Transcriptional regulation of the human ferritin gene by coordinated regulation of Nrf2 and protein arginine methyltransferases PRMT1 and PRMT4

Abstract

Antioxidant genes such as ferritin are transcriptionally activated in oxidative stress via the antioxidant responsive element (ARE), to which nuclear factor-E2-related factor 2 (Nrf2) binds and activates transcription. Histone modification plays a cooperative and essential role in transcriptional regulation; however, its role in antioxidant gene transcription remains elusive. Arsenic exposure activated ferritin transcription via the ARE concomitant with increased methylation of histones H4Arg3 (H4R3) and H3Arg17 (H3R17). To test our hypothesis that histone H4R3 and H3R17 methylation regulates ferritin transcription, H4R3 and H3R17 protein arginine (R) methyltransferases 1 and 4 (PRMT1 and PRMT4) were investigated. Arsenic exposure of human HaCaT keratinocytes induced nuclear accumulation of PRMT1 and PRMT4, histone H4R3 and H3R17 methylation proximal to the ARE, but not to the non-ARE regions of ferritin genes. PRMT1 or PRMT4 knockdown did not block Nrf2 nuclear accumulation but inhibited Nrf2 binding to the AREs by ∼40% (P<0.05), thus diminishing ferritin transcription in HaCaT and human primary keratinocytes and fibroblasts, causing enhanced cellular susceptibility to arsenic toxicity as evidenced by 2-fold caspase 3 activation. Focused microarray further characterized several oxidative stress response genes are subject to PRMT1 or PRMT4 regulation. Collectively, PRMT1 and PRMT4 regulate the ARE and cellular antioxidant response to arsenic.—Huang, B.-W., Ray, P. D., Iwasaki, K., Tsuji, Y. Transcriptional regulation of the human ferritin gene by coordinated regulation of Nrf2 and protein arginine methyltransferases PRMT1 and PRMT4.

Cells are constantly exposed to harmful xenobiotics from the environment, as well as endobiotics and reactive oxygen species produced during various metabolic activities. To combat and detoxify such injurious chemicals and oxidants, cells evolved antioxidant systems and detoxification signaling pathways to biotransform them to less toxic molecules (1). On exposure to such compounds, cells transcriptionally induce a battery of antioxidant detoxification genes such as hemeoxygenase-1 (HO-1), NAD(P)H quinone oxidoreductase-1 (NQO1), and glutathione S-transferases (GSTs) (2). In addition to these metabolic enzymes, we previously reported that ferritin, the major iron storage protein composed of multimeric H and L subunits, is transcriptionally and post-transcriptionally up-regulated under oxidative stress conditions (3–6). The H and L subunits of ferritin play key roles in iron storage through the ferroxidase activity of the H subunit and structural stabilization imparted by the L subunit, which facilitate the oxidation and efficient incorporation of Fe²⁺ into a multimeric ferritin shell (7). Iron is an essential element for a wide variety of cellular activities including metabolism, proliferation, and differentiation; however, excess free iron is toxic to cells because it catalyzes production of the highly reactive hydroxyl radical through the Fenton reaction, resulting in damage to macromolecules, including DNA, proteins, and lipids (7, 8). Therefore, transcriptional up-regulation of ferritin under oxidative stress is an important cellular defense mechanism by chelating excess intracellular free iron, thereby minimizing hydroxyl radical formation. It should be noted that when iron levels are excessive, ferritin is up-regulated at the translational level by iron via the well-characterized IRE-IRP system, while in cells under oxidative stress, ferritin is up-regulated at the transcriptional level in an iron-independent manner (8–10). Transcriptional activation of ferritin and other antioxidant detoxification genes is regulated via a conserved enhancer element, termed the antioxidant responsive element (ARE; ref. 11). The core ARE sequence is an AP1-like TGACnnnGCA motif (11, 12), to which nuclear factor-E2-related factor 2 (Nrf2) and small Maf proteins are recruited and thus activate transcription of antioxidant genes (2, 13).

Since DNA is wrapped around core histones (an octamer of H2A, H2B, H3, and H4) and tightly packed as nucleosomes, dynamic and reversible changes in chromatin structure and conformation through post-translational modifications of core histones is necessary to allow transcription factors access to their specific cis-acting elements and carry out their functions properly (14). One well-studied coactivator that collaborates with various transcription factors is the histone acetyltransferase (HAT). With regard to the ARE enhancer, we and others have reported that p300, CREB binding protein (CBP; refs. 15–17), and monocytic leukemia zinc-finger protein (MOZ; ref. 18) are involved in transcriptional activation of ferritin H and GST genes through the ARE under oxidative stress conditions. N-terminal tails of core histones have multiple lysine (Lys), arginine (Arg), and serine/threonine (Ser/Thr) residues that are subject to reversible post-translational modifications such as acetylation, methylation, and phosphorylation (14, 19). Indeed, acetylation of histone H3 Lys9 and Lys18 (H3K9 and H3K18) in conjunction with recruitment of p300 and CBP HATs were associated with t-BHQ (tert-butylhydroquinone)-induced ARE activation (15). These HATs may play a role in acetylation of H3K9 and H3K18, as well as direct acetylation of Nrf2 that was shown to activate Nrf2 transcription function (20, 21). Accumulating evidence indicates that post-translational modifications of histones play a crucial role in transcriptional regulation; however, particular histone modifications and enzymes involved in antioxidant gene regulation under oxidative stress remain largely uncharacterized.

The protein methyltransferases (PMTs), composed of protein lysine methyltransferases (PKMTs) and protein arginine methyltransferases (PRMTs), have been characterized as important regulators of gene transcription by facilitating the transfer of methyl groups to specific Lys and Arg residues, respectively, in both histones and nonhistone proteins (22). N-terminal histone tails contain Lys residues that may be mono-, di-, or trimethylated, or Arg residues that are mono- or dimethylated (either symmetric or asymmetric), thus providing a platform for interaction with methyl-Lys or methyl-Arg binding proteins. These proteins contain such binding motifs as the Chromo domain (binding to methyl-Lys) or Tudor domain (binding to methyl-Arg) (23), allowing new protein-histone interactions that either activate or repress gene transcription in a context-dependent manner (24). For instance, methylation at Lys 4 and Lys 9 on histone H3 by several PKMTs were characterized as marks of transcriptional activation and repression, respectively (14). Similarly, Arg methylation by PRMTs on histones causes either transcriptional activation or repression (25). The mammalian PRMT family comprises at least 9 members (PRMT1–PRMT9), in which type I enzymes, such as PRMTs 1, 3, 4, 6, and 8, catalyze monomethylation and asymmetric dimethylation of Arg, whereas the type II enzymes, such as PRMT5, catalyze monomethylation and symmetric dimethylation of Arg (25–27). Among them, PRMT1 is the major type I enzyme that catalyzes methylation of histone H4 Arg 3 (H4R3; ref. 28), which is a mark of transcription activation (29). PRMT4 [coactivator-associated arginine methyltransferase 1 (CARM1)] is another PRMT involved in transcriptional activation that was originally shown to bind steroid receptor coactivators (SRCs; ref. 30) and was subsequently characterized as an enzyme involved in methylation of histone H3 Arg 17 (H3R17; ref. 31).

Arsenic exposure is associated with toxicity in various cell types, as well as several cancers, including skin, bladder, and lung. A potential mechanism underlying arsenic-induced toxicity and carcinogenicity is that of oxidative stress, which mediates macromolecular damage of nucleic acids and proteins (32, 33). Oxidative stress is implicated in various diseases, such as cancer and neurodegeneration; therefore, gaining insight into the molecular mechanism through which antioxidant-detoxifying genes are regulated is important in understanding these diseases. In this study, we report that arsenic treatment induced transcription of the ferritin H and L genes via the AREs in human HaCaT keratinocytes and primary culture cells. During arsenic treatment, we observed that PRMT1 and PRMT4 transiently accumulated in the nucleus, and chromatin immunoprecipitation (ChIP) assays revealed that methylation of H4R3 and H3R17 was enriched in the ARE enhancer regions. Knocking down PRMT1 or PRMT4 inhibited arsenic-induced Nrf2 recruitment to the ARE, as well as ferritin mRNA expression, suggesting the importance of PRMT1 and PRMT4 in arsenic-induced ferritin transcription through the ARE. Furthermore, human oxidative stress signaling pathway polymerase chain reaction (PCR) array revealed that several oxidative stress response genes such as glutamate-cysteine ligase (GCLC) are subject to PRMT1 or PRMT4 regulation.

MATERIALS AND METHODS

Cell culture and chemical reagents

HaCaT human keratinocytes (34) were cultured in DMEM containing 10% FBS (Mediatech, Manassas, VA, USA). K562 human erythroleukemia cells (American Type Culture Collection, Manassas, VA, USA) were cultured in RPMI1640 medium supplemented with 25 mM Hepes, 0.3 mg/ml l-glutamine, and 10% FBS. Human primary cells and culture media were purchased from Life Technologies (Carlsbad, CA, USA). Human primary dermal fibroblasts were cultured in medium 106 supplemented with low serum growth supplement. Human primary epidermal keratinocytes were cultured in EpiLife medium with human keratinocyte growth supplement. Cells were incubated at 37°C in a humidified 95% air, 5% carbon dioxide atmosphere. Sodium arsenite (NaAsO₂; Thermo Fisher, Miami, OK, USA) and adenosine-2′,3′-dialdehyde (AdOx; Sigma-Aldrich, St. Louis, MO, USA) were dissolved in distilled water and DMSO, respectively.

Plasmids and antibodies

The −4.5-kb ARE(+) and −4.4-kb ARE(−) human ferritin H 5′ upstream enhancer and promoter fused to a firefly luciferase reporter gene were described previously (6). Antibodies used in this work were purchased from the following companies: anti-ferritin H (sc-25617) and anti-Nrf2 (sc-13032x) and anti-rabbit immunoglobulin G (IgG; sc-2027), Santa Cruz Biotechnology (Santa Cruz, CA, USA); anti-histone H4 (ab7311), Abcam (Cambridge, MA, USA); anti-lamin B (Ab-1), Oncogene (EMD Biosciences, La Jolla, CA, USA); anti-lactate dehydrogenase (LDH; AB1222), Chemicon (Temecula, CA, USA); anti-caspase 3 (8G10) and anti-histone H3 (9715), Cell Signaling Technology (Danvers, MA, USA); anti-human ferritin (F5012, used for detection of ferritin light chain) and anti-β-actin (A5441), Sigma-Aldrich; anti-PRMT4 (A300-421), Bethyl Laboratories (Montgomery, TX, USA); anti-dimethyl histone H4R3 antibody for ChIP assays (39706), Active Motif (Carlsbad, CA, USA); anti-dimethyl histone H4R3 antibody for Western blotting (07–213), Millipore (Billerica, MA, USA); anti-dimethyl histone H3R17 for ChIP assays (07–214), Millipore; and anti-dimethyl histone H3R17 for Western blots (ab8284), Abcam.

DNA transfection and luciferase reporter assay

HaCaT cells were transiently transfected by electroporation (Gene Pulser X-Cell; Bio-Rad, Hercules, CA, USA) with −4.5-kb ARE(+) or −4.4-kb ARE(−) human ferritin H luciferase reporter plasmids. pRL-null (10 ng; Promega, Madison, WI, USA) was cotransfected as an internal control and incubated for 48 h. Cells were then treated with sodium arsenite for 20–24 h and subjected to luciferase assays using Dual Luciferase reagents (Promega); Renilla luciferase activity was used to normalize firefly luciferase expression driven by the ferritin H promoter.

Histone extraction

Histone extraction was carried out according to the published method with minor modifications (35). Briefly, HaCaT cells were suspended and washed with buffer C (20 mM HEPES, pH 7.9; 0.1% Triton X-100; 1.5 mM MgCl₂; 1 mM PMSF; and 1 mM DTT), followed by resuspension in buffer D (10 mM HEPES, pH 7.9; 10 mM KCl; 1.5 mM MgCl₂; 1 mM PMSF; and 1 mM DTT). Then H₂SO₄ was added to a final concentration of 0.4 N and rocked for 30 min. Extracted histone proteins were precipitated with TCA, washed once with 0.1% HCl in acetone, then with 100% acetone, and dissolved in 2 M urea.

Western blotting

Whole-cell lysates or nuclear and cytosolic fractions were loaded on 7.5–15% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (Mini-Protean 3, Bio-Rad), transferred to a polyvinylidene difluoride (PVDF) membrane (Thermo Scientific, Rockford, IL, USA), and incubated with a primary antibody at 4°C overnight, followed by a secondary antibody-conjugated with horseradish peroxidase at room temperature for 1.5 h. Proteins were visualized using HyGLO detection reagent (Denville Scientific, Metuchen, NJ, USA), ECL prime (GE HealthCare, Waukesha, WI, USA), or Femtomax (Rockland Immunochemicals, Gilbertsville, PA, USA). Cell fractionation was carried out with a nuclear isolation kit (Active Motif).

Northern blotting

Total RNA was isolated by using TRI Reagent RT (Molecular Research Center, Cincinnati, OH, USA). Total RNA (2–10 μg) was separated on 0.7% agarose gel with 5% formaldehyde in 3-(N-morpholino)-propanesulfonic acid buffer, followed by capillary transfer to an Immobilon-NC nitrocellulose membrane (Millipore). ³²P-labeled human ferritin H or ferritin L cDNA probe was hybridized with membranes at 42°C overnight and washed with washing buffer (0.1% SDS in 0.5× SSC) at 52°C. The dried membranes were subjected to autoradiography. Staining RNA with ethidium bromide was used for evaluation of equal RNA loading and positions of ribosomal 18S and 28S RNA. Quantitation of autoradiography was performed with Multi Gauge software (Fujifilm, Tokyo, Japan).

Small interfering RNA (siRNA) transfection

Ninety percent confluent HaCaT cells in 100-mm culture dishes were electroporated with 100 pmol of nontargeting siRNA (siControl; D-001210-01), siPRMT1 (J-010102-07), or siPRMT4 (J-004130-05) from Thermo Fisher or siFerritin H (S100300251) or siFerritin L (S100300258) from Qiagen (Valencia, CA, USA) by Gene Pulser Xcell in 100 μl of siRNA transfection medium (sc-36868; Santa Cruz Biotechnology). After electroporation, cells were incubated at room temperature for 10 min, transferred into cell culture dishes with growth medium, and incubated at 37°C for 48 h. Cells were then subjected to sodium arsenite treatment and Western blotting, Northern blotting, reverse transcription (RT)-PCR, or ChIP assay.

Real-time quantitative PCR (qPCR)

RNA was reverse transcribed to cDNA with iScript cDNA Synthesis Kit (Bio-Rad). cDNA was subjected to SYBR Green qPCR with iQ SYBR Green Supermix (Bio-Rad) by using the primer pairs specific for detection of target genes (ferritin H: forward, 5′-ACTGATGAAGCTGCAGAACC-3′ and reverse, 5′-GTCACCCAATTCTTTGATGG-3′; ferritin L: forward, 5′-CAGCCTGGTCAATTTGTACCT-3′ and reverse, 5′-CGGTCGAAATAGAAGCCCAGAG-3′) in the CFX96 PCR system (Bio-Rad) with the parameters of 95°C for 3 min followed by 40 cycles of 95°C for 10 s and 60°C for 45 s. The relative efficiency of each primer set was determined from RNA standard curve by sequential 10-fold dilutions. Expression of mRNA level of each gene was normalized to β-microglobulin (ΔC_t) and defined as the change of C_t in treated samples relative to untreated control or siControl (ΔΔC_t). Exponential ΔΔC_t values were converted to linear values (2^−ΔΔCt) for fold induction.

ChIP assay

ChIP assays were carried out according to the fast ChIP method (36). Briefly, HaCaT cells were fixed with 1.42% formaldehyde for chromatin cross-linking and quenched with 125 mM glycine. Cell lysates were sonicated to shear chromatin DNA, as described previously (37). ChIP by IgG or specific antibodies against interested proteins were performed at 4°C in a sonication bath (Branson 2510, 40 mHz; Branson Ultrasonics Corp., Danbury, CT, USA) for 15 min and incubated with protein A agarose/ssDNA bead slurry (16–157; Millipore). After washing and decrosslinking, the genomic DNA was subjected to SYBR Green qPCR with iQ SYBR Green Supermix (Bio-Rad) by using primer pairs for ferritin H and L AREs or non-ARE regions (ferritin H ARE: forward, 5′-TCCAGGTCTTATGACTGCTC-3′ and reverse, 5′-GATGAGAGAAGAGCCAAGC-3′; ferritin L ARE: forward, 5′-TAGTTCTGAGGGTCCCACCA-3′ and reverse, 5′-GGTATCTGGGGTCCTTGTTG-3′; ferritin H non-ARE: forward, 5′-TAGTCCCTGGCTGCTGATCT-3′ and reverse, 5′-AGTGCCTCCTCATGGAAATG-3′; and ferritin L non-ARE: forward, 5′-ACCTCAGGAGGCCATACCTT and reverse, 5′-TCCCTTCCTCTCTGTCCTCA-3′). The relative efficiency of each primer set was determined using input genomic DNA. The DNA in each immunoprecipitated sample was normalized to input (ΔC_t) and defined as the change of C_t in treated samples relative to control (ΔΔC_t). Exponential ΔΔC_t values were converted to linear values (2^−ΔΔCt) for fold induction.

Human oxidative stress signaling pathway RT² profiler PCR array

Total RNA was isolated using RNeasy Plus Mini Kit (Qiagen) from HaCaT cells transfected with siControl, siPRMT1, or siPRMT4 and treated with 10 μM sodium arsenite. RNA was reverse transcribed to cDNA using the RT² First Strand Kit (SABiosciences, Frederick, MD, USA) and the Human Oxidative Stress Plus Pathway RT² Profiler PCR Array (SABiosciences) was performed with the CFX96 real-time PCR system (Bio-Rad). Expression of 84 oxidative stress pathway genes for each condition was profiled with the parameters of 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The complete array data obtained from 3 independent experiments were analyzed with the company's online software RT² Profiler PCR Array Data Analysis 3.5. To normalize gene expression (2^−ΔCt) and determine the fold change between groups (2^−ΔΔCt), 2 housekeeping genes (GAPDH and RPLP0) were used as internal controls. We used the following cutoff criteria: >1.5-fold induction or repression of expression with sodium arsenite treatment was considered to represent significantly up- or down-regulated gene expression, and >10% expression changes with PRMT1 or PRMT4 knockdown were considered PRMT1- or PRMT4-regulated genes.

RESULTS

Arsenic induced transcription of the human ferritin H and L genes through the Nrf2-ARE activation

It has been demonstrated that arsenic generates oxidative stress (32) and induces several antioxidant detoxification genes, such as HO-1 and NQO1 (38). As ferritin encapsulates excess iron and protects cells from oxidative stress (8, 15, 39, 40), and the fact that the cellular response to arsenic in regard to ferritin expression remains largely uncharacterized, we tested whether arsenic induces ferritin expression. We used human HaCaT keratinocytes for the relevance of arsenic causing skin cancer. They were treated with 10 μM sodium arsenite for 2–24 h to measure ferritin H and L mRNA expression by qPCR, in which arsenic induced both ferritin H and L mRNA in a time-dependent manner (Fig. 1A). Increasing concentrations of arsenic induced both mRNA and protein levels of ferritin H and L (Supplemental Fig. S1), as well as NQO1 and HO-1 expression, as previously reported (38). The human ferritin H gene contains an ARE, a key oxidative stress responsive enhancer comprising 1 AP1-like and 1 AP1/NFE2 site located 4.5 kb upstream from the transcription start site (6). We therefore tested whether the ARE is responsible for the transcriptional activation of ferritin H following arsenic exposure. HaCaT cells were transiently transfected with luciferase reporters containing or lacking the ARE [4.5-kb ARE (+) or 4.4-kb ARE (−)] and treated with 0.5 or 1 μM arsenic for 24 h for luciferase assays. Arsenic induced luciferase expression driven by the 4.5-kb ARE (+) but not by the 4.4-kb ARE (−) (Fig. 1B), suggesting that arsenic induces transcription of the ferritin H gene through the ARE.

Figure 1.

Arsenic-induced transcription of the human ferritin genes through the Nrf2-ARE activation. A) HaCaT cells were treated with 10 μM sodium arsenite for 2, 4, 8, 12, and 24 h. Ferritin H (FH) or ferritin L (FL) mRNA expression was analyzed with real-time qPCR. Ferritin mRNA expression in nontreated cells was set to 1. B) HaCaT cells were transfected with luciferase reporter plasmids containing 4.5-kb ARE(+) or 4.4-kb ARE(−) 5′-promoter region of the human ferritin H gene and treated with sodium arsenite (0.5 or 1 μM) or t-BHQ (5 μM) for 24 h. pRL-null Renilla luciferase reporter was cotransfected and used as an internal transfection efficiency control. Firefly luciferase activity normalized with Renilla luciferase activity from cells with no treatment (0) was defined as 1. C) ChIP assays for Nrf2 binding to the ferritin H or L ARE were performed in HaCaT cells treated with 10 μM sodium arsenite for the indicated time. Level of Nrf2 bound to ferritin H ARE (solid bars) and ferritin L ARE (gray bars) with no treatment was defined as 1 (0). Open bars represent control IgG from the same ChIP assays. D) K562 cells were transfected with nontargeting (SiControl) or Nrf2-targeting (SiNrf2) siRNA and incubated for 48 h and then treated with 10 μM sodium arsenite for 24 h. Real-time qPCR was performed for Ferritin H or L mRNA expression. Expression level of ferritin H or L from cells transfected with SiControl and treated with arsenite was defined as 100%. Results are means ± se of duplicate samples from 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.0001 vs. no treatment (0) or SiControl; Student's t test.

As Nrf2 has been shown to be the major transcription factor that regulates the AREs of antioxidant genes, including the ferritin genes (37, 41), we tested whether Nrf2 mediates ferritin gene transcription after arsenic treatment. A time course was conducted wherein HaCaT cells were treated with 10 μM sodium arsenite, and recruitment of Nrf2 to the ferritin AREs was measured by ChIP assay. Nrf2 recruitment to the ferritin H and L AREs commenced at 30 min after exposure to arsenic, plateauing at 2 to 4 h (Fig. 1C). Subsequent to Nrf2 recruitment to the AREs, ferritin H and L mRNA were induced 4 h after arsenic treatment (Fig. 1A), suggesting that Nrf2 regulates arsenic-mediated ferritin gene transcription. Knockdown of Nrf2 attenuates the induction of ferritin H and L genes after arsenic treatment by 50% (Fig. 1D), demonstrating the importance of Nrf2 for arsenic-induced ferritin H and L expression.

Histone H4R3 and H3R17 methylation, along with nuclear accumulation of PRMT1 and PRMT4, is involved in arsenic-mediated ferritin transcription

To investigate the molecular mechanism by which arsenic regulates Nrf2 in the activation of ferritin gene transcription, we researched the roles of histone modifications and their upstream enzymes. Histone arginine methylation by protein arginine methyltransferases has been shown to play an important role in transcriptional regulation (27, 42). Among at least 9 PRMT members, PRMT1 and PRMT4 (CARM1) have been characterized as transcriptionally activating PRMTs through their role in catalyzing histone methylation on H4R3 and H3R17, respectively (28, 30, 43, 44). To test whether PRMT1, PRMT4, and methylation of H4R3, and H3R17 are involved in arsenic-mediated ferritin transcription, Western blots were performed, and demonstrated that arsenic treatment increased dimethylation of H4R3 and H3R17 in HaCaT cells (Fig. 2A). Furthermore, we observed that PRMT1 and PRMT4, localized in both the cytoplasm and nucleus in HaCaT cells, transiently accumulated in the nucleus 1 h after arsenic treatment (Fig. 2B). These results suggest that PRMT1 and PRMT4 are recruited to the nucleus to induce methylation of histone H4R3 and H3R17, thereby regulating the transcription of ferritin genes after exposure to arsenic. To confirm this observation, we pretreated HaCaT cells with 50 μM AdOx, a general inhibitor of methyltransferases, and treated them with arsenic to examine ferritin H mRNA expression. Indeed, AdOx inhibited the effect of arsenic on ferritin H mRNA expression and histone H4R3 and H3R17 methylation (Fig. 2C).

Figure 2.

Induction of histone arginine methylation and nuclear accumulation of PRMT1 and PRMT4 are involved in arsenic-mediated ferritin transcription. A) HaCaT cells were treated with 10 μM sodium arsenite for 0, 0.5, 1, 2, and 4 h, and extracted histones were analyzed for asymmetric dimethylation of histone H4R3 (H4R3me2a) and H3R17 (H3R17me2a) by Western blotting. Histone H4 and H3 Western blots are shown for loading control. B) HaCaT cells were treated with 10 μM sodium arsenite for 0, 0.25, 0.5, 1, 2, 4, and 6 h, and cytosolic and nuclear fractions were analyzed for PRMT1 and PRMT4 expression by Western blotting. Lamin B and LDH were used for nuclear and cytoplasmic markers, respectively, as well as for loading controls. C) Total RNA and protein lysates were isolated from HaCaT cells pretreated with a methyltransferase inhibitor, AdOx, for 1 h, followed by 10 μM sodium arsenite treatment for 14 h. RNA was subjected to Northern blotting (top 2 panels) and hybridized with ³²P-labeled human ferritin H probe. An equal amount of RNA loading was confirmed by staining RNA with ethidium bromide, and positions of 28S and 18S ribosomal RNA positions are indicated. Protein lysates were measured in Western blotting (bottom 4 panels) by anti-H4R3me2a and H3R17me2a; histone H4 and H3 were used as loading controls.

Methylation of histone H4R3 and H3R17 preceded or occurred concomitantly with Nrf2 recruitment to the ferritin AREs following arsenic treatment

We then asked whether methylated histone H4R3 and H3R17 are enriched in the proximity of the ferritin ARE, and if so, at what time relative to the recruitment of Nrf2 to the ARE after arsenic treatment. To address these issues, HaCaT cells were treated with 10 μM arsenic for 15–120 min, and ChIP assays for ferritin H and L AREs were carried out using anti-dimethyl H4R3 or anti-dimethyl H3R17 antibodies. The enrichment of dimethylated H4R3 at the ferritin H and L AREs was seen by 15 min after arsenic treatment, and the enrichment of dimethylated H3R17 was observed by 1 h only at the ferritin H ARE (Fig. 3). Non-ARE regions of the ferritin H and L genes showed no enrichment of dimethylated H4R3 or H3R17 (Fig. 3). As Nrf2 recruitment to the ferritin H and L AREs was induced by 0.5–2 h after arsenic treatment (Fig. 1C), these results suggest that there may be a causal association between Nrf2 activation and methylation of histone H4R3 and/or H3R17.

Figure 3.

Methylation of histone H4R3 and H3R17 precedes or occurs concomitantly with Nrf2 recruitment to the ferritin AREs following arsenic treatment. HaCaT cells treated with 10 μM sodium arsenite for 15–120 min were analyzed by primer sets flanking the region of the ferritin H ARE (FH ARE), ferritin L ARE (FL ARE), or non-AREs (FH non-ARE or FL non-ARE) in ChIP assays using rabbit IgG, dimethylated histone H4R3 (H4R3me2a), or H3R17 (H3R17me2a) antibody. Histone arginine methylation in untreated cells with each specific antibody was set to 1 (0). Dimethylated histone H4R3 or H3R17 in FH promoter regions are shown as solid bars, FL promoter as gray bars, and control IgG as open bars. Results are means ± se of duplicate samples from 3 independent experiments. *P < 0.05, **P < 0.01 vs. no treatment (0); Student's t test.

Nrf2 binding to the ferritin ARE, but not Nrf2 nuclear accumulation, was inhibited after PRMT1 or PRMT4 knockdown

To explore further, we transfected siRNA to knock down PRMT1 or PRMT4 in HaCaT cells, then treated with arsenic and measured the nuclear accumulation and recruitment of Nrf2 to the ferritin H and L AREs. Knocking down PRMT1 or PRMT4 showed only marginal effects on arsenic-induced nuclear accumulation of Nrf2 (Fig. 4A); however, ChIP assay showed inhibition of Nrf2 binding to the ferritin H and L AREs (Fig. 4B). We then measured expression of ferritin under PRMT1- or PRMT4-deficient condition. Indeed, induction of mRNA and protein expression of ferritin H and L by arsenic treatment was inhibited by either PRMT1 or PRMT4 knockdown in HaCaT (Fig. 5A) and K562 human chronic myelogenous leukemic cells (Fig. 5B) that express the BCR-ABL oncoprotein being a potential therapeutic target of arsenic trioxide, as well as human primary keratinocytes and fibroblasts (Supplemental Fig. S2). Collectively, these results suggest that PRMT1 and PRMT4 are important for Nrf2 binding to the ferritin H and L AREs in various cell types, which increases ferritin mRNA levels in cells exposed to arsenic.

Figure 4.

Knocking down PRMT1 or PRMT4 does not affect Nrf2 nuclear accumulation but inhibits Nrf2 binding to the ferritin AREs. Nontarget siRNA (siControl), siPRMT1, or siPRMT4 was transfected into HaCaT cells. A) Cells were treated with 10 μM sodium arsenite for 12 h. Nuclear fractions were analyzed by Western blotting with anti-Nrf2 or anti-lamin B; cytosolic fractions with anti-PRMT1, anti-PRMT4, or anti-β-actin antibody. B) Cells were treated with 10 μM sodium arsenite for 2 h, and Nrf2 binding to the ARE of ferritin H (FH ARE) or ferritin L (FL ARE) was analyzed by ChIP assay. Binding of Nrf2 to each ARE in arsenite-treated (solid and gray bars) or untreated cells (open bars) is shown as percentage compared with nontarget siRNA (siControl) treated with arsenic as 100%. Results are means ± se from 3 independent experiments. *P < 0.05, **P < 0.01 vs. SiControl treated with arsenic; Student's t test.

Figure 5.

Knocking down PRMT1 or PRMT4 inhibits arsenic-induced ferritin expression. Nontarget siRNA (siControl), siPRMT1, or siPRMT4 was transfected into HaCaT cells (A) or K562 cells (B), followed by treatment with 10 μM sodium arsenite for 12 h. Protein and mRNA expression of ferritin H (FH) and ferritin L (FL) were measured by Western blotting (top panels) or Northern blotting (middle panels). Knockdown of PRMT1 and PRMT4 were validated in Western blotting with anti-PRMT1 or anti-PRMT4 antibody. β-Actin and staining RNA with ethidium bromide were used for loading controls. Graphs show quantitative results of Northern blots from 3 independent experiments (bottom panels). Ferritin mRNA levels in cells transfected with siControl treated with 10 μM arsenite was defined as 100%. Solid (FH) and gray (FL) bars represent mRNA from cells treated with arsenic; open bars represent no treatment. *P < 0.05, **P < 0.01, ***P < 0.0001 vs. SiControl treated with arsenic; Student's t test.

PRMT1- and PRMT4-mediated ferritin gene transcription is involved in cytoprotection and cellular susceptibility to arsenic toxicity

Ferritin, as well as other antioxidant genes regulated by Nrf2, was demonstrated to be cytoprotective against oxidative stress. Arsenic-mediated oxidative stress induces apoptosis (33). To understand the role of PRMT1- and PRMT4-mediated ferritin transcription in the cellular defense against oxidative stress, we knocked down PRMT1, PRMT4, ferritin H, or ferritin L in HaCaT cells and assessed cellular susceptibility to arsenic toxicity by examining caspase 3 cleavage, a hallmark of apoptosis. Under significant knockdown of PRMT1 or PRMT4 in HaCaT cells (Fig. 6A), we observed increased cleavage of caspase 3 following arsenic treatment for 12 h (Fig. 6A). Similar results of enhanced cleavage of caspase 3 were also observed in knockdown of ferritin H or L in HaCaT cells, either with or without arsenic treatment (Fig. 6B). These results indicate that PRMT1 or PRMT4 deficiency sensitizes cells to arsenic-mediated apoptosis, suggesting that these histone methyltransferases play a cytoprotective role by up-regulating ferritin transcription.

Figure 6.

PRMT1- and PRMT4-mediated ferritin gene induction protects cells from arsenic toxicity HaCaT cells transfected with nontarget siRNA (siControl), siPRMT1, or siPRMT4 (A) or siControl, siFerritin H, or siFerritin L (B) were treated with 10 μM sodium arsenite for 12 h, and whole-cell lysates were isolated and analyzed by Western blotting with anti-caspase 3 antibody. Knockdown of PRMT1, PRMT4, ferritin H, and ferritin L were verified by Western blot with specific antibodies. β-Actin was used as a loading control.

Focused microarray expression analysis of oxidative stress-related genes following arsenic treatment and the effect of PRMT1 or PRMT4 knockdown

We further investigated broader gene expression changes induced by arsenic treatment and asked whether PRMT1 and/or PRMT4 are involved in the regulation of gene expression. For this purpose, we performed human oxidative stress signaling pathway PCR array analysis and characterized expression profiles of 84 oxidative stress-related genes in PRMT1- or PRMT4-knockdown HaCaT cells treated with 10 μM sodium arsenite for 12 h. cDNA samples synthesized from 4 groups of RNA samples isolated from HaCaT cells in siControl/no treatment, siControl/arsenic, siPRMT1/arsenic, and siPRMT4/arsenic were analyzed by qPCR. As shown in Fig. 7, arsenic treatment induced up-regulation of 23 genes and down-regulation of 5 genes out of 84 genes we analyzed. Among 23 up-regulated genes, 4 genes (FTH1: ferritin H, GCLC: glutamate-cysteine ligase, SPINK1: Serine peptidase inhibitor, Kazal type 1 and HSP90AA1: Heat shock protein 90 kDa alpha (cytosolic), class A member 1) were regulated by PRMT1 and/or PRMT4, showing significant gene expression changes after PRMT1 or PRMT4 knockdown (Table 1). Similarly, several genes with either no change or down-regulated by arsenic treatment were also subject to PRMT1 or PRMT4 regulation (Table 1). Furthermore, expression of many genes in this array (73 of 84 genes; Fig. 7, Table 1, and Supplemental Table S1) did not show significant expression changes in PRMT1 or PRMT4 knockdown. These results also confirmed that transcriptional activation of ferritin by arsenic through PRMT1 and PRMT4 is not due to the result of general histone methylation and subsequent chromatin decondensation effects.

Figure 7.

Gene expression profiling of oxidative stress-relevant genes after arsenic treatment. Nontarget siRNA (siControl), siPRMT1, or siPRMT4 was transfected into HaCaT cells and then treated with 10 μM sodium arsenite for 12 h. A) Volcano plot in human oxidative stress signaling pathway PCR array represents fold regulation of 84 genes associated with oxidative stress in HaCaT cells transfected with siControl and treated with 10 μM sodium arsenite compared to no treatment. The log₂ of fold regulation of arsenic-induced expression changes from 3 independent experiments is shown on the x axis, and the negative log₁₀-transformed P value on the y axis. Data points indicate increased gene expression with the positive numbers and decreased expression with negative numbers on the x axis. Vertical lines to the left and right of the midline (x=0) highlight a 1.5-fold change. Genes with up- or down-regulation >1.5-fold are shown by solid circles and squares, respectively; and genes with no significant changes are shown by open circles. Horizontal line in the y axis represents a value of P = 0.05. Heme oxygenase I is not shown in the plot but exhibited 300 fold up-regulation (P<10⁻⁶). Genes A to K correspond to genes listed in Table 1.

Table 1.

Genes up-regulated, down-regulated, or not significantly changed after arsenic treatment, along with the effect of PRMT1 or PRMT4 knockdown on mRNA expression

Arsenic effect	Gene	NCBI accession	Arsenic treatment
			siControl (fold)	siPRMT1 (%)	siPRMT4 (%)
Up-regulated
A	FTH1	NM_002032	1.7**	83**	79*
B	GCLC	NM_001498	2.4***	104***	81*
C	SPINK1	NM_003122	8.0***	89*	156***
D	HSP90AA1	NM_001017963	1.7**	90**	102
Down-regulated
E	KRT1	NM_006121	−3.1**	119	51***
F	EPHX2	NM_001979	−1.8**	112	82*
Not significantly changed
G	ALOX12	NM_000697	1.3	77**	94
H	PRDX3	NM_006793	−1.2*	89**	99
I	MB	NM_005368	−1.4	110	57*
J	TRAPPC6A	NM_024108	−1.2	99	83*
K	BNIP3	NM_004052	1.2	111	90*

Genes A to K correspond to genes plotted in Fig. 7. For the effect of PRMT1 or PRMT4 knockdown on mRNA expression, siControl treated with arsenic was defined as 100%. NCBI, U.S. National Center for Biotechnology Information; FTH1, ferritin heavy polypeptide 1; GCLC, glutamate-cysteine ligase catalytic subunit; SPINK1, serine peptidase inhibitor Kazal type 1; HSP90AA1, heat-shock protein 90 kDa α (cytosolic) class A member 1; KRT1, keratin 1; EPHX2, epoxide hydrolase 2 cytoplasmic; ALOX12, arachidonate 12-lipoxygenase; PRDX3, peroxiredoxin 3; MB, myoglobin; TRAPPC6A, trafficking protein particle complex 6A; BNIP3, BCL2/adenovirus E1B 19-kDa interacting protein 3.

^*P ≤ 0.05,

^**P ≤ 0.01,

^***P ≤ 0.001.

DISCUSSION

Arsenic is an environmentally ubiquitous human carcinogen that causes oxidative cellular damage. To understand the cellular defense program elicited by arsenic exposure, we characterized the molecular mechanism through which ferritin and other antioxidant genes are coordinately regulated by the Nrf2 transcription factor, along with histone modifications proximal to the ARE enhancer element of ferritin H and L genes. In this study, we found that arsenic is a potent inducer of the major intracellular iron-storage protein ferritin in various cell types. The induction of ferritin H by arsenic was regulated at the transcriptional level via activation of the ARE (Fig. 1), which we previously identified as both a basal and an oxidative stress-responsive enhancer of the mouse (3, 45) and human ferritin H gene (6). We also observed ferritin L induction by arsenic exposure (Fig. 1), which may be regulated in the same manner as ferritin H, because the ferritin L gene was also shown to be transcriptionally regulated through an ARE (12, 46). Iron plays an essential role in cell proliferation and transformation (47); therefore, up-regulation of ferritin in arsenic-treated cells appears to be an adaptive response to alleviate arsenic toxicity by chelating more intracellular free iron into ferritin, thereby decreasing the risk of iron-catalyzed hydroxyl radical formation, as well as cell proliferation and transformation. Indeed, knockdown of PRMT1or PRMT4 mimicked the effect of ferritin H or L deficiency, which increased susceptibility to arsenic toxicity (Fig. 6). However, it was shown that human osteogenic sarcoma, induced after long-term exposure of arsenic, exhibited higher total iron levels than the normal counterpart; in addition, ferritin expression in arsenic-transformed cells was not increased but decreased (48). This observation suggests that iron homeostasis appears to be adjusted during the long-term arsenic exposure in transformed cells through alteration of ferritin expression, as well as other genes involved in iron metabolism. We also observed that expression of HO-1 and NQO1 was induced by arsenic in HaCaT cells (Supplemental Fig. S1), consistent with our understanding of these well-characterized antioxidant genes, which are known to be induced by arsenic (38, 49, 50). Since arsenic was shown to activate an ARE (38), our finding characterizes ferritin as another new antioxidant gene transcriptionally induced by arsenic.

Arsenic is known to activate Nrf2 (38), the major transcription factor involved in ARE-regulated gene transcription. However, despite accumulating evidence demonstrating the importance of the chromatin environment in multiple stages of gene transcription, how Nrf2 and histone modifications proximal to an ARE regulate each other is a question that remains unanswered. Our results showing that AdOx, a broad histone methyltransferase inhibitor, blocked both arsenic-induced ferritin H expression and histone H4R3 and H3R17 methylation (Fig. 2) led us to investigate the role of PRMTs in ARE regulation by arsenic, in particular PRMT1 and PRMT4, which are transactivation-associated PRMTs and mediate histone H4R3 and H3R17 methylation, respectively (25). In our experiments, PRMT1 and PRMT4 proteins were localized in both the cytoplasm and nucleus in HaCaT cells, and PRMT1 and PRMT4 accumulated transiently in the nucleus following arsenic exposure (Fig. 2). To our knowledge, this is the first report showing increased levels of these PRMTs in the nucleus when cells were placed under a particular stress condition. However, it should be noted that pregnane X receptor was shown to interact with PRMT1, and their interaction determines PRMT1 subcellular compartmentalization (51). In addition, phosphorylation of PRMT4 was shown to cause PRMT4 cytoplasmic localization (52). Further investigation will be necessary to better understand the molecular mechanism by which nuclear accumulation of PRMT1 or PRMT4 is regulated in response to arsenic exposure.

In conjunction with nuclear accumulation of PRMT1 and PRMT4 following arsenic treatment, we observed increased methylation of H4R3 and H3R17 proximal to the AREs in the ferritin H and L genes (Fig. 3) but not in the non-ARE regions. We also analyzed expression and histone methylation status of the promoter regions of p21 and PTGES (prostaglandin E synthase) genes that were reported to be regulated through histone H4R3 and H3R17 methylation by PRMT1 and PRMT4 (53, 54), but our results showed that arsenic treatment induced no changes either in the status of histone H4R3 and H3R17 methylation of their promoter regions or in their gene expression after knocking down PRMT1 or PRMT4 (Supplemental Fig. S3), suggesting that transcriptional activation of ferritin by arsenic through PRMT1 and PRMT4 is not due to the result of general histone arginine methylation leading to chromatin relaxation effects. The methylation of histone H4R3 or H3R17 by PRMT1 or PRMT4 appears to enable Nrf2 to bind more efficiently to the ARE and activate transcription of the ferritin gene. This is supported by our results showing increased methylation of histone H4R3 and H3R17 preceding or occurring concomitantly with Nrf2 recruitment to the ARE following arsenic treatment (Fig. 1 and Fig. 3 time-course results); more notably, knocking down PRMT1 or PRMT4 with siRNA did not affect arsenic-induced Nrf2 nuclear accumulation, but did block the binding of Nrf2 to the ferritin H and L AREs (Fig. 4). In contrast, we did not obtain evidence for the contribution of PRMT1 or PRMT4 to the activation of the HO-1 ARE following arsenic treatment (unpublished results). Consistently, our results of human oxidative stress PCR array demonstrated that expression of HO-1 mRNA was induced by arsenic treatment, but PRMT1 or PRMT4 knockdown did not affect HO-1 expression (Fig. 7, Table 1, and Supplemental Table S1). This focused microarray approach identified another ARE-regulated gene, GCLC, as a PRMT4-regulated gene after arsenic treatment (Fig. 7B). Although the ARE sequences are highly conserved, and its regulatory mechanism appears to be largely shared (12), chromatin environment of each ARE may differ, and as a result, it may significantly affect the transcriptional process. Indeed, there are several cases representing a particular regulatory mechanism applied only to specific ARE sequences (21, 55–57).

Our results suggest that PRMT1- and PRMT4-mediated histone methylation is likely to be involved in facilitating Nrf2 binding to the ARE. However, another possibility cannot be ruled out, which is the PRMT1- or PRMT4-mediated protein Arg methylation of nonhistone proteins. It was demonstrated that PRMT1 methylates and activates signal transducers and activators of transcription 1 (STAT1) transcriptional activity by altering its interaction with the inhibitor protein PIAS1 (58). PRMT1 also methylates FOXO1 within an Akt phosphorylation motif, resulting in inhibition of FOXO1 phosphorylation by Akt, thereby blocking FOXO1 degradation (59). In addition, activity of the transcriptional coactivator peroxisome proliferator-activated receptor γ coactivator 1α (PGC1α) was increased by multiple Arg methylation in the C-terminal region of PGC1α by PRMT1 (60). In the context of ARE regulation, several transcription factors are involved, including Nrf2 and other b-zip transcription factors, such as small Maf proteins (MafF, MafK, MafG), Bach1, and AP1 family members (2). In addition, several transcriptional coactivators were shown to be involved in the regulation of ARE enhancer activity, including p300/CBP (15–17), MOZ (18), BRG1 (55), and p160RAC3/SRC3 (61, 62). There is accumulating evidence for cooperative functions of PRMT-mediated histone arginine methylation and acetylation of histones by p300/CBP (63, 64), as well as the p160 SRC family that interacts with PRMT4 and enhances transcription of target genes (30, 65). Therefore, in regard to the possibility of the involvement of nonhistone protein methylation, these ARE-interacting transcription factors and/or coactivators might also be regulated by PRMT1 or PRMT4 in response to arsenic exposure.

Arsenic is metabolized by oxidative methylation along with glutathione conjugation (66). The methylation step catalyzed by arsenic methyltransferases utilizes S-adenosylmethionine (SAM) as a methyl donor, resulting in decreased SAM pools. Thus, arsenic causes global hypomethylation in genomic DNA (67); however, it is also known that arsenic induces gene-specific DNA hypermethylation by unknown mechanisms that are thought to contribute to gene expression changes (68). We believe that the latter situation may be similar to arsenic-induced histone Arg methylation, in which cells exposed to arsenic recruit PRMT1 and PRMT4 to the nucleus and increase histone H4R3 and H3R17 methylation even under conditions of decreased SAM pools; this appears to be a cellular antioxidant response to arsenic toxicity by allowing Nrf2 to up-regulate specific ARE-regulated antioxidant genes, such as ferritin and GCLC.

In summary, this work demonstrated that ferritin is a new arsenic-responsive gene, transcriptionally regulated through activation of the ARE by Nrf2; arsenic induces histone H4R3 and H3R17 methylation, along with nuclear accumulation of PRMT1 and PRMT4; methylation of histone H4R3 and H3R17 precedes or accompanies Nrf2 recruitment to the ARE following arsenic treatment; PRMT1 or PRMT4 deficiency inhibits Nrf2 recruitment to the ferritin H and L AREs and their transcriptional activation by arsenic treatment; and PRMT1 or PRMT4 knockdown increased cellular susceptibility to arsenic-mediated apoptosis. Collectively, these results suggest that PRMT1 and/or PRMT4 is important for arsenic-mediated ARE activation and the antioxidant cellular defense program.