The SNF2h (sucrose non-fermenting protein 2 homologue)-containing chromatin-remodelling complex NoRC silences a fraction of ribosomal RNA genes (rDNA) by establishing a heterochromatic structure at the rDNA promoter1,2,3. Here we show that the acetyltransferase MOF (males absent on the first) acetylates TIP5, the largest subunit of NoRC, at a single lysine residue, K633, adjacent to the TIP5 RNA-binding domain, and that the NAD+-dependent deacetylase SIRT1 (sirtuin-1) removes the acetyl group from K633. Acetylation regulates the interaction of NoRC with promoter-associated RNA (pRNA), which in turn affects heterochromatin formation, nucleosome positioning and rDNA silencing. Significantly, NoRC acetylation is responsive to the intracellular energy status and fluctuates during S phase. Activation of SIRT1 on glucose deprivation leads to deacetylation of K633, enhanced pRNA binding and an increase in heterochromatic histone marks. These results suggest a mechanism that links the epigenetic state of rDNA to cell metabolism and reveal another layer of epigenetic control that involves post-translational modification of a chromatin remodelling complex.
Plasticity in transcription regulation can be achieved by dynamic regulation of the chromatin structure, involving ATP-dependent changes in nucleosome positions, modification of histone tails or DNA methylation. Acetylation of K16 on histone H4 (H4K16ac) is critical for chromatin decondensation, transcriptional activity and dosage compensation in Drosophila melanogaster4,5,6,7. Many transcriptional regulators, such as p300/CBP, GCN5 and MOF, have been proven to be histone acetyltransferases8,9. Drosophila MOF acetylates histone H4 (at K16) and other proteins, including MSL3, a subunit of the dosage compensation complex. Acetylation modulates the interaction of MSL3 with roX2 RNA7,10,11.
Here, we present data showing that the activity of NoRC, which consists of TIP5 and SNF2h (ref. 1), is regulated by mammalian MOF- and SIRT1-dependent acetylation and deacetylation. NoRC mediates silencing of a fraction of rDNA by recruiting histone-modifying enzymes and DNA methyltransferases, leading to heterochromatin formation and transcriptional silencing1,2,3,12. NoRC also shifts the promoter-bound nucleosome into a position that prevents transcription complex formation13. As binding of NoRC to rDNA requires the interaction of TIP5 with acetylated H4K16, which is acetylated by MOF14,15,16,17, we surmised that MOF might regulate the epigenetic state of rDNA. Indeed, chromatin immunopreciptitation (ChIP) experiments showed that MOF is associated with rDNA, the level of MOF and H4K16ac being higher at the promoter region than in the pre-rRNA coding region (Supplementary Information, Fig. S1a). Depletion of MOF decreased the association of H4K16ac and mouse TIP5 with rDNA and impaired nucleolar localization of NoRC (Supplementary Information, Fig. S1b, c). Digestion of immunoprecipitated DNA with the methylation-sensitive restriction enzyme HpaII revealed that DNA polymerase I (Pol I) was exclusively found at active, unmethylated rRNA genes, whereas mTIP5 and the heterochromatic trimethylated histone mark H4K20me3 were associated with methylated rDNA copies (Fig. 1a). In contrast, MOF and H4K16ac occupied both methylated and unmethylated rDNA copies, suggesting that they may have roles both in activating and silencing rRNA genes.
To examine whether MOF is involved in NoRC-dependent rDNA silencing, we assayed the interaction of MOF with TIP5 in vivo and in vitro. As shown in Fig. 1b, MOF but not Pol I co-precipitated with TIP5, at the region adjacent to its TAM (TIP5, ARBP and MBD) domain, required for MOF binding (Supplementary Information, Fig. S2a). In an in vitro assay with radiolabelled acetyl-CoA, MOF efficiently acetylated truncated human (h) TIP5 (Fig. 1c). Mass spectrometric analysis identified K649, a residue adjacent to the RNA binding domain, as a target of MOF-dependent acetylation (Supplementary Information, Fig. S2b). K649 in hTIP5 corresponds to K633 in mTIP5. In vitro, MOF acetylated K633, as monitored with a peptide-specific antibody raised against acetylated (ac) mTIP5 (Fig. 1d; Supplementary Information, Fig. S2c). In vivo overexpression of MOF enhanced acetylation of mTIP5, but not of the mTIP5K633R mutant. Acetylation of mTIP5 did not occur in the presence of the HAT domain-deficient mutant MOFG627D (Fig. 1e).
Cell fractionation experiments and quantitative western blots revealed that a minor fraction (7–11%) of cellular TIP5 was acetylated (data not shown) and that most of this acetylated TIP5 was associated with chromatin (Fig. 2a). Both mTIP5 and mTIP5K633 localized to nucleoli and were tightly associated with chromatin, indicating that acetylation is not required for recruiting NoRC to rDNA (Supplementary Information, Fig. S3a–c). Overexpression of mTIP5 reduced euchromatic histone marks, such as H4ac and H3K4me2, and increased the level of repressive histone modifications, such as H3K9me2, H3K27me3 and H4K20me3 (Fig. 2b). In contrast, the acetylation-deficient mutant mTIP5K633R did not affect any of these histone modifications. In addition, mTIP5K633R was not capable of triggering de novo CpG (cytosine and guanine separated by a phosphate) methylation (Fig. 2c) or transcriptional repression of a co-transfected Pol I reporter plasmid (Fig. 2d), emphasizing the importance of K633 acetylation for heterochromatin formation and silencing. Consistently, short hairpin RNA (shRNA)-mediated knockdown of MOF led to hypoacetylation of TIP5 (Fig. 2e), abrogated transcriptional repression (Fig. 2f) and prevented NoRC-dependent de novo methylation of rDNA (Fig. 2g).
As NoRC function requires the association of TIP5 with pRNA (150–250 nucleotide RNAs that are complementary in sequence to the rDNA promoter)18,19, we examined the effect of TIP5 acetylation on pRNA binding. Contrary to our expectations, overexpression of MOF strongly decreased the level of pRNA associated with NoRC (Fig. 3a). Similarly, in pulldown experiments, acetylation by MOF strongly reduced the pRNA-binding activity of truncated mTIP5 (TIP51-731), whereas three times more pRNA was associated with the acetylation-deficient mutant mTIP5K633, regardless of MOF coexpression (Fig. 3b). These results indicate that acetylation impairs the interaction of TIP5 with pRNA, an unexpected result given that pRNA binding is essential for NoRC function18. As NoRC is targeted to rDNA not only by binding to pRNA but also by interaction with TTF-I bound to its promoter-proximal target site T01, acetylation may also weaken the association of TIP5 with TTF-I. Indeed, co-immunoprecipitation experiments demonstrate enhanced interaction of TIP5K633R with TTF-I, supporting the notion that unacetylated NoRC is recruited to rDNA (Supplementary Information, Fig. S4).
To decipher the crosstalk between TIP5 acetylation, pRNA binding and NoRC function, we sought to artificially lower the affinity of mTIP5K633R for pRNA by introducing two point mutations (WY531,532GA) into its TAM domain. These two amino acid exchanges have been shown to impair the TIP5 interaction with pRNA and heterochromatin formation18. However, this mutant (mTIP5WY) was still capable of triggering DNA methylation and of silencing rDNA transcription. As expected, pRNA binding of mTIP5WYK633R, a mutant that is deficient in both RNA binding and K633 acetylation, was just as low as that of mTIP5WY (Fig. 3c; Supplementary Information, Fig. S5). Both mutants were less tightly associated with chromatin than wild-type TIP5, highlighting the importance of pRNA association for the binding of NoRC to chromatin (Supplementary Information, Fig. S3b). Surprisingly, mTIP5WYK633R efficiently triggered de novo DNA methylation and transcriptional silencing, implying that a weakened RNA binding enabled the acetylation-deficient mutant to function properly (Fig. 3d, e). The observation that impaired RNA binding rescues the silencing activity of mTIP5K633R demonstrates that both binding and release of pRNA are indispensable for NoRC function, and that reversible acetylation of TIP5 regulates its association and dissociation with pRNA.
Silencing of rRNA genes has been shown to require ATP-dependent nucleosome remodelling, which causes the promoter-bound nucleosome downstream of the transcription start site to shift into a translational position that is incompatible with transcription complex assembly13. To examine whether TIP5 acetylation is necessary for NoRC-dependent nucleosome repositioning, we digested crosslinked chromatin from cells overexpressing wild-type or mutant TIP5 with MNase and mapped the 5′-ends of mononucleosomal DNA by ligation mediated (LM)-PCR using linker- and rDNA-specific primers (Fig. 3f). Amplification of the reaction products yielded two PCR fragments, the shorter one corresponding to promoter-bound nucleosomes at active genes (from −157 to −2) and the more slowly migrating one corresponding to nucleosomes at silent genes (from −132 to +22; ref. 13). In mock-treated cells, the ratio of the two PCR fragments was similar with ∼60% of nucleosomes in the active and ∼40% in the silent position. On overexpression of mTIP5, the levels of the two fragments changed to 20% and 80%, respectively, demonstrating that enhanced levels of NoRC increased the proportion of the downstream nucleosome position at the expense of the upstream one. Significantly, the RNA binding-deficient mutants mTIP5WY531,532 and mTIP5WYK633R, but not the acetylation-deficient mutant TIP5K633R, were capable of shifting the nucleosome into the silent position. This result supports the notion that TIP5 acetylation promotes its dissociation from pRNA, a prerequisite for NoRC-dependent nucleosome positioning.
To identify the enzyme(s) that counteract MOF-dependent acetylation of K633, we monitored TIP5 acetylation after treating cells with specific deacetylase inhibitors. Trichostatin A (TSA) did not affect TIP5 acetylation, whereas inhibitors of NAD+-dependent deacetylases, including nicotinamide and splitomycin, enhanced K633 acetylation. Resveratrol, an activator of the NAD+-dependent deacetylase SIRT1, decreased K633 acetylation, suggesting that SIRT1 counteracts MOF-dependent acetylation of TIP5 (Fig. 4a). Consistently, overexpression of SIRT1, but not the inactive mutant SIRT1H355Y, decreased K633 acetylation (Fig. 4b). In contrast, knockdown of SIRT1 enhanced K633 acetylation (Fig. 4c). Significantly, upon SIRT1 knockdown, TIP5 did not promote transcriptional repression and de novo DNA methylation (Fig. 4d, e), demonstrating the relevance of reversible acetylation for proper NoRC function.
Given that energy-dependent changes in the NAD+/NADH ratio regulates the activity of SIRT1, we examined NoRC acetylation and epigenetic changes at the rDNA promoter on glucose deprivation. Glucose depletion reduced pre-rRNA synthesis and decreased Pol I occupancy at the rDNA promoter (Fig. 4f, g). In addition, euchromatic histone marks, such as H4ac and H3K9ac, were decreased, whereas heterochromatic marks, such as H3K9me2 and H4K20me3, were increased (Fig. 4g). Elevated levels of heterochromatic histone modifications correlated with decreased binding of Pol I and enhanced binding of TIP5 to rDNA, in accord with a limited energy supply promoting NoRC-dependent heterochromatin formation (Fig. 4g). Consistent with glucose deprivation activating SIRT1 (refs 20, 21), the level of acetylated TIP5 dropped ∼5-fold, whereas the amounts of SIRT1 and MOF were not altered (Fig. 4h). Significantly, deacetylation of TIP5 in glucose-depleted cells led to a significant increase in the level of NoRC-associated pRNA (Fig. 4i), again demonstrating the functional link between reversible TIP5 acetylation, pRNA binding and rDNA silencing.
If reversible acetylation regulates NoRC function, then changes in TIP5 acetylation should be most obvious during late S phase, that is, when the silent copies of rRNA genes are replicated and the heterochromatic state is inherited22. Indeed, the overall level of TIP5 was elevated in late S phase (5–6 h after release from G1/S), whereas TIP5 acetylation peaked earlier and declined to background levels in late S phase, consistent with a decrease in MOF and an increase in SIRT1 levels (Fig. 5a, b). TIP5, HDAC1 and H4K20me3 occupancy on rDNA was also most pronounced 5 h after release from G1/S arrest (Fig. 5c). Importantly, the level of NoRC-associated pRNA was also elevated in late S phase, that is, when TIP5 acetylation was lowest (Fig. 5d). The inverse correlation between the acetylation state of NoRC and its association with pRNA and chromatin, reveals a mechanism by which reversible acetylation modulates NoRC binding to pRNA and propagates the heterochromatic state of silent rRNA genes through cell division.
The following model illustrates the role of acetylation and pRNA binding in NoRC-dependent heterochromatin formation and gene silencing (Fig. 5e). NoRC is recruited to rDNA by interaction with TTF-I bound to a target sequence upstream of the transcription start site1 and by interaction with pRNA that matches the rDNA promoter18. Preliminary data indicate that binding to pRNA traps NoRC at the rDNA promoter. Therefore, transient dissociation of TTF-I and the release of pRNA is required for ATP-dependent chromatin remodelling and transcriptional silencing. Acetylation by MOF weakens the interaction between NoRC and both pRNA and TTF-I, and this change in RNA binding seems to be required for shifting the promoter-bound nucleosome downstream of the transcription start site into a translational position that is unfavourable for transcription complex formation13. In turn, NoRC-dependent re-positioning of the nucleosome at the rDNA promoter is required for DNA methylation by NoRC-associated DNA methyltransferases (DNMTs). Methylation of a critical CpG residue within the upstream control element (UCE) has been shown to prevent binding of UBF to chromatin templates and to abrogate transcription complex formation23. To maintain and propagate the silent state of rDNA repeats through cell division, SIRT1 has to remove the acetyl group from K633 in late S phase. Deacetylation is required for re-association with pRNA, a prerequisite for heterochromatin formation and transcriptional silencing. Thus, acetylation by MOF and deacetylation by SIRT1 is instrumental for NoRC function. Together, our results reveal a crucial control mechanism that regulates epigenetic gene silencing by reversible acetylation of a chromatin remodelling complex, highlighting the close interrelationships between the acetyltransferase MOF, the NAD+-dependent deacetylase SIRT1 and the chromatin remodeller NoRC. The involvement of the NAD+-dependent deacetylase SIRT1 in the silencing process provides an opportunity to link cellular energy status to rDNA silencing, a process required for nucleolar integrity and cell survival24,25.
Plasmids, cell lines and cell synchronization.
Expression vectors encoding TIP5 and truncation mutants of TIP5 (refs 1, 13), MOF (ref. 14), SIRT1 (ref. 26) and the reporter plasmid pMr131-BH12 have been described previously. The RNA binding-deficient mutant TIP5WY531,532 contains two amino acid exchanges within the TAM domain18. To generate the acetylation-deficient mutant TIP5K633R, K633 was replaced by arginine. Mutant TIP5WYK633R is a fusion of TIP5WY531,532 and TIP5K633R. shRNA against MOF (5′-gtgatccagtctcgagtga-3′) and SIRT1 (5′-gatgaagttgacctcctca-3′) was cloned into the retroviral vector pQXCIP27. Mouse L1210 cells were synchronized at G1/S by aphidicolin (2 μg ml−1 for 12 h), released into aphidicolin-free medium for 10 h, and then cultured for another 14 h in the presence of aphidicolin. For glucose deprivation, cells grown in DMEM (with 25 mM glucose) were washed twice with PBS and cultured in glucose-free medium for 24 h.
Chromatin immunoprecipitation (ChIP) and DNA methylation assays.
Crosslinked chromatin was incubated with antibodies and immunoprecipitated proteins were bound to protein A/G Sepharose. After reversal of the crosslink and digestion with proteinase K, DNA was extracted and amplified using quantitative (q)PCR. The ratio of rDNA in the immunoprecipitates versus rDNA in the input chromatin was normalized to that from control reactions in mock-treated cells. Data from ChIP assays using antibodies for specific histone modifications were normalized to data from histone H3 ChIP assays . To monitor CpG methylation, DNA was digested with HpaII (20 U per μg DNA) before PCR amplification. The relative resistance to HpaII digestion was normalized to mock- and MspI-digested DNA23. Primers used to amplify human or mouse rDNA have been described previously28.
Transfection and RNA analysis.
Mouse NIH 3T3 cells (3 × 105) were transfected with the Pol I reporter plasmid pMr131-BH (1 μg) and different amounts of expression vectors encoding wild-type or mutant Flag- or HA-tagged TIP5. For stable expression of TIP5, NIH 3T3 cells were infected with a retrovirus encoding wild-type or mutant mTIP5 and cultured for 4–5 days in the presence of puromycin (5 μg ml−1) before collection. Transcripts from the reporter plasmid were monitored on northern blots by hybridization to a 32P-labelled riboprobe that is complementary to pUC9 sequences inserted between the 5′- and 3′-terminal rDNA fragment in pMr131-BH. Alternatively, transcripts were analysed by RT-qPCR and normalized to GAPDH mRNA.
Purification of recombinant proteins and cell fractionation.
For acetylation assays, histidine-tagged hTIP5 was expressed in Escherichia coli BL21(DE3) cells and purified on Ni-NTA agarose (Qiagen). For functional studies, TIP5 was immunopurified from human HEK293T cells overexpressing Flag- and HA-tagged TIP5. To assay the association of NoRC with chromatin, nuclei were extracted twice with 10 mM Pipes at pH 7.0, 420 mM NaCl, 3 mM MgCl2 0.2 mM EDTA and 0.5 mM dithiothreitol, and lysed; chromatin was dissolved in 10 mM Pipes at pH 7.0, 420 mM NaCl, 10 mM MgCl2 and 0.5% Triton X-100. After sonication, chromatin was incubated with benzonase (for 30 min, 30 U), cleared by centrifugation and immunoblotted to determine the distribution of TIP5 in the chromatin-bound and soluble fractions.
RNA binding assays and RNA immunoprecipitation (RIP).
HEK293T cells overexpressing Flag- and HA-tagged truncated TIP51-731 were lysed in 500 μl of L-buffer (20 mM Tris-HCl at pH 7.4, 200 mM NaCl, 2 mM EDTA, 2 mM EGTA, 1% Triton X-100, 3.5 mM sodium butyrate, 5 ng ml−1 TSA and 10 mM nicotinamide) and incubated for 3 h with M2-agarose beads. Bead-bound TIP5 was incubated for 30 min on ice with in vitro transcribed radiolabelled pRNA or non-specific RNA derived from the multiple cloning site (MCS-RNA) of pBluescript-KS/EcoRI18. Beads were washed successively in buffer A (20 mM Tris-HCl at pH 7.6, 5 mM MgCl2 and 0.2 mM EDTA) containing 200 mM KCl, 400 mM KCl or 600 mM KCl. Retained proteins were quantified by liquid scintillation counting or analysed on 8% denaturing polyacrylamide gels. For RIP experiments, lysates from HEK293T cells were incubated with M2-agarose beads for 3–5h. Beads were washed twice, first with a buffer containing 400 mM NaCl and then with a buffer containing 200 mM NaCl. pRNA and control RNAs (β-actin, GAPDH and 28S rRNA) levels were quantified by RT-qPCR; reactions without reverse transcriptase served as negative controls13.
Analysis of nucleosome positions by LM-PCR.
Analysis of nucleosome positions at the rDNA promoter has been described previously13. Briefly, cells were crosslinked with 1% formaldehyde and permeabilized with 0.05% lysolecitin for 1 min. DNA was digested with MNase (15 U ml−1; Roche) in 150 mM sucrose, 50 mM Tris-HCl at pH 7.4, 50 mM NaCl and 2 mM CaCl2 for 30 min at room temperature. After reversal of crosslinking, mononucleosome-sized DNA was isolated from 2% agarose gels and ligated to linker oligonucleotides (linker S: 5′-gaattcagatc-3′ and linker L: 5′-gcggtgacccgggagatctgaattc-3′) phosphorylated with T4 polynucleotide kinase. Purified DNA was amplified by PCR using a linker primer and a 5′-labelled rDNA-specific primer (−105/−87). PCR products were separated on 8% polyacrylamide gels and analysed by autoradiography.
In vitro acetylation by MOF.
Acetylation reactions contained 0.2 μg HA–MOF, 1 μg hTIP5510-732 and 0.05 μCi [3H]-labelled acetyl-CoA in 20 mM Tris-HCl at pH 8.8, 1.5 mM MgCl2 and 20 mM NaCl. After incubation for 1 h at 30°C, proteins were separated by 15% SDS–PAGE and analysed by fluorography. For mass spectrometric analysis of TIP5 acetylation, the assays were performed in the presence of unlabelled acetyl-CoA (10 μM).
Knockdown of MOF and SIRT1.
Cells (6 × 105) were transfected twice with 100 pmol of short interfering RNA (siRNA) against MOF (5′-gugauccagucucgaguga-3′)14 using the TransIT-TKO transfection Reagent (Mirus). Cells were collected 3 days after the first knockdown. NIH 3T3 cells were infected with a retrovirus encoding shRNA against MOF or SIRT1 and collected after 5 days.
Antibodies against TIP5 (ref. 1) and MOF14 have been described previously. To generate antibodies that specifically recognize mTIP5 acetylated at K633 (anti-mTIP5K633ac), rabbits were immunized with an acetylated mTIP5 peptide (PPKIacKMPELC, acK representing acetylated K633). Commercial antibodies were purchased from the following companies: anti-Flag (M2) antibody from Sigma, anti-H4K20me3, anti-HDAC1 and anti-HA antibodies from Abcam, and antibodies against SIRT1, acetylated histone H4 (H4ac), H4K16ac, H3K4me2, H3K9me2 and H3K27me3 from Upstate Biotechnology.
Error bars denote mean ± s.d. from three to four independent experiments.
Bands were excised from the gel and proteins were digested with trypsin according to standard protocols. Extracted peptides were separated by Nano-HPLC with a linear gradient of water/acetonitrile/formic acid at a flow rate of 60 nl min−1 on a reversed phase column. Lysine-acetylated peptides were detected by precursor-ion-like scanning. Acetylated amino acids were verified by subsequent peptide sequencing using HPLC-MS/MS.
Note: Supplementary Information is available on the Nature Cell Biology website.
We thank S. Smale for providing the vector pQXCIP, L. Guarente for expression vectors encoding SIRT1H355Y, G. Xouri and M. Gentzel for their contribution at the early stages of this work and M. Wilm for his support in the mass spectroscopic analysis. This study was funded by the Deutsche Forschungsgemeinschaft (SFB/Transregio 5, SP 'Epigenetics'), the EU-Network 'Epigenome' and the Fonds der Chemischen Industrie.
About this article
Dosage compensation in Drosophila melanogaster: epigenetic fine-tuning of chromosome-wide transcription
Nature Reviews Genetics (2012)