Alterations of the Human CSF Proteome in Alzheimer's Disease
Richard Margolin1*, Christopher Whelan1, Michael Nagle1, Shorena Janelidze2, Erik Stomrud2, Tarek Samad1, Niklas Mattsson2, Anders Malarstig1, Oskar Hansson2
1Pfizer, Inc; 2Lund University
Introduction: Until lately, the search for treatments for Alzheimer's disease (AD) has largely targeted amyloid β (Aβ) and tau pathology. However, evidence from human genetics, transcriptomics, and neuroimaging has implicated multiple additional pathophysiological processes in AD, including impaired immune signaling, cerebrovascular dysfunction, inefficient phospholipid metabolism, oxidative DNA damage, and changes in autophagy, cell adhesion, endocytosis and mitochondrial function. Large-scale, multiplex proteomics assays have not been widely used in AD drug discovery because of technical challenges and cost but may powerfully aid in identifying novel therapeutic targets or valuable biomarkers associated with such processes.
Methods: We measured a diverse panel of 270 proteins in human CSF obtained from elderly participants in the Swedish BioFINDER study (www.biofinder.se). Subjects were grouped by clinical diagnosis (cognitively normal [CN], mild cognitive impairment [MCI] and AD dementia [AD]) and CSF Aβ levels (amyloid-negative: Aβ-; amyloid-positive: Aβ+). We compared protein levels from 420 Aβ- CN to levels from 285 Aβ+ CN, 53 Aβ- MCI, 78 Aβ+ MCI, and 176 AD using the well-validated, highly sensitive OLINK proximity extension assay (www.olink.com). Analysis involved multiple linear regressions in R, corrected for multiple comparisons using false discovery rate (FDR).
Results: Compared to Aβ- CN, the AD group showed significant decreases of 17 proteins including N2DL2, CCL19, SHPS1, KYNU, FGF19, IGFB7, CD200, VEGFA, TRAIL, tPA, CDH6, IL17A, RGMB, LDLR, UNC5C, TR-AP, and ALCAM, and significantly increases of 16 proteins, including CHIT1, MMP10, HAGH, ADAM22, SMOC2, ITGB2, EZR, OPG, 4EBP1, IL1A, YKL-40, CCL3, CASP8, STAMPB, TNFRSF12A, and TFPI. 7 proteins were increased in Aβ+ MCI, including CHIT1, ITGB2, SMOC2, MMP10, TNF-R2, OPN, and GFR-α1. 5 proteins were increased in Aβ- MCI, including RGMB, DRAXIN, ROBO2, NTRK3, and GCP5, and 2 were decreased, including PAI and OPG. SMOC2, ITGB2, and YKL-40 were all increased in Aβ+ CN (p<0.05, FDR-corrected).
Discussion: This study was among the largest human CSF proteomic studies of AD performed to date. We identified highly significant changes across many proteins involved in inflammation, angiogenesis, and other functions in the CSF of patients with MCI and AD. Future work will explore relationships between these findings and various neuropsychological and neuroimaging measurements, genetic risk factors for AD, and plasma protein levels. The use of human CSF and possibly plasma proteomics may aid drug and biomarker discovery for AD.