Redox Signaling, Alkylation (Carbonylation) of Conserved Cysteines Inactivates Class I Histone Deacetylases 1, 2, and 3 and Antagonizes Their Transcriptional Repressor Function*

Cells use redox signaling to adapt to oxidative stress. For instance, certain transcription factors exist in a latent state that may be disrupted by oxidative modifications that activate their transcription potential. We hypothesized that DNA-binding sites (response elements) for redox-sensitive transcription factors may also exist in a latent state, maintained by co-repressor complexes containing class I histone deacetylase (HDAC) enzymes, and that HDAC inactivation by oxidative stress may antagonize deacetylase activity and unmask electrophile-response elements, thus activating transcription. Electrophiles suitable to test this hypothesis include reactive carbonyl species, often derived from peroxidation of arachidonic acid. We report that α,β-unsaturated carbonyl compounds, e.g. the cyclopentenone prostaglandin, 15-deoxy-Δ12,14-PGJ2 (15d-PGJ2), and 4-hydroxy-2-nonenal (4HNE), alkylate (carbonylate), a subset of class I HDACs including HDAC1, -2, and -3, but not HDAC8. Covalent modification at two conserved cysteine residues, corresponding to Cys261 and Cys273 in HDAC1, coincided with attenuation of histone deacetylase activity, changes in histone H3 and H4 acetylation patterns, derepression of a LEF1·β-catenin model system, and transcription of HDAC-repressed genes, e.g. heme oxygenase-1 (HO-1), Gadd45, and HSP70. Identification of particular class I HDACs as components of the redox/electrophile-responsive proteome offers a basis for understanding how cells stratify their responses to varying degrees of pathophysiological oxidative stress associated with inflammation, cancer, and metabolic syndrome.

We report that RCS covalently modified a subset of class I HDACs including HDAC1, -2, and -3 as well as the yeast homologue RPD3, but not HDAC8. Covalent modification of two conserved cysteines, Cys 261 and Cys 273 in HDAC1, occurred in a dose-and time-dependent manner that led to attenuation of histone deacetylase, changes in histone H3 and H4 acetylation patterns, derepression of LEF1⅐␤-catenin transactivation, and activation of RCS-sensitive genes repressed by HDACs. Our results imply that HDAC1, -2, and -3 isoenzymes are components of a redox/electrophile-responsive proteome that governs redox signaling in eukaryotes. Our data also suggest a mechanism by which cells might stratify their responses and adapt to varying intensities of oxidative stress and may explain, in part, how maladaptation can influence the etiology of cancer, atherosclerosis, or other chronic diseases.
Cell Lines and Culture-All mammalian cell lines were propagated in medium containing penicillin and streptomycin and 2 mM L-glutamine. SuperTOPflash (STF) and STF3a (HEK293) cell lines were maintained in advanced Dulbecco's minimum essential medium and 2% FCS (HyClone); A549 cells were grown in F-12 medium with 10% FCS; HL-60 cells were grown in RPMI with 10% FCS; and HCT116 cells were grown in McCoy's 5A medium with 10% FCS. Yeast cells were a gift from Dr. David Stillman, University of Utah (strain DY6092), grown in YP medium containing 2% glucose and at 30°C.
Immunochemical Analysis (Western Blots) and NeutrAvidin Pulldown of Biotinylated HDACs-Adherent mammalian cell lines were grown to ϳ90% confluency in 35-mm plates containing 1-2% serum. Cells were treated with biotinylated prostaglandins for the times and concentrations indicated. Following treatment, cells were washed twice with PBS, pH 7.4, at 4°C and harvested in 140 l of lysis buffer. Cells are completely lysed with one freeze/thaw cycle, and cellular debris was pelleted at 10,000 ϫ g. Between 100 and 200 g of total protein was incubated overnight at 4°C in 1 ml of total volume of PBS, pH 7.4, with 0.4% Tween 20 and 40 l of NeutrAvidin bead slurry to sequester biotinylated (protein⅐prostaglandin-biotin) proteins. Following incubation, beads were pelleted and washed four times with PBS, pH 7.4, containing 0.4% Tween 20. Proteins were released from beads by boiling in 25 l of 1.5ϫ Laemmli buffer containing 3.75% ␤-mercaptoethanol for 5 min and assayed by Western immunoblot for individual HDACs. Histone acetylation was analyzed by Western blotting of 10 g of total protein from whole cell lysates, separated on 10 -20% SDS-PAGE gels.
Saccharomyces cerevisiae cells harboring stably inserted RPD3-Myc (strain DY6092) were treated with 15d-PGJ 2 -B at A 600 of ϳ0.30 -0.35 in 10 ml of medium at 30°C for the time indicated. Yeast were pelleted and lysed using 0.7-mm zirconia beads (Biospec Products, Inc.) in lysis buffer. RPD3-Myc⅐15d-PGJ 2 -B conjugates were precipitated from lysates with NeutrAvidin beads as above and assayed by immunoblot for Myc tag. Total histone H4 acetylation was assayed by separating 15 g of total protein from whole cell lysates on 10 -20% SDS-PAGE gels followed by immunoblotting.
Cloning and Site-directed Mutagenesis-MAD⅐LEF and MAD(Pro)⅐LEF chimeras, from Dr. Don Ayer, University of Utah School of Medicine, were prepared by cloning the first 105 bp of MAD upstream and in frame of full-length LEF1 (isoform 1) using EcoRI in expression vector pcDNA3.1. The MAD(Pro)⅐LEF mutant was prepared by mutating t35c (Leu 12 to Pro 12 ) and g46c (Ala 16 to Pro 16 ). We generated the Cys to Ser mutants of HDAC1-FLAG (from Eric Verdin, University of California San Francisco) via site-directed mutagenesis using a QuikChange site-directed mutagenesis kit (Stratagene) with the following forward primers and their reverse complement primers: C261S, 5Ј-CGGTGGTCTTACAGAGTGGCTCAG-ACTCC-3Ј; C273S, 5Ј-TATCTGGGGATCGGTTAGG-TAGCTTCAATCTAACTATC-3Ј.
HDAC Activity Assays-We prepared 3 H-acetylated histone substrate from RKO cells as described previously (24). Radiolabeled histones (ϳ500 dpm/g) were purified by acid extraction and dialysis (25) and lyophilized before use. Inhibition of HDAC activity was measured by treating recombinant HDAC3⅐N-CoR1 (5 ng/l) in 60 l of reaction buffer (50 mM Tris/Cl, pH 8.0, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , and 0.25 mg/ml BSA) with vehicle, HDAC inhibitors TSA, sodium butyrate, or 15d-PGJ 2 -B and 4HNE for 15 min followed by the addition of 3 H-acetylated histone substrate (15,000 dpm/reaction dissolved in distilled H 2 O) and incubation at 37°C for 30 min. The reaction was quenched with 10 l of 10 N HCl/2.5 N acetic acid at 4°C. Samples were extracted with 600 l of ethyl acetate by vortexing for 10 s and centrifuged at 16,000 ϫ g for 10 min. Released [ 3 H]acetate was counted by adding 500 l of the organic fraction to scintillation fluid. HDAC activity is measured as (counts (from test sample) (dpm) Ϫ counts (from vehicle control)(dpm))/((time (h)/volume of sample (ml)).
WNT/␤-Catenin Signaling in STF and STF3a Cells-STF cells grown to ϳ60 -70% confluency in 35-mm wells were treated with vehicle or 15d-PGJ 2 (0 -30 M) in 2% FCS for 18 h. Cells were washed twice with PBS, pH 7.4, at 4°C and lysed in 200 l of reporter lysis buffer. Cell debris was pelleted at 10,000 ϫ g, and 20 g of total protein was used in a luciferase assay system (Promega). Vehicle signal was normalized to a value of 1 luciferase count per second.
STF3a cells (Dr. David Virshup, Duke NUS Graduate Medical School) were used as described (26). STF3a cells (6 ϫ 10 5 cells/well) were plated in 35-mm wells and grown overnight in antibiotic-free Dulbecco's minimum essential medium. 15d-PGJ 2 -sensitive Gene Expression by Reverse Transcription-Quantitative PCR-A549 cells were grown to ϳ70% confluency in 35-mm wells in antibiotic-free F-12 medium. HDAC1 WT and HDAC1 C261S/C273S plasmids were transfected (4 g of DNA and 10 l of Lipofectamine 2000), and medium was changed to 10% FCS after 6 h according to the manufacturer's directions. After 24 h, medium was changed to 1% FCS, and cells were treated with either vehicle or 15d-PGJ 2 (10 M). After 1 h, FCS was added to the cells to 10%, and the cells were incubated for 23 h before harvesting. Total RNA was isolated by using the RNeasy kit (Qiagen) according to manufacturer's directions followed by on-column DNase digestion. Total RNA was measured using a nanodrop spectrophotometer, and 2.5 g of RNA was reverse-transcribed to cDNA using a first-strand cDNA synthesis kit (Fermentas) with oligo(dT) 18 primers. The cDNA (1 l) was used directly for quantitative PCR using SYBR Green master mix (Roche Applied Science) and the following primer pairs:

Alkylation (Carbonylation) of Conserved Cysteine Residues in HDAC1, -2 and -3 Antagonizes Their Deacetylase Activity and
Transcriptional Co-repressor Function-Differential alkylation (carbonylation) of some, but not every class I HDAC, implies that their covalent modification is not indiscriminate and may involve particular, conserved cysteinyl thiol residues. Our data (Fig. 1) argue against the cysteine residues corresponding to Cys 151 of HDAC1 because these catalytic cysteines are shared by all class I HDACs, including HDAC8. Protein sequence alignment revealed two other cysteine residues in HDAC1, -2, and -3 and RPD3 that we considered candidates for modification ( Fig. 2A). These correspond to Cys 261 and Cys 273 residues of HDAC1, which appear to be surface-accessible, according to homology models (35). Notably, HDAC8, which was not susceptible to alkylation by 15d-PGJ 2 -B, has a Leu substituted for Cys at one conserved position and a displaced Cys with a different flanking residue at the other conserved position. The motif surrounding Cys 273 of HDAC1 shares appreciable homology with motifs surrounding Cys residues alkylated by cyclopentenone prostaglandins in other proteins, such as the T-loops of the IB kinase ␤ and LKB1 Ser/Thr kinases, h-Ras, phosphatase and tensin homolog (PTEN), and the p50 and p65 subunits of NFB. To determine whether any of these cysteine residues reacted with 15d-PGJ 2 -B, we used site-directed mutagenesis of an HDAC1-FLAG construct to make Cys 3 Ser mutants, C261S and C273S. Plasmids expressing these HDAC1-FLAG mutants were transfected into HEK293 cells, which were subsequently treated with 5 M 15d-PGJ 2 -B. When compared with the wild type HDAC1-FLAG, each of these single-mutant HDAC1-FLAG proteins formed ϳ70% less covalent adduct with 15d-PGJ 2 -B (Fig. 2B, lanes 2 and 3 versus lane  1). Each showed comparable alkylation, consistent with reaction at Cys 261 and Cys 273 . An HDAC1-FLAG double-mutant, C261S/C273S, showed ϳ95% less alkylation by 5 M 15d-PGJ 2 -B (Fig. 2B, lane 4 versus lane 1), supporting the conclusion that both Cys 261 and Cys 273 are susceptible to alkylation (carbonylation) by ␣,␤-unsaturated carbonyl compounds. A corresponding set of Cys 3 Ala mutants gave similar results. There was a low but detectable formation of 15d-PGJ 2 -B adduct with the C261S/C273S HDAC1 double mutant, ϳ5%, implying that other sites may be alkylated to a minor degree.
To better understand the functional consequences of HDAC alkylation by RCS, we tested in vitro deacetylase activity of HDAC3⅐N-CoR1 treated with 4HNE, an enal that can be generated by oxidative stress (36), and 15d-PGJ 2 -B. Inhibition of the HDAC3⅐N-CoR1 complex by these RCS was concentrationdependent and intermediate in potency, relative to TSA and butyrate. Half-maximal inhibition (IC 50 ) required 10 nM TSA, 13 M 15d-PGJ 2 -B, 98 M 4HNE, and 140 M butyrate (Fig. 3).  Maximal inhibition of HDAC3⅐N-CoR1 by 15d-PGJ 2 -B and 4HNE is 50 and 70%, respectively. The lack of complete inhibition upon alkylation suggests that inhibition of deacetylase activity is not likely the major consequence of HDAC alkylation yet implies that its modifiable thiolate residues are near the catalytic site or that they are allosterically linked to the catalytic site.
Class I HDACs in co-repressor complexes play a major role in transcriptional silencing (18,37). Thus, we hypothesized that alkylation and inhibition of HDACs by 15d-PGJ 2 would antagonize their co-repressor function in HEK293 cells harboring an STF luciferase reporter with seven LEF/TCF-binding sites (38). LEF1 recruits HDAC1 as a co-repressor of transcription in this system. Others have reported that inhibition of HDAC1 by TSA relieves LEF1-mediated repression and converts it to a transcriptional activator (23). If 15d-PGJ 2 -B or other RCS inhibited HDAC1 in an analogous experiment, LEF1 would be similarly activated, leading to enhanced luciferase expression. Consistent with this hypothesis, luciferase expression rose 20 -80-fold as a function of 15d-PGJ 2 concentration in HEK293 cells stably expressing an STF reporter construct (Fig. 4A).
To determine whether 15d-PGJ 2 , and related RCS, act independently of other, upstream WNT signaling events, we also used STF3a cells, a cell line that stably expresses both the STF luciferase reporter gene and the wnt3a gene. In these cells, constitutively secreted WNT3a binds to FRZL, its cognate receptor, leading to nuclear accumulation of ␤-catenin and transactivation of the STF luciferase reporter gene (26). In conjunction, we used plasmids expressing chimeric MAD⅐LEF1 genes. The MAD component of the chimera associates with mSin3a in a multiprotein complex that recruits HDAC1 as a co-repressor (Fig. 4C) (18). In transfected STF3a cells, the MAD⅐LEF1 chimera protein caused maximal repression of the luciferase reporter gene (Fig. 4B, M:L bar), relative to the basal luciferase activity in mock-transfected STF3a cells (Fig. 4B, Vehicle bar). Conversely, transfection and expression of a mutant, dysfunctional MAD variant, MAD(Pro)⅐LEF1, which cannot recruit HDAC1 to co-repressor complexes, caused maximal derepression (i.e. induction) of the luciferase reporter gene ( HDAC Alkylation by 15d-PGJ 2 Promotes Changes in Histone Acetylation and Promotes Activation of Genes That Are Induced by 15d-PGJ 2 -To substantiate our observation that RCS inhibited HDAC activity and to determine the biological outcome of this inhibition, we first measured changes in histone acetylation in yeast and mammalian cells exposed to 15d-PGJ 2 -B. Immunoblot analysis showed that total histone H4 acetylation increased, coincident with covalent modification of RPD3 in S. cerevisiae (Fig. 5, A and B). Moreover, 15d-PGJ 2 -B differentially altered the acetylation pattern at specific lysine residues of histone H3 in A549 cells concurrent with alkylation of HDACs in A549 cells (Fig. 5D). For instance, levels of H3K9Ac rose, whereas H3K14Ac remained constant (Fig. 5C).
Because H3K9Ac is associated with transcriptionally active chromatin (39,40), we sought to determine whether carbonylation of HDACs regulates genes that are sensitive to 15d-PGJ 2 (Fig. 6B). Various cell lines increase expression of genes associated with redox regulation (e.g. ␥-glutamylcysteine synthetase), heat shock response (e. g. HSP70, HO-1), and p53-responsive  , HEK293 cells, harboring a stably integrated STF luciferase reporter construct that is transactivated by nuclear LEF/␤-catenin, were treated with 0 -30 M 15d-PGJ 2 for 18 h, which led to increased luciferase expression, in the absence of WNT ligand. LCPS, luciferase counts per second. B, STF3a cells, which harbor the STF luciferase reporter construct, plus a construct for stably secreting WNT3a, were transfected with a MAD⅐LEF chimera or a mutant form MAD(Pro)⅐LEF for 24 h followed by treatment for 18 h with 15d-PGJ 2 ,⌬12-PGJ 2, 4HNE,or the known HDAC inhibitors TSA or butyrate. The bar labeled Vehicle represents basal expression of luciferase; the bar labeled M(P):L represents a fully derepressed system; the bar labeled M:L represents a repressed system. Luciferase activity in all conditions is inversely proportional to LEF1 repression of the pTOPFLASH gene. The enones (15d-PGJ 2 , ⌬12-PGJ 2 ) and the enal (4HNE) antagonized the HDAC-mediated transcriptional repression in cells expressing the MAD⅐LEF1 chimera, comparably to the HDAC inhibitors TSA and butyrate (bars labeled M:L plus agent). C, scheme depicting the effect of MAD⅐LEF and mutant MAD(Pro)⅐LEF chimeras on HDAC-mediated co-repression at the pTOPFLASH promoter. Left panel, MAD⅐LEF chimera enables LEF repression by associating with mSin3a containing HDAC1. Inhibition of HDAC1 in this system leads to LEF activation. Right panel, the MAD(Pro)⅐LEF mutant chimera cannot associate with mSin3a, and therefore, LEF is in a minimally repressed state, i.e. it enables LEF/␤catenin (␤-Cat.) transactivation.
genes (e.g. Gadd45)) when treated with the cyclopentenone prostaglandin 15d-PGJ 2 (41). The mechanisms of 15d-PGJ 2mediated gene up-regulation are not fully understood, yet transcriptional repression of these genes is governed in part by altering the post-translational acetylation of histones by HDACs and activation of appropriate transcription factors. Therefore, the presence of 15d-PGJ 2 may not only promote activation of transcription factors but may also lead to coordinate disruption of HDAC-mediated gene repression. To test this hypothesis, we transiently transfected A549 cells with either wild type HDAC1 (HDAC1 WT ) or 15d-PGJ 2 -insensitive mutant HDAC1 (HDAC1 C261S/C273S ). Both were treated with either vehicle or 15d-PGJ 2 (10 M) for 24 h. We then measured the mRNA levels of HO-1, Gadd45, and HSP70 via reverse transcription-quantitative PCR. Consistent with previous reports, 15d-PGJ 2 induced expression of HO-1, Gadd45, and HSP70 genes in cells harboring HDAC1 WT when compared with cells treated with vehicle alone (41). Consistent with our hypothesis, A549 cells expressing HDAC1 C261S/C273S are less responsive to 15d-PGJ 2 treatment, inducing less expression of HO-1 (42% Ϯ 4%), Gadd45 (43% Ϯ 1%), and HSP70 (28% Ϯ 1%) when compared with 15d-PGJ 2 -treated cells harboring HDAC1 WT (Fig.  6A). The observed expression of these genes in cells expressing HDAC1 C261S/C273S is likely due to endogenous wild type HDACs competing with the transfected mutant.

DISCUSSION
We conclude that RCS, typified by cyclopentenone prostaglandins of the J-and A-series, alkylate a subset of mammalian class I HDACs that include HDAC1, -2, -3, as well as RPD3, a yeast class I HDAC homolog. Alkylation (carbonylation) occurred at two conserved cysteine residues, corresponding to Cys 261 and Cys 273 in HDAC1. HDAC8, which differs from other class I HDACs in this regard, was not susceptible to alkylation. Our results concur with a recent mass spectrometry survey (42), which found HDAC1 and -2, but not HDAC8, among 417 cellular proteins modified by 100 M 4HNE (43). HDAC2 can also be nitrosylated at Cys 262 and Cys 274 (44,45). Because residues sensitive to ROS and reactive nitrogen species are often sensitive to modification by RCS, our results suggest that the homologous cysteine residues in HDAC1 and HDAC3 as well as yeast RPD3 could be sensitive to nitrosylation and oxidation.
RCS (15d-PGJ 2 and 4HNE) that carbonylated HDACs also inhibited deacetylation of [ 3 H]acetyl-histones, for example, HDAC3⅐N-CoR1. Conversely, they did not inhibit deacetylation of synthetic, fluorescent tetrapeptide substrates. These results imply that carbonylation of redox-sensitive cysteine residues, e.g. Cys 261 and Cys 273 in HDAC1, disrupts interactions between HDACs and their endogenous histone substrates, rather than disrupting their catalytic site per se. Moreover, these data indicate that a catalytic cysteine, corresponding to Cys 151 in HDAC1, was not susceptible to alkylation by RCS. If this catalytic cysteine were affected, one would expect inhibition of catalysis, regardless of substrate type. Likewise, one would expect to see alkylation of all class I HDACs that share the conserved active site cysteine, including HDAC8 (Cys 153 ); however, this was not the case.
Disruption of cellular deacetylase activity by RCS coincided with changes in chromatin dynamics, the function of HDACs situated in co-repressor complexes, and derepression/activation of RCS-sensitive genes. For instance, alkylation of HDACs in mammalian cells: (i) altered the so called histone code, e.g. differentially elevating the level of H3K9Ac, but not H3K14Ac alkylation; (ii) antagonized the transcriptional co-repressor function of HDAC1 in a LEF1⅐␤-catenin model system; and (iii) promoted transcription of HDAC1-repressed genes, HO-1, Gadd45, and HSP70.
Overall, our results suggest that some DNA-binding sites (response elements) for redox-sensitive transcription factors exist in a latent state, which is maintained by co-repressor complexes containing class I HDAC enzymes that favor heterochromatin formation via their deacetylation of histones. Carbonylation of HDAC1, -2 and -3 antagonizes their deacetylase activity, disrupts their transcriptional repression, modulates chromatin dynamics, and unmasks electrophile-response elements latent within heterochromatin. Thus, cellular redox signaling can involve liberation of latent transcription factors (e.g. KEAP⅐NRF2, NFB, LEF/␤-catenin, p53) coordinated with liberation of their DNA recognition sites in gene promoters. Oxidative stress from various pathophysiological conditions is not uniformly cytotoxic. Our model of redox signaling suggests a mechanism to coordinate the activity of co-repressor complexes containing HDAC1, -2, and -3 and latent, redox-sensitive transcription factors. This may enable cells to stratify their anti-oxidant response according to the gravity, or nature, of the oxidant threat. For example, inflammation consists of a "wounding" phase where redox stress annihilates pathogens and a "healing" phase where redox stress influences repair and regeneration of damaged host tissue (46). The transition between phases depends on gradual exhaustion of inflammatory mediators and conversion of certain pro-inflammatory mediators, e.g. PGD 2 , into anti-inflammatory metabolites (32,47,48). Elements of the inflamed site itself, e.g. ROS, albumin, fibroblasts and neutrophils, orchestrate this conversion (27, 49 -51). ROS cause non-enzymatic peroxidation of essential fatty acids, like arachidonic acid (52). Arachidonic acid hydroperoxides transform readily into RCS (6, 53) that include acrolein (2-propenal (54, 55), 4HNE (43), and cyclopentenone prostaglandins, PGA 2 , ⌬12-PGJ 2 , and 15d-PGJ 2 (56). Covalent modification of NFB and IB kinase ␣/␤ proteins by these RCS (protein carbonylation) seems to be a "switch" to terminate inflammation (12,29,(57)(58)(59). Could carbonylation also be a switch to initiate repair and regeneration of tissue damaged by inflammation? If so, which proteins would be carbonylated, and how might their involvement sometimes favor tumorigenesis? We investigated HDACs, based first, on reports that cyclopentenone prostaglandins accumulated in the nucleus and associated irreversibly with chromatin and other nuclear proteins (13), and second, on reports that HDACs mediate WNT/␤-catenin signaling (23), a determinant of cell growth, differentiation, and disease.
Redox stress and inflammatory processes create a milieu rich in ROS and RCS. As an investigative tool to study HDAC modification by RCS, we have used cyclopentenone prostaglandins and 4-hydroxynonenal to represent RCS generated during lipid oxidation and arachidonic acid metabolism, although other RCS are undoubtedly present during redox stress and inflammation (2). Exposure of cells in vitro to a single, bolus addition of separate RCS (1-10 M) is difficult to relate directly to progressive, cumulative exposures that occur in vivo during various types of oxidative stress. Quantification of RCS is complicated by their reactivity. Because RCS conjugate with peptides (GSH) and proteins, analysis of the level of "free" compound does not necessarily reflect, and more likely underestimates, the concentrations and fluxes of various RCS. Carbonylation is a marker of oxidative stress, and chronic exposure or elevated concentrations of RCS have been implicated in many diseases including, but not limited to, Alzheimer disease, chronic obstructive pulmonary disorder (60), arthritis, diabetes, and Parkinson disease, as reviewed by Dalle-Donne et al. (61). Although aberrant modification and dysregulation of proteins have been the major focus of most protein carbonylation studies, it is becoming more evident that RCS may play an active role in redox signaling in ways analogous to ROSmediated signal transduction (62)(63)(64)(65).