Delineating three distinct spatiotemporal patterns of brain atrophy in Parkinson's disease
Yusuke Sakato,Atsushi Shima,Yuta Terada,Kiyoaki Takeda,Haruhi Sakamaki-Tsukita,Akira Nishida,Kenji Yoshimura,Ikko Wada,Koji Furukawa,Daisuke Kambe,Hiroki Togo,Yohei Mukai,Masanori Sawamura,Etsuro Nakanishi,Hodaka Yamakado,Yasutaka Fushimi,Tomohisa Okada,Yuji Takahashi,Yuji Nakamoto,Ryosuke Takahashi,Takashi Hanakawa,Nobukatsu Sawamoto
DOI: https://doi.org/10.1093/brain/awae303
IF: 14.5
2024-10-25
Brain
Abstract:The clinical manifestation of Parkinson's disease exhibits significant heterogeneity in the prevalence of non-motor symptoms and the rate of progression of motor symptoms, suggesting that Parkinson's disease can be classified into distinct subtypes. In this study, we aimed to explore this heterogeneity by identifying a set of subtypes with distinct patterns of spatiotemporal trajectories of neurodegeneration. We applied Subtype and Stage Inference (SuStaIn), an unsupervised machine learning algorithm that combined disease progression modelling with clustering methods, to cortical and subcortical neurodegeneration visible on 3 T structural MRI of a large cross-sectional sample of 504 patients and 279 healthy controls. Serial longitudinal data were available for a subset of 178 patients at the 2-year follow-up and for 140 patients at the 4-year follow-up. In a subset of 210 patients, concomitant Alzheimer's disease pathology was assessed by evaluating amyloid-β concentrations in the CSF or via the amyloid-specific radiotracer 18 F-flutemetamol with PET. The SuStaIn analysis revealed three distinct subtypes, each characterized by unique patterns of spatiotemporal evolution of brain atrophy: neocortical, limbic and brainstem. In the neocortical subtype, a reduction in brain volume occurred in the frontal and parietal cortices in the earliest disease stage and progressed across the entire neocortex during the early stage, although with relative sparing of the striatum, pallidum, accumbens area and brainstem. The limbic subtype represented comparative regional vulnerability, which was characterized by early volume loss in the amygdala, accumbens area, striatum and temporal cortex, subsequently spreading to the parietal and frontal cortices across disease stage. The brainstem subtype showed gradual rostral progression from the brainstem extending to the amygdala and hippocampus, followed by the temporal and other cortices. Longitudinal MRI data confirmed that 77.8% of participants at the 2-year follow-up and 84.0% at the 4-year follow-up were assigned to subtypes consistent with estimates from the cross-sectional data. This three-subtype model aligned with empirically proposed subtypes based on age at onset, because the neocortical subtype demonstrated characteristics similar to those found in the old-onset phenotype, including older onset and cognitive decline symptoms ( P < 0.05). Moreover, the subtypes correspond to the three categories of the neuropathological consensus criteria for symptomatic patients with Lewy pathology, proposing neocortex-, limbic- and brainstem-predominant patterns as different subgroups of α-synuclein distributions. Among the subtypes, the prevalence of biomarker evidence of amyloid-β pathology was comparable. Upon validation, the subtype model might be applied to individual cases, potentially serving as a biomarker to track disease progression and predict temporal evolution.
neurosciences,clinical neurology