Cell-death pathways and tau-associated neuronal vulnerability in Alzheimer's disease

Abstract

Neuronal loss is the ultimate driver of neural system dysfunction in Alzheimer's disease (AD). We used single-nucleus RNA sequencing and neuropathological phenotyping to elucidate mechanisms of neurodegeneration in AD by identifying vulnerable neuronal populations and probing for their differentially expressed genes. Evidenced by transcriptomic analyses and quantitative tau immunoassays of human AD and non-AD brain tissue, we identified a neuronal population especially vulnerable to tau pathology. Multiplexed immunohistochemistry and in situ hybridization (CBLN2 and LINC00507) validated the presence of the tau-vulnerable neuronal population and revealed a propensity of this population to bear tau pathology. Differentially expressed genes associated with phospho-tau pathology in these neurons revealed genes involved in apoptosis, cell-component dissociation (e.g., autophagosome maturation and actin filament depolymerization), and regulation of vesicle-mediated transport.

Keywords: Alzheimer’s disease; CP: Neuroscience; multiplex FISH/IHC; neurodegeneration; phospho-tau ELISA; selective vulnerability; single-nucleus RNA-seq; tau pathology.

Figure 1.. Single-nuclei RNA-seq results from 30 control and AD donors

(A) Cortical samples were taken from five brain regions corresponding to the spread of tau pathology during typical AD progression (EC, entorhinal cortex; ITG, inferior temporal gyrus; PFC, prefrontal cortex; V2, secondary visual cortex; V1, primary visual cortex). Nuclei derived from these samples were sorted for NeuN positivity before snRNA-seq. Cortical tissue samples were analyzed by pT231/Total Tau ELISA and HEK seeding assays. Fresh frozen slides were cut from each sample for later in situ analyses. (B) Annotated neuronal subtype from scVI integration of 125 samples across five brain regions (Table S1) followed by clustering with Leiden algorithm, showing 20 discrete neuronal clusters (Table 1 and Figure S12). (C) (Top) Nuclei counts for each detected neuronal subtype. Bar color indicates the brain region of origin for the nuclei comprising each subtype. (Bottom) Heatmap of neuronal subtype marker expression calculated from 2,000 random nuclei sampled from each subtype. (D) (Top) Tau pathology as quantified by pT231/Total Tau ELISA across brain regions and stratified by Braak stage. (Bottom) Assessment of tau seeding activity across brain regions as measured by HEK seeding assay and stratified by Braak stage. Data are presented as mean ± SEM. See also Figure S2; Tables S1, S8, and S11.

Figure 2.. Transcriptomic changes associated with Braak stages

(A) Multidimensional scaling (MDS) plot based on the Braak associated log₂(fold change) between each pair of samples from 14 neuronal subtypes and four brain regions (ITG, PFC, V2, and V1). Inhibitory (triangles) and excitatory (circles) subtypes are indicated by shape and brain region origin of subtype by color, and the ellipses show the grouping of the Braak-associated log₂(fold change) by brain region. (B) All GO terms that have significant −log10 (adjusted p value) differences with Braak-stage-associated gene-expression fold change between the ITG and V1. Colorbar denotes −log10 (adjusted p value). (C) Heatmap showing pairwise correlation between Braak-associated fold changes in each neuron subtype from ITG. Rows and columns are clustered using ward.D2. Inhibitory (Inh.) or excitatory (Exc.) status of the neurons is indicated next to each neuronal subtype label. Colorbar denotes Spearman’s rank correlation coefficient. (D) Boxplot showing difference in statistically significant (p < 0.001), Braak-associated log₂(fold change) for selected enriched GO terms between inhibitory and excitatory neurons in ITG. Data are represented as mean ± SEM. See also Figures S2 and S3; Tables S2, S3, and S10.

Figure 3.. Transcriptomic and histological evidence of an association of L2/3 IT ITG neurons with pTau

(A) Dot-and-whisker plot showing the estimated pT231/Total Tau final parameter and 95% confidence interval derived from scCODA analysis (STAR Methods) for each neuron subtype in ITG. (B) Scatterplot showing the Pearson correlation (r = −0.51815, p value = 0.00474) of L2/3 IT neurons (number of L2/3 neurons/number of all neurons) with pT231 abundance for each sample in the ITG. (C) Tau-vulnerable neurons (L2/3 IT neurons detected based on positivity of NeuN [antibody] and the number of copies of LINC00507 [mRNA probe] and CBLN2 [mRNA probe]) (STAR Methods). Tau pathology was measured by AT180 (pT231; antibody) immunopositivity within each detected neuronal cell body. Scale bars, 50 μm. (D) Simple linear correlation (r = −0.4355, p value = 0.0262) of the abundance of L2/3 IT neurons and AT180 integrated intensity within the analyzed image area (STAR Methods). (E and F) (E) AT180 integrated intensity of L2/3 IT neurons and all other neurons demonstrate that L2/3 IT neurons have higher probability of containing tau pathology (two-way Wilcoxon ranked pairs test, p value < 0.0001) and (F) present with significantly higher pTau burden compared to all other neurons on average (two-way Wilcoxon ranked pairs test, p value = 0.011). (G) (Left) Spatial plot images from one AD and one control donor of medial temporal gyrus tissue with layer annotations derived from stereotypical cortical layer makers expressed in each spot (data from Chen et al.; STAR Methods). Expected abundance of L2/3 IT neurons in each spot after cell-type annotation of spatial transcriptomics data to our annotation. The color bar depicts the abundance of L2/3 IT neurons in each spot. (Right) Expression of CBLN2 and LINC00507 transcripts in donor tissue sections. The color bar depicts abundance of transcripts in each spot. Scale bars, 1.3 mm. See also Figures S1–S8 and Table S4.

Figure 4.. Transcriptional profiling of L2/3 IT neurons

(A) MA plot showing 641 genes with a significant positive pT231/Total Tau coefficient and 592 genes with significant negative pT231/Total Tau coefficient as identified in L2/3 IT neurons in ITG. Significance based on q value < 0.05, denoted in legend by color of data point. (B) Dotplot showing −log10(p value) and enrichment score (number of genes in gene set/total number of differentially expressed genes) for top ten GO terms from an enrichment analysis of the 641 genes with a significant positive pT231/Total Tau coefficient. Colorbar denotes −log10(p value), size of data point denotes count. (C) Dot-and-whisker plot showing the mean and 95% confidence interval of the log odds ratio of AT8⁺ vs. AT8⁻ neurons of each neuronal subtype in the Otero-Garcia et al. dataset, based on our annotation. Log odds ratio is calculated for every sample as log(proportion of neuron subtype in AT8⁺ fraction/proportion of neuron subtype in AT8⁻ fraction). (D) Density scatterplot of the AT8⁺ vs. AT8⁻ log₂ (fold change) of L2/3 IT neurons in the Otero-Garcia et al. dataset vs. pT231/Total Tau regression coefficient of L2/3 IT neurons in this study (r = 0.56). See also Figure S7. (E) SBGNview^, graphical representation of the apoptosis signaling pathway (PANTHER: P00006) from the PANTHER database. Genes producing protein products in white boxes were not represented in L2/3 IT ITG neurons (percent expression <3%). Genes producing protein products in gray boxes held no association to pTau in L2/3 IT ITG neurons. Genes producing protein products in red boxes were positively associated and genes in green boxes negatively associated with pTau in L2/3 IT ITG neurons. Color bar denotes log₂(fold change) (Table S4). See also Figures S7, S9, S10, and S11; Tables S4, S5, and S7.