Increased post-mitotic senescence in aged human neurons is a pathological feature of Alzheimer's disease

Abstract

The concept of senescence as a phenomenon limited to proliferating cells has been challenged by growing evidence of senescence-like features in terminally differentiated cells, including neurons. The persistence of senescent cells late in life is associated with tissue dysfunction and increased risk of age-related disease. We found that Alzheimer's disease (AD) brains have significantly higher proportions of neurons that express senescence markers, and their distribution indicates bystander effects. AD patient-derived directly induced neurons (iNs) exhibit strong transcriptomic, epigenetic, and molecular biomarker signatures, indicating a specific human neuronal senescence-like state. AD iN single-cell transcriptomics revealed that senescent-like neurons face oncogenic challenges and metabolic dysfunction as well as display a pro-inflammatory signature. Integrative profiling of the inflammatory secretome of AD iNs and patient cerebral spinal fluid revealed a neuronal senescence-associated secretory phenotype that could trigger astrogliosis in human astrocytes. Finally, we show that targeting senescence-like neurons with senotherapeutics could be a strategy for preventing or treating AD.

Keywords: Alzheimer’s disease; SASP; aging; induced neurons (iNs); inflammation; senescence; senolytics.

Figure 1.. The human AD brain harbors an increased proportion of senescent neurons

(A) Expression of CDKN2A in the brains of Alzheimer’s disease (AD) patients and cognitively normal (NC) controls in the pre-frontal cortex (PCx) and frontal white matter (FWM) (Wald test DESeq2, n = AD 30; 99 NC). (B) Fluorescence microscopy analysis of p16 expression in post-mortem PCx tissue from AD and NC patients, left 20X, right 63X, scale bar = 16 μm. (C) Quantification of p16 expression in NeuN⁺ cells (left) and DAPI⁺ nuclei (right) (n = AD 10; 10 NC, unpaired t test). (D) Clustering analysis of p16 signal adjacent to other p16⁺ loci (top left), DAPI⁺ loci (top right), NeuN⁺ loci (bottom left), and NeuN adjacency compared with DAPI adjacency (bottom right) (n = AD 10; 10 NC, paired t test). (E) Example cumulative distribution function for p16 signal near NeuN⁺ and DAPI⁺ loci used to calculate the adjacency AUC in (D). *p < 0.05, **p < 0.01, ****p < 0.0001.

Figure 2.. Senescence changes in the transcriptome and epigenome of aged AD neurons

(A) Experimental schema. (B) GSEA of differential expression of AD and CTL in four cell types: iNs (aged induced neurons n = 35), iPSC-iNs (rejuvenated young neurons, n = 20), iPSCs (n = 20), and fibroblasts (n = 35); indicates a significant enrichment exclusively (q value = 0.012) in AD iNs. (C) Volcano plot of genes used for GSEA in (B) in iNs. Genes above the significance threshold (padj < 0.05, Log2FC > 1) are colored red. (D) Histogram of peaks averaged from AD 10 and 10 CTL iN samples across senescence genes, including the transcription start site (TSS) and transcription end site (TES) and 2kb window (left). Peak counts in senescence genes are significantly higher in AD iNs than in CTLs (right); each point represents the number of peaks in a single gene in our senescence gene list (Mann-Whitney test). (E) Increased promoter methylation is negatively associated with senescence gene expression in CTL iNs and positively associated with expression in AD iNs (n = AD 8; 8 CTL). (F) GO analysis of genes with low promoter methylation and high expression in AD iNs. (G) H2AFJ is expressed significantly higher in AD iNs (right) and shows significantly brighter staining by ICC (left); scale bar = 16 μm. (H) The binding motif for the transcription factor ETS1 is significantly enriched in open chromatin in AD (target) relative to CTL (background) iNs. (I) There are no significant differences between telomere erosion in AD and CTL parent fibroblast lines or in total telomere content, as measured by qPCR. *p < 0.05, **p < 0.01.

Figure 3.. iNs present bona fide markers of cellular senescence

(A) AD iNs have increased p16 protein measured by chemiluminescence in a capillary western blot analysis (n = AD 9; 9 CTL, unpaired t test). (B) Example confocal images of p16 co-stained with Tuj in PSA-NCAM-sorted iNs; scale bar = 16 μm (left). AD iNs have a significantly increased population and intensity of p16⁺ cells (right) (n = AD 9; 8 CTL, unpaired t test). (C) Example picture of PSANCAM-sorted iNs stained for SA-B-Gal and co-stained for mature neuron marker NeuN (left); scale bar = 8 μm. Quantification of BGal⁺ rates in AD and CTL iN cultures (right) (n = AD 17, 16 CTL, unpaired t test). (D) Example FACS gating to separate C12⁺ from C12⁻ cells (left). AD iNs have a significantly larger C12⁺ population, as indicated by a higher geometric mean of C12 signal (right) (n = AD 4; 4 CTL, unpaired t test). (E) Example images of CTL and AD nuclei stained with DAPI (left). AD iNs have increased nuclear area measured by integrated DAPI signal (right) (n = AD 10 and 10 CTL patients, unpaired t test). (F) Example images of p16 and UNG:GFP in iPSC-derived AD iNs; scale bar = 16 μm (left). Quantification of p16 signal intensity in iPSC-derived AD iNs (right) (n = 3 iPSC-iN; 3 iN, unpaired t test). (G) Example BGal staining in iPSC-iN cultures; scale bar = 8 μm (left). Quantification of BGal⁺ rates in AD iPSC-iNs or fibroblast iNs (right) (n = 3 iPSC-iN; 3 iN, unpaired t test). (H) P16ink4a is not expressed in iPSC-iNs. (I) Senescent-like phenotypes manifest only in aged human neurons, but young rejuvenated iPSC-iNs do not present senescence markers under baseline conditions. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 4.. scRNA-seq indicates oncogenic challenges and metabolic dysfunction in senescent neurons

(A) ScRNA-seq allows us to discriminate senescent gene expression from the majority of non-senescent cells that mask expression in bulk RNAseq. (B) UMAP clustering of all unsorted fibroblast to iN cultures after 3 weeks of conversion from AD 4 and 3 CTL patients. (C) Mean module score for each patient in iNs. (D) Volcano plot of genes differentially expressed between sen-low and sen-high iNs. (E) GSEA of gene sets enriched (right) or depleted (left) in sen-high cells. (F) Gene network analysis of genes enriched in AD iNs in sen-low (left) or sen-high (right) populations. (G) UMAP projection of iNs colored by their rank in pseudotime analysis tracking the expression of CDKN2A. (H) Density plot of patient proportions across CDKN2A pseudotime and gene expression changes across pseudotime.

Figure 5.. AD-conditioned media triggers reactive astrogliosis

(A) Experimental paradigm. Cerebral spinal fluid (CSF) and patient iN conditioned media (CM) were submitted for a multiplex inflammation panel, and CM was applied to human astrocytes to evaluate reactivity. (B) Overlap of log2FC of inflammatory probes (AD/CTL) from CSF and iN CM. Co-upregulated in AD in red and co-downregulated in blue. (C) Differential expression analysis between AD and CTL CM-treated astrocytes reveals 981 genes significantly differentially expressed (n = 3 astrocyte cell lines). (D) Expression of reactive astrocyte markers in AD- or CTL-treated astrocytes (n = 3, Wald test DESeq2). (E) A larger population of senescent-like neurons triggered a hyperreactive state in human astrocytes. (F) RRHO mapping of the transcriptional overlap between astrocytes from post-mortem patients with or without AD pathology and between astrocytes treated with AD or CTL CM in vitro. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 6.. Senescence in neurons can be triggered by oncogenes and eliminated with senotherapeutics

(A) Ras overexpression experimental scheme. (B) Representative image of Ras- or GFP-treated iNs and p16 staining (left). Ras-treated iNs are significantly more likely to be p16⁺ than GFP controls (right) (n = 8 biological replicates per condition, paired t test). (C) AD iN cultures experience a dose-dependent increase in number of TUNEL⁺ cells after exposure to senolytic cocktail dasatinib +10 μM quercetin (n = AD 3; 3 CTL, two-way ANOVA). (D) Representative field of iN cultures from AD or CTL patients treated with 1.5 μM dasatinib +10 μM quercetin. (E) AD cultures have significantly reduced p16⁺ cells after D + Q treatments that are comparable to CTL p16 levels (n = AD 3; 3 CTL, one-way ANOVA). (F) Theory cartoon. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 7.. Neuronal senescence manifests in impaired electrical activity

(A) Experimental schema. Sorted iNs were treated with p16 or control lentivirus and cultured on multi electrode array (MEA) plates for one month of recording (left). Example images of synpasin:dsRed-labeled iNs on an MEA plate at 24 h and one month after replating (right). (B) Ribbon plot of mean firing rate of p16 or GFP iNs. The slope of activity change overtime was significantly higher in GFP than in p16-expressing iNs (n = 18 recording wells for each condition; two-tailed t test). (C) The cumulative mean firing rate across all 30 days of recording in GFP iNs was significantly higher than p16 iNs. Measurements across all wells were averaged into a single point for a given day (n = 30, paired t test). (D) Activity of iNs could be ablated with treatment of 1 uM tetrodotoxin (n = 18 recording wells, paired t test). (E) There were no significant differences in the mean firing rate of GFP and p16 iNs following electrical stimulation (Stim) (n = 18 recording wells for each condition, paired t test). *p< 0.05, **p< 0.01.