BACKGROUND: The heterogeneity of PD suggests the existence of different subtypes. While some motor clusters have been consistently identified, little is known about non-motor PD subtypes and motor-non-motor interplay. Research in this regard has produced somewhat contradictory results, which might be biased by the inclusion of treated patients. PATIENTS AND METHODS: We performed a non-hierarchical cluster analysis using both motor and non-motor data on 398 newly diagnosed untreated PD patients enrolled in the Parkinson's Progressive Marker Initiative (PPMI) study. We further evaluated whether dopaminergic dysfunction, as measured by (123)[I]-FP-CIT SPECT scan, could explain, at least partially, the observed difference between the clusters. RESULTS: Three clusters were identified. Group 1 was characterized by the lowest motor and non-motor burden, whereas group 2 and 3 had similar motor disability, but different non-motor involvement, especially with regards to apathy and hallucinations. (123)[I]-FP-CIT binding values paralleled motor disability burden among the 3 clusters, but further multivariate analyses also revealed a negative correlation with depression. DISCUSSION: Our results confirm the motor as well as non-motor heterogeneity of PD, suggesting the existence of 3 different subtypes. Dopaminergic dysfunction only marginally explains the non-motor variability of PD. Identification of such clusters can have important implications for generating novel pathophysiological hypotheses and therapeutic strategies.