Published online 4 December 2019 Nucleic Acids Research, 2020, Vol. 48, No. 2 665–681 doi: 10.1093/nar/gkz1127 SIRT1/2 orchestrate acquisition of DNA methylation and loss of histone H3 activating marks to prevent premature activation of inflammatory genes in macrophages Tianlu Li 1,2 , Antonio Garcia-Gomez1,2 , Octavio Morante-Palacios1,2 , Laura Ciudad1,2 , Sevgi Özkaramehmet2 , Evelien Van Dijck2 , Javier Rodrı́guez-Ubreva1,2 , Alejandro Vaquero3 and Esteban Ballestar 1,2,* 1 Epigenetics and Immune Disease Group, Josep Carreras Leukaemia Research Institute (IJC), 08916 Badalona, Barcelona, Spain, 2 Chromatin and Disease Group, Cancer Epigenetics and Biology Programme (PEBC), Bellvitge Biomedical Research Institute (IDIBELL), 08908 L’Hospitalet de Llobregat, Barcelona, Spain and 3 Chromatin Biology Group, Josep Carreras Leukaemia Research Institute (IJC), 08916 Badalona, Barcelona, Spain Received May 10, 2019; Revised November 14, 2019; Editorial Decision November 15, 2019; Accepted November 18, 2019 ABSTRACT novel role for SIRT1/2 to restrict premature activation of proinflammatory genes. Sirtuins 1 and 2 (SIRT1/2) are two NAD-dependent deacetylases with major roles in inflammation. In addition to deacetylating histones and other pro- INTRODUCTION teins, SIRT1/2-mediated regulation is coupled with Macrophages (MACs) are required to respond to a wide other epigenetic enzymes. Here, we investigate the range of environmental stimuli which specify their func- links between SIRT1/2 activity and DNA methy- tions. Historically classified as both pro-inflammatory and lation in macrophage differentiation due to their anti-inflammatory, MACs provide versatile and dynamic relevance in myeloid cells. SIRT1/2 display dras- responses as part of the innate immune system. In order tic upregulation during macrophage differentiation to acquire the corresponding phenotypes of each cell type, and their inhibition impacts the expression of many MACs undergo very specific changes in gene expression that inflammation-related genes. In this context, SIRT1/2 are mediated by the complex interplay between signalling, inhibition abrogates DNA methylation gains, but transcriptional and epigenetic machineries. Deregulation of does not affect demethylation. Inhibition of hyper- these processes results in abnormal MAC function which ultimately forms the basis for many immune diseases. methylation occurs at many inflammatory loci, which Sirtuins, highly conserved proteins that belong to the results in more drastic upregulation of their expres- family of class III histone deacetylases, are key regulators sion upon macrophage polarization following bac- of transcriptional and epigenetic landscape. This family of terial lipopolysaccharide (LPS) challenge. SIRT1/2- proteins has been implicated in a wide range of biologi- mediated gains of methylation concur with de- cal and pathological processes, including metabolism, ag- creases in activating histone marks, and their inhibi- ing and inflammation. One important member of the sir- tion revert these histone marks to resemble an open tuin family, SIRT1, regulates inflammation in myeloid cells chromatin. Remarkably, specific inhibition of DNA (1,2). Originally reported to deacetylate histones H3 and methyltransferases is sufficient to upregulate inflam- H4, SIRT1 substrates have now been expanded to sev- matory genes that are maintained in a silent state by eral transcription factors (TFs), including the p65 sub- SIRT1/2. Both SIRT1 and SIRT2 directly interact with unit of NF-␬B and p53. SIRT1 also determines the epige- netic landscape through interactions with other chromatin- DNMT3B, and their binding to proinflammatory genes modifying enzymes (3–6). SIRT1 is induced in mature is lost upon exposure to LPS or through pharmaco- macrophages by anti-inflammatory conditions, such as the logical inhibition of their activity. In all, we describe a exposure to Th2-cytokines and glucocorticoids (7). In fact, * To whom correspondence should be addressed. Tel: +34 935572800 (Ext. 4250); Fax: +34 934978654; Email: eballestar@carrerasresearch.org  C The Author(s) 2019. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 666 Nucleic Acids Research, 2020, Vol. 48, No. 2 SIRT1 has been extensively described to be integral to MATERIALS AND METHODS macrophage biology through several distinct mechanisms. MO isolation, differentiation and culture For instance, SIRT1 plays a key role in the self-renewal of murine macrophages through cell cycle and longevity Buffy coats used in the study were obtained from anony- pathways (8). Also, in a murine model of atherosclerosis, mous donors via the Catalan Blood and Tissue Bank (Banc in vivo treatment with SIRT1-specific inhibitor EX-527 re- de Sang i Teixits; BST). All donors signed an informed sulted in increased atherosclerotic lesion size through in- consent form prior to donation, and all donations were creased intraplaque macrophage infiltration and impaired in accordance with the guidelines of the World Medical autophagy (9). Finally, macrophages isolated from SIRT1 Association (WMA) Declaration of Helsinki. Peripheral transgenic mice exhibited enhanced polarization toward the blood mononuclear cells (PBMCs) were extracted by layer- M2 axis, coupled with decreased expression of TNF␣ and ing buffy coats on Lymphocytes Isolation Solution (Rafer, IL-1␤ (10). Zaragoza, Spain) and centrifuged at 800 g for 30 min in the Another member of the sirtuin family, SIRT2, tran- absence of braking. After collection, PBMCs were washed siently shuttles to the nucleus during G2 /M transition and with phosphate-buffered saline (PBS) and MOs were pu- shares redundant roles with SIRT1 in the deacetylation of rified using CD14 MicroBeads (Miltenyi Biotec, Bergisch H4K16 and p65 (11,12). Although less described, SIRT2 Gladbach, Germany) following the manufacturers’ instruc- also plays a role in macrophage biology, as SIRT2 ame- tions. Following cell separation, CD14+ MOs were attached liorates LPS-induced iNOS expression in bone marrow to plates by incubation with serum-free medium and then macrophages (13) and its activities are required for the cultured in ␣-minimal essential medium (␣-MEM; Invit- hypo-inflammation phase of sepsis in a mouse model (14). rogen, Carlsbad, CA, USA) containing 10% fetal bovine DNA methylation is another crucial regulator of MAC serum, 1% Penicillin Streptomycin (Gibco Thermo Fisher differentiation, and many key genes have been identified to Scientific, MA, USA) and supplemented with 25 ng/ml hu- undergo rapid demethylation during terminal myeloid dif- man M-CSF (PeproTech EC, London, UF) for MAC differ- ferentiation (15,16), whereas others undergo slower gains of entiation. For MAC activation, 10 ng/ml of LPS (from E. methylation. In addition, key enzymes in maintaining DNA coli O111:B4, Sigma-Aldrich, Darmstadt, Germany) were methylation balance, such as DNA methyltransferase 3A added at day 5 of differentiation for 18 hrs. To inhibit (DNMT3A) (17) and Ten-Eleven-Translocation 2 (TET2), SIRT1, SIRT2 and in combination, 70 ␮M of EX-527 are frequently mutated in myeloid leukaemia (18,19), re- (Sigma-Aldrich, Missouri, USA), 4 ␮M of AGK2 (Sigma- inforcing the importance of DNA methylation in myeloid Aldrich) and 50 ␮M of Cambinol (BioVision, California, differentiation. Furthermore, in specific contexts of ter- USA) were added, respectively, to MOs at day 0 in the pres- minal differentiation, DNMTs are required to yield the ence of M-CSF. Cytokines and inhibitors were refreshed ev- final functional phenotype, as such that downregulation ery two days until the end of the experiment. Corresponding of DNMT3A abolishes immune-suppressive properties of volumes of DMSO were used as control. myeloid-derived suppressor cells (20). In humans, MACs arise from circulating or resident Flow cytometry analysis monocytes (MOs) which are largely present in the blood, spleen and bone marrow. MAC differentiation can be Cells were detached mechanically using cell scrapers and achieved in vitro by the addition of M-CSF to isolated pe- resuspended in staining buffer (PBS containing 2 mM of ripheral blood MOs. M-CSF MACs can be further po- EDTA and 4% FBS), followed by incubation with hu- larized into a pro-inflammatory or anti-inflammatory phe- man FcR Blocking Reagent (Miltenyi Biotec). Cells were notype when exposed to lipopolysaccharide (LPS) or IL- counted and 1 × 105 cells were stained with fluorochrom- 4/IL-10 respectively. The plasticity of these MACs render conjugated antibodies against CD163-FITC (BioLegend, them responsive to further polarization depending on the California, USA), CD86-APC (Miltenyi Biotec), CD83- environmental stimuli encountered, hence they are coined APC (Miltenyi Biotec) and CD80-PE (Miltenyi Biotec) for as M0 MACs. Despite the relevance of sirtuins and DNA 20 min on ice protected from light. Cells were washed methylation in myeloid differentiation and function, their twice with staining buffer and fixed with PBS containing interplay has not been explored in this context until now. 4% paraformaldehyde (Electron Microscopy Sciences, PA, In the present study, we observed a rapid upregulation of USA). 8000 events were acquired using Gallios Flow Cy- SIRT1/2 proteins upon the addition of M-CSF to MOs to tometer (Beckman Coulter, CA, USA) and analyzed using stimulate MAC differentiation, prior to further polariza- the Kaluza software (Backman Coulter). tion. We studied the effects of SIRT1/2 inhibition on global gene expression, DNA methylation and changes in various Quantitative real-time PCR histone marks during macrophage differentiation and ac- tivation. In brief, we report a novel role for SIRT1/2 in Total RNA was isolated utilizing Maxwell® RSC sim- the sequential and hierarchical deposition of DNA methy- plyRNA Cells Kit (Promega, Wisconsin, USA) and re- lation, involving a direct interaction with DNMT3B, and verse transcription was performed using Transcriptor First the inhibition of active histone H3 modifications in pro- Strand cDNA Synthesis Kit (Roche, Basel, Switzerland) inflammatory genes to prevent their premature activation all according to manufacturers’ instructions. LightCycler® during differentiation and prior to the encounter with bac- 480 SYBR Green I Master mix (Roche) and forward and re- terial antigens. verse primers (sequences annotated in Supplementary Table Nucleic Acids Research, 2020, Vol. 48, No. 2 667 S1) were added to the cDNA. Quantitative PCR was carried Co-immunoprecipitation (Co-IP) assays out using LightCycler® 480 II System (Roche) and results Co-IP assays were performed using M-CSF macrophages were analyzed using the corresponding LightCycler® 480 differentiated from CD14+ monocytes for 5 days. Cell ex- Software (Roche). Relative quantification of target genes tracts were prepared in lysis buffer [50 mM Tris–HCl, pH was calculated by the standard double delta Ct method, and 7.5, 1 mM EDTA, 150 mM NaCl, 1% Triton-X-100, pro- all genes were normalized against the housekeeping gene tease inhibitor cocktail (cOmplete™, Merck)] with corre- ribosomal protein L38 (RPL38). All statistical tests were sponding units of Benzonase (Sigma) and incubated at 4◦ C performed using Ct values in which statistical significance for 4 h. 50 ␮l of supernatant was saved as input and diluted was defined as P < 0.05 by paired Student’s t-test. with 2× Laemmli sample buffer (4% SDS, 20% glycerol, 120 mM Tris–HCl, pH 6.8). Supernatant was first incubated Western blotting with PureProteome™ Protein A/G Mix Magnetic Beads Protein expression was visualized by western blotting, (Millipore) for 1 h to remove background signal. Samples in which electrophoresis and transfer were performed as then incubated overnight at 4◦ C with corresponding anti- previously described (16). Membranes were then incubated bodies against DNMT1 (NB100-56519, Novus), DNMT3A with primary antibodies against SIRT1 (07-131, Millipore), (3598, Cell Signaling), DNMT3B (sc-20704, Santa Cruz), SIRT2 (ab51023, Abcam), NF-␬B p65 (sc-372, Santa SIRT1 (07-131, Millipore), and SIRT2 (ab51023, Abcam) Cruz Biotechnologies, TX, USA), p65K310ac (ab19840, according to the specifications of each antibody. Negative Abcam, Cambridge, UK), DNMT1 (NB100-56519, controls were incubated with rabbit (12-370; Merck Milli- Novus), DNMT3A (3598, Cell Signaling), DNMT3B pore) and mouse (12-371; Merck Millipore) IgGs. Subse- (sc-20704, Santa Cruz) and ␤-Actin (ab6276; Abcam). quently, samples were incubated with magnetic beads at 4◦ C HRP-conjugated mouse and rabbit secondary antibodies for 2 h, and beads were then washed three times with lysis were from Thermo Fisher (Massachusetts, USA) and buffer. For sample elution, 100 ␮l of 1× Laemmli sample Abcam respectively. Membranes were developed by Amer- buffer was added to beads. Samples and inputs were de- sham™ ECL™ Western Blotting Detection Reagents (GE natured at 95◦ C in the presence of 1% ␤-mercaptoethanol. Healthcare, Illinois, USA) and signals were detected by Whole-cell extracts and IP samples were visualized by west- Amersham™ Imager 600 (GE Healthcare). Band intensities ern blotting, as described above. were quantified using the Fiji Software, and relative protein expression was calculated by normalizing against ␤-Actin expression. Transfection of primary human MOs For RNA knockdown experiments, we transfected iso- Enzyme-linked immunosorbent assay (ELISA) lated monocytes with SIRT1 Silencer® pre-designed siRNA (AM16708-136457, Ambion, TX, USA), and ON- Released cytokines IL-10 and IL-1␤ were detected using TARGETplus SMARTpool siRNAs against SIRT2 (L- ELISA kits provided by BioLegend and Invitrogen (Cali- 004826-00), DNMT3A (L-006672-01) and DNMT3B (L- fornia, USA), respectively, according to manufacturers’ in- 006395-00), with non-targeting siRNA as control, all pro- structions. Absorbance was measured at 450 nm using Pow- vided by Dharmacon (Colorado, USA). Transfections were erWave™ XS microplate reader (BioTek® Instruments, Ver- carried in MAC differentiation media using Lipofectamine mont, USA). 3000 Reagent (Thermo Fisher Scientific Co., Carlsbad, CA, USA) according to manufacturer’s instructions. Media con- Chromatin immunoprecipitation (ChIP) assays taining siRNA and Lipofectamine were removed after 6–8 ChIP assays were performed as previously described (16). h following transfections, and this was repeated every 48 h Briefly, MOs, DMSO- and cambinol-treated day 5 and in M-CSF differentiation media for 5 days, in which MACs LPS-stimulated MACs were crosslinked with 1% methanol- were subsequently stimulated with LPS for 18 h. Transfec- free formaldehyde (Thermo Fisher) for 15 min and sub- tion efficiencies were examined by western blotting. jected to immunoprecipitation after sonication. ChIP ex- periments were performed using the LowCell# ChIP kit™ Bisulfite modification and pyrosequencing protein A (Diagenode, Liège, Belgium). We used antibod- ies against acetylated lysine 16 of histone H4 (H4K16ac; Total DNA was extracted by the proteinase K protocol. 07-329, Millipore), monomethylated lysine 4 of histone H3 Briefly, cells were lysed using lysis buffer (50 mM Tris pH (H3K4me1; ab8895, Abcam), trimethylated lysine 4 of hi- 8.8, 10 mM EDTA pH 8.3, 100 mM NaCl, 1% SDS) in stone H3 (H3K4me3; CS200580, Millipore), acetylated ly- the presence of Proteinase K (Roche), and nucleic acids sine 27 of histone H3 (H3K27ac; 07-360 Millipore), SIRT1 and lipids were separated by repeated centrifugation. DNA (07-131, Millipore), SIRT2 (ab51023, Abcam), PU.1 (PA5- was precipitated using isopropanol and washed with 75% 17505, Invitrogen) and DNMT3B (sc-20704, Santa Cruz). ethanol. 300 ␮g of isolated DNA were bisulfite (BS)- Corresponding rabbit IgG (Diagenode) is used as con- converted using EZ DNA Methylation-Gold™ Kit (Zymo trol. Protein binding was analyzed by real-time quantitative Research, CA, USA) according to manufacturers’ instruc- PCR, and data are represented as ratio of the enriched frac- tions. BS-converted DNA (∼20 ng) was used as template tion with respect to input. ChIP primers were designed for for further pyrosequencing. Primers for PCR amplification the areas flanking differentially methylated CpGs and their and sequencing were designed using the PyroMark® As- sequences are shown in Supplementary Table S1. say Design Software 2.0 (QIAGEN, Hilden, Germany) and 668 Nucleic Acids Research, 2020, Vol. 48, No. 2 sequences are listed in Supplementary Table S1. Conven- edu/public/html) (21) by applying the basal plus extension tional PCR was carried out using IMMOLASE™ DNA settings to annotate genomic regions. Annotated CpGs in Polymerase kit (Bioline, London, UK) and PCR products the EPIC array were used as background, and GO terms were pyrosequenced with PyroMark® Q24 System (QIA- with P-value < 0.01 and FC > 2 were considered signifi- GEN). cantly enriched. GO analysis of gene expression was carried out using the online tool DAVID (https://david.ncifcrf.gov) Gene expression array and analysis and annotated genes in the Clariom S array were used as background. The Functional Annotation tool was used and The quality of extracted total RNA was first analyzed by the GO terms with P-value < 0.01 was considered significantly 2100 Bioanalyzer System (Agilent, CA, USA), and 100 ng enriched. of excellent quality RNA (RNA Integrity number of > 9) Motif enrichment analyses were performed using was then hybridized on Human Clariom S™ Assay arrays HOMER motif discovery software (22) and analyses were (Thermo Fisher) carried out at/by the High Technology carried out using the statistical programming language Unit (UAT), at Vall d’Hebron Research Institute (VHIR) Unix. Annotated CpGs in EPIC array and genes in Clar- following manufacturers’ instructions. Downstream data iom S array were used as background for methylation and normalization and analysis were performed using statisti- expression analyses respectively. For DNA methylation, cal analysis language R in combination with Bioconductor a window of ±250 bp around the interrogated CpGs was (https://www.bioconductor.org) and CRAN (https://cran.r- applied, and for gene expression, a window of –2000 bp project.org) repository packages. Firstly, background cor- upstream of the TSSs of interrogated genes was applied. A rection was performed using Robust Microarray Analy- P-value < 0.01 defines a motif as significantly enriched. sis (RMA) normalization provided by oligo package. Each Pathway enrichment analyses of gene expression were probe was then annotated using clariomshumantranscript- carried out using GSEA (Gene Set Enrichment Analy- cluster.db package, and the average expression level was cal- sis) software. Genes were ranked by expression log2 FC of culated if more than one probe was mapped to the same Cambinol-treated versus untreated M-CSF MACs, and a gene. Downstream statistical analyses were performed using Normalized Enrichment Score (NES) was calculated for an eBayes-moderated paired t-test provided by limma pack- each hallmark as described in the original article (23). age to obtain log2 fold change (log2 FC), P-value and ad- justed P-value (Benjamini Hochberg-calculated FDR) be- tween sample groups. Genes that displayed statistically sig- Analyses of DNase I hypersensitivity and ChIP-seq of histone nificant tests (P-value < 0.01 and FDR < 0.05) were con- modifications sidered differentially expressed. Heatmaps were generated using calculated z-scores. DNase I hypersensitivity and ChIP-seq data of hi- stone modifications H3K27ac, H3K4me1, H3K4me3, H3K27me3, H3K36me3 and H3K9/14ac of MO and DNA methylation profiling and analysis MACs were downloaded from the BLUEPRINT portal 500 ng of BS-converted DNA were hybridized on In- (http://dcc.blueprint-epigenome.eu/). More than three inde- finium MethylationEPIC BeadChip array (Illumina, Inc., pendent datasets were downloaded for each histone mark San Diego, CA, USA) following manufacturers’ instruc- and ChIP-seq peaks were consolidated, filtering the peaks tions. This array covers over 850 000 (850K) CpG methy- with a q-value < 0.01 and a fold change ≥2 and using the lation sites per sample at single-nucleotide resolution, cov- MSPC program (24) with the parameters -w = 1E−4 , -s = ering 99% of RefSeq genes. Probe fluorescence was detected 1E−8 , and -c = 3. by BeadArray Reader (Illumina, Inc.), and image process- ing and intensity data extraction were performed as previ- ously described (16). For downstream normalization, filter- RESULTS ing and analysis, standard pipelines from minfi and limma SIRT1/2 become rapidly upregulated during macrophage packages were used and data was processed using statis- differentiation and their inhibition upregulates many tical analysis language R. Probes were filtered by detec- inflammation-related genes tion P-value, in which a P-value < 0.01 was considered trustworthy. Background correction was carried by quan- MOs isolated from PBMCs by CD14 positive selection were tile normalization and CpGs in SNPs were excluded. M differentiated in vitro into MACs by the addition of M- and beta values were calculated using minfi, and M val- CSF for 5 days (16). In the absence of further stimuli, these ues were subsequently used for downstream statistical anal- MACs retain a plastic phenotype, in which the cells are sus- ysis, using the limma package, to obtain P-value and ad- ceptible to polarization to either an inflammatory (M1) or justed P-value (Benjamini–Hochberg-calculated FDR) be- anti-inflammatory (M2) phenotype through the addition of tween sample groups by an eBayes-moderated paired t-test. LPS or IL-10/IL-4 respectively (25–27). We observed an in- P-value of < 0.01 and FDR of < 0.05 was considered sta- crease of the SIRT1 protein levels after 24 h following M- tistically significant. CSF addition that continued during the 5 following days, suggesting an essential role of this deacetylase in M-CSF MAC differentiation (Figure 1A). SIRT1 upregulation was Gene ontology, motif and pathway enrichment analyses concomitant with a decrease in its substrate H4K16Ac (Fig- Gene ontology (GO) analysis of DNA methylation was car- ure 1A). The upregulation of SIRT1 occurred in parallel ried out using GREAT online tool (http://great.stanford. with SIRT2 (Figure 1B). The two sirtuins did not experience Nucleic Acids Research, 2020, Vol. 48, No. 2 669 A B M-CSF SIRT1 SIRT2 Monocyte Macrophage SIRT1/β-Actin H4K16ac/H4 5 10 M0 10 * Relative RNA expression 0 1 3 5 LPS (days) Relative band intensity 0 1 3 5 (days) 1.0 4 8 8 SIRT1 SIRT1 6 3 6 SIRT2 β-Actin 4 0.5 β-Actin 2 4 H4K16ac 2 1 2 H4 0 0.0 0 0 0 1 3 5 0 1 3 5 (days) O AC S O AC S LP LP M M M M C IL-10 D CD163 CD86 CD80 CD83 800 10 Secreted protein (pg/ml) 25 1.5 8 Mean Fluorescence 600 * * ** * Intensity (MFI) 400 8 20 6 200 1.0 6 15 * * 4 60 * 4 10 * 0.5 40 2 2 5 20 0 0 0.0 0 0 - - + - + - - + - + - - + - + Cambinol - - + - + Cambinol - + - + LPS - - - + + - - - + + - - - + + - - - + + LPS - - + + MO MAC MO MAC MO MAC MO MAC E IL-1β IL1B IL1A TNF CD163 Relative RNA expression 8 15000 200000 100 ** * Secreted protein (pg/ml) * 150000 * 400 * * 10000 ** 100000 80 6 5000 *** 50000 300 ** ** 60 4 20 300 ** 40 200 15 200 2 10 20 100 100 5 ns 0 0 0 0 0 Cambinol - + - + Cambinol - - + - + - - + - + - - + - + - - + - + LPS - - + + LPS - - - + + - - - + + - - - + + - - - + + MO MAC MO MAC MO MAC MO MAC F M-CSF G Mo M-CSF +Camb Cambinol-downregulated genes Cambinol-upregulated genes Cell cycle Immune response Cambinol- Response to stimulus Chromosome downregulated Cell cycle Immune system process 533 genes Signal transduction Mitosis Cellular response to cytokine stimulus Cell division Category Diabetes type 2 Nucleosome organization Rheumatoid arthritis GAD_DISEASE Chromatin assembly or disassembly Celiac disease GAD_DISEASE_CLASS Systemic lupus erythematosus Atherosclerosis GOTERM_BP_ALL DNA replication Hematological GOTERM_BP_DIRECT Negative regulation of gene expression Systemic lupus erythematosus Interleukin−1 production KEGG_PATHWAY Cambinol- Cancer Immune UP_KEYWORDS upregulated Immune system process 303 genes Myeloid leukocyte activation Regulation of immune system process Innate immune response p53 signaling pathway Toll-like receptor signaling pathway Phagosome Interferon−gamma−mediated signaling pathway Negative regulation of immune system Hematopoietic cell lineage 150 100 50 0 0 5 10 15 20 −2 0 2 z−score −log2(pvalue) H P-value Cambinol-upregulated genes Cambinol-downregulated genes < 0.05 & >= 0.01 < 0.01 & >= 1e−4 NFkB−p65−Rel Fold enrichment of motif Fold enrichment of motif < 1e−4 1.75 Srebp2 E2F 2.0 TF Family N/A 1.50 E2F7 bHLH Usf2 CCAAT Brn1 FXR 1.5 E2F1 DM E2F6 NFY E2F 1.25 NPAS2 ETS COUP−TFII CLOCK Homeobox Ptf1a POU Eomes 1.0 Zf 0 25 50 75 100 0 25 50 75 100 NR RHD Windows containing motif (%) Windows containing motif (%) T−box Figure 1. Inhibition of SIRT1/2 with cambinol during MO-to-MAC differentiation results in an aberrant pro-inflammatory phenotype. (A) A represen- tative western blotting analysis showing SIRT1 and acetylated lysine 16 of histone H4 (H4K16ac) protein expression of peripheral blood CD14+ MOs 670 Nucleic Acids Research, 2020, Vol. 48, No. 2 further changes in their protein levels following LPS stimu- revealed the participation of factors such as NF-kB, which lation (Figure 1B). SIRT1 mRNA levels remained relatively is a known target of SIRT1/2 and relevant to inflammation, stable during this process (Figure 1B), indicating that the in- in the set of upregulated genes (Figure 1H). crease at the protein level was due to post-transcriptional Altogether, SIRT1/2 inhibition during macrophage dif- mechanisms. Contrastingly, SIRT2 RNA became signifi- ferentiation resulted in the acquisition of a more pro- cantly upregulated during MO-to-MAC differentiation in inflammatory signature, suggesting a role for these deacety- parallel to its protein levels (Figure 1B). The rapid upregu- lases in the silencing of inflammatory genes prior to MAC lation of these sirtuins during MAC differentiation are con- polarization. sistent with their described roles in MAC biology, as well as suggesting a potential role during MAC differentiation. SIRT1/2 activity is dispensable for active demethylation Subsequently, we inhibited SIRT1/2 activity utilizing but determines de novo gains of DNA methylation during cambinol, and inhibition efficiency was ascertained by macrophage differentiation the increase in acetylated p65, a target of SIRT1/2 (4), in cambinol-treated MACs (Supplementary Figure S1A). MO to MAC differentiation is accompanied by large-scale Consequently, we observed that cambinol-treated MACs DNA methylation changes, both gains and losses of methy- underwent various phenotypic changes, including a de- lation, which are critical to the acquisition of the final phe- crease in the release of anti-inflammatory cytokine IL-10 notype of MACs (16,28,29). Given that SIRT1/2 partici- both before and after LPS stimulation, and increased re- pates in macrophage differentiation and their inhibition re- lease of IL-1␤ after activation (Figure 1C). We also ob- sults in broad changes in gene expression, we investigated served that cambinol treatment resulted in the downregu- their potential interplay with DNA methylation changes. lation of the macrophage scavenger receptor CD163 and Utilizing high density bead arrays, we observed that 7475 upregulation of inflammatory surface markers including CpGs became demethylated (FDR < 0.05; hypomethy- CD86, CD80 and CD83 (Figure 1D). Further gene expres- lated cluster) upon MAC differentiation from MO, and sion analysis revealed that upon cambinol treatment, MACs this demethylation was not affected when SIRT1/2 were upregulated such proinflammatory genes as IL1B, IL1A inhibited by cambinol, suggesting that they do not par- and TNF before and in some cases after LPS activation, and ticipate in active demethylation (Figure 2A and B, top). downregulated CD163 expression after LPS treatment (Fig- The hypomethylated cluster included CpGs of genes asso- ure 1E). Analyses utilizing specific inhibitors against SIRT1 ciated with innate responses, such as TM7SF4, ACP5 and (EX527) and SIRT2 (AGK2) suggested a stronger effect ADAM12, and displayed enrichment in motifs of the bZIP upon SIRT1 inhibition compared to SIRT2 inhibition (Sup- family of TFs, in concordance to a previous study (Supple- plementary Figure S1B); however, both inhibitors, even in mentary Figure S2B-D) (16). Conversely, 2652 CpGs gained combination, yielded weaker effects than cambinol. de novo methylation (FDR < 0.05; hypermethylated clus- To better characterize the impact of SIRT1/2 inhibi- ter) during MAC differentiation, and surprisingly, inhibi- tion, we hybridized expression microarrays with MOs, M- tion of SIRT1/2 in turn inhibited almost all CpG methy- CSF MACs and cambinol-treated M-CSF MACs and ob- lation gains, which suggested that SIRT1/2 activity is re- served that cambinol treatment during MO-to-MAC differ- quired for MAC-associated DNA hypermethylation (Fig- entiation resulted in the significant (FDR < 0.05) upreg- ure 2A and B, bottom). Analyzing variance between sam- ulation and downregulation of 303 and 533 genes respec- ples using principal component analysis (PCA), we showed tively (Figure 1F). More specifically, many genes associated that for demethylated CpGs, MACs and cambinol-treated with the inflammatory niche were transcriptionally upreg- MACs displayed overlapping PC dimensions. In contrast, ulated when SIRT1/2 was inhibited (Supplementary Table for hypermethylated CpGs, the three cell populations sep- S2). Further gene ontology analyses revealed that upregu- arated along the axis of PC1, in which cambinol-treated lated genes were enriched in categories including immune MACs showed less PC1 variance with MOs than MACs response, myeloid activation, and Toll-like receptor and IL- (Figure 2C). Further analyses of CpG positioning revealed 6 signalling. Regarding downregulated genes, we observed that both hyper- and hypomethylated CpGs were enriched enrichment in categories such as cell cycle, DNA replica- in open seas with a slight enrichment in 5 -UTR compared tion, nucleosome organization, and E2F targets (Figure 1G, to background (Supplementary Figure S2A). GO analysis Supplementary Figure S1C). TF motif enrichment analyses revealed that most of the genes associated with hyperme- thylated CpGs were enriched in functional categories re- ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− stimulated with 25 ng/mL of M-CSF for 0, 1, 3 and 5 days. Relative band intensities of SIRT1 and H4K16ac were normalised against ␤-Actin and H4 re- spectively (right panel) from two independent experiments. (B) Western blot analysis of protein expression of SIRT1 and SIRT2 in MO exposed to M-CSF for 0, 1, 3 and 5 days and activated by the addition of LPS for 18 hrs. RNA expressions of SIRT1 and SIRT2 were normalised against RPL38. (C) Secreted levels of IL-10 and IL-1␤ were analysed by ELISA in MACs differentiated in the presence of DMSO or treated with 50 nM of cambinol. (D) FACS analysis of surface markers CD163 (FITC), CD86 (APC), CD80 (PE) and CD83 (APC) in MOs and M-CSF MACs before and after treatment with LPS for 18 hrs in the presence and absence of cambinol (50 nM). (E) Gene expression analysis of IL1B, IL1A,TNF and CD163 in MOs, cambinol- or DMSO-treated M-CSF MACs differentiated for 5 days and MACs treated with LPS for an additional 18 hrs. (A-E) Statistical significance of at least three independent experiments was calculated by paired student t-test (* p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001). (F) Heatmap representation of z-scores obtained from log2 expression of genes that undergo significant changes (FDR < 0.05) in cambinol-treated (M-CSF + Camb) compared to DMSO-treated (M-CSF) MACs, in which 533 and 303 genes become downregulated (log2FC < 0 and FDR < 0.05) and upregulated (log2FC > 0 and FDR < 0.05) re- spectively. (G) GO analyses of cambinol-downregulated (left) and -upregulated (right panel) genes using the DAVID tool (https://david.ncifcrf.gov). (H) Bubble plot representation of HOMER TF motif enrichment analysis. A window of -2000 bp upstream of TSS was used for all genes analysed. Nucleic Acids Research, 2020, Vol. 48, No. 2 671 A Mo vs M-CSF B C M-CSF FDR < 0.05 Mo M-CSF +Camb Hypomethylated Hypomethylated Normalized DNA methylation -2 0 2 Hypo- MO 7475 CpGs 1.0 0.2 Normalized MAC PC2 (6.68 %) DNA methylation MAC+Camb 0.5 0.0 0.0 −0.2 −0.5 −0.4 −1.0 −0.6 MO MAC MAC −0.4 −0.2 0.0 0.2 +Camb PC1 (81 %) Hypermethylated Hypermethylated Normalized DNA methylation 1.0 0.25 PC2 (8.5 %) 0.5 0.00 0.0 Hyper- −0.25 2652 CpGs −0.5 −0.50 −1.0 MO MAC MAC −0.25 0.00 0.25 0.50 +Camb PC1 (75.44%) D E Hypermethylated ADORA2A RUNX3 IL2RA TNFAIP3 leukocyte migration inflammatory response cg02237342 cg19746802 cg11733245 cg06066908 0.4 0.3 0.25 0.4 leukocyte chemotaxis cytokine receptor activity 0.20 0.3 0.3 0.2 β values myeloid leukocyte activation 0.15 C−X−C chemokine receptor activity 0.2 0.2 0.10 phagocytosis 0.1 negative regulation of histone H3−K36 methylation 0.1 0.1 0.05 T cell activation Ontology 0.0 0.0 0.00 0.0 negative regulation of stem cell differentiation GO Biological Process Genes involved in MyD88:Mal cascade GO Molecular Function JAK3 SLC1A2 ADAMDEC1 CD83 C−MYB transcription factor network MSigDB Pathway NOTCH1 Intracellular Domain Regulates Transcription cg03272225 cg08896849 cg11695852 cg15159503 TRAF6 mediated induction of NFkB upon TLR7/8/9 activation 0.25 0.20 0.20 0.15 IL12 and Stat4 Dependent Signaling Pathway in Th1 0.20 0.15 0.15 β values Genes involved in Toll Receptor Cascades 0.10 histone deacetylase binding 0.15 0.10 0.10 METS affect on Macrophage Differentiation 0.10 0.05 Hematopoietic cell lineage 0.05 0.05 0.05 IL 6 signaling pathway Genes involved in Activated TLR4 signalling 0.00 0.00 0.00 0.00 MO MAC MAC MO MAC MAC MO MAC MAC MO MAC MAC 0 10 20 30 40 50 +Camb +Camb +Camb +Camb −log2(FDR) F G Hypermethylated Hypermethylated DNAse I 2.4 Ets1−distal H3K9me3 FDR P-value Fold Enrichment of motif 1 1e−10 to 1e−40 −100 H3K9/14ac 1e −40 −100 1e to 1e −200 EWS:FLI1−fusion 1e < 1e−100 H3K4me3 2.0 1e−300 EWS:ERG−fusion TF Family H3K4me1 Etv2 RUNX RUNX2 Fli1 bZIP Odds Ratio CEBP ETS H3K36me3 10 1.6 RUNX−AML ETS1 Runt PU.1 GABPA ETV1 0 H3K27me3 RUNX1 EHF −10 ERG H3K27ac −20 0 25 50 75 100 MO MAC Windows containing motif (%) Figure 2. Inhibition of SIRT1/2 inhibition with cambinol affects DNA hypermethylation during MO-to-MAC differentiation. (A) Heatmap representa- tion of significant (FDR < 0.05) DNA methylation changes during MO-to-MAC differentiation comparing DMSO-treated MACs (M-CSF) with MOs. 672 Nucleic Acids Research, 2020, Vol. 48, No. 2 lated to macrophage biology including inflammatory re- we first mapped the hypermethylated CpGs to the near- sponse, cytokine receptor activity and leukocyte migration, est gene, yielding 1612 unique genes (Figure 3A). Surpris- and these genes include ADORA2A, RUNX3, IL2RA, TN ingly, only 63 genes (3.9%) displayed upregulation and 61 FAIP3, JAK3, SLC1A2, ADAMDEC1 and CD83 (Figure genes (3.8%) displayed downregulation mediated by cam- 2D and E). These results suggested that gains of methyla- binol treatment, whereas 92.3% of genes showed no sig- tion at genes associated with these biological processes were nificant gene expression changes despite the inhibition of altered when SIRT1/2 were inhibited. Inspection of the pro- DNA hypermethylation (Figure 3B, right panel). In con- tein levels of DNMTs revealed a significant downregulation trast, majority of genes from the hypermethylated CpG of DNMT3A protein in cambinol-treated MACs, with both cluster did display a significant change in expression dur- DNMT1 and DNMT3B displaying similar trends with- ing the MO-to-MAC differentiation step; however, both out reaching statistical significance (Supplementary Figure upregulation and downregulation were observed (Figure S2E). 3A, left panel), which emphasized the complex relation- Sequences associated with hypermethylated CpGs were ship between DNA methylation and gene expression. Since enriched in a distinct set of TFs, namely members of the the pro-inflammatory phenotype observed in cambinol- ETS family (Figure 2F). Previously, PU.1 and other mem- exposed MACs appeared more pronounced following LPS- bers of the ETS TF family was linked to hypermethylated mediated activation (Figure 1C–E), it is therefore of interest regions in terminal myeloid differentiation model, in which to interrogate the expression of genes associated to hyper- PU.1 was observed to recruit DNMT3B to mediate de novo methylated CpGs following LPS stimulation. Interestingly, DNA methylation (30). Therefore, given these results, it is upon LPS-stimulation, genes such as ADORA2A, RUNX3, conceivable that SIRT1/2 activity is required for ETS TF- IL2RA, TNFAIP3, JAK3, SLC1A2, ADAMDEC1, TLR2, mediated recruitment of DNMTs to hypermethylated CpG INHBA, JUN, TNFAIP8, RUNX1, previously described to sites. be upregulated upon LPS stimulation (38), displayed fur- SIRT1 determines the epigenetic landscape not only ther upregulation in cells exposed to cambinol (Figure 3C), through its ability to directly deacetylate H1, H3 and H4 indicating that SIRT1/2 determined, at least in part, their (31–33), but also through the post-translational regulation repression, and DNA methylation might participate in tight of other histone-modifying enzymes (34). Given the role control of their repressive status prior to polarization to a of SIRT1 in histone modifications and its effect on DNA pro-inflammatory phenotype. methylation, we then analyzed the enrichment of various histone marks in both MOs and MACs, obtained from SIRT1/2 play concurrent roles in de novo gains of DNA publically available Blueprint ChIP-seq data (http://dcc. methylation and histone modifications in macrophage differ- blueprint-epigenome.eu; refer to Materials and Methods entiation and activation for the consolidation of data), in the hypermethylated CpG cluster. We observed that there was a significant enrich- Our results showed that the activities of SIRT1/2 deter- ment in DNase I hypersensitivity in MOs which decreased mined de novo DNA methylation during MO-to-MAC dif- in MACs, coupled with significant enrichment of H3K4me1 ferentiation. It is likely that acquisition of DNA methy- in both cell types (Figure 2G). These observations were in lation is accompanied by additional histone modifica- accordance to what has been previously described in regards tions. Hence, we analyzed the enrichment of histone to the relationship between DNA methylation and chro- marks H3K9/14ac, H3K4me3, H3K4me1, H3K36me3, matin states, in which gains of DNA methylation in regions H3K27me3 and H3K27ac, as well as DNase I hypersen- of enhancers correspond to heterochromatin (35,36). sitive sites, all obtained from the Blueprint database, in MOs and MACs within a window of 4000 bp centring around hyper- and hypomethylated CpGs (Figure 4A). SIRT1/2 activity is required to prevent premature expression We observed that for CpGs whose hypermethylation dur- of pro-inflammatory genes upon LPS stimulation through ing MO-to-MAC differentiation was blocked in cambinol- premeditated DNA hypermethylation treated MACs, they appeared to undergo a dramatic de- The relationship between DNA methylation and gene ex- crease in DNase I hypersensitivity and reduced H3K4me3 pression has been nothing short of controversial. It is and H3K27ac histone marks, which were in accordance to widely accepted that DNA methylation inhibits gene ex- a more closed chromatin profile (39,40). We also observed pression, however mounting evidence suggest a more com- a slightly higher enrichment of H3K4me1 in MACs with plicated reciprocal relationship (reviewed by Jones et al. respect to MOs, which marks primed enhancers (41). Con- (37)). To evaluate the relationship between gene expression versely, we observed that for CpGs whose MO-to-MAC and DNA methylation in relation to SIRT1/2 inhibition, demethylation was not affected by SIRT1/2 inhibition, they ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− (B) Violin plots depicting the distribution and probability density of z-scores of hypo- (upper) and hypermethylated (lower panel) CpGs. (C) Principal component analysis (PCA) of hypo- and hypermethylated CpGs of the three cell populations (Mo, M-CSF and M-CSF + Camb). (D) GO analysis of hypermethylated CpGs, in which CpGs were mapped to genes using GREAT online tool (http://great.stanford.edu/public/html) by applying the basal plus extension settings, using EPIC array as background. (E) Representation of the beta values of selected hypermethylated CpGs that map to relevant genes. (F) HOMER TF motif enrichment analysis of hypermethylated CpGs were carried out using a window of ±250 bp centring around the CpGs. CpGs annotated in the EPIC 850K array were used as background. (G) Histone marks enrichment analysis of hypermethylated CpGs were carried out by crossing DNase-seq and ChIP-seq data of H3K9me3, H3K9/14ac, H3K4me3, H3K4me1, H3K36me3, H3K27me3 and H3K27ac of MOs and MACs downloaded from the Blueprint database (refer to Materials and Methods for data processing). Nucleic Acids Research, 2020, Vol. 48, No. 2 673 A B M-CSF Mo M-CSF +Camb Mo vs M−CSF M-CSF vs M−CSF + Camb Hyper- 4 1612 genes 5 2 0 0 log2 FC expression −2 −5 −4 −0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4 differential β value −2 0 2 z−score C ADORA2A RUNX3 IL2RA TNFAIP3 JAK3 SLC1A2 Relative gene expression 15 50 250 100 25 250 * * * ** 40 200 80 * 20 200 10 30 150 60 15 150 *** * 20 100 40 10 100 5 ** 10 50 * 20 5 50 * ** ** 0 0 0 0 0 0 ** Cambinol - + - + - + - + - + - + - + - + - + - + - + - + LPS - - + + - - + + - - + + - - + + - - + + - - + + MO MAC MO MAC MO MAC MO MAC MO MAC MO MAC ADAMDEC1 TLR2 INHBA JUN TNFAIP8 RUNX1 Relative gene expression 40 100 150 2.5 50 20 * * ** * * 80 2.0 * 40 30 15 100 60 1.5 30 20 10 * 40 1.0 * 20 * 50 * ** 10 5 20 ** 0.5 10 ** ** 0 0 0 0.0 0 0 Cambinol - + - + - + - + - + - + - + - + - + - + - + - + LPS - - + + - - + + - - + + - - + + - - + + - - + + MO MAC MO MAC MO MAC MO MAC MO MAC MO MAC Figure 3. Treatment with cambinol upregulates pro-inflammatory genes associated to hypermethylated CpGs following MAC activation with LPS. (A) Color-coded heatmap representation of gene expression of hypermethylated CpGs of the three cell populations (MO, M-CSF and M-CSF + Camb). 2652 hypermethylated CpGs were mapped to 1612 unique genes utilizing GREAT online tool with the setting of mapping to the nearest gene. (B) Scatter plots representing log2 FC of gene expression (y-axis) and difference in beta (beta; x-axis) of M-CSF in respect to MOs (left panel) and M-CSF + Camb in respect to M-CSF MACs (right panel). Genes that show significant (FDR < 0.05) upregulation or downregulation are depicted in red and blue respectively. (C) Gene expression analyses of ADORA2A, RUNX3, IL2RA, TNFAIP3, JAK3, SLC1A2, ADAMDEC1, TLR2, INHBA, JUN, TNFAIP8, RUNX1 in MOs, MACs and LPS-activated MACs in the presence and absence of 50 ␮M cambinol. Bar graphs represent mean and standard deviation of relative expression of each gene of four independent experiments normalized against RPL38. Statistical significance was calculated using paired Student’s t-tests (*P-value < 0.05, **P-value < 0.01 and ***P-value < 0.001). 674 Nucleic Acids Research, 2020, Vol. 48, No. 2 A Hypermethylated CpGs Hypomethylated CpGs DNAseI H3K27ac DNAseI H3K27ac 4 6 4 Cell Type 3 1.2 4 3 MO 2 1.0 MAC 0.8 2 2 1 0.6 1 H3K4me1 H3K4me3 H3K4me1 H3K4me3 1.0 0.8 5 0.8 9 0.7 Odds Ratio 4 0.6 0.6 6 3 0.5 0.4 3 0.4 2 -2000 -1000 0 1000 2000 -2000 -1000 0 1000 2000 -2000 -1000 0 1000 2000 -2000 -1000 0 1000 2000 Distance (bp) B Scale 50 kb Scale 100 kb chr22 24,400,000 24,450,000 chr1 24,850,000 24,900,000 24,950,000 25,000,000 DHS DHS H3K27ac H3K27ac H3K4me1 H3K4me1 H3K4me3 H3K4me3 CpG CpG ChIP amplicon ChIP amplicon ADORA2A RUNX3 Scale 50 kb Scale 50 kb chr10 6,050,000 6,100,000 chr19 17,800,000 17,850,000 DHS DHS H3K27ac H3K27ac H3K4me1 H3K4me1 H3K4me3 H3K4me3 CpG CpG ChIP amplicon ChIP amplicon IL2RA JAK3 C H4K16ac H3K4me1 H3K4me3 H3K27ac 800 4000 100 80 * Relative enrichment * IgG 80 * ADORA2A 600 3000 * 60 H4K16ac 60 H3K4me1 400 2000 40 40 H4K4me3 200 1000 20 H3K27ac 20 0 0 0 0 1500 4000 200 * 200 ** Relative enrichment 3000 150 RUNX3 1000 150 * 2000 100 * 100 500 1000 50 50 0 0 0 0 600 2500 400 80 Relative enrichment 2000 300 60 IL2RA 400 1500 * 200 40 1000 200 100 20 500 0 0 0 0 800 4000 150 80 Relative enrichment * * * * *** 600 3000 60 JAK3 100 400 2000 40 50 * 200 1000 20 0 0 0 0 MO MAC MO MAC MO MAC MO MAC Cambinol - - + - + - - + - + - - + - + - - + - + LPS - - - + + - - - + + - - - + + - - - + + Figure 4. MO-to-MAC DNA methylation gains occur concomitantly with a loss of DNase I hypersensitivity and activating histone marks, and SIRT1/2 inhibition upregulates activating histone marks. (A) DNase-seq and ChIP-seq data of H3K27ac, H3K4me1 and H3K4me3 of MOs and MACs were downloaded from the Blueprint database (refer to Materials and Methods for details). Odds ratios were calculated for bins of 10 bp up to ±2500 bp centering around hypo- (left panel) and hypermethylated (right panel) CpGs in which CpGs annotated in the EPIC 850K array were used as background. (B) Schematic representation of DNase I hypersensitivity, H3K27ac, H3K4me3 and H3K4me1 marks near hypermethylated CpGs in ADORA2A, RUNX3, IL2RA and JAK3. MOs are represented in red and MACs are represented in blue. (C) Chromatin immunoprecipitation (ChIP) of H4K16ac (yellow), H3K4me1 (green), H3K4me3 (light blue) and H3K27ac(dark blue) MOs, MACs and LPS-activated MACs in the presence and absence of 50 ␮M cambinol. ChIP amplicon positions in relation to hypermethylated CpGs and genes ADORA2A, RUNX3, IL2RA and JAK3 are depicted in (B). Three independent experiments were performed and statistical significance was calculated using paired Student’s t-tests (*P-value < 0.05, **P-value < 0.01 and ***P-value < 0.001). Nucleic Acids Research, 2020, Vol. 48, No. 2 675 underwent an increase in DNase I accessibility, concur- ated in the presence of cambinol and DAC displayed an ef- rently with higher enrichment in H3K4me3, H3K27ac and fective inhibition in gains of DNA methylation both before H3K4me1 in MACs compared to MOs. No significant dif- and after LPS exposure (Figure 5A). Secondly, in parallel ferences were observed for H3K27me3, H3K36me3 and to what was observed in cambinol-treated cells, we also ob- H3K9/14ac, which was due, in part, to lack of data to served a significant upregulation of such LPS-induced genes generate sufficient and reliable reads (Supplementary Fig- as ADORA2A, RUNX3, TNFAIP3 and SLC1A2 in MACs ure S3A). Altogether, these results suggest that gains and treated with DAC following LPS stimulation (Figure 5B). losses of DNA methylation during MO-to-MAC differen- Other genes such as IL2RA, JAK3 and ADAMDEC1 dis- tiation occur concomitantly with changing histone marks played a trend in their upregulation without reaching sta- and chromatin accessibility, in which SIRT1/2 inhibition tistical significance (Figure 5B). Interestingly, we did not specifically inhibited MO-to-MAC gains of methylation. observe an upregulation in these genes prior to LPS stim- Hence, it is not unbefitting to envision that SIRT1/2 may ulation (Supplementary Figure S4A), suggesting that, sim- have additional implications in the modification of the chro- ilar to cambinol, LPS stimulus was required for their aber- matin landscape. rant expression. Finally, DAC-treated MACs secreted sig- To further understand the hierarchy between the estab- nificantly less IL-10 following LPS stimulus (Figure 5C), lishment of histone marks by SIRT1/2 and DNA methyla- however no changes in IL-1␤ secretion were observed. Al- tion and their effect on gene expression, we performed ChIP together, these results suggested that the inhibition of DNA analyses of H3K4me1, H3K4me3, H3K27ac and H4K16ac methylation alone by DAC is sufficient to induce tran- in MOs, MACs and cambinol-treated MACs, both before scriptional activation of proinflammatory genes in the pres- and after LPS activation, in LPS-responsive genes with ence of functional sirtuins, and, to some extent, mediated a hypermethylated CpGs. First, we observed a general loss change in the final functional phenotype of MACs. of active histone marks in the MO-to-MAC differentia- tion process, which was in accordance with what was ob- Both SIRT1 and SIRT2 interact with DNMT3B and are served in the Blueprint data (Figure 4B). Second, we ob- recruited to genes that become hypermethylated during served that SIRT1/2 inhibition by cambinol not only re- macrophage differentiation sulted in enrichment of H3K16ac but also in H3K4me3 and H3K27Ac at several proinflammatory loci, both be- To better understand the relationship between SIRT1/2 and fore and following LPS challenge (Figure 4C), and these the DNA methylation machinery in MACs in relation to included ADORA2A, RUNX3 and JAK3. A similar trend the control of proinflammatory genes, we then performed was observed in IL2RA, however it did not reach statisti- co-immunoprecipitation experiments to assess the potential cal significance. These results indicated that SIRT1/2 are interaction between SIRT1/2 with DNMTs. Immunopre- required to maintain a closed chromatin structure at proin- cipitation, performed using M-CSF MACs differentiated flammatory loci through DNA hypermethylation and the from MOs for 5 days, showed that both SIRT1 and SIRT2 inhibition of activating histone marks. Interestingly, cambi- co-immunoprecipitated with DNMT3B, but not DNMT1 nol treatment already results in histone modifications prior and DNMT3A (Figure 6A). Interestingly, DNMT3B and to MAC activation, however, aberrant upregulation of these DNMT3A also seemed to interact with each other (Fig- genes was mostly observed following LPS stimulation (Fig- ure 6A). Such interaction has been previously described in ure 3C). These results suggest that alteration of the histone other cell types (43,44). Furthermore, the interaction be- modification status was not sufficient to induce aberrant ex- tween DNMT3B and SIRT1/2 was not affected by cambi- pression, hence an additional stimulus was required. In all, nol treatment (Figure 6B). Subsequently, we tested the re- our results proved that SIRT1/2 are essential in the estab- cruitment of sirtuins to inflammatory loci that become hy- lishment of chromatin structure at pro-inflammatory genes, permethylated during MAC differentiation in the presence as their functions are not only critical to achieve correct of cambinol in unstimulated and LPS-activated MACs. We gains of DNA methylation during MO-to-MAC differenti- first observed an increase in SIRT1 and SIRT2 binding ation but is also required for both the loss of active histone to SLC1A2, TNFAIP3 and ADORA2A loci upon MO-to- marks and their maintenance during differentiation and ac- MAC differentiation, in which the complex appeared to tivation respectively. dissociate following LPS stimulation. Second, we observed that their binding was abrogated upon inhibition with cam- binol both before and after LPS stimulation (Figure 6C). Inhibition of DNA methylation alone is sufficient to aber- Finally, and remarkably, we also observed binding of PU.1 rantly upregulate proinflammatory genes in LPS-activated to these genes, in which the pattern of binding of PU.1 was MACs very similar to SIRT1 and 2 (Figure 6C). We were unable to The strict association between the cambinol-mediated aber- detect binding of DNMT3A/B to these sequences, however, rant upregulation of genes following encounter with LPS we could not discard that this was a technical limitation due stimulus and their previous inhibition of DNA methylation to the quality of the DNMT antibodies. led us to speculate on the direct relevance of DNA methyla- Utilizing specific pharmacological inhibitors targeting tion in the transcriptional status of pro-inflammatory genes SIRT1 and SIRT2, we observed that inhibition of SIRT1 following activation. To test such hypothesis, we differen- may have a stronger effect on gene expression than SIRT2 tiated MACs in the presence of a bona fide DNMT in- (Supplementary Figure S1C). Thus in order to assess the hibitor, 5-aza-2 -deoxycytidine (DAC) (42). Firstly, utiliz- specific contribution of sirtuins to DNA hypermethylation ing pyrosequencing, we confirmed that MACs differenti- and expression of inflammatory genes, we downregulated 676 Nucleic Acids Research, 2020, Vol. 48, No. 2 A MO Untreated DMSO Cambinol (50 μM) 5’-aza (10 μM) TNFAIP3 RUNX3 IL2RA ADORA2A ADORA2A JAK3 cg06066908 cg19746802 cg11733245 cg23763137 cg02237342 cg03272225 40 25 25 30 30 ** * ** ** % DNA methylation 30 ** 20 * 10 20 * * * ** *** * 20 * 20 ** ** 15 ** 15 ** ** * 20 * ** 10 5 10 10 10 10 5 5 0 0 0 0 0 MO MAC LPS MO MAC LPS MO MAC LPS MO MAC LPS MO MAC LPS MO MAC LPS B TNFAIP3 RUNX3 IL2RA ADORA2A JAK3 SLC1A2 ADAMDEC1 100 200 40 8 30 30 ** * * Relative gene expression * 100 *** ** * * ** 20 80 150 30 6 50 20 10 60 100 20 4 10 4 40 8 3 10 6 50 10 2 20 2 4 2 1 0 0 0 0 0 0 0 C IL-1β IL-10 25 4000 Secreted protein (pg/mL) *** ** 20 * 3000 15 2000 10 5 1000 0 0 Figure 5. Inhibition of DNA methylation alone is sufficient to upregulate gene expression following LPS exposure. (A) Pyrosequencing of hypermethylated CpGs mapped to TNFAIP3, RUNX3, IL2RA, ADORA2A, and JAK3 in MOs (gray), MACs and LPS-activated MACs in the presence of DMSO (green), 50 ␮M cambinol (orange) and 10 ␮M DAC (light orange) or remained untreated (light green). Untreated MACs were controls for DAC and DMSO- treated MACs were controls for cambinol. Three independent experiments were performed. (B) Gene expression analysis of TNFAIP3, RUNX3, IL2RA, ADORA2A, JAK3, SLC1A2, and ADAMDEC1 of MACs differentiated in the presence of 50 ␮M of cambinol or 10 ␮M of DAC and subsequently activated with 10 ng/mL of LPS for 18 hrs. Gene expression were normalized against housekeeping gene RPL38 and represented as fold changes against control (MACs treated with corresponding volumes of DMSO for cambinol or untreated for DAC). Four independent experiments were performed. (C) ELISA analyses of IL-1␤ and IL-10 secretion in M-CSF MACs differentiated for 5 days and activated with 10 ng/mL LPS for an additional 18 hrs in the presence of 50 ␮M of cambinol or 10 ␮M of DAC. DMSO-treated and untreated MACs were used as controls. Four independent experiments were performed. All statistical analyses were performed using paired Student’s t-tests (*P-value < 0.05, **P-value < 0.01 and ***P-value < 0.001). SIRT1 and SIRT2 using siRNAs. We were able to downreg- SIRT2. These observations suggest that SIRT1 is the main ulate SIRT1 and SIRT2 with distinct efficiency, achieving player mediating the acquisition of DNA methylation dur- 65% and 95% reduction in protein expression respectively ing MAC differentiation. (Figure 6D). We observed that upon SIRT1 knockdown, there was significant inhibition in the acquisition of MO-to- DISCUSSION MAC DNA methylation for CpGs annotated to TNFAIP3, JAK3, RUNX3 and ADORA2A. Conversely, SIRT2 knock- In this study, we report a novel role for SIRT1/2 in down did not appear to have an effect on DNA methyla- macrophage differentiation, in which these deacetylases act tion, except in the case of JAK3 (Figure 6E). Knockdown as a key element in establishing the hierarchy of epige- of either sirtuins did not have an effect on hypomethy- netic modifications involving acquisition of de novo DNA lated CpGs (Supplementary Figure S4A). Moreover, the methylation through direct DNMT3B interaction, linked inhibition of DNA methylation gain observed in ΔSIRT1 to H3K4me3 demethylation and H3K27ac deacetylation, MACs correlated with an aberrant upregulation of the same to prevent premature activation of pro-inflammatory genes genes following LPS stimulation, albeit some did not reach during MAC differentiation and prior to encounter with statistical significance (Figure 6F). Similar to what was bacterial antigens. observed with pharmacological inhibitors, SIRT1 down- Our initial observation showed that both SIRT1 and regulation appeared to have stronger impact compared to SIRT2 undergo a drastic upregulation concomitant with Nucleic Acids Research, 2020, Vol. 48, No. 2 677 A B Input Pull-down Input Pull-down B B T3 T3 T1 T2 T1 T2 A B M M T1 T3 T3 R R R R G T1 T2 N N M M M SI SI SI SI Ig D D R R G - + - - - - + + + N N N SI SI Camb Ig D D D DNMT1 170kDa DNMT3B DNMT3A 130kDa SIRT1 DNMT3B 100kDa SIRT1 110kDa C D ADORA2A TNFAIP3 SLC1A2 ChIP amplicon CpG site TSS IgG A 5 8 10 N SIRT1 Relative enrichment R T N si 4 SIRT2 8 6 PU.1 SIRT1 3 6 β-actin 4 2 4 1.00 0.35 2 1 2 0 0 0 A N R T 1.5 1.5 1.5 N si Relative enrichment SIRT2 1.0 1.0 1.0 β-actin 1.00 0.05 0.5 0.5 0.5 0.0 0.0 0.0 3 6 6 Relative enrichment 2 4 4 1 2 2 0 0 0 MO MAC MO MAC MO MAC Cambinol - - + - + - - + - + - - + - + LPS - - - + + - - - + + - - - + + E SLC1A2 TNFAIP3 JAK3 RUNX3 ADORA2A ADORA2A cg08896849 cg06066908 cg03272225 cg19746802 cg23763137 cg02237342 25 15 15 20 20 15 * * % DNA methylation 20 * * * 15 * 15 10 10 10 15 10 10 10 5 5 5 5 5 5 0 0 0 0 0 0 O T T1 T2 O T T1 T2 O T T1 T2 O T T1 T2 O T T1 T2 O T T1 T2 N N N N N M M M M M N M R R R R R IR R R R R IR IR SI SI SI SI SI SI SI SI SI S S S si si si si si si si si si si si si F SLC1A2 TNFAIP3 JAK3 RUNX3 ADORA2A 1.0 8 8 2.5 10 Relative gene expression 0.8 * 2.0 8 * 6 6 * 0.6 1.5 6 4 4 0.4 1.0 4 2 2 0.2 0.5 2 0.0 0 0 0.0 0 M O N T T1 T2 M O N T T1 T2 M O N T T1 T2 M O N T T1 T2 M O N T T1 T2 R R R R R R R R R R SI SI SI SI SI SI SI SI SI SI Figure 6. SIRT1 and 2 interact with DNMTs to drive DNA hypermethylation at inflammatory loci. (A) Co-immunoprecipitation assays were per- formed in MOs differentiated to MACs in the presence of M-CSF for 5 days. Protein extracts were immunoprecipitated utilising anti-DNMT1, - 678 Nucleic Acids Research, 2020, Vol. 48, No. 2 MAC differentiation, and that the effects of SIRT1/2 in- pluripotentency enhancers (51). Finally, DNA methylation hibition during this process impact the expression levels of has been described to inversely correlate with acetylated many genes related to the immune properties of these cells, H3K27 deposition, a classical marker for enhancers (39), with a particular effect on proinflammatory genes mediat- however, recent studies also identified bivalent regions, in ing their aberrant upregulation. These observations support which H3K27ac co-exists with DNA methylation, although a major role of SIRT1/2 in maintaining a repressed status this accounts for a minority of CpGs (38%) (52). Our for these genes, most likely through the direction interaction findings reinforce the connection between DNA methyla- with DNMTs, which is in concordance to what has been tion and H3 modifications, in which DNA methylation lev- observed in other studies. Notably, SIRT1 has been previ- els inversely correlate with H3K4me3 and H3K27ac (Fig- ously described by Park et al. to exert anti-inflammatory ure 4), as such that the inhibition of SIRT1/2 mediates effects in RA macrophages through the inhibition of NF- higher enrichment of these two histone marks concurrently ␬B signalling, and SIRT1 Tg-mice displayed reduced M1 with a decrease in DNA methylation prior to encounter polarization (10). Individual siRNA knockdown of SIRT1 with bacterial antigen. Furthermore, CpGs that become and SIRT2 showed that the former mediated loss of DNA demethylated upon SIRT1/2 inhibition also gain H4K16ac, methylation and upregulated proinflammatory gene expres- which is coherent with the role of SIRT1/2 in its deacetyla- sion while the latter had weaker significant effects. It is tion. Altogether, these results indicate that the activities of therefore likely that SIRT1 is the major contributor to the SIRT1/2 are essential in the maintenance of a silent chro- phenotype of M-CSF macrophages while SIRT2 may offer matin, defined by DNA hypermethylation and low levels of little or no contribution. H3K4me3, H3K27ac and H4K16ac, and their inhibition re- Interestingly, the analysis of DNA methylation changes sults in chromatin aperture. in relation to inhibition and downregulation of SIRT1/2 The dynamics between SIRT1/2 and DNMTs has been shows that only gains of DNA methylation are affected, partly elucidated by our results, in which we have observed in contrast to DNA demethylation, which is identical to a direct interaction between SIRT1/2 and DNMT3B, which those without SIRT1/2 inhibition. Remarkably, the group in turn interacts with DNMT3A. Other studies had pre- of genes that gain methylation during macrophage differ- viously linked SIRT1 activity with DNMTs. For instance, entiation, almost all of which are affected by SIRT1/2 in- O’Hagan et al. showed that SIRT1 recruitment to induced hibition, associate with additional epigenetic features in- double-strand breaks was dependent on DNMT1, and that cluding a decrease in H3K27ac and H3K4me3 and an DNMT1, DNMT3B and SIRT1 form part of a large multi- increase in H3K4me1. These observations indicate that protein complex following oxidative DNA damage (43). these three histone marks are linked to the observed gains We observe that the binding of SIRT/DNMT3B complex of DNA methylation. Different studies have established to chromatin appears to be dependent on SIRT1/2 ac- links between DNA methylation and histone modifications. tivity, and it is possible that other transcription factors, First, the presence of DNA methylation has originally been such as PU.1, may target the complex to repress pro- described to prevent methylation of H3K4 (45). Second, inflammatory genes in M0 MACs. Furthermore, the inter- treatment with DNMT inhibitors were shown to increase action of SIRT1/2 with DNMTs may also participate in the H3K4me2 and H3K4me3 occupancy at target promoters direct deacetylation of these enzymes to modify their ac- and gene bodies, as well as resulting in a dramatic increase tivities in a similar way that has been observed for SIRT1 in enhancer activity (46). Furthermore, structural and bio- and DNMT3L (53). Additionally, SIRT1 has also been in- chemical studies have shown a direct interaction between directly linked to DNA methylation through its activities the catalytic domain of de novo DNA methyltransferases of modifying histones. One such example was described by with unmethylated H3K4 histone peptide leading to the cat- Aguilar-Arnal et al. where SIRT1 was observed to directly alytic activation of DNMT3A (47,48). This interaction was deacetylate MLL1 and modulate H3K4 tri-methylation in turn inhibited by H3K4 methylation and histone acety- along the circadian cycle (54). Since H3K4 methylation play lation (49). Taken together, evidence suggests an inverse re- concerted roles with DNA methylation, it is therefore plau- lationship between DNA methylation and H3K4 methyla- sible that SIRT1 can modulate DNA methylation through tion. Third, histone remodelers of H3K4 have been linked histone modifications. to DNMT3 recruitment and activity, in which MLL4 en- The hierarchy of DNA methylation, histone modifica- hances DNMT3A-catalyzed DNA methylation at super- tions and gene expression has been proven to be extremely enhancers (50), and contradictorily LSD1 has been de- complex, in which contradictory literature has described scribed to function in a complex with DNMT3 to medi- DNA methylation as both a cause and a consequence of ate both H3K4 demethylation and DNA methylation at changes in gene expression. While it is generally accepted ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− DNMT3A, -DNMT3B, -SIRT1 and –SIRT2 antibodies, in which IgG was used as a negative control and total protein extract was used as input. (B) Co-immunoprecipitation of MACs differentiated in the presence of DMSO or 50 ␮M cambinol for 5 days. (C) Chromatin immunoprecipitation of two independent experiments of SIRT1 (pink), SIRT2 (red) and PU.1 (blue) was performed in isolated monocytes and MACs differentiated in the presence of M-CSF for 5 days in the presence and absence of cambinol, before and after LPS stimulation for 18 hrs. (D) MOs were transfected with 100nM of corre- sponding siRNA and differentiated to MACs in the presence of M-CSF for 5 days, as described in Materials and Methods. Annotated numbers indicate relative band intensities compared to non-targeting control (NT). (E) Pyrosequencing of hypermethylated CpGs in MOs and MACs transfected with NT, siSIRT1 and siSIRT2 prior to LPS activation. (F) Gene expression of SCL1A2, TNFAIP3, JAK3, RUNX3 and ADORA2A in MOs and LPS-activated MACs transfected with NT, siSIRT1 and siSIRT2, as normalised against RPL38. (D-F) Statistical significance of at least three independent experiments was calculated using paired student t-tests (* p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001). Nucleic Acids Research, 2020, Vol. 48, No. 2 679 that DNA methylation results in gene repression, however, FUNDING recent studies have provided conflicting evidence in their CERCA Programme/Generalitat de Catalunya; E.B. was relationship. One study involving 62 unrelated individuals funded by the Ministry of Science, Innovation and Uni- concluded that DNA methylation contributes to only 8% versities (MCIU) [grant numbers: SAF2014- 55942-R, of inter-individual gene expression variation (55). Similarly, SAF2017-88086-R; AEI/FEDER, UE]. Funding for open another study shows that the presence of DNA methyla- access charge: Ministry of Science, Innovation and Univer- tion is insufficient to transcriptionally repress promoters sities (MCIU) [SAF2017-88086-R; AEI/FEDER, UE]. (56). This conclusion was, however, recently refuted by Ko- Conflict of interest statement. None declared. rthauer et al., as the authors show that, by utilizing differ- ent statistical strategies that include statistical inference of differentially methylated regions and additional normaliza- REFERENCES tion techniques, promoter DNA methylation indeed corre- late negatively with gene expression (57). Nevertheless, the 1. Schug,T.T., Xu,Q., Gao,H., Peres-da-Silva,A., Draper,D.W., Fessler,M.B., Purushotham,A. and Li,X. (2010) Myeloid deletion of role of DNA methylation during a cell activation event, such SIRT1 induces inflammatory signaling in response to environmental as encounter with bacterial antigens, is still unknown. In a stress. Mol. Cell Biol., 30, 4712–4721. study conducted in our laboratory, Vento-Tormo et al. show 2. Woo,S.J., Lee,S.-M., Lim,H.S., Hah,Y.-S., Jung,I.D., Park,Y.-M., that a previous demethylation step, together with histone Kim,H.-O., Cheon,Y.-H., Jeon,M.-G., Jang,K.Y. et al. (2016) Myeloid modifications, is required for correct upregulation of in- deletion of SIRT1 suppresses collagen-induced arthritis in mice by modulating dendritic cell maturation. Exp. Mol. Med., 48, e221. flammatory genes during dendritic cell activation (28). Con- 3. Vaziri,H., Dessain,S.K., Ng Eaton,E., Imai,S.I., Frye,R.A., trastingly, a more recent article shows that upregulation of Pandita,T.K., Guarente,L. and Weinberg,R.A. (2001) hSIR2(SIRT1) proinflammatory genes precedes DNA demethylation fol- functions as an NAD-dependent p53 deacetylase. Cell, 107, 149–159. lowing dendritic cell activation, however, this behavior only 4. Yeung,F., Hoberg,J.E., Ramsey,C.S., Keller,M.D., Jones,D.R., Frye,R.A. and Mayo,M.W. (2004) Modulation of NF-␬B-dependent accounts for less than 10% of all differentially expressed transcription and cell survival by the SIRT1 deacetylase. EMBO J., genes and differentially methylated CpGs (58). In this study, 23, 2369–2380. we provide a direct link between DNA methylation, histone 5. Zhao,X., Sternsdorf,T., Bolger,T.A., Evans,R.M. and Yao,T.-P. modifications and gene expression. We show that SIRT1/2- (2005) Regulation of MEF2 by histone deacetylase 4- and SIRT1 dependent establishment of correct chromatin state prior to deacetylase-mediated lysine modifications. Mol. Cell Biol., 25, 8456–8464. LPS challenge is essential for proper and controlled expres- 6. Bosch-Presegué,L., Raurell-Vila,H., Marazuela-Duque,A., sion of proinflammatory genes following macrophage acti- Kane-Goldsmith,N., Valle,A., Oliver,J., Serrano,L. and Vaquero,A. vation. In fact, we demonstrate that the majority of DNA (2011) Stabilization of Suv39H1 by SirT1 is part of oxidative stress methylation gains and histone modifications are established response and ensures genome protection. Mol. Cell, 42, 210–223. 7. Gratchev,A., Kzhyshkowska,J., Kannookadan,S., Ochsenreiter,M., in the MO-to-MAC differentiation step, similar to what Popova,A., Yu,X., Mamidi,S., Stonehouse-Usselmann,E., were observed by Vento-Tormo and colleagues, and their Muller-Molinet,I., Gooi,L. et al. (2008) Activation of a disruption by cambinol results in aberrant post-LPS upreg- TGF-beta-specific multistep gene expression program in mature ulation of proinflammatory genes. Furthermore, we demon- macrophages requires glucocorticoid-mediated surface expression of strate the essentiality of DNA methylation in gene expres- TGF-beta receptor II. J. Immunol., 180, 6553–6565. 8. Imperatore,F., Maurizio,J., Vargas Aguilar,S., Busch,C.J., Favret,J., sion by specifically inhibiting MO-to-MAC DNA hyperme- Kowenz-Leutz,E., Cathou,W., Gentek,R., Perrin,P., Leutz,A. et al. thylation utilizing DAC, and we show that although DAC (2017) SIRT1 regulates macrophage self-renewal. EMBO J., 36, inhibits methylation prior to LPS challenge, aberrant upreg- 2353–2372. ulation of gene expression is not observed until after MAC 9. Yang,X., Wei,J., He,Y., Jing,T., Li,Y., Xiao,Y., Wang,B., Wang,W., Zhang,J. and Lin,R. (2017) SIRT1 inhibition promotes atherosclerosis activation. through impaired autophagy. Oncotarget, 8, 51447–51461. In conclusion, our work directly identifies SIRT1/2, 10. Shi,G., Gong,S., Chiurchiù,V., Dae Kim,C., Youn Park,S., Won through direct interaction with DNMTs, as essential fac- Lee,S., Yeob Lee,S., Whan Hong,K., Sik Bae,S. and Kim,K. (2017) tors in the establishment of gains of methylation, as well as sirT1/adenosine monophosphate-activated protein kinase ␣ signaling additional epigenetic modifications that prevent premature enhances macrophage polarization to an anti-inflammatory phenotype in rheumatoid arthritis. Immunol, 8, 1135. expression of proinflammatory genes during differentiation 11. Vaquero,A., Scher,M.B., Lee,D.H., Sutton,A., Cheng,H.-L., to MACs and prior to their encounter to bacterial antigens. Alt,F.W., Serrano,L., Sternglanz,R. and Reinberg,D. (2006) SirT2 is a histone deacetylase with preference for histone H4 Lys 16 during mitosis. Genes Dev., 20, 1256–1261. 12. Rothgiesser,K.M., Erener,S., Waibel,S., Luscher,B. and Hottiger,M.O. (2010) SIRT2 regulates NF- B-dependent gene DATA AVAILABILITY expression through deacetylation of p65 Lys310. J. Cell Sci., 123, 4251–4258. Expression and methylation array data for this publica- 13. Lee,A.S., Jung,Y.J., Kim,D., Nguyen-Thanh,T., Kang,K.P., Lee,S., tion have been deposited in the NCBI Gene Expression Park,S.K. and Kim,W. (2014) SIRT2 ameliorates Omnibus and are accessible through GEO Series accession lipopolysaccharide-induced inflammation in macrophages. Biochem. numbers GSE131115 and GSE131177. Biophys. Res. Commun., 450, 1363–1369. 14. Wang,X., Buechler,N.L., Martin,A., Wells,J., Yoza,B., McCall,C.E. and Vachharajani,V. (2016) Sirtuin-2 regulates sepsis inflammation in ob/ob mice. PLoS One, 11, e0160431. 15. Klug,M., Schmidhofer,S., Gebhard,C., Andreesen,R. and Rehli,M. SUPPLEMENTARY DATA (2013) 5-Hydroxymethylcytosine is an essential intermediate of active DNA demethylation processes in primary human monocytes. Supplementary Data are available at NAR Online. Genome Biol., 14, R46. 680 Nucleic Acids Research, 2020, Vol. 48, No. 2 16. Garcia-Gomez,A., Li,T., Kerick,M., Català-Moll,F., Comet,N.R., 32. Shankaranarayana,G.D., Motamedi,M.R., Moazed,D. and Rodrı́guez-Ubreva,J., de la Rica,L., Branco,M.R., Martı́n,J. and Grewal,S.I.S. (2003) Sir2 regulates histone H3 lysine 9 methylation Ballestar,E. (2017) TET2- and TDG-mediated changes are required and heterochromatin assembly in fission yeast. Curr. Biol., 13, for the acquisition of distinct histone modifications in divergent 1240–1246. terminal differentiation of myeloid cells. Nucleic Acids Res., 45, 33. Vaquero,A., Sternglanz,R. and Reinberg,D. (2007) NAD+-dependent 10002–10017. deacetylation of H4 lysine 16 by class III HDACs. Oncogene, 26, 17. Yang,L., Rau,R. and Goodell,M.A. (2015) DNMT3A in 5505–5520. haematological malignancies. Nat. Rev. Cancer, 15, 152–165. 34. Vaquero,A., Scher,M., Erdjument-Bromage,H., Tempst,P., 18. Moran-Crusio,K., Reavie,L., Shih,A., Abdel-Wahab,O., Serrano,L. and Reinberg,D. (2007) SIRT1 regulates the histone Ndiaye-Lobry,D., Lobry,C., Figueroa,M.E., Vasanthakumar,A., methyl-transferase SUV39H1 during heterochromatin formation. Patel,J., Zhao,X. et al. (2011) Tet2 loss leads to increased Nature, 450, 440–444. hematopoietic stem cell self-renewal and myeloid transformation. 35. Richards,E.J. and Elgin,S.C.R. (2002) Epigenetic codes for Cancer Cell, 20, 11–24. heterochromatin formation and silencing: rounding up the usual 19. Quivoron,C., Couronné,L., Della Valle,V., Lopez,C.K., Plo,I., suspects. Cell, 108, 489–500. Wagner-Ballon,O., Do Cruzeiro,M., Delhommeau,F., Arnulf,B., 36. Skvortsova,K., Masle-Farquhar,E., Luu,P.-L., Song,J.Z., Qu,W., Stern,M.-H. et al. (2011) TET2 Inactivation results in pleiotropic Zotenko,E., Gould,C.M., Du,Q., Peters,T.J., Colino-Sanguino,Y. hematopoietic abnormalities in mouse and is a recurrent event during et al. (2019) DNA hypermethylation encroachment at CpG island human lymphomagenesis. Cancer Cell, 20, 25–38. borders in cancer is predisposed by H3K4 monomethylation patterns. 20. Rodrı́guez-Ubreva,J., Català-Moll,F., Obermajer,N., Cancer Cell, 35, 297–314. Álvarez-Errico,D., Ramirez,R.N., Company,C., Vento-Tormo,R., 37. Jones,P.A. (2012) Functions of DNA methylation: islands, start sites, Moreno-Bueno,G., Edwards,R.P., Mortazavi,A. et al. (2017) gene bodies and beyond. Nat. Rev. Genet., 13, 484–492. Prostaglandin E2 leads to the acquisition of DNMT3A-Dependent 38. Martinez,F.O., Gordon,S., Locati,M. and Mantovani,A. (2006) tolerogenic functions in human myeloid-derived suppressor cells. Cell Transcriptional profiling of the human monocyte-to-macrophage Rep., 21, 154–167. differentiation and polarization: new molecules and patterns of gene 21. McLean,C.Y., Bristor,D., Hiller,M., Clarke,S.L., Schaar,B.T., expression. J. Immunol., 177, 7303–7311. Lowe,C.B., Wenger,A.M. and Bejerano,G. (2010) GREAT improves 39. King,A.D., Huang,K., Rubbi,L., Liu,S., Wang,C.-Y., Wang,Y., functional interpretation of cis-regulatory regions. Nat. Biotechnol., Pellegrini,M. and Fan,G. (2016) Reversible regulation of promoter 28, 495–501. and enhancer histone landscape by DNA methylation in mouse 22. Heinz,S., Benner,C., Spann,N., Bertolino,E., Lin,Y.C., Laslo,P., embryonic stem cells. Cell Rep., 17, 289–302. Cheng,J.X., Murre,C., Singh,H. and Glass,C.K. (2010) Simple 40. Malin,J., Aniba,M.R. and Hannenhalli,S. (2013) Enhancer networks combinations of lineage-determining transcription factors prime revealed by correlated DNAse hypersensitivity states of enhancers. cis-regulatory elements required for macrophage and B cell identities. Nucleic Acids Res., 41, 6828–6838. Mol. Cell, 38, 576–589. 41. Calo,E. and Wysocka,J. (2013) Modification of enhancer chromatin: 23. Subramanian,A., Tamayo,P., Mootha,V.K., Mukherjee,S., what, how, and why? Mol. Cell, 49, 825–837. Ebert,B.L., Gillette,M.A., Paulovich,A., Pomeroy,S.L., Golub,T.R., 42. Gore,S.D., Jones,C. and Kirkpatrick,P. (2006) Decitabine. Nat. Rev. Lander,E.S. et al. (2005) Gene set enrichment analysis: a Drug Discov., 5, 891–892. knowledge-based approach for interpreting genome-wide expression 43. O’Hagan,H.M., Wang,W., Sen,S., DeStefano Shields,C., Lee,S.S., profiles. Proc. Natl. Acad. Sci. U.S.A., 102, 15545–15550. Zhang,Y.W., Clements,E.G., Cai,Y., Van Neste,L., Easwaran,H. et al. 24. Jalili,V., Matteucci,M., Masseroli,M. and Morelli,M.J. (2015) Using (2011) Oxidative damage targets complexes containing DNA combined evidence from replicates to evaluate ChIP-seq peaks. methyltransferases, SIRT1, and polycomb members to promoter Bioinformatics, 31, 2761–2769. CpG Islands. Cancer Cell, 20, 606–619. 25. Chistiakov,D.A., Myasoedova,V.A., Revin,V. V., Orekhov,A.N. and 44. Li,J.-Y., Pu,M.-T., Hirasawa,R., Li,B.-Z., Huang,Y.-N., Zeng,R., Bobryshev,Y. V. (2018) The impact of interferon-regulatory factors to Jing,N.-H., Chen,T., Li,E., Sasaki,H. et al. (2007) Synergistic macrophage differentiation and polarization into M1 and M2. function of DNA methyltransferases Dnmt3a and Dnmt3b in the Immunobiology, 223, 101–111. methylation of Oct4 and nanog. Mol. Cell Biol., 27, 8748–8759. 26. Tarique,A.A., Logan,J., Thomas,E., Holt,P.G., Sly,P.D. and 45. Hashimshony,T., Zhang,J., Keshet,I., Bustin,M. and Cedar,H. (2003) Fantino,E. (2015) Phenotypic, functional, and plasticity features of The role of DNA methylation in setting up chromatin structure classical and alternatively activated human macrophages. Am. J. during development. Nat. Genet., 34, 187–192. Respir Cell Mol. Biol., 53, 676–688. 46. Leadem,B.R., Kagiampakis,I., Wilson,C., Cheung,T.K., Arnott,D., 27. Zajac,E., Schweighofer,B., Kupriyanova,T.A., Juncker-Jensen,A., Trojer,P., Classon,M., Easwaran,H., Baylin,S.B. and Baylin,S. (2018) Minder,P., Quigley,J.P. and Deryugina,E.I. (2013) Angiogenic A KDM5 inhibitor increases global H3K4 trimethylation occupancy capacity of M1- and M2-polarized macrophages is determined by the and enhances the biological efficacy of 5-Aza-2 -deoxycytidine. levels of TIMP-1 complexed with their secreted proMMP-9. Blood, Cancer Res, 78, 1127–1139. 122, 4054–4067. 47. Otani,J., Nankumo,T., Arita,K., Inamoto,S., Ariyoshi,M. and 28. Vento-Tormo,R., Company,C., Rodrı́guez-Ubreva,J., de la Rica,L., Shirakawa,M. (2009) Structural basis for recognition of H3K4 Urquiza,J.M., Javierre,B.M., Sabarinathan,R., Luque,A., Esteller,M., methylation status by the DNA methyltransferase 3A Aran,J.M. et al. (2016) IL-4 orchestrates STAT6-mediated DNA ATRX–DNMT3–DNMT3L domain. EMBO Rep., 10, 1235–1241. demethylation leading to dendritic cell differentiation. Genome Biol., 48. Ooi,S.K.T., Qiu,C., Bernstein,E., Li,K., Jia,D., Yang,Z., 17, 4. Erdjument-Bromage,H., Tempst,P., Lin,S.-P., Allis,C.D. et al. (2007) 29. Kallin,E.M., Rodrı́guez-Ubreva,J., Christensen,J., Cimmino,L., DNMT3L connects unmethylated lysine 4 of histone H3 to de novo Aifantis,I., Helin,K., Ballestar,E. and Graf,T. (2012) Tet2 facilitates methylation of DNA. Nature, 448, 714–717. the derepression of myeloid target genes during CEBP␣-induced 49. Li,B.-Z., Huang,Z., Cui,Q.-Y., Song,X.-H., Du,L., Jeltsch,A., transdifferentiation of pre-B cells. Mol. Cell, 48, 266–276. Chen,P., Li,G., Li,E. and Xu,G.-L. (2011) Histone tails regulate DNA 30. de la Rica,L., Rodrı́guez-Ubreva,J., Garcı́a,M., Islam,A.B., methylation by allosterically activating de novo methyltransferase. Urquiza,J.M., Hernando,H., Christensen,J., Helin,K., Cell Res., 21, 1172–1181. Gómez-Vaquero,C. and Ballestar,E. (2013) PU.1 target genes 50. Dhar,S.S., Zhao,D., Lin,T., Gu,B., Pal,K., Wu,S.J., Alam,H., Lv,J., undergo Tet2-coupled demethylation and DNMT3b-mediated Yun,K., Gopalakrishnan,V. et al. (2018) MLL4 is required to methylation in monocyte-to-osteoclast differentiation. Genome Biol., maintain broad H3K4me3 peaks and Super-Enhancers at tumor 14, R99. suppressor genes. Mol. Cell, 70, 825–841. 31. Vaquero,A., Scher,M., Lee,D., Erdjument-Bromage,H., Tempst,P. 51. Petell,C.J., Alabdi,L., He,M., Miguel,San, Rose,P. and Gowher,H. and Reinberg,D. (2004) Human SirT1 interacts with histone H1 and (2016) An epigenetic switch regulates de novo DNA methylation at a promotes formation of facultative heterochromatin. Mol. Cell, 16, subset of pluripotency gene enhancers during embryonic stem cell 93–105. differentiation. Nucleic Acids Res., 44, 7605–7617. Nucleic Acids Research, 2020, Vol. 48, No. 2 681 52. Charlet,J., Duymich,C.E., Lay,F.D., Mundbjerg,K., 56. Ford,E., Grimmer,M.R., Stolzenburg,S., Bogdanovic,O., DalsgaardSørensen,K., Liang,K. and Jones,P.A. (2016) Bivalent Mendoza,A. de, Farnham,P.J., Blancafort,P. and Lister,R. (2017) regions of cytosine methylation and H3K27 acetylation suggest an Frequent lack of repressive capacity of promoter DNA methylation active role for DNA methylation at enhancers. Mol. Cell, 62, 422–431. identified through genome-wide epigenomic manipulation. bioRxiv 53. Heo,J., Lim,J., Lee,S., Jeong,J., Kang,H., Kim,Y., Kang,J.W., doi: https://doi.org/10.1101/170506, 20 September 2017, preprint: not Yu,H.Y., Jeong,E.M., Kim,K. et al. (2017) Sirt1 regulates DNA peer reviewed. methylation and differentiation potential of embryonic stem cells by 57. Korthauer,K. and Irizarry,R.A. (2018) Genome-wide repressive antagonizing Dnmt3l. Cell Rep., 18, 1930–1945. capacity of promoter DNA methylation is revealed through 54. Aguilar-Arnal,L., Katada,S., Orozco-Solis,R. and Sassone-Corsi,P. epigenomic manipulation. bioRxiv doi: (2015) NAD(+)-SIRT1 control of H3K4 trimethylation through https://doi.org/10.1101/381145, 01 August 2018, preprint: not peer circadian deacetylation of MLL1. Nat. Struct. Mol. Biol., 22, reviewed. 312–318. 58. Pacis,A., Mailhot-Léonard,F., Tailleux,L., Randolph,H.E., 55. Wagner,J.R., Busche,S., Ge,B., Kwan,T., Pastinen,T. and Yotova,V., Dumaine,A., Grenier,J.-C. and Barreiro,L.B. (2019) Gene Blanchette,M. (2014) The relationship between DNA methylation, activation precedes DNA demethylation in response to infection in genetic and expression inter-individual variation in untransformed human dendritic cells. Proc. Natl. Acad. Sci. U.S.A., 116, 6938–6943. human fibroblasts. Genome Biol., 15, R37.