Nitric Oxide Modifies Global Histone Methylation by Inhibiting Jumonji C Domain-containing Demethylases*♦

Background: The methylation status of histone tails is a balance between methylation and demethylation. Results: Nitric oxide inhibits lysine demethylase 3A and alters cellular histone methylation patterns. Conclusion: Nitric oxide can significantly modify the epigenetic landscape. Significance: These results establish nitric oxide as a physiological epigenetic regulator acting through a nonclassical cell signaling mechanism. Methylation of lysine residues on histone tails is an important epigenetic modification that is dynamically regulated through the combined effects of methyltransferases and demethylases. The Jumonji C domain Fe(II) α-ketoglutarate family of proteins performs the majority of histone demethylation. We demonstrate that nitric oxide (•NO) directly inhibits the activity of the demethylase KDM3A by forming a nitrosyliron complex in the catalytic pocket. Exposing cells to either chemical or cellular sources of •NO resulted in a significant increase in dimethyl Lys-9 on histone 3 (H3K9me2), the preferred substrate for KDM3A. G9a, the primary methyltransferase acting on H3K9me2, was down-regulated in response to •NO, and changes in methylation state could not be accounted for by methylation in general. Furthermore, cellular iron sequestration via dinitrosyliron complex formation correlated with increased methylation. The mRNA of several histone demethylases and methyltransferases was also differentially regulated in response to •NO. Taken together, these data reveal three novel and distinct mechanisms whereby •NO can affect histone methylation as follows: direct inhibition of Jumonji C demethylase activity, reduction in iron cofactor availability, and regulation of expression of methyl-modifying enzymes. This model of •NO as an epigenetic modulator provides a novel explanation for nonclassical gene regulation by •NO.

The accumulation of whole genomic sequence from thousands of individuals has made it clear that epigenetic regulation is as important as nucleotide sequence for determining final phenotypic outcomes (1,2). For the past 50 years, major advances in our understanding of gene expression have focused on the central dogma of molecular biology (DNA to RNA to protein). More recently, however, the emphasis has shifted toward investigating the importance of epigenetics on pheno-typic outcomes. One example is post-translational modifications of histones that have major influences on chromatin structure and gene transcription (3). The most studied of these modifications is acetylation, but it is now recognized that methylation, phosphorylation, ubiquitination, and SUMOylation all have regulatory functions (4,5). Methylation of lysine residues, once thought to be an irreversible process, has gained prominence over the past decade as one of the most important histone modifications involved in genetic regulation (6). This is due, in part, to the discovery of two families of histone demethylases that catalyze the removal of methyl groups from arginine and lysine residues. The first family is composed of the lysine-specific demethylases LSD1 and LSD2, which are FADdependent amine oxidases (7,8). The other type belongs to a much larger family of the Jumonji C (JMJC) 2 domain-containing demethylases. This family is composed of 30 proteins, twothirds of which target specific Lys residues on the N-terminal tail of histone 3 (H3) (9 -18).
The JMJC class of lysine-specific demethylases (KDMs) are mononuclear Fe(II)-dependent dioxygenases that use ␣-ketoglutarate (␣-KG) and oxygen (O 2 ) as cosubstrates to oxidatively demethylate specific Lys residues (11). The common structural motif of this protein family is a 2-histidine-1-carboxylate facial triad that coordinates a non-heme Fe(II) at the catalytic center (19,20). Although general catalytic mechanisms for these enzymes have been proposed, the details are an ongoing area of research and believed to be enzyme-specific (9). Methylation status of lysine residues on histones can positively or negatively influence gene transcription. These lysine residues can be mono-, di-, or tri-methylated, and the functional consequences depend upon both the location and degree of this methylation. The primary targets on H3 are lysines 4,9,27,36, and 79 with 4 and 9 being the most heavily modified. Methylation at Lys-4 is generally associated with transcriptional activation, although Lys-9 methylation is typically found in silenced chromatin. H4 is also commonly methylated, especially at Lys-20. Methylation is carried out by lysine methyltransferases (KMTs), and methylation status is therefore a dynamic process that represents a balance of KMT and KDM activities. This resulting balance can have both acute and long term impacts on chromatin structure and mRNA expression (6,9,(21)(22)(23).
Nitric oxide ( ⅐ NO, nitrogen monoxide) is a biological free radical signaling molecule that regulates many physiological processes, including angiogenesis, smooth muscle tone, immune response, apoptosis, and synaptic communication (24). Like iron chelators and divalent metals (Co 2ϩ and Ni 2ϩ ), ⅐ NO has been shown to inhibit mononuclear non-heme iron oxygenases that contain the 2-His-1-carboxylate facial triad structural motif. One of the most studied examples of this interaction occurs with the HIF prolyl hydroxylase (25)(26)(27). Its activity is inhibited by Co 2ϩ and Ni 2ϩ by replacing the iron at the catalytic site (28). ⅐ NO and iron chelators have also been shown to inhibit prolyl hydroxylase activity by either binding to or removing this catalytic iron (29). Previously, it has been shown that the JMJC domain-containing histone demethylase KDM3A is highly sensitive to inhibition by carcinogenic nickel ions (28,30,31). Others have demonstrated that various dioxygenase inhibitors such as ␣-KG analogs, hypoxia, ⅐ NO, reactive oxygen species, CoCl 2 , and iron chelators like desferrioxamine (DFO) all increase histone methylation, most probably by reducing demethylase activity (32)(33)(34)(35)(36)(37).
The common structural similarities between JMJC demethylases and other mononuclear non-heme iron oxygenases spurred us to investigate whether ⅐ NO might similarly inhibit this class of demethylases resulting in alterations in histone methylation patterns. We chose to focus on KDM3A as it is the predominant demethylase acting on the di-and monomethylated states of the heavily modified H3K9 residue (12). We demonstrate that ⅐ NO directly inhibits KDM3A leading to the accumulation of the H3K9me2. The contribution of methyltransferases to these increases was ruled out. Our studies are the first to provide direct mechanistic evidence for epigenetic regulation by ⅐ NO through histone modifications. This is an important finding as it is well known that the multitude of gene expression changes caused by ⅐ NO cannot be solely explained by classical ⅐ NO-heme interactions or S-nitrosation of protein thiols. Further investigations into ⅐ NO-driven changes in histone methylation status at specific promoter sites will provide a model for examining many of the unexplained changes in ⅐ NOassociated gene expression. [3-aminopropyl]-N-[4-(3-aminopropylammonio)butyl]-amino]diazen-1-ium-1,2-diolate (Sper/NO), and 1-(hydroxy-N-NO-azoxy)-L-proline (Proli/NO) were the generous gifts from Dr. Joseph Hrabie (NCI, National Institutes of Health). Iron(II) sulfate heptahydrate (FeSO 4 ), ␣-ketoglutaric acid sodium salt, ferric ammonium citrate, DFO, aminoguanidine hydrochloride, lipopolysaccharides (LPS), and 1H- [1,2,4]oxadiazolo [4,3a]quinoxalin-1-one (ODQ) were purchased from Sigma. All cell culture reagents were purchased from Invitrogen with the exception of methionine-free DMEM (AthenaES).

EXPERIMENTAL PROCEDURES
Cell Culture-MDA-MB-231 human breast carcinoma cells and RAW 264.7 murine macrophages were cultured in DMEM, and Jurkat T lymphocyte cells were grown in RPMI 1640 medium. In all cases, media were supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Additionally, media for Jurkat T cells contained 50 M 2-mercaptoethanol. Cells were grown to 80% confluence, and serum-starved overnight prior to treatment. Cells were incubated in serum-free, methionine-free DMEM for 4 h before addition of DETA/NO wherever necessary. For coculture experiments, RAW 264.7 cells were activated with LPS (1 g/ml) for 6 h to induce inducible nitric oxide (iNOS) expression. Subsequently, media were aspirated, and Jurkat T cells were counted and added in serumfree media in the presence or absence of 2 mM aminoguanidine. For iron supplementation experiments, cells were treated with ferric ammonium citrate (150 g/ml) for 16 h and washed with PBS to remove excess iron prior to treatment.
KDM3A Activity Assay-Recombinant KDM3A was purchased from BPS Biosciences. 10-l reactions were conducted at 25°C in Tris, pH 8.0, containing 80 M FeSO 4 , 1 mM ␣-ketoglutarate, 220 M O 2 , 2 mM ascorbate, 40 ng of recombinant KDM3A protein, and 125 nM histone 3 peptide fragment containing a dimethyl lysine 9 (Anaspec). Positive ion MALDI-TOF mass spectra of peptide reactions were acquired using a Voyager DE Pro mass spectrometer (Applied Biosystems, Foster, CA). Preceding the mass spectrometric analysis, each sample was processed with a C18-ZipTip (Millipore, Cambridge, MA) cleanup. A 1-l aliquot of the desalted peptide solution was mixed with 1 l of matrix solution composed of a saturated ␣-cyano-4-hydroxycinnamic acid solution made up in acetonitrile/water (1:1, v/v) acidified with 0.1% (v/v) trifluoroacetic acid. A 1-l aliquot of the mixture was then spotted on a MALDI-TOF sample stage and air-dried prior to analysis. For each sample, mass spectra from 500 laser shots were acquired in linear mode and signal averaged over the range m/z 500 to 5000.
Electron Paramagnetic Resonance (EPR)-Measurements were performed on a Bruker X-band EMX Plus EPR spectrometer. Samples were frozen and read in liquid N 2 . Dinitrosyliron complexes (DNIC) were detected at g ϭ 2.03. Settings were as follows: centerfield 3335.25 G, microwave frequency 9.460544 GHz, modulation amplitude 10G, 200 G scan range, 90-s scan time, one scan. For quantification, the double integral of the first derivative spectra was compared with a standard curve generated with synthetic diglutathione DNIC as described previously (38). The CIP was estimated by treating cells with 1 mM DFO for 4 h. The resulting Fe 3ϩ -DFO g ϭ 4.3 signal was read with the following settings: center field 1575.65 G, microwave frequency 9.449528 GHz, modulation amplitude of 10 G, 200 G scan range, 30-s scan time, four scans. For quantification, the double integral of the first derivative spectra was compared with that of a standard curve generated with Fe 3ϩ -DFO as described previously (38). Settings for measurement of the recombinant KDM3A⅐ ⅐ NO complex were as follows: centerfield 3328.15 G, microwave frequency 9.440474 GHz, modula-tion amplitude 7.5 G, 1000 G scan range, 60-s scan time, 16 scans.

Quantification of NO 3
Ϫ /NO 2 Ϫ , S-Nitrosothiols, and ⅐ NO-Quantification was performed by chemiluminescence using the Sievers Nitric Oxide Analyzer 280i. For coculture experiments, 25-l media aliquots were drawn and injected into a reaction chamber containing vanadium chloride in HCl to determine total NO 3 Ϫ /NO 2 Ϫ concentrations as described previously (39). S-Nitrosothiols were measured as described previously (40,41). For determining steady-state ⅐ NO concentrations, [ ⅐ NO] ss , 25-l aliquots were injected with a Hamilton air-tight syringe into a reaction vessel containing 5 M NaOH.
Statistics-Significance was determined using either t test or one-way analysis of variance with Bonferoni post hoc analysis. Error bars indicate S.E.

RESULTS
Nitric Oxide Inhibits KDM3A Activity in Vitro-KDM3A is the predominant demethylase acting upon lysine 9 of histone 3, and the mono-and dimethyl modifications of this residue (H3K9me1 and H3K9me2, respectively) are its preferred substrates. Hypoxia and inducers of "chemical hypoxia," including nickel (Ni 2ϩ ), cobalt (Co 2ϩ ), as well as dioxygenase inhibitors such as N-(methoxyoxoacetyl)glycine methyl ester, are recognized as strong inhibitors of JMJC demethylase activity (28, 30 -32, 34, 36). Because ⅐ NO is known to inhibit structurally similar mononuclear non-heme iron oxygenases containing the 2-His-1-carboxylate facial triad structural motif, we tested ⅐ NO's ability to inhibit KDM3A in vitro. We incubated a histone Although ⅐ NO inhibited KDM3A enzymatic activity, one possibility is that ⅐ NO is simply inducing nonspecific oxidative or nitrosative damage to the protein. To investigate this, we repeated the reaction in the presence of a relatively high amount of ⅐ NO, Ϸ20 M, given as a bolus (10 M Proli/NO). This amount of ⅐ NO was approximately the same as the total amount used in Fig. 1C, but the exposure time was dramatically shorter (5 s versus 60 min). No inhibition of KDM3A was observed (Fig. 1D). Because the total amount of ⅐ NO was similar but the exposure time was different, we took this to indicate that the inhibitory effects of ⅐ NO on KDM3A activity are reversible, they require sustained ⅐ NO exposure, and that critical protein modifications or destruction of the enzyme most likely do not explain our results. Moreover, in parallel experiments of KDM3A treated with ⅐ NO, we could not detect any protein S-nitrosothiols by chemiluminescence, further supporting Fe(II)⅐ ⅐ NO complex formation as the mechanism of inhibition (data not shown).
To determine temporal and concentration effects of ⅐ NO on enzyme activity, we measured the demethylation products of KDM3A at four different ⅐ NO concentrations over time (0 -120 min) (Fig. 1E). In solution, ⅐ NO reacts with O 2 (autooxidation), and O 2 is a substrate for the enzyme. Therefore, to eliminate the confounding effect of substrate loss due to O 2 scavenging by ⅐ NO, we chose concentrations of ⅐ NO that would decrease the O 2 concentration by no more 10% through the course of the experiment. Under these conditions, even the lowest steadystate concentration of ⅐ NO (Ϸ 200 nM; 25 M Sper/NO) was able to inhibit product formation by as much as 40%, and Ϸ2 M [ ⅐ NO] ss (200 M Sper/NO) inhibited this by Ϸ70%. These experiments were conducted at ambient O 2 concentrations (Ϸ220 M). Although direct comparisons between steady-state ⅐ NO concentrations and dissolved O 2 concentrations are inexact, we approximate that the ⅐ NO to O 2 ratios may be as high as 1:1000 under these conditions. Although further experiments will be needed to confirm this, our data suggest a higher enzyme affinity for ⅐ NO than for O 2 . These results indicate both concentration-and duration-dependent effects of ⅐ NO exposure on enzyme activity.
Nitric Oxide Forms a Histidyl⅐Iron⅐Nitrosyl Complex in the Catalytic Pocket of KDM3A-To gain further insight into the mechanism of KDM3A inhibition by ⅐ NO, we conducted an EPR analysis of the enzyme. Fig. 2 is an EPR spectrum of KDM3A in the presence of substrates, cofactors, and ⅐ NO. Spectra were not observed in the absence of ⅐ NO nor was there a signal from ⅐ NO combined with all the cofactors in the absence of KDM3A. There appeared to be a superposition of spectra of more than three lines. The major spectrum has g values of 2.17, 2.06, and 2.00 or possibly 2.09, 2.06, and 2.00, with the line at g ϭ 2.17 not assigned. There are weak signals around g ϭ 2.04 assigned to Fe(NO) 2 (-SR) 2 and a line at 1.92 not assigned. Examples of EPR signals assigned to a histidyl⅐iron⅐nitrosyl complex have g values of 2.055, 2.033, and 2.015 for the signal from the reaction of ferritin and ⅐ NO (42). Furthermore, d7 and d9 forms of histidyl⅐iron⅐ nitrosyl⅐aconitase complexes with g values of 2.05 and 2.01 for d7 and 2.032 and 2.004 for d9 were reported (43). g values of 2.07, 2.02, and 1.98 were assigned to a substrate-bound ironnitrosyl-cysteine dioxygenase site (44). Although the anisotropy in g values is greater than in the examples, an ⅐ NO⅐Fe⅐KDM3A complex with the 2-His-1-carboxylate facial triad is the most probable source for these spectra under our conditions. The enzymatic inhibition seen in Fig. 1, likely results from the direct binding of ⅐ NO to the Fe(II) coordinated by the facial triad.
Physiological ⅐ NO Concentrations Alter Histone 3 Lysine 9 Methylation in Cultured Cells-Having established that ⅐ NO inhibits KDM3A activity in vitro, we extended these findings into a cellular system to determine whether differential methylation patterns at H3K9 would similarly be observed subsequent to ⅐ NO exposure. Initially, we set out to examine changes in H3K9me2 as it is the most abundant of the three methylation states at H3K9 (9), and it is also the primary substrate for KDM3A. We compared differences in methylation patterns between cells exposed to ⅐ NO and those exposed to other known inhibitors of protein activity (Ni 2ϩ and DFO). Similar increases in methylation at H3K9me2 were noted for all treatments (Fig. 3A). A 24-h time course of ⅐ NO exposure demon-   MAY 31, 2013 • VOLUME 288 • NUMBER 22

Nitric Oxide Inhibits Histone Demethylases
strated that within 8 h increases in H3K9me2 could be observed relative to untreated controls, and at 24 h the degree of methylation was still increasing (1.4-fold, Fig. 3B). Since we observed robust and reproducible responses to ⅐ NO at 24 h, this time point was chosen for all subsequent investigations. Next, we investigated concentration-dependent effects of ⅐ NO and determined that as little as 100 M DETA/NO resulted in detectable increases in H3K9me2, but maximal induction was achieved with 500 M DETA/NO (Fig. 3C). Importantly, these amounts of DETA/NO generated concentrations of ⅐ NO well within the physiological range (Fig. 3) (38, 39). Although dimethyl is the principal methyl modification at H3K9, it was also important to measure the relative changes in the monoand trimethyl states as change in one may be linked to changes in the others. In response to ⅐ NO, H3K9me3 was unchanged, whereas H3K9me1 was decreased with respect to untreated controls (Fig. 3D).
Nitric Oxide Increases Demethylase Gene Expression-In addition to altering histone methylation patterns, hypoxia and inducers of chemical hypoxia have been shown to strongly induce the expression JMJC domain-containing demethylases. For this reason, we compared the ability of ⅐ NO to up-regulate KDM3A to that of Co 2ϩ , Ni 2ϩ , hypoxia (1% O 2 , and the metal chelator DFO. Fig. 4A demonstrates that in cells treated with ⅐ NO, KDM3A was up-regulated to the same extent as hypoxia, DFO, Co 2ϩ , and Ni 2ϩ . In response to ⅐ NO, KDM3A was up-regulated in both a time-and concentration-dependent manner (Fig. 4, B and C). Interestingly, these responses did not translate into decreases in H3K9me2 levels. In addition to KDM3A, we also examined the steady-state mRNA levels of eight other demethylases that are known to be present in this cell type. Of the nine demethylases, KDM3A and KDM7A were strongly upregulated in response to 24 h of ⅐ NO exposure, and there was a modest up-regulation of the others (Fig. 4D). All of these enzymes can demethylate one of the three methylation states of H3K9 to some extent. The majority of demethylation of H3K9me2, however, is believed to be carried out by KDM3A.
Methyltransferase Activity Does Not Account for Increased H3K9 Methylation by ⅐ NO-An increase in histone methylation can result from either a relative decrease in demethylation or a relative increase in methylation. To determine the contribution of increased H3K9 methylation by methyltransferases in  response to ⅐ NO, we cultured cells in methionine-free media. S-Adenosylmethionine, which is synthesized from methionine and ATP, is the sole methyl donor for methyltransferase reactions. Histone methylation changes detected in cells grown in media depleted of methionine will predominantly reflect alterations in demethylation as opposed to methylation (33,34,36,37). Fig. 5A demonstrates that, even in methionine-free media, H3K9me2 was strongly increased in response to ⅐ NO exposure compared with untreated controls.
Although active methylation did not explain changes in methylation patterns, we went on to examine methyltransferase gene expression in response to ⅐ NO. G9a is the major methyltransferase that acts on H3K9 (45,46). We noted a moderate down-regulation in G9a protein within 2 h of ⅐ NO exposure and a marked decrease by 24 h (Fig. 5B). As was seen with KDM3A, changes in G9a expression did not translate into expected changes in H3K9me2. In addition to G9a, at least six more methylating enzymes have been reported to act on this residue. For this reason, we examined changes in steady-state mRNA levels after cellular exposure to ⅐ NO for all reported methyltransferases (Fig. 5C). Significant increases were noted for SETDB2 and SUV39h2. These enzymes, however, are primarily responsible for trimethylation of H3K9 (47)(48)(49), and increasing their activity would tend to decrease, rather than increase, the dimethyl state of H3K9. PRDM2 was also up-regulated; however, little is known about its substrate specificity (50).

Increasing the Chelatable Iron Pool Can Abrogate the Effect of ⅐ NO on Histone Methylation-Iron chelators inhibit 2-His-1-
carboxylate non-heme iron oxygenases by depleting the availability of Fe(II) necessary for catalysis. These compounds react with iron in the CIP, which is considered the major source of iron for JMJC domain dioxygenases (51-53). We and others have shown previously that upon cellular exposure to ⅐ NO, the CIP is quantitatively converted into paramagnetic DNIC (39,40,54). Therefore, in addition to direct enzyme inhibition by ⅐ NO, it is likely that KDM3A activity could be further impaired by the loss of iron availability via DNIC formation. To test this, we augmented the CIP, exposed cells to ⅐ NO, and measured changes in H3K9 methylation.
The CIP was increased by exposing cells to media supplemented with ferric ammonium citrate for 16 h. The cells were then treated with two different concentrations of ⅐ NO (250 and 500 M DETA/NO). The resulting increases in CIP and DNIC formation after iron supplementation are shown in Fig. 6A. The CIP was measured by changes in the g ϭ 4.3 signal as detected by EPR (Fig. 6B). DNIC were also quantified with EPR by measuring their characteristic signal centered at g ϭ 2.03 (Fig. 6, C  and D). In the absence of iron supplementation, treatment with both 250 and 500 M DETA/NO formed DNIC at concentrations roughly equal to the CIP. For both concentrations of  DETA/NO, supplementation of iron greatly increased the EPRdetectable signal (Fig. 6A). Under iron-supplemented conditions, the higher dose of ⅐ NO was able to convert 100% of the CIP into DNIC. At the lower ⅐ NO dose, however, only Ϸ50% the CIP was sequestered in the form of DNIC (Fig. 6A). Changes in H3K9me2 were also measured under identical conditions of iron supplementation and ⅐ NO exposure. During iron supplementation, increases in H3K9me2 were only observed with the higher dose of ⅐ NO (Fig. 6E). ⅐ NO-mediated changes in H3K9me2 were therefore only observed when the CIP was completely converted into DNIC. These results support our hypothesis that sequestration of chelatable iron via DNIC formation contributes to the inhibitory effects of ⅐ NO on histone demethylases.
Cellular Production of ⅐ NO Changes Global Methylation Status-Having established that histone methylation patterns were altered by physiological ⅐ NO concentrations, we sought to replicate these findings under conditions of endogenous ⅐ NO synthesis. We developed a coculture assay that allowed for measurements of global levels of H3K9me2 in Jurkat T cells grown in suspension over a monolayer of ⅐ NO-producing RAW 264.7 macrophages. Using this technique, we observed significant time-dependent increases in macrophage-derived ⅐ NO production (NO 3 Ϫ /NO 2 Ϫ ) and in Jurkat T cell H3K9me2 levels (Fig. 7A). Fig. 7B demonstrates that increases in H3K9me2 correlated to increases in the total amount of ⅐ NO to which the Jurkat T cells were exposed. By changing the ratio of Jurkat T/RAW 264.7 cells, we were able to modulate the amount of ⅐ NO exposure. Addition of the iNOS inhibitor aminoguanidine prevented NO 3 Ϫ /NO 2 Ϫ accumulation in the media and confirmed that changes in Jurkat T H3K9me2 only occurred in the presence of RAW 264.7-derived ⅐ NO (Fig. 7C). Finally, we measured changes in KDM3A protein in Jurkat T cells following 24 h of coculture (Fig. 7D). The total amount of KDM3A increased in response to ⅐ NO and was comparable with the change seen with 100 M DETA/NO. This effect was also abrogated by aminoguanidine. Overall, H3K9me2 and KDM3A changes in Jurkat T cells correlated strongly with ⅐ NO synthesis, and they were indistinguishable from what was observed utilizing ⅐ NO-donor compounds.
HIF-1␣ and sGC Are Not Required for ⅐ NO-mediated Regulation of Histone Modifications-Several reports indicate that KDM3A is up-regulated by hypoxia-inducible factor 1␣ (HIF-1␣) accumulation (55)(56)(57). Other studies have demonstrated that G9a is up-regulated under hypoxia (33). It is also well known that ⅐ NO can strongly induce HIF-1␣ accumulation under normoxic conditions (58). To investigate this pathway, we treated cells with ⅐ NO at 1 and 21% O 2 and measured KDM3A expression (Fig. 8A). ⅐ NO up-regulated KDM3A regardless of oxygen concentration. In fact, expression was higher at 1% O 2 when ⅐ NO is predicted to destabilize HIF-1␣ (59). Using a HIF-1␣ knockdown cell line (38), we further probed the role of HIF-1␣ in the up-regulation of KDM3A and down-regulation of G9a by ⅐ NO. Fig. 8B demonstrates that in the absence of HIF-1␣, KDM3A was still up-regulated to the same extent, and G9a was down-regulated in response to ⅐ NO. These data strongly suggest that ⅐ NO effects these changes in a HIF-1␣-independent manner. When ⅐ NO-mediated changes in histone methylation patterns were examined in the HIF-1␣ knockdown cell line, they were indistinguishable from what was observed in the wild type (Fig. 8C).
The activation of soluble guanylyl cyclase (sGC) by ⅐ NO is largely considered its most important biological function (24). To determine whether changes in KDM3A and G9a protein expression were a result of sGC activation by ⅐ NO, we pretreated cells with the sGC inhibitor ODQ. Inhibition of sGC did not affect the ability of ⅐ NO to up-regulate KDM3A or downregulate G9a (Fig. 8D) FIGURE 7. Paracrine regulation of H3K9me2 methylation and KDM3A expression in Jurkat T cells cocultured with ⅐ NO-synthesizing RAW 264.7 cells. RAW 264.7 cells in monolayer were stimulated with LPS for 6 h. Thereafter, media were replaced with LPS-free media followed by the addition of Jurkat T cells in suspension. Nitric oxide synthesis was verified in each experiment by chemiluminescent measurements of NO 3 Ϫ /NO 2 Ϫ accumulation in the media of the cocultured cells. Immunoblot and densitometric quantifications are shown in A. Temporal changes in H3K9me2 from Jurkat T cell total histone extracts after 0 -24 h coculture are shown. B, H3K9me2 measurements from Jurkat T cell total histone extracts after 24 h coculture with increasing concentrations of activated RAW 264.7 cells. U indicates incubation with nonactivated RAW 264.7 cells 1:3. C, changes in H3K9me2 from Jurkat T cell total histone extracts after 24 h coculture Ϯ the iNOS inhibitor aminoguanidine. D, changes in KDM3A protein in whole cell lysates of Jurkat T cells after 24 h coculture Ϯ the iNOS inhibitor aminoguanidine. All are representative immunoblots and chemiluminescent measurements of n Ն3. ** indicates p Ͻ 0.01 with respect to untreated controls, which are set arbitrarily to 1.0. AU, arbitrary unit.

DISCUSSION
The identification of endogenously produced, epigenetic regulatory molecules would significantly alter our understanding of gene expression in health and disease. This study provides compelling evidence that ⅐ NO falls into this category of molecules. Until now, direct inhibition of histone demethylase activity by any endogenously produced small molecule has not been reported. Our data indicate that ⅐ NO is capable of directly inhibiting the lysine-specific histone demethylase KDM3A, and likely other JMJC demethylases, resulting in variations of histone methylation patterns. As these changes dramatically influence chromatin structure and gene transcription, this model provides a direct mechanism for the regulation of a multitude of genes by ⅐ NO. Classical mechanisms of ⅐ NO signaling result from its ability to regulate enzyme function by binding to the heme center of proteins or through covalent or redox modifications of key protein residues. Our results highlight the importance of non-heme, iron⅐nitrosyl complexes and provide a direct link between ⅐ NO and significant epigenetic modifications. Thus, this alternative mode of ⅐ NO signaling is unique in that it does not invoke much of the commonly relied upon complex chemistry necessary to explain phenotypic consequences of ⅐ NO synthesis.
KDM3A, like all JMJC domain mononuclear Fe(II)-dependent dioxygenases, contains the 2-His-1-carboxylate facial triad. Unlike heme proteins, this enzyme coordinates iron at only three sites, leaving three additional sites available for sub-strate and cofactor binding (19,20). Although there is no crystal structure for KDM3A, it is thought that His-1120, Asp-1122, and His-1249 are most likely to participate in coordinating the iron. During catalysis, ␣-KG is bound in a bidentate fashion, which results in the formation of a five-coordinate Fe(II) center. At this stage, the enzyme is poised to bind O 2 (19,20) or, as we propose, ⅐ NO at the sixth coordination site (Fig. 9, inset). Although KDM3A contains 39 cysteine residues, S-nitrosothiol formation was not detected following ⅐ NO exposure. Moreover, removal of ⅐ NO restored KDM3A activity indicating that it binds to the enzyme reversibly. The inhibitory effect of ⅐ NO is therefore caused by neither covalent modification nor oxidative destruction of the enzyme. This reversibility may be important under physiological conditions by tying the regulation of ⅐ NO synthesis to the simultaneous fine-tuning of demethylase activity.  (60,61). This is agreement with our results that suggest that the affinity of this enzyme for ⅐ NO is greater than for O 2 . Although this requires further confirmatory studies, we hypothesize that, over a range of physiological O 2 tensions, the inhibitory effect of ⅐ NO would be even more profound.
Cells treated with physiological ⅐ NO concentrations demonstrated changes in H3K9 methylation patterns in a time-and concentration-dependent manner. Collectively, we have identified three unique mechanisms that explain how ⅐ NO can mediate these changes (Fig. 9). The JMJC class of demethylases contains a non-heme iron as a cofactor, and the chelatable iron pool is generally considered the source of this iron (51)(52)(53). Similarly, the iron utilized for DNIC assembly is also derived from this pool (39,40,54). Just as chemical iron chelators dramatically inhibit enzyme function by starving the cell of iron (62), we observed analogous effects of DNIC formation that likely result from shifts in the enzyme equilibrium toward its iron-free apoenzyme form (Fig. 9). Therefore, in addition to direct inhibition of demethylase activity by ⅐ NO, our data suggest that formation of DNIC indirectly inhibits enzyme activity by reducing the overall pool of available iron. This is significant as functional roles for DNIC assembly have remained obscure. With continued ⅐ NO exposure, KDM3A protein levels increased while the methyltransferase G9a protein levels decreased. Logic would predict that increasing demethylating enzymes and decreasing methylating enzymes would favor an overall decrease in histone methylation. What we observed, however, was significant increases in methylation at H3K9. We suspect that increased methylation subsequent to KDM3A inhibition by ⅐ NO results in compensatory changes in gene expression that attempt to override this inhibitory effect by modulating the relative concentrations of methyl-modifying enzymes.
Whether JMJC demethylase inhibition by ⅐ NO is an important physiological regulatory mechanism or associated more with ⅐ NO pathologies is also not clear. Both changes in histone methylation patterns and changes in methyl-modifying enzymes occurred in response to low physiological concentrations of ⅐ NO. Although both responses exhibited significant dose dependence on ⅐ NO concentration, as little as 16 Ϯ 8 nM steady-state ⅐ NO was sufficient. Under both normal and disease conditions, there are numerous locations where ⅐ NO is synthesized. Although ⅐ NO may be derived from enzymatic, dietary (NO 3 Ϫ /NO 2 Ϫ ), and pharmacological origins, the biological activities of ⅐ NO are independent of its source. As JMJC demethylases are ubiquitous in nature, the manner by which ⅐ NO inhibits these enzymes in a tumor cell will be identical to how it inhibits these enzymes in a neuron, myocyte, hepatocyte, macrophage, or bacteria. Thus, like most ⅐ NO-mediated biological responses, epigenetic effects will largely be a function of the cellular microenvironment that ultimately dictates the concentration and duration of ⅐ NO exposure (63). Therefore, although our studies examined global methylation patterns, it is likely that subtle changes in methylation will be localized around areas of basal ⅐ NO synthesis.
It is becoming increasing clear that the regulation of methylmodifying enzymes and post-translational histone modifications play important roles in tumor biology. For example, KDM3A expression was found to be elevated in certain types of cancer (64,65). In one particular study, immunohistochemical analysis revealed higher levels of KDM3A near the vessels of renal cell carcinomas, implicating a possible role in regulation of VEGF (65). Furthermore, inhibition of KDM3A improved outcomes of the anti-angiogenic VEGF antibody, bevacizumab (66). Thus, paradoxically, a combination of VEGF inhibitors and ⅐ NO might prove beneficial despite the fact that ⅐ NO drives up-regulation of angiogenesis through VEGF (67). Expression of KDM2B promotes differentiation of more aggressive forms of pancreatic cancer (68). KDM4A promotes cellular transformation through transcriptional repression of the tumor suppressor CHD5 (69), and KDM6B has been found to promote the epithelial-mesenchymal transition (70). Conversely, others have demonstrated that KDM6B acts as a tumor suppressor by regulating p53 nuclear stabilization (71). In colorectal cancer, depletion of KDM5B induced cellular senescence (72). In breast cancer, the methyltransferase G9a plays a critical role in the epigenetic regulation of epithelial-mesenchymal transition (73). G9a has also been shown to specifically methylate p53 at Lys-373, which correlates with inactivation of the protein (74). Thus, down-regulation of G9a by ⅐ NO may serve a tumor-suppressive function. As a whole, these studies underscore the complexity of epigenetic regulation in cancer biology and emphasize the importance of tumor type and specific epigenetic makeup. Furthermore, these novel epigenetic regulatory effects of ⅐ NO may have an important yet unrealized influence on cancer etiology.
Methyl-modifying enzymes are also involved in numerous other physiological and pathological conditions. During neurogenesis, methylation of H3K27 by KDM6B plays a critical role in the maintenance of the embryonic respiratory neuronal net- FIGURE 9. Mechanisms of ⅐ NO-mediated epigenetic regulation. 1, nitric oxide directly binds to the iron atom that is coordinated by the 2-His-1-carboxylate facial triad in the active center of KDM3A (inset). This inhibits demethylase activity by preventing the binding of O 2 . 2, nitric oxide reacts with iron in the CIP and two anions (typically thiols, "R-," like glutathione or cysteine) to form DNIC. Because the CIP is the major source of iron required as a cofactor by JMJC demethylases like KDM3A, the accumulation of DNIC results in iron sequestration and indirectly contributes to enzyme inhibition. 3, major H3K9 methylmodifying enzymes are the methyltransferase, G9a, and the demethylase, KD3MA. G9a adds a methyl group from the donor compound, S-adenosylmethionine, to yield mono-and dimethylated products. Nitric oxide up-regulates KDM3A and concomitantly down-regulates G9a. Expression changes in methylmodifying enzymes will influence the overall steady-state levels of H3K9 methylation status.
work (75). H3K4 demethylation by KDM5A drives differential transcriptional silencing during development (76). In human mesenchymal stem cells derived from bone marrow, histone demethylases KDM4B and KDM6B promote osteogenic differentiation (77). In hematopoietic stem cells, the H3K27 demethylase KDM6A regulates stem cell migration and hematopoiesis (78). These are just a few of the numerous examples that demonstrate the critical importance and the diversity of function of the methyl-modifying enzymes and suggest a potential role for regulation by ⅐ NO as well. In fact, a recent study has shown dendritic outgrowth in neuronal cells arising from the decreased trimethylation at H3K9 was due to nitric oxide-dependent degradation of the methyltransferase SUV39h1 (79).
Although we have focused on lysine 9 methylation of histone 3, this is just one of a vast number of possible modifications. Understanding how histone methylation status is controlled at specific DNA sequences by ⅐ NO and how these modifications affect local protein interactions will be required to elucidate the ultimate role of ⅐ NO in epigenetic regulation. This study revealed the following three distinct mechanisms whereby ⅐ NO could affect histone methylation patterns in general: direct inhibition of JMJC demethylases, reduction in iron cofactor availability, and regulation of KDM and KMT gene expression (Fig. 9). As we have discussed, both the methylation status as well as changes in steady-state levels of methyl-modifying enzymes result in a multitude of phenotypic consequences. Our results revealed the dramatic effects ⅐ NO could have on both of these aspects suggesting that the magnitude of ⅐ NO-attributable responses may be much greater than previously thought. Moreover, in addition to methylation, other histone lysine modifications such as acetylation can occur on the same residue. This is a reciprocal relationship, however, and each histone modification is mutually exclusive for a specific residue (i.e. a methylated lysine cannot be acetylated and vice versa). This means that ⅐ NO-mediated increases in methylation may have additional unrealized and far-reaching consequences such as precluding the formation of other critical regulatory histone marks.
Based on our current data, we classify ⅐ NO as an endogenously produced, epigenetic regulatory molecule. Although these epigenetic mechanisms do not eliminate or contradict currently accepted, well defined paths of ⅐ NO signaling, they may turn out to be equally as important under a diverse set of cellular conditions. Furthermore, these mechanisms address some fundamental problems in ⅐ NO chemical biology by providing a simple and plausible alternative hypothesis explaining the pleiotropic nature of a complex signaling molecule. The ability of ⅐ NO to change the topography of the epigenetic landscape reveals a new dimension to its list of regulatory functions and allows us to rethink classical ⅐ NO-signaling mechanisms.