a-b, Illustration of the experimental protocol showing the stimulation electrode in the PFC and recording electrode in the NAcore. Stimulation of the PFC was targeted to the ventral prelimbic cortex (PrL; circle). The field potentials attained maximum amplitude when stimulation in PFC was delivered 3.5 mm ventral to the surface of the brain. Representative traces (average of 10 consecutive field potentials) are shown at different depths of stimulation in the PFC; arrow indicates the stimulation artifact. Current ejection of pontamine blue spot marked the recording site in the NAcore. ac- anterior commissure, Cg-cingulate cortex, IL- infralimbic cortex, NAshell- accumbens shell. c, Left, sample trace demonstrating that recorded fields are sensitive to CNQX (100 μM) and TTX (10 μM), but independent of GABAergic currents (bicuculline, Bic, 100 μM). Right, representative traces (average of 5 consecutive field potentials) before (1) and after (2) CNQX and a subtraction of the 2 traces (1-2) equivalent to the AMPA receptor dependent current. d, Effect of different drugs on field amplitude measured as an average of 5 min bins, immediately before and 25 min after drug perfusion. CNQX (100 μM, p=0.002), bicuculline (100 μM), picrotoxin (1.0 mM), TTX (10 μM, p= 0.029), compared to baseline using a Mann-Whitney test. n is shown in the bars. In all figures, scale bars correspond to 0.2 mV, 10 msec, and error bars indicate mean ± s.e.m.

a, Active vs. inactive lever pressing throughout self-administration training days (SA; 10 days), and extinction (E; 21 days), n=42. A two-way repeated measures ANOVA was used to determine that active lever pressing was greater than inactive lever pressing over last 3 days of SA, F(1, 164) = 15.40, p<0.001. b, Input-output curves from rats extinguished from self-administered cocaine vs. control animals. A two-way ANOVA was used to show that the input-output curve for cocaine was different from the control group (group F(1,374)=14.43, p< 0.001). Inset shows representative evoked field potentials for an input-output curve (0.3-1.5 mA) in a cocaine trained animal. All data shown as mean ± s.e.m.

a, Upper panel: Representative experiment inducing LTP in a control rat after HFS (50 Hz). HFS was delivered at the arrow. Insets show an average of 5 consecutive fEPSP before (1) and after (2) HFS. Middle panel: Representative experiment of LTP failure in a cocaine rat after HFS (50 Hz). Lower panel: HFS (50 Hz) failed to induce LTP in the cocaine (n=13) compared to control animals (n= 8 yoked saline, n= 8 naive). A two-way repeated measures ANOVA revealed a significant effect of cocaine treatment: F(1,1458)= 11.05, p= 0.003. b, Upper and middle panels: LTP was induced in control animals only at 50 HZ (30Hz, 50Hz, 100Hz, group F(2,247)= 4.094, p= 0.042, time F(19,247)= 2.23, p= 0.003, and LTP in the cocaine animals could not be induced by any stimulation protocol (30Hz, 50Hz, 100Hz). Lower panel: Percentage potentiation from baseline of 30, 50, and 100 Hz HFS in cocaine and control goups. Each bar represents an average of the 5 min time bin immediately before (baseline) and 15-20 min after HFS. n is shown in the bars. A one-way ANOVA revealed effect of frequency only in control rats. F(3,31)= 6.243, p< 0.001. Data shown as mean ± s.e.m.
* p< 0.05 compared to baseline using a Bonferronni post hoc analysis

a, Upper panel: Representative experiment of LTD in a control rat after LFS. LFS was delivered at the arrows. Insets show an average of 5 consecutive fEPSP before (1) and after (2) LFS. Middle panel: Representative experiment of LTD impairment in a cocaine rat after LFS. Lower panel: LTD is impaired in the cocaine rats compared to controls (3 bursts LFS protocol) (n= 3 yoked saline, n= 10 naive) using a 2-way repeated measures ANOVA, F(1,1180)= 15.21, p= 0.001. LFS was delivered at the arrows, and the dotted line corresponds to the average amplitude before LFS. b, Upper and middle panels: In control animals, 5 and 3 bursts were not different; in cocaine rats, 5 bursts LFS resulted in robust LTD compared to 3 bursts (F(1,826)= 7.95, p = 0.014). Lower panel: Percentage depression from baseline of 3, 5, 7 or 9 bursts in cocaine and control groups. Each bar represents an average of 5 min time bin immediately before (baseline) and 15-20 min after LFS. n is shown in the bar and data are shown as means ± s.e.m. A one way ANOVA revealed significant LTD in both cocaine F(4,43)= 36.88, p< 0.001 and control subjects F(4,45)= 36.56, p< 0.001. c, Cocaine withdrawn subjects underwent 3-burst LFS to induce LTD, followed by HFS to induce LTP-like potentiation (F(7,87)= 3.72, p< 0.001).
* p< 0.05, compared to baseline using a Bonferroni multiple comparisons.
+ p< 0.05, comparing 5 to 3 bursts

a, Experimental paradigm: NAC (100 mg/kg, ip) was injected 2.5 hrs before collecting baseline data (B), and 3 hrs prior to inducing LTP or LTD (S). LY345491 (1 mg/kg, ip) was administered with NAC and 2 times thereafter (arrows), and MPEP (10 mg/kg, ip) was administered with NAC and once thereafter. b, Current-voltage relationship for different groups showing that NAC normalized the input-output curve in cocaine animals without affecting control rats. (2-way ANOVA, treatment F(3,489) =8.321, p< 0.001). c-d, Representative experiments (c) and average of all experiments (d) showing that NAC restored the ability to induce LTP in cocaine rats (NAC, cocaine n=8) resulting in equivalent potentiation to LTP in control rats injected with NAC (NAC, control, n=6). NAC also restored the ability to induce LTD in cocaine rats (NAC, cocaine n=5) resulting in equivalent depression to LTD in control rats injected with NAC (NAC, control, n=4). The cocaine group from figure 2a is shown for comparison (grey). A two-way ANOVA comparing NAC, cocaine with NAC, control revealed no effect of group or interaction. All data are shown as mean ± s.e.m.

a, LY345491 (1 mg/kg, ip) prevented NAC-induced restoration of LTP (treatment F(1,374)= 6.10, p= 0.031), but did not block NAC restoration of LTD. See figure 5a for treatment protocol. b, The mGluR5 antagonist MPEP (10 mg/kg, ip) blocked the NAC restoration of impaired LTD in cocaine rats (treatment F(1,540)= 8.93, p =0.014), but did not block NAC restoration of LTP. c, Percentage change from baseline in control and cocaine animals after NAC and LY345491 or MPEP. Each bar represents an average of 5 min time bin immediately before (baseline) and 15-20 min after LFS or HFS. One-way ANOVA: LTP F(4,69)= 8.77, p< 0.001; LTD F(4,58)= 40.31, p< 0.001. n for each treatment group is shown in the bars, and data are shown as mean ± s.e.m.
* p< 0.05, compared with baseline using a Bonferroni multiple comparisons.
+ p< 0.05, compared with NAC in cocaine animals (NAC+cocaine)

a, Neither compound induces lever pressing when administered alone. b, When administered 20 min prior to a cocaine priming injection (10 mg/kg, ip) MPEP inhibited cocaine-induced lever pressing (F(3,31)= 4.76, p= 0.011). c, The administration of NAC inhibits cocaine-induced lever pressing and this effect of NAC was abolished by pretreatment with CDPPB (F(2,32)= 16.04, p< 0.001). Self-administration refers to the mean ± sem active lever presses over the last 2 days of cocaine training. In panels b and c all animals received all treatments in random order separated by a 72 hr intertrial interval. n for each treatment group is shown in the bars, and data are shown as mean ± s.e.m.
* p< 0.05, compared to extinction using a Bonferroni multiple comparisons
+ p< 0.05, comparing NAC + Vehicle with NAC + CDPPB