Subtype classification of Parkinson's disease based on cortical functional gradient heterogeneity and its neurotransmitter basis

Technical Summary: Subtype Classification of Parkinson's Disease Based on Cortical Functional Gradient Heterogeneity and Its Neurotransmitter Basis

Problem Statement
Parkinson's disease (PD) is characterized by significant clinical heterogeneity, with patients exhibiting diverse patterns, severities, and progression rates of both motor and non-motor symptoms. This variability poses a critical challenge for precision diagnosis and treatment. While previous studies have utilized motor symptom staging (e.g., Hoehn-Yahr) for classification, these methods often fail to predict non-motor outcomes or quality of life accurately. Furthermore, the neurobiological mechanisms linking macroscopic functional network disruptions to microscopic neurotransmitter deficits in PD subtypes remain unclear. The study addresses the need for a scientifically grounded classification system based on neural activity patterns and seeks to elucidate the underlying neuromodulatory basis of these functional alterations.

Methodology
The study employed a cross-sectional design involving 99 PD patients and 104 age-matched normal controls (NC). All participants underwent T1-weighted structural MRI and resting-state blood oxygen level-dependent (BOLD) functional MRI scanning while in the "off" medication state.

1. Data Preprocessing: Resting-state BOLD data were preprocessed using fMRIPrep, including slice timing correction, motion correction, distortion correction, and registration to the MNI template. Spurious variance was removed via linear regression, and data were band-pass filtered (0.01–0.08 Hz) and smoothed.
2. Gradient Analysis: Functional connectivity matrices were constructed using the Schaefer 400-parcellation atlas. Cortical functional gradients were derived using diffusion embedding (via the BrainSpace toolbox). The study focused on the first two principal gradients, which represent the primary axes of functional hierarchy from unimodal to transmodal regions.
3. Subtype Classification: Individual gradient heterogeneity was calculated as the deviation of each PD patient's gradient from the mean NC gradient. Hierarchical clustering analysis (Ward's method) was applied to these heterogeneity vectors to identify optimal PD subgroups.
4. Neurotransmitter Basis: Post-mortem gene expression data from the Allen Human Brain Atlas (AHBA) were utilized. Expression profiles for 31 pre-selected genes encoding dopaminergic, cholinergic, noradrenergic, and serotonergic receptors were mapped to the 400 cortical regions.
5. Statistical Analysis: Group differences (PD vs. NC, and between subtypes) were analyzed using General Linear Models (GLM) with age and gender as covariates. Correlations between cortical gradient changes and gene expression were assessed using Spearman correlation, with False Discovery Rate (FDR) correction applied.

Key Results

• Global and Network-Level Gradient Alterations: PD patients exhibited abnormal global gradient patterns compared to NCs. Specifically, Gradient 1 showed a coexistence of increases and decreases across different cortices.
  • Decreases: Observed primarily in the middle frontal gyrus, supramarginal gyrus, cingulate gyrus, and precuneus (associated with dorsal attention, ventral attention, and default mode systems).
  • Increases: Observed in the precentral gyrus, paracentral lobule, cuneus, calcarine sulcus, and lingual gyrus (associated with visual and sensorimotor systems).
  • Network Specificity: The most pronounced reductions in Gradient 1 were found in the dorsal attention and ventral attention networks. No significant differences were found in Gradient 2.
• Subtype Classification: Hierarchical clustering successfully divided the PD cohort into two distinct subgroups:
  • Subtype I (Non-NPPD): 39 patients.
  • Subtype II (Neuropsychiatric PD or NPPD): 60 patients.
  • Clinical Correlation: While motor scores (UPDRS III, Hoehn-Yahr) did not differ significantly between subtypes, Subtype II (NPPD) demonstrated significantly lower Mini-Mental State Examination (MMSE) scores and higher Hamilton Anxiety (HAMA) and Hamilton Depression (HAMD) scores compared to Subtype I.
• Neurotransmitter Associations:
  • PD vs. NC: Gradient 1 alterations in PD were significantly correlated with the expression of genes for dopaminergic (DRD2), cholinergic (CHRNB2, CHRNA2-4, CHRM4), noradrenergic (ADRA1B), and serotonergic (HTR1A, HTR2A, HTR7, HTR1F) receptors.
  • NPPD vs. Non-NPPD: The specific gradient changes distinguishing the NPPD subtype were significantly correlated with the expression of the cholinergic receptor gene CHRM5.

Key Contributions

1. Framework for Classification: The study provides a novel framework for classifying PD patients into neuropsychiatric (NPPD) and non-neuropsychiatric subtypes based on individual cortical functional gradient heterogeneity, rather than relying solely on clinical symptom staging.
2. Mapping Functional Hierarchy Disruption: It identifies specific disruptions in the cortical functional hierarchy of PD, particularly the weakening of segregation between unimodal and transmodal regions and specific deficits in attention networks.
3. Linking Macro- and Micro-Scales: The research establishes a direct link between macroscopic functional gradient abnormalities and microscopic neurotransmitter receptor gene expression, suggesting that the observed functional heterogeneity is driven by dysregulation in dopaminergic, cholinergic, noradrenergic, and serotonergic systems.

Significance and Claims
The authors claim that these findings offer a significant theoretical foundation for understanding the intricate interplay between macroscopic functional gradient changes and microscopic neuromodulation roles in PD. By demonstrating that CHRM5 expression is specifically associated with the NPPD subtype, the study suggests a potential molecular target for addressing neuropsychiatric symptoms in PD. The work supports the view of PD as a complex "disease spectrum" and advocates for precision medicine approaches that account for individual neural heterogeneity. The authors modestly note that while the "off" medication state mitigates acute drug effects, long-term medication impacts and the need for longitudinal studies to track temporal patterns remain limitations.