Normal human somatic cells have a finite replicative lifespan, and after prolonged replication they undergo cellular senescence as a result of telomere dysfunction6,7
. Despite the ageing-like phenotype of SIRT6-deficient mice, no effect of SIRT6 deficiency on cellular lifespan has been reported. To determine whether SIRT6 influences cellular senescence, retroviral transduction of short hairpin RNAs (shRNAs) was used to stably knock down SIRT6
expression in WI-38 human fibroblasts (). SIRT6
knockdown (S6KD) cells have a strikingly shortened replicative lifespan, undergoing premature cellular senescence about ten population doublings before control cells, and show increased levels of senescence-associated β-galactosidase (SA-β-gal) staining (, and Supplementary Fig. 2a
). These kinetics of premature senescence are similar to those associated with the inactivation of telomere accessory factors (Supplementary Fig. 3a, b
. Premature cellular senescence was also observed with two independent SIRT6
shRNAs, ruling out off-target shRNA effects, but not with irrelevant shRNAs (Supplementary Fig 2b–d
and Supplementary Fig 3c
). We conclude that SIRT6 is crucial in maintaining a normal replicative life-span and in preventing the premature senescence of human cells.
SIRT6 knockdown leads to premature cellular senescence and telomere dysfunction
Replicative cellular senescence can result from dysfunctional telomeres, which are recognized by DNA damage response factors and are detected as telomere dysfunction-induced foci (TIFs)9,10
. Analysis of TIFs revealed elevated telomere dysfunction in S6KD cells (, and Supplementary Fig. 4a
). The telomere signals at TIFs in S6KD cells are weak compared with non-TIF telomere signals, suggesting a subpopulation of telomeres that have undergone significant sequence loss (see below). However, mean telomere length was not significantly reduced in S6KD cells (Supplementary Fig. 5
). Together, these observations suggest that S6KD cells undergo accelerated senescence and telomere dysfunction in response to stochastic telomere sequence loss, without increased global telomere erosion.
Loss of proper telomeric protective end structures can lead to dicentric chromosomes as a result of chromosomal end-to-end fusions11
. We therefore scored chromosomal end fusions in S6KD and control metaphases in several independent cytogenetic analyses. Non-recurrent chromosomal end-to-end fusions were observed in S6KD cells and were more pronounced at later population doublings, but they were rarely observed in control cells ( and Supplementary Fig. 4b, c
). These observations indicate that SIRT6 is critical for maintaining functional telomeres to avert chromosomal instability due to aberrant chromosomal end-to-end fusions.
Several experiments provide further evidence that the premature senescence of S6KD cells is due to telomere dysfunction and not to defective base excision repair (BER), which was previously implicated in the phenotypes of SIRT6
knockout (S6KO) mouse cells3
, or to oxidative stress brought on by supraphysiological oxygen conditions of ambient cell culture conditions12
. First, telomere stabilization (by the ectopic expression of telomerase (hTERT)) reversed the premature senescence of S6KD, whereas augmenting BER (by the ectopic expression of the DNA polymerase-β dRP lyase domain) did not ( and Supplementary Fig. 6
). This DNA polymerase-β domain was previously shown to rescue the hypersensitivity of S6KO mouse cells to alkylating DNA damage agents3
. Second, S6KD cells underwent premature senescence even when cultured under low (physiological) oxygen conditions (Supplementary Fig. 7
). Together, these findings demonstrate that telomere dysfunction, not BER defects or oxidative stress, underlie the premature senescence phenotype of S6KD cells.
The premature cellular senescence, telomere dysfunction and chromosomal fusions observed in S6KD cells are reminiscent of the cellular phenotype of Werner syndrome (WS)13–16
, a hereditary dis-order associated with signs of premature ageing4,5
. The WS-defective protein WRN associates with telomeres (in primary human IMR90 cells and U2OS osteosarcoma cancer cells) and regulates telomere processing during S phase13,17
. We formed the hypothesis that SIRT6 might function in a similar context. To investigate this possibility, we first examined whether SIRT6 associates with telomeres during S phase by telomere chromatin immunoprecipitation (T-ChIP)18
. Cell synchronization, release, and analysis by bromodeoxyuridine/propidium iodide staining were performed to enrich for specific cell-cycle phases (Supplementary Fig. 8
). T-ChIP analysis at different time points after release from cell synchronization revealed that SIRT6, like WRN, preferentially associates with telomeric chromatin in S-phase-enriched cultures (Supplementary Fig. 9
). SIRT6 occupancy at telomeric chromatin was observed for both recombinant Flag-tagged SIRT6 () and endogenous SIRT6 (), and in both U2OS osteosarcoma cells () and primary IMR90 cells (). As controls, the association of Alu repeat sequences with SIRT6 ChIPs was not above background (), and the SIRT6 T-ChIP signal was abolished in S6KD cells, validating the specificity of the signal (). Together, these data locate SIRT6 at telomeric chromatin in S phase and suggest a potential role for SIRT6 in regulating replication-associated dynamics in telomere structure.
SIRT6 associates with telomeric chromatin
Chromatin at telomeres is enriched for hypoacetylated histone tails19
. Although no physiological enzymatic activity for SIRT6 on histones or other trans
substrate has yet been observed, we proposed that SIRT6 might regulate chromatin at telomeres by deacetylating a specific histone tail residue. We therefore used mass spectrometry to screen for NAD-dependent SIRT6 deacetylase activity in vitro
, on a collection of acetylated histone tail peptides. SIRT6 manifested modest deacetylation activity on peptides containing acetylated H3K9 (H3K9Ac) ( and Supplementary Fig. 10
). This activity was highly specific for H3K9Ac, because no deacetylation was detected for 12 other acetylated histone peptides ( and Supplementary Fig. 11
). SIRT6 also deacetylated H3K9Ac, but not several other acetylated residues, in the context of purified full-length histone H3, and mutation of a conserved catalytic residue (H133Y) of SIRT6 markedly decreased this activity ( and data not shown). Finally, SIRT6 efficiently and specifically deacetylated H3K9Ac (but not other acetylated histone residues) in 293T cells, whereas the mutant SIRT6 protein did not ( and Supplementary Fig. 12
). Together, these observations indicate that SIRT6 is an NAD-dependent deacetylase with specificity for H3K9Ac.
SIRT6 deacetylates lysine 9 of histone H3 at telomeric chromatin
We next sought to identify the physiological context in which SIRT6 deacetylates H3K9Ac. Western analysis of S6KD and control cells did not reveal significant differences in global H3K9Ac levels (data not shown). In contrast, investigation of histone acetylation status at telomeres in S-phase-enriched cultures by T-ChIP revealed H3K9 hyperacetylation in S6KD cells (). SIRT6 is therefore required for the maintenance of the low physiological levels of H3K9 acetylation at telomeric chromatin in S phase, and hyperacetylation of H3K9 in the absence of SIRT6 correlates with telomere dysfunction. H3K9 was also hyperacetylated at telomeric chromatin in S6KO mouse cells, providing in vivo
evidence for a physiological role for SIRT6 in deacetylating this histone residue (Supplementary Fig. 13a–c
We next proposed that in SIRT6-deficient cells, hyperacetylation of H3K9 at telomeres in S phase might interfere with the association of WRN. WRN occupancy at telomeres was compared in S-phase-enriched S6KD and control cultures by T-ChIP analysis. In both U2OS and IMR90 cells, SIRT6
knockdown significantly inhibited the association of WRN with telomeric chromatin (, and data not shown). We conclude that SIRT6 is required for stabilization of WRN at telomeric chromatin. We note that additional functional or physical interactions between SIRT6 and WRN might exist. However, the association of SIRT6 with telomeres was independent of WRN (Supplementary Fig. 13e
), and we have not observed a direct interaction between SIRT6 and WRN at chromatin (data not shown). Thus, our data are consistent with the hypothesis that SIRT6 deacetylation of H3K9Ac at telomeric chromatin leads to an altered chromatin state that is required for efficient WRN association in S phase.
SIRT6 stabilizes WRN at telomeric chromatin and prevents replication-associated telomere defects
WS cells exhibit specific defects that reflect problems with the replication-associated processing and metabolism of telomeres, and analysis of S6KD cells revealed similar abnormalities. First, S6KD metaphases showed elevated levels of missing telomere signals (sister telomere loss) and extra telomere signals (telomere doublets) (), defects observed in WS cells and associated with aberrant telomere processing during replication8,13,20
. In addition, chromosome fusions in S6KD cells, as in WRN-defective cells14,15
, have weak or no telomere signals at the fusion sites (). In WS cells this phenotype is proposed to result from stochastic replication-associated telomere loss5,15
, and it contrasts with the strong telomere signals observed at sites of chromosome fusions resulting from deficiency for telomere end-capping factors such as telomeric repeat binding factor 2 (TRF2)11
. Aberrant replication-associated telomere processing by WRN is proposed to contribute to a delayed completion of S phase observed in WS cells13,14,21,22
. Similarly, SIRT6
knockdown in U2OS cells resulted in delayed completion of S phase (Supplementary Fig. 14
), which is consistent with a role for SIRT6 in modulating telomeres during replication. Together, these findings suggest that SIRT6 collaborates with WRN at telomeric chromatin to ensure efficient telomere replication and to prevent the accrual of structural abnormalities at telomeres.
Telomeres are specialized structures that function to shield linear chromosome ends from DNA repair, degradation and fusion7
. Mammalian telomeres are packaged in an unusual chromatin structure with features of heterochromatin23,24
, but relatively little is understood about the role of chromatin modifications on telomere metabolism. In this study we show that human SIRT6 deacetylates H3K9Ac at telomeres to prevent telomere dysfunction. Inactivation of SIRT6 leads to H3K9 hyperacetylation at telomeric chromatin in both human and mouse cells, but the mouse cells do not display the downstream cellular defects observed in human cells, which is consistent with the large functional reserve of mouse telomeres (Supplementary Fig. 13d
; data not shown)14,25
. Our findings regarding SIRT6 provide a direct link between mammalian telomere dysfunction and a histone modifying enzymatic activity. Our results indicate that deacetylation of H3K9Ac by SIRT6 is important during S phase at replicating telomeres. We propose a model (Supplementary Fig. 1
) in which the deacetylation of H3K9Ac by SIRT6 promotes the formation of a specialized telomeric chromatin state that is required for the stable association of S-phase-dependent telomere-processing factors such as WRN, to prevent aberrant sequence loss or metabolism of telomeres. The resulting telomere dysfunction then contributes to premature cellular senescence. Our study has identified a crucial function for SIRT6 in chromatin regulation at mammalian telomeres and provides a new mechanism by which the regulation of chromatin is linked to telomere function, cellular senescence and, potentially, organismal ageing.