Cortical thickness subtypes in cognitively unimpaired individuals: Differential network and transcriptomic vulnerability to cortical thinning.

INTRODUCTION

The emergence, stability, and contributing factors of Alzheimer's disease (AD) gray matter subtypes remain unclear.

METHODS

We analyzed data from 1323 individuals without a diagnosis of dementia (CDR < 1) with T1w-MRI and amyloid-PET, including 622 with longitudinal data (3.66 ± 1.78 years). Cortical thickness subtypes were identified using a non-negative matrix factorization (NMF) clustering algorithm. We examined clinical and demographic differences, subtype stability, and longitudinal thinning patterns using brain network models and imaging-transcriptomic analysis. Replication was performed with an alternative clustering approach and a validation cohort.

RESULTS

Two stable subtypes emerged: limbic-predominant and hippocampal-sparing. Limbic-predominant participants were older, had higher amyloid burden, and faster memory decline, while hippocampal-sparing individuals showed greater attention and executive function decline. Distinct thinning patterns were linked to specific network properties and gene expression profiles.

DISCUSSION: These MRI-based subtypes reflect underlying pathophysiological mechanisms and may aid in prognostication and clinical trial stratification.

HIGHLIGHTS: Two gray matter thickness subtypes can already be identified in preclinical stages, exhibiting distinct clinical characteristics and progression patterns. Individual subtype assignment remains stable over time. Longitudinal cortical thinning patterns follow distinct network- and transcriptomic-based mechanisms within each subtype.