Neuroimmune proteins can differentiate between tauopathies

Abstract

Background: Tauopathies are a group of neurodegenerative diseases where there is pathologic accumulation of hyperphosphorylated tau protein (ptau). The most common tauopathy is Alzheimer's disease (AD), but chronic traumatic encephalopathy (CTE), progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and argyrophilic grain disease (AGD) are significant health risks as well. Currently, it is unclear what specific molecular factors might drive each distinct disease and represent therapeutic targets. Additionally, there is a lack of biomarkers that can differentiate each disease in life. Recent work has suggested that neuroinflammatory changes might be specific among distinct diseases and offers a novel resource for mechanistic targets and biomarker candidates.

Methods: To better examine each tauopathy, a 71 immune-related protein multiplex ELISA panel was utilized to analyze anterior cingulate grey matter from 127 individuals neuropathologically diagnosed with AD, CTE, PSP, CBD, and AGD. A partial least square regression analysis was carried out to perform unbiased clustering and identify proteins that are distinctly correlated with each tauopathy correcting for age and gender. Receiver operator characteristic and binary logistic regression analyses were then used to examine the ability of each candidate protein to distinguish diseases. Validation in postmortem cerebrospinal fluid (CSF) from 15 AD and 14 CTE cases was performed to determine if candidate proteins could act as possible novel biomarkers.

Results: Five clusters of immune proteins were identified and compared to each tauopathy to determine if clusters were specific to distinct disease. Each cluster was found to correlate with either CTE, AD, PSP, CBD, or AGD. When examining which proteins were the strongest driver of each cluster, it was observed the most distinctive protein for CTE was CCL21, AD was FLT3L, and PSP was IL13. Individual proteins that were specific to CBD and AGD were not observed. CCL21 was observed to be elevated in CTE CSF compared to AD cases (p = 0.02), further validating the use as possible biomarkers. Sub-analyses for male only cases confirmed the results were not skewed by gender differences.

Conclusions: Overall, these results highlight that different neuroinflammatory responses might underlie unique mechanisms in related neurodegenerative pathologies. Additionally, the use of distinct neuroinflammatory signatures could help differentiate between tauopathies and act as novel biomarker candidate to increase specificity for in-life diagnoses.

Keywords: Biomarkers; CSF; Immune; Neurodegeneration; Neuroinflammation; Tau; Tauopathies.

Fig. 1

Partial Least Square regression identifies clusters of proteins that correlate with each tauopathy. Using the partial least square regression analysis, five clusters of ELISA proteins were identified and compared against each tauopathy to determine degree of correlation. Each comparison was a contrast of a single tauopathy against all other neuropathologies. Blue represents low correlation and red represents high correlation. Comparisons are adjusted for age at death and gender. Cluster 1 was most correlated with CTE, Cluster 2 with AD, Cluster 3 with PSP, Cluster 4 with CBD, and Cluster 5 with AGD

Fig. 2

Correlation of each cluster to ELISA proteins identifies which immune factors are most specific to each disease. Singular value decomposition (SVD) was conducted on a partial correlation matrix between each cluster and ELISA proteins after adjusting for age and gender, where each column corresponds to a cluster and each row corresponds to an ELISA protein. The resulting heatmap demonstrates which immune proteins best correlate to each cluster/disease. From this heatmap, the top 5 candidate proteins that are distinct for each cluster/disease were selected. Blue represents low correlation and red represents high correlation

Fig. 3

Receiver operator characteristics (ROC) curve demonstrates the top 5 candidate proteins for each cluster are specific and sensitive to identify tauopathies. Using the PLS results and top 5 candidate proteins for each cluster, ROC curve was performed for A) CTE, B) AD, C) PSP, D) CBD, and E) AGD. The area under the curve (AUC), standard error, and significance is displayed below each graph. AUC values over 0.5 suggest positive association with each disease, while values under 0.5 represent negative association. The black line is the reference line. Multiple comparison correction set the significance threshold to p < 0.01

Fig. 4

CCL21 is elevated in the CSF in CTE. To validate PLS results and determine if candidate proteins can be used as biomarkers as fluid biomarkers, the top distinct protein for CTE, CCL21, was measured in postmortem CSF from individuals with AD and CTE. CCL21 concentrations were higher in CTE compared to CTE as measured with a Mann–Whitney test (*p < 0.05). Each dot represents 1 case. Error bars represent mean ± SEM