Article Text

Download PDFPDF

Original article
Gut microbiota shape ‘inflamm-ageing’ cytokines and account for age-dependent decline in DNA damage repair
  1. Avital Guedj1,
  2. Yael Volman1,
  3. Anat Geiger-Maor1,
  4. Julia Bolik2,
  5. Neele Schumacher2,
  6. Sven Künzel3,
  7. John F Baines3,4,
  8. Yuval Nevo5,
  9. Sharona Elgavish5,
  10. Eithan Galun1,
  11. Hagai Amsalem6,
  12. Dirk Schmidt-Arras2,
  13. Jacob Rachmilewitz1
  1. 1 Goldyne Savad Institute of Gene Therapy, Hadassah Medical Center, Hebrew University of Jerusalem, Jerusalem, Israel
  2. 2 Institute of Biochemistry, Kiel University, Kiel, Germany
  3. 3 Institute for Evolutionary Biology, Max Planck, Plön, Germany
  4. 4 Institute for Experimental Medicine, Kiel University, Kiel, Germany
  5. 5 Bioinformatics Unit of the I-CORE Computation Center, The Hebrew University and Hadassah Hebrew University Medical Center, Jerusalem, Israel
  6. 6 Department of Obstetrics and Gynecology, Hadassah University Hospital-Mount Scopus, Jerusalem, Israel
  1. Correspondence to Professor Jacob Rachmilewitz, Goldyne Savad Institute of Gene Therapy, Hadassah Medical Center, Hebrew University of Jerusalem, Jerusalem 91120, Israel; rjacob444{at}gmail.com

Abstract

Objective Failing to properly repair damaged DNA drives the ageing process. Furthermore, age-related inflammation contributes to the manifestation of ageing. Recently, we demonstrated that the efficiency of repair of diethylnitrosamine (DEN)-induced double-strand breaks (DSBs) rapidly declines with age. We therefore hypothesised that with age, the decline in DNA damage repair stems from age-related inflammation.

Design We used DEN-induced DNA damage in mouse livers and compared the efficiency of their resolution in different ages and following various permutations aimed at manipulating the liver age-related inflammation.

Results We found that age-related deregulation of innate immunity was linked to altered gut microbiota. Consequently, antibiotic treatment, MyD88 ablation or germ-free mice had reduced cytokine expression and improved DSBs rejoining in 6-month-old mice. In contrast, feeding young mice with a high-fat diet enhanced inflammation and facilitated the decline in DSBs repair. This latter effect was reversed by antibiotic treatment. Kupffer cell replenishment or their inactivation with gadolinium chloride reduced proinflammatory cytokine expression and reversed the decline in DSBs repair. The addition of proinflammatory cytokines ablated DSBs rejoining mediated by macrophage-derived heparin-binding epidermal growth factor-like growth factor.

Conclusions Taken together, our results reveal a previously unrecognised link between commensal bacteria-induced inflammation that results in age-dependent decline in DNA damage repair. Importantly, the present study support the notion of a cell non-autonomous mechanism for age-related decline in DNA damage repair that is based on the presence of ‘inflamm-ageing’ cytokines in the tissue microenvironment, rather than an intrinsic cellular deficiency in the DNA repair machinery.

  • liver
  • DNA damage
  • kupffer cell
  • inflammation
  • ageing
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Age is associated with accumulation of unrepaired DNA breaks that leads to increased frequency of mutations and genomic instability that can accelerate ageing and age-related diseases.

  • Age is associated with reduced DNA damage repair (DDR) capacity in ‘aged’ cells due to changes in the DNA repair machinery.

  • Age is associated with changes in the composition of the gut microbiota in both human and mice.

  • Age-related changes in the gut microbiota result in the induction of chronic, low tone inflammatory response known as ‘inflamm-ageing’.

What are the new findings?

  • Increase in inflamm-ageing cytokines coincides with the decline in DDR already at a relatively early age, these two intriguing correlates of early ageing are linked, that is, the decline in DDR stem from inflammatory ageing.

  • Age-related DDR is improved via antibiotic treatment, MyD88 ablation and in germ-free mice.

  • High-fat diet induces ‘inflamm-ageing’ and premature decline in DDR.

  • Proinflammatory cytokines interfere with double-strand breaks (DSBs) rejoining.

Significance of this study

How might it impact on clinical practice in the foreseeable future?

  • Our data highlight the dynamic interrelationship between tissue inflammatory ‘tone’ and the efficiency of DNA damage repairDDR that are particularly relevant in the context of the tumortumour microenvironment. Chemotherapy and ionizingionising radiation therapy are based on induction of cytotoxic levels of DNA DSBs in the tumortumour and are expected to improve in tumorstumours with DDR impairment. We suggest that the immunosuppressive tumortumour microenvironment creates a major barrier for both chemotherapy and radiation therapy as it allows for better damage resolution, and can be a target for therapeutic intervention.

Introduction

Ageing is characterised by progressive loss of many functions at the cellular and tissue level leading to organismal decline. Age-dependent decline in DNA damage repair and accumulation of damaged DNA are likely critical contributors to the manifestation of ageing (reviewed by Hoeijmakers1). The presence of double-strand breaks (DSBs) alone can drive the ageing phenotype and can account for ageing-associated fitness decline.2 DNA damage, specifically in the form of DSBs, can lead to mutations, chromosomal aberrations and eventually loss of genome integrity. Lesions in the double helix activate elaborate processes collectively termed DNA damage response (DDR) that repair the DNA lesions and simultaneously lead to transient cell cycle arrest, or in the case of heavily damaged and unrepairable genome, to cellular senescence or apoptosis.3

A tacit assumption in this field has been that ‘aged’ cells have an inherent imperfection in the DNA repair system and are defective in their DNA damage capabilities due to reduced expression and/or function of various components of the DNA repair machinery.4 5 However, our recent study6 demonstrated that the decline in DNA damage repair in the tissue is evident at relatively early age, before intrinsic DNA damage repair deficiencies are documented in ‘aged’ cells. Moreover, given that this early age decline in DDR precedes ‘old age’, it may be a driving force of the ageing process rather than merely a phenotypic consequence of old age. Yet, the mechanism underlying the decline in DDR at early age is not yet known.

Several studies demonstrate that the composition of the gut microbiota gradually changes with age and eventually the gut microbiota of older people and mice differ from that of younger adults.7 8 In turn, these age-dependent changes in gut microbiota, termed microbial dysbiosis, impact the ageing process.9 For example, loss of microbiota diversity is associated with impairment in cognitive performance, reduced health and overall increased frailty on ageing.10 11 These changes might be attenuated by probiotics that, at least in part, have been successfully used in elderly people for the treatment of various age-related phenotypes.9

Moreover, these age-related changes in the gut microbiota result in ‘inflamm-ageing’, whereby with age the innate immune system becomes dysregulated and is characterised by a chronic and systemic low tone inflammatory response (reviewed by Shaw et al 12). Inflamm-ageing is associated with several clinical syndromes that typically occur in older adults, including cachexia, frailty, neurodegenerative diseases, dementia, depression and cancer.13 14

Given the above-mentioned literature, linking age-dependent microbial dysbiosis and chronic inflammation to ageing, in the present study, we tested whether inflammatory ageing may provide a mechanism that compromises DNA damage repair efficiency.

Methods

Animal model

Induction of DNA damage in vivo was performed as previously described.6 15 For more details see online supplementary methods.

Immunohistochemistry and immunofluorescence

Immunohistochemical analysis of liver tissues was performed as described previously.6 15 For more details see online supplementary methods.

RNAseq and gene expression analysis

RNA extraction, RNAseq analysis, bioinformatics and quantitative PCR (qPCR) assay were performed as described previously.6 For more details see online supplementary methods. The GEO accession numbers for RNAseq data sets reported in this paper are GSE103069 and GSE103200. Primers used for qPCR analysis are listed in the online supplementary table S1.

Faecal collection, DNA extraction and 16S rRNA qPCR

Fresh faecal pellets were collected from individual mouse and were immediately snap frozen in liquid nitrogen and stored at −80°C. Bacterial DNA was extracted using the PowerLyzer PowerSoil DNA Isolation Kit according to the manufacturer’s protocols. The concentration of DNA was measured using a Nanodrop spectrophotometer ND-1000 (Thermo scientific, Wilmington, Delaware, USA). 16S rRNA gene was amplified by qPCR in triplicates, using previously published primers targeting specific bacterial species, genera and phyla (online supplementary table S2). Relative concentrations of specific bacterial 16S rRNA were normalised to the total amount of faecal bacteria amplified with 16S rRNA gene-based universal primers.

Intestinal permeability assay

Intestinal permeability studies using tracer fluorescein isothiocyanate-labelled dextran (4 kDa; Sigma-Aldrich) were performed as described previously.16

Cells

Primary normal human dermal fibroblasts (HF) were provided by the Department of Surgery, Hadassah Hebrew-University Medical Center, Jerusalem, Israel. The in vitro experimental protocol is adapted from our previous work.15 For more details, see online supplementary methods.

Statistical analysis

Data are expressed as means±SEM. Differences were analysed by Student’s t-test or analysis of variance, where appropriate. P values ≤0.05 were considered significant.

Results

Age-dependent deregulation of innate immunity in the liver

We have recently demonstrated that the efficiency of diethylnitrosamine (DEN)-induced DNA damage repair declines at the relatively early age of 6 months.6 DEN is metabolically activated by cytochrome P450-enzymes (predominantly CYP2E1) into metabolites that alkylate DNA.17 Previously, we demonstrated that the expression levels of CYP2E1 or other cytochrome P450 did not significantly change between mice at various ages, and there was no significant difference in the extent of initial damage noted in the liver 2 days after DEN injection in mice of all ages.6 In addition, to determine DEN clearance, we performed mass-spectrometry analysis for livers of 1-month-old mice and 6-month-old mice following DEN injection. After 5 hours, an average of 3.357±2.03 ng and 0.338±0.16 ng DEN is detected in 1-month-old mice and 6-month-old mice, respectively. Significantly, in both 1-month-old mice and 6-month-old mice livers, DEN was not detected 24 hours after injection, ruling out the possibility that metabolism of the drug or its clearance is reduced in aged mice. The finding that DEN is cleared from both young and aged mouse livers within 24 hours imply that in aged mice DNA damage repair is reduced17 rather than DNA damage is ongoing.

To look for age-related changes in the liver that may account for this early age decline in DNA damage resolution, we used RNAseq gene expression analysis to characterise mRNA expression profile in livers from 1-month-old mice versus 6-month-old mice (the earliest age where we noted a decline in DNA damage repair) either before or 2 days (representing the peak of DDR) and 6 days (representing the resolution phase of DNA damage) after DEN treatment. While we have previously detected significant changes in the response to DEN between the two ages in mice,6 we also observed that most of the differentially expressed genes were age related irrespective of the DEN treatment (figure 1A–B). Canonical pathway analysis revealed age-dependent upregulation of genes related to inflammation, antigen presentation and adaptive immune responses (online supplementary figure S1A) and downregulation of genes that were mainly related to cholesterol, fatty acid and glucose homeostasis, most prominently downregulation of the liver X receptors and retinoid X receptors activation pathway (not shown).

Figure 1

Age-dependent deregulation of innate immunity in the liver. (A) Differential transcript expression scatter plot depicting mean counts in control livers of 1-month-old versus 6-month-old mice (n=9 for each age). Significant differentially expressed transcripts with padj<0.1 are shown (red for upregulated and green for downregulated transcripts). In general, more genes were upregulated than downregulated (416 and 313 transcripts, respectively) in the older liver. (B) Heat map of hierarchical clustering analysis of significantly enriched genes in 1-month-old compared with 6-month-old mouse livers. (C) Upstream regulatory analysis revealed a strong enrichment for genes regulated by lipopolysaccharide, toll-like receptors, TNFα, IL-1β, nuclear factor kappa-light-chain-enhancer of activated B cells and other immune pathways (highlighted in yellow) that are activated in 6-month-old mice. The total number of target genes is indicated in parentheses in each row. Selected genes common to most regulators are specified. (D) Quantitative reverse transcription PCR analysis of TNFα, IL-1β and RANTES expression in RNA extracts from frozen steady state liver tissues harvested from mice at the indicated age (n=4 for each age). One of two independent experiments is shown. (E) Representative immunofluorescence images showing TNFα or IL-1β (red) and F4/80 (green; right panels) with DAPI counterstain (blue) of paraffin-embedded liver sections at the indicated ages. Right panels demonstrate the expression of TNFα and IL-1β by F4/80-positive cells (n=3 for each age). Graphs at lower panels show quantitation of the number of cytokine-expressing cells (average±SD). *P<0.05; **P<0.005. DAPI, 4′,6-diamidino-2-phenylindole; IL-1β, interleukin-1 beta; TNFα, tumour necrosis factor alpha.

In addition, we performed upstream regulator analysis of differentially expressed genes and identified the bacterial product, lipopolysaccharide (LPS) as the top upstream regulator. Of the top 30 regulators, 14 relate to several toll-like receptors (TLR 3, 4, 9 and the adaptor protein MyD88) and their ligands as well as components of the downstream transcription factor, nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) (figure 1C). Additionally, two key innate cytokines, tumour necrosis factor alpha (TNFα) and interleukin-1 beta (IL-1β) were identified as upstream regulators (figure 1C) and are also differentially expressed genes themselves, as confirmed by quantitative reverse transcription-PCR analysis (figure 1D). We further extended this analysis to mice at the age of 3 months, representing fully mature young mice that in our previous study demonstrated similar DNA damage repair capacity as 1-month-old mice. The elevated inflammatory response seen in 6-month-old mice extended to 12 months of age. We also validated the expression of RANTES, a chemokine that was highly upregulated with age (figure 1D). Immunofluorescence staining for either TNFα or IL-1β established an age-dependent increase in the expression of those cytokines in the liver (figure 1E) and demonstrated an almost exclusive expression by resident Kupffer cells (figure 1E). Moreover, high levels of TNFα and IL-1β transcripts were detected in macrophages purified from mouse livers before and after DEN injection (online supplementary figure S1B), demonstrating that resident Kupffer cells predominantly express these cytokines. Of note, the overall levels of F4/80 and CD11b expression and the number of Kupffer cells was only slightly increased with age (online supplementary figure S1C,D), and the expression of the chemokine CCL2 that mediates monocytes/macrophage recruitment, increased moderately with age (online supplementary figure S1C).

Taken together, the present study and previous work by others18 19 reveals that the most prevalent age-related phenotype in the liver is chronic inflammation with elevated levels of basal proinflammatory cytokines. This increased low grade but chronic inflammation has long been associated with ageing and was termed ‘inflamm-ageing’ (reviewed by Shaw et al 12).

Given the two intriguing correlates of early ageing: the phenomenon of ‘inflamm-ageing’ and the reduced DNA damage repair capacity at the age of 6 months,6 we postulated that the two may be linked, that is, that the decline in DDR may stem from inflammatory ageing.

Oral administration of antibiotics modulates cytokine expression and improves DNA damage resolution in DEN-treated mouse livers

The gut microbiota shapes the immune system by providing tonic immune stimulatory signals. These signals are important for ‘priming’ the peripheral innate immune system and inducing elevated levels of inflammatory cytokines and drive age-associated inflammation,20 both locally at the GI mucosa and systemically at distant tissues.20–23 This is specifically true for the liver that serves as the portal for the soluble factors exported from the GI tract and thus senses the intestinal microenvironment.

Several independent studies have demonstrated that the composition of gut microbiota gradually changes as a function of age in both humans10 24 25 and mice8 26 in what is termed ‘microbiota dysbiosis’. Using 16S ribosomal RNA (rRNA) gene sequencing, Kulecka et al, detected changes in the composition of gut bacteria already in 6-month-old as compared with 3-month-old C57BL6 mice.26 Based on this study, we further corroborated changes in gut microbiota composition in 6-month-old mice using qPCR (online supplementary figure S2A).

Apart from changes in the microbiota composition, impaired barrier function of the gut epithelium observed in aged individuals16 27 may lead to bacterial components, such as LPS, to cross the epithelial barrier and induce inflammation (reviewed by Nicoletti28). We tested mice at ages 1, 3 and 6 months and demonstrated that at the age of 6 months the barrier function already decreased when compared with younger mice (online supplementary figure S2B).

To test whether gut microbiota may modulate DDR as consequence of its effect on innate immune functions, 6-month-old mice received an antibiotic cocktail (ABX)23 for 2 weeks prior to DEN injection. Mice treated with this antibiotic regime exhibited impaired innate immunity and specifically lower TNF production by myeloid cells.21 23 Antibiotic activity was analysed by macroscopic changes observed at the level of swelling of the colon and reduced bacteria load as revealed by cultivating the faecal pellets on agar plates (online supplementary figure S2C). In accordance with the notion that commensal-derived signals primes inflammatory cytokine production21 23 and with the reported reduction in gut permeability on antibiotic treatment,29 the level of liver TNFα, IL-1β and RANTES were decreased in ABX-treated 6-month-old mice as compared with control age-matched mice but were comparable to 1-month-old mice (figure 2A and online supplementary figure S2D).

Figure 2

Gut microbiota reduce DNA damage repair and modulate cytokine expression Six-month-old mice received either water (CNT) or an ABX of vancomycin, imipenem and neomycin in drinking water for 2 weeks and continuing throughout the experiment. (A) Quantitative reverse transcription-PCR analysis of TNFα, IL-1β and RANTES expression in RNA extracts from frozen steady state (before DEN injection) liver tissues harvested from 1-month-old mice, 6-month-old mice given H2O (CNT) or ABX and 6-month-old MyD88−/− mice. An average of three mice in each group is shown. One of two independent experiments is shown. (B) Representative images of residual γH2AX staining of liver tissue sections 6 days after DEN treatment of 1-month-old mouse, 6-month-old control mouse and ABX-treated mouse or 6-month-old MyD88−/− mouse are shown (×4 objective). (C–D) Graphs show γH2AX staining areas 2 and 6 days after DEN injections in CNT, ABX-treated and in 6-month-old MyD88−/− mice (average±SD). The percentage of DNA damage repair was calculated by dividing the amount of DNA damage at the peak of damage after 48 hours and the residual damage after the resolution phase 6 days after DEN treatment. The percentage of residual damage in 6-month-old CNT mice compared with ABX-treated mice and that of WT versus MyD88−/− mice are shown (n=4 for each group). One of two independent experiments is shown for ABX and MyD88 experiments. *P<0.05; ***P<0.0005; ****P<0.0001. ABX, antibiotic cocktail; CNT, control; DEN, diethylnitrosamine; IL-1β, interleukin-1 beta; TNFα, tumour necrosis factor alpha; WT, wild type; γH2AX, phosphorylated histone H2AX.

To test the efficiency and kinetics of DNA damage repair, we induced DNA damage in mouse livers by a single injection of DEN, as previously described.6 15 We performed immunofluorescence staining to detect cells containing phosphorylated histone H2AX (γH2AX) in liver tissue sections at days 2 and 6 following DEN injection. As was previously shown,6 15 DSBs resolution in 6-month-old mice was impaired and significant numbers of residual γH2AX foci were detected in livers 6 days following DEN treatment. In contrast, while there were no significant differences in the extent of initial damage in the livers 2 days after DEN injection in ABX-treated mice, the extent of residual DSBs 6 days after DEN injection was significantly lower as compared with control mice that were treated with H2O2 and equivalent to that in 1-month-old mice (figure 2B–C). These findings demonstrate that antibiotic treatment restores DNA damage repair capacity in 6-month-old mice and in parallel reduces inflammatory cytokine expression.

We further corroborated γH2AX-staining findings by demonstrating that KRAB-associated protein (KAP-1) phosphorylated on serine 824 forms foci overlapping with γH2AX at sites of DSBs, indicating active ongoing DNA-damage repair (White et al and Ziv et al 30 31; online supplementary figure S2E).

Bacterial products are recognised by pattern recognition receptors, TLRs and the downstream adaptor protein MyD88 under normal steady-state conditions32–34 and, as suggested, induce chronic ‘low-grade’ inflammation.21 23 To strengthen the link between increased inflammatory tone, as a result of bacteria sensing and age-dependent decline in DDR, we next recapitulated the above experiments using 6-month-old mice with systemic ablation of MyD88.

MyD88 deficiency led, as expected, to lower TNFα, IL-1β and RANTES expression (figure 2A and online supplementary figure S2F). Importantly, MyD88 deficiency resulted in reduced residual γH2AX in DEN-treated livers in 6-month-old mice compared with WT mice (figure 2B,C).

Taken together, these findings suggest a link between commensal-derived signals sensed by TLRs, inflammation and DNA damage repair activity and suggest that the decline in DNA damage repair can be reversed by physiological manipulations.

Germ-free (GF) mice have reduced cytokine expression and improved DNA damage resolution

To further study the relationship between gut microbiota, DDR and innate immune functions, we used GF mice. As expected, no bacteria were detected in the GF mice faecal pellets cultivated on agar plates (online supplementary figure S3A). Previous reports demonstrated lower age-related inflammation in GF mice.16 In agreement with these findings, GF mice expressed lower levels of TNFα and IL-1β as compared with their specific pathogen-free (SPF) counterparts (figure 3A), with no difference in the number of F4/80+ cells (online supplementary figure S3B). Importantly, 6 days after DEN treatment, GF mouse livers had significantly reduced residual γH2AX foci as compared with SPF mice (figure 3B).

Figure 3

GF mice have markedly reduced cytokine expression and improved DNA damage resolution. (A) Representative immunofluorescence images showing TNFα or IL-1β (red) with DAPI counterstain (blue) of paraffin-embedded liver sections of the indicated groups. Graphs on the right show quantitation of the number of cytokine expressing cells (n=5; average±SD). (B) Representative images of residual γH2AX staining of liver tissue sections before (ctrl) and 2 and 6 days after DEN treatment of 6-month-old SPF and GF mice (×4 objective). (C) Graphs show residual γH2AX staining areas 6 days after DEN injections in control (SPF) and GF mice (n=5 for each group; average±SD). *P<0.05. DAPI, 4′,6-diamidino-2-phenylindole; GF, germ free; IL-1β, interleukin-1 beta; SPF, specific pathogen free; TNFα, tumour necrosis factor alpha; γH2AX, phosphorylated histone H2AX.

We further performed RNAseq analysis on RNA isolated from liver tissues of SPF and GF mice. Principle component analysis reveals separate grouping of SPF and GF mice regardless of DEN treatment (online supplementary figure S3C). Similarly, analysis for conserved transcription factor binding sites revealed that the NF-kB pathway is over-represented in SPF mice at all time points (online supplementary figure S3D). Moreover, the transcription factors Kruppel-Like Factor 4, Early Growth response 1, SP1 and TP53, that, among others, play a role in cell cycle regulation and response to DNA damage, were also over-represented in our differentially expressed gene sets (online supplementary figure S3D).

Gene set enrichment analysis of differentially expressed genes revealed that genes linked to inflammation are enriched in SPF mice and TNF receptor 1 (TNFR1)-induced proapoptotic signalling, regulation of TNFR1 signalling and TNFR1-induced NF-kB signalling pathway were significantly downregulated in GF mice 2 days after DEN treatment (online supplementary figure S3E). Moreover, at day 6 after DEN injection, genes associated with cell cycle, and cell cycle checkpoints and DNA damage repair as well as cellular senescence are significantly enriched in SPF as compared with GF mice (online supplementary figure S3F). This ongoing gene expression (at day 6) of genes associated with cell cycle regulation and DNA damage repair is in accordance with the reduced damage resolution capacity observed in SPF mice.

Taken together, these findings reinforce the link between commensal-derived signals, inflammation and DNA damage repair activity.

High-fat diet (HFD) increases gut microbiota load, elevates cytokine expression and delays DNA damage resolution in 3-month old mice

We next intended to accelerate the decline in DNA damage repair in young mice. Ingestion of HFD was shown to alter gut microbiota composition, increase intestinal permeability and plasma LPS concentrations that, in turn, induce chronic ‘low-grade’ systemic inflammation.35–37

We, therefore, hypothesised that HFD may increase inflammation in the liver and accelerate age-dependent decline in DDR. To test our hypothesis, 1-month-old mice were fed a normal diet (ND) or HFD (60% kcal fat) for 9 weeks (figure 4A). Cultivating the faecal pellets on agar plates revealed an increase in bacterial load in HFD compared with ND (online supplementary figure S4A). HFD mice progressively gained significantly more weight than ND mice (figure 4B), accompanied with substantial accumulation of lipids in the liver (figure 4C) and upregulation of TNFα and IL-1β expression (figure 4D), but with no significant effect on the number of F4/80+ cells (not shown). In order to confirm a role for gut microbiota, a control group received ABX treatment for the last 3 weeks of the HFD feeding (figure 4A). As expected ABX treatment significantly reduced faecal bacterial load (online supplementary figure S4A) and downregulated cytokine expression (figure 4D). Weight gain and lipid accumulation were slightly reduced on ABX treatment compared with HFD (figure 4B–C).

Figure 4

HFD elevates proinflammatory cytokine expression and accelerates age-dependent decline in DNA damage repair in a microbiota dependent manner. (A) Timeline of the experimental procedure. One-month-old mice received either normal diet (ND; n=8) or HFD (60% kcal fat; n=16) for 9 weeks. Half of the mice on HFD received antibiotics for the last 3 weeks (HFD+ABX). (B) Graph depicting weight gain during the 9 weeks of HFD (ND: n=8; HFD: n=8; HFD+ABX: n=8). (C) Representative images of H&E staining of liver sections demonstrating lipid accumulation in HFD mice. (D) Representative immunofluorescence images showing TNFα or IL-1β (red) with DAPI counterstain (blue) of paraffin-embedded liver sections of the indicated groups. (E) Representative images of γH2AX staining of liver tissue sections 2 and 6 days after DEN treatment of the various mouse groups (×4 objective). ABX, antibiotic cocktail; DAPI, 4′,6-diamidino-2-phenylindole; HFD, high-fat diet; IL-1β, interleukin-1 beta; ND, normal diet; TNFα, tumour necrosis factor alpha; γH2AX, phosphorylated histone H2AX.

Finally, mice were injected with DEN and the level of DNA damage was tested. The resolution of foci containing γH2AX (as well as pKAP-1; online supplementary figure S4D) 6 days after DEN treatment was considerably impaired in HFD as compared with ND, but the effect of HFD was completely reversed by ABX treatment (figure 4E; online supplementary figure S4D).

Interestingly, these results not only resemble that of 6-month-old mice but also are reminiscent of macrophage-depleted 1-month-old mice or those treated with the inhibitor (CRM197) that blocks the activity of macrophage-derived heparin-binding epidermal growth factor-like growth factor (HB-EGF),38 a factor we have demonstrated plays a role in facilitating DNA damage resolution.15 Hence, reinforcing the notion that the age-related decline is not cell intrinsic and may stem from a deficiency in the systemic regulation of DDR by macrophages.

Kupffer cell replenishment or inactivation by gadolinium chloride (GdCl3) improves DNA damage resolution in 6-month old mice

Commensal gut microbiota was shown to mediate its effects mainly by modulating myeloid-derived cell functions.21 23 In addition, our data indicate that Kupffer cells are the major source of both TNFα and IL-1β in the liver (figure 1E and online supplementary figure S1B). On the other hand, immunohistochemistry staining suggest that it is unlikely that T cells play a major role in facilitating DNA damage repair (online supplementary figure S5A). Hence, we hypothesised that in situ production of proinflammatory cytokines by Kupffer cells may be responsible for ablating the DDR in the liver.

Depletion of Kupffer cells could not be applied since we have recently demonstrated that macrophages and Kupffer cells are necessary for efficient DNA damage repair and that depletion of macrophages by clodronate liposomes abrogated DEN-induced DNA damage repair.15 Instead, we attempted to replace ‘inflammatory’ macrophages with fresh ‘inactivated’ ones using what we termed a ‘reset’ experiment. In this experiment, mice were treated with clodronate-liposomes in order to deplete peripheral macrophages and Kupffer cells. After a recovery period of 3–4 weeks, the time required for macrophages and Kupffer cell numbers to fully restore (data not shown), mice were treated with DEN and DNA damage repair was monitored (figure 5A). Replenished cells expressed lower levels of TNFα and IL-1β as compared with Kupffer cells in control mice (figure 5B). Notably, in clodronate-treated mice, DNA damage resolution 6 days after DEN injection was significantly improved as compared with control mice that were treated with phosphate buffered saline liposomes (figure 5C).

Figure 5

Age-dependent decline in DNA damage repair and cytokine expression are reversed by Kupffer cell replenishment or inactivation by gadolinium chloride in 6-month-old mice. (A) Timeline of the experimental procedure. To deplete liver Kupffer cells, 6-month-old mice received two injections of clodronate-loaded liposomes. PBS-loaded liposomes were used as control. The mice were kept for 4 weeks (the time required for macrophages to repopulate the liver) and then were injected with DEN and the extent of DNA damage was determined by γH2AX staining. F4/80 immunostaining in the lower panels were used to assess macrophage depletion and recovery. (B) Representative immunofluorescence images showing TNFα or IL-1β (red) with DAPI counterstain (blue) of paraffin-embedded liver sections of mice 4 weeks after either PBS or clodronate treatment. Graphs on the right show quantitation of the number of cytokine-expressing cells (n=4; average±SD). (C) Graphs show γH2AX staining areas 6 days after DEN injections in PBS and clodronate-treated (Clod) 6-month-old mice. An average (±SD) of four mice in each group is shown. Similar results were obtained in two independent experiments. (D) Mice were treated with gadolinium a day before DEN treatment. The number of F4/80 cells was not affected by gadolinium treatment as determined by immunostaining. (E) Graphs show γH2AX staining areas 2 and 6 days after DEN injections in 6-month-old gadolinium-treated mice. The percentage of residual damage in 6-month-old CNT mice compared with gadolinium-treated mice is shown. (F) Quantitative reverse transcription-PCR analysis of TNFα, IL-1β, RANTES and HB-EGF expression in RNA extracts from frozen liver tissues harvested from 1-month-old mice and 6-month-old mice untreated or treated with gadolinium (Gadol). An average (±SD) of four mice in each group is shown. One of two independent experiments is shown. *P<0.05; **P<0.005. CNT, control; DAPI, 4′,6-diamidino-2-phenylindole; DEN, diethylnitrosamine; HB-EGF, heparin-binding epidermal growth factor-like growth factor; IL-1β, interleukin-1 beta; PBS, phosphate buffered saline; TNFα, tumour necrosis factor alpha.

Given the macrophage ‘reset’ finding, we postulated that Kupffer cell replacement may not be necessary per se, but rather, it may be sufficient to simply down-modulate their proinflammatory activity. For that end, mice were treated with GdCl3, which selectively inactivates Kupffer cells,39 without affecting their numbers (figure 5D). While depletion of Kupffer cells by clodronate-liposomes abrogated DNA damage repair,15 inactivation by GdCl3 had no adverse effect on DNA damage resolution in 1-month-old mice (not shown) and had no effect on HB-EGF expression (figure 5F), a factor we have previously shown to mediate macrophage-assisted DDR.15 Notably, in 6-month-old mice treated with GdCl3, the extent of residual DSBs 6 days after DEN injection was significantly lower as compared with control mice (figure 5E) and was accompanied with reduced expression of inflamm-ageing cytokines without effecting HB-EGF expression (figure 5F and online supplementary figure S5C).

To sum up, the data suggest a critical role for ‘inflamm-ageing’ cytokine production by Kupffer cells in the observed decline in DDR.

TNFα and IL-1β are detrimental to HB-EGF-assisted DNA damage repair

Having demonstrated a link between in situ production of proinflammatory cytokines and the decline in DDR, we postulated that the presence of proinflammatory cytokines in-and-of themselves may be sufficient for impairing DNA damage repair. We first examined the effect of IL-1β signalling deficiency in IL-1 receptor (IL-1R) knockout (KO) mice on DNA damage repair and found that it did not improve DNA damage resolution in 6-month-old mice (figure 6A). We next tested TNF KO mice and found that it was insufficient to improve DNA damage resolution (figure 6B). We validated that KO mice do not express TNFα but as expected revealed that these mice express a normal IL-1β level (figure 6C), raising the possibility that ablating only one cytokine at a time is not sufficient to rescue DNA damage repair.

Figure 6

TNFα and IL-1β are detrimental to HB-EGF-assisted DNA damage repair. (A–B) Graphs show γH2AX staining areas 2 and 6 days after DEN injections (left panels; average±SD) and the calculated percentage of residual damage (right panels) in 6-month-old IL-1R-KO (A) or TNF-KO (B) mice. An average of four (for IL-1R-KO) and five (for TNF-KO) mice in each group is shown. One of two experiments is shown. (C) Representative immunofluorescence images showing TNFα or IL-1β (red) with DAPI counterstain (blue) of paraffin-embedded liver sections of TNF-KO mice. (D) Schematic illustration of in vitro DNA damage repair experiment. HF grown on a cover slide were treated with 50 µM H2O2, and then were cultured for 24 hours in the presence or absence of HB-EGF (5 ng/mL) orTNFα (5 ng/mL), IL-1β (2 ng/mL), RANTES (10 ng/mL), IP10 (500 ng/mL), LPS (10 ng/mL) or the combination of the two. Then, HF were fixed and immunostained using anti-γH2AX and anti-53BP1 mAbs. Lower panel: representative confocal immunofluorescence image showing γH2AX foci (green), 53BP1 (red) with DAPI counterstain (blue). (E) Graph showing the number (mean±SEM) of γH2AX foci per cell treated as indicated (at least 50 nuclei counted for each treatment). One representative of three independent experiments is shown. ****P<0.0001. DAPI, 4′,6-diamidino-2-phenylindole; HB-EGF, heparin-binding epidermal growth factor-like growth factor; HF, human fibroblasts; IL-1β, interleukin-1 beta; IP10, interferon gamma-induced protein 10; KO, knockout; LPS, lipopolysaccharide; s, non-significant; TNFα, tumour necrosis factor alpha; WT, wild type.

To test whether each individual cytokine can interfere with DSBs resolution, we turned to an in vitro experimental system (figure 6D). In a previous study, we demonstrated that the high number of γH2AX foci observed in human fibroblasts (HF) following peroxide treatment significantly declined after 24 hours, leaving only few distinct foci in each cell. These distinct nuclear foci also displayed co-staining with 53BP1, an established player in the cellular response to DNA damage that is also recruited to sites of DSBs (figure 6D). Moreover, we have demonstrated that the average number of residual γH2AX foci per cell was significantly reduced in the presence of recombinant HB-EGF in the culture media.15 Therefore, we tested the effect of TNFα or IL-1β on the efficiency of DSBs resolution in the presence or absence of HB-EGF. Interestingly, while the addition of TNFα or IL-1β had no significant effect on the number of γH2AX foci per cell after 24 hours, it significantly abrogated the beneficial effect of HB-EGF (figure 6E). Similar results were obtained with the chemokine RANTES (figure 6E). In contrast, the addition of the chemokine IP-10 or LPS, used as control for a bacterial product that induces the expression of pro-inflammatory cytokines, had no detrimental effect on damage resolution in the presence of HB-EGF (figure 6E).

We suggest that whereas HB-EGF enhances DDR and facilitates DSBs rejoining,15 the presence of Kupffer cell-derived ‘inflamm-ageing’ cytokines interferes with HB-EGF-assisted DDR resulting in delayed DSBs resolution (online supplementary figure S6).

Discussion

The DDR is a cell-intrinsic signalling cascade that originates from the DNA lesion and coordinates DNA repair. Today’s research on ageing focuses on age-related changes in activation of key DNA repair pathways, such as intrinsic deficiencies in the expression and activity of the DDR machinery taking place in ‘aged’ cells.4 5 In the present study, we propose a cell non-autonomous mechanism for age-related decline in DNA damage repair that is based on the presence of ‘inflamm-ageing’ cytokines in the tissue microenvironment.

Our study proposes the following chain of events (that emerges in mice already at the age of 6 months): age-related changes in gut microbiota composition are accompanied with a decline in barrier function of the gut epithelium. These events result in increased leakiness of bacterial soluble factors that act as immune-activating signals.16 This is specifically evident in the liver that is directly exposed to the intestinal microenvironment through the portal system. Eventually, commensal bacteria-derived factors prime liver Kupffer cells for inflammatory cytokine production and induce low grade, chronic inflammation. In turn, the continuing presence of low-level cytokines in the tissue microenvironment during genotoxic injury accounts for the decline in DNA damage repair efficiency, as it is detrimental to HB-EGF-assisted DSBs rejoining (see proposed model in online supplementary figure S6).

Inflammation and cytokines can directly induce DNA damage, mainly via reactive oxygen species production. Therefore, it can be argued that ‘inflamm-ageing’ induce excess DNA damage rather than interfered with DNA damage repair. However, in our experiments, chronic low-tone ‘inflamm-ageing’ on its own or the cytokines in-and-of themselves (in in-vitro experiments) did not induce detectable DNA damage or enhance DEN/peroxide-induced DNA damage.

We showed that deliberate manipulation of environmental factors such as commensal bacteria or inflammation in the tissue microenvironment reversed the detrimental effects of ageing on the cellular response to DNA damage, reinforcing the importance of extrinsic pathways in this age-related phenomenon. Previous studies have similarly linked low grade, chronic inflammation with normal and pathological ageing. Baruch et al identified elevated expression of the type I interferon (IFN-I) pathway in the choroid plexus of old mice that underlie loss of function in the ageing brain.40 Jurk et al, induced low-grade inflammation and demonstrated that that-alone was sufficient to aggravate telomerase dysfunction and cell senescence and prompt premature ageing.41 Another study provided a direct link between constitutive productions of IL-1β as a result of NLRC4 inflammasome chronic activation and cardiovascular disease in aged individuals.42 Finally, flies with constitutively active NF-kB signalling suffered from extensive neurodegeneration and reduced lifespan, whereas flies with reduced levels of NF-kB resulted in improved function in old age and extended lifespan.43

Here we propose the existence of two paths in the DDR that are blunted during ageing. The first, the well-established intrinsic DDR machinery that as mentioned above may fail in ‘aged’ cells. The second is the extrinsic regulatory mechanism that is based on macrophage-derived HB-EGF-assisted DDR. This regulatory mechanism that enhances and facilitates DNA damage resolution is diminished at relatively early age once chronic inflammation is established as a result of age-dependent microbial dysbiosis. In the context of DNA damage and ageing, the cause and effect are intricately linked. On the one hand, with age, DNA damage repair declines and damaged DNA accumulates. On the other hand, DNA damage itself can drive and accelerate ageing and contribute to age-related diseases.44 Therefore, it is possible that the early age decline in DNA repair efficiency is a cause, and not merely a symptom, of ageing, and precedes the intrinsic deficiency in damage repair observed in ‘aged’ cells.

Our study reveals an unexpected link between diet, commensal bacteria, inflammation and the efficiency of DNA damage repair in ageing. Specifically, we demonstrated that HFD increases inflammatory cytokine expression and facilitates the decline in DNA damage repair. In contrast, there is considerable evidence that dietary calorie restriction extends the lifespan of many species probably by slowing down many age-dependent processes. Calorie restriction has been shown to exert anti-inflammatory effects preventing increase in ‘inflamm-ageing’ cytokines.45–47 Moreover, it has been demonstrated that calorie restriction reduces the rate of the age-dependent decrease in the DNA repair mechanisms, base excision repair, nucleotide excision repair (NER)48 49 as well as non-homologous end joining,50 thus leading to a better maintenance of genomic integrity. In contrast, mice deficient in the NER gene, Ercc1, show symptoms of accelerated ageing and have much shorter lifespans than wild-type siblings.51 Interestingly, the authors found that these mice have a median lifespan of 20 weeks surprisingly similar to the age of mice in which increased gut permeability, microbiota dysbiosis,26 increase in ‘inflamm-ageing’ cytokines and decline in DNA damage repair6 were observed. Strikingly, the average lifespan of these mice (as well as Xpg−/− mice, another NER-deficient mice) tripled on dietary restriction. These mice remained healthier and had reduced level of DNA damage as measured by the number of γH2AX foci, compared with their counterparts that eat freely.52

Based on our data, it is possible that Ercc-deficient mice, which lack important intrinsic DNA damage repair machinery, on reaching the age of 6 months and ‘inflamm-ageing’ is established further lose the ‘compensatory’ pathway that enhance DNA repair. Consequently, persistent DNA damage causes premature age-related disorders and a shortened lifespan. However, given that calorie restriction reduces the expression of ‘inflamm-ageing’ cytokines, the decline in DNA damage resolution is delayed resulting in reduced endogenous DNA damage and an extended lifespan.

In conclusion, we demonstrate for the first time a link between age-related microbial dysbiosis that disrupts cytokine networks resulting in impaired DNA damage repair function that ultimately compromises tissue integrity resulting in functional deterioration with age.

Acknowledgments

We thank Dr G Zamir from the Department of Surgery, Hadassah Hebrew-University Medical Center, Jerusalem, Israel, for providing the primary normal human dermal fibroblasts (HF). We thank Matthias Stechele for assisting in DEN clearance experiments and M Tykocinski for the critical discussion and reading of the manuscript.

References

View Abstract

Footnotes

  • Contributors JR and AG designed the research and analysed the data. JR wrote the manuscript. AG prepared the manuscript. YV and AG-M performed experiments. YN and SE analysed the RNAseq data. HA and EG contributed to experimental design and data analysis. SK and JFB provided germ-free mice, performed the experiments and RNA sequencing. DS-A supervised experiments, analysed RNAseq data and contributed to writing the manuscript.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval Experimental protocol was approved by the Hebrew University Institutional Animal Care and Ethical Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.