Synapse protein signatures in cerebrospinal fluid and plasma predict cognitive maintenance versus decline in Alzheimers disease

Key results

1. 01

   ShownAfter adjusting for PTau181:Aβ42, synapse proteins were the strongest CSF correlates of cognitive impairment (CI). LASSO with recursive feature elimination reduced 214 synapse proteins to a normalized YWHAG minus NPTX2 ratio, which replicated in cohorts not used for discovery.

   Discovery in Knight-ADRC and ADNI (7,289 SomaScan measurements): 675 up and 721 down proteins at BH q<0.05, with SynGO synapse enrichment; 2 proteins retained 83% of full model performance; the ratio correlated with CI at r=0.54 (ADNI), 0.55 (Knight-ADRC), 0.62 (Stanford) and 0.66 (DIAN). · CSF YWHAG:NPTX2 ratio explains a substantial…; Linear regression; Derivation of CSF YWHAG:NPTX2 ratio; Figure legends

2. 02

   ShownThe ratio explains CI variance beyond the standard amyloid, tau and neurodegeneration markers. This includes tau PET (tested with the closely correlated YWHAZ:NPTX2 in BioFINDER2) and CSF neurofilament light (NfL).

   A+T1+ individuals: PTau181:Aβ42 explains 10% of CI variance and the ratio adds 27%; BioFINDER2 Aβ+: Aβ42:Aβ40 plus tau PET explain 35% and YWHAZ:NPTX2 adds 10% (YWHAZ vs YWHAG r=0.94); Stanford A+T1+ (n=31): ratio adds 50% beyond NfL (ratio vs NfL r=0.65); 62% of low-ratio vs 4% of high-ratio A+T1+ cases were cognitively normal. · CSF YWHAG:NPTX2 ratio explains a substantial…; Figure legends

3. 03

   ShownBaseline CSF YWHAG:NPTX2 predicts later cognitive decline and dementia onset beyond amyloid/tau status in three longitudinal cohorts. Stratifying by the ratio separates risk far better than AT1 status alone.

   ADNI, Knight-ADRC and Stanford (n=1,365): per-SD meta hazard ratio (HR) 2.34 for stage progression in A+T1+ MCI-mild dementia; 1.92 for normal-to-dementia conversion in A+T1+ cognitively normal people (Stanford excluded for few events); top vs bottom 25th percentile HR 15.36, rising to 18.87 when combined with A+T1+ status, vs 3.96 for A+T1+ vs A–T1– alone (no covariates). · CSF YWHAG:NPTX2 ratio predicts future tau…; Abstract; Cox proportional hazards regression

4. 04

   ShownA LASSO plasma signature built on 745 preselected proteins was developed after plasma YWHAG:NPTX2 itself showed no significant correlation with CI. The signature partly mirrors the CSF ratio, explains CI independently of autopsy pathology, and predicts both decline and dementia onset.

   Stanford test r=0.51 with CI (training: Knight-ADRC 0.66, ROSMAP 0.62); r=0.58 and 0.53 with the CSF ratio in 503 paired samples; in 126 ROSMAP autopsy-confirmed AD cases it explains 30% of CI variance, matching plaque plus tangle load (30%); per-SD meta HR 1.49 (progression) and 2.03 (conversion); extreme-quartile HR 4.51. · Plasma proteomic signature of cognitive…; Derivation of plasma signature of cognitive…; CODE AVAILABILITY

5. 05

   ShownIn cognitively normal people the ratio rises with age from early adulthood, long before PTau181:Aβ42 changes. It rises faster in autosomal dominant AD (ADAD) mutation carriers and in APOE4 carriers, and the slope scales with mutation severity.

   DIAN cognitively normal participants under 55: carriers higher (p=7.21×10^−13) with a steeper age slope (interaction p=0.032); slope vs estimated onset age across onset bins Spearman r=–0.9, p=0.037; APOE4 dose by age interaction p=0.016, with APOE4 carriers rising 33% faster than APOE3/3; rise of 1.9 SD over 60 years of aging vs 2.7 SD from aged to dementia. · CSF YWHAG:NPTX2 ratio increases with normal…