Data-Driven Biological Subtypes of Parkinson's Disease.

The heterogenous clinical and pathophysiological presentation of idiopathic Parkinson's Disease (iPD) suggests the existence of underlying distinct biological subtypes. A better definition of PD subtypes is crucial for the development of target disease modifier approaches. Using data from the Parkinson's Precision Medicine Initiative (PPMI) cohort, we aimed to identify distinct biological subtypes of iPD through data-driven cluster analysis leveraging currently available biomarker data. We included n=225 de novo iPD subjects. Individuals with known genetic forms of PD were excluded. A total of 22 biomarkers reflecting different biological pathways (neurodegeneration, proteinopathy, neuroinflammation, mitochondrial impairment, and lipid metabolism/autophagy-lysosomal dysfunction) and measured in cerebrospinal fluid (CSF), whole blood, serum, plasma, or urine were selected. Hierarchical clustering was performed using ward.D2 method and Gower dissimilarity. Clinical-demographic characteristics of the identified clusters were compared at baseline and 5-year follow-up. We identified three clusters: Cluster I ("Mitochondrial-Predominant" subtype); Cluster II ("Cryptic" subtype); Cluster III ("Mixed Pathology" subtype). The "Mitochondrial-Predominant" and "Mixed Pathology" subtypes showed higher levels of CSF mitochondrial DNA deletions and ND1 compared to the "Cryptic" subtype. The "Mixed Pathology" subtype was also characterized by higher levels of serum NfL, CSF p-tau, CSF GFAP, CSF YKL-40, and lower levels of plasma total ceramide, glucosylceramide, and sphingomyelin compared to the other subtypes. In the "Cryptic" subtype no clear dominant biological profile was detected. Clinically, subjects belonging to the "Mixed Pathology" subtype were older and predominantly male. No differences in 5-year clinical progression were found between clusters. While most of the previous research has performed a biological characterization of the clinical subtypes of PD, with our work we propose a biology-driven clustering in a large cohort if subjects with iPD. We identified three biological subtypes. The "Mitochondrial-Predominant" subtype was characterized by alterations exclusively of biomarkers reflecting mitochondrial dysfunction. The "Mixed Pathology" subtype, in contrast, was characterized by biomarker alterations across all explored pathological pathways. Finally, the "Cryptic" subtype included subjects whose biomarker signature did not indicate predominant alterations in any of the explored pathways. The clusters did not differ in terms of baseline clinical-demographic characteristics, except for age and sex. No difference in mid-term clinical progression was observed between subtypes.