Loss of Adaptive DNA Breaks in Alzheimer’s Disease Brains

Human samples

Autopsy brain (frontal cortex) tissues were obtained from the University of Kansas Alzheimer’s Disease Research Center (KU ADRC) Neuropathology Core. For information on the human samples used in this study (AD, n = 3 males; controls, n = 3 males), please refer to Table 1.

Immunohistochemistry

For immunohistochemistry, FFPE histological sections were dewaxed, rehydrated in an ethanol series (95%, 85%, and 70%) and then subjected to microwave antigen retrieval in 0.01 M citrate (pH 6.0; 0.05% Tween-20) and methanol/H₂O₂ treatment. After blocking with 5% goat serum, the slides were sequentially incubated with anti-poly-ADP-ribose binding reagent (Millipore-Sigma, MABE1031) at a 1:1,000 dilution and HRP-labeled secondary antibodies. A NovaRed kit (Vector, SK-4800) was used for visualization.

Preparation of nuclei from human brain specimens

Nuclei were extracted from frozen human brain tissues by using Nuclei EZ Prep (Millipore-Sigma, NUC-101) following the manufacturer’s manual. The detailed procedures are described in [20]. Briefly, after thawing on ice, human brain tissues were minced by blades and homogenized in a Dounce homogenizer. The homogenate was then filtered through a 70-μm mesh strainer and centrifuged at 500×g for 5 min at 4°C. After removing the supernatant, the nuclei were resuspended in 1.5 ml of Nuclei EZ Lysis Buffer and incubated for another 5 min on ice. The nuclei were centrifuged at 500×g for 5 min at 4°C. After removing the supernatant, the nuclei were washed and then filtered through a 40-μm mesh strainer. Intact nuclei were counted after counterstaining with Trypan blue in a standard cell counter.

CUT&RUN

For CUT&RUN, 500,000 nuclei were washed 1x with CUT&RUN wash buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 0.5 mM spermidine), bound to activated ConA beads, permeabilized in wash buffer (wash buffer + 0.002% digitonin), incubated with anti-poly-ADP-ribose binding reagent, washed in wash buffer, incubated with pA-MN (EpiCypher 15–1016), and washed in wash buffer. Following the final wash, the cells were washed with ice-cold low-salt wash buffer and digested using MNase digestion buffer for 25 min on ice. Solubilized chromatin was released using an isosmotic stop buffer and collected using a PCR cleanup kit column.

CUT&RUN Library Prep was performed using an Illumina NovaSeq 6000 Sequencing System at the University of Kansas Medical Center Genomics Core (Kansas City, KS). Fragmented input and immunoprecipitated chromatin (5 ng) were used to initiate the TruSeq ChIP Sample Prep Kit library preparation protocol with modifications for CUT&RUN sample input (Illumina Cat# IP-202-1012). The fragmented chromatin underwent end repair and 3’ adenylation prior to Illumina indexed adapter ligation. No gel size selection of the ligation product was performed. Ten cycles of PCR amplification with a modified extension time of 10 s were performed using Illumina adaptor-specific priming with final library purification using KAPA Pure magnetic bead purification (KAPA Cat# KK8002).

Library validation was performed using a DNA 1000 Assay Kit (Agilent Technologies 5067-1504) on an Agilent TapeStation 4200. The concentration of each library was determined by qPCR using a Roche LightCycler 96 using FastStart Essential DNA Green Master Mix (Roche 06402712001) and KAPA Library Quant (Illumina) DNA Standards 1–6 (KAPA Biosystems KK4903). The libraries were pooled based on equal molar amounts to 1.85 nM for multiplexed sequencing.

The pooled libraries were denatured with 0.2 N NaOH (0.04 N final concentration) and neutralized with 400 mM Tris-HCl pH 8.0. Dilution of the pooled libraries to 370 pM was performed in the sample tube on the instrument, after which onboard clonal clustering of the patterned flow cell was performed using a NovaSeq 6000 S1 Reagent Kit v1.5 (200 cycle) (Illumina 20028318). A 2×101 cycle sequencing profile with dual index reads was completed using the following sequence profile: Read 1 –101 cycles×Index Read 1 –6 cycles×Index Read 2 –0 cycles×Read 2 –101 cycles. Following collection, the sequence data were converted from the.bcl file format to the.fastq file format using bcl2fastq software and demultiplexed into individual sequences for data distribution using a secure FTP site or Illumina BaseSpace for further downstream analysis.

CUT&RUN data processing and analysis

TrimGalore was used to trim the raw.fastq files to remove adaptors. The trimmed.fastq files were then mapped to the hg19 genome utilizing Bowtie2. The same procedure was run to align the.fastq files to a masked Saccharomyces cerevisiae v3 (sacCer3) genome for spike-in control DNA, which was also downloaded from the University of California Santa Cruz (UCSC) (https://genome.ucsc.edu/). Sambamba was then used to remove duplicates. For IGV visualization, deepTools was used with the “bamCoverage” function to generate normalized CPM.bw files. For peak calling, the recently developed SEACR was utilized and run in “relaxed’ mode to produce peak files, as the BED files used were already normalized to the number of yeast spike-in reads. DeepTools was further applied for heatmap visualization with the functions “computeMatrix” and “plotHeatmap”. The “dba.peakset” function of the Diffbind R package was further applied to identify overlapping peaks on the basis of the bound peaks. The links of CUT&RUN peaks and their related genes were established with the “annotatePeaks.pl” function in Homer. Motif enrichment analyses were performed using the “findMotifsGenome.pl” function in Homer, leading to known enrichment results and de novo enrichment results, and the latter were chosen in this study.

Loss of DSB sites in nervous system genes in AD

In our previous CUT&RUN targeting γH2AX [20], in addition to predominantly observing an increase in the number of γH2AX peaks in AD brains, there were also 12,167 γH2AX peaks that were downregulated. These downregulated γH2AX peaks were distributed in 3’ and 5’ untranslated regions (UTRs), exons, intergenic regions, introns, and promoters (Fig. 1A, B). Interestingly, in gene ontology (GO) analysis, γH2AX peaks in the intronic regions and intergenic regions were enriched for nervous system functions, e.g., dendrite development (SHANK3, CAMK2A), synapse organization (ARC, MECP2) (Fig. 1C). These data suggest that while DNA DSBs are formed in the intronic and intergenic regions of nervous system function genes in ND brains, presumably to facilitate their expression, these DSBs are lost in AD brains.

OPEN IN VIEWER

Fig. 1

Loss of CUT&RUN γH2AX peaks in AD brain. A) Heatmaps showing distribution of γH2AX peaks at 3’UTR, 5’UTR, exon, intergenic region, intron, and promoter. B) Distribution of γH2AX peaks in a±2 kb window at 3’UTR, 5’UTR, exon, intergenic region, intron, and promoter. C) GO analysis for PAR peaks at 3’UTR, 5’UTR, exon, intergenic region, intron, and promoter.

Mapping DNA break sites using CUT&RUN targeting PAR

Our prior CUT&RUN measured DSBs indirectly through adjacent γH2AX [20]. However, γH2AX can be found up to 50 kb away from both sides of DSBs in euchromatin [22]. Thus, to confirm our results, we performed CUT&RUN targeting ADP-ribose (PAR) by using an anti-PAR binding reagent. Various types of DNA damage, including DSBs, SSBs, and single-strand gaps, activate PAR polymerases (PARPs). PARP activity signals the presence of DNA damage by modifying adjacent proteins with PAR. Thus, CUT&RUN for PAR can be used to detect DNA breaks and map their genomic location [18].

We obtained autopsy brain tissue from the frontal cortex of postmortem human ND and AD patients through the University of Kansas Alzheimer’s Disease Research Center (KU ADRC) Neuropathology Core. The age, sex, and diagnostic information is shown in Table 1. First, we examined PAR-associated DNA damage in these samples. Immunohistochemistry (IHC) for PAR using histological sections from these patients showed a few cells positive for PAR in ND samples; in contrast, a significant increase in the number of cells positive for PAR was observed in AD samples, especially in neurons (74.4 ± 10.5 in AD samples versus 16.8 ± 6.9 in ND samples, p < 0.05) (Fig. 2A).

Following CUT&RUN, differential binding analysis of CUT&RUN and hierarchical clustering revealed that all 3 ND samples (ADC3, ADC5 and ADC14) were clustered together, while all 3 AD samples (ADC7, ADC11, and ADC13) were clustered together, indicating a high degree of conservation in the DNA break profile between ND and AD samples (Fig. 2B). There was a 19.9-fold increase in the number of PAR peaks detected in AD versus ND samples (Fig. 2C). Thus, consistent with previous reports, our CUT&RUN for PAR data shows an accumulation of DNA damage in AD. Furthermore, distribution analysis shows that many PAR peaks are in the intergenic and intronic regions in both ND and AD samples (Fig. 2D), except ADC7, which shows a profound loss of CUT&RUN PAR peaks in the intergenic and intronic regions (γH2AX peaks are not lost in ADC7 in these genomic regions [20]).

OPEN IN VIEWER

Fig. 2

CUT&RUN for PAR analysis in AD brain. A) Increased PAR staining by IHC in AD samples. Scale bar, 100μm. PAR-positive cells were quantified from multiple view areas in 3 ND and 3 AD samples. Data represent mean±SD. B) Diffbind analysis show clustering of CUT&RUN for PAR in AD and ND samples. C) Elevated PAR peaks (DNA breaks) in AD versus ND samples. D) Distribution of differential PAR peaks in genome.

Altered genomic distribution of PAR peaks in AD brains

CUT&RUN for PAR shows strong central peaks, which presumably indicate DNA break sites (Fig. 3A, B, D, and E). GO analysis shows that these DNA break sites upregulated in AD occur at genes that encode regulators of proteasome-mediated ubiquitin process (e.g., UBE2J2, DVL1, PARK7), small molecule catabolic process (e.g., ENO1, PGD, ALDH4A1), and histone modification (e.g., NOC2L, PRDM2, PADI2), etc. (Fig. 3C). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis identified protein processing in endoplasmic reticulum (e.g., FBXO2, FBXO6), carbon metabolism (e.g., SDHB, PKLR), etc. (Fig. 3C).

OPEN IN VIEWER

Fig. 3

PAR peaks in ND and AD samples. A, D) Heatmaps showing distribution of AD-enriched (A) and ND-enriched (D) PAR peaks. B, E) Distribution of±5 kb window of AD-enriched (B) and ND-enriched (E) PAR peaks. C, F) GO and KEGG analysis of AD-enriched (C) and ND-enriched (F) PAR peaks.

Interestingly, GO analysis of DNA break sites downregulated in AD revealed genes that encode regulators of synapse organization (e.g., NFIA, ARHGAP22, ANK3), cognition (e.g., PRKCZ, FOXO6, SORCS3), glutamatergic synaptic transmission (e.g., GRIK3, LRRK2, SYT1), etc. (Fig. 3F). KEGG analysis identified calcium signaling pathway (e.g., PDE1C, CAMK2D, PLCE1), neuroactive ligand-receptor interaction (e.g., PRL, CHRNA7, CHRM5), Axon guidance (SEMA6A, CAMK2D, SLIT2), etc. (Fig. 3F).

Together, these data suggest DNA breaks that form in genes mediating nervous system function are lost in AD. Instead, AD-associated DNA breaks are strongly accumulated at ectopic genomic sites associated with novel functions, such as ubiquitin and catabolic processes.

PAR peaks upregulated in AD samples

We examined the enrichment profiles of PAR peaks in ND and AD samples, which showed distribution around the 3’-UTR, 5’UTR, exons, and promoters (Fig. 4A, B). Interestingly, our data show that CUT&RUN PAR peaks at these genomic regions are enriched for different functions in functional pathway identification analysis (Fig. 4C). While PAR peaks at the promoters and exons are uniquely enriched for pathways related to cadherin binding (e.g., CDH11, CTNND1) and ATP hydrolysis activity (e.g., ASCC2, FBX01), PAR peaks at 3’-UTR and 5’-UTRs encode pathways related to SNARE binding (e.g., UVRAG, VAMP7) and ligase activity (e.g., ASCL4, SCL27A2). Moreover, functional term identification analysis (Fig. 3D) showed that PAR peaks at promoters and exons are uniquely enriched for many functions, such as protein processing in endoplasmic reticulum (e.g., CUL3, CTAGE4) and regulation of actin cytoskeleton (e.g., ROCK1, WASHC5). Only 248 PAR peaks were upregulated in the intron and intergenic regions and functional pathway identification analysis shows that they are enriched for functions, such as cellular response to nutrient levels (e.g., BCL2, LAMP2) and cellular response to extracellular stimulus (e.g., ATG7, GSDMD) (data now shown).

OPEN IN VIEWER

Fig. 4

Genomic distribution of PAR peaks upregulated in AD samples with functional analysis. A) Heatmaps showing distribution of PAR peaks at promoter, intron, exon, 3’UTR and intergenic. B) Distribution of PAR peaks in a±2 kb window of γH2AX binding sites at promoter, intron, exon, 3’UTR, and intergenic. C) GO analysis for PAR peaks at promoter, exon, 3’UTR and 5’UTR. D) KEGG analysis for PAR peaks at promoter, exon, 3’UTR and 5’UTR.

PAR peaks downregulated in AD samples

We further examined the enrichment profiles of CUT&RUN PAR peaks in ND and AD samples, which showed distribution around the promoters, introns, intergenic regions, and exons (Fig. 5A, B). Interestingly, functional pathway identification analysis demonstrates that PAR peaks at the promoters are enriched for functions, such as protein localization to nucleus (e.g., CDK1, CTNNA1), at the exons are enriched for functions, such as inorganic cation import across plasma membrane (e.g., ABCC9, ATP1A1), and at intergenic regions are enriched for functions, such as regulation of protein catabolic process (e.g., ABCA2, ADAM8) (Fig. 5C). Importantly, similar to the intronic gH2AX peaks downregulated in AD (Fig. 1), PAR peaks at introns downregulated in AD are also enriched for nervous system-related functions (e.g., ADCY1, CHD7) (Fig. 5C).

OPEN IN VIEWER

Fig. 5

Genomic distribution of PAR peaks downregulated in AD samples with functional analysis. A) Heatmaps showing distribution of PAR peaks at promoter, intron, intergenic, and exon. B) Distribution of PAR peaks in a±2 kb window of γH2AX binding sites at promoter, intron, intergenic, and exon. C) GO analysis for PAR peaks at promoter, intron, intergenic, and exon. D) KEGG analysis for PAR peaks at promoter, intron, intergenic, and exon.

Differential PAR peaks in AD samples

We performed differential peak analysis on DNA breaks located near transcription start sites (TSSs) (Fig. 6A) and used Hypergeometric Optimization of Motif EnRichment (HOMER) reference to examine sequences related to transcription factor (TF)-binding sites. Our analysis showed that while the binding motifs for the CEBP and CREB TFs were among the most enriched in ND-associated peaks (Fig. 6B), the binding motifs for the Olig2 and RARα TFs were among the most enriched in AD-associated peaks (Fig. 6C).

OPEN IN VIEWER

Fig. 6

Differential PAR peak analysis and correlation with gene expression. A) Differentially enriched PAR peaks between AD and ND. Number of differential peaks and peaks with > 2-fold change is labeled. B) Sequences and significance of enrichment of DNA motifs at PAR in ND samples. C) Sequences and significance of enrichment of DNA motifs at PAR in AD samples. D) GO analysis of biological processes for AD and ND enriched peaks. E) Representative image showing identified peaks binding in the SHMT1 gene locus from AD and ND. F) Representative image showing identified peaks binding in the APP gene locus from AD and ND. G) Violin plots showing the expression of representative ND-associated gene signature levels. H) Violin plots showing the expression of representative AD-associated gene signature levels.

After linking peaks to respective genes using the Genomic Regions Enrichment of Annotations Tool (GREAT) [23], we identified 8,326 genes with upregulated- and 638 genes with downregulated PAR occupancy in AD samples, suggesting that these genes have increased and decreased numbers of DNA breaks in AD samples, respectively. Next, we analyzed their functions. GO analysis showed that the genes with upregulated PAR occupancy in AD samples were enriched for functions such as the organic acid catabolic process; in contrast, genes with downregulated PAR occupancy in AD samples were enriched for nervous system-related functions such as axon development, synapse organization, etc. (Fig. 6D). For example, SHMT1 encodes a folate-dependent enzyme that mediates neuronal and cognitive function by regulating de novo thymidylate biosynthesis [24]. PAR peaks were detected at the SHMT1 gene in ND but not AD samples (Fig. 6E), suggesting that DNA breaks are formed at the SHMT1 gene in ND samples but not in AD samples. Conversely, PAR peaks were detected in the APOE gene in some AD samples but not in any of the ND samples (Fig. 6F).

To identify the potential impact of the observed altered DNA breaks on gene expression, we examined the gene expression profile in an RNA-seq dataset of 10 ND samples and 12 AD samples [25]. We found that genes associated with nervous system function-related GO terms (“synapse organization”, “axon development”, etc.) was significantly downregulated in AD samples (Fig. 6G). In contrast, the expression of genes with GO terms (“organic acid catabolic process”, “small molecule catabolic process”) with AD-associated DNA breaks was upregulated in AD samples (Fig. 6H). These data suggest that AD-related loss and gain of DNA breaks in the genes bearing these GO functions result in their down- and upregulated gene expression, respectively.

Downregulation of TOP2B expression in AD brains

DNA topoisomerases catalyze DNA break formation. To test whether altered DNA breaks results from dysregulated DNA topoisomerase expression in AD, we initially examined the expression of DNA topoisomerases (TOP1, TOP2A, TOP2B, TOP3A, and TOP3B) using the RNA-seq dataset [25]. Our data show that only the expression of TOP2B and TOP3B significantly change in AD. Specifically, TOP2B expression is downregulated in AD (Fig. 7A), while TOP3B is aberrantly upregulated in AD (data not shown).

OPEN IN VIEWER

Fig. 7

TOP2B expression in ND and AD brains. A) TOP2B expression is reduced in AD in the indicated RNA-seq dataset. B) TOP2B expression is reduced in AD in three published RNA-seq datasets. C) scRNA-seq: TOP2B is downregulated in INH and EX neurons but upregulated in ODC neurons. *p < 0.05. D) Reduced TOP2B IHC staining in AD versus ND brains. Scale bar, 50μm. TOP2B-positive cells were quantified from multiple view areas in 3 ND and 3 AD samples. Data represent mean±SD.

TOP2B plays a critical role in producing DSBs to facilitate IEG expression, which mediates synaptic plasticity following neuronal activity [16, 17]. Thus, downregulated TOP2B expression is consistent with a loss of DNA breaks near genes associated with nervous system function genes. We further examined TOP2B expression in three published large-scale RNA-seq studies using ND and AD samples [26–28]. In all three datasets, TOP2B expression was significantly downregulated in AD brains (Fig. 7B). By investigating a single cell RNA sequencing (scRNA-seq) dataset [29], we found that TOP2B expression was significantly downregulated in AD brain inhibitory (INH) and excitatory (EX) neurons, and upregulated in oligodendrocytes (ODC) (Fig. 7C). To examine TOP2B expression at the protein level we conducted IHC staining, which showed that there is a significantly decreased number of cells expressing TOP2B in AD brains (21.7±9.3 in AD samples versus 82.4±7.0 in ND samples, p < 0.05) (Fig. 7D). Together, these data are consistent with the possibility that loss of DNA breaks in genes critical nervous system function may reflect TOP2B expression in AD brains.