Early stage Parkinson's disease patients and normal volunteers: comparative mechanisms of sequence learning

Abstract

Early-stage nondemented Parkinson's disease (PD(es)) patients can learn short but not long sequences as well as controls. We have previously shown that to achieve normal performance, PD(es) patients activated the same right-sided cortical regions as controls plus the homologous left sided cortex and bilateral cerebellum. In this study, we evaluated two related hypotheses to explain the behavioral abnormalities and the increased bilateral brain activation observed in the PD(es) group. Hypothesis 1 proposed that PD(es) patients recruit regions from a normal bilateral network specialized for sequence learning that healthy controls would activate if performing difficult tasks. Thus, PD(es) patients can learn short sequences as well as controls. Hypothesis 2 proposed that information processing within the network in the PD(es) group is impaired. Thus, PD(es) patients cannot learn as difficult a sequence as controls. To test hypothesis 1, we increased task difficulty and statistical power in the control group and showed that the control and the PD(es) groups activated the same regions. To test hypothesis 2, we analyzed the equal performance data using two partial least squares (PLS) multivariate analyses. The task-PLS analysis showed that to perform equally with controls, the PD(es) group expressed the normal bilateral network more than the control group. The behavior-PLS analysis showed that the correlation between learning performance and regional activation was significantly different between the groups. We conclude that PD(es) patients have near normal learning if task difficulty is moderate because they can recruit additional regions from a normal bilateral network specialized for sequence learning. However, when a difficult task would normally require bilateral activation, PD(es) patients fail to learn because information processing within the network is impaired. Hum. Brain Mapp. 20:246-258, 2003.

Figure 1

Brain regions activated during sequence learning (TEseq‐TEseq_c) in Parkinson's Disease and Control groups are color coded and overlaid on a structural MRI that conforms to the space of the Talairach and Tourneaux [ 1988] atlas. yellow, PD but not Control; blue, PD and Control; red, Control but not PD. A: The two hypotheses tested in this report were generated from data that are printed from Mentis et al. [ 2003]. Panel A demonstrates brain regions activated when PD_(es) patients were in the process of achieving equal performance with controls. That is, when performance between the groups was equal by the end of the PET scan. Power to detect regional activation was equivalent between groups. B: Testing hypothesis 1. The normal bilateral sequence learning circuit was observed when the control group learned longer sequences and when the power to detect activation was increased (Supporting prediction a). The PD_(es) group activated a network similar to the normal bilateral sequence learning network (supporting prediction b). Numbers represent level of axial slice in mm relative to the anterior‐posterior commissure line. Left side of the brain is on left of image. Figure reproduced with permission of Lippincott, Williams & Wilkins [Mentis et al., 2003].

Figure 2

Testing hypotheses 1 and 2. A: The same brain network is associated with sequence learning in both groups (supporting hypothesis 1 prediction b). The network is overlaid on horizontal sections from a structural MRI that conforms to the space of Talairach and Tourneaux [ 1988]. Numbers represent level of axial slice in mm relative to the anterior–posterior commissure line. Left side of the brain is on left of image. B: Degree to which each group expresses the brain network during TEseq and TEseq_c. To achieve equal performance, the PD_(es) group had to express the normal bilateral sequence learning network 15% more than the control group. This suggests less efficient utilization of the network by the PD_(es) group (supporting hypothesis 2).

Figure 3

Testing hypothesis 2. a: In both groups, this brain network is significantly correlated with the number of “correct” movements. The network is overlaid on horizontal sections from a structural MRI that conforms to the space of Talairach and Tourneaux [ 1988]. Numbers represent level of axial slice in mm relative to the anterior‐posterior commissure line. Left side of the brain is on left of image. b: The brain network‐behavior correlation is significant for each group, but the direction of the correlation is significantly different across groups. This suggests information processing within the bilateral sequence learning network is different in the PD_(es) than in the Control group (supporting hypothesis 2).