Anterior cingulate cortex hypoactivations to an emotionally salient task in cocaine addiction

Behavior.

Whole-Brain Activations/Hypoactivations and Correlation Analyses.

For all subjects and compared with the fixation baseline, the fMRI task produced brain activation and hypoactivation patterns similar to those reported using other blocked cognitive (7, 8) and emotional tasks (19, 20) (Table 1 and Fig. S1). Most importantly, despite a lack of significant group differences in both objective and subjective task-related behavioral measures (Table S1), there were significant whole-brain blood-oxygenation-level-dependent (BOLD) fMRI differences between the groups (Table 1). Notably, these group differences included both the cdACC [Brodmann Area (BA) 32] and rvACC/medial OFC (BA 10, 11), our 2 a priori regions of interest (ROIs). These ACC subdivisions were engaged by this fMRI task in our previous study that included another cohort of CUD (n = 14) and no controls (24). Other regions that showed significant group differences were the left dorsolateral and dorsomedial prefrontal cortex, right inferior parietal lobule, middle occipital gyrus and cuneus, and the left cerebellum.

Follow-up ROI analyses confirmed the group main effect in all these regions (F_1,32 > 4.6, P < 0.05), showing hypoactivations in the CUD compared with controls in both ACC subdivisions (Fig. 1). Interestingly, follow-up t tests showed that, in the CUD, the cdACC had the most hypoactivations in the least salient condition (neutral words at 0¢ condition; reaching significance as compared with the drug words at the 0¢ condition, t₁₆ = 2.2, P < 0.05) (Fig. 1A). In all of the other regions, the pattern was reversed such that CUD showed higher activations than controls (Fig. S2). In addition to the group main effect, the cuneus also showed a significant money main effect (F_1,32 = 4.2, P < 0.05) and a word by group interaction (F_1,32 = 5.3, P < 0.05) (Fig. S2E).

To inspect intercorrelations between both ACC subdivisions, we performed ROI correlations that were then followed by a whole-brain simple regression analysis. The ROI analyses showed positive correlations across all subjects for the most salient condition (drug words at 50¢, r = 0.34, P < 0.05) with a similar trend for the neutral words at the same monetary condition (r = 0.31, P < 0.1). The whole-brain regression analysis, where subjects' cdACC % signal change scores to the most salient condition (from a fixation baseline) were entered as seed values against their own contrast (activation) maps during the same condition, confirmed the first correlation (with rvACC, peak voxel: x = −3, y = 51, z = 0, 31 voxels, T = 3.3, threshold: P < 0.001 voxelwise uncorrected with 20 voxels minimum). Follow-up analyses using this peak voxel showed this correlation to be driven by the controls (r = 0.69, P < 0.01) and not by CUD (r = 0.21, P > 0.4). The test of coincidence of these regression lines was significant (F_2,30 = 5.4, P = 0.01), confirming that these specific ACC intercorrelations differed significantly between the study groups (this result is especially sound given that the whole-brain regression analysis was conducted across all study subjects, enhancing the possibility for uncovering similar group results but decreasing the possibility for differential group results, which were nevertheless significant in the current study).

Correlations Between the ACC and Drug Use in the CUD.

No correlations were significant at the nominal significance level (P < 0.01). At the reduced threshold for our a priori ACC region, 2 correlations were significant: (i) the more the cdACC hypoactivations, the more frequent the current cocaine use (number of days used in the past 30 days, Spearman r (r_S) = −0.56, P < 0.05). This correlation was strongest for the least salient condition (neutral words at 0¢: r_S = −0.47, P = 0.057; the other 3 conditions: r_S < −0.28, P < 0.3) and it was confirmed with a whole-brain regression analysis, where the frequency of current cocaine use by individuals was entered as seed value against their own contrast maps during the least salient condition (peak voxel: x = −3, y = −6, z = 42, 48 voxels, T = 4.2, threshold: P < 0.034 cluster-level corrected with a small volume correction (26) and a voxelwise uncorrected level of P < 0.005 with 20 voxels minimum) (Fig. 1C); (ii) the rvACC showed a trend for a correlation with the fMRI-induced change in craving (r_S = 0.36, P < 0.15), which reached significance for the most salient condition (drug words at 50¢, r_S = 0.51, P < 0.05; the other 3 conditions: r_S < 0.28, P < 0.3). This correlation was also confirmed with a whole-brain regression analysis, where fMRI-induced craving change scores of individuals were entered as seed value against their own contrast maps during the most salient condition (peak voxel: x = −6, y = 54, z = −3, 41 voxels, T = 3.7, threshold: P < 0.041 cluster-level corrected with a small volume correction (26) and a voxelwise uncorrected level of P < 0.005 with 20 voxels minimum) (Fig. 1D).

Subjects.

Fifty-one right-handed native English speakers were recruited using advertisements in local newspapers and by word of mouth. Subjects underwent a full physical and neurological examination by a neurologist and a diagnostic interview by a clinical psychologist. This interview included the Structured Clinical Interview for DSM-IV Axis I Disorders (research version) (36, 37), the Addiction Severity Index (38), the Cocaine Selective Severity Assessment Scale (39), and the Cocaine Craving Questionnaire (40). All subjects were able to understand and give informed consent. Subjects were healthy individuals, not taking any medications, and further excluded for (i) history of head trauma or loss of consciousness (>30 min) or other neurological disease of central origin (including seizures); (ii) abnormal vital signs at time of screening and history of major medical conditions, encompassing cardiovascular (including high blood pressure), endocrinological (including metabolic), oncological, or autoimmune diseases; (iii) history of major psychiatric disorder (other than substance abuse or dependence for CUD and/or nicotine dependence for both groups); (iv) history of gambling as assessed with the South Oaks gambling questionnaire (cutoff <5) (41); (v) except for cocaine in the CUD subjects, positive urine screens for other psychoactive drugs or their metabolites (phencyclidine, benzodiazepines, cannabis, opiates, barbiturates, and inhalants); (vi) pregnancy as confirmed with a urine test in all female subjects; (vii) contraindications to the MRI study; and (viii) because of the verbal nature of the fMRI task (24) education of less than 12 years (or equivalent) and verbal intelligence of ≤85.

Of these 51 subjects, 38 subjects (17 CUD, 21 controls) completed the fMRI task without loss of data due to motion or technical difficulties (see later thresholds discussion; there were no group differences in these subject exclusions, χ₁² = 1.1, P > 0.3). Four more controls were excluded to numerically equate the groups and for demographic matching that included sex, age, education, intellectual functioning, and socioeconomic status (Table S1). Race and history of cigarette smoking nevertheless differed between the groups. Their potential contribution to results was inspected as described in SI Text. Note that, of the 34 subjects included in the current study, results of 26 (15 CUD and 11 controls) were previously reported (42) as further described later (under BOLD-fMRI analyses).

All 17 CUD used crack/cocaine (mostly smoked route) in the past 30 days and met DSM-IV criteria for current cocaine dependence (n = 15) or abuse (n = 2; these 2 subjects met criteria for cocaine dependence under remission). Three of the CUD also reported current alcohol abuse (n = 1), moderate alcohol dependence (n = 1), or marijuana abuse (n = 1). Current use or dependence on other drugs was denied and corroborated by the prescan urine tests in all subjects (urine was positive for cocaine in 12 CUD; urine was negative for all drugs in all other subjects). Subjects were fully informed of all study procedures and risks and provided written consent in accordance with the local institutional review board.

Task.

The fMRI task (developed in E-prime, Psychology Software Tools, Inc.) uses 4 counterbalanced monetary amounts (50¢, 25¢, 1¢, or 0¢), obtained for correct performance on a blocked neuropsychological task for up to $75, received at the completion of this study (Fig. S3). For simplicity and clarity, in the current manuscript we report results of 2 of these monetary conditions: the highest reward available versus no monetary gain (50¢ and 0¢). Each of these monetary amounts was received twice, under a drug versus neutral cue. The drug cues were 40 regular drug words; non-English or slang drug words were not used (as they may have not been recognized by the control subjects) (24). The neutral cues were 40 household words matched to the drug words on length, frequency in the English language (43), and part of speech (noun, adjective, adverb, verb) (24). Similarly to other fMRI tasks of emotion, the 2 word types were presented in a blocked on-off or off-on order (i.e., drug-neutral or neutral-drug) (44), counterbalanced between subjects (Fig. S3A). Across all 3.4-min task conditions (Fig. S3B), subjects had to press 1 of 4 buttons (yellow, blue, red, green) on a commercially available response pad (Cedrus brand Lumina model LP-400), matching the color of the word they had just read; word color order was pseudorandomized across all sequences (Fig. S3C). The task was presented via MRI-compatible goggles.

Behavioral Measures.

Reaction time and accuracy data were collected across all trials using E-prime. Note that minimum accuracy was ≥50%, a threshold that was met by all 34 subjects included in this study. We also collected ratings of drug craving (“how much do you want cocaine right now?” from not at all to very much, 0 to 10) outside the scanner (just before the task on a computer monitor) and then immediately before, once during, and at the completion of all fMRI sequences (inside the scanner this question was presented through MRI compatible goggles). A similar procedure was followed for the motivation to gain money (“how much do you want money right now?”). Finally, immediately after completion of the MRI session (and outside the scanner), all subjects rated all words on valence (“how negative or positive” a word is, from extremely negative to extremely positive, −5 to +5). All ratings were obtained using custom programs written in C++.

MRI Data Acquisition.

MRI scanning was performed on a 4T whole-body Varian/Siemens MRI scanner. The BOLD responses were measured as a function of time using a T2*-weighted single-shot gradient-echo EPI sequence (TE/TR = 20/1600 ms, 4 mm slice thickness, 1 mm gap, typically 33 coronal slices, 20 cm FOV, 64 × 64 matrix size, 90°-flip angle, 200 kHz bandwidth with ramp sampling, 128 time points, and 4 dummy scans to be discarded to avoid nonequilibrium effects in the fMRI signal). Padding was used to minimize subject motion, which was also monitored immediately after each fMRI run (45). Earplugs (−28 dB sound attenuation; Aearo Ear TaperFit 2; Aearo Company) and headphones (−30 dB sound attenuation; Commander XG MRI Audio System, Resonance Technology Inc.) were used to minimize the interference effect of scanner noise during fMRI (46). Anatomical images were collected using a T1-weighted 3D-MDEFT sequence (47) (TE/TR = 7/15 ms, 0.94 × 0.94 × 1 mm spatial resolution, axial orientation, 256 readout and 192 × 96 phase-encoding steps, 16 min scan time) and a modified T2-weigthed hyperecho sequence (48) (TE/TR = 42/10,000 ms, echo train length = 16, 256 × 256 matrix size, 30 coronal slices, 0.86 × 0.86 mm in-plane resolution, 5 mm thickness, no gap, 2 min scan time), both reviewed to rule out gross brain morphological abnormalities.

MRI Data Processing.

Subsequent analyses were performed with the statistical parametric mapping package SPM2 (Wellcome Trust Centre for Neuroimaging). A 6-parameter rigid body transformation (3 rotations, 3 translations) was used for image realignment and to correct for head motion; 2 mm displacement and 2° rotation in any of the axes in any of the task sequences were used as criteria for acceptable motion. The realigned datasets were spatially normalized to the standard frame (Talairach) with a 12-parameters affine transformation (49), using a voxel size of 3 × 3 × 3 mm³. An 8-mm full-width-half-maximum Gaussian kernel was used to smooth the data.

BOLD-fMRI Analyses.

A general linear model (50) and a box-car design convolved with a canonical hemodynamic response function (HRF) and low-pass (HRF) and high-pass (cut-off frequency: 1/256 Hz) filters were used to calculate individual BOLD-fMRI maps. Four contrast maps per subject were calculated, reflecting % signal change from a fixation baseline for each of the 4 task conditions. These individual contrast maps were included in a second-order (random-effects) repeated measures ANOVA SPM2 model with 2 groups (CUD and control), 2 word conditions (drug and neutral), and 2 monetary conditions (50¢ and 0¢). Brain activation clusters with at least 20 voxels (540 mm³) and P < 0.001, cluster-level corrected for multiple comparisons using the continuous random field calculation implemented in SPM2, were considered significant. This higher than usual (P < 0.05 corrected) statistical threshold was selected as we were particularly interested in the most robust task activations/hypoactivations and the group main effect [reported exclusively in the current manuscript; for results of interaction effects (in 30 subjects, 26 from the current sample), which necessitated a lower statistical threshold (P < 0.005 uncorrected) and revealed the dopaminergic midbrain, see ref. 42]. In all SPM analyses, anatomical specificity was corroborated with a coplanar stereotaxic atlas of the human brain (51).

To confirm the voxel-based analyses, functional ROIs with an isotropic volume of 27 voxels (729 mm³) were defined at all of the significant coordinates (Table 1) to extract (with a custom program written in IDL, ITT Visual Information Solutions) the average (and variability) BOLD-fMRI signal amplitudes in these regions. These ROIs offer precise spatial localization of the functional responses as we previously reported (e.g., refs. 7 and 8). These ROI measures were used in follow-up analyses (e.g., ANOVA with Bonferroni corrections, t tests, correlations) conducted in SPSS 11.5 (SPSS Inc.). Statistical significance for these ROI analyses was defined at P < 0.05, uncorrected (note that here we only inspected the regions that were significant at P < 0.001 cluster-level corrected at the whole-brain analyses, providing protection against Type I error). In all analyses, the appropriate corrections were used in cases of violation of homogeneity of variance (e.g., as tested with Levene's test or Mauchly's test of sphericity).

We conducted correlations between the selected ROIs (averaged across the 4 task conditions) with 6 selected cocaine use variables (Table S1). These correlations in SPSS were inspected with a familywise corrected threshold (P < 0.01). These exploratory correlations were conducted to inspect associations between the brain regions that differed between the group and severity of drug use (including time since last use). For our a priori region, the ACC correlations were reported if they reached the more lenient statistical threshold (P < 0.05). All ROI correlations were validated with SPM whole-brain, voxel-based correlation analyses (simple regressions) as described in Results.

The behavioral measures were similarly analyzed. Note that all continuous and normally distributed variables were inspected with parametric tests (within groups: paired t test; between groups: independent t tests; correlations: Pearson r). Self-report measures (including depression and the nonnormally distributed cocaine use variables) were inspected with the respective nonparametric tests (Wilcoxon, Mann–Whitney U, or Spearman r = r_S).