Thyroid Hormone–mediated Histone Modification Protects Cortical Neurons From the Toxic Effects of Hypoxic Injury

Abstract

Context

Thyroid hormone has been shown to have a protective role in neuronal injury, although the mechanisms have not been established. The cellular response to stress that promotes adaptation and survival has been shown to involve epigenetic modifications.

Objective

We hypothesized that the neuroprotective role of thyroid hormone was associated with epigenetic modifications of histone proteins. We used hypoxic neurons as a model system for hypoxia-induced brain injury.

Methods

Mouse primary cortical neurons were exposed to 0.2% oxygen for 7 hours, with or without, treatment with triiodothyronine (T3). We analyzed the expression of histone-modifying enzymes by RNA-seq and the post-translationally modified histone 3 proteins by enzyme-linked immunosorbent assay (ELISA) and Western blot.

Results

We found that methylation of H3K27, associated with inactive promoters, was highly induced in hypoxic neurons, and this histone methylation was reduced by T3 treatment. H3K4 methylation is the hallmark of active promoters. The expression of 3 (Set1db, Kmta2c, and Kmt2e) out of 6 H3K4 methyltransferases was downregulated by hypoxia and expression was restored by T3 treatment. H3K4me3 protein, measured by ELISA, was increased 76% in T3-treated hypoxic neurons compared with the levels without T3 treatment. H3K56ac plays a critical role in transcription initiation and was markedly increased in T3-treated hypoxic neurons compared with those without T3 treatment, indicating stimulation of gene transcription. Additionally, T3 treatment restored hypoxia-induced downregulation of histone acetyltransferase, Kat6a, Kat6b, and Crebbp, which function as transcription factors.

Conclusion

These findings indicate that T3 treatment mitigates hypoxia-induced histone modifications and protects neurons from hypoxia-induced injury.

Histone 3 (H3) modification is a key epigenetic mechanism that regulates chromatin accessibility of transcription factors and subsequent gene transcription, and allows adaptation of the cellular phenotype in response to a range of internal and external signals [1]. H3 modification is important in intrinsic cellular programs, such as development and memory, as well as responding to extrinsic factors, such as diet, injury, and disease [2]. H3 is subject to various modifications, including methylation, demethylation, acetylation, and deacetylation, as well as phosphorylation. These modifications are catalyzed by histone methyltransferase/demethylase, histone acetyltransferase (HAT)/deacetylase (HDAC) and protein kinases/phosphatases. Several enzymes are involved in modifying each lysine in the H3 tail, which may be specific to the tissue, cell type, and developmental stage. For example, in self-renewing embryonic stem (ES) cells, H3K4 is methylated by Set1a (set domain-containing 1a), and in differentiating ES cell it is methylated by Kmt2 (lysine methyltransferase 2, or MLL2 (mixed lineage leukemia 2)) [3].

H3 methylation is associated with both gene activation and repression, depending on which lysine site in the H3 tail is modified. Methylation of H3K4, H3K36, and H3K79 is frequently found in euchromatin and linked to gene transcription. Methylation of H3K9 and H3K27 is the hallmark of heterochromatin and linked to gene silencing [4]. Each lysine residue may be mono-, di-, or trimethylated, producing different chromatin structures and variable impact on gene transcription. A functional study of H3K9 methylation showed that mono- and dimethylated H3K9 is specifically localized to silent domains of chromatin, and trimethylated H3K9 is concentrated at pericentric heterochromatin [5]. H3K4me3 is concentrated in transcription start sites (TSSs), H3K4me2 is located downstream of the promoter, and H3K4me1 is in the distal region of the promoter. The frequency of H3K4 mono-, di-, and trimethylation is not evenly distributed in an active gene. High-frequency transcription is correlated with a high level of H3K4me3 in the TSS [6]. Additionally, high levels of H3K4me3 are correlated with better transcription fidelity and increased elongation rates [7].

H3 acetylation is an epigenetic hallmark of gene activation. Acetylation of histone lysine increases negative charge to the histone tail and reduces chromatin condensation, opening the nucleosome to allow transcription factors to bind to DNA and promote transcription [8, 9]. In contrast, histone deacetylation tightens the nucleosome binding on DNA and condenses chromatin, resulting in transcriptional repression. Acetylated H3K56 (H3K56ac) forms a complex with the bromodomain of CREB-binding protein (CBP) [10], which demonstrates that H3K56ac directly regulates CBP/p300-mediated transcription.

The expression of histone-modifying enzymes is tightly regulated and an imbalance of enzyme expression results in aberrant epigenetic modifications, which have been linked to neurological diseases. In Alzheimer's disease, elevated HDAC2 blocks acetylation of genes important for learning and memory [11]. Mutations in lysine methyltransferase (Kmt) 2d decreases trimethylation of H3K4 (H3K4me3) and are seen in Kabuki syndrome 1, a congenital multisystem disorder [12, 13]. Ezh2 (enhancer of Zeste 2 polycomb repressive complex 2 subunit) specifically methylates H3K27. Mutations of Ezh2 result in reduced di- and trimethylation of H3K27 and are associated with Weaver syndrome, characterized by pre- and postnatal rapid growth and intellectual disability [12, 14, 15].

Environmental changes, such as hypoxia, are a regulator of histone modification [16, 17]. Histone lysine demethylases, Kdm4a, Kdm6a, Kdm5a, and muscle-specific histone methyltransferase Smyd 1 have been identified as oxygen sensors in cells [18–24]. Kdm4a specifically targets H3K9me3 and is responsible for activation of HIF1α transcription in hypoxia [22]. Hypoxia decrease Kdm6a and leads to hypermethylation of H3K27me3 and repression of gene transcription [21].

Thyroid hormone receptor is known to play a significant role in histone modification. Ligand binding to thyroid hormone receptor results in recruitment of HAT (P300, Proto-oncogene tyrosine-protein kinase Sarcoma (SRC) family member), resulting in opening of the chromatin and promotion of transcription [25]. The type 2 deiodinase (Dio2)–mediated local T3 production transiently modifies histone and promotes chromatin accessibility for gene transcription during liver development [26]. Regulation of TSHα gene by triiodothyronine (T3) involves T3-mediated histone modification of H3K9me3, H3K18ac, and H3K4me3 [27].

Thyroid hormone is known to have a role in neuroprotection [28–34]. In vitro studies in primary mouse cortical neuronal cells showed that T3 modulate DNA methylation by increasing 5-hydroxymethylcytosine (5hmc), a product of the first step in DNA demethylation and reduces DNA methyltransferase 3α (Dnmt3α), as well as increasing DNA demethylase Tet1 (ten-eleven translocation 1) and Tet2 [28]. Single cell genomic studies showed that thyroid hormone treatment post-traumatic brain injury (TBI) in animals mitigates brain injury–associated genomic and behavioral abnormalities [34]. We hypothesize that the neural protective role of thyroid hormone involves epigenetic modulation of histones. In the present study, we used cultured primary cortical neurons exposed to hypoxia (0.2% oxygen), with or without T3 treatment, to evaluate the changes in expression of histone-modifying enzymes and the resulting histone modifications. We found that T3 reduced hypoxia-induced hypermethylation of H3K9 and H3K27. In addition, T3 markedly induced H3K4me3 and H3K56ac in neurons exposed to hypoxia. Our data demonstrate that T3 treatment of hypoxic neurons promotes histone 3 modifications that enhance gene transcription, maintain cellular function, and protect from hypoxia-induced neuronal injury.

Materials and Methods

Primary Mouse Cortical Neuronal Culture

Cortical neurons (Thermofisher Scientific) were isolated from C57BL/6 embryonic day 17 fetuses. The preparation is free of other cell types (>98% are cortical neurons) and viability is in the range of 50% to 80%, according to the manufacturer’s description. The neurons were plated on a culture dish and chamber slides, as previously described [28]. In brief, the dish and chamber slides were precoated with poly-D-lysine at 4.5 mg/m². Cells were plated at a density of 0.3 million cells per 10-cm dish or to a 4-chamber slide at 5 × 10⁴ cells per chamber. Cells were grown in Neurobasal® medium containing 200 mM GluMAX™-1 and B27 (2% v/v) supplemented with 50% medium and changed every 3 days. B27 is a proprietary serum-free supplement that includes in its list of components the hormones corticosterone, progesterone, and T3. T3-induced and T3-repressed gene expression were seen at the expected magnitude, with and without addition of T3, indicating that any T3 supplement in the media was negligible and did not significantly influence measurement of gene expression. Cells were used for experiments after 6 days of growth.

Hypoxia Exposure and Thyroid Hormone Treatment

The method was described previously [28]. In brief, cortical neurons were placed in a Hypoxia Chamber (HypOxygen station H45, HypOxygen Inc.) with a gas mixture composed of 0.2% oxygen, 5% CO₂, and 94.8% N₂. Medium was pre-equilibrated in a 0.2% oxygen environment, and medium changes were performed in the hypoxic chamber to avoid fluctuations in the oxygen level. In treated cells, T3 was added to the medium to a final concentration of 10 nM. The period of hypoxia exposure was 7 hours prior to the collection of cells.

Histone Isolation and Western Blot

Histone proteins were prepared from cortical cells (10⁶ cells) using EpiQuik total histone extraction kit (Epigentek Inc). Several extractions were combined in order to obtain a sufficient amount of protein. The isolated histone protein (3 μg) was loaded in a 12% sodium dodecyl sulfate gel and separated by electrophoresis. The protein was transferred to PVDF membrane and blotted with antibodies. Each gel was loaded with an identical amount of protein for all antibody detection studies. The antibodies used were antitotal histone H3, anti-H3K9me3 (Epigentek Cat# A-4036, RRID:AB_2920606), anti-H3K27me3 (Epigentek Cat# A-4039, RRID:AB_2722776), anti-H3K79me3 (Epigentek Cat# A-4045, RRID:AB_2861209), anti-H3K4me3 (Epigentek Cat# A-4033, RRID:AB_2920607), anti-H3K9ac (Epigentek Cat# A-4054, RRID:AB_2920609), anti-H3K14ac (Epigentek Cat# A-4023, RRID:AB_2891336), anti-H3K56ac, and antitotal H3 (Epigentek Cat# A-4026, RRID:AB_2636966) at a dilution of 1:1000. Western blots were imaged and quantified using the Li-Cor Studio Lite. Each modified histone protein was presented as pixel density % relative to total H3 pixel density.

Enzyme-linked Immunosorbent Assay Analysis of Histone Modifications

The histone modification was analyzed using a multiplex enzyme-linked immunosorbent assay (ELISA) kit from EpigenTek Inc. (Catalog # P-3100-96, RRID:AB_2920611) following the manufacturer's instructions. In brief, the total histone was isolated and used in the ELISA with 4 replicates. In each assay well, 100 ng of histone extract was used. After antibody binding and washing, the colorimetric signals were detected using a Microplate reader (Promega Inc.) at wavelength 450 nm. The data are presented as the reading at OD 450 nm.

Immunofluorescence Staining of Trimethylated Histone 3 in Cortical Neurons and Imaging

Cortical neurons plated in chamber slides were described previously [28]. In brief, cells were plated at a density of 1 × 10⁵ per chamber slide that was precoated with Poly-D-Lysine. Cell were maintained in chamber slides for 3 days. The culture with 50% medium was changed prior to exposure to hypoxia in the HypOxygen system. Hypoxia exposure was for 7 hours. Cells were fixed with 4% paraformaldehyde and followed regular processes for immunofluorescence (IF) staining. Cells were incubated with primary antibodies overnight and then with Alexa fluor 488 (green fluorescence) for MAP2 conjugation and Alexa fluor 594 (red fluorescence) for trimethylated histones for 1 hour before being mounted with Prolong Gold containing DAPI. The primary antibodies used were anti-MAP2 (Santa Cruz Biotechnology Cat# sc-20172, RRID:AB_2250101), anti-H3K9me3, anti-H3K4me3, and anti-H3K56ac. The cells were imaged with a Zeiss LSM 770 confocal microscope. The pixel intensity of red fluorescence for histone trimethylation was quantified by ImageJ. Two consecutive frames were quantified and shown with mean value ± SD.

RNA-seq

Each group (normoxia, hypoxia, and hypoxia + T3) contains 3 replicates. After 7 hours of hypoxic exposure (0.2% oxygen), cortical neurons were collected and RNA was isolated using RNeasy Kit (Qiagen Inc). RNA integrity and quality was determined by Qubit and the TapeStation system. The RNA library was constructed and sequenced at a depth of 1 × 50 using the Hiseq 3000 system. The RNAseq data was analyzed by UCLA Genomic Bioinformatics Core, utilizing Partek Flow edgeR software. Trimmed mean of the M-values was used to normalize data before sample data comparison, part of the EdgeR algorithm for the statistical comparison of differential gene expression. The P value was generated by edgeR Partek Flow. RNA-seq data were submitted to the GEO depository, accession #GSE205801.

Statistical Analysis

One-way analysis of variance (ANOVA) was used for all comparisons, except RNA-seq data analysis, which used the edge R package.

Results

Effects of T3 on the Expression of Histone-modifying Enzymes in Hypoxic Neurons

Brain oxygen levels under normal physiological conditions range from 0.55% in the midbrain to 8% in the pia mater and depend on the brain region assessed [35]. We used previously determined hypoxic conditions, 0.2% O₂, that induced neuronal damage in primary cortical neurons [28]. The axons of the cortical neuron were visibly shortened and fragmented after 7 hours of hypoxic exposure (Fig. 1A). This indicates that cellular stress and injury were induced by low oxygen tension, although cell death was not observed. T3 treatment during hypoxic exposure preserved neurons and resulted in maintenance of axon integrity and appeared similar to the control. The observed differences in cell morphology in hypoxic neurons, with and without T3 treatment, were associated with epigenetic changes, including DNA and histone modifications. We found that hypoxia and T3 differentially affected the expression of histone-modifying enzymes in cortical neurons, as determined by RNA-seq (Fig. 1B-1D). The expression of histone lysine demethylase and methyltransferase (known as methylase) family genes in T3-treated hypoxic cells was significantly changed compared with hypoxia alone. In particular, T3 treatment increased demethylase Kdm6b 0.67 log₂ fold (or 1.59-fold, P < .01) and decreased Kdm6a 0.6 log₂ fold (or 1.51-fold, P < .05), as well as increased methylase Kmt2a 0.35 log₂ fold (or 1.27-fold, P < 5 × ^–4) and decreased nuclear receptor binding SET domain protein 1 (Nsd1) 0.4 log₂ fold (or 1.32-fold, P < .04), compared with hypoxia exposure without T3 treatment.

Figure 1.

Hypoxia induces a change in histone modification enzyme landscape in mouse cortical neurons partially reversed by T3 treatment. Cortical neurons were exposed to hypoxia (0.2% oxygen) for 7 hours with/without T3 (10 nM) added to the culture medium and compared with control cortical neurons without hypoxic exposure. Cells were collected and RNA was isolated for RNA-seq analysis. (A) Cells were stained with anti-MAP2 and Alexa fluor 488. The nuclei were stained with DAPI. Cells were imaged with confocal microscopy. Bar = 50 μm. (B, C) Heatmaps of the mRNA expression of histone-modifying enzymes (expression level CPM ≥ 2). (D) Log₂FC – log₁₀P volcano plot of all known histone-modifying enzymes in hypoxic neurons treated with or without T3. Red dots depict P < .05 and black color indicate no significant differences comparing T3 treatment vs no treatment in hypoxic neurons. (E) Volcano plot of all gene expression (CPM ≥ 2). F, mRNA level of the known hypoxia-induced genes. (G) Genes required for T3 action in neurons. (H) mRNA level of T3-target genes. *P < .05 hypoxia compared with hypoxia/ + T3.

The gene expression pattern of HAT showed that T3 treatment increases HAT gene expression, including Crebbp (0.25 log₂ fold or 1.2-fold, P < .003), Ep300 (0.35 log₂ fold or 1.27-fold, P < .02), Kat2a (0.46 log₂ fold or 1.28-fold, P < .004), and Kat6b (0.4 log₂ fold or 1.32-fold, P < .002) and HAT component Ncoa1 Ncoa2, compared with hypoxic neurons without T3 treatment (Fig. 1C). In contrast to the increased expression of HAT mRNA, Hdac9 decreased 0.7 log₂ fold (or 1.62-fold, P < 1.3 × ^–8), Hdac11 0.43 log₂ fold (or 1.35-fold, P < .001), and Hdac4 1.15 log₂ fold (or 2.22-fold, P < .001). The core component of HDAC, Ncor1, was increased 0.7 log2 fold (or 1.62-fold, P < .01) and Ncor2 0.5 log2 fold (or 1.41-fold, P < .02). The volcano plot indicated that the changes of known H3-modifying enzymes in gene expression were within approximately ±1 log₂ fold (or 2-fold), except hairless (Hr), which was induced 2.8 log₂ fold (or 6.96 fold, P < 2.8 × ^–9) (Fig. 1D). The changes in histone modification enzymes were associated with a global change in gene expression (Fig. 1E). The expression of a total of 2690 genes was significantly changed, with a threshold P <.05, of which 1253 genes were upregulated and 1437 genes were downregulated in T3-treated cells, compared with cells without treatment (GSE205801). Known hypoxia-regulated genes, including hypoxia inducible factor 1a (Hif1α), HIF1α protein regulators, Egln2 (also known as PHD1), and Vhl (Von Hippel–Lindau) and Hmox1 (heme oxygenase 1), a sensor of hypoxia-induced oxidative stress, had no significant change in mRNA levels with T3 treatment compared with hypoxia alone (Fig. 1F). Hif2a, a T3-regulated gene, was increased 1.6 log₂ fold (or 3.03-fold, P < .04). Egln1 (PHD2), a part of Hif2a complex, was increased 0.36 log₂ fold (or 1.28-fold, P < .04). The significant downregulation of Nos1 (neuronal nitric oxide synthesis 1) suggested that both NO synthesis and NO cytotoxicity were lower in T3-treated hypoxic neurons. Glucose transporter 1 (Glut-1/Scl2a1) and 3 (Glut3/Slc2a3) are the 2 primary glucose transporters in the brain. The expression of Glut1 and Glut3 in hypoxic neurons was stimulated 0.5 log₂ fold (or 1.41-fold, P < 1.4 × ^–5) and 0.3 log₂ fold (or 1.23-fold, P < 4 × ^–3), respectively, by T3 (Fig. 1F). T3 is known to regulate glucose transport in the heart, and brown and white adipose tissues [36–38]. T3 stimulation of Glut1 and Glut3 may enhance glucose uptake in response to hypoxia-induced metabolic change and support neuronal cell adaptation to hypoxia.

We examined the genes that mediate T3 action and metabolism and found that thyroid hormone receptor beta (Thrb) and 5′ iodothyronine deiodinase 2 (Dio2) mRNA were both downregulated 1 log₂ fold (P < .001) with T3 treatment compared with hypoxia alone (Fig. 1G). In neurons, T3 action is primarily mediated by the thyroid hormone receptor alpha (Thra) receptor, whose expression was not affected by T3 treatment. Expression of Slc16a2 (also known as MCT8), the principle thyroid hormone membrane transporter, was not affected, and the other 2 transporters, Slc7a5 (also known as LAT1) and Slc16a1 (also known as MCT1), were stimulated 0.2 log₂FC (fold-change) (or 1.15-fold, P < .03) and 0.38 log₂FC (1.3-fold, P < 2 × ^–4), respectively, by T3 compared with hypoxia alone. We analyzed known T3 target genes (Fig. 1H). T3-induced genes, such as Hairless (Hr), was stimulated 2.8 log₂ fold (or 6.96-fold, P < 2.8 × ^–9) and Krüppel like factor 9 (Klf9) 0.87 log₂ fold (or 1.83-fold, P < 5.7 × ^–4). T3-repressed genes, Kelch-like protein 14 Klhl14 and sialytransferase St8sia IV (St8sia4) [39], were downregulated 0.6 log₂ fold (or 1.52-fold, P < .01) and 0.5 log₂ fold (or 1.41-fold, P < 7.1 × ^–5) respectively. Interestingly, early growth response1 (Egr1), which transcriptionally regulates Hif1α transcription in hypoxia [40], was stimulated 1.7-fold (or 3.25-fold, P < 1.1 × ^–4) by T3. EGR1 protein recruits TET1 to remove methylation marks on its own gene for transcription [41]. We previously reported that T3 stimulates TET1 mRNA expression in hypoxic neurons [28], suggesting that T3 may stimulate Egr1 expression through stimulation of TET1 expression.

T3 Alleviated Hypoxia-induced H3K27 and H3K9 Methylation in Hypoxic Neurons

Modification of lysines in the H3 tail influences chromatin condensation and transcriptional activity. We compared the mRNA level of the methylase and demethylase enzymes under control condition (normoxia), hypoxia, and hypoxia/+ T3 in primary cortical neurons. H3K9 and H3K27 methylation are found in the promoter of inactive genes. There are 2 methylases that are specific for methylation of H3K27, Enhancer of Zeste homolog 1(Ezh1) and Ezh2, which catalyze mono-, di-, and trimethylation of H3K27. In hypoxic neurons, Ezh1 and Ezh2 mRNA were significantly induced, 0.38 log₂ fold (or 1.3-fold, P < .05) and 0.6 log₂ fold (or 1.52-fold, P < .02), respectively, compared with normoxia (control, log₂FC = 0). T3 treatment restored Ezh1 to the control level of expression and partially restored Ezh2 expression (Fig. 2A). There are 4 H3K27 demethylases: Kdm6a and Kdm6b catalyze the demethylation of H3K27me2/me3 and Kdm7a and Phf8 the demethylation of H3K27me2. Hypoxia with T3 treatment downregulated Kdm6a mRNA expression but upregulated Kdm6b mRNA. A similar pattern of expression was observed for Kdm7a and Phf8 mRNA. The outcome of differential mRNA expression of the methylase and demethylase enzymes was determined by evaluation of the H3K27 protein methylation state. Hypoxia induced H3K27me1 1.89-fold (2.27-fold, P < .001), H3K27me2 1.4-fold (2.64-fold, P < .01), and H3K27me3 1.7-fold (3.25-fold, P < 1.0 × ^–4) compared with control (Fig. 2B). T3 treatment reduced hypoxia-induced H3K27me1 by 21% (P < .001), H3K27me2 10% (P < .07), and H3k27me3 17% (P < .001). Western blot data showed that the H3K27me3 protein was increased in hypoxic neurons and reduced to near control level with T3 treatment (Fig. 2C). The results indicate that T3 treatment of hypoxic neurons protected against hypoxia-induced H3K27 methylation.

Figure 2.

T3 differentially modulates H3K27 and H3K9 methylation. Cell culture conditions are the same as described in the Fig. 1 legend. Cells were exposed to 7 hours of hypoxia, with or without T3 treatment, and were then lysed and total histone isolated for ELISA of mono-, di-, and trimethylated proteins and for Western blot (WB) detection of trimethylated proteins. (A, D) mRNA level of histone-modifying enzymes specific for H3K27 and H3K9. Dot color code: blue, monomethylation, magenta, dimethylation, and black, trimethylation [1]. (B, E) Mono-(me1), di-(me-2) and tri-methylated (me-3) H3K27 and H3K9 protein content. C and F, Western blot detection of H3K27me3 and H3K9me3 proteins. The protein quantification was carried out using Li-Cor Studio Lite. The data was normalized to total H3 protein density in the Western Blot (WB) and expressed as % total H3. (G) Immunofluorescent (IF) detection of H3K9me3 protein in cortical neurons with representative merged images shown. (H) Quantification of H3K9me3 fluorescence intensity. H3K9me3 antibody was conjugated with Alexa Fluor 594 (red). Two consecutive images in the red channel were quantified with ImageJ. To quantify H3K9me3 (red fluorescence), the red channel was isolated from the color-merged image and then the red fluorescence intensity was quantified in each cell nuclei. A total of 3 images were quantified and the fluorescence intensity of all H3K9me3 positive nuclei from 3 slides is shown. Anti-H3K9me3 (red) was diluted at 1:200 (Epigentek Inc) and anti-AMP2 (green) at 1:500 dilution (Abcam Inc). DAPI (blue) for nuclear staining was from Prolong Gold seal (Thermofisher Sci Inc). Scale bar = 100 μm. One-way ANOVA was used for statistical analysis. *P < .05 compared with control; **P < .05, hypoxia/ + T3 compared with hypoxia.

There are 4 known H3K9 methylases expressed in mouse cortical neurons. Prdm2 (PR/Set domain 2) and Prdm4 catalyze mono-methylation of H3K9me1, and their expression was not affected by hypoxia or T3 treatment. Di- and trimethylases, Suv39h1 (Su(var)3-9 homolog1) and Setd1b (Set domain containing 1B) were significantly reduced in hypoxic neurons compared with control. T3 treatment further reduced Setd1b mRNA 70% (P < .005) compared with hypoxia alone (Fig. 2D). Kdm3a and Phf8 catalyze the removal of a methyl group from H3K9me1/me2. Hypoxia and T3 noticeably upregulated Kdm3a and downregulated Phf8. Direct evaluation of the protein indicated that H3K9 methylation states were similarly induced across H3K9me1/me2/me3 by hypoxia. T3 treatment decreased H3K9me1 by 15% (P < .02), H3K9me2 15% (P < .06) and H3K9me3 10% (P < .06) compared with hypoxia alone (Fig. 2E). Western blot confirmed H3K9me3 was induced by hypoxia 26% (P < .03) and T3 treatment 23% (P < .05), compared with control (Fig. 2F). Hypoxia-induced H3K9me3 was observed in IF-stained cortical neurons with anti-H3K9me3 antibody, which showed greater intensity of H3K9me3 in hypoxic neurons compared with control (Fig. 2G and 2H). These data indicate that hypoxia was the primary factor promoting the increase in H3K9me3. T3 treatment significantly reduced the protein level of H3K9me1 and had no effect on H3K9me2/3. H3K9 methylation is the hallmark of heterochromatin and increases in H3K9me2/me3 indicated that more genes were silent in hypoxic neurons. H3K9me1, in contrast to H3K9me2/me3, is enriched at TSSs of active genes in the human genome [42]. Interestingly, although the mRNA levels of H3K9 methylases, Suv39h1, and Setd1b, were markedly downregulated and the other 2 (Prdm2 and Prdm4) had no change in expression, H3K9me1/me2/me3 protein levels were significantly elevated, suggesting there were unidentified methylases in cortical neurons which methylated H3K9.

T3 Treatment Promotes H3K4 Methylation in Hypoxic Neurons

Histone H3K4 methylation is well known for its role in activating transcription. Methylated H3K4 is highly enriched in active promotor region and at TSSs [43]. There are at least 10 methylases expressed in humans that ensure sufficient H3K4 methylation. We found that 7 H3K4 methylases were expressed in mouse cortical neurons. Set domain containing (Setd)1a, setd1b, Kmt2a (also known as myeloid/lymphoid or mixed-linage leukemia 1 [MLL1]) and Kmt2d (MLL2) were capable of mono-, di-, and trimethylation of H3K4. Kdm2c (MLL3) kdm2b (MLL4), and Kdm2e (MLL5) catalyze mono- and dimethylation. Hypoxia significantly reduced gene expression of Setd1b (P < 1.0 × ^–5), Kmt2c (P < .01), and Kmet2e (P < .01) compared with control. T3 treatment completely reversed hypoxia-induced downregulation of these genes (Fig. 3A). On the other hand, T3 significantly increased Kmt2a (P < .008). H3K4 demethylase Kdm5a was significantly downregulated by both T3 (P < .01) and hypoxia (P < .03). Other H3K4 demethylases did not respond to hypoxia or hypoxia with T3 treatment. H3K4 methylation is the most important histone mark for active gene transcription. The vast majority, 90%, of H3K4me3 are present at RNA polymerase II binding regions [44]. All 3 methylation states of H3K4 were associated with transcription enhancers at promoters [42]. ELISA of H3K4me1/me2/me3 in cortical neurons indicated differential effects of hypoxia on H3K4 methylation. Hypoxia, but not T3, was a dominant inducer of H3K4me1 and stimulated H3K4me1 by 49% (P < .005) compared with control. On the other hand, hypoxia reduced H3K4me2 protein by 20% (P < .01) and T3 treatment completely reversed the effect of hypoxia. H3K4me3 protein was reduced 10% (P < .05) by hypoxia, but was dramatically increased 66% (P < .001) in T3-treated hypoxic neurons (Fig. 3B). In addition, IF staining of H3K4me3 protein in cortical neurons showed distinct high intensity of H3K4me3 staining (Fig. 3C and 3D). Hypoxia with T3 treatment reversed the observed increase in H3K4me3 from hypoxia alone and was confirmed by Western blot. H3K4me3 was reduced 22% in hypoxic neurons and increased 27% in T3-treated hypoxic neurons compared with control (Fig. 3E). The data demonstrate that T3 stimulated gene transcription in hypoxic neurons, as seen Fig. 1, is associated with a T3-induced increase in the level of H3K4me3.

Figure 3.

T3 treatment promotes H3K4 trimethylation. (A) mRNA expression of modifying enzymes specific to H3K4. Dot color code: blue, monomethylation, magenta, dimethylation, and black, trimethylation. (B) ELISA of H3K4me1, me2, and me3 protein level in control (normoxia) cells, hypoxic cells, and hypoxic cells treated with T3. (C) Immunofluorescent detection H3K4me3 in cortical neurons with anti-H3K4me3 (red) at dilution 1:200, anti-AMP2 (green) at dilution 500 and DAPI (blue for nuclei) and imaged using confocal microscope. Antibody source were the same as described in Fig. 2. Scale bar = 100 μm. (D) Quantification of H3K4me3 fluorescence intensity using imageJ as described in Fig. 2. (E) Western blot detection of H4K4me3. Total histone protein (15 μg) was utilized in each 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The protein was transferred to a PVDF membrane and blotted with anti-H3K4me3 or anti-H3 antibody. Both antibodies were diluted 1:500 (Epigentek Inc). Western blot detected protein were quantified using Licor Studio Lite. The data were normalized to total H3 protein density in the Western blot and expressed as percentage of total H3. One-way ANOVA was used for statistical analysis. *P < .05 compared with control; **P < .05; hypoxia/ + T3 was compared with hypoxia.

Hypoxia Reduces H3K36me3

In the human genome, H3K36me3 is enriched in transcribed regions of active genes [42, 45]. In stress conditions it is important to maintain methylase Setd2 levels to preserve H3K36me3, since H3K36me3 directly and quickly recruits DNA repair machinery to the chromatin [46]. There are 8 methylases for H3K36 methylation in mammals. In mouse cortical neurons, 5 H3K36 methylases were detected (Fig. 4A). Setd2 is the only enzyme in mammals that catalyzes the trimethylation of H3K36, and the other enzymes catalyze either mono- and/or dimethylation of H3K36 [47]. Surprisingly, neither hypoxia nor T3 had an impact on Setd2 expression (Fig. 4A). There are 5 demethylases for H3K36me1/me2/me3. The mRNA level of these genes was mainly affected by hypoxia. T3 did not have additional significant effects compared with hypoxia alone (Fig. 4A). The protein level of H3K36me1/me2 was not significantly changed. H3K36me3, however, was reduced 22% (P < .03) in hypoxic neurons and 13% (P < .001) in T3-treated hypoxic neurons (Fig. 4B). Western blot analysis showed that H3K36me3 protein was reduced 27% in hypoxic neurons and 21% in hypoxic neurons treated with T3 compared with control (Fig. 4C). H3K36 has an important role to promote gene transcription and reduction in H3K36me3 may promote the downregulation of gene expression in hypoxia.

Figure 4.

T3 and hypoxia differentially influence H3K79 and H3K36 modifications. Cortical neuronal cell treatment conditions were the same as described in Fig. 2. (A, D) mRNA expression of histone modification enzymes specific for H3K36 and H3K79. (B, E) ELISA of methylated H3K36 and H3K79 protein at each methylation state. (C, F) Western blot detection of H3K36me3 and H3K79me3. The antibodies used were anti-H3K36me3, H3K79me, and anti-H3. All antibodies were used at a dilution 1:500 (Epigentek Inc). Detected proteins were quantified using Li-Core Studio Lite. H3K36me3 and H3K79me3 proteins were normalized to total H3 protein and expressed as percentage of total H3 protein. One-way ANOVA was used for statistical analysis. *P < .05 compared with control; **P < .05; hypoxia/ + T3 was compared with hypoxia.

H3K79me3 Was Induced by Hypoxia and T3

H3K79 methylation, which belongs to the group of H3K4 and H3K36, is a mark of active transcription. Dotl1 (disruptor of telomeric silencing-1, also known as Kmt4 in human) is the only methylase identified for H3K79 methylation and catalyzes mono-, di-, or trimethylation of H3K79. Deletion of Dotl1 increases the spread of heterochromatin [9, 48, 49]. T3 treatment stimulated Dotl1 0.54 log₂ fold (or 1.45-fold, P < .01), whereas hypoxia alone did not have a significant effect on Dotl1 expression (Fig. 4D). The only known demethylase for H3K79 is Kdm2b [50], whose mRNA level was reduced 0.33 log₂ fold (or 1.26-fold, P < .05) by hypoxia and 0.4 log₂ fold (or 1.32-fold, P < .05) by T3 treatment compared with control. With increasing Dotl1 and decreasing Kdm2b, methylated H3K79 in all states (me1/me2/me3) was significantly greater in hypoxic neurons compared with control cells (Fig. 4E). However, with T3 treatment, H3K79me1 and H3K79me3 were further increased 27% (P < .05) and 35% (P < .03), respectively, compared with the levels seen with hypoxia alone (Fig. 4E). Moreover, the Western blot showed that H3K79me3 protein was increased 1.68-fold (P < .005) in hypoxic neurons and 2.9-fold (P < .001) in T3-treated hypoxic neurons (Fig. 4F), a 73% increase. Methylation states of H3K4, H3K36 and H3K79 belong to the same group of histone marks that were positively correlated with gene transcription. In this group, H3K79 was the only one whose methylation state was positively modulated by both hypoxia and T3, suggesting that methylation of H3K79 may play an important role in stress-induced gene expression.

Effects of Hypoxia and T3 on HDACene Expression

In humans and mice, there are 18 HDACs, including 11 from the HDAC family and 7 from the sirtuin (Sirt) HDAC family. Gene expression of the majority of Hdacs was not affected by hypoxia and T3. However, Hdac9 and Hdac4, which belong to class II HDAC, were downregulated. Hdac9 was downregulated 1.0 log₂ fold (or 2-fold, P < .01) in hypoxic neurons and 1.9 log₂ fold (or 3.73-fold, P < .01) in T3-treated hypoxic neurons (Fig. 5A). In the brain, Hdac9 is exclusively expressed in postmitotic and mature neurons [51]. Silencing Hdac9 benefits neuronal survival by suppressing oxygen-glucose deprivation–induced neuronal apoptosis via downregulation of NeuroD1 [52]. Indeed, we found that downregulation of Hdac9 was accompanied by reduction of NeuroD1 mRNA, 0.63 log₂FC (or 1.55-fold, P < .003) in hypoxic neurons, and 1.3 log₂FC (2.46-fold, P < .003) in T3-treated hypoxic neurons compared with control (Fig. 5B), consistent with our previously reported reduction of apoptosis marker genes in T3-treated hypoxic neurons [28].

Figure 5.

T3 treatment inhibits HDAC, partially restores HAT, and robustly stimulates H3K56 acetylation. Cell treatment conditions were the same as described in Fig. 2. (A-D) mRNA level by RNA-seq analysis of (A) HDAC family, (B) Neuro D1, (C) Sirt family, and (D) HAT family. (E) ELISA of protein level of H3K9ac, H3K14ac, H3K18ac, and H3K56ac using whole-cell histone extract. (F) Western blot detection of H3K9ac, H3K14ac, H3K56ac, H3K18ac, and total histone 3 (H3), using total histone extracts. (G) Quantification of protein shown in Western blot (F) using Li-Cor Studio lite. The quantified pixel density is expressed as a percentage of the total H3 protein. The antibodies used in Western blot were at 1:1000 dilution (Epigentek Inc.). (H) Immunofluorescent detection of H3K56ac protein in cortical neurons. Anti-H3K56ac (red) was diluted at 1:200 dilution and anti-MAP2 (green) at 1:500 dilution. DAPI (blue) was for staining nucleus. Antibody source as is described in “Materials and Methods.” (I) Quantification of H3K56ac fluorescence (red) intensity using ImageJ. The anti-H3K56ac stained image was used in the quantification and the remainder of the method is the same as described in Fig. 2. Scale bar = 100 μm. One-way ANOVA was used for statistical analysis. *P < .05 compared with control; **P < .05; hypoxia/ + T3 was compared with hypoxia.

HDAC4 is highly expressed in mouse brain. Its role, however, remains controversial, as shown in Hdac4-selective knockout under different promoters (CamkII gene promoter and Thy1 or Nestin gene promoter) [53–55]. Regardless, the primary role of Hdac4 remains as a transcription repressor. In hypoxia, HDAC4 mRNA was downregulated 2.2 log2 fold (or 4.6-fold, P < 0.01) in cortical neurons, and was restored 50% by T3 treatment. HDAC11, the sole member of class IV HDAC and located in both nucleus and cytoplasm, was highly expressed in neurons and oligodendrocytes in mouse brain [56]. HDAC 11 mRNA was elevated (0.5 log₂ fold or 1.41-fold) in hypoxic neurons, and was restored to control level by T3 treatment (Fig. 5A). The Sirtuin family (Sirt1 to Sirt7) are Class III HDAC and have a wide range of biological functions. In cortical neurons, Sirt1 was stimulated 0.72 log₂ fold (or 1.65-fold, P < 0.03) by T3. Other Sirt family members were not affected by hypoxia or T3 (Fig. 5C). Previous studies showed a concomitant increase of Hif2a protein and Sirt1 mRNA in hypoxic hepatic cells, due to Hif2a bound to Sirt1 promoter and augmented Sirt1 gene transcription [57]. In our study, we observed concomitant increase of Hif2a and Sirt1 in cortical neurons. Sirt1 has been shown to reduce neuroinflammation by deacetylation of p65 subunit of NFKB complex and antagonize NFKB signaling [58].

T3 Restored Hypoxia-associated Downregulation of HAT MRNA Levels in Hypoxic Neurons

HATs are transcription factors and respond to cellular cues and hormonal changes in modulation of gene transcription. Well-characterized HATs expressed in cortical neurons include Kat family members Kat2a (known as GCn5), Kat5, Kat6a, Kat6b, Kat7, and Kat8 and Crebbp (CBP) and EP300 (known as P300). The mRNA levels of Kat5, Kat7, Kat8, and EP300 were not significantly changed in hypoxic neurons, with or without T3 treatment (Fig. 5D). Kat2b, known as p300/CBP-associated factor (PCAF), was detectable at very low levels in cortical neurons. The Kat2a, a paralog of Kat2b, was stimulated 1.3 log₂ fold by T3 compared with hypoxia alone. In hypoxic neurons, Kat6a, Kat6b, and Crebbp were significantly downregulated, 0.49 log₂ fold (or 1.4-fold, P < .003), 0.83 log₂FC (or 1.78-fold, P < .001), and 0.4 log₂ fold (or 1.32-fold, P < .01), respectively. T3 treatment completely restored Kat6a and Crebbp and partially restored Kat6b to the control levels (Fig. 5D), which contributed to genes in active transcription in T3-treated hypoxic neurons.

Hyperacetylation of H3K9ac, H3K18ac, and H356ac

T3 treatment significantly increased HAT mRNA expression, which may contribute to higher HAT activity. We analyzed histone acetylation of H3K9ac, H3K14ac, H3K18ac, and H3K56ac in cortical neurons, with or without treatment, by ELISA and Western blot, and compared with control. H3K9ac was the most abundant acetylated histone among those tested (Fig. 5E-5G). ELISA indicated that hypoxia alone induced H3K9ac 1.4-fold (or 2.64-fold, P < .001) and T3 treatment 1.7-fold (3.25-fold, P < .001) compared with control. Western blots confirmed the ELISA results of H3K9ac. Neither hypoxia nor T3 influenced the H3K14ac level. H3K18ac, as measured by ELISA, was stimulated by hypoxia, 2.3-fold, and 3-fold by T3. Interestingly, acetylation of H3K56ac was significantly reduced 24% (P < .05) in hypoxic neurons. In contrast, H3K56ac was dramatically increased 201% (P < .01) by T3 treatment compared with hypoxia, and increased 153% (P < .01) compared with control (Fig. 5E-5G). H3K56ac had increased intensity of IF staining in cortical neurons compared with control and hypoxia alone, which was consistent with ELISA and Western blot analysis (Fig. 5H and 5I). Since the principle role of H3K56 is through association with transcription cofactors, the global hyperacetylation of H3K56ac by T3 may increase chromatin accessibility, mediating the T3-protective effects.

Discussion

We have shown that thyroid hormone treatment of hypoxic neurons altered the mRNA expression of histone modification enzymes and histone modifications. T3 treatment had no effect on hypoxia-induced H3K9 di- and trimethylation and the least effect on H3K36 mono- and dimethylation. T3 treatment markedly increased H3K4me3 and H3K79me3, and alleviated hypoxia-induced H3K27 me1/me2/me3. HDAC and HAT act as transcription factors in chromatin remodeling and modulating transcription. Eighteen HDACs have been identified in mammals and can be detected in mouse primary cortical neurons. Four HDACs (Hda4, Hdac9 Hdac11 and Sit1) responded to hypoxia and T3 treatment. T3 treatment of hypoxic neurons restored the hypoxia-downregulated HATs (Kat6a, Kat6b, and Crebbp) and markedly increased acetylation of H3K9ac and H3K56ac (Fig. 6). These T3-responsive epigenetic marks (HDACs and HATs) are important for mediation of the neuroprotective effects of thyroid hormone. We, and others, previously reported that a single dose of thyroid hormone injection post-TBI reduced brain edema in acute TBI mouse/rat models [29]. In the hypoxic neural cell culture model, T3 treatment preserved neurofilament length, reduced apoptosis, and induced a high level of expression of genes (Junb, Vegfa, tyrosine hydroxylase) associated with hypoxic adaptation at the early stage of hypoxia exposure [28, 29, 59–61]. In addition, T3 treatment of hypoxic neurons induced 5hmc and reduced Dnmt3a, which are implicated in reduction of hypoxia-induced DNA methylation [28]. These T3-responsive epigenetic markers have a broad range of biological functions, although the direct role in neurons needs to be elucidated.

Figure 6.

Summary of the effects of hypoxia and T3 treatment on histone 3 post-translational modifications. Mouse primary cortical neurons were cultured and exposed to hypoxia (0.2% oxygen), with or without T3 (10 nM) in the culture medium. Cells were collected after 7 hours of exposure and total histone was isolated. The histone 3 modification was analyzed using ELISA (Epigentek Inc). Each histone assays contained quadruplicates. The data show are the mean value. PTM, post-translational modification.

HDACs and HTAs regulate acetylation/deacetylation of histone or nonhistone transcription factors. A few HDAC4 substrates have been identified, including HIF1α. The inability of HDAC4 to deacetylate HIF1α reduces its stability [62]. In muscle, PGC-1a, myosin heavy chain (MyHC) and heat shock 70 (Shc70) are HDAC4 substrates. Blocking HDAC4 activity by inhibitors or gene knockout, increases acetylation of these proteins and rescues muscle from denervation-induced muscle atrophy [63]. In this study, we showed that Hdac4 mRNA was largely decreased in hypoxic neurons and partially rescued by T3 treatment. We hypothesize that potential protein substrates of HDAC4 may contribute to the thyroid hormone neuroprotective function. The substrates of HATs, such as CBP/P300, are both histone and transcription factors. CBP/P300 acetylates H3 (K56, K9, and K14) and H4 (K5, K8, K12, and K16) for chromatin remolding, which facilitates transcription [64–70]. The crystal structure of CBP showed that the bromodomain recognizes H3K56 with high affinity [10]. H3K56ac increases in S phase and serves as a marker for new nucleosomes in DNA duplication. In human ES cells, H3K56ac is recruited by Oct4 to the target gene promoters, which are essential to maintain ES cell pluripotency [71]. Studies also show that acetylation of H3K56acis is associated with DNA repair in various human cell types [72–75]. In neural cells, H3K56ac function is less well characterized. Global induction of hyperacetylation of H3K56ac by T3 may mediate some of the T3 protective effects. Further study is needed in this area. CBP/P300 is the only known factor to acetylate H3K56ac. Although the mRNA level of CBP/P300 was not induced by T3 in hypoxic neurons, their acetylation activity, which was not analyzed in this study, may be increased by other mechanisms. Additionally, it is conceivable that there are other HATs capable of acetylation of H3K56ac.

CBP/P300 is a dual acetylase, which acetylates histone and transcription factors. In T3-dependent gene expression, CBP/P300 is recruited by SRC-1 family members to the promoters, and acetylates THRA at lysine K128, K132, and K134 in the DNA binding domain and enhances DNA binding [76]. In addition, CBP/P300 remodels chromatin by acetylation of H3 and H4 lysine targets for making the euchromatin state, which may explain why T3-regulated gene expression was not more influenced by hypoxia. In fact, most nuclear receptors are acetylated [77]. Further investigation of the CBP/P300 substrates may identify the beneficial factors that promote cell survival under hypoxia.

H3K4me3 marks active gene promoters. Genome-wide analysis of H3K4 methylation identified H3K4me3 as associated with the 5′ region of virtually all transcriptionally active genes [78]. In this study, we found that T3 treatment highly promoted trimethylation of H3K4, which led to T3-induced gene transcription. H3 K4 methylation is tightly regulated by specific histone-modifying enzymes during brain development. Dysregulation of H3K4 methylation causes intellectual disability and neuropsychiatric disorder in patients [79–81]. Modulation of H3K4me3 level by methylases may lead to memory change, since H3K4me3 modulates the transcription and accuracy of memory formation–related genes [80, 82]. There are 6 methylases for H3K4 in cortical neurons and we found that 3 of the methylases (Setd1b, Kmt2c, and Kmt2e) were downregulated by hypoxia and restored to normal levels by T3 treatment, which may be especially important targets for T3 in recovery after brain injury.

Our study design does not distinguish the potential influence of hypoxia-activated pathways on conferring T3 responsiveness to genes that are not normally responsive in the absence of hypoxia. Our study did identify some T3-responsive genes previously known to be induced in neurons in the absence of hypoxia, such as Hairless and the glucose transporter and others, such as the chromatin modifiers, which were not previously reported. The other challenge is modeling these actions in cell culture, where the standard oxygen concentration of 22% is far in excess of the oxygen concentrations in human tissues, such as lung (4-14%), liver (5.4%), and kidney (9.5%) [83]. The brain is naturally low in oxygen tension, in rat models ranging from 0.5% in the midbrain to 5.33% in the cortex (gray and white matter) [84]. It is likely that the T3-responsive genes we reported included those that are the result of factors induced by hypoxia and those that are T3 responsive independent of hypoxia, but this will require further study to clarify.

The in vitro cortical neuronal cell model used is appropriate for these mechanistic studies we report, and the T3 effects are consistent with findings from in vivo models of the role of T3 in brain development and maintaining normal function in the adult brain. Hypoxia, however, is one of several factors that contribute to the effects of brain injury and recovery of function. Further evaluation of T3-induced modifications of histones will need to be evaluated in an in vivo rodent TBI model. These identified mechanistic targets will allow for a more sensitive evaluation of the optimal form of thyroid hormone or analog, dose, and treatment timing when promoting brain recovery after TBI.

Abbreviations

ANOVA

analysis of variance

CBP

CREB-binding protein

ELISA

enzyme-linked immunosorbent assay

ES

embryonic stem

H3

histone 3

H3K56ac

acetylated H3K56

HAT

histone acetyltransferase

HDAC

deacetylase

MLL

mixed-linage leukemia

T3

triiodothyronine

TBI

traumatic brain injury

TSS

transcription start site

Conflict of Interest

All of authors confirm that there is no conflict of interest in relation to this work.

Data Availability

Data generated and analyzed in this study is included in this published article and in NCBI GEO repository, accession #GSE205801.