Inhibition of HDAC3 prevents diabetic cardiomyopathy in OVE26 mice via epigenetic regulation of DUSP5-ERK1/2 pathway

Diabetes mellitus is a global health problem, as evidenced by the facts that the number of global diabetic patients has been up to 415 million in 2015 and that from 2012 to 2015, diabetes and its complications have resulted in approximately 1.5–5.0 million deaths each year [1]. In terms of cardiac complications, diabetes leads to cardiac vascular injury, the formation of atherosclerotic plaque, and even myocardial infarction. Among these complications, a myocardial disorder without coronary artery disease and hypertension, named as diabetic cardiomyopathy (DCM), has received much attention in recent years [2]. DCM has been noticed in both human and animal models with Type 1 or Type 2 diabetes [3]. In diabetic mouse models that were generated by either streptozotocin (STZ) or transgenic approach, mice displayed decreased cardiac diastolic and/or systolic dysfunction and a variety of morphological changes, including cardiac hypertrophy and interstitial fibrosis [4–6]. Some mechanisms have been proposed for contributing to the pathogenesis of DCM [7–9], including the alterations in ion homeostasis, especially calcium transients, mitochondrial dysfunction, accumulation of advanced glycated end-products, accumulation of toxic lipids, trace metal dyshomeostasis, and hyperlipidemia and inflammatory cytokines. In the clinic, patients who were randomized initially to the more intensified insulin regimen and then returned to regular glycemic control continued to have better prevention of various microvascular and macrovascular events comparing with those who were only regular glycemic control. This phenomenon was coined as ‘metabolic memory’ [10,11] and recently was attributed to epigenetic modifications [11,12].

Histone deacetylases (HDACs), a class of epigenetic modification enzymes, remove acetyl groups (O=C–CH₃) from a ε-N-acetyl lysine amino acid on a histone, making the DNA chain to twine the histones more tightly, and resulting in inhibition of gene expression. The 18 HDACs are encoded by distinct genes and are divided into four classes. Class I, II, and IV HDACs are zinc-dependent enzymes, whereas class III HDACs, which are also known as sirtuins, require nicotinamide adenine dinucleotide (NAD⁺) for catalytic activity [13]. Recent studies have shown that activation of HDACs was associated with DCM since total HDAC inhibition attenuated diabetes-induced cardiac injury in the STZ-induced diabetic animal models [14,15]. In these two studies, it is unclear which specific HDAC really involves in the pathogenesis of DCM because they only used non-specific HDAC inhibitor to inhibit total HDAC activity.

Several previous studies reveal that extracellular signal-regulated kinases (ERK) 1/2 are involved in the development of cardiac hypertrophy and interstitial fibrosis [16,17]. For instance, Ferguson et al. [18] found that class I HDAC inhibition repressed cardiac hypertrophy induced by several pathological insults in vitro via inhibition of ERK1/2 by dual specificity phosphatase 5 (DUSP5). Furthermore, HDAC3 was reportedly involved in different aspects of heart pathologic change, including heart failure [19], hypertrophy [20], and abnormal energy metabolism [21], but so far no information is available for its role in DCM.

In the present study, therefore, we tested our hypothesis that specific inhibition of HDAC3 with HDAC3 selective inhibitor, RGFP966, may prevent the development of DCM with OVE26 mice that spontaneously develop Type 1 diabetes. We also explored whether its cardioprotection from diabetes is mediated by stimulating DUSP5 to inactive ERK1/2 signaling pathway. In addition, considering the existence of metabolic memory due to epigenetic modification for the development of diabetic cardiovascular complications, we treated diabetic mice for the first 3 months and then observed the treatment effect immediately and 3 months later after the end of 3-month treatment to explore the existence of ‘cardiac protective memory’.

Male OVE26 and wild-type (WT) FVB mice were housed at the University of Louisville Research Resources Center with a 12-h light/dark cycle at 22°C. All animal experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Louisville, and were performed complying with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996).

The specific HDAC3 inhibitor RGFP966 (10 mg/kg, Abcam, Cambridge, MA, U.S.A.) and pan-HDAC inhibitor valproic acid (VPA, 200 mg/kg, Santa Cruz, Dallas, TX, U.S.A.) were subcutaneously injected into the mice every other day for 3 months [22]. The specific HDAC inhibitor RGFP966 was dissolved in DMSO and diluted in 30% hydroxypropyl β-cyclodextrin, and the final DMSO concentration was no more than 1%. VPA was dissolved in phosphate buffer solution. In the present study, VPA was included as a total HDAC inhibition control. Cardiac protection from DCM will be compared between HDAC total inhibition of VPA and HDAC3-specific inhibition with RGP966 in order to define the contribution of HDAC3 to HDAC total inhibition-mediated cardiac protection from DCM.

Animals were divided into six experimental groups: WT + vehicle, WT + RGFP966, WT + VPA, OVE + vehicle, OVE + RGFP966, and OVE + VPA. At the end of the 3-month treatment, five to seven mice from each group were killed as the time point of 3 months (labeled as 3 M) cohort. The remaining animals (n=5 at least) in each group were kept for another 3 months without HDAC inhibitors’ treatment (labeled as 6 M). Mice were intraperitoneally anesthetized with Avertin (tribromoethanol, 350 mg/kg). The chest cavity was opened, and heart was removed for weighing, and then derived into three pieces. One piece was for protein and RNA analysis, one for frozen section, and the rest was fixed in 10% neutral buffered formalin for histological analysis.

Echocardiography was performed as we described previously [23]. In brief, heart rate (HR), parasternal long-axis and short-axis views in mice anesthetized with isoflurane were recorded by transthoracic echocardiography (Vevo 770, Visual Sonics, Canada) equipped with a high-frequency ultrasound probe (RMV-707B). Left ventricular (LV) dimensions and wall thicknesses were assessed on the basis of parasternal short-axis M-mode images. Ejection fraction (EF), fractional shortening (FS), LV mass, and LV end volume were instantaneously computed by Vevo770 software (Table 1).

Mice were anesthetized and the hearts were collected, fixed in 10% formalin, dehydrated in graded alcohol series, cleared with xylene, embedded in paraffin, and sectioned at 5 μm thickness. FITC-conjugated WGA staining (Alexa Fluor-488, Invitrogen, Carlsbad, CA, U.S.A.) was performed to evaluate the cardiomyocyte size under the confocal microscope. Sirius Red staining was performed to detect collagen deposition in cardiac tissues [23].

To assess the production of reactive oxygen species (ROS), frozen heart tissue segments were cut into 8-μm sections in a cryostat and placed on a microslide. Sections were washed in phosphate buffer saline (PBS) for three times (5 min/time) at 37°C. After incubated for 1 h at 37°C with 10 μM dihydroethidium (DHE) (Thermo Fisher, Grand Island, NY, U.S.A.), fluorescence was detected using a 585-nm filter with fluorescence microscopy (XI 71 Olympus, Tokyo, Japan). Fluorescence intensity was analyzed by the use of ImageJ software.

HDAC and HDAC3 activities were determined using HDAC Activity Fluorometric Assay Kit (K330-100 and K343-100, BioVision, CA, U.S.A.) based on the protocol provided by the manufacturer. Forty micrograms of protein lysate was diluted to 85 µl (final volume) of ddH₂O in each well along with a well of ddH₂O only as the background. HDAC Assay Buffer (10 µl of 10×) and 5 µl of the HDAC fluorometric substrate were added to each well with thorough mixing. Then, the plates were incubated at 37°C for 1 h before adding 10 µl of Lysine Developer to stop the reaction. The plate was incubated at 37°C for another 30 min. All the reacted samples were analyzed in the fluorescence plate reader at E_x/E_m = 350/450 nm. The HDAC activity can be expressed as the relative fluorescence units per µg protein. For the measurement of HDAC3 activity, first, the standard curve was prepared by following the protocol in HDAC3 Activity Fluorometric Assay Kit. In details, 10 μM AFC (7-amino-4-trifluoromethyl coumarin) at 0, 1, 2, 4, 6, 8, and 10 μl was added to individual wells in a 96-well plate and adjust the volume to 100 μl/well with HDAC3 assay buffer to generate 0, 10, 20, 40, 60, 80, and 100 pmol/well of AFC standard. Read fluorometrically at E_m/E_x = 380/500 nm after mixing. Then, for sample preparations, samples were prepared as replicates of up to 25 μl/well with HDAC3 Assay Buffer in a 96-well plate. One sample replicate added 2 μl of HDAC3 Inhibitor (Trichostatin A), was used as background control. After mix well and incubating for 10 min at 37°C, 25 μl of substrate solution (23 μl of HDAC3 assay buffer and 2 μl of HDAC3 substrate) was added into each test sample and background control wells except standard surve wells and mix well, then incubated at 37°C for 30 min. After then, 10 μl of Developer and 40 μl HDAC3 assay buffer were added to each well and incubated another 5 min at 37°C. The plates were read in the fluorescence plate reader at E_x/E_m = 380/500 nm for each background well (RB) and sample well (RS) respectively. The RFU of fluorescence generated is ∆ RFU = RS – RB. Apply the ∆ RFU to the standard curve to get B pmol of AFC. At last, the final HDAC3 activity was calculated by the following formula:

HDAC3 activity = (B × sample dilution factor)/(30 × V) = pmol/min/ml = U/ml

B is the AFC amount from the standard curve (in pmol); 30 is the sample/substrate incubation time (in min); V is the sample volume added into the reaction well (in ml). One unit is defined as the amount of HDAC3 able to generate 1.0 pmol of AFC per minute at 37°C when incubated with the HDAC3 Substrate (R-H-K-K(Ac)-AFC).

The lipid peroxidation accumulation in the hearts was evaluated by the content of malondialdehyde (MDA) as described previously [24].

As described in our previous study [23], nuclear extracts were purified with the nuclear extraction kit (Abcam, Cambridge, MA, U.S.A.).

Western blot was carried out as we detailed previously [23]. Primary antibodies used for this assay include atrial natriuretic peptide (ANP), connective tissue growth factor (CTGF), GAPDH, Lamin B1, DUSP5, Histone H3 and β-actin (Santa Cruz Biotechnology, Dallas, TX, U.S.A.), 3-nitrotyrosine (3-NT), acetyl-Histone H3 (Millipore, Billerica, CA, U.S.A.), 4-hydroxy-2-nonenal (4-HNE, Alpha Diagnostic International, San Antonio, TX, U.S.A.), phosphorylated c-Jun N-terminal protein kinase (p-JNK), tumor necrosis factor-α (TNF-α), glucose transporter type 1 and 4 (GLUT1 and GLUT4) (Abcam, Cambridge, MA, U.S.A.), p-ERK1/2, ERK1/2, JNK, p-p38, p38, insulin receptor substrate 1 (IRS1), p-Akt, Akt, 5’ AMP-activated protein kinase α (AMPKα), p-AMPKα, HDAC3 and acetyl-Histone H3 (Lys9/Lys14) (Cell Signaling Technology, Danvers, MA, U.S.A.), and plasminogen activator inhibitor-1 (PAI-1, BD Bioscience, San Jose, CA, U.S.A.). The protein blot on the nitrocellulose membrane was captured by the ChemiDoc Touch Imaging System (Bio-Rad, Hercules, CA, U.S.A.) and was normalized to that of GAPDH or β-actin.

TRIzol Reagent (Invitrogen, Carlsbad, CA, U.S.A.) was used to purify the total RNA from heart tissues. RT-qPCR was performed as previously described [23]. Primers [ANP: Mm01256744_m1; collagen I: Mm01302043_g1; fibronectin-1 (FN-1): Mm01255747_g1; DUSP5: Mm01266106; GAPDH: Mm99999915_g1] for PCR were obtained from Thermo Fisher (Grand Island, NY, U.S.A.).

As described in our previous study [25], EpiQuik™ Tissue ChIP Kit (P-2012; Epigentek Group Inc., Farmingdale, NY, U.S.A.) was used to perform the chromatin immunoprecipitation (ChIP) assay according to the manufacturer’s protocol. In brief, the acetylated histone H3 antibody (1 μg, Epigentek Group Inc.) or 1 μl of normal mouse IgG (as a negative control) was used to precoat the assay wells. Meanwhile, 30 mg of heart tissue was cut into little pieces and cross-linked with 1% formaldehyde. The cross-link was stopped by glycine solution (1.25 M). After tissue disaggregation and the nuclei isolation, the DNA was sheared by sonication (S-450-Dwithmicro-tip probe; Emerson Industrial, St. Louis, MO, U.S.A.) with five pulses of 20 s each separated by 40 s rest on ice (output control: 2). After centrifugation, 5 μl of the diluted supernatants were used as input DNA. The other diluted supernatant (100 μl) was added to the acetylated histone H3 antibody-coated wells followed by incubation at room temperature for 60 min. ChIP-enriched DNA fragments were precipitated, purified, and assayed by quantitative PCR (qPCR) with the following primers: DUSP5 forward 5′-CTGACACTCCACCGGTAGTC-3′, reverse 5′-GGGCGTCTTAGAGCGAGAAA-3′. The value was computed relative to a calibrator (2^−ΔΔC_t) and normalized to the input.

Data were expressed as means ± S.D. (n=5–7 per group). One-way ANOVA followed by post-hoc pairwise repetitive comparisons with Turkey test was used for data comparison among groups. Statistical analysis was performed with Origin 8.0 Lab data analysis and graphing software. P values less than 0.05 were considered to be statistically significant.

We first examined the general features of WT and OVE26 mice treated with vehicle, VPA, or RGFP966. Before the treatment, the baseline cardiac function measured by echocardiography showed no significant difference between OVE26 and WT mice at 2 months of age (Figure 1A). The peripheral blood glucose concentration was measured every 4 weeks with collecting blood from the tail tip vein. Quintessentially, mice with blood glucose readings of >250 mg/dl were diagnosed diabetic. In this project, the blood glucose of diabetic mice was more than 400 mg/dl, indicating the presence of diabetes. Three months treatment with RGFP966 or VPA did not significantly affect the blood glucose levels in both OVE26 and WT mice (Figure 1B). Next, the activity of HDAC3 and HDAC was measured, showing that in OVE26 mice activity of both HDAC3 and HDAC was significantly increased compared with WT mice at the 3 M time point, which persisted to the 6 M time point. Treatment with either RGFP966 or VPA for 3 months significantly reduced the diabetes up-regulated HDAC3 and HDAC activities, even at 6 M time point (Figure 1C and D). In addition, the acetylated histone H3, especially at Lysine 9/14 (Figure 1E–G), was increased in RGFP966 and VPA-treated group at two time points, which confirmed that the increased acetylated histone H3 induced by inhibiting HDAC could persist 3 months after cessation of HDAC inhibitor therapy although the increased levels were reduced at 6 M than those at 3 M.

To define whether HDAC3 inhibition by RGFP966 protected against diabetes-induced cardiac dysfunction in OVE26 mice, we examined cardiac function with echocardiography and found that OVE26 mice had significantly increased LVID; d, LVID; s, LV vol; d, LV vol; s, LV mass, and decreased EF and FS compared with WT mice (Table 1). Three-month treatment with RGFP966 improved these cardiac functional parameters in diabetic mice, which lasted to the 6 M time point. There was no significant difference between RGFP966 and VPA treatment groups in terms of cardiac function improvement.

We next determined the structural changes after HDAC3 inhibition. Compared with WT mice, OVE26 mice exhibited a higher heart weight to tibia length ratio (Figure 2A) and enlarged cardiomyocyte size as determined by FITC-conjugated wheat germ agglutinin (WGA) staining (Figure 2B and C). The mRNA and protein expression of ANP, a cardiac disease marker, was also increased in the OVE26 mouse heart (Figure 2D–F) at both 3 and 6 M time points. However, all these cardiac hypertrophic changes were significantly ameliorated by the treatment of RGFP966, which still sustained until 3 months after the end of 3-month treatment. There was no significant difference between RGFP966 and VPA treatment groups in terms of cardiac hypertrophy improvement.

Compared with WT mice, OVE26 mice exhibited increased fibrosis as revealed by the Sirius Red staining (Figure 3A and B), which was supported by the increased expression of collagen I and FN-1 (Figure 3C and D). Evidence of cardiac fibrosis was further consolidated by the changes in the expression of the profibrotic mediator, CTGF and FN-1 as determined by Western blot (Figure 3E–G). RGFP966 treatment decreased the collagen accumulation and the expression of CTGF and FN-1 in the diabetic heart. There was no significant difference between RGFP966 and VPA treatment groups in terms of cardiac fibrosis improvement. Collectively, these findings suggest that 3-month administration of RGFP966 significantly prevents the diabetes-related cardiac remodeling at both 3 and 6 M time points.

Oxidative stress and inflammation play important roles in diabetes-induced cardiac pathogenesis. Consistent with this notion, ROS production, detected by staining with DHE (a superoxide indicator, Figure 4A and B) and lipid peroxidation, reflected by MDA contents (Figure 4C), were increased in the heart of OVE26 diabetic mice, but not significantly in VPA- or RGFp966-treated OVE26 diabetic mice, compared with that of in the WT heart. Furthermore, Western blot of 4-HNE (Figure 4D and E) and 3-NT (Figure 4F and G) as oxidative and nitrative proteins respectively were significantly increased in OVE26 diabetic mice, which was significantly reduced by RGFP966 at both 3 and 6 M time points. Similarly, the expressions of inflammatory factors PAI-1 and TNF-α (Figure 5A–C) were also substantially increased in the OVE26 diabetic heart compared with that of in the WT heart, which were suppressed in the RGFP966-treated diabetic mouse at 3 and 6 M time points. There was no significant difference between RGFP966 and VPA treatment groups in terms of cardiac oxidative stress and inflammation improvement.

Insulin resistance also plays an important role in the Type 1 diabetes disease process that is commonly recognized [26]. We checked whether the insulin receptor signaling pathway is reduced in OVE26 mice and improved by the treatment of HDAC inhibition. As shown in Figure 6, HDAC3 inhibitor reversed the diabetes-induced classical insulin pathway suppression, including the up-regulation of IRS1 expression, p-Akt to total Akt ratio, and GLUT4 expression (Figure 6A–F), as did VPA, at both two time points. Interestingly, the expression of GLUT1 was not affected by the treatment of HDAC inhibitors in the OVE26 diabetic mouse heart (Figure 6G and H).

It is reported that AMPK is associated with the process of glucose metabolism in the high fat diet-fed diabetic mice [27]. In the OVE26 diabetic mouse model, we also found that p-AMPKα decreased in the diabetic heart, which was reversed by the treatment of RGFP966 and VPA at both two time points without difference between two kinds of inhibitors (Supplementary Figure S1A and S1B).

Previous studies have shown a connection between DCM and the mitogen-activated protein kinase (MAPK) pathway [28], so we decided to examine whether HDAC3 inhibitor could block the activation of major MAPK family members: ERK1/2, JNK, and p38 MAPK. As shown in Figure 7 and Supplementary Figure S1, phosphorylated ERK1/2, JNK, and p38 MAPK protein levels were significantly increased in the hearts of OVE26 diabetic mice compared with those of WT hearts at both 3 and 6 M time points. RGFP966 treatment blocked the activation of cardiac ERK1/2 (Figure 7A and B), but not JNK or p38 MAPK in OVE26 diabetic mice (Supplementary Figure S1A–C). These inhibitory effects imposed by RGFP966 persisted for another 3 months after termination of treatment. There was no significant difference between RGFP966 and VPA treatment groups in terms of the MAPK expression.

A previous study showed that increased nuclear ERK1/2 signaling was linked to the development of cardiac hypertrophy [18]. Therefore, we further examined the expression levels of ERK1/2 in the nucleus by Western blot. Indeed, the increased expression of nuclear ERK1/2 in the OVE26 diabetic mouse heart was observed, which was suppressed by RGFP966 (Figure 7C and D). As a nuclear ERK1/2 phosphatase, DUSP5 specifically dephosphorylates and inactivates ERK1/2 in the nucleus [29]. To address whether HDAC3 inhibitor could stimulate DUSP5 expression in the diabetic mouse heart, the DUSP5 expression was investigated at both mRNA and protein levels. The DUSP5 expression in the OVE26 mouse heart was inhibited at both 3 and 6 M time points, which was reversed by the treatment of RGFP966 (Figure 7E–G). Next, to further understand how the HDAC3 inhibitor increased the expression of DUSP5 in the diabetic mouse heart, we performed ChIP assay with an acetylated histone H3 antibody to determine the transcriptional management of DUSP5 in chromatin. We found that RGFP966 increased the acetylated level of histone H3 on the cardiac DUSP5 gene promoter compared with the OVE26 diabetic group at both 3 and 6 M time points (Figure 7H). There was no significant difference between RGFP966 and VPA treatment groups with regard to the expression of nuclear ERK1/2, DUSP5, and acetylated histone H3.

A few studies have shown the association of total HDAC inhibition with DCM, which is substantially advanced by the present work in the following aspects: (1) We are the first study to specifically define the pivotal role of HDAC3 inhibition in the prevention of DCM; (2) Both previous two studies have used STZ-induced Type 1 diabetes model that may involve in the STZ’s effect on the total HDAC activity. Here, we have used well-appreciated genetically OVE26 spontaneous type 1 and its potential epigenetic cardioprotection against DCM in the Type 1 diabetic mice; (3) There was no significant difference between HDAC inhibition and HDAC3 inhibition groups in terms of cardiac function and pathology improvement in the OVE26 diabetic mouse, which suggests the predominant role of HDAC3 in the development of DCM; (4) Mechanistically, the most important protection by HDAC3 inhibition against DCM probably is mediated by its suppression of diabetes-induced activation of ERK1/2 in the nucleus, but not that of JNK and p38 MAPK, and also increases in the expression of DUSP5, a nuclear ERK1/2-specific phosphatase and the acetylation of histone H3 on the promoter region of DUSP5 gene. All these results indicate the crucial role for HDAC3 in the development of DCM.

As epigenetic regulators, HDACs catalyze the removal of acetyl groups from lysine residues in a variety of proteins. Classically, HDACs deacetylate nucleosome histones and alter the electrostatic properties of chromatin in a manner that represses gene expression [30,31]. Inhibition of HDACs using small molecules such as butyrate, VPA, or trichostatin A exerts multiple cardioprotective effects, such as the anti-hypertrophy [32,33], anti-fibrosis [34,35], anti-oxidative stress [36], anti-inflammation [37,38], and anti-apoptosis [39] under non-diabetic conditions, and also anti-DCM [14,15]. Although HDACs are the promising therapeutic target for multiple cardiovascular diseases, the diversity of HDAC isoforms constrain the practice. It is urgently needed to clarify the association of specific HDAC and disease, and to develop HDAC inhibitors with high specificity and selectivity in order to reduce unwanted off-target effects.

In the present study, we used RGFP966, an HDAC3 inhibitor [40], to examine the epigenetic effect of HDAC3 on DCM. In addition, VPA was used as the total HDACs inhibitor to detect the cardioprotection in the OVE26 Type 1 diabetes mice too. Although butyrate was used in the previous studies as total HDAC inhibitor [15], pharmacokinetic studies have confirmed its lower cardiac distribution, compared with VPA [41,42]. In addition, phase II/III clinical trials have confirmed that VPA is a safe medicine for the therapy of some cancers [43,44]. More importantly, VPA is also an established drug for the long-term therapy of epilepsy because of its HDAC inhibition activity [45]. Up to date, a number of studies have shown that VPA plays an important role in the prevention of multiple cardiac pathological changes induced by different stimuli [36,46–48], including Type 2 diabetes [49].

As shown in the results, HDAC3 specific inhibitor RGFP966 did not affect the blood glucose levels, but inhibited the activity of HDAC3 and preserved cardiac function in diabetic mice. Meanwhile, RGFP966 prevented the major pathological manifestations including cardiac hypertrophy as evidenced by decreased expression of hypertrophic markers, the heart weight to tibia length ratio, and cardiomyocyte size. It also significantly diminished the cardiac fibrosis in diabetic mice as evidenced by reduced expression of the fibrosis markers and collagen accumulation. Pan-HDAC inhibitor VPA similarly diminished cardiac injuries induced by diabetes as HDAC3 specific inhibitor RGFP966. Together, HDAC3 inhibition could protect the heart from the injury induced by diabetes in different perspectives.

Ferguson et al. [18] found that Trichostatin A, a non-selective HDAC inhibitor, suppressed the ERK1/2 phosphorylation induced by different stimuli. A prior study revealed that HDAC3 governed the TGF-β-induced phosphorylation of ERK1/2 in C3H10T1/2 cells [50]. In the present study, we found that HDAC3 inhibitor RGFP966 not only decreased the cardiac nuclear ERK1/2 but also reduced DCM in the OVE26 Type 1 diabetes mice, which indicates that down-regulation of cardiac nuclear ERK1/2 might be a key mechanism of hampering DCM by HDAC3 inhibitor on the basis of the argument that nuclear ERK1/2 signaling plays a major role in pathological cardiac hypertrophy [51,52] and the development of DCM [16,53].

As a nuclear ERK1/2 specific phosphatase, DUSP5 dephosphorylates the activated ERK1/2 and inhibits the hypertrophic signaling pathway induced by phosphorylated ERK1/2. Meanwhile, based on the study of Ferguson and his group, the MEK-ERK signaling promotes class I HDAC-mediated suppression of DUSP5 expression, thereby providing a self-reinforcing mechanism to sustain ERK1/2 signaling in response to stress stimuli [18]. Our results showed that the expression of DUSP5 was decreased in the diabetic heart, which was prevented by RGFP966. Mechanistically, this prevention was achieved potentially through enhancing histone H3 acetylation on the DUSP5 gene promoter region. HDAC3 specific inhibitor RGFP966 similarly diminished nuclear ERK1/2 expression, up-regulated DUSP5 and histone H3 acetylation in the OVE26 diabetic heart as did Pan-HDAC inhibitor VPA. These results indicate that epigenetic regulation of the DUSP5-ERK1/2 pathway via inhibiting HDAC3 might be the key component for HDAC-target therapy for DCM.

The most impressive feature in this study was that we set up two time points to evaluate the protective effects of the HDAC3 inhibitor on DCM mice, in which mice were treated with the agent for 3 months followed by non-treatment for another 3 months. The inspiration of this setup came from the ‘metabolic memory’ phenomenon that associated with epigenetics [10,11]. We found that the activity of both HDAC3 and total HDAC is increased in the heart of diabetic mice. These diabetes-induced cardiac damages were significantly inhibited for the first 3 months by treatment with RGFP966 or VPA, which sustained for another 3 months after the end of 3-month treatment. Such phenomenon is called as ‘cardiac protective memory’, which is similar to ‘metabolic memory’, observed in previous studies for diabetic patients with intensified glycemic control [10,11]. In the present work, the level of acetylated histone H3 on the DUSP5 gene promoter region was still elevated by the HDAC3 inhibitor at 6 M time point, which might contribute to the maintenance of ERK1/2 inactivation even during the withdrawal period and produce the ‘cardiac protective memory’ by epigenetic up-regulation of DUSP5.

As shown in Figure 6, however, HDAC3 inhibitor reversed diabetes-induced classical insulin resistance in the heart, shown by up-regulating IRS1 expression, Akt phosphorylation, and GLUT4 expressions, as does VPA, at both two time points. These results are consistent with previous studies on the effect by HDAC inhibition on insulin resistance in different models [54–56], including in the heart of diabetic mice [15,57,58]. Therefore, reducing insulin resistance may contribute, at certain levels, to the cardioprotection of HDAC3 inhibitor in OVE26 diabetic mouse model although the exact mechanism needs to be further explored in our future work. Here what calls for special attention is that the lowering blood glucose was not found, although previous studies indicated that HDAC inhibitors could reduce blood glucose through improving insulin resistance-related signaling pathway. In a low dose STZ- and high fat diet-induced Type 2 diabetic mouse model, HDAC inhibition could reduce blood glucose through improving the insulin receptor signaling pathway [57]. In an STZ-induced Type 1 diabetic mouse model (single dose of STZ at 200 mg/kg), a complete elimination of pancreatic β-cell model that is similar to what we used OVE26 mice [59], HDAC inhibition didn’t decreased blood glucose although it reversed the reduction of cardiac GLUT1 and GLUT4 induced by diabetes [15], a similar pattern as what we show in OVE26 mice (Figure 1B, Figure 6). Interestingly, HDAC inhibition could reduce the blood glucose slightly in STZ-induced diabetic rat model where a single dose of STZ at 60 and 65 mg/kg respectively was given [14,56]. Therefore, in multiple low doses of STZ-induced diabetic mouse model, and a single dose of STZ-induced diabetic rat model, pancreatic β cells may not have been completed destroyed, which were probably rescued by inhibition of HADCs in certain unknown reasons, resulting in certain small amount of insulin to reduce blood glucose. For this discrepancy, we will further explore in the future studies.

For the result of GLUT4 expression, we only examined the change in the heart, which partly consistent with the study mentioned above [15], but didn’t know whether GLUT4 was up-regulated in other organs. Therefore, based on our own results and a previous work [15], we may conclude that, even though the extent of the cardiac IRS1-Akt-GLUT4 change were improved in certain extends, it remains not enough to reduce blood glucose in the severe Type 1 diabetic models without insulin level improvement (OVE26 mice and one single large dose of STZ-induced mode).

Several studies revealed the activation of AMPK-related signaling pathway with non-specific HDAC inhibitors, MPT0E014, and VPA, in both Type 1 diabetic rats and obese mice with several other metabolic variables [58,60]. However, the effect of the HDAC3 inhibitor on AMPK in the diabetic heart remains unclear. As shown in Supplementary Figure SA, we found that AMPKα phosphorylation was decreased in the diabetic heart but not significantly in the heart of diabetic mice treated with both HDAC inhibitors VPA and RGFP966. However, whether activation of AMPK-mediated metabolism by inhibition of HDACs or HDAC3 contributes to the cardiac protection from DCM remains unclear and will be further explored in the future.

There remain a few limitations in the present study. For example, we did not perform the examination for insulin resistance regrettably although we provide the result of insulin signaling pathway. In addition, we did not further confirm the findings in the cardiac specific ERK1/2 knockout mouse, which is not available for the time being. We did not directly define the pivotal role of ERK1/2 in the development of DCM though nuclear ERK1/2 activity was previously reported to modulate cardiac hypotrophy in other heart disease models [18]. The more detailed mechanisms on how HDAC3 mediates the expression of DUSP5 also remains to be explored in the future.

In summary, our study reveals for the first time that HDAC3 inhibition by RGFP966 prevents diabetes-induced cardiac dysfunction and remodeling in the OVE26 diabetic mice. The cardiac protection by inhibition of HDAC3 from DCM is persistent at least for 3 months after the end of treatment, like a ‘protective memory’ phenomenon. Mechanistically, this cardioprotection of HDAC3 inhibition may be mainly mediated by the epigenetic regulation of DUSP5-ERK1/2 pathway via acetylation of histone H3 on the DUSP5 gene promoter. However, we are also aware that HDAC3 inhibition-mediated stimulation of insulin intracellular signalings and the activation of AMPK may also have certain contributions. The present study suggests the potential application of HDAC3 inhibitor for the prevention of DCM.

• Although a few studies showed the association of total HDAC inhibition with DCM, we are the first to specifically define the pivotal role of HDAC3 inhibition in preventing DCM.

• HDAC3 inhibition prevents Type 1 diabetes-induced cardiac inflammation, remodeling, and dysfunction, which reaches to the same effect of total HADC inhibition. The cardiac protection by HDAC3 inhibition from DCM is persistent even until 3 months after the end of treatment, like a phenomenon of ‘protective memory’.

• Mechanistically, the cardioprotection of HDAC3 inhibition is probably mediated by the epigenetic regulation of DUSP5-ERK1/2 pathway via acetylation of histone H3 on the DUSP5 gene promoter.

• The present study suggests that HDAC3 may be an important target potentially for clinical application to prevent DCM in Type 1 diabetic patients.

J.S., Q.T., Z.Z., Y.Z., S.-Y.C., Q.L., and L.C. contributed to the initial experimental discussion and designs. J.S. and L.C. consistently directly guided the research project progression. Z.X. set up the animal model, analyzed the data, and wrote the manuscript. Z.X., S.W., and L.Q. performed experiments. J.S., S.-y. C, and L.C. wrote or revised the manuscript. All authors have reviewed the final manuscript and approved the submission to this journal.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported by the National Science Foundation of China [grant numbers 81370318, 81570339 (to Y.Z.), 81570338 (to Z.Z.), 81302860 (to Q.L.), and 81573493 (to Q.T.)]; the Jilin Province Science and Technology Development Project [grant number 20130206021SF (to J.S.)]; the Scholarship of China Scholarship Council [grant number 201508330753 (to L.-B.Q.)]; the American Diabetes Association [grant number 1-15-BS-018 (to L.C.)]; and the National Institute of Health [grant numbers AA021434, AA024337, and AA020265 (to S.-Y.C)].

3-NT

3-nitrotyrosine

4-HNE

4-hydroxy-2-nonenal

AMPK

5΄ AMP-activated protein kinase

ANP

atrial natriuretic peptide

ChIP

chromatin immunoprecipitation

CTGF

connective tissue growth factor

DCM

diabetic cardiomyopathy

DHE

dihydroethidium

DUSP5

dual specificity phosphatase 5

EF

ejection fraction

ERK

extracellular signal-regulated kinases

FN-1

fibronectin-1

FS

fractional shortening

GLUT1 and 4

glucose transporter type 1 and 4

HDAC

histone deacetylase

IRS1

insulin receptor substrate 1

JNK

c-Jun N-terminal protein kinase

LV

left ventricular

LV vol; d

LV end-diastolic volume

LV vol; s

LV end-systolic volume

LV

left ventricular

LVID; d

LV internal diastolic diameter

LVID;s

LV internal systolic diameter

MAPK

mitogen-activated protein kinase

MDA

malondialdehyde

PAI-1

plasminogen activator inhibitor-1

ROS

reactive oxygen species

RT-qPCR

reverse transcription quantitative polymerase chain reaction

STZ

streptozotocin

TNF-α

tumor necrosis factor-α

VPA

valproic acid

WGA

wheat germ agglutinin

WT

wild-type