CSF Proteomic Alzheimer's Disease-Predictive Subtypes in Cognitively Intact Amyloid Negative Individuals

Abstract

We recently discovered three distinct pathophysiological subtypes in Alzheimer's disease (AD) using cerebrospinal fluid (CSF) proteomics: one with neuronal hyperplasticity, a second with innate immune system activation, and a third subtype with blood-brain barrier dysfunction. It remains unclear whether AD proteomic subtype profiles are a consequence of amyloid aggregation, or might exist upstream from aggregated amyloid. We studied this question in 127 older individuals with intact cognition and normal AD biomarkers in two independent cohorts (EMIF-AD MBD and ADNI). We clustered 705 proteins measured in CSF that were previously related to AD. We identified in these cognitively intact individuals without AD pathology three subtypes: two subtypes were seen in both cohorts (n = 49 with neuronal hyperplasticity and n = 44 with blood-brain barrier dysfunction), and one only in ADNI (n = 12 with innate immune activation). The proteins specific for these subtypes strongly overlapped with AD subtype protein profiles (overlap coefficients 92%-71%). Longitudinal p₁₈₁-tau and amyloid β 1-42 (Aβ42) CSF analysis showed that in the hyperplasticity subtype p₁₈₁-tau increased (β = 2.6 pg/mL per year, p = 0.01) and Aβ42 decreased over time (β = -4.4 pg/mL per year, p = 0.03), in the innate immune activation subtype p₁₈₁-tau increased (β = 3.1 pg/mL per year, p = 0.01) while in the blood-brain barrier dysfunction subtype Aβ42 decreased (β = -3.7 pg/mL per year, p = 0.009). These findings suggest that AD proteomic subtypes might already manifest in cognitively normal individuals and may predispose for AD before amyloid has reached abnormal levels.

Keywords: Alzheimer’s disease; amyloid beta; cerebrospinal fluid proteomics; cognitive functioning; risk factors; tau.

Figure 1

(a) Subject scores which reflect how well they match each of the three subtypes, individuals were assigned to the subtype on which they showed the highest loading; (b) protein levels averaged across individuals for each subtype (see Supplementary Table S5a,b for statistics of protein level comparisons between subtypes); (c) proportion of subtype-specific proteins that were labelled to be specific for a particular cell type (left: EMIF-AD MBD; right: ADNI); (d) selection of pathways enriched for subtype-specific proteins (see Supplemental Table S6 for complete list of enriched pathways). For (c,d): bars going up represent pathways associated with increased proteins, bars going down represent pathways associated with decreased proteins, and in (d) absolute numbers represent log(pFDR) * −1, * is times.

Figure 2

Subtype comparisons on biological characteristics. (a) Comparison of subtypes for the proportion of females and APOE e4 carriers, as well as age, t-tau, and p-tau levels; (b) comparisons of subtypes on other CSF markers that were not included in clustering (see Supplemental Table S8 for statistics).