yoo et al., 2011), the latter being the precursor for μ-peptide β-endorphin and ACTH. These peptides are decreased in the arcuate nucleus of the hypothalamus, which is involved in feeding behavior (Lopaschuk et al., 2010). These findings suggest that maternal separation produces a long-lasting dysregulation of the endogenous opioid component of the stress system.

Central δ-opioid systems also are altered by maternal separation. Met-enkephalin and preproenkephalin mRNA levels decrease in the nucleus accumbens (Vazquez et al., 2005b), and this change is coupled with a decrease in enkephalin metabolic enzymes (Irazusta et al., 1999) and δ-opioid receptors (Ploj et al., 2003b) in amygdala. In contrast, met-enkephlin-Arg6-Phe7 (MEAP), a marker of proenkephalin, is increased in the prefrontal cortex, nucleus accumbens, hypothalamus, and periaqueductal gray (Ploj et al., 2001, 2003b). After a history of alcohol self-administration, MEAP is decreased in medial prefrontal cortex and periaqueductal gray area, although increases in MEAP are evident in the striatum and hypothalamus (Gustafsson et al., 2007). In addition, a history of alcohol self-administration is linked to increases in MEAP in the hypothalamus, striatum, and amygdala (Gustafsson et al., 2005, 2007), suggesting an overlap of δ-opioid-mediated effects of maternal separation and alcohol exposure.

κ-Opioid systems also play a role in maternal separation. Dynorphin B, a marker of prodynorphin, is increased in several regions, with more widespread effects observed with brief maternal separation (i.e., pituitary gland, striatum, periaqueductal gray, hippocampus, substantia nigra, medulla) and more restricted effects observed with prolonged maternal separation (Ploj et al., 1999, 2003b). Findings in the hypothalamus have been mixed, likely the result of sex differences, as maternally separated females exhibit a decrease in dynorphin B (Gustafsson et al., 2005), whereas maternally separated males exhibit an increase in dynorphin B (Ploj et al., 1999, 2003b). Additionally, with brief maternal separation, dynorphin B is decreased in the frontal cortex and amygdala (Ploj et al., 1999, 2001, 2003b). However, opposite effects are observed in animals after alcohol self-administration. For instance, in frontal cortex, hypothalamus, and substantia nigra, alcohol-experienced MS360 rats exhibit an increase in dynorphin B, whereas MS360 nondrinkers show no change (Gustafsson et al., 2005, 2007). Taken together, endogenous opioid-mediated modulation of the HPA stress axis during maternal separation appears especially susceptible to long-lasting changes important for mediating alcohol effects, although the implications of opioid modulation of the HPA axis on other abused drugs are less certain.

In summary, there is a vast literature on the neurobehavioral effects of maternal separation during development on later sensitivity to abused drugs. Some inconsistencies have been reported, likely based on how and when the maternal separation is applied, as well as the methods used to measure drug sensitivity. One overall conclusion from these studies is that brief separation (≤15 min per day) reduces drug sensitivity, whereas prolonged separation (≥ 60 min per day) potentiates drug sensitivity. In cases where effects are observed, these effects are relatively long-term, and they tend to generalize across drug classes. The long-term behavioral effects of prolonged maternal separation involve dopamine and 5-HT systems, as well as enhanced sensitivity of the HPA axis. The altered HPA axis may be due to long-term changes in μ-, δ-, and κ-opioid receptor systems.

2. Housing Conditions

Environmental enrichment is a manipulation in which animals are exposed to housing conditions that differ in the amount of stimulus novelty and opportunity to engage in social activity (Bennett et al., 1969). In the typical procedure, rodents are housed in either an enriched condition (EC) with novel objects and conspecifics or an isolated condition (IC) without objects or conspecifics. To determine the relative contribution of novel objects and conspecifics, a separate group is often housed in a social condition (SC) with conspecifics, but without novel objects. Some studies have varied the number of conspecifics with or without novel objects (Zakharova et al., 2009). In general, neurobehavioral differences are greater when comparing EC and IC rodents, with SC rats falling intermediate between these two extreme groups. Although enrichment is often applied during the periadolescent period of development, it produces effects across the life span, and these effects are reversible (Renner and Rosenzweig, 1987). In the current review, IC animals are considered the treated group, and EC animals are considered the comparison group because IC rats are most vulnerable to drug self-administration, which is similar to animals that score high on the novelty responder, novelty seeking, and impulsivity tests described earlier.

The hyperactivity produced by acute stimulant treatment is blunted in IC rats relative to EC rats (Bowling et al., 1993; Bowling and Bardo, 1994; Bardo et al., 1995); one exception, however, is that nicotine-induced hyperactivity is increased in IC rats (Green et al., 2003). In contrast to acute treatment, repeated stimulant treatment, including nicotine, is increased in IC rats relative to EC rats (Bardo et al., 1995; Funk et al., 2005; Solinas et al., 2009; Wooters et al., 2011). One caveat in these findings is that IC rats have higher baseline activity than EC rats before any drug treatment, which can complicate interpretation of the findings (Smith et al., 2009). That is, similar to inherent differences in activity between HR and LR rats, the conclusions drawn for IC and EC rats may differ depending on whether drug-induced hyperactivity is expressed as an absolute value or as a percent change relative to the different baselines.

Similar to enhanced sensitization to repeated stimulants, IC rats show greater propensity to self-administer amphetamine and cocaine relative to EC rats (Bardo et al., 2001; Miczek et al., 2008; Green et al., 2010; Gipson et al., 2011a). This difference is greater at low unit doses than at high unit doses delivered on either FR or PR schedules. Although baseline differences in lever pressing between IC and EC rats need to be considered as described previously (Smith et al., 2009), evidence indicates that differential housing alters sensitivity to the reinforcing effects of abused drugs independent of baseline rates of lever pressing. For example, although IC rats have higher baseline (nonreinforced) rates of lever pressing, they acquire food-reinforced lever pressing at a slower rate than do EC rats (Bardo et al., 2001). It is possible that IC rats self-administer more drug in an attempt to compensate for a relative insensitivity to the reinforcing effect of stimulants, but this is unlikely because IC rats respond for lower unit doses on an FR schedule and show higher breakpoints on a PR schedule. Instead, isolation rearing likely enhances sensitivity to stimulant reinforcement.

In contrast to the effects on self-administration, isolation housing decreases amphetamine and cocaine CPP (Bowling and Bardo, 1994; Bardo et al., 1995). These findings seem paradoxical because isolation housing decreases stimulant CPP and increases stimulant self-administration. However, self-administration and CPP measure different aspects of reward, with self-administration measuring directly the reinforcing effect of the drug and CPP measuring the rewarding value of contextual cues associated with drug (Bardo and Bevins, 2000). This suggests that isolation housing may produce differential effects on the primary and secondary reinforcement associated with abused drugs. Alternatively, since isolation rearing increases impulsive lever pressing (Engler et al., 2005; Perry et al., 2008), this change may enhance the propensity to lever press for drug specifically, without altering the strength of CPP. Further, since isolation rearing also disrupts contextual fear conditioning (Weiss et al., 2004; Gresack et al., 2010), isolation-induced deficits in general Pavlovian learning processes inherent in the CPP paradigm cannot be ruled out.

In addition to the discrepancy between self-administration and CPP results using either amphetamine or cocaine, isolated housing does not produce consistent alterations in self-administration and CPP using other drugs. For example, neither methamphetamine- nor morphine-induced CPP is altered reliably by isolation housing (Smith et al., 2005; Thiriet et al., 2011), although one study found that SC mice display greater heroin CPP than EC mice (El Rawas et al., 2009). In the case of alcohol, isolation housing reliably increases free-access intake in both rats and rhesus monkeys (Kraemer and McKinney, 1985; Procopio-Souza et al., 2011), and similar results are obtained with diazepam (Wolffgramm and Heyne, 1991). Self-administration of oral alcohol also is higher in IC rats compared with EC rats (Bienkowski et al., 2001); when given a choice between response-contingent water or 10% alcohol, IC rats exhibit a preference for the alcohol lever over the water lever. Free-access intake and PR breakpoints also are higher in IC alcohol-preferring rats (Bisaga and Kostowski, 1993), although an isolation-induced increase is not obtained with standard outbred rats (Gingras and Cools, 1995), suggesting a gene x environment interaction. In sum, although isolation rearing increases self-administration of amphetamine, cocaine, and alcohol, this effect does not generalize to all drug classes and does not generalize to the reward accrued to drug-associated cues measured by CPP.

To determine the influence of conspecifics per se, several studies have compared IC and SC animals. Early work indicated that IC rats display greater self-administration and CPP relative to SC rats using cocaine or heroin, but not amphetamine (Schenk et al., 1983, 1985, 1988); however, more recent work indicates that IC rats also self-administer more amphetamine when a low unit dose (0.03 mg/kg per i.v. infusion) is used (Bardo et al., 2001). IC rats also display a leftward shift in the cocaine dose response curve compared with SC rats (Howes et al., 2000), indicating that isolation housing enhances sensitivity to stimulant reinforcement compared with social housing. However, the presence of novel objects also plays a role, as the behavioral responses to stimulants, opiates, and alcohol differ between SC and EC animals (Bardo et al., 2001; Green et al., 2003; Coolon and Cain, 2009; El Rawas et al., 2009; de Carvalho et al., 2010).

Isolation housing produces profound alterations in various neural systems (Renner and Rosenzweig, 1987). Since much of the early work in this area focused on learning and memory deficits, changes in cortical function have been examined extensively. IC rats have decreased cortical thickness compared with EC rats (Diamond et al., 1964), primarily as a result of decreases in astrocytic branching and in brain capillaries associated with a lowering of mitochondrial metabolic activity (Sirevaag and Greenough, 1991). Neuronal cytoarchitecture is altered, characterized by a decrease in the number and size of dendrites, as well as a decrease in dendritic branching and number of dendritic spines (Rosenzweig and Bennett, 1996; Diamond, 2001). Although these changes are most prominent in cortical regions, particularly the visual and auditory cortices, subcortical regions also are affected. For example, in the hippocampus, both nerve growth factor and brain-derived neurotropic factor (BDNF) are reduced by isolation housing (Torasdotter et al., 1998; Thiel et al., 2012). There also is a decrease in number of dendritic spines found on type I spiny neurons in striatum (Comery et al., 1996), as well as a decrease in dendritic arborization on spiny neurons in the nucleus accumbens (Kolb et al., 2003). These last results indicate that reward-relevant dopamine systems may not function optimally after isolation housing.

As shown in Fig. 2, isolation housing alters the dopamine systems involved in hyperactivity and drug reward. The brain circuitry implicated in the behavioral effects of isolation is remarkably similar to that described previously in HR rats (see Fig. 1), suggesting that isolated and HR rats share overlapping neural mechanisms. One exception to this conclusion is that DAT function in nucleus accumbens and medial prefrontal cortex differs between isolated and HR rats (see Figs. 1 and 2). Despite this, there is a similar increase in accumbal dopamine in both isolated and HR rats, which may explain the hyperactivity observed in both groups.

Although initial in vitro studies did not observe differences between IC and EC rats in electrically or amphetamine-stimulated dopamine release in striatal or accumbal tissue slices (Bowling et al., 1993; Bardo et al., 1995;), ex vivo and in vivo studies show differences in accumbal function (Jones et al., 1992). Isolation rearing increases dopamine in the nucleus accumbens, striatum, and medial prefrontal cortex (Bowling et al., 1993) and also increases the transcription factor cAMP response element-binding in the nucleus accumbens (Green et al., 2010), suggesting an isolation-induced overactivity of basal mesolimbic function. This overactivity may be due to a decrease in DAT protein in the nucleus accumbens of IC rats (Zakharova et al., 2009), thus potentiating transynaptic dopamine signaling. In contrast to the deficit in DAT functioning in the nucleus accumbens, IC rats have enhanced DAT function in the prefrontal cortex as is evident by an increase in the maximum velocity of [³H]dopamine uptake compared with EC rats (Zhu et al., 2004), an effect that involves protein kinase C (Wooters et al., 2011). In the prefrontal cortex, IC rats also have more DAT protein at the cell surface compared with EC rats (Zhu et al., 2005). As part of a complex neurocircuitry, the medial prefrontal cortex has direct inputs to the nucleus accumbens (Sesack et al., 1989), as well as indirect inputs into other structures within the mesocorticolimbic system (Tzschentke and Schmidt, 2000). Thus, the isolation-induced increase in dopamine reuptake via DAT in the prefrontal cortex may diminish the influence of this inhibitory cortical structure over limbic reward processes.

In addition to changes in basal dopamine function, isolation housing alters drug-induced changes in mesocorticolimbic function. After amphetamine, IC rats have decreased levels of the dopamine precursor dihydroxyphenylalanine (DOPA) in striatum and decreased levels of the dopamine metabolite dihydroxyphenylacetic acid (DOPAC) in striatum and the medial prefrontal cortex compared with either EC or SC rats (Bowling et al., 1993). In contrast, IC rats have increased amphetamine-stimulated levels of extracellular dopamine in striatum and the nucleus accumbens using in vivo microdialysis compared with SC rats (Jones et al., 1992), although this effect is not obtained under chloral hydrate anesthesia (Bardo et al., 1999). IC rats also show greater cocaine-stimulated release of accumbal dopamine and greater alterations in the immediate-early genes c-fos and zif-268 in striatum and the central nucleus of the amygdala and, in some cases, in the nucleus accumbens and frontal cortex (Howes et al., 2000; Solinas et al., 2009; Thiel et al., 2010). Housing-induced differences in neurochemical response to stimulant drugs do not likely reflect pharmacokinetic changes, as brain levels of [³H]amphetamine are similar in IC and EC rats after systemic injection (Bardo et al., 1999). Instead, although biosynthetic and metabolic processes are diminished, isolation housing likely enhances the readily releasable stores of dopamine from vesicular or nonvesicular presynaptic pools.

Similar to abuse-prone HR rats, described earlier, isolated animals have an overactive stress axis (Serra et al., 2005). Basal corticosterone is higher in IC than in EC rats, and the amphetamine-stimulated increase in corticosterone is greater in IC rats (Stairs et al., 2011). However, the ability of the glucocorticoid receptor antagonist RU-486 to reduce amphetamine self-administration is blunted in IC rats (Stairs et al., 2011), suggesting that chronic elevation of stress hormones may desensitize glucocorticoid receptors. Further, IC rats show reduced CB1 receptor mRNA levels in the hypothalamus and basolateral amygdala (El Rawas et al., 2011), suggesting that endocannabinoid systems may contribute to the enhanced response to stress in IC rats. Regardless of the precise mechanism, the isolation-induced increase in corticosterone would be expected to increase stimulant reinforcement (Piazza et al., 1991a; Goeders, 2002).

Beyond dopamine, other neurotransmitter systems are affected by isolation housing. In the nucleus accumbens, IC rats have decreased 5-HIAA and decreased levels of the acetylcholine synthetic enzyme choline acetyltransferase (Jones et al., 1991). 5-HT levels in the medial prefrontal cortex and hippocampus also are reduced by isolation housing (Brenes et al., 2008), whereas hippocampal norepinephrine is increased (Galani et al., 2007). Further, glutamate signaling is blunted by isolation housing. For example, IC rats have decreased glutamatergic tone mediated via mGluR2 receptors in the dorsal medial prefrontal cortex compared with EC rats (Melendez et al., 2004). Amphetamine-stimulated extracellular glutamate levels in the nucleus accumbens also are blunted in IC rats compared with EC rats (Rahman and Bardo, 2008). Pretreatment with MK-801, a noncompetitive NMDA receptor antagonist, prevents the acute amphetamine-induced increase in extracellular glutamate levels in the nucleus accumbens, thus implicating an accumbal glutamatergic mechanism in the environment-dependent effects of amphetamine.

In summary, isolate-housed animals are hypersensitive to abused drugs relative to animals raised in enriched or social housing conditions. Isolated rats self-administer lower doses of stimulants compared with EC rats. Isolation rearing also enhances the self-administration of alcohol. Although these effects indicate enhanced sensitivity to the primary reinforcing effect of abused drugs, they do not generalize to contextual conditioning measured by CPP. The isolation-induced changes in sensitivity to abused drugs are associated with overactive dopamine and HPA systems. Recent evidence also indicates that isolation rearing blunts glutamate systems in the nucleus accumbens and frontal cortex.

3. Social Reward

Initiation of drug taking nearly always occurs in a social context in which peers provide social reward for the behavior, yet surprisingly little research has been conducted with animal models to investigate the influence of social interactions at the time of drug taking (i.e., social context). The presence of a nonthreatening conspecific is highly salient and rewarding, especially in adolescent rats (Vanderschuren et al., 1997; Spear, 2000); for instance, nonthreatening conspecifics 1) elicit approach (Panksepp et al., 1984), 2) elicit ultrasonic vocalizations thought to be indicative of positive affect (Burgdorf et al., 2008), 3) are positive reinforcers (Angermeier et al., 1959; Werner and Anderson, 1976; Evans et al., 1994), and 4) produce CPP (Calcagnetti and Schechter, 1992; Crowder and Hutto, 1992; Douglas et al., 2004). Research examining drug effects in a social context suggests that this is an important variable that influences sensitivity to the rewarding effects of drugs.

Social context (i.e., being with a conspecific at the time of the drug experience) alters stimulant reward. Social reward interacts synergistically with both cocaine and nicotine to produce CPP in adolescent male rats since conditioning parameters that fail to support social or drug reward when given alone can produce CPP when given together (Thiel et al., 2008, 2009). Social reward competes with drug reward in young male rats since pairing a conspecific with one compartment and cocaine with the other compartment fails to produce CPP, even though either reward alone produces CPP (Fritz et al., 2011b,d). After establishing cocaine CPP, pairing a conspecific with the previously saline-paired compartment shifts the preference away from the cocaine-paired compartment in favor of the social reward-paired compartment (Fritz et al., 2011b). Further, young adult male mice show social enhancement of methamphetamine CPP, but only when mice paired together are both under drug influence and not when one mouse is under drug influence and the other is not (Watanabe, 2011). This last finding suggests that sharing of the drug experience may be critical for drug-social reward synergistic interactions.

The degree to which social context influences drug effects and the degree to which drugs influence social behavior appears to be related for some drugs but not for others. For instance, 3,4-methylenedioxymethamphetamine (MDMA) increases social interactions in rats (Thompson et al., 2009), and MDMA-induced locomotor sensitization is greater in a social context than in a nonsocial context (Procopio-Souza et al., 2011). Under similar conditions, amphetamine has no effect on social interaction, and amphetamine-induced sensitization is not affected by social context (Procopio-Souza et al., 2011). In contrast, morphine increases social interactions, yet morphine-induced sensitization is attenuated in a social context (Procopio-Souza et al., 2011). Thus, the relationship between social context and drug effects varies depending on the drug.

Some abused drugs may decrease the effects of social interaction. For example, exposure to a conspecific or methylphenidate alone produces CPP, whereas no CPP is produced when these stimuli are combined (Trezza et al., 2009), suggesting an inhibitory interaction between social reward and methylphenidate. One interpretation of these findings is that methylphenidate decreases social reward, perhaps by reducing play behavior (Vanderschuren et al., 2008). However, nicotine and cocaine also reduce play behavior, yet enhance social CPP (Thiel et al., 2008, 2009), suggesting that a reduction in play does not necessarily result in an inhibitory interaction between all stimulants and social reward. Further research is needed to investigate whether drug x social reward interactions depend on the drug or environmental determinants that vary across studies.

Stimulant self-administration may be enhanced by the presence of a nonthreatening conspecific. Acquisition of nicotine self-administration paired with oral delivery of a sweet-scented solution is facilitated in male and female adolescent rats when a conspecific drinks the scented solution in an adjacent cage (Chen et al., 2011). In that experiment, the divider between the cages was perforated to allow the transfer of olfactory cues. In the absence of the conspecific, rats failed to increase licking on an active spout that resulted in solution + nicotine delivery compared with an inactive spout. The social facilitation of nicotine self-administration was blocked by the nicotinic receptor antagonist mecamylamine and was stronger in the presence of a familiar rat versus a novel rat, suggesting that social cues, rather than novel cues, drive the receptor-mediated facilitation. Similarly, a conspecific in an adjacent cage sharing a clear Plexiglas wall with the self-administration cage facilitates intake of a high (0.1 mg/kg per infusion), but not a low (0.01 mg/kg per infusion) unit dose of amphetamine (Gipson et al., 2011b). This latter effect is selective for amphetamine, as rats reinforced with sucrose pellets show a social-induced disruption of intake.

In addition to stimulant drugs, opiate-induced locomotor activity and reward are altered by social context. Morphine facilitates social interaction, including play behavior (Normansell and Panksepp, 1990), yet sensitization of morphine-induced locomotor activity is attenuated in mice that receive morphine in a social context (Procopio-Souza et al., 2011). Similarly, using a highly social strain of mice (C57BL/6J), morphine CPP is inhibited in male mice when conditioning occurs in a social context compared with an isolated context (Kennedy et al., 2012). No influence of social context is observed in male mice that are socially housed.

There is strong support for modulation of alcohol intake by social context, and in general, social context facilitates alcohol intake and preference. Exceptions where social context is protective against alcohol consumption are studies comparing chronic isolation versus social housing for an extended period, which show that chronic isolation stress increases alcohol consumption and preference for high concentrations of alcohol compared with social housing (Wolffgramm, 1990; Wolffgramm and Heyne, 1991; Ehlers et al., 2007). In contrast, studies have found social facilitation of drinking using pre-exposure to a demonstrator rat that had been exposed to alcohol (Hunt et al., 2001, 2000). Males are more sensitive than females to familiarity of social partners, with a familiar demonstrator rat being the best facilitator of drinking, followed by a familiar sober rat. Less drinking occurs with an unfamiliar demonstrator and sober rat. In contrast, females exhibit comparable enhancements of drinking after exposure to either familiar or unfamiliar demonstrator rats (Maldonado et al., 2008). Adolescent rats given access to alcohol (i.e., forced consumption) consume more alcohol when pair-housed rather than individually housed; however, when these groups are later given a choice between alcohol and water, there are no group differences in alcohol consumption (Thorsell et al., 2005). Using highly social prairie voles, alcohol preference is higher under duplex housing (i.e., living quarters separated by mesh separation barrier allowing some social contact) compared with single housing, whereas there is no difference across conditions for saccharin preference (Anacker et al., 2011a). Voles predetermined to be high drinkers decrease alcohol intake when paired with a low drinker, and the reduced drinking persists even if the previously high drinker is isolated (Anacker et al., 2011b). Thus, social influence can have protracted effects on alcohol intake.

Social context also modulates the effects of alcohol-associated cues. For instance, alcohol alone produces conditioned place aversion, but if alcohol is given in the presence of an intoxicated or sober conspecific, it fails to produce conditioned place aversion (Gauvin et al., 1994), suggesting that the social context attenuates the aversive effects of alcohol. When an alcohol scent is paired with either a conscious or unconscious rat, conditioned approach occurs only to the scent paired with the conscious rat (Fernandez-Vidal and Molina, 2004). Similarly, cues paired previously with social interaction facilitate subsequent alcohol intake (Tomie et al., 2004).

The facilitation of drug intake by social context can be mediated by reduced stress reactivity compared with that experienced in isolation. Blunted and delayed plasma corticosterone levels are observed in response to amphetamine in social versus isolate housed rats (Stairs et al., 2011). Blunted plasma corticosterone responses to nicotine also are observed in adolescent male rats tested in a social context relative to those tested in isolation (Pentkowski et al., 2011), indicating that stress-related social influences exist across different stimulants.

Neural processes associated with appetitive social interactions overlap with those implicated in drug abuse, including involvement of nucleus accumbens and amygdala (Burgdorf and Panksepp, 2006; Alcaro et al., 2007). Activation of accumbal shell, amygdala (central and basolateral nuclei), and ventral tegmental area as measured by expression of zif268 in response to a cocaine-conditioned environment is blunted by the presence of social reward–conditioned cues (Fritz et al., 2011b). Lesion experiments also suggest dissociable contributions of some of these regions to drug versus social reward. Under CPP conditions in which social interaction is paired with one compartment and cocaine is paired with the other compartment, lesions of accumbal core or basolateral amygdala produce a preference for the social compartment, whereas lesions of accumbal shell produce a preference for the cocaine compartment (Fritz et al., 2011a).

Although the influence of abused drugs on social behaviors has been examined extensively (e.g., for review, see Vanderschuren et al., 1997; Burgdorf and Panksepp, 2006), relatively little is known about the neurobiology of appetitive social context at the time of drug initiation on vulnerability to continued drug use. However, some recent studies demonstrate a relationship between dopamine, opiate, and oxytocin involvement in social behavior and drug abuse–related behaviors. In male prairie voles that have formed pair bonds with their mates, amphetamine fails to produce CPP under conditioning parameters that support amphetamine CPP in sexually naïve males that have not yet formed pair bonds (Liu et al., 2011), suggesting that pair bonding reduces sensitivity to amphetamine reward. Amphetamine also increases D1 receptor binding in the nucleus accumbens and produces CPP in sexually naïve voles, but it has the opposite effect of decreasing D1 receptor binding in pair-bonded voles that also fail to acquire CPP (Liu et al., 2011). Moreover, amphetamine CPP in sexually naïve voles is altered by intra-accumbens manipulations of D1, but not D2, receptors (Liu et al., 2011). As for opioid systems, Burgdorf et al. (2007) found that stimulation of brain reward pathways produces high-pitch (50 kHz) ultrasonic vocalizations and that there is a relationship between the ability of μ-opiate receptor agonist administration into the ventral tegmental area to support CPP and to elicit ultrasonic vocalizations. Finally, oxytocin systems play a role in various appetitive social behaviors, including pair bonding, mother-infant bonding, and social approach and recognition (Insel, 1992; Young et al., 2001; McGregor et al., 2008; Young et al., 2011). Male rats given oxytocin during early adolescence later exhibit enhanced social behavior, less anxiety-like behavior, and upregulated levels of oxytocin and oxytocin mRNA (Bowen et al., 2011). After oxytocin administration in adulthood, they also exhibit less alcohol drinking (Bowen et al., 2011), suggesting that upregulation of oxytocin may have a protective effect against alcohol intake (Bowen et al., 2011). Such systems are prime candidates for investigating the neural basis of social context at the time of drug initiation and on vulnerability to development of further drug use and dependence.

In summary, social contexts are rewarding alone, and they can alter the effects of abused drugs. In the presence of social interactions, sensitivity to the rewarding effects of abused drugs is enhanced using either CPP or self-administration, and these effects are observed with both stimulants and alcohol. In a choice situation, a social context can reduce the preference for a drug-paired context. Several neural systems appear altered, including the HPA axis and dopamine systems in nucleus accumbens and amygdala. Recent evidence suggests that social reward also is influenced by oxytocin systems.

4. Social Defeat

Social defeat (SD) is used to model the effects of physical and psychologic stress on drug abuse–related behaviors in animals. SD is typically investigated in rodents selected as either dominant or subordinate, or aggressive and nonaggressive, to facilitate fighting and defeat during encounters (Hilakivi-Clarke et al., 1991; Kudryavtseva et al., 1991; Marrow et al., 1999; Shimamoto et al., 2011). The resident intruder procedure is another common method for examining SD effects on drug abuse–related behavior (Yap and Miczek, 2007). In this procedure, a male rodent (the intruder) is placed in the cage of a resident male rodent that is pair-housed with a female rodent without the female present. The resident rodent attacks the intruder, and after meeting a criterion of defeat, the intruder is removed. Sometimes before or after SD, the intruder is placed into a smaller cage within the resident’s cage, which may have small openings that allow transmittance of social cues indicative of threat (i.e., vocalizations, smells). In addition to housing with a female to stimulate territorial aggression, the resident rat is often larger (Miczek and Mutschler, 1996), from a more aggressive strain (Kabbaj et al., 2001), or selected for aggressive behavior toward intruders (de Jong et al., 2005). SD has been studied in female rodents by using a lactating female as the resident and a nonlactating, less aggressive female as the intruder (Haney et al., 1995).

In general, SD enhances subsequent effects of stimulant drugs. For instance, acute or repeated SD typically cross sensitizes animals to cocaine- and amphetamine-induced locomotion (Nikulina et al., 1998; Marrow et al., 1999; Miczek et al., 1999b; Covington and Miczek, 2001; Nikulina et al., 2004; Boyson et al., 2011; de Jong et al., 2005; Yap et al., 2005; Yap and Miczek, 2007; Dietz et al., 2008; Quadros and Miczek, 2009). In rats bred for low responsivity to amphetamine-induced locomotion, SD cross-sensitizes these animals to acute amphetamine-induced hyperactivity such that these rats do not differ from rats bred for high responsivity to amphetamine (Dietz et al., 2008). These findings suggest that SD eliminates the individual differences normally observed between high versus low amphetamine responders. However, studies examining adolescent rats and hamsters found either no effect or decreased sensitivity to stimulant-induced locomotion after SD (Kabbaj et al., 2002; Trzcinska et al., 2002; Burke et al., 2011), suggesting that cross-sensitization with stimulants occurs when SD is experienced in adulthood but not when it is experienced during adolescence. The time course of cross-sensitization between SD and stimulants also differs across studies, with some studies finding persistent effects (Nikulina et al., 1998, 2004; Covington et al., 2005) and other studies finding only transient effects (de Jong et al., 2005). Further research directly comparing effects of SD in adolescent versus adult rodents on stimulant-induced locomotion tested at varying intervals is needed to address these issues.

Sensitivity to stimulant reward also is enhanced as a result of SD (Table 4). In contrast to the results with locomotor activity, SD enhances sensitivity to stimulant CPP in both adolescents and adults (McLaughlin et al., 2006; Burke et al., 2011). The effect observed in adolescents is specific to SD, as cross-sensitization is not observed in adolescent rats pre-exposed to foot shock stress (Burke et al., 2011). SD also enhances cocaine-primed reinstatement of extinguished cocaine CPP in mice (Ribeiro Do Couto et al., 2009).

Several studies have found that SD enhances the stimulant self-administration. An increase in acquisition of cocaine self-administration occurs with continuous access to cocaine, with SD rats acquiring more rapidly than controls (Tidey and Miczek, 1997). With short access daily acquisition sessions, however, SD effects are not reliably observed early during acquisition (Haney et al., 1995; Covington and Miczek, 2001; Quadros and Miczek, 2009), which may be due to individual differences. Rats classified as high responders to amphetamine exhibit delayed self-administration acquisition on short access sessions after SD compared with nondefeated high responders (Kabbaj et al., 2001). During later acquisition sessions, however, rats classified as low responders exhibit an increase in cocaine intake after SD relative to nondefeated rats (Kabbaj et al., 2001). A similar increase in cocaine intake is observed during late acquisition sessions in nonclassified male and female rats that experience SD (Haney et al., 1995), indicating that the effect of SD is most prominent with extended self-administration training.

In animals that have acquired cocaine self-administration, SD increases cocaine intake during maintenance of self-administration (Boyson et al., 2011), and this effect is greater when low doses of cocaine (0.03–0.125 mg/kg per infusion) are available (Miczek and Mutschler, 1996). An increase in cocaine intake as a result of SD is observed consistently in animals given binge access (i.e., 24 h) to cocaine (Covington and Miczek, 2001, 2005; Covington et al., 2005; Quadros and Miczek, 2009; Boyson et al., 2011; Miczek et al., 2011b). This effect reflects a shortened inter-reinforcer interval at high doses of cocaine (>0.25 mg/kg per infusion) earned during the late portion of the binge, as well as a longer persistence of responding before stopping the binge. Most studies find that SD increases cocaine intake on a PR schedule of reinforcement in rats (Covington et al., 2005; Covington and Miczek, 2005; Covington et al., 2008; Quadros and Miczek, 2009), although this effect might not generalize to mice (Yap and Miczek, 2007).

Although relatively little is known about the effects of SD on opiate abuse–related behaviors, some results suggest that SD enhances sensitivity to opiates similar to its effects on stimulant-induced behaviors. For instance, SD cross-sensitizes rats to morphine- and heroin-induced hyperactivity (Stohr et al., 1999; Cruz et al., 2011), and acute SD reinstates extinguished morphine CPP (Ribeiro Do Couto et al., 2006). Animals subjected to SD also exhibit analgesia (Miczek et al., 1982), which is reversed by a κ-opiate receptor antagonist, demonstrating that opiate systems are affected by SD (McLaughlin et al., 2006). It is unclear whether SD increases the reinforcing effects of opiates, however, since intake of heroin is not altered, but intake of a speedball mixture of cocaine and heroin is increased (Cruz et al., 2011). Further research is needed to address more thoroughly the effects of SD on opiate self-administration.

The effects of SD on alcohol abuse–related behaviors are more complex than with other drugs, likely as a result of variation across studies in the duration between SD and drug treatment, as well as the influence of individual differences. Alcohol intake decreases when rats are tested immediately after SD or when the threat of SD is present for 6 h postdefeat (van Erp et al., 2001; Funk et al., 2005), although one study found no change when the threat of SD occurred continuously in the normal housing condition (Keeney and Hogg, 1999). The decrease in alcohol consumption is consistent with the general suppressant effects that stressors have on ongoing behavior in rats (e.g., Meerlo et al., 1996). In contrast, when access to alcohol is delayed 2 h post-SD, rats exhibit an increase in alcohol intake (Caldwell and Riccio, 2010), in contrast to the decrease in alcohol consumption immediately after SD. Also, Croft et al. (2005) found a delayed increase in alcohol preference in mice with a history of repeated SD that emerged about 2 weeks after the last defeat. This study used mice that had a low alcohol preference, and it is unclear whether an increase in preference also would be observed in high alcohol-preferring mice. Similarly, CRH knock-out mice lacking CRH1 receptors exhibit a delayed increase in alcohol consumption weeks after repeated SD episodes (Sillaber et al., 2002). Individual differences in the effects of SD on alcohol consumption involve genetic influences, as SD increases alcohol consumption in C57BL/J6 mice but not in CBA/Lac mice (Kudryavtseva et al., 1991). Individual differences in fighting experience also influence SD effects on consumption, as increases in alcohol intake are observed in mice that are defeated but not in mice that are successful in defeating another mouse (Kudryavtseva et al., 1991; Hilakivi-Clarke and Lister, 1992). Finally, Funk et al. (2005) found a transient increase in alcohol consumption in animals deprived of alcohol before SD, suggesting that alcohol deprivation interacts with SD effects on alcohol consumption.

Among the various mechanisms of action, the HPA axis has received considerable attention. SD increases plasma corticosterone (Pich et al., 1993; Ribiero Do Couto et al., 2006), in some cases for several hours (Koolhaas et al., 1997). SD also enhances corticosterone release in response to subsequent mild stressors, such as exposure to a novel environment. This effect is observed in both male and female rats, although SD females exhibit a higher initial increase in corticosterone than female controls, whereas SD males exhibit a longer lasting elevation of corticosterone relative to male controls (Haney et al., 1995). With repeated exposures to SD, corticosterone release does not habituate, unlike that typically observed with nonsocial stressors (Covington and Miczek, 2005). Enhanced activation of the HPA axis also is involved in SD-induced increases in cocaine self-administration, and this increase is blocked by a CRH1 receptor antagonist administered either systemically or into the ventral tegmental area (Boyson et al., 2011). By contrast, a blunted response of the HPA axis is observed in rats that experience SD during adolescence and then are tested for amphetamine-induced elevation in corticosterone as adults (Burke et al., 2010), suggesting that SD produces differential effects on the HPA axis across the life span.

The mesocorticolimbic dopamine system also is implicated in the SD-induced increase in stimulant abuse–related behaviors. SD or exposure to a cage where social defeat has been experienced previously increases dopamine in nucleus accumbens and enhances cocaine-induced increases in dopamine in the nucleus accumbens (Tidey and Miczek, 1997; Miczek et al., 1999a, 2011a). The enhancement of cocaine-induced dopamine may be related to the accelerated acquisition of cocaine self-administration and increased intake during a binge (Tidey and Miczek, 1997; Miczek et al., 2011a), perhaps because of an SD-induced elevation in BDNF in the ventral tegmental area (Miczek et al., 2011a). BDNF is thought to facilitate dopamine release (Altar et al., 1992; Cordeira et al., 2010), and BDNF in mesolimbic dopamine neurons has been linked to increased cocaine-taking and cocaine-seeking behaviors (Grimm et al., 2003; Graham et al., 2007).

The SD-induced changes in dopamine D2 receptors also are related to behavioral changes in response to stimulants. As mentioned previously, HR rats show greater sensitization to amphetamine and lower dopamine D2 receptor binding in striatum compared with LR rats. However, LR rats exposed to SD exhibit cross-sensitization to amphetamine and have D2 receptor levels similar to HR rats exposed to SD (Dietz et al., 2008). Thus, SD appears to reduce HR and LR individual differences in sensitivity to amphetamine, perhaps as a result of downregulation of striatal dopamine D2 receptors. In contrast, when SD is experienced during adolescence, amphetamine challenge later in life no longer downregulates dopamine D2 receptors (Burke et al., 2011), even though the SD treatment produces enhanced dopamine levels in nucleus accumbens tissue samples (Burke et al., 2010). Further research is needed to determine whether altered responses of mesocortical dopamine neurons play a role in the different behavioral outcomes of SD when experienced during adolescence versus adulthood.

Glutamate systems also are implicated in the effects of SD on stimulant abuse-related behaviors. For instance, SD cross-sensitization with amphetamine, but not amphetamine sensitization alone, is blocked by pretreatment with the glutamate mGluR5 receptor noncompetitive antagonist MPEP during induction (Yap et al., 2005), suggesting a dissociation between the mechanisms involved in SD and drug sensitization effects. SD cross-sensitization with amphetamine-induced locomotion and enhanced cocaine self-administration on a PR schedule or during binge access are blocked by either MK-801 or the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor antagonist AP-5 given during induction before SD episodes (Yap et al., 2005; Covington et al., 2008). These effects are mimicked by administration of AP-5 into the ventral tegmental area, suggesting that AMPA receptors are involved in SD-induced increases in cocaine intake (Covington et al., 2008). Further support for this idea is that SD increases AMPA GluR1 receptor subunits, but it has no effect on NMDA NR1 receptor subunits (Covington et al., 2008). Glutamate NMDA NR2 receptor subunit upregulation has been implicated in SD-induced increases in alcohol consumption in CRH1 receptor knock-out mice (Sillaber et al., 2002). Further research is needed to determine the role of NR2-containing NMDA receptors in SD-induced vulnerability to drug abuse–related behaviors.

The effect of SD on alcohol intake has been dissociated from its anxiogenic effects. Croft et al. (2005) examined two different anxiolytics given before SD and found that a cholecystokinin B receptor antagonist blocked SD effects on alcohol consumption, whereas the benzodiazepine receptor agonist diazepam had no effect.

Studies examining immediate early gene expression support a role for activity in mesocorticolimbic structures, as well as other brain regions. Amphetamine challenge increases zif268 mRNA in the medial amygdala of nonstressed rats but decreases it in SD rats (Covington et al., 2005). SD also increases zif268 mRNA relative to nonstressed controls in the central nucleus of the amygdala, and this effect is reversed by an amphetamine challenge. In the infralimbic prefrontal cortex, SD decreases zif268 mRNA, whereas amphetamine challenge has no effect. After an acute cocaine challenge before a single SD episode, the immediate early gene protein Fos and its related antigen (i.e., Fos-Li) are attenuated in the periaqueductal gray, dorsal raphe, and locus coeruleus compared with nonstressed, cocaine-challenged controls (Nikulina et al., 1998). This time point corresponds to the time of induction of cocaine sensitization. Interestingly, acute cocaine administration or SD increases Fos-Li in the periaqueductal gray, but when given together, a decrease is observed, suggesting that Fos likely plays a role in plasticity underlying interactions between these two stressors (Miczek et al., 1999b). When either cocaine or amphetamine challenges are given 7 days after SD, there is an increase in Fos-Li in the periaqueductal gray and locus coeruleus, and when given 17–70 days later, there is an increase in ventral tegmental area as well (Miczek et al., 1999b; Nikulina et al., 2004). There also is an increase in Fos-Li in the central amygdala 17–70 days after SD, regardless of amphetamine challenge (Nikulina et al., 2004). Collectively, these findings suggest that plasticity in response to SD within cell bodies of monoamine neurons, as well as within the defense-relevant periaqueductal gray, is likely involved in subsequent changes in responsivity to stimulants.

In summary, social defeat is a profound stressor that produces long-lasting effects of sensitivity to abused drugs, most notably with stimulants and alcohol. SD produces cross-sensitization to the locomotor and rewarding effects of stimulant drugs. In contrast, repeated SD reduces alcohol self-administration, indicating pharmacological specificity. The increase in stimulant reward likely reflects enhanced CRF1 function within the HPA axis, although alterations in dopamine and glutamate systems also are implicated.

5. Social Rank

Social dominance or subordination can be viewed as either a state or trait. In the previous section, social defeat was considered a state induced when an animal interacts with a dominant conspecific, thus leading to a transient alteration in drug effect in the defeated animal. In the present section, social dominance or subordination is considered a trait that alters drug effects across multiple domains. As a trait, social dominance generally increases resource allocation (food, water, mating opportunities) compared with social submission. In rodents, social dominance can be measured by various behaviors, including aggressive attacks and threatening postures relative to submissive rodents (Blanchard et al., 1993). With nonhuman primates, dominance also is measured by fighting and posturing, as well as increases in general activity, the amount of grooming received, and the latency to approach a novel object (Miczek and Gold, 1983; Czoty et al., 2010).

Social submission is a known risk trait for drug abuse in laboratory animals. In rats and mice, subordinate individuals consume more alcohol and diazepam than dominant individuals (Ellison and Potthoff, 1984; Hilakivi-Clarke and Lister, 1992; Pohorecky, 2006). The elevated intake in subordinate rats is not due to state-based defeat from dominant rats because the same relationship holds when rats are tested after long-term isolation (Wolffgramm and Heyne, 1991). Similarly, subordinate monkeys show greater activity and reinforcement with cocaine or amphetamine (Miczek and Gold, 1983; Morgan et al., 2000, 2002). The difference between subordinate and dominant monkeys is blunted by chronic cocaine self-administration, but the difference re-emerges with abstinence (Czoty et al., 2010), demonstrating the stability of this trait. Subordinate monkeys also consume more alcohol than dominants (McKenzie-Quirk and Miczek, 2008). However, consumption of sucrose fluid is not altered by dominance rank (McKenzie-Quirk and Miczek, 2008), demonstrating that the effect is specific to abused drugs rather than a nonspecific alteration in sensitivity to reinforcement.

Relatively little is known about the precise neural mechanisms underlying differences in social rank. Whereas state-based social defeat activates the HPA axis (Pich et al., 1993; Ribeiro Do Couto et al., 2006), trait-based differences between dominant and subordinate animals are not associated with serum corticosterone levels (Morgan et al., 2000; Stavisky et al., 2001; Ribeiro Do Couto et al., 2006; Czoty et al., 2009; Riddick et al., 2009). Platelet monoamine oxidase (MAO) levels are reduced in subordinate rhesus monkeys (Fahlke et al., 2002). Since MAO is the major enzyme that metabolizes monoamine neurotransmitters such as dopamine and 5-HT, these results suggest that the activity of monoamine brain systems may differ between dominant and subordinate individuals. However, this conclusion is speculative since platelet MAO activity is not a reliable index of MAO activity in brain (Young et al., 1986).

Tied more closely to brain function, CSF assays suggest that altered brain activity is associated with social rank. In monkeys, the CSF level of the dopamine metabolite homovanillic acid is reduced in subordinates (Kaplan et al., 2002), whereas 5-HIAA is increased (Howell et al., 2007; Riddick et al., 2009). These CSF metabolite differences between dominant and subordinate monkeys may be greater in males (Kaplan et al., 2002), although females also exhibit this relationship (Riddick et al., 2009). Regardless of any sex differences, the enhanced vulnerability for drug abuse in subordinate individuals may reflect enhanced dopamine activity (lower metabolism), combined with blunted 5-HT activity (higher metabolism). This conclusion is consistent with evidence in humans suggesting that low levels of impulsive aggression, which typifies subordinate behavior, is associated with the combined effect of dopamine hyperfunctioning and 5-HT hypofunctioning (Seo et al., 2008).

More direct analyses of postmortem monoamine neurochemistry also reveal region-specific alterations in brain function related to social rank. A study in birds revealed that steady-state levels of striatal dopamine are decreased in subordinates compared with dominants (McIntyre and Chew, 1983); however, this dominance-subordination difference is not observed in rodents (Blanchard et al., 1991). More important, indices of functional brain activity are altered by social rank in rodents. Consistent with the results from CSF sampling, subordinate rats show elevated 5-HIAA levels in several brain regions, including the preoptic area, basal hypothalamus, hippocampus, and amygdala, although no changes in DOPAC are evident (Blanchard et al., 1991). In contrast to the 5-HT–associated effects observed in rodents, PET evidence from monkeys shows a decrease in dopamine D2-like receptor availability in subordinates, with no change in 5-HT transporter availability (Morgan et al., 2002; Riddick et al., 2009). The decrease in D2 receptor availability may be related to the enhanced sensitivity of subordinates to the reinforcing effect of cocaine compared with dominants (Morgan et al., 2002). However, PET technology using D2 selective radiotracers such as [¹¹C]raclopride or [¹⁸F]fluoroclebopride for imaging does not readily distinguish between receptor levels and alterations in dopamine release, nor does it readily distinguish between autoreceptors and heteroreceptors. In any case, an important aspect of these preclinical findings is that they parallel studies in humans showing that decreased D2 receptor availability is associated with drug reinforcement and abuse (Volkow et al., 1999, 2009).

In summary, social subordination enhances the sensitivity to abused drugs, including stimulants, benzodiazepines, and alcohol; however, little is known about social submission and dominance with opiates. The neural mechanisms of dominance and subordination have been explored primarily using nonhuman primates. These studies reveal that social subordination is associated with decreased D2 receptor availability.