The SETD8/PR-Set7 Methyltransferase Functions as a Barrier to Prevent Senescence-Associated Metabolic Remodeling
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 1
The Loss of SETD8/PR-Set7 Induces Senescence
(A) Schematic representation of senescence models used in this study. SA-β-Gal staining of growing, oncogene-induced senescence (OIS), and replicative senescence (RS) cells are shown. Numbers indicate the percentage of SA-β-Gal-positive cells (each n > 300 cells). Scale bars, 100 μm.
(B) Western blot analysis of SETD8 and p16INK4A in growing, OIS, and RS cells.
(C) Growth curves for Ctr- or SETD8-KD IMR-90 cells.
(D) EdU incorporation assay on day 3 of Ctr- or SETD8-KD cells. EdU-positive and EdU-negative cells were distinguished by measuring the fluorescence intensity of EdU-labeled DNA in each cell (each n > 1,600 cells).
(E) SA-β-Gal staining on days 3–15 of Ctr- or SETD8-KD (each n > 300 cells). Scale bars, 100 μm.
(F) SAHF on days 3–15 of Ctr- or SETD8-KD (each n > 300 cells). Scale bars, 10 μm.
(G) Western blot analysis of SETD8, p16INK4A, and p21CDKN1A on days 3 and 6 of Ctr- or SETD8-KD.
(H) Growth curves for DMSO-treated or the SETD8 inhibitor UNC0379-treated IMR-90 cells.
(I) SA-β-Gal staining on day 6 of DMSO or UNC0379 treatment (each n > 300 cells). Scale bars, 100 μm.
(J) SAHF on day 6 of DMSO or UNC0379 treatment (each n > 300 cells). Scale bars, 10 μm.
Values are means ± SD (n = 3). p values were calculated using the Student’s t test (∗p < 0.05, ∗∗p < 0.01).
Highlights
- •
SETD8 protein is downregulated in senescent cells
- •
The loss of SETD8 triggers cellular senescence
- •
SETD8 represses p16INK4A and ribosome-associated genes through H4K20 methylation
- •
SETD8 regulates the structure and function of nucleoli and mitochondria
Summary
Cellular senescence is an irreversible growth arrest that contributes to development, tumor suppression, and age-related conditions. Senescent cells show active metabolism compared with proliferating cells, but the underlying mechanisms remain unclear. Here we show that the SETD8/PR-Set7 methyltransferase, which catalyzes mono-methylation of histone H4 at lysine 20 (H4K20me1), suppresses nucleolar and mitochondrial activities to prevent cellular senescence. SETD8 protein was selectively downregulated in both oncogene-induced and replicative senescence. Inhibition of SETD8 alone was sufficient to trigger senescence. Under these states, the expression of genes encoding ribosomal proteins (RPs) and ribosomal RNAs as well as the cyclin-dependent kinase (CDK) inhibitor p16INK4A was increased, with a corresponding reduction of H4K20me1 at each locus. As a result, the loss of SETD8 concurrently stimulated nucleolar function and retinoblastoma protein-mediated mitochondrial metabolism. In conclusion, our data demonstrate that SETD8 acts as a barrier to prevent cellular senescence through chromatin-mediated regulation of senescence-associated metabolic remodeling.
Introduction
Cellular senescence is induced by various cellular stresses such as oncogene expression, telomere attrition, and genome-scale perturbation of chromatin. It is characterized by irreversible cell cycle arrest, senescence-associated β-galactosidase (SA-β-Gal) activity, and the senescence-associated secretory phenotype (SASP), as well as alterations of gene expression and chromatin (Benayoun et al., 2015, Campisi and d’Adda di Fagagna, 2007, Kuilman et al., 2010). These changes in senescent cells contribute to tumor suppression, tissue repair, and developmental processes, as well as age-related deterioration of tissue functions in vivo (Baker et al., 2016, Muñoz-Espín and Serrano, 2014, van Deursen, 2014). The senescent cells also undergo metabolic remodeling as indicated by enlarged cell size and increased protein content. Various metabolic pathways, including protein synthesis and degradation, autophagy, glycolysis, and mitochondrial oxidative phosphorylation (OXPHOS), are essential for the establishment of senescence (Salama et al., 2014, Wiley and Campisi, 2016). However, it is not clear how senescent cells remodel their metabolic status in combination with other senescence-associated features.
Chromatin-modifying factors play a fundamental role in gene regulation and are involved in DNA methylation, histone modification, and the formation of higher-order chromatin structures. The epigenomic landscapes in senescent cells differ from those of proliferating cells (Chandra et al., 2012, Chicas et al., 2012, Criscione et al., 2016, Cruickshanks et al., 2013, Hirosue et al., 2012, O’Sullivan et al., 2010, Shah et al., 2013), indicating that chromatin regulators play an important role in establishing and maintaining the senescent state. Growing evidence suggests that there is a reciprocal relationship between epigenetic regulation and cellular metabolism (Gut and Verdin, 2013, Hino et al., 2012, Hino et al., 2013). Most chromatin-modifying enzymes use substrates or cofactors derived from various metabolites (such as S-adenosylmethionine for many methyltransferases), while biochemical reactions depend on coordinated expression of many enzyme-encoding genes in metabolic pathways (Desvergne et al., 2006). So far, we do not understand how epigenetic and metabolic mechanisms cooperate to establish cellular senescence.
SETD8, also known as PR-Set7 or SET8, is a nucleosome-specific methyltransferase that is responsible for mono-methylation of histone H4 lysine 20 (H4K20me1) (Nishioka et al., 2002). SETD8 is involved in various genomic functions including DNA replication, mitosis, DNA repair, and gene expression via H4K20 methylation (Beck et al., 2012, Jørgensen et al., 2013). The protein levels of SETD8 are precisely controlled through proteasomal degradation, resulting in the lowest level during the S phase and the highest level during mitosis, which ensures proper cell cycle progression. However, the precise role of SETD8 in gene expression is rather confusing. Genome-wide profiling indicated that H4K20me1 is predominantly enriched downstream of the transcription start site of actively transcribed genes (Barski et al., 2007, Beck et al., 2012). Indeed, the levels of H4K20me1 in gene bodies are positively correlated with gene expression levels during cell differentiation (Cui et al., 2009). In contrast, loss-of-function analysis of SETD8 suggested that enrichment of H4K20me1 represses transcription of target genes regardless of their basal expression levels (Congdon et al., 2010, Kalakonda et al., 2008, Karachentsev et al., 2005). In addition, depletion of either of the H4K20me1 demethylases PHF8 or LSD1n resulted in reduced gene expression and elevated H4K20me1 levels at their respective target genes (Liu et al., 2010, Qi et al., 2010, Wang et al., 2015a). The repressive function of H4K20me1 is at least partially mediated by a polycomb protein, L3MBTL1, which binds this mark and compacts nucleosomes (Kalakonda et al., 2008, Trojer et al., 2007). Thus, it is conceivable that H4K20me1 is a transcriptionally repressive mark at actively transcribed genes. These studies suggest that SETD8 plays a fundamental role in gene expression as well as cell proliferation, but the role of SETD8 in cellular senescence remains unknown.
In this report, we show that SETD8 suppresses nucleolar and mitochondrial activities to prevent cellular senescence. SETD8 is downregulated in senescent cells induced by oncogenic and replicative stresses. Consistently, depletion of SETD8 directly induced senescence in cultured human fibroblasts. In the senescent state, the H4K20me1 landscape was largely remodeled at gene loci, including genes encoding ribosomal proteins (RPs) and RNAs, as well as p16INK4A, resulting in nucleolar and mitochondrial coactivation. Our data shed light on the evidence that SETD8 links the epigenomic dynamics to energy metabolism during cellular senescence.
Results
SETD8 Is Commonly Downregulated in Senescent Cells
To explore a specific chromatin regulator involved in cellular senescence, we used two established senescence models: oncogene-induced senescence (OIS) and replicative senescence (RS) (Figure 1A). OIS cells were prepared by expression of 4-hydroxytamoxifen (4-OHT)-inducible Ras (H-RasV12) for 6–8 days in IMR-90 ER:Ras or MRC5 ER:Ras human fibroblast cells. For RS, IMR-90 cells were repeatedly passaged for 10–12 weeks (shown by late passage). These models expressed senescence markers including SA-β-Gal and reduced incorporation of the thymidine analog 5-ethynyl-2’-deoxyuridine (EdU) (Figures 1A and S1A). We found that levels of SETD8 protein were specifically decreased in both OIS and RS cells, compared with growing cells (Figure 1B). Time-course analysis revealed that SETD8 was steeply decreased on and after day 4 during establishment of OIS (Figure S1B). qRT-PCR analysis indicated that levels of SETD8 mRNA did not significantly change in either OIS or RS cells (Figure S1C). By contrast, the OIS cells treated with the proteasome inhibitor MG-132 resulted in increased levels of SETD8 protein (Figure S1D). There were two SETD8 bands of ∼45 kDa, which could be post-translationally modified or processed. These data indicate that downregulation of SETD8 is a hallmark of the senescent state, probably due to selective proteasomal degradation.
Loss of SETD8 Induces Senescence in Human Fibroblasts
To elucidate the role of SETD8 during senescent processes, we performed specific knockdown (KD) of SETD8 in IMR-90 cells with three independent small interfering RNAs (siRNAs). Compared with the control (Ctr), the proliferation of SETD8-KD cells was reduced within 2 days of siRNA treatment (Figures 1C and 1D and S2A). Cell cycle analysis revealed that the population of cells in both the S and G2/M phases were increased in SETD8-KD cells (Figure S2B). After synchronization by double thymidine block, SETD8-KD cells showed delayed progression through the G2/M phase (Figures S2C and S2D), in agreement with the role of SETD8 in mitotic progression (Karachentsev et al., 2005). Unexpectedly, we found that SETD8-depleted cells exhibited SA-β-Gal marking and senescence-associated heterochromatic foci (SAHF), as well as enlarged cell size, beginning at KD day 3 (Figures 1E and 1F and S2E). Nearly all treated cells were SA-β-Gal positive after 9 days of siRNA treatment, and this state was maintained without additional siRNA treatment (Figure 1E and S2F). Western blot analysis revealed that the cyclin-dependent kinase inhibitors p16INK4A and p21CDKN1A were also increased in SETD8-KD cells (Figure 1G).
To further examine the functional implication of SETD8 in senescence, we used the selective SETD8 inhibitor UNC0379, which clearly reduced the levels of H4K20me1 in a dose-dependent manner (Figure S2G) (Ma et al., 2014). The proliferation of UNC0379-treated cells was significantly reduced (Figure 1H), while SA-β-Gal and SAHF appeared in a dose-dependent manner (Figures 1I and 1J). In another human fibroblast cell line, MRC5 ER:Ras, SETD8 inhibition also significantly increased SA-β-Gal and SAHF without oncogene activation (Figures S2H–S2K). Collectively, loss of SETD8 induces growth arrest and senescence in human fibroblasts.
SETD8 Represses RP and Senescence-Associated Genes
To investigate how SETD8 depletion induces senescence, we performed genome-wide microarray analyses of Ctr and SETD8-KD IMR-90 cells at 24 hr, as confirmed by western blot analysis (Figure S3A). Gene set enrichment analysis (GSEA) using the KEGG platform revealed that 12 gene sets were significantly upregulated by SETD8-KD (Figure S3B), while no gene sets were downregulated. These data support that SETD8 is involved in transcriptional repression of the gene sets (Congdon et al., 2010, Kalakonda et al., 2008, Karachentsev et al., 2005). Among the upregulated gene sets, the ribosome, the p53 signaling pathway, and the cytokine-cytokine receptor interaction could be associated with senescence and/or metabolic activities (Figure 2A). Notably, most of the RP gene probes as well as other highly expressed gene probes (those with a signal intensity > 20,000 in Ctr-KD cells, which was in the top 1.5% of probes tested) were concurrently upregulated by SETD8-KD (Figures 2B and 2C and S3C–S3E). We confirmed the upregulation of a set of RP genes by qRT-PCR in SETD8-KD cells (Figure 2D). Although the ratio of upregulation was relatively small, the global increase of the highly expressed RPgenes could affect the ribosome biogenesis and protein synthesis in SETD8-KD cells (as shown later). In contrast, expression of the mitochondrial RP genes was comparable between Ctr- and SETD8-KD cells (Figure S3F). We then confirmed upregulation of senescence markers such as p15INK4B, p16INK4A, p21CDKN1A, and interleukin (IL)-8 on day 1 and later of SETD8-KD (Figures 2E and S3G and S3H). Taken together, loss of SETD8 upregulates many genes, including a set of the RPgenes and senescence-associated genes in human fibroblasts.
SETD8 Regulates RP and Senescence-Associated Genes via H4K20 Mono-methylation
SETD8 is capable of regulating genes on its own by adding H4K20me1. To identify genes specifically targeted by SETD8, we examined the genome-wide H4K20me1 distribution in adult human dermal fibroblasts (NHDF-Ad cells) using ENCODE chromatin immunoprecipitation sequencing (ChIP-seq) datasets. Peak analysis using the SICER algorithm identified 8,967 genomic regions as H4K20me1-enriched islands, of which 94% (8,416 of 8,967) were deposited on intragenic regions of RefSeq-annotated genes. Functional annotation clustering analysis using Database for Annotation, Visualization, and Integrated Discovery (DAVID) revealed that H4K20me1 was preferentially enriched at a set of the RP genes (p = 6.12E-15; false discovery rate [FDR] = 8.10E-14) (Figures 3A, S4, and S5A). In fact, among the 86 RP genes detected in IMR-90 cells (Figure 2C), 83% (71 of 86) were enriched for H4K20me1 and 69% (59 of 86) were also upregulated by SETD8-KD (Figure 3A). In addition, p16INK4A and p21CDKN1A were enriched for H4K20me1 and were upregulated by SETD8-KD (Figures 2E, 3A, and S4).
Next, we tested whether loss of SETD8 affects the levels of H4K20me1. As shown in Figure 3B, global levels of H4K20me1 were decreased 23%–40% at 24 hr after SETD8-KD without changing the cell cycle distribution (Figures S5B and S5C). Consistent with previous reports, the basal levels of H4K20me1 were low at the S phase and high at the G2/M phase, compared with the G1 phase (Jørgensen et al., 2013), but depletion of SETD8 significantly decreased H4K20me1 levels at all cell cycle stages (Figure S5D). Thus, the overall reduction in H4K20me1 levels was not due to specific cell cycle stages.
We then performed ChIP to assess the levels of H4K20me1 at RP gene loci at 24 hr after SETD8-KD. RPL5, RPS24, RPL27A, and RPS18 were chosen as representative genes upregulated by SETD8-KD (Figure 2D). Interestingly, H4K20me1 was highly enriched at the intragenic regions of the RP genes, compared with the promoter regions, and depletion of SETD8 decreased gene-body H4K20me1 levels by 17%–31% (Figures 3C and S5E). In contrast, the levels of H4K20me3 did not change at these loci. Recently, it was reported that deposition of H4K20me1 at gene bodies is associated with repression of transcriptional elongation (Wang et al., 2015a). Consistent with this, we found that the levels of H3K36me3, a transcriptional elongation mark, were increased at gene bodies of most RP genes in SETD8-KD cells (Figures 3D and S5F). In addition, p21CDKN1A and p16INK4A genes were also enriched for H4K20me1 at gene bodies, and loss of SETD8 decreased H4K20me1 levels at these loci, together with an increase of H3K36me3 (Figures 3E and 3F and S5G and S5H). Peak analysis of H4K20me1 in three human cell lines, HeLa-S3, H1-human embryonic stem cells (hESC), and NHDF-Ad, indicated that enrichment of H4K20me1 at RP genes is conserved among different cell types (Figures S5I and S5J). Taken together, these results suggest that SETD8 is required for H4K20me1 at RP genes, p16INK4A and p21CDKN1A, and that depletion of SETD8 dampens H4K20me1-associated transcriptional repression of these genes.
SETD8 Regulates Ribosomal RNA Genes and Nucleolar Function via H4K20me1
The global upregulation of RP genes and enlarged cell size in SETD8-KD cells suggest that SETD8 depletion activates ribosome biogenesis during the establishment of senescence (Figures 2 and S2E). Ribosome biogenesis is tightly associated with cell growth and protein synthesis, especially due to the coordinate synthesis of RP proteins and ribosomal RNAs (rRNAs). Several hundred copies of rRNA genes are transcribed, and abundant rRNAs are processed by specific RPs at multiple steps, resulting in rRNA-protein assembly at the nucleolus (Thomson et al., 2013). Notably, as rRNA synthesis is a critical step in ribosome biogenesis, the rate of rRNA transcription is finely tuned by various signaling pathways and epigenetic mechanisms (Grummt and Längst, 2013, Moss et al., 2007). Interestingly, we found that rRNA gene bodies were modestly enriched for H4K20me1, and loss of SETD8 reduced 41%–62% of H4K20me1 levels, while H4K20me3 was rather stable at these loci (Figure 4A). In addition, the observed decrease of H4K20me1 coincided with increased deposition of H3K36me3 at gene bodies (Figure 4B), suggesting that SETD8 is involved in epigenetic regulation of rRNA genes. Next, we assessed the rate of rRNA transcription by measuring incorporation of the 5-ethynyl uridine (EU) in nucleoli coimmunostained with nucleophosmin (B23). Consistent with ChIP results, the rate of rRNA transcription was increased by SETD8-KD (Figure 4C). In agreement with the activation of ribosome biogenesis, the protein synthesis augmented in SETD8-KD cells (Figures S5K and S5L). Previously, it was reported that aberrant activation of rRNA genes evokes nucleolar stress, together with enlarged nucleolar morphology (Nishimura et al., 2015). In agreement with this, we found that SETD8-KD cells had enlarged nucleolar morphology, where the number of nucleoli per cell was decreased while the total area of nucleolus per cell was increased (Figures 4D and 4E). Collectively, our data suggest that SETD8 is responsible for H4K20me1 deposition at rRNA gene loci, and that loss of SETD8 upregulates rRNA transcription and induces nucleolar stress during senescence.
Loss of SETD8 Activates Mitochondrial OXPHOS via Retinoblastoma
Given that ribosome biogenesis is a major consumer of energy in cells (Warner, 1999) and that loss of SETD8 coordinately upregulates RPs and rRNAs, we speculated that depletion of SETD8 remodels the balance of energy metabolism for senescent state. In fact, several lines of evidence suggest that senescent cells exhibit higher mitochondrial activity than proliferating cells (Dörr et al., 2013, Hutter et al., 2004, Kaplon et al., 2013, Quijano et al., 2012, Takebayashi et al., 2015). Interestingly, we found that SETD8-KD cells increased mitochondrial oxygen consumption rate (OCR) at both the basal and carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP)-uncoupled maximal states (Figure 5A). In contrast, glycolytic activity, as indicated by the extracellular acidification rate (ECAR), did not largely change under the assay conditions (Figure S6A). Therefore, the ratio of basal OCR/ECAR increased by SETD8-KD (Figure 5B). In addition, depletion of SETD8 increased mitochondrial mass and membrane potential, as determined by fluorescent JC-1 staining (Figure 5C). Further, intracellular levels of reactive oxygen species (ROS) were increased approximately 2-fold in SETD8-KD cells (Figure 5D). These increases in mitochondrial activities were observed on day 3 of SETD8-KD when senescence markers were expressed only to a small extent (Figures 1E and 1F and S6B–S6F), suggesting that persistent upregulation of OXPHOS activity would contribute to establishment of senescence.
We and other groups previously reported that retinoblastoma (RB) protein plays a critical role in mitochondrial activity (Moiseeva et al., 2009, Nicolay et al., 2015, Takebayashi et al., 2015), as well as in cellular senescence (Chicas et al., 2010, Helman et al., 2016, Narita et al., 2003). Of note, the loss of SETD8 upregulated p16INK4A and p21CDKN1A (Figures 1G, 2E, 3E, and S5G), which could activate RB by inhibiting the cyclin-dependent kinases. To clarify whether RB is involved in mitochondrial activation in SETD8-KD cells, we performed simultaneous knockdown of RB and SETD8 (Figures S6G and S6H). Interestingly, depletion of RB reduced the S-phase population observed in SETD8-KD cells and abolished the activation of mitochondrial respiration (Figures 5E and 5F and S6I and S6J). Moreover, increases in the mitochondrial membrane potential and ROS accumulation in SETD8-KD cells were partially abolished by RB-KD (Figures 5G and 5H and S6K). These observations support the metabolic role of RB during establishment of senescence (Takebayashi et al., 2015). Further, depletion of RB completely abolished SA-β-Gal and SAHF formation in SETD8-KD cells, while the expression of p16INK4A and p21CDKN1A was consistently increased (Figures 5I and 5J and S6L). As was the case of RB, KD of p16INK4A also blocked mitochondrial activation and senescent features in SETD8-KD cells as well as in OIS cells (Figures S6M–S6S). Together, these data show that loss of SETD8 promotes mitochondrial OXPHOS activity and ROS accumulation, at least in part through RB function, contributing to senescence.
The Senescent State Remodels the H4K20me1 Landscape at SETD8 Target Genes
Our primary observations showed that SETD8 protein was downregulated during the establishment of senescence (Figure 1B). To investigate whether SETD8 target genes acquire altered H4K20me1 during senescence, we performed expression microarray analyses of OIS, RS, and SETD8-KD cells. Commonly upregulated gene sets included ribosome, lysosome, cancer, and p53 signaling, which are each closely associated with senescence (Figure 6A). Interestingly, a set of RP genes were consistently upregulated in both OIS and RS (late passage), as well as in SETD8-KD cells (Figures 6B and S7A). Subsequent ChIP analysis revealed that H4K20me1 at RP gene bodies was decreased in both OIS and RS cells, while the levels of H3K36me3 were partially increased at these loci (Figures 6C and 6D and S7B and S7C). Further, H4K20me1 decreased at rRNA gene loci in OIS cells, together with upregulation of rRNA transcription and enlarged nucleolar morphology (Figures 6E and 6F and S7D and S7E).
Under the downregulation of SETD8, we observed that global amount of H4K20me1 was decreased in OIS cells (Figure S7F). It was previously reported that the global distribution of H4K20me1 is dynamically regulated during cell differentiation, where this mark is gained at upregulated genes and lost at repressed genes (Cui et al., 2009). Thus, the status of H4K20me1 is likely to be regulated in cell- and gene-specific manners. We found that the levels of H4K20me1 in the gene bodies of HMGA2 and IL8 were increased in OIS cells, where both gene expression and H3K36me3 levels were increased (Figure S7G and S7H). On the contrary, as was observed at the RP and rRNA genes, H4K20me1 decreased at p16INK4A gene in OIS cells, even though its expression was increased (Figures 6G and 6H and S7I). Collectively, our data suggest that SETD8 plays an essential role in remodeling the H4K20me1 landscape at RP, rRNA, and p16INK4A gene loci in senescent cells.
Discussion
This study demonstrates that the SETD8 methyltransferase plays a pivotal role in senescence-associated metabolic remodeling, which is characterized by nucleolar and mitochondrial coactivation (Figures 4, 5, and 7A). Under OIS and RS conditions, SETD8 protein was downregulated, resulting in decreased H4K20me1 at target gene loci such as the RP, rRNA, and p16INK4A genes. As SETD8 exerts transcriptional repression via H4K20me1 (Congdon et al., 2010, Kalakonda et al., 2008, Karachentsev et al., 2005), the decrease in H4K20me1 levels led to transcriptional derepression of these genes (Figure 7B). Further, loss of SETD8 function using siRNAs or a specific inhibitor UNC0379 caused senescent phenotypes in proliferating fibroblasts, suggesting that SETD8 is essential for normal cell growth and that loss of SETD8 is sufficient to establish cellular senescence.
Based on the following two findings, we conclude that SETD8 is involved in the conversion of energy metabolism in cellular senescence. First, the activation of nucleolar function resulting in increased ribosome biogenesis is one feature of senescent cells. For example, abnormal activation of rRNA transcription, a critical process of ribosome biogenesis, is observed in OIS cells (Nishimura et al., 2015). Some oncogenes as well as tumor suppressors are involved in regulation of rRNAgenes (Moss et al., 2007). Our findings demonstrate that SETD8 regulates rRNAgenes by the establishment of H4K20me1 (Figure 4). This was corroborated by an observed loss of H4K20me1 at RP genes and their consequent upregulation in the absence of SETD8 (Figures 2 and 3). Therefore, our data suggest that SETD8 depletion upregulates ribosome biogenesis through coordinate activation of both rRNA and RP genes. Notably, a previous report suggested that aberrant activation of rRNA genes evokes nucleolar stress, which promotes p53/p21-dependent senescence (Nishimura et al., 2015). Moreover, translational control of protein synthesis is a key step in rapid production of SASP factors that accelerate senescence (Dörr et al., 2013, Herranz et al., 2015, Laberge et al., 2015). Thus, loss of SETD8 may promote senescence through activation of nucleolar function, owing to upregulation of rRNA and RP genes.
Second, increased mitochondrial OXPHOS activity is another feature of the senescent state (Dörr et al., 2013, Hutter et al., 2004, Kaplon et al., 2013, Quijano et al., 2012, Takebayashi et al., 2015). Recently, we and other groups reported that RB plays a critical role in mitochondrial activation (Franco et al., 2016, Moiseeva et al., 2009, Takebayashi et al., 2015). In agreement with this, we found that loss of SETD8 increased OXPHOS activity in an RB-dependent manner (Figure 5). It is well known that RB function is enhanced by p16INK4A and p21CDKN1A via inhibition of cyclin-dependent kinases (Campisi and d’Adda di Fagagna, 2007). Interestingly, SETD8 depletion upregulates p16INK4A and p21CDKN1A via decreased H4K20me1 deposition at their gene loci (Figures 1G, 2E, 3E, and S5G). Thus, RB is activated in SETD8-depleted cells, which in turn activates OXPHOS metabolism. As reported earlier, abnormal activation of mitochondrial OXPHOS leads to ROS accumulation, which promotes the DNA damage response and senescence (Correia-Melo et al., 2016). Indeed, we detected the presence of DNA damages, as indicated by γH2AX in SETD8-KD cells (Figure S7J). Thus, it is likely that loss of SETD8 accelerates senescence at least in part through mitochondrial OXPHOS activation, by upregulation of p16INK4A and/or p21CDKN1A.
It would be of great interest whether overexpression of SETD8 could prevent or suppress senescence. As previously reported (Abbas et al., 2010, Hartlerode et al., 2012, Sims and Rice, 2008), we found that overexpression of SETD8 rather decreased H4K20me1 levels in both IMR-90 and HeLa cells (Figures S7K–S7M). In addition, the transient expression of SETD8 did not affect the mitochondrial activation and the cell cycle arrest in OIS and RS cells (Figures S7N–S7Q). Mechanistically, it is important to understand how SETD8 and H4K20me1 become downregulated during senescence. SETD8 protein is decreased via proteasome degradation rather than transcriptional control under stress conditions (Figure S1). Previous studies revealed that SETD8 is a substrate of the E3 ubiquitin ligases CRL4Cdt2 and SCFβ-TRCP, which mediate degradation of SETD8 during the DNA damage response (Beck et al., 2012, Wang et al., 2015b). Similarly, another ubiquitin ligase, APCCdh1, is also involved in SETD8 degradation (Beck et al., 2012). This ligase is activated via ROS-mediated DNA damage in OIS cells and is also involved in degradation of the H3K9 methyltransferases G9a and GLP (Takahashi et al., 2012). As senescent cells are subjected to persistent DNA damage caused by telomere attrition, replication stress, and/or ROS accumulation (Kuilman et al., 2010), these ligases may be cooperatively involved in SETD8 degradation and accelerate senescence.
It also remains unresolved how H4K20me1 is removed from specific gene loci after SETD8 inhibition. It has been reported that PHF8 (PHD finger protein 8) demethylates H4K20me1 as well as H3K9me1/2 at the promoters of target genes, and consequently activates gene expression (Liu et al., 2010, Qi et al., 2010). Interestingly, PHF8 acts on the rRNA gene loci at both promoter and coding regions and upregulates rRNA transcription, although it has not been determined whether its demethylase activity against H4K20me1 is implicated in this process (Feng et al., 2010). Alternatively, H4K20 is known to be specifically mono-methylated at the G2/M phase when SETD8 levels are high, and the majority of H4K20me1 modifications are subsequently di-/tri-methylated during the M/G1 phases (Jørgensen et al., 2013). In fact, the levels of H4K20me1 were stable for at least 100 days in SETD8-null non-dividing hepatocytes of mice, whereas H4K20me1 became lost once these cells divided, with DNA damage and ROS accumulation (Nikolaou et al., 2015). Thus, there are three possibilities for reducing H4K20me1 in the absence of SETD8: active removal by demethylases, active conversion from the mono- to di-/tri-methylated form by methyltransferases, or passive demethylation through cell division.
Several lines of evidence suggest that metabolic interventions could prevent senescence and the aging process. For example, experimental reduction of ribosome biogenesis or protein translation delayed replicative senescence in human fibroblasts and even extended lifespan in some organisms, including mice (Hofmann et al., 2015, MacInnes, 2016, Takauji et al., 2016). Similarly, reduction of ROS accumulation also attenuated senescent phenotypes under oncogene and replication stresses, prevented stem cell aging, and increased longevity in mice (García-Prat et al., 2016, Lu and Finkel, 2008, Sun et al., 2016). Further, caloric restriction in aged mice reduced the population of senescent cells in the small intestine and liver (Wang et al., 2010). Therefore, understanding the molecular basis of senescence-associated metabolic remodeling and the epigenomic features in cellular senescence might explain why the senescent state presents such hypermetabolic phenotypes, leading to the development of novel treatments in senescence- and age-related disorders.
Author Contributions
H.T., S.T., A.S., and M.N. designed and conducted the experiments, together with support from T.I., Y.N., N.S., and S.H. H.T., S.H., and M.N. prepared the manuscript.
————-
The p53 tumor suppressor protein is regulated by multiple post-translational modifications, including lysine methylation. We previously found that monomethylation of p53 at lysine 382 (p53K382me1) by the protein lysine methyltransferase (PKMT) SET8/PR-Set7 represses p53 transactivation of target genes. However, the molecular mechanism linking p53K382 monomethylation to repression is not known. Here we show in biochemical and crystallographic studies the preferential recognition of p53K382me1 by the triple malignant brain tumor (MBT) repeats of the chromatin compaction factor L3MBTL1. We demonstrate that SET8-mediated methylation of p53 at Lys-382 promotes the interaction between L3MBTL1 and p53 in cells, and the chromatin occupancy of L3MBTL1 at p53 target promoters.
In the absence of DNA damage, L3MBTL1 interacts with p53K382me1 and p53-target genes are repressed, whereas depletion of L3MBTL1 results in a p53-dependent increase in p21 and PUMA transcript levels. Activation of p53 by DNA damage is coupled to a decrease in p53K382me1 levels, abrogation of the L3MBTL1-p53 interaction, and disassociation of L3MBTL1 from p53-target promoters. Together, we identify L3MBTL1 as the second known methyl-p53 effector protein, and provide a molecular explanation for the mechanism by which p53K382me1 is transduced to regulate p53 activity.
The p53 tumor suppressor protein is regulated by multiple post-translational modifications, including lysine methylation. We previously found that monomethylation of p53 at lysine 382 (p53K382me1) by the protein lysine methyltransferase (PKMT) SET8/PR-Set7 represses p53 transactivation of target genes. However, the molecular mechanism linking p53K382 monomethylation to repression is not known. Here we show in biochemical and crystallographic studies the preferential recognition of p53K382me1 by the triple malignant brain tumor (MBT) repeats of the chromatin compaction factor L3MBTL1.
We demonstrate that SET8-mediated methylation of p53 at Lys-382 promotes the interaction between L3MBTL1 and p53 in cells, and the chromatin occupancy of L3MBTL1 at p53 target promoters.
In the absence of DNA damage, L3MBTL1 interacts with p53K382me1 and p53-target genes are repressed, whereas depletion of L3MBTL1 results in a p53-dependent increase in p21 and PUMA transcript levels.
Activation of p53 by DNA damage is coupled to a decrease in p53K382me1 levels, abrogation of the L3MBTL1-p53 interaction, and disassociation of L3MBTL1 from p53-target promoters. Together, we identify L3MBTL1 as the second known methyl-p53 effector protein, and provide a molecular explanation for the mechanism by which p53K382me1 is transduced to regulate p53 activity.