Multi‐cohort cerebrospinal fluid proteomics identifies robust molecular signatures for asymptomatic and symptomatic Alzheimer's disease

Abstract

Background: Alzheimer’s disease (AD), the most common form of dementia, is characterized by the accumulation of Amyloid‐β (Aβ 42 ) and hyperphosphorylated Tau 181 (p‐tau 181 ) proteins in the brain. Changes in these proteins in AD brains and cerebrospinal fluid (CSF) are detected years before AD symptoms. Studying AD proteome in CSF can reflect its diverse underlying pathophysiology and pave the way for reliable diagnostic and therapeutic advancements. Method: We present one of the largest CSF AD proteomic profiles (7,029 protein analytes) in CSF of 3,065 individuals in three stages. Discovery was performed in 836 samples from the Knight ADRC and 618 samples from the FACE cohorts using the ATN framework (AT‐ = 680 and AT+ = 490). The identified proteins were tested in 832 individuals (AT‐ = 235 and AT+ = 358) from the ADNI and Barcelona‐1 cohorts. The proteins that passed multiple Bonferroni corrections on the meta‐analysis were utilized for AD prediction models and pathway enrichment analysis to gain mechanistic insights into AD pathophysiology. Result: In discovery, we identified 3,565 proteins to be significantly (FDR < 0.05) altered. Of these, 2,543 passed FDR in replication with a consistent direction. In the meta‐analysis, we identified 2,233 proteins to be significantly (P‐Bonferroni < 0.05) altered in the AD CSF proteome. Some of the highly significant proteins included YWHAG (P‐Bonf < 4.5×10 −219 ), SMOC1 (P‐Bonf < 5.8×10 −208 ), NRGN (P‐Bonf < 3.2×10 −119 ), and NEFL (P‐Bonf < 1.8×10 −37 ). By using lasso, we identified a set of 39 proteins with high predictive power (Discovery AUC = 1.0; Replication AUC = 0.99) representing a robust and precise AD diagnostic biomarker. Enrichment analysis highlighted several neurological disorders (e.g. AD, tauopathy, and synucleinopathy) and neuronal functions (neuron projection morphogenesis, synapse assembly and organization, and neuron differentiation) as significantly enriched in the altered AD CSF proteome. Conclusion: Our findings show the promising potential of AD CSF proteomics as a reliable and robust AD prediction model. We identified proteins and biological pathways that are compromised in AD, thereby, increasing our understanding of AD biology. Our findings may accelerate the development of effective intervention therapies that target the earliest molecular triggers of AD.