CHD4 regulates PADI1 and PADI3 expression linking pyruvate kinase M2 citrullination to glycolysis and proliferation

CHD3 and CHD4 are mutually exclusive ATPase subunits of the Nucleosome Remodelling and Deacetylation (NuRD) complex that regulates gene expression. CHD4 is essential for growth of multiple patient derived melanoma xenografts and for breast cancer. Here we show that CHD4 regulates expression of PADI1 (Protein Arginine Deiminase 1) and PADI3 in multiple cancer cell types modulating citrullination of three arginines of the allosterically-regulated glycolytic enzyme pyruvate kinase M2 (PKM2). PKM2 citrullination lowers sensitivity to the inhibitors Tryptophan, Alanine and Phenylalanine shifting the equilibrium towards the activator Serine. Citrullination thus bypasses normal physiological regulation by low Serine levels to promote excessive glycolysis defining a novel pathway regulating proliferation of melanoma and other cancer cells. We provide unique insight as to how conversion of arginines to citrulline impacts key interactions within PKM2 reprogramming its activity as an additional layer of complexity to the mechanisms regulating this important enzyme. Significance This study describes the mechanism by which citrullination of key arginines in pyruvate kinase M2 reprograms its allosteric regulation by the activating amino acid serine. PKM2 citrullination defines a novel pathway regulating glycolysis and cancer cell proliferation.


Abstract.
CHD3 and CHD4 are mutually exclusive ATPase subunits of the Nucleosome Remodelling and Deacetylation (NuRD) complex that regulates gene expression. CHD4 is essential for growth of multiple patient derived melanoma xenografts and for breast cancer. Here we show that CHD4 regulates expression of PADI1 (Protein Arginine Deiminase 1) and PADI3 in multiple cancer cell types modulating citrullination of three arginines of the allostericallyregulated glycolytic enzyme pyruvate kinase M2 (PKM2). PKM2 citrullination lowers sensitivity to the inhibitors Tryptophan, Alanine and Phenylalanine shifting the equilibrium towards the activator Serine. Citrullination thus bypasses normal physiological regulation by low Serine levels to promote excessive glycolysis defining a novel pathway regulating proliferation of melanoma and other cancer cells. We provide unique insight as to how conversion of arginines to citrulline impacts key interactions within PKM2 reprogramming its activity as an additional layer of complexity to the mechanisms regulating this important enzyme.
Significance. This study describes the mechanism by which citrullination of key arginines in pyruvate kinase M2 reprograms its allosteric regulation by the activating amino acid serine.
PKM2 citrullination defines a novel pathway regulating glycolysis and cancer cell proliferation.

Introduction.
A hallmark of cancer cells is the high glycolysis and lactic acid production under aerobic conditions, a metabolic state known as the Warburg effect (1). Tumour tissues accumulate increased amounts of glucose used not only for energy production, but also for anabolic reactions. Glycolytic intermediates are notably used for de novo synthesis of nucleotides or amino acids like glycine and serine produced in large amounts to sustain high rates of cancer cell proliferation (2,3). Coupling of energy production via glycolysis to the availability of the intermediates required for nucleotide and amino acid synthesis is controlled in large part by an alternatively spliced isoform of the enzyme pyruvate kinase called PKM2 expressed in proliferating embryonic and cancer cells (4,5). Unlike the PKM1 isoform that is constitutively active, PKM2 activity is positively regulated by serine (Ser), fructose 1,6-biphosphate (FBP) an intermediate of the glycolytic pathway and succinylaminoimidazole-carboxamide riboside (SAICAR), an intermediate in de novo purine nucleotide synthesis (4,6,7). High levels of these molecules stimulate PKM2, but when their levels are lowered through excessive glycolysis, PKM2 activity is inhibited by amino acids such as tryptophan (Trp), alanine (Ala) and phenylalanine (Phe) that compete with Ser to allosterically regulate PKM2 activity (8)(9)(10).
Through this complex feedback loop, PKM2 couples glycolytic flux to the level of critical intermediate metabolites. PKM2 activity is also regulated by post-translational modifications, such as tyrosine phosphorylation or proline hydroxylation under hypoxia (11,12).
Melanoma cells are no exception to the Warburg effect, showing high levels of aerobic glycolysis induced by transformation with oncogenic BRAF or NRAS (13). Treatment with vemurafenib, an inhibitor of oncogenic BRAF, down-regulates aerobic glycolysis, regained in resistant cells (14). Transcription factor MITF (Microphthalmia associated transcription factor) regulates many parameters of melanoma cell physiology including metabolism (15). MITF directly regulates PPARGC1 and cells with high MITF expression show elevated oxidative phosphorylation compared to cells with low MITF with higher glycolysis (16,17).
Bossi et al (18) performed an shRNA dropout screen to identify proteins involved in epigenetics and chromatin functions essential for patient derived melanoma xenograft (PDX) growth. This screen identified the ATPase subunit of the PBAF chromatin remodelling complex BRG1 along with CHD3 and CHD4, the catalytic ATPase subunits of the Nuclesome Remodelling and Deacetylation (NuRD) complex, as essential for tumour formation by all tested melanoma PDX. NuRD, is an epigenetic regulator of gene expression, acting in many, but not all (19), contexts as a co-repressor that remodels chromatin through its ATPase subunits and deacetylates nucleosomes through its HDAC1 and HDAC2 subunits (20) (21) (22) (23). CHD4 has also been reported to be essential in breast cancer (24). Here, we describe a novel pathway where CHD4 regulates expression of the PADI1 (Protein Arginine Deiminase 1) and PADI3 enzymes that convert arginine to citrulline. Increased PADI1 and PADI3 expression enhances citrullination of three arginines of the key glycolytic enzyme PKM2 attenuating repression of its activity by the allosteric regulators Phe, Trp and Ala, leading to excessive glycolysis, lowered ATP levels and slowed cell growth. CHD4 therefore links epigenetic regulation of PADI1 and PAD3 expression to glycolytic flux and the control of cancer cell growth.

CHD4 regulates expression of PADI1 and PADI3 in melanoma cells.
We performed siRNA CHD3 and CHD4 silencing in a collection of melanoma cells.
Silencing was specific for each subunit as measured by RT-qPCR and confirmed by immunoblot ( Fig. 1A-B). Loss of CHD3 also mildly reduced MITF expression, whereas that of SOX10 was unchanged. In agreement with the results of the previous shRNA dropout screen, siRNA-mediated CHD3 or CHD4 silencing reduced clonogenic capacity, increased the proportion of slow proliferating cells (Fig. 1A-D), but did not induce apoptosis (Fig. 1E). The effects were less dramatic than seen upon silencing of MITF that induces a potent cell cycle arrest and senescence (25) (26). RNA-seq following CHD4 silencing in melanoma cells identified more than 1000 up-regulated genes compared to 364 down-regulated genes showing that CHD4 was primarily acting as a transcriptional repressor ( Fig. 1F-G, and Supplementary Dataset S1). In contrast, similar numbers of genes were up or down-regulated by CHD3 silencing (Fig. 1F-G and Supplementary Dataset S1), but no significant overlap between the CHD3 and CHD4 up and down-regulated genes and the pathways they regulate were observed ( Fig. 1H-I). These data accord with previous reports showing that CHD3 and CHD4 form NuRD complexes with mainly distinct functions (27). De-regulated expression of selected genes upon CHD4 silencing was confirmed by RT-qPCR on independent RNA samples in both 501Mel and MM117 melanoma cells (Fig. 1J-K).
Amongst the genes potently up-regulated by CHD4, but not CHD3, silencing are PADI1 (Protein Arginine Deiminase 1) and PADI3 encoding enzymes that convert arginine to citrulline (28) (Fig. 1J-K, Fig. 2A-B). In all tested melanoma lines, PADI3 expression was almost undetectable and potently activated by CHD4 silencing, whereas some others had low basal PADI1 levels also strongly stimulated by CHD4 silencing (Fig. 1H). RNA-seq further showed that expression of PADI2 and PADI4 was low to undetectable in multiple melanoma cells and their expression was not induced by siCHD4 silencing (Supplemental Fig. S1A). The PADI1 and PADI3 genes are located next to each other (Supplemental Fig. S1B). ChIP-seq in melanoma cells revealed that CHD4 occupied an intronic regulatory element in PADI1 immediately adjacent to sites occupied by transcription factors CTCF and FOSL2 (AP1). This element is predicted to regulate both the PADI1 and PADI3 genes (Supplemental Fig. S1B) and is further marked by H2AZ, H3K4me1, BRG1 and ATAC-seq, but not by the lineage-specific transcription factors MITF and SOX10. CHD4 therefore appears to repress the activity of this element to prevent activation by CTCF, AP1 or other transcription factor that may bind to it.
Arginine citrullination by PADI enzymes is a calcium-dependent reaction. CHD4 silencing also strongly up-regulated a set of genes involved in calcium signalling including calcium channel subunits, calcium binding proteins and calcium-dependent protein kinases (Fig.   2C) consistent with CHD4 silencing inducing PADI enzyme expression and activity.

PADI1 and PADI3 citrullinate glycolytic enzymes and stimulate glycolysis.
To identify potential PADI1 and PADI3 substrates in melanoma cells, we made protein extracts from siC and siCHD4 cells, performed immunoprecipitation (IP) with a pan-citrulline antibody and analysed precipitated proteins by mass-spectrometry ( We focussed on PKM2, a highly regulated enzyme playing a central role in integrating control of glycolysis with cellular metabolic status and cell cycle (29). PKM2 converts phosphoenolpyruvate (PEP) to pyruvate then converted to lactic acid. To investigate PKM2 citrullination, melanoma cells were transfected with siC, siCHD4 or vectors allowing ectopic expression of PADI1 and PADI3 (Fig. 2G). While CHD4 silencing or PADI1/3 expression did not alter overall PKM2 levels ( Fig. 2H), strongly increased amounts of PKM2 were detected in the pan-citrulline IP following siCHD4 compared to siC in both 501Mel and MM117 melanoma cells and after ectopic PADI1 and PADI3 expression, particularly upon co-expression of both enzymes ( Fig. 2I-J).
To determine if siCHD4 silencing and enhanced PKM2 citrullination altered glycolysis, we profiled melanoma cell metabolism in real time. CHD4 silencing in all tested melanoma lines increased the basal OCR (oxygen consumption rate) and ECAR (extracellular acidification rate), markedly increased maximum OCR and ECAR and decreased the OCR/ECAR ratio due to the increased ECAR values (Fig. 3A-D). ECAR was blocked using 2-deoxy-D-glucose confirming that it was due to increased glycolysis (Fig. 3C). Increased glycolysis and lactic acid production diverts pyruvate from oxidative metabolism a more efficient ATP source.
The increased glycolysis seen upon CHD4 silencing was strongly diminished when PADI1 and PADI3 were additionally silenced ( Fig. 3F and J). In contrast, exogenous expression of PADI1, PADI3 or both stimulated glycolysis ( Fig. 3G and K). Consistent with increased glycolysis, PADI1/3 expression led to reduced intracellular ATP levels ( Fig. 3H and L) and reduced cell proliferation (Fig. 3I). PADI1 and PADI3 were therefore necessary and sufficient for increased glycolysis accounting for the effect seen upon CHD4 silencing.
It has previously been shown that treatment of melanoma cells with BRAF inhibitors induces metabolic reprogramming, strongly reducing glycolysis (14). Moreover, dependence on glycolysis sensitizes melanoma cells to the effects of BRAF inhibition (30). Consistent with these observations, CHD4 silencing or ectopic PADI1/3 expression that increased glycolysis sensitized Sk-Mel-28 cells to the effects of the BRAF inhibitor vemurafenib (Supplemental Fig.   S2). Hence, by regulating glycolysis CHD4 silencing or PADI1 and PADI3 expression acts to modulate melanoma cell sensitivity to BRAF inhibition. Nevertheless, the effect of CHD4 silencing had more potent effects on vemurafenib sensitivity that ectopic PADI1 and PADI3 expression suggesting additional pathways are affected.

PADI1 and PADI3 stimulate glycolysis in a variety of cancer cell types.
As mentioned above CHD4 may control PADI1/3 expression not via melanoma-specific factors, but through a regulatory element binding more ubiquitous factors and therefore regulate their expression in non-melanoma cancer cells. SiCHD4 silencing in SiHa cervical carcinoma cells strongly diminished their clonogenic capacity ( Therefore, in cell lines from distinct cancer types, CHD4 silencing or ectopic PADI1/3 expression increased glycolysis and negatively impacted cell proliferation.

Citrullination reprograms PKM2 allosteric regulation.
As described in the introduction, PKM2 isoform activity is positively regulated by serine (Ser), and FBP and negatively by Trp, Ala and Phe, thus coupling glycolytic flux to the level of critical intermediate metabolites (4)(5)(6). PKM2 allosteric regulation involves three distinct enzyme conformations [ (8,9,31) and Supplemental Fig. S3A]. In the apo (resting) state, in absence of small molecules and ions, the PKM2 N-terminal and A domains adopt an active conformation, but the B domain is in an inactive conformation. In the activated R-state, binding of FBP or Ser and magnesium, stabilizes the N and A domains in their active conformation, and rotates the B domain towards the A domain that together form the active site. In the inactive T-state, upon binding of inhibitory amino acids (Trp, Ala and Phe), the B domain adopts a partially active conformation, but the N and A domains undergo structural changes and disorganize the active site. The structural changes observed between the different PKM2 states are reinforced allosterically by organisation into a tetramer that is essential for enzyme function.
In siCHD4 extracts, 3 citrullinated arginine residues, R106, R246 and R489 enriched in the siCHD4 extracts were identified by mass-spectrometry ( Transition between the R-and T-states is finely regulated by changes in the relative concentrations of Ser versus Trp, Ala and Phe that compete for binding to the pocket (9). Loss of R106 positive side chain charge upon citrullination will diminish its ability to interact with the carboxylate group of the free amino acids. Due to its extended network of hydrogen bonds within the pocket and as it does not modify the active conformations of the N and A domains, it is possible that binding of Ser is less affected than that of the hydrophobic amino acids that induce important structural changes within the N and A domains. Consequently, R106 citrullination could reduce the inhibitory effect of Trp, Ala and Phe thereby shifting the equilibrium towards activation by Ser.
To test the above hypotheses, we asked if citrullination modulated glycolysis under different conditions. When cells were grown in absence of Ser, basal glycolysis was reduced and was no longer stimulated upon siCHD4 or PADI1/3 expression (Fig. 5B). On the other hand, exogenous Ser stimulated basal glycolysis to a level that was not further increased by siCHD4 (Fig. 5C). In contrast, basal glycolysis was reduced by exogenous Trp, but remained stimulated by siCHD4 and by PADI1/3 expression (Fig. 5D). Similarly, glycolysis was stimulated by siCHD4 in presence of increasing Phe concentrations (Fig. 5E), an effect particularly visible in MM117 cells where despite strongly inhibited basal glycolysis, stimulation was seen upon siCHD4 (Fig. 5F). PADI1/3 expression also stimulated glycolysis in presence of exogenous Ala (Fig. 5G). PKM2 citrullination did not therefore bypass the requirement for Ser, while excess Ser mimicked stimulation seen by siCHD4. In contrast, siCHD4 or PADI1/3 expression diminished inhibition by Trp/Ala/Phe, consistent with the idea that R106 citrullination modified the equilibrium in favour of the activator Ser.
R489 is directly involved in FBP binding with strong interactions between its guanidino group and the FBP 1' phosphate group (Fig. 6A). Despite its extensive interaction network with PKM2, FBP binding is severely reduced upon mutation of R489 into alanine (8) (10). Hydrogen bonding with R489 therefore plays a critical role in FBP binding that should be diminished by loss of its side chain charge upon citrullination, hence suggesting that activation of PKM2 by citrullination required weakening of its interaction with FBP.
In agreement with this idea, increasing concentrations of exogenous FBP had little effect on basal glycolysis, but blocked stimulation by siCHD4 (Fig. 6B). Addition of exogenous Ser at low FBP concentration (0.5 mM) augmented basal glycolysis and re-established stimulation by siCHD4. In contrast, at higher FBP concentration (2.0 mM), no increase in basal or siCHD4stimulated glycolysis was seen in presence of exogenous Ser. Increasing FBP therefore inhibited Ser and citrullination-dependent stimulation of glycolysis.
At low concentrations of exogenous FBP, increasing concentrations of exogenous Phe lowered basal glycolysis, whereas higher FBP concentrations overcame the Phe-induced repression (Fig. 6C), consistent with the known antagonistic effect of these ligands (9,10). At low FBP concentrations and in presence of Phe, siCHD4 stimulated glycolysis, but to a lower level than seen in absence of FBP, whereas higher FBP concentrations blocked stimulation.
These observations indicated that increased FBP inhibited the ability of Ser to stimulate glycolysis under basal conditions and following citrullination. This is further supported by the observation that the ability of citrullination to overcome Phe inhibition through shifting the equilibrium towards Ser was also diminished by FBP. Together these observations support the idea that by disrupting its hydrogen bonding, R489 citrullination acted to lower FBP binding and its ability to inhibit Ser, while citrullination of R106 reduced inhibition by Phe/Ala/Trp shifting the equilibrium towards Ser. Through these two concerted events, citrullination therefore reprogramed PKM2 to be principally regulated by Ser (Fig. 6E).

Discussion
Citrullination; a novel regulator of PKM2 activity, glycolysis and cancer cell proliferation.
Here we describe a regulatory pathway by which PKM2 citrullination regulates glycolysis and cancer cell proliferation. PKM2 is an allostatic regulator integrating a finely balanced feedback mechanism that modulates its activity over a wide range of absolute and relative amino acid concentrations (9). When Ser levels are lowered through glycolysis, the amino acid binding pocket of PKM2 is more readily occupied by inhibitory amino acids reducing glycolysis and allowing accumulation of metabolic intermediates required for Ser and nucleotide synthesis. Synthetic small molecules that increase PKM2 activity and override this feedback stimulate excessive glycolysis resulting in Ser auxotrophy and reduced cell proliferation (5,6,32,33).
Our data show that excess Ser stimulated glycolysis, whereas excess FBP did not.
Moreover, while FBP and Ser each stimulate PKM2 activity by stabilising the active R-state (9), our data show that FBP inhibits stimulation of glycolysis by Ser. This is unexpected given that each stabilizes the active state and that FBP and Ser can bind PKM2 simultaneously at least under in vitro conditions used for crystallography. Our observations rather suggest that in vivo, the active R-state is stabilised by one or the other, but not by both simultaneously. Exogenous FBP did not stimulate glycolysis, an observation in agreement with the report of Macpherson et al., (10) that exogenous FBP did not stimulate PKM2 activity as intracellular FBP concentrations are sufficient to near-saturate PKM2. They also reported that FBP and Phe can simultaneously bind PKM2 and that Phe prevents maximal activation of the FBP bound tetramer (10) contributing to maintaining PKM2 in a lower activity state (4). Stabilisation of the active state by Ser, whose binding is mutually exclusive with Phe/Trp/Ala, would therefore lead to higher PKM2 activity compared to FBP. This mechanism would be in agreement with our observations that exogenous Ser stimulates glycolysis, pushing the equilibrium towards Ser bound PKM2, whereas FBP that is already saturating and antagonized by Phe did not (Fig. 6E).
Excess FBP did however counteract inhibition of glycolysis by exogenous Phe again consistent with reported antagonism between these molecules. How FBP binding disfavours that of Ser remains to be determined.
That the active state stabilised by Ser has a higher activity than that of a FBP is further supported by the effects of citrullination. R489 citrullination that should diminish FBP binding serves to overcome FBP-inhibition of Ser-dependent stimulation of glycolysis. This effect is amplified by R106 citrullination that lowered PKM2 sensitivity to Trp/Ala/Phe shifting the equilibrium towards Ser. These two modifications would therefore act in concert to promote PKM2 regulation by Ser leading to increased glycolysis, analogous to addition of excess Ser and in agreement with the observation that stimulation of glycolysis by citrullination requires Ser (Fig. 6E).
Our data provide unique insight as to how conversion of arginine to citrulline impacts their key interactions. Unlike other post-translational modifications such as phosphorylation, or methylation, and to some extent acetylation, that often act positively to create new interactions with proteins the specifically recognize the modified amino acids, citrullination acts negatively due to loss of side chain charge and weakened hydrogen bonding ability. In the case of PKM2, our data illustrate how weakening of two interactions paradoxically translates into a positive reprograming and stimulation of glycolysis.
Under most normal conditions, expression of PADI enzymes in general and PADI1 and PADI3 in particular is tightly regulated with low or no expression. Analysis of TCGA data shows that their expression in many tumour types is also very low. Tight control of PADI expression is perhaps essential given that several tumour types have been shown to be auxotrophic for arginine (34). A majority of melanomas were reported to be auxotrophic for arginine because of the lack of Argininosuccinate Synthetase 1 (ASS1) (Dillon et al., 2004), the first of the two enzymes involved in arginine biosynthesis. Therefore, it cannot be excluded that induction of PADI1 and PADI3 and the subsequent increase in citrullinated proteins, so far believed to be irreversible, exacerbates arginine auxotrophy through reduced ability to turn over the citrullinated proteins with non-citrullinated forms.
There are however well documented situations where PADI enzyme expression is upregulated. Expression of PADI enzymes can be induced under hypoxic conditions, for example in glioblastoma (35). In hypoxia, it has been shown that PKM2 undergoes proline hydroxylation and acts as a co-factor for HIF1A to increase expression of glycolytic enzymes (11). Our data further suggest that PADI1 and PADI3 expression in hypoxic tumour cells would stimulate glycolysis through PKM2 citrullination.
PADI enzyme expression is also de-regulated in pathological situations such as rheumatoid arthritis (RA) where the production of antibodies against aberrantly citrullinated proteins contributes to the chronic inflammatory state (36)(37)(38). Moreover, citrullination of glycolytic enzymes including PKM2 was observed in RA (38). The RA-associated environment is characterised by hypoxia and heterogeneous availability of nutrients, resembling that of some tumours (39). Our data suggest that in addition to contributing to the inflammatory state through production of antibodies against citrullinated proteins, PADI1 and PADI3 expression in RA and subsequent PKM2 citrullination may account for the increased glycolysis seen in activated RA-associated fibroblast-like synoviocytes, another hallmark of the disease (39)(40)(41).
In conclusion, we identify a novel pathway regulating melanoma cell proliferation where PADI1 and PADI3 citrullinate key arginines in PKM2 involved in its allosteric regulation to modulate glycolysis and cell proliferation. This pathway is shared in other cancer cells indicating a more general mechanism for regulating cell proliferation and may be active in other cellular pathological contexts associated with increased glycolysis.

Cell culture, siRNA silencing and expression vector transfection
Melanoma cell lines 501Mel and SK-Mel-28 were grown in RPMI 1640 medium supplemented with 10% foetal calf serum (FCS). MM074 and MM117 were grown in HAM-

Proliferation, viability and senescence analyses by flow cytometry
To assess proliferation after siRNA treatment, cells were stained with Cell Trace Violet (Invitrogen) on the day of transfection. To assess cell viability, cells were harvested 72 hours after siRNA transfection and stained with Annexin-V (Biolegend) following manufacturer instructions. Cells were analysed on a LSRII Fortessa (BD Biosciences) and data were analysed using Flowjo software.

ATP measurement
The concentration of ATP was determined 72h after siRNA transfection using the luminescent ATP detection system (Abcam, ab113849) following the manufacturer's instructions.

Protein extraction and Western blotting
Whole cell extracts were prepared by the standard freeze-thaw technique using LSDB 500 buffer (500 mM KCl, 25 mM Tris at pH 7.9, 10% glycerol (v/v), 0.05% NP-40 (v/v), 16mM DTT, and protease inhibitor cocktail). Cell lysates were subjected to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and proteins were transferred onto a nitrocellulose membrane.

Immunoprecipitation and mass-spectrometry
Citrullinated proteins were immunoprecipitated from whole cell extracts with an antipan-citrulline antibody (Abcam, ab6464). Samples were concentrated on Amicon Ultra 0.5 mL columns (cutoff: 10 kDa, Millipore), resolved by SDS-PAGE and stained using the Silver 7 Quest kit (Invitrogen).

Mass spectrometry and analysis
Mass-spectrometry was performed at the IGBMC proteomics platform (Strasbourg, France). Samples were reduced, alkylated and digested with LysC and trypsin at 37°C overnight.

Peptides were then analyzed with an nanoLC-MS/MS system (Ultimate nano-LC and LTQ
Velos ion trap, Thermo Scientific, San Jose Califronia). Briefly, peptides were separated on a C18 nano-column with a 1 to 30 % linear gradient of acetonitrile and analyzed in a TOP20 CID data-dependent MS method. Peptides were identified with SequestHT algorithm in Proteome Discoverer 2.2 (Thermo Fisher Scientific) using Human Swissprot database (20347 sequences).
Precursor and fragment mass tolerance were set at 0.9 Da and 0.6 Da respectively. Trypsin was set as enzyme, and up to 2 missed cleavages were allowed. Oxidation (M) and Citrullination (R) were set as variable modifications, and Carbamidomethylation (C) as fixed modification.
Peptides were filtered with a 1 % FDR (false discovery rate) on peptides and proteins. For statistical analyses data was re-analysed using Perseus (42).

Chromatin immunoprecipitation and sequencing
CHD4 ChIP experiments were performed on 0.4% Paraformaldehyde fixed and sonicated chromatin isolated from 501Mel cells according to standard protocols as previously described (43). MicroPlex Library Preparation kit v2 was used for ChIP-seq library preparation.
The libraries were sequenced on Illumina Hiseq 4000 sequencer as Single-Read 50 base reads following Illumina's instructions. Sequenced reads were mapped to the Homo sapiens genome assembly hg19 using Bowtie with the following arguments: -m 1 --strata --best -y -S -l 40 -p 2.
After sequencing, peak detection was performed using the MACS software (44). Peaks were annotated with Homer (http://homer.salk.edu/homer/ngs/annotation.html) using the GTF from ENSEMBL v75. Peak intersections were computed using bedtools and Global Clustering was done using seqMINER. De novo motif discovery was performed using the MEME suite (memesuite.org). Motif enrichment analyses were performed using in house algorithms as described in (45).

RNA preparation, quantitative PCR and RNA-seq analysis
RNA isolation was performed according to standard procedure (Qiagen kit). qRT-PCR was carried out with SYBR Green I (Qiagen) and Multiscribe Reverse Transcriptase (Invitrogen) and monitored using a LightCycler 480 (Roche). RPLP0 gene expression was used to normalize the results. Primer sequences for each cDNA were designed using Primer3 Software and are available upon request. RNA-seq was performed essentially as previously described (46). Gene ontology analyses were performed with the Gene Set Enrichment Analysis software GSEA v3.0 using the hallmark gene sets of the Molecular Signatures Database v6.2 and the functional annotation clustering function of DAVID.

Analysis of oxygen consumption rate (OCR) and glycolytic rate (ECAR) in living cells
The ECAR and OCR were measured in an XF96 extracellular analyzer (Seahorse Bioscience).
A total of 20000 cells per well were seeded and transfected by siRNA or expression vector 72h and 24h hours respectively prior the experiment. The cells were incubated in a CO2-free incubator at 37°C and the medium was changed to XF base medium supplemented with 1mM pyruvate, 2 mM glutamine and 10mM glucose for an hour before measurement. For OCR profiling, cells were sequentially exposed to 2 µM oligomycin, 1 µM carbonyl cyanide-4-(trifluorome-thoxy) phenylhydrazone (FCCP), and 0.5 µM rotenone and antimycin A. For ECAR profiling, cells were sequentially exposed to 2 µM oligomycin and 150 mM 2-        represented as a tetramer, is predominantly bound by the activator FBP in equilibrium with the inhibitors Trp/Phe/Ala. Increased Ser shifts the equilibrium to a Ser-bound form, in presence or absence of FBP, that has higher activity due to mutually exclusive occupancy by Ser or Trp/Phe/Ala accounting for the observed increase in glycolysis. Citrullination diminishes FBP binding (R489<C represented by -C) alleviating its negative effect on Ser and shifts the mutually exclusive Ser vs Trp/Phe/Ala binding in favour of Ser. The net result is to promote a predominantly Ser-bound form accounting for the observed increased in glycolysis.