Early age decline in DNA repair capacity in the liver: in depth profile of differential gene expression

Age-dependent decline in DNA damage repair

To test the efficiency and kinetics of DNA damage repair in vivo, we induced DNA damage in mouse livers by a single injection of DEN. We performed immunofluorescence staining to detect cells containing phosphorylated histone H2AX (γH2AX) in liver tissue sections at various time points following the DEN injection. Using this mouse model we have previously demonstrated that cells positive for γH2AX appear 24h after DEN injection and persist up to three days post-injection. The DNA damage was resolved by day six post DEN treatment, and at this point in time the number of cells harboring γH2AX foci and their relative intensity significantly declined, leaving only a few cells with detectable foci [25]. One month-old mice were used in these previous experiments. In order to test whether DNA damage response declines with age, we compared the efficiency of DNA damage repair following DEN injection in 1-, 3-, 6- and 12-month-old mice.

Since DEN itself does not exert hepatocyte toxicity, it needs to be metabolically activated by cytochrome P450 enzymes (CYP) in the liver, predominantly by the cytochrome P450 2E1 (CYP2E1), resulting in DNA-adducts formed through an alkylation mechanism [26]. We, therefore, analyzed expression of this cytochrome P450 isoform by quantitative real time PCR. We did not detect any significant alterations in expression levels of CYP2E1 between mice at various ages (Supplementary Fig. S1A). Furthermore, out of 117 CYP isoforms that were detected in an RNAseq analysis only 15 were differentially expressed between 1 month and 6 month-old mouse livers, and only 4 were upregulated with age. Consistent with similar metabolic activation of DEN at all ages, we detected comparable levels of DNA double strand breaks, as indicated by γH2AX immunofluorescence, 48h after DEN insult. There was no significant difference in the extent of initial damage noted in the liver two days after DEN injection in mice of all ages, however, the extent of residual DSBs six days after DEN injection was significantly higher in 6- and 12-month old mice as compared to younger mice (Fig. 1A). The extent of DNA damage repair was calculated by dividing the area of DNA damage at the peak of damage (after 48h) and the residual damage after the resolution phase (6 days post DEN treatment). While, approximately 80% of the initial damage was resolved in 1- and 3-month-old mice by day 6, only ∼25% of the damage was resolved in 6- and 12 month old-mice although the intensity of γH2AX staining was somewhat lower compared to 48h post DEN treatment (Fig. 1A).

Interestingly, the level of residual damage in 6-month-old mice further declined to around 40% at days 9 and 12 after DEN treatment and the intensity of residual foci were further reduced as compared to day 6 (Fig. 1B), suggesting that with age DNA damage repair process is both impaired and significantly delayed.

The heterochromatin protein KRAB-associated protein (KAP-1), a co-repressor of gene transcription, is phosphorylated by Ataxia telangiectasia mutated (ATM) at serine 824 in response to DNA damage. Phosphorylated KAP-1 (pKAP-1) forms foci overlapping with γH2AX at sites of DSBs [27,28] and was suggested to control DNA repair in heterochromatin [29,30]. We, therefore, performed immunofluorescence staining to detect pKAP-1 in liver tissue sections from various ages and at various time points following DEN injection. Like γH2AX, pKAP-1 foci were clearly evident after DEN injection, and were co-localized with relatively intense γH2AX-positive nuclei (Fig. 1C). The time-dependent appearance and recovery of pKAP-1 followed the same pattern and kinetics as γH2AX at all mouse ages (Supplementary Fig. S1B and not shown), thus confirming γH2AX results. However, pKAP-1 foci seemed to resolve better than γH2AX foci that are more persistent (Supplementary Fig. S1B).

DEN induces age-dependent changes in gene expression profiles

To more sensitively detect potential age-related changes in DEN-induced DNA damage response in the liver we used RNAseq analysis. We specifically looked for genes that are differentially expressed two and six days after DEN treatment in 1- and 6-month-old mice (the earliest age where DNA damage resolution decline was detected). Two days after DEN-treatment 89 differentially expressed genes were identified in livers of 1 month old mice that were classified into two clusters based on the expression profile using hierarchal clustering. One cluster contained 47 genes that were downregulated as compared to the control and the second cluster contained 42 upregulated genes (Fig. 2A). Overall, most differentially expressed genes (both up- and down-regulated) returned to baseline levels by day 6, thus demonstrating that the response to DEN was largely resolved at the transcriptome level. This hypothesis was reinforced by the finding that only eight of the 89 differentially expressed genes at day 2 were also differentially expressed at day 6, and all together only 18 genes were found to be differentially expressed in day 6 as compared to the control (Fig 2D). Remarkably, most downregulated genes in 1-month-old mice were not differentially expressed in 6-month old mice. In contrast, most upregulated genes in 1 month old mice were also upregulated in 6-month old mice, however they did not return to basal levels at day 6 (Fig. 2A).

In 6-month-old mouse livers, 116 genes were significantly differentially expressed at day 2 in contrast to 1-month-old mice where the majority of these genes (both up- and down-regulated) did not resume basal expression levels at day 6 (Fig. 2B). Surprisingly, despite the comparable response to DEN at day 2 (Fig. 1A), 1 month and 6 month old mice shared only 26 differentially expressed genes at day 2 (Fig. 2E). Comparing 6 days post DEN to the control revealed 80 differentially expressed genes, 67 of which were shared with the differentially expressed genes at day 2 (Fig. 2F). Surprisingly, the fold change in the expression levels of these shared genes did not significantly reduce at day 6 as compared with day 2 (Supplementary Fig. S2A), demonstrating that between day 2 and day 6 the response did not significantly decline.

In keeping with the observed age-related decline in DNA damage repair, the data demonstrated a marked difference in gene expression pattern in response to genotoxic damage between 1 and 6 month old mice and the extent to which the response is resolved.

To validate RNAseq results, we confirmed the transcriptional levels of genes representing different expression patterns by qRT-PCR (Supplementary Fig. S2B). These genes exhibited differential expression in qRT-PCR that was consistent with the RNAseq data, indicating good concordance of both methods. In addition, we tested the pattern of expression of these genes in 3- and 12-month old mice and demonstrated that they followed similar pattern of expression as 1- and 6-month-old mice, respectively, thus closely paralleling their DNA damage resolution kinetics. Moreover, for genes that did not resume basal levels in 6-month-old mice by day 6 we further analyzed their expression at day 9 and 12. In accordance with the reduced levels of DSBs in the liver at these time points (Fig. 1B), the level of expression was also slightly reduced at day 12 (but did not return to basal levels).

Functional analysis of differentially expressed genes

After identifying the profiles of differentially expressed genes for each comparison, we performed functional enrichment analysis to reveal transcripts putatively involved in potential relevant biological processes, signaling pathways and networks using ClueGO in Cytoscape. The Venn diagram analysis shown in the previous section allowed us to uncover common and exclusively differentially expressed genes between the two mouse ages, underscoring potential common and exclusive biological functions regulated in each case.

Genes associated with p53 signaling pathway and intrinsic apoptotic signaling pathway in response to DNA damage by p53 were significantly enriched 2 days after DEN injection in both 1-month and 6-month old mice (Fig. 3A-B). However, in 6-month old mice a more robust response was observed. In addition to these two pathways other related pathways and biological functions were also activated including the response to toxic substance, response to UV and gamma radiations and others (Fig. 3B). Consequently, the 26 common genes between the two groups were also associated with signal transduction by p53 (Fig. 3C). The fold change in expression of these shared genes was significantly higher in 6-month-old mice as compared to 1-month-old mice (Supplementary Fig. S3). No significant enriched signaling pathway or biological function was identified among genes that are unique to 1-month-old mice, whereas genes unique to 6-month old mice were enriched for regulation of execution of apoptosis, regulation of fibroblast proliferation, cellular oxidant detoxification, and positive regulation of transcription from polymerase II promoter in response to stress (Fig. 3C). Although the number of common genes between the two ages is surprisingly low, they share a similar functional response based on p53 module with an overall more robust response in 6-month old mice.

At day 6 after DEN treatment no significantly enriched pathway was detected in 1-month old mice. The most relevant biological processes found in genes differentially expressed in 6 month-old mice 6 days after DEN were p53-signaling pathway, intrinsic apoptotic signaling pathway in response to DNA damage by p53 and cellular response to UV (Fig. 4A). These pathways were shared between the two time points and were enriched in differentially expressed genes common to day 2 and 6 (Fig. 4B).

Ingenuity® Pathway Analysis (IPA), pathway and upstream regulator analysis of differentially expressed genes, further corroborated these findings demonstrating a more robust response in 6-month-old mice that was not resolved by day 6 after DEN treatment. In general, most upstream regulators identified by this analysis were associated with p53 signaling pathway and agents inducing DNA damage (Supplementary Fig. S4A-C).

Additionally, we combined differentially expressed genes from all 4 groups and analyzed putative protein interaction network analysis. This analysis once again confirmed the results and highlighted a common interaction network that related to the genes involved with p53 signaling and intrinsic apoptotic pathways (Supplementary Fig. S4D).

Collectively, the transcriptional data correlates well with the relative inefficient resolution of DSBs in 6 month-old mice. Surprisingly, despite the reduced DNA damage response in this age it is accompanied with a robust early (2 days after DEN) and prolonged p53 signaling response.

Animal model

Experimental protocol was approved by the Hebrew University Institutional Animal Care and Ethical Committee. For the induction of DNA damage in vivo, 1,3,6 and 12-month-old C57BL/6 male mice (Harlan, Israel) were injected intraperitoneally (i.p.) with DEN (25 mg/g body wt; Sigma-Aldrich; St. Louis, MO). Mice were sacrificed at the indicated time points and liver tissues were fixed with 4% PFA or snap frozen in liquid nitrogen.

Immunofluorescence

Livers were fixed in 4% formaldehyde, dehydrated, embedded in paraffin, and sectioned (5 μm-thick). To estimate DNA damage, sections were immunofluorescence stained with anti-phospho-histone H2AX (γH2AX) monoclonal antibody (Millipore, Billerica, MA) and rabbit anti-phospho-KAP-1 (s824; Bethyl Laboratories, Montgomery, TX), followed by the secondary antibodies Alexa-488 goat anti mouse IgG (Molecular Probes, Eugene, OR) and Alexa-647 donkey anti rabbit IgG (Jackson ImmunoResearch,West Grove, PA). Nuclei were counterstained with DAPI (Calbiochem, Darmstadt, Germany). An Olympus BX61 microscope (Olympus, Tokyo, Japan) was used for low field image acquisition and a laser scanning confocal microscope system (FluoView-1000; Olympus) with a 40X UPLAN-SApo objective and 2x digital zoo for high field acquisition. Positive γH2AX areas in mice livers were calculated using ImageJ software.

RNA extraction

Total RNA was isolated from liver samples using Trizol reagent (Life Technologies, Paisley, UK) and purified using MaXtract High Density (QIAGEN, Redwood City, CA). RNA yield and quantity was determined using a Nanodrop spectrophotometer ND-1000 (Thermo scientific, Wilmington, DE). The RNA quality was tested using a ND-1000 V3.7.1, according to the manufacturer’s instructions and assigned an RNA integrity number (RIN). Only samples with a RNA integrity number (RIN) of 8 or greater were employed.

RNAseq and bioinformatic analysis

Each sample represents equal amount of total RNA from three mice that were pooled prior to library construction. RNAseq libraries were constructed in the center for Genomic Technologies using Trueseq RNA Library preparation kit according to Illumina protocol and sequenced with the Illumina Nextseq 500 System to an average depth of around 30 million reads. The bioinformatics’ analyses were performed in the Bioinformatics Unit of the I-CORE Computation Center at the Hebrew University and Hadassah. The NextSeq base-calls files were converted to fastq files using the bcl2fastq (v2.17.1.14) program. The processed fastq files were mapped to the mouse transcriptome and genome using TopHat (v2.0.13). The genome version was GRCm38, with annotations from Ensembl release 84. Normalization and differential expression were performed with the DESeq2 package (version 1.10.1). Differential expression was calculated using a design, which included the age factor, the days post-treatment factor and the interaction between them, compared with a reduced model that lacked the interaction term, and using the LRT test (all other parameters were kept at their defaults). The significance threshold for all comparisons was taken as the padj<0.1. Results were then combined with gene details (such as symbol, Entrez accession, etc.) taken from the results of a BioMart query (Ensembl, release 84) to generate a final Excel file. The ClueGO application (http://www.cytoscape.org) was applied to gene lists that were significantly differentially expressed. ClueGO visualizes the non-redundant biological terms for large clusters of genes in a functionally grouped network and performs network visualization of biological function. Gene ontology of biological process (July 2016) was used. Analysis for enriched pathways, upstream regulators and networks were also performed using QIAGEN’s Ingenuity® Pathway Analysis (IPA®, QIAGEN, http://www.qiagen.com/ingenuity). In addition a protein interaction network was constructed for the differentially expressed genes of the various comparisons. Since protein-protein interaction data is about 10-fold larger for human than for mouse (based on BioGRID statistics, http://wiki.thebiogrid.org/doku.php/statistics), we built the network using the human orthologs. Protein interaction data was extracted from the Human Integrated Protein-Protein Interaction Reference (HIPPIE 2.0, June 2016 [40];). The network was visualized using Cytoscape.

Validation of RNAseq Results by quantitative Real-time PCR (qRT-PCR)

To validate the results of RNAseq analysis, we confirmed by qRT-PCR the differential expression of some representative genes. Total RNA was reverse transcribed to cDNA using oligo(dT) primers and M-MLV reverse transcriptase (Thermo scientific). We used PerfeCTa SYBR Green FastMix ROX (Quanta Biosiences) for real-time PCR according to the manufacturer’s protocol and all the samples were run in triplicate on CFX384 Touch Real-Time system c1000 thermal cycler (Bio-Rad, Hercules, CA). Cycling conditions were 95°C for 20 sec, followed by 40 cycles of 95°C for 1 sec, and 60°C for 20sec, 65°C for 5sec. Gene expression levels were normalized to Hprt gene. Primers used for qPCR analysis can be found in Supplementary table 1.

Statistical analysis

All data were subjected to statistical analysis using the Excel software package (Microsoft) or GraphPad Prism6 (GraphPad Software Inc., La Jolla, CA, USA). Two-tailed Student’s t- test was used to determine the difference between the groups. Data are given as mean ± SD or ±SEM, and are shown as error bars for all experiments. Differences were considered significant at P < 0.05.