Abstract
Although the clinical importance of heart failure with preserved ejection fraction (HFpEF), which makes up half of heart failure, has been extensively explored, most therapeutic regimens, including nitric oxide (NO) donors, lack therapeutic benefit1-12. Here we report that neuronal nitric oxide synthase (nNOS, also known as NOS1) induces HFpEF by S-nitrosylation of histone deacetylase 2 (HDAC2). HFpEF animal models—SAUNA (SAlty drinking water/Unilateral Nephrectomy/Aldosterone)13,14 and mild transverse aortic constriction (TAC) mice14,15—showed increased nNOS expression and NO production, which resulted in the S-nitrosylation of HDAC2. HFpEF was alleviated in S-nitrosylation-dead HDAC2 knock-in mice. Pharmacologic intervention by either nNOS inhibition or HDAC2 denitrosylation attenuated HFpEF. Our observations are the first to demonstrate a completely new mechanistic aspect in HFpEF, which may provide a novel therapeutic approach to HFpEF. In addition, our results provide evidence for why conventional NO-enhancement trials have not been effective for improving HFpEF.
Introduction
Heart failure refers to an insufficient circulation of blood to peripheral organs. The well-known type is systolic heart failure, the characteristics of which are a loss of effective myocardium for contraction16,17. In recent decades, in part due to the prevalence of metabolic diseases such as diabetes, the patient population with HFpEF has grown, with a parallel increase in the clinical and social burdens of this disease5,6,8,10. To date, 50% of total heart failure is regarded as HFpEF6. Unfortunately, none of the therapies that have been proven effective for heart failure with reduced ejection fraction (HFrEF) have been proven beneficial for increasing the survival rate of patients with HFpEF2,11,12,18-21. It is surprising that even NO donors, which are potent vasodilators, do not have any benefit in the treatment of HFpEF2,11, and this raises fundamental questions as whether our strategy of replenishing NO is appropriate. Indeed, in the current work, we propose that NOS/NO-mediated nitrosylation of protein may worsen HFpEF, suggesting critical reconsideration of the therapeutic effect of NO.
Results
Although no perfect mouse model for HFpEF is available14, we applied the SAUNA model (Methods, Fig. 1a), which is known to be most similar to human HFpEF13,14. At 30 days after initiation of the protocol, we measured cardiac function: parameters of exercise capacity such as ejection fraction (EF), early diastole (E), and mitral valve annulus movement (E’); and performance on a rotarod treadmill test (Methods). Systolic function was well conserved in the SAUNA group. However, the E/E’ ratio was aberrantly increased (Fig. 1b, Extended Data Fig. 1a). Furthermore, locomotive activities were impaired in SAUNA mice (Extended Data Fig. 1c). Thus, we confirmed that the SAUNA model successfully induced HFpEF. As an alternative model, mild TAC (Methods) was introduced. Four weeks after mild TAC, we assessed cardiac function (Extended Data Fig. 1b, 1c, 1d). We found that mild TAC could also induce HFpEF in less than 1 month.
To delineate the mechanism of HFpEF, we measured posttranslational modifications of proteins in both mouse models. Among the posttranslational modifications tested, protein S-nitrosylation was significantly increased (Fig. 1d), whereas no notable alterations in acetylation, methylation, or phosphorylation were observed (Extended Data Fig. 1e, 1f, 1g). NO can be generated by three nitric oxide synthases: nNOS, iNOS2 (NOS2), and eNOS (NOS3)22. To understand the source of the NO, we measured transcript amounts of nNOS, iNOS, or eNOS by quantitative real-time PCR. Both nNOS and iNOS were dramatically increased in mild TAC mice, whereas only nNOS was activated in the SAUNA model (Fig. 1e, 1f). We confirmed that nNOS was predominantly increased in cardiomyocytes isolated from SAUNA-treated mice (Methods) (Fig. 1g).
Several pharmacologic agents specifically inhibit nNOS23-28. Thus, to confirm the involvement of nNOS, we used N(ω)-propyl-L-arginine (NPLA)28 in our HFpEF models. Injection of NPLA every other day (Methods) improved both diastolic dysfunction and exercise capacity (Fig. 1h, 1i, 1j)
Next, we looked for nNOS-specific S-nitrosylation targets in the heart by using a biotin-switching assay29 (Methods). Infection of H9c2 cells with adenovirus nNOS-HA generated de novo S-nitrosylation (Fig. 2a). By mass spectrometry, we identified that nNOS S-nitrosylated of Hdac2 and Gapdh (Fig. 2a). Incubation of S-nitrosoglutathione (GSNO), a nonselective NO donor, with heart lysates increased S-nitrosylation of both proteins (Extended Data Fig. 2a). S-Nitrosylation of Hdac2 and Gapdh was further confirmed by western blot (Fig. 2b, Extended Data Fig. 2b, 2c).
To test whether SAUNA stresses induced S-nitrosylation of Hdac2, we performed the biotin-switching assay with SAUNA heart lysates and observed that S-nitrosylation of Hdac2 was significantly increased (Fig. 2c, 2d). S-Nitrosylation of Hdac2 was also observed in the mild TAC model (Extended Data Fig. 2d). S-Nitrosylation of Hdac2 was nearly completely abolished in SAUNA mice with NPLA administration (Fig. 2e).
nNOS is mainly located in the cytoplasmic membrane, whereas HDAC2 is tethered in the nucleus. Hence, we assumed that nNOS ‘indirectly’ S-nitrosylates Hdac2 by transferring NO from membranous nNOS to nuclear HDAC2. Kornberg et al30 reported that GAPDH may work as a shuttling mediator of NO after its S-nitrosylation at Cys150 and after redistribution in the cell. In our experimental models, Gapdh was also S-nitrosylated by nNOS (Fig. 2a), which then resulted in its nuclear redistribution (Fig. 2f). It is also noteworthy that physical interaction between Gapdh and Hdac2 increased in the presence of nNOS (Fig. 2g). When a nonselective nitrosylation-inducer, GSNO, was incubated with heart lysates, Gapdh also physically interacted with Hdac2 (Extended Data Fig. 2e).
Two cysteine residues, Cys262 and Cys274, are known to be responsible for S-nitrosylation of HDAC231. These residues are highly conserved throughout species (Fig. 3a). We generated S-nitrosylation-resistant mutant mice by gene knocking-in technology and substituting those two cysteines with alanines (hereafter, HDAC2 2CA). We could not observe S-nitrosylation of HDAC2 in HDAC2 2CA mouse embryonic fibroblasts even in the presence of adenovirus nNOS-HA (Fig. 3b).
We then applied the SAUNA model to HDAC2 2CA mice. SAUNA failed to exaggerate diastolic dysfunction in HDAC2 2CA mice. The aberrant increase in the E/E’ ratio seen in wild type was not observed in HDAC2 2CA (Fig. 3c). HDAC2 2CA mice also did not develop SAUNA-induced impairment of exercise tolerance (Fig. 3d). As expected, systolic function was well conserved in HDAC2 2CA mice even in SAUNA stresses (Extended Data Fig. 3a). SAUNA-induced HDAC2 S-nitrosylation was not detected in HDAC2 2CA hearts (Fig. 3e). Mild TAC also failed to induce diastolic dysfunction in HDAC2 2CA mice (Extended Data Fig. 3b). Thus, we concluded that HDAC2 2CA mice can tolerate HFpEF.
We next questioned whether S-nitrosylation affects HDAC2 function in HFpEF by measuring HDAC2 activity (Methods). The deacetylase activity of HDAC2 from heart lysates was not changed by GSNO-induced nitrosylation (Extended Data Fig. 3c). Thus, rather than its deacetylase activity, other functions of HDAC2 such as intracellular localization or alterations in complex formation may contribute to the development of HFpEF. This remains to be clarified.
Heretofore, we delineated the roles of nNOS in association with HDAC2 C262/C274 S-nitrosylation in the development of HFpEF. Hence, we questioned whether pharmacologic intervention to reduce the nitrosylation of HDAC2 could also alleviate HFpEF. As reported previously32, we observed that NRF2 (also known as NFE2L2) can denitrosylate HDAC2 in cardiac cells. Adenovirus NRF2 significantly activated its target gene, Glutathione S-Transferase Alpha 3 (GSTA3), in H9c2 cells (Extended Data Fig. 4a, 4b). Adenovirus nNOS-HA infection in H9c2 cells successfully induced HDAC2 S-nitrosylation, which was attenuated by infection of adenovirus NRF2 (Fig. 4a).
To test the effect of NRF2-induced denitrosylation on HFpEF, we designed an AAV9-NRF2 delivery scheme (Fig. 4b). Direct cardiac injection of AAV9-NRF2 ameliorated diastolic dysfunction in SAUNA mice (Fig. 4c) without any significant changes in systolic function (Extended Data Fig. 4c). Locomotor intolerance was also improved by cardiac overexpression of NRF2 (Fig. 4d). As expected, S-nitrosylation of HDAC2 was notably attenuated in the AAV9-NRF2 injection group (Fig. 4e). AAV9-NRF2 infection was confirmed by transcription amounts of the target gene, GSTA3 (Extended Data Fig. 4d).
Although the specific mechanism remains unclear, dimethyl fumarate (DMF) has been reported to induce NRF2 transcription33. We tested whether DMF could alleviate the HFpEF phenotype (Methods). Daily intake of DMF activated NRF2 and subsequently induced GSTA3 (Extended Data Fig. 4e, 4f). Like AAV9-NRF2, HDAC2 S-nitrosylation was not observed in DMF-administered mice (Fig. 4f). DMF also improved both diastolic dysfunction (Fig. 4g) and exercise tolerance (Fig. 4h), which suggests DMF as a novel therapeutic strategy for HFpEF.
To study whether S-nitrosylation of HDAC2 is also involved in HFrEF, we induced severe TAC in HDAC2 2CA mice or wild type littermates (Methods) and measured survival for 15 weeks. Interestingly, HDAC2 2CA mice showed no survival benefit compared with their wild type littermates in this model (Extended Data Fig. 4g). Severe impairment of EF, a surrogate indicator of HFrEF, was also observed as either in the wild type mice or in DMF-administered mice (Extended Data Fig. 4h). Similarly, DMF did not improve the survival rate of wild type mice exposed to severe TAC. These findings suggested that nitrosative stress specifically induces HFpEF rather than general heart failure.
No established regimens are currently available for the patients with HFpEF3,6,10,17,34. Whereas NO has beneficial effects like vasodilation and reduction of afterload in HFrEF, NO donors are not at all effective for improving the survival of HFpEF2,11. Although preliminary, our new findings that NO-mediated nitrosylation provokes HFpEF may explain why NO donors do not work in the clinical setting. Our study is not the only work to suggest NO as an inducer of HFpEF; indeed, a recent report also implied NOS/NO-mediated nitrosylation of protein in the development of HFpEF35. Thus, it is very likely that S-nitrosylation of key proteins may lead the scientific world to revisit NO as a potential therapeutic target of HFpEF. Of note, unlike in HFrEF, however, inhibition of NOS or removal of S-nitrosylation could be considered in HFpEF.
Methods
Antibodies
Antibodies used were as follows: mouse monoclonal anti-HDAC2 (Abcam, 1:5,000, 12169), rabbit polyclonal anti-HDAC2 (Invitrogen, 1:1,000, 51-5100), mouse monoclonal GAPDH (Bio-Rad, 1:1,000, VMA00046), rabbit polyclonal anti-GAPDH (Bio-Rad, 1:1,000, VPA00187), and mouse monoclonal anti-HA (Sigma, 1:30,000, H9658).
Animal model
The use of animal experiments for disease models was approved by the Chonnam National University Medical School Research Institutional Animal Care and Use Committee (CNU IACUC-H-2019-3). For the HFpEF model, 8-week-old male C57BL/6 mice underwent the SAUNA (SAlty drinking water/Unilateral Nephrectomy/Aldosterone) model. The mice were anesthetized with 2,2,2-tribromoethanol (300 mg·kg-1, Sigma, T48402) and placed in the right lateral decubitus position. The left kidney was removed and a micro-osmotic pump (Alzet®, Durect Corp, 1004) containing d-aldosterone (Sigma, 0.30 μg·h-1, A9477) was implanted under the back skin. Starting 1 day after the operation, the mice were given drinking water with 1% NaCl for 30 days. Exercise capacity and cardiac function were assessed at 30 days. N(ω)-Propyl-L-arginine (NPLA, Cayman, 80587) (50 mg·kg-1·day-1, every other day) was injected to inhibit nNOS until the end of the study. To induce Nrf2, DMF (Sigma, 0.5 g·l-1 in drinking water) was administered together with drinking water for 30 days. AAV9-NRF2 infection (Vector Biolabs, AAV-216638) was carried out by direct cardiac injection. C57BL/6 mice were anesthetized with 2,2,2-tribromoethanol and maintained with an artificial ventilator. The intercostal space between the 4th and 5th rib was cut and widened. 1×1012 genome copies of AAV9-NRF2 was injected in the left ventricular free wall. To guarantee successful expression of infected-AAV9-NRF2, the AAV virus was injected 2 weeks before the SAUNA operation. For an alternative HFpEF animal model, mild TAC was carried out to induce isolated diastolic dysfunction without a concomitant change in systolic function. A partial thoracotomy was performed at the second rib and two loose knots were made between the brachiocephalic artery and the left common carotid artery with a 26-gauge needle. For severe TAC, a 27.5-gauge needle was utilized.
Adult cardiomyocyte isolation
Adult mouse ventricular cardiomyocytes were obtained from C57BL/6 mouse heart. Mice were injected with 50 units of heparin and were euthanized by cervical dislocation. The heart was quickly harvested and cannulated with calcium-free Tyrode buffer (10 mM HEPES pH 7.4, 137 mM NaCl, 5.4 mM KCl, 1 mM MgCl2, 10 mM glucose, 5 mM taurine, and 10 mM 2,3-butanedione monoxime) for 3 min with 100% oxygen. The enzyme digestion was carried out by digestion buffer (calcium-free Tyrode buffer supplemented with hyaluronidase [Worthington, 0.1 mg·ml-1, LS005477] and collagenase type B [Roche, 0.35 U·ml-1, 11088807001]). The left ventricle was collected after 10 minutes’ digestion and cut into small pieces. Further digestion was performed with gentle stirring for 10 minutes. The cells were allowed to stand briefly for large parts to settle and supernatants were filtered by use of a 100-mm pre-cell strainer. The filtered cells were plated on culture dishes for 2 hours to remove cardiac fibroblasts.
Biotin-switching assay
Protein S-nitrosylation was measured after biotin-switching29 with slight modification. All reactions were carried out in dark amber tubes to avoid UV exposure. Protein lysates were prepared with HENS buffer (250 mM HEPES, pH 7.7, 1 mM EDTA, 0.1 mM neocuproine [Sigma, P9599], 1% SDS) and protease inhibitor cocktail. Lysates were prepared in 200 μL of HENS buffer containing 100 μg (cell lysate) or 500 μg (heart lysate) of protein. Free thiol groups were blocked with blocking solution (250 mM HEPES, pH 7.7, 1 mM EDTA, 0.1 mM neocuproine [Sigma, P9599], 1% SDS, and 25 mM S-methylmethane thiosulfonate [Sigma, 208795]) for 1 hour at room temperature. The blocking step was terminated by adding 1 mL of ice-cold acetone. Protein was precipitated by centrifugation at 6,000 RCF for 15 minutes. After removing acetone, the pellet was dissolved in 170 μL of HENS buffer. S-NO was removed by adding 10 μL of 1 M sodium ascorbate for 20 minutes, and subsequent biotin labeling was carried out by adding 20 μL of 4 mM biotin-HPDP (Thermo, 21341) for 40 minutes. The reaction was stopped by ice-cold acetone precipitation. Biotin-labeled protein was dissolved in 400 μL of HENS buffer and neutralized with 800 μL of 1% NP buffer (1% Igepal CA-630, 50 mM Tris pH 8.0, 150 mM NaCl, 1 mM EDTA) and protease inhibitor cocktail. The biotin-switching protein was collected with 30 μL of streptavidin-agarose beads and continuous rotation at 4°C overnight. The beads were washed three times with 1% NP buffer and the samples were analyzed by SDS-PAGE.
Echocardiogram
Cardiac function was measured by ultrasonography (General Electric Company, Vivid S5). Mice were anesthetized with 2,2,2-tribromoethanol and checked for lack of response with a light touch. At the papillary muscle level, 2-dimension M-mode was acquired from the parasternal long-axis view or parasternal short-axis view. Ejection fraction was determined by the Teichholz formula: EF(%)=(Vd-Vs)/Vd, where Vd indicates LV volume at end diastole and Vs is at end systole, and Vd=[7/(2.4+LVIDd)]×LVIDd^3, Vs=[7/(2.4+LVIDs)]×LVIDs^3, where LVIDd is LV interventricular dimension at end diastole and LVIDs is at end systole. To assess diastolic function, E, and E’ were measured. After measurement of the parasternal short-axis view, the sonographic probe was tilted 45 degrees to visualize the parasternal 4-chamber view. Mitral E waves were recorded with pulse-wave mode at the mitral valve opening. Mitral annulus movement, also known as E’ wave, was assessed from the medial mitral valve annulus with tissue velocity image mode.
Genetically engineered mice
HDAC2 S-nitrosylation-dead knock-in mice (HDAC2 2CA) were generated by a commercial company (Toolgen). Mouse HDAC2 genomic sequences flanking cysteine 262 and cysteine 274 were exchanged as follows: TGT GGC GCA GAC TCC CTG TCT GGG GAC AGG CTT GGT TGT to GCT GGA GCC GAT AGC CTT AGC GGA GAT CGC CTGGGA GCT. Protein sequences were not changed except for two targeted cysteines (CGADSLSGDRLGC to AGADSLSGDRLGA). Genotype was determined by T7E1 (NEB, M0302). Direct Sanger sequencing of PCR products was requested to further confirm the genotype of the homozygote. The oligomer set for genotyping was as follows: sense: 5’-TGCTGTCAATTTTCCCATGA-3’, antisense: 5’-AGAGTTTGGCATCGAGTTGG-3’.
Histone deacetylase activity assay
A commercial kit (HDAC-Glo™ 2 Assays, Promega, G9590) was used to measure HDAC2 activity. Heart lysates were prepared with 1% NP buffer without EDTA to avoid zinc chelation. Fifty micrograms of heart lysates were mixed with HDAC2 assay substrate and incubated for 15 minutes at room temperature. To check the S-nitrosylation effect of HDAC2, 500 μM of GSNO (Cayman, 82240) or 500 μM of GSH (L-Glutathione, Cayman, 10007461) was added to the assay buffer. The deacetylase activity of HDAC2 was measured with a luminometer. The vehicle-treated condition was regarded as 1 and fold changes were calculated.
Quantitative real-time polymerase chain reaction
Total mRNA was extracted with TRIzol (Invitrogen, 15596026). cDNA was synthesized by use of random hexamer (M-MLV reverse transcriptase, Invitrogen, 28025013). Quantitative real-time PCR was carried out by using QuantiTect SYBR Green kits (Qiagen, 204143) with a Rotor-Gene Q (Qiagen). PCR analysis was performed in triplicate and the average was regarded as a single result. The relative contents of mRNA transcripts were normalized to those of Gapdh. Specific oligomer sets were as follows:
Mouse GAPDH, sense: 5’-GCATGGCCTTCCGTGTTCCT-3’, antisense: 5’-CCCTGTTGCTGTAGCCGTAT-3’
Mouse nNOS, sense: 5’-ACTGACACCCTGCACCTGAAGA-3’, antisense: 5’-GTGCGGACATCTTCTGACTTCC-3’
Mouse iNOS, sense: 5’-CAGCTGGGCTGTACAAACCTT-3’, antisense: 5’-CATTGGAAGTGAAGCGGTTCG-3’
Mouse eNOS, sense: 5’-CCTCGAGTAAAGAACTGGGAAGTG-3’, antisense: 5’-AACTTCCTTGGAAACACCAGGG-3’
Mouse GSTA3, sense: 5’-TACTTTGATGGCAGGGGAAG-3’, antisense: 5’-GCACTTGCTGGAACATCAGA-3’
Mouse 18s, sense: 5’-GTAACCCGTTGAACCCCATT-3’, antisense: 5’-CCATCCAATCGGTAGTAGCG-3’
Human NRF2, sense: 5’-CACATCCAGTCAGAAACCAGTGG-3’, antisense: 5’-GGAATGTCTGCGCCAAAAGCTG-3’
Rat NRF2, sense: 5’-AGAAGCACACTGAAGGCACGG-3’, antisense: 5’-GAATGTGTTGGCTGTGCTTTAGGTC-3’
Rat GSTA3, sense: 5’-AGTCCTTCACTACTTCGATGGCAG-3’, antisense: 5’-CACTTGCTGGAACATCAAACTCC-3’
Rat GAPDH, sense: 5’-ATGACATCAAGAAGGTGGTG-3’, antisense: 5’-CATACCAGGAAATGAGCTTG-3’
Rotarod treadmill test
The locomotor tolerance of mice was measured by using a rotarod system (ENV-577M, Med Associate Inc). Mice were pre-adapted to the rod at a fixed speed of 10 rpm for 5 minutes before measurement. Mice were placed on the round rod (35-mm diameter) with a continuous acceleration from 4 to 40 rpm by 1 rpm per 5 seconds. Motor ability was determined as the time of falling off or passive rolling with grasp rod, up to 5 minutes. The best latency time of 3 tests was used for the study. Time interval between each trial was 30 minutes.
Statistics
Statistical significance was analyzed with PASW Statistics 25 (SPSS, IBM corp). For two independent groups, two-tailed unpaired Student’s t-test was applied after checking for a normal distribution. In the case of more than two groups, one-way analysis of variance was used. Bonferroni’s post hoc test was applied for multiple comparisons when equal variance was assumed by use of Levene statistics, whereas the Dunnett T3 test was used in case of unequal variance. Survival rate was visualized and calculated by use of Prism 8.0 (GraphPad). Significance was determined at values of p < 0.05.
Author contributions
S.Y. performed most of the experiments. M.K. performed fractional western blot, immunoprecipitation assays, and quantitative real-time PCR. H.L. carried out the rotarod treadmill test. G.K. statistically analyzed the numeric data. K-I.N. performed histology staining for cardiac geometry. S.Y, H.K., and G.H.E. designed the whole study and wrote the manuscript. H.K. and G.H.E. financially supported the study.
Competing interests
The authors declare no competing interests.
Acknowledgments
This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT) (NRF-2019004521, 20191058992, 2018R1A2B3001503), Republic of Korea.