Low-frequency signal drift was removed using a high-pass filter (cutoff = 128 sec) and an autoregressive modeling (AR [1]) of temporal autocorrelations was applied. At the first level, subject-specific Cilengitide nmr contrast images were generated for each working-memory load condition versus baseline. Each working-memory load versus baseline contrast was then entered into second-level GLM ANOVAs to obtain SPM-F maps that investigated: (1) the main effect Inhibitors,research,lifescience,medical of task; (2) the main effect of group (PD-Off, Controls); (3) the group by task interaction (PD-Off, Controls × high-, medium-, and low-load working memory); (4) the main effect of treatment (PD-Off, PD-On); and

(5) the treatment by task interaction (PD-Off, PD-On × high-, medium-, and low-load working memory). Furthermore, to account for possible effects of behavioral variability on brain activations, analyses (2), (3), (4), and (5) were repeated including RT and accuracy as variables of no interest. Of note,

Inhibitors,research,lifescience,medical we also tested for linear and quadratic interactive effects between medication and DAT-BPND values Inhibitors,research,lifescience,medical on striatal BOLD responses in the PD group. The SPM model included two separate regressors for each patient: (1) the DAT-BPND values (testing for linear effects); (2) the square of these values (testing for quadratic functions). This way, it is possible to investigate linear fits (excluding quadratic ones) and vice versa (i.e., testing for quadratic effects factoring out linear ones). This method has been used before Inhibitors,research,lifescience,medical in fMRI studies exploring linear and nonlinear relations between drug effects on clinical variables in PD patients (Rowe et al. 2008). The same method was also used to study linear and quadratic effects of disease duration. Analyses exploring activations within the whole brain were thresholded at P < 0.05, family-wise error (FWE), whole-brain correction. In addition, given our strong a priori hypotheses on specific Inhibitors,research,lifescience,medical PFC and striatal regions, we employed a ROI approach. ROIs included

the anterior cingulate cortex (ACC); the superior, MTMR9 middle, and inferior frontal gyrus (SFG, MFG, IFG); the caudate; and putamen and were created using the “aal.02” atlas (http://marsbar.sourceforge.net/) (Tzourio-Mazoyer et al. 2002). The statistical threshold for ROI analyses was set at P < 0.05, FWE, small volume correction (svc) (Worsley et al. 1996; Friston 1997). Because 12 ROIs (six on the left, six on the right) with different size were defined, we treated them as separate hypotheses and further adjusted the significance for multiple comparison testing using Dunn–Sidak correction (Howell et al. 2007). Finally, for explorative purposes only, we also report brain regions not predicted a priori but that met a threshold of P < 0.001, uncorrected, >10 contiguous voxels.