Principal component analysis of synaptic density measured with [11C]UCB-J PET in Alzheimer's disease

Abstract

<h3>Introduction</h3> Synaptic loss has been cited as an early pathological event and major structural correlate of cognitive impairment in Alzheimer's disease (AD). However, this observation is based on limited neuropathological studies. Here, we examined the <i>in vivo</i> relationship between synaptic density and cognitive performance in early AD using [¹¹C]UCB-J PET and an extensive neuropsychological battery. We used principal components analysis (PCA) to identify regional patterns of variance in synaptic density and then assessed the association between principle components (PC) scores with cognitive performance in AD. <h3>Methods</h3> [¹¹C]UCB-J binding was measured in 45 amyloid+ participants with AD (17 amnestic MCI and 28 mild dementia) and 19 amyloid– cognitively normal participants aged 50-85. Synaptic density was calculated as the distribution volume ratio (<i>DVR</i>) in a composite of AD-affected regions, using the whole cerebellum as reference region. A validated neuropsychological battery assessed performance across five cognitive domains, and an estimate of global cognition was made by averaging z-scores from each of these domains. PCA was applied to the pooled population (n=64) using <i>DVR</i> values centered and standardized by region from each of 42 bilateral cortical and subcortical regions of interest (ROI). Pearson's correlations were used to determine relationships between principle component subject scores and measures of cognitive functioning. <h3>Results</h3> After application of PCA, parallel analysis determined three significant PCs explaining 75.2% of the total variance (PC1=58.1%, PC2=11.4%, PC3=5.7%). PC1 was characterized by only positive loadings of equally low contributions across the majority of ROIs (Figure 1A, Figure 2A&B). PC2 was characterized by positive and negative loadings with the strongest contributions generally corresponding to ROIs from subcortical vs. cortical regions, respectively (Figure 1B, Figure 2C), while PC3 was characterized by positive and negative loadings with the strongest contributions generally corresponding to ROIs from caudal vs. rostral cortical regions, respectively (Figure 1C, Figure 2C). Within the AD group (n=45), component scores for PC1 were positively correlated with performance across all cognitive domains (Pearson <i>r</i>=0.33-0.44, <i>P</i>=0.03-0.003; Table 1) with the exception of verbal memory (Pearson <i>r</i>=0.24, <i>P</i>=0.11), as well as with a measure of global synaptic density (Pearson <i>r</i>=0.97, <i>P</i><0.0001) but not global Aβ deposition (Pearson <i>r</i>=-0.20, <i>P</i>=0.20). Scores for PC2 were inversely correlated with executive functioning (Pearson <i>r</i>=-0.32, <i>P</i>=0.03) and age (Pearson <i>r</i>=-0.45, <i>P</i>=0.002), while PC3 component scores demonstrated no significant correlations with either cognitive or demographic measures within the AD group. Effect sizes to determine group differences in [¹¹C]UCB-J PET were strongest in PC1 (Cohen's d=1.15), which was weaker than the effect size using a hippocampal region alone (Cohen's d=1.51). <h3>Conclusions</h3> We present the first <i>in vivo</i> evidence confirming previous neuropathologic studies, demonstrating a significant association between synaptic density and cognitive performance. Our findings reinforce synaptic density as a robust biomarker of disease severity and suggest this correlation extends to the early stages of AD. Without the use of diagnostic identifiers, a data-driven approach defined specific spatial patterns of synaptic density, which were in turn correlated with unique participant characteristics within the AD group. <h3>Funding</h3> P50-AG047270 (CHvD), P30-AG066508 [CHvD], K23-AG057794 (APM), R01-AG052560 (REC, CHV), R01-AG062276 (CHvD), Dana Foundation (MKC), T32-MH019961 (RSO), and Thomas P. Detre Fellowship (RSO).