Loss of dendrite stabilization by the Abl-related gene (Arg) kinase regulates behavioral flexibility and sensitivity to cocaine

Abstract

Adolescence is characterized by increased vulnerability to developing neuropsychiatric disorders and involves a period of prefrontal cortical dendritic refinement and synaptic pruning that culminates in cytoskeletal stabilization in adulthood. The Abl-related gene (Arg) acts through p190RhoGAP to inhibit the RhoA GTPase and stabilize cortical dendritic arbors beginning in adolescence. Cortical axons, dendrites, and synapses develop normally in Arg-deficient (arg^−/−) mice, but adult dendrites destabilize and regress; thus, arg^−/− mice present a model of adolescent-onset dendritic simplification. We show that arg^−/− mice are impaired in a reversal task and that deficits are grossly exacerbated by low-dose cocaine administration. Although ventral prefrontal dopamine D2 receptor levels predict “perseverative” error counts in wild-type mice, no such relationship is found in arg^−/− mice. Moreover, arg^−/− mice are insensitive to the disruptive effects of the D2/D3 antagonist haloperidol in reversal but show normal sensitivity to its locomotor-depressant actions. Arg deficiency and orbitofrontal cortical Arg inhibition via STI-571 infusion also enhance the psychomotor stimulant actions of cocaine. These findings provide evidence that stabilization of dendritic structure beginning in adolescence is critical for the development of adaptive and flexible behavior after cocaine exposure.

Adolescence is a critical period of maturation for cortico-striatal neurocircuitry and is characterized by increased risk taking, novelty seeking, and vulnerability to addictive drugs that, with repeated use, are associated with cognitive dysfunction in adulthood (1–3). In utero cocaine exposure has limited effects on cognitive performance in adult rodents (4–6), but adolescent and adult cocaine use is associated with poor inhibitory control, inflexible behavior, and drug-related compulsivity. Here, we use an animal model of late-onset dendritic simplification to test the hypothesis that the transition from dendritic refinement and synaptic pruning processes that occur in the adolescent prefrontal cortex (PFC), to the dendrite stabilization processes occurring in the adult PFC, is a period of vulnerability to cocaine.

Numerous proteins regulate cortical dendrite development; to date, however, only a few have been implicated in cytoskeletal stability in adolescence and adulthood, serving primarily to antagonize inherent cytoskeletal retraction machinery. For example, activation of the master cytoskeletal regulator, RhoA (Rho), induces neurite retraction, a process required for pruning and refining neuronal connections during development. Rho activation in mature neurons, however, leads to synapse loss and dendritic regression in a wide variety of experimental systems (7–9). Recent studies have identified a Rho inhibitory pathway involving integrin adhesion receptors, the Abl family tyrosine kinase, Arg, and the Rho inhibitor, p190RhoGAP, that acts in dendritic spines of late adolescent mice to maintain synapses and promote dendrite stability during maturation into adulthood. Cortical axons, dendrites, and synapses develop normally in Arg-deficient (arg^−/−) mice; but dendrites destabilize with age, leading to significant atrophy by 6 weeks (10, 11), corresponding to adolescence in humans (12).

Repeated psychostimulant exposure also causes dendritic spine atrophy in the adult orbital PFC (oPFC) (13), implicating drug-induced neuronal morphological decline in addiction pathology. Whether the mechanisms that ensure cortical dendrite complexity also influence sensitivity to drugs of abuse or the development of drug-related cognitive deficits is not known, in part because of the paucity of animal models of late-onset (i.e., adolescent) dendritic simplification. Because Arg-deficient mice represent such a model, the aim of this study was to test whether arg^−/− mice were more vulnerable than wild-type (wt) mice to developing i) heightened psychostimulant sensitivity, and ii) cocaine-induced deficits in inhibitory control processes, which are thought to contribute to the development of compulsive behavior in addiction (1, 14–16).

Results

Prefrontal Postsynaptic Density Protein 95 and Dopamine Receptor Levels Are Altered in arg^−/− Mice.

We first confirmed that the distribution of postsynaptic density protein 95 (PSD95)—a postsynaptic marker—was grossly normal in the adult arg^−/− PFC (Fig. 1A). To verify the expected decline in total PSD95 expression levels between pre- and postadolescent periods (10), we also analyzed PSD95 by Western blot in ventral PFC tissue samples collected at postnatal day (P) 21 or ≈210 from wt and arg^−/− mice. Here, adult arg^−/− mice had less PSD95 than all other groups, as expected based on previous morphometric characterizations [age × genotype interaction F_{(1, 33)} = 5.1, P = 0.03, post hoc P <0.04] (Fig. 1B) (11).

Fig. 1.

Ventral PFC PSD95 and DA receptor expression in arg^−/− mice. (A) Distribution of PSD95 in coronal slices collected from adult arg^−/− mice appeared grossly normal. Sections of oPFC, indicated by the white squares at left, are shown at higher magnification at Right, with green representing NeuN and red representing PSD95. Genotypes are indicated. (B) Western blot analysis revealed a down-regulation of total PSD95 expression in ventral prefrontal tissue collected from adult (P210) arg^−/− mice, relative to both young (P21) arg^−/− mice and adult wt mice, consistent with cortical dendritic regression in adulthood. (C) DA D1 and D2 receptor expression was characteristically higher in young mice than in older mice, regardless of genotype. Inset: Difference scores calculated based on variation from each animal's respective genotypic P21 mean suggest the age-related decline in receptor expression is exaggerated in arg^−/− mice. Representative blots loaded in the same order as the bars above are shown. Bars represent means (± SEM)/group (*, P < 0.05). Atlas image reprinted from 38. N = 5–14/group.

We also evaluated PFC dopamine (DA) D1 and D2 receptor expression. Expression was characteristically higher in young animals relative to older animals, regardless of genotype [main effect of age F_{(1, 27)} = 10.1, P = 0.004; F_{(1, 33)} = 88.4, P < 0.001] (17, 18), and fold change calculations indicated adult arg^−/− mice experienced greater age-related pruning of D1 and D2 than wt mice (t₁₆ = −3.1, P = 0.007; t₂₂ = −3.5, P = 0.002) (Fig. 1C). This finding is consistent with age-related spine and basal dendrite regression in arg^−/− mice (11, 19, 20).

Arg Deficiency Impairs Inhibitory Control.

Adult arg^−/− mice were first trained to respond (nose poke) on one of three nose poke recesses in operant conditioning chambers to obtain food reinforcement. Mice were trained to asymptote, at which point wt and arg^−/− mice did not differ in responses made or reinforcements earned [supporting information (SI) Fig. S1]. Next, the location of the reinforced aperture was reversed, such that mice trained to respond on the left-side aperture, for example, were required to respond on the right-side aperture to receive reinforcement. Here, arg^−/− mice performed more perseverative errors, i.e., responses on the previously reinforced aperture, than wt mice [main effect F_{(1, 13)} = 4.2, P = 0.05] (Fig. 2A). Nonperseverative, nonreinforced responses were unchanged [Fs <1] (Fig. 2C), indicating that the deficit in response inhibition in arg^−/− mice was selective to the previously reinforced aperture, as occurs with oPFC lesions (21).

Fig. 2.

Selective hypersensitivity to cocaine in arg^−/− mice. (A) Overall, arg^−/− mice performed more “perseverative” responses in a reversal task than wt control mice. Arrows indicate reversal of the response requirement with a corresponding peak in perseverative responding before mice acquired the appropriate response. (B) After repeated low-dose cocaine injection, the same arg^−/− mice performed more perseverative responses during a final reversal, with greater errors particularly during the initial session. The wt mice were unaffected by cocaine. (C and D) Nonperseverative, nonreinforced responses were unchanged before (C), as well as after (D), cocaine administration. (E) Response attenuation in the absence of reinforcement (extinction) was also unaffected by genotype when tested before or after cocaine. The break in the x axis indicates the passage of 1 day. Symbols represent means (± SEM)/group (*, P ≤ 0.05; **, P < 0.001). Asterisks next to legends indicate a main effect of genotype. N = 6–10/group.

High doses of cocaine (30 mg/kg; ref. 14) also impair reversal learning. To evaluate whether arg^−/− mice were particularly vulnerable to developing cocaine-related cognitive deficits, we subjected mice to six low-dose (10 mg/kg, i.p.) cocaine injections. After washout, these mice performed one final reversal. Here, the main effects of genotype and session on perseverative responses and an interaction between these factors were identified [F_{(1, 13)} = 11, P = 0.003; F_{(3, 39)} = 10.6, P < 0.001; F_{(3, 39)} = 7, P < 0.001] (Fig. 2B). Despite previous experience with the task, arg^−/− mice exhibited dramatically more perseverative responses during the first postcocaine session (relative to wt mice, P < 0.001). We also compared total perseverative responses during the final reversal before cocaine (final four sessions, starting at the third arrow in Fig. 2A) to those after cocaine (four sessions in Fig. 2B). This analysis suggested that wt mice were not affected by low-dose cocaine, as indicated by no change in perseverative responding, but that arg^−/− mice increased perseverative responding after cocaine [cocaine × genotype interaction F_{(1, 13)} = 6.6, P = 0.001; post hoc P = 0.003). Nonperseverative, nonreinforced responses remained unaffected (Fs <1) (Fig. 2D).

In arg^−/− mice, oPFC-dependent task performance was observed to be modestly impaired at baseline and vulnerable to further impairment by cocaine exposure, raising the possibility that cortical dendritic regression could also impair performance, or confer hypersensitivity to cocaine, in tasks that depend on more dorsal regions of the PFC (15). However, when reinforcement was withheld in extinction testing in a separate group of mice, all animals decreased responding across sessions as expected [before cocaine: F_{(5, 70)} = 5.3, P < 0.001; after cocaine: F_{(5, 70)} = 13.4, P < 0.001], and no genotype or interaction effects were found [all Fs <1] (Fig. 2E).

D2 Expression and Response Inhibition.

What might be responsible for selective behavioral vulnerabilities in arg^−/− mice? To address this question, we quantified ventral PFC D2 expression in two independent cohorts of reversal-experienced mice. Using genotype and perseverative response group (high vs. low based on a median split of response number) as factors, we found that wt mice with few perseverative responses had higher PFC D2 levels than all other groups [group × genotype interaction F_{(1, 24)} = 6.5, P < 0.02; post hoc P ≤0.02]. D2 levels in arg^−/− mice did not differ between response groups (P > 0.9) (Fig. 3A). In regression analyses, D2 expression predicted perseverative responses made by wt (Spearman's r = −6.0; P = 0.03) but not arg^−/− mice (Spearman's r = 0.2; P = 0.48) (Fig. 3B) in an initial reversal test session. This relationship between ventral PFC D2 and perseverative response number in wt mice appeared to be selective to D2, as PFC D1 expression had no relationship with perseverative responses made [Fs <1] (Fig. 3C). Dorsal striatal D1 and D2 also had no significant relationships with this outcome (group × genotype interaction ps≥0.1) (Figs. 3 D and E).

Fig. 3.

Ventral PFC D2 reflects reversal performance in wt but not arg^−/−, mice. (A) Ventral PFC D2 was analyzed by Western blot in high vs. low perseverative responders (original group sizes, 13 and 14). The wt mice that exerted fewer perseverative responses had higher prefrontal D2 expression than all other groups, but D2 failed to predict performance in arg^−/− mice. Dashed line indicates 100%, the wt mean. (B) Regression analyses revealed significant covariation between perseverative responses in an initial reversal session and D2 expression only in the wt group. (C) Ventral PFC D1 expression had no perceptible relationship with perseverative responses made by either group, nor did dorsal striatal D1 (D) or D2 (E). (F) In a separate group of mice, the D2/D3 antagonist haloperidol increased perseverative responding in wt mice performing a reversal task, but arg^−/− mice were insensitive to the compound (n = 7–11/group). (G) This was despite equivalent sensitivity to the locomotor-suppressive effects of haloperidol, as measured by photobeam breaks in a clean cage. Bars/symbols represent means (± SEM)/group (*, P < 0.05).

Together, these data suggest that there is a predictive relationship between ventral PFC D2 expression and performance in this classically oPFC-dependent task, and that this relationship is disorganized in arg^−/− mice, in agreement with selective regression of basal dendrites in these animals (11, 19). An alternative hypothesis is that no relationship exists, i.e., that the correlation between responding and D2 expression in wt mice occurred by chance. To clarify this issue, we chronically treated a group of trained mice with haloperidol, a D2/D3 receptor antagonist, and then tested mice in a reversal task. Here, drug-naive arg^−/− mice performed more perseverative responses than drug-naive wt mice, as before [interaction F_{(1, 23)} = 4.7, P = 0.04; post hoc P = 0.04] (Fig. 3F). Haloperidol increased perseverative responding within the wt group (P = 0.006), in agreement with previous reports (22). By contrast, haloperidol did not affect arg^−/− mice (P = 0.47), despite locomotor counts from mice of both genotypes being equally suppressed by haloperidol [main effect of haloperidol F_{(1, 25)} = 94.7, P < 0.001; genotype × haloperidol interaction F = 1] (Fig. 3G). This insensitivity lends support to the hypothesis that dendritic simplification in arg^−/− mice disrupts D2 regulation of inhibitory control processes.

Arg Knockout Enhances Locomotor Sensitivity to Cocaine.

We also characterized acute sensitivity to cocaine in arg^−/− mice. During habituation, total photobeam breaks in the novel field did not significantly differ between genotypes (t₁₃ = −1.9, P = 0.08) (Fig. 4A, left of break). Mice were next given five daily low-dose (10 mg/kg) cocaine injections. Analysis of photobeams broken on days 1 and 5 indicated that arg^−/− mice were more active than wt mice after cocaine [F_{(1, 13)} = 7.2, P = 0.02] (Fig. 4A). Interactions between group and day were not detected here or in the experiments described below and represented in Fig. 4 B and C, or when counts from all days were included in the ANOVAs (all Ps ≥0.06).

Fig. 4.

Orbitofrontal Arg regulates locomotor sensitivity to cocaine. (A) In a test of locomotor sensitivity to cocaine, wt and arg^−/− mice did not differ in the number of photobeams broken during habituation to a novel field (left of axis break), but arg^−/− mice show increased locomotor activity after repeated low-dose cocaine injection. (B) This profile is recapitulated in adolescent, compared with adult, mice. (C) As a control measure, cocaine sensitivity in abl^−/− mice was evaluated and found to be unaffected. (D) oPFC microinfusion of the Abl/Arg inhibitor, STI-571, recapitulated the effects of constitutive arg^−/− knockout, potentiating locomotor activity upon repeated cocaine administration above baseline levels (indicated as 100%). (E) One week later, mice previously exposed to 30 mg/kg cocaine and STI-571 showed a larger fold increase in photobeams broken after a “challenge” injection of low-dose cocaine. Bars/symbols represent means (± SEM)/group (*, P < 0.05). Sal, saline. N = 5–9/group.

The arg^−/− mice show greater sensitivity to the locomotor-activating effects of cocaine. The same phenotype has been observed in adolescent rodents tested before or during the period when the dendritic stabilizing effects of Arg are first detectable (10, 23). To evaluate the degree to which Arg deficiency recapitulates the adolescent response to cocaine, we tested locomotor activity after repeated cocaine injection in P42 (peri-adolescent) and P180 (adult) mice. As before, photobeam breaks during a 1-h initial habituation period were unaffected by group (t₁₅ = −1, P = 0.3) (Fig. 4B, left of break). After cocaine, however, adolescent mice broke more photobeams than adult mice [F_{(1, 15)} = 5.8, P = 0.03] (Fig. 4B).

As a control measure, we conducted this experiment in adult Abl-deficient mice. Although Abl and Arg share extensive homology, the loss of Abl function does not lead to significant reductions in cortical dendrites (11). Abl-deficiency did not influence photobeam breaks made during habituation (t₁₁ = −1.2, P = 0.3) (Fig. 4C, left of break). Genotype also did not affect the number of photobeams broken after repeated cocaine injection [F_{(1, 13)} = 2.1, P = 0.2] (Fig. 4C). Sensitivity to cocaine in the reversal task was also unaffected in these knockouts (Fig. S2), and we confirmed the absence of conditioned locomotor activation in response to injection and the testing chamber in all three groups of mice described in this section (Fig. S3).

oPFC Arg Inhibition Enhances Cocaine Sensitivity.

Reversal task deficits in cocaine-exposed arg^−/− mice recapitulated classical oPFC lesion effects in reversal, suggesting that Arg deficiency in this region contributed to the observed impairments. However, locomotor sensitivity to cocaine involves multiple cortical and subcortical regions in addition to the oPFC (24). To evaluate whether locomotor hypersensitivity to cocaine in Arg-deficient mice could in part be attributed to diminished Arg signaling within this region, we microinfused STI-571, an Abl/Arg inhibitor, or saline into the oPFC of adult mice and evaluated the potentiation of locomotor activity by cocaine (10 or 30 mg/kg). STI-571 mimicked constitutive arg knock-out by enhancing locomotor activity after cocaine [main effects of STI-571 and cocaine dose, F_{(1, 30)} = 15.1, P = 0.005; F_{(2, 30)} = 63.3, P < 0.001] (Fig. 4D). Mice in the STI-571 + 10 mg/kg cocaine group were more active than mice in the saline + 10 mg/kg group (P < 0.05) and indistinguishable from mice in the saline + 30 mg/kg group. Mice in the STI-571 + 30 mg/kg group had higher activity levels than all other groups (P <0.05). In other words, oPFC STI-571 infusion induced a leftward shift in the cocaine dose–response curve. Because locomotor sensitivity to cocaine was indistinguishable between wt and abl^−/− mice (Fig. 4C), we believe that the effects of STI-571 are unlikely to be caused by actions on Abl signaling.

The pattern of locomotor activity after oPFC Arg inhibition was suggestive of escalating cocaine sensitivity [STI-571, cocaine, and session interactions F_{(2, 30)} = 4.4, P = 0.03; F_{(4, 120)} = 2.6, P = 0.04; F_{(8, 120)} = 7.7, P < 0.001]. Therefore, we next tested sensitivity to a “challenge” injection of cocaine after a 1-week drug washout period. At the start of the test session, all animals were injected with saline. This was followed 1 h later by an injection of low-dose (10 mg/kg) cocaine. Photobeam breaks are represented as fold-change increases in activity. This “challenge” revealed a sensitized locomotor response in mice previously administered STI-571 and high-dose cocaine, relative to mice in the corresponding saline + 30 mg/kg group and other STI-571–treated mice [cocaine × STI-571 interaction F_{(2, 29)} = 3.4, P < 0.05, post hoc ps ≤0.02] (Fig. 4E). These data provide further evidence for regulation of cocaine sensitivity by oPFC Arg activity.

Discussion

Repeated cocaine use impairs oPFC-mediated inhibitory control, and functional neuroimaging studies in individuals addicted to cocaine consistently point to a fundamental disruption of cortico-striatal networks. These and other findings have led to the hypothesis that such disruptions may be both consequences of, and predisposing factors for, addiction. For example, ventral striatal DA D2 receptor expression levels predict impulsive and cocaine self-administration behaviors in rodents and humans (25, 26). Cortical risk factors are, however, less well characterized. Our results indicate that arresting dendrite stabilization processes that occur during adolescence—a critical period for the development of addiction—increases susceptibility to cocaine-induced oPFC-mediated behavioral deficits and psychomotor sensitivity.

We show here that adult arg^−/− mice, which exhibit 20–30% loss of cortical dendritic branch points (11) and PSD95 expression within the prefrontal cortex (including the oPFC), are deficient in a reversal task, performing more “perseverative” responses than wt mice. This response pattern recapitulates oPFC lesion effects in the same type of spatial reversal task (20). Moreover, perseverative responding after cocaine in arg^−/− mice was profoundly enhanced at a dose that left wt mice unaffected. Together with previous work indicating that higher-dose or chronic cocaine exposure similarly impairs an animal's ability to use information about response contingencies to inhibit a learned response in reversal (14, 27), our data suggest that cocaine exposure and cortical dendritic regression additively impair behavioral plasticity dependent on the oPFC. Moreover, we provide evidence that reduced D2 expression and function may in part account for behavioral deficits in arg^−/− mice, in concordance with a well-established role for D2 receptor binding in behavioral inhibitory processes (28).

A general involvement of DA systems in reversal learning is known, but increasing evidence indicates that activation of D2-like receptors (i.e., D2, D3, and D4) is particularly critical for task performance. For example, the D2/D3 antagonists, haloperidol (29) and raclopride (30), but not D1/D5 antagonists, impair monkeys' ability to reverse reinforced responses. D2/D3 antagonism (22) or D2 gene knockout (31) in rodents also retards reversal learning, suggesting diminished receptor availability impairs behavioral flexibility. In our study, perseverative errors were inversely associated with ventral PFC (including oPFC) D2 expression in genetically intact mice using two analytic approaches. That the same pattern was not detected in the arg^−/− mice suggests a disorganization of D2 receptor influence on behavioral flexibility. In support of this hypothesis, haloperidol had no effects on adult arg^−/− mice in reversal, but it impaired performance in wt mice as expected.

Consistent with the regression of basal dendritic arbors and spines in adult arg^−/− mice, we also found evidence of an exaggerated age-related loss of D1/D2 protein from high expression levels at P21 to lower levels in adulthood in arg^−/− mice (11, 19, 20). We would not necessarily expect to detect an overall D2 deficit because neuronal D2 receptors account for only a subset of total PFC D2 expression (32).

Locomotor sensitivity to cocaine was also enhanced in arg^−/− mice, and this phenotype was recapitulated by inhibiting Arg signaling within the oPFC, directly implicating this region in sensitivity to the drug (24). Although previous work might predict dendritic simplification within more dorsal and medial regions of the PFC could also contribute to heightened cocaine sensitivity (33), performance on a task sensitive to infralimbic PFC lesions (extinction) was unaffected by Arg knockout. Other medial PFC-dependent processes (e.g., sustained attention) were not tested. Whether oPFC-dependent “inhibitory control” processes are particularly vulnerable to cytoskeletal insult, as these data imply, will be a topic of future research.

This report demonstrates that pharmacological or genetic suppression of a cortical dendrite stability effector heightens cocaine sensitivity. This experimental approach is distinct from, and complementary to, investigations of the post hoc effects of psychostimulant exposure on oPFC spine regression (13) or function (26), and therefore contributes to a more complete view of structural attributes of drug-related oPFC dysfunction. Moreover, our data show that natural variation in ventral PFC D2 receptor expression levels predict perseverative responding in a reversal task in genetically intact mice. That arg^−/− mice lack this phenotype, and sensitivity to haloperidol, raises the possibility that cortical structural variations or vulnerabilities that emerge in adolescence may play a key role in addiction by disrupting D2-mediated behavioral flexibility. Further characterization of the molecular mechanisms that regulate cortical development and their impact upon neuronal responses to drugs of abuse may provide a significant advance toward a more comprehensive view of the cortical involvement in cyclical behavioral patterns in addiction.

Methods

Subjects.

Knockout mice were generated as previously described (34), housed in groups of two to five, and restricted to 120-min/day access to food to motivate responding during instrumental conditioning. Otherwise, mice were fed ad libitum. Mice were maintained on a 12-h light cycle (07:00 on). Procedures were approved by the Yale University Animal Care and Use Committee.

Immunostaining/Blotting.

For staining, adult arg^−/− mice were deeply anesthetized with pentobarbital and transcardially perfused with 4% paraformaldahyde. Brains were sliced into 40 μm-thick sections on a microtome set at −15 °C ± 1. Every third section was immunostained for NeuN (Ms; 1:5000; Chemicon, Billerica, MA) and PSD95 (Rb; 1:500; Cell Signaling, Danvers, MA). AlexaFluor goat IgGs (1:300; Invitrogen, Carlsbad, CA) served as secondary antibodies.

For Western blotting, arg^−/− mice were rapidly decapitated at P21 or ≈P210, and frozen brains were cut into 1-mm coronal slices. Bilateral ventral PFC (including oPFC) and dorsal striatal samples were collected and sonicated in lysis buffer [137 mM NaCl, 20 mM Tris-HCl (pH = 8), 1% igepal, 10% glycerol] and stored at −80 °C. Protein concentrations were determined using a Bradford colorimetric assay (Pierce, Rockland, IL). A 20-μg quantity per sample was added to 10 μl Laemmli buffer (20% glycerol, 2% sodium dodecyl sulfate [SDS], bromophenol blue) and boiled for 10 min. Samples were separated by SDS–polyacrylamide gel electrophoresis (PAGE) on 8–16% gradient Tris-glycine gels (Invitrogen, Carlsbad, CA). Primary antibodies were anti-GAPDH (Ms; 1:20K; Advanced Immunochemical Inc., Long Beach, CA), anti-PSD95 (Rb; 1:1000; Cell Signaling), anti-D1 (Rb; 1:1000; Abcam, Cambridge, MA), and anti-D2 (Rb; 1:800; Abcam). Membranes were incubated for 1 h or overnight and then incubated with IRDye 700 Dx Anti-Rb IgG and IRDye 800 Dx Anti-Ms IgG for 1 h (1:5000; Rockland Immunochemicals, Gilbertsville, PA). Bands were quantified using fluorescent densitometry analysis (LI-COR Odyssey Imaging System). D1 expression was unexpectedly undetectable in six initial samples, which were excluded.

D1, D2, and PSD95 fluorescence values were normalized to corresponding GAPDH signals; these ratios were converted to a percentage of the adult wt mean from the same membrane to control for fluorescence variance between gels. Variability in this group was generated by converting each animal's value to a percentage of its own group mean. The resulting dataset was analyzed by analysis of variance (ANOVA) with genotype ± age as factors, as appropriate. All post hoc comparisons were made with Tukey's t tests; all analyses were 2-tailed, with P ≤ 0.05 indicating a difference between groups. For Fig. 1B, difference scores were also calculated based on each adult animal's fold-change variation from its respective genotypic P21 mean; group means were compared by t test.

Instrumental Conditioning.

Experimenters used lit operant conditioning chambers (16 × 14 × 12.5 cm) controlled by MedPC software (Med Associates Inc., Georgia, VT) housed in sound-attenuating outer chambers. Head entries into the nose poke apertures and magazine were detected by photocell. A pellet dispenser delivered grain-based food pellets (20 mg; Bio-Serv, Frenchtown, NJ) upon completion of the response contingency, and a 2-s 2.9 kHz tone signaled reinforcement. The house light was also extinguished, and a light above the magazine was lit.

During training and testing, inserting the nose into one of three holes was reinforced; responding in the other holes had no consequences. The first 10 reinforcers were obtained on a fixed ratio 1 schedule, followed by a variable ratio 2 schedule, in which one, two, or three responses were reinforced in a 15-min session (35). Mice were required to retrieve each pellet before further reinforcement. The position of the active nose poke (left, right, or center) was counterbalanced. Each mouse displayed stable responding for 2 consecutive days before proceeding to the next testing phase.

Mice were matched and split into three experimental groups: One group performed a reversal task, was then exposed to cocaine, and then performed another reversal after cocaine. The second group was treated with haloperidol (Sigma Aldrich, St. Louis, MO) for 3 weeks [40 μg/ml, p.o., dissolved in tap water plus 2.4% acetic acid, resulting in ≈3 mg/kg/day (36); fluids were changed every 48 h] and then tested in reversal. A third group was used for extinction testing.

Reversal Test.

Here, mice were required to respond in a previously nonreinforced aperture to acquire reinforcement, i.e., mice trained to respond in the left aperture were required to respond in the right. One aperture was reinforced per 15-min session; for experiments with multiple reversals (Fig. 2), after the acquisition of the first reversal, mice were reversed back to the previously reinforced nose poke. Mice were then reversed to the third aperture, then exposed to low-dose cocaine. A fourth reversal occurred 1 week after the last injection.

Perseverative, nonperseverative nonreinforced, and reinforced responses were analyzed by two-factor (genotype × session) repeated-measures (RM) ANOVA. For haloperidol-treated mice, perseverative responses are represented as a percentage of total responses made, as haloperidol was sedative, reducing total responding in both treated groups. These values were analyzed by three-factor (genotype × treatment × session) RM ANOVA. The sedative effects of haloperidol were quantified using the Omnitech Digiscan Micromonitor system (Columbus, OH) equipped with 16 photocells. Photobeam breaks made in 5-min time bins over 1 h were analyzed by two-factor (genotype × treatment) ANOVA with RM.

Extinction.

To assay infralimbic PFC function and cocaine sensitivity, we tested sensitivity to non-reinforcement (extinction) before and 1 week after cocaine exposure. Test sessions were identical to those during training, except that the magazine was disconnected, and cues signaling reinforcement were withheld. Responses in the reinforced aperture were analyzed by three-factor (genotype × session × cocaine) RM ANOVA. Instrumental responding was restored with several reinforced sessions after the first extinction test and before cocaine exposure.

Cocaine Administration.

To evaluate sensitivity to cocaine in arg^−/−, abl^−/−, and adolescent mice, we used a repeated low-dose cocaine administration protocol that was not necessarily expected to affect adult wt mice. Mice were injected with cocaine hydrochloride (10 mg/kg, i.p., 1 ml/100 g; generously provided by NIDA) for 5 consecutive days after 1-h habituation to a large, clean cage normally used to house rats. Photobeams broken on days 1 and 5 were analyzed by ANOVA with RM.

We confirmed the absence of injection-associated or testing chamber–associated conditioned locomotor activation 1 week after the last session in these mice. Here, each animal was monitored for 3 h. First, mice were allowed to habituate to the chamber. Mice then received a saline injection, followed by a second injection containing 10 mg/kg cocaine (representing the animals' sixth total cocaine injection). Photobeam counts during each of the 3 h were compared by two-factor (genotype or age × hour) ANOVA with RM. Issues inherent in breeding knockout animals prohibited the inclusion of saline-only control groups, but the use of saline and cocaine injections in the same test session allowed within-subjects comparisons; findings are provided in Fig. S3.

STI-571 Microinfusion.

For surgery, wt mice of at least 12 weeks of age bred on the same strain background as arg^−/− and abl^−/− mice were anesthetized with 1:1 2-methyl-2-butanol and tribromoethanol (Sigma) diluted 40-fold with saline. The head was shaved and placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA). The scalp was incised, skin retracted, bregma and lamda identified, the head leveled, and coordinates located using Kopf's digital coordinate system with resolution of 1/100 mm. Three burr holes were drilled, and sterile saline or the Abl/Arg inhibitor, STI-571 [10 mM as described by Cancino (37); kindly provided by Bill Bornmann], was infused over 2 min. The lateral oPFC was targeted with 0.15 μl/side infused at +2.6AP, ±1.2ML, −2.8DV (38, 39). The medial oPFC was targeted with a midline infusion of 0.20 μl at + 2.8AP, −2.3DV (S.L.G., J. Howell, and J.R.T., unpublished). Needles were left in place for 2 additional min before withdrawal.

Mice were sutured and allowed to recover for 4 days, at which point cocaine-induced locomotor activity was monitored by photobeam in six experimental groups (saline+saline, STI-571+saline, saline + 10 mg/kg cocaine, STI-571 + 10 mg/kg cocaine, saline + 30 mg/kg cocaine, and STI-571 + 30 mg/kg cocaine). Because STI-571 infusions had a suppressive effect on baseline locomotor activity, activity counts were normalized to the mean generated by respective saline control groups to quantify the potentiation of locomotor activity by cocaine (e.g., photobeam breaks generated by mice in the STI-571 + 10 mg/kg cocaine group were normalized to the mean photobeam breaks made by mice in the STI-571+saline group). Variability in the control groups was generated by calculating the degree to which each animal varied from its own group mean. Photobeams broken during the five daily test sessions were analyzed by three-factor (infusion × cocaine × session) RM ANOVA.

Next, animals had a 1-week cocaine washout, then each mouse was allowed to habituate to the chamber for 1 h. Mice then received saline and were monitored for 1 h. Finally, all mice were administered a 10 mg/kg cocaine challenge. Fold-change increases in photobeam breaks after cocaine (relative to saline) were compared by two-factor (infusion × cocaine) ANOVA.