Brain proteomic analysis implicates actin filament processes and injury response in resilience to Alzheimer's disease

Abstract

Resilience to Alzheimer's disease is an uncommon combination of high disease burden without dementia that offers valuable insights into limiting clinical impact. Here we assessed 43 research participants meeting stringent criteria, 11 healthy controls, 12 resilience to Alzheimer's disease and 20 Alzheimer's disease with dementia and analyzed matched isocortical regions, hippocampus, and caudate nucleus by mass spectrometry-based proteomics. Of 7115 differentially expressed soluble proteins, lower isocortical and hippocampal soluble Aβ levels is a significant feature of resilience when compared to healthy control and Alzheimer's disease dementia groups. Protein co-expression analysis reveals 181 densely-interacting proteins significantly associated with resilience that were enriched for actin filament-based processes, cellular detoxification, and wound healing in isocortex and hippocampus, further supported by four validation cohorts. Our results suggest that lowering soluble Aβ concentration may suppress severe cognitive impairment along the Alzheimer's disease continuum. The molecular basis of resilience likely holds important therapeutic insights.

Fig. 1. Workflow of this study.

a Samples (N = 155) from up to four matched brain regions were donated by 43 research participants who were assigned to three clinico-pathologic groups: HC (healthy control), RAD (cognitive resilience to Alzheimer’s disease), or AD dementia (ADD). Samples were quantified by data independent tandem mass spectrometry and data analyzed by differential expression and co-expression network analyses. Results were compared to four independent data sets that most closely approximated our study design. b Illustration of differential expression analysis and summary of the final number of RAD-associated differentially expressed proteins (RAD DEPs). DLPFC dorsolateral prefrontal cortex, PC precuneus. Source data are provided as a Source Data file.

Fig. 2. The process of deriving differentially expressed proteins (DEPs).

a Venn diagram shows overlap in proteins quantified across four brain regions. b Summary of detected proteins in multiple brain regions. c 85 proteins were differentially expressed (FDR cut-off = 0.05) among the three clinico-pathologic groups across the four brain regions. d Of the 85 DEPs, 43 were significantly different between ADD versus HC, and 42 were RAD-associated, meaning significantly different between RAD and either HC or ADD in one or more regions. e 9 proteins were differentially expressed in RAD versus HC, and 33 proteins were differentially expressed in RAD versus ADD. Aβ and IF5 were differentially expressed in both RAD versus HC and RAD versus ADD (overlapping region in figure). We excluded RAD DEPs that were not detected in all regions, yielding 33 unique RAD DEPs measured in all four regions. CAUD caudate, HIPP hippocampus, IPL inferior parietal lobule, SMTG superior and middle temporal gyrus, HC healthy control, RAD resilience to AD, ADD Alzheimer’s disease and dementia. Source data are provided as a Source Data file.

Fig. 3. RAD DEPs among four brain regions and four external validation datasets.

Results of corrected multiple comparisons among four brain regions in the study set (a), and the four external validation sets of which only BLSA examined two brain isocortical regions (b). Each column for the study set consists of three comparisons: RAD vs. HC, ADD vs. RAD, and ADD vs. HC. Two-sided Student’s t-test was used with the Benjamin–Hochberg procedure (FDR = 0.05) to adjust the P-values. For external datasets, each column consists of three comparisons: AsymAD vs. Ctrl, ADD vs. AsymAD, and ADD vs. Ctrl (non-significant comparisons were colored in gray; insufficient data are white and annotated with NA; fold change (FC) < 1 was colored in blue, FC > 1 was colored in red; and colors are the same for -log₁₀(Adjusted P-value) ≥ 3). Annotations: * Adjusted P-value < 0.05; ** Adjusted P-value < 0.01, *** Adjusted P-value < 0.001. Correlation of expression of each RAD DEP with hallmark AD protein expression in the same brain region (c) and with clinical, genetic, or pathologic features of the individual (d). Spearman correlation test with two-sided P-values was used with the Benjamin-Hochberg procedure (FDR = 0.05) to adjust the P-values. Note: Aβ was a RAD DEP in both HIPP and SMTG, and CAPG was a RAD DEP in both HIPP and IPL. e Boxplots of selected RAD DEP expression. Two-sided Student’s t-test was used to derive P-values, followed with the Benjamin–Hochberg procedure (FDR = 0.05). Number of samples within each region and group are displayed under x-axes. For the boxplots, the interior horizontal line represents the median value, the upper and lower box edges represent the 75th and 25th percentile, and the upper and lower bars represent the 90th and 10th percentiles, respectively. f Principal component analysis (PCA) for 33 RAD DEPs in all 4 regions of each brain (original dimension: 33 × 4 = 132) colored by clinico-pathologic groups, visualized in principal dimensions 1 and 2. Variable contributions to the principal dimension 1 (g) and principal dimension 2 (h) with dashed lines in red showing variable contributions and their expected average. CAUD caudate, HIPP hippocampus, IPL inferior parietal lobule, SMTG superior and middle temporal gyrus, HC healthy control, RAD resilience to AD, ADD Alzheimer’s disease and dementia, DLPFC dorsolateral prefrontal cortex, PC precuneus. Source data of exact P-values are provided as a Source Data file.

Fig. 4. Consensus protein co-expression analysis and enrichment analysis results.

a Consensus protein co-expression analysis identified 9 modules across four brain regions. Pearson correlation with two-sided P-values was used to evaluate the relationships between clinico-pathologic groups and eigenprotein expression. Exact P-values are provided in the Source Data file. b Three co-expression modules contained the 33 RAD DEPs. c The number of expected and observed RAD DEPs in each module, and enrichment analysis via two-sided hypergeometric test. d Module 1 and 5 eigenprotein expressions in HC, RAD, and ADD for the study set and in Ctrl, AsymAD, and ADD for external validation sets. The number of samples in the study cohort: HC = 11, RES = 12, ADD = 20. The number of samples in the external validation cohorts: Banner (Ctrl = 42, AsymAD = 45, ADD = 92), ROS/MAP (Ctrl = 78, AsymAD = 89, ADD = 162), UPP (Ctrl = 26, AsymAD= 20, ADD = 49), and BLSA (DLPFC: Ctrl = 11, AsymAD = 13, ADD = 17; PC: Ctrl = 13, AsymAD = 13, ADD = 19). For the boxplots, the interior horizontal line represents the median value, the upper and lower box edges represent the 75th and 25th percentile, and the upper and lower bars represent the 90th and 10th percentiles, respectively. e Top 3 enriched GO biological process categories in M5 and their enrichment analysis results. f Patterns of the change in M5 z-scores. Font sizes of clinico-pathologic groups reflect average z-score changes within the GO categories. Abbreviations: CAUD caudate, HIPP hippocampus, IPL inferior parietal lobule, SMTG superior and middle temporal gyrus, HC healthy control, RAD resilience to AD, ADD Alzheimer’s disease and dementia, DLPFC dorsolateral prefrontal cortex, PC precuneus, GO gene ontology. Source data are provided as a Source Data file.