Highly expressed periostin in diabetic mouse heart, HG-induced CF and patients with DCM

Firstly, we obtained a mouse cytokine antibody array to probe the potential therapeutic targets in DCM. The protein profiling assay with this cytokine antibody array showed the levels of 32 proteins were higher in diabetic hearts when compared with those in normal heart. Of these, the protein expression of periostin was increased by 14.3 fold in diabetic hearts, making it the top one upregulated protein in DCM (Fig. 1A). This drove us to explore the clinical relevance of periostin in DCM. Compared with healthy controls, serum levels of periostin were significantly enhanced in patients suffering from DCM (Fig. 1B), and serum periostin levels were negatively correlated with ejection fraction in enrolled participants (Fig. 1C). Importantly, the analysis of receiver-operating characteristic (ROC) curve revealed that the changes in serum periostin levels might serve as a possible biomarker for cardiomyopathy in diabetic patients (Fig. 1D, E). In consistence with clinical data, serum levels of periostin were also raised in DCM mice (Fig. 1F). RT-PCR, ELISA, immunoblotting and immunofluorescence consistently revealed higher periostin level in heart tissues of diabetic mice (Fig. 1G–J). Intriguingly, chronic hypercemia upregulated periostin transcription level in cultured cardiac fibroblasts (CF) from diabetic mice, without changing periostin mRNA level in isolated cardiomyocytes (CM) and cardiac endothelial cells (EC) from diabetic mice (Fig. 1K). To establish which cardiac cell type contributed to the upregulated expression levels of periostin, we investigated the expression profiles of periostin in mouse heart with the aid of the Tabula Muris database. Single cell sequencing of mouse heart tissues showed that periostin was mainly expressed in CF among the top 5 mouse heart cell types (Fig. 1L). In parallel, the mRNA and protein levels of periostin were significantly elevated in primary CF upon exposure of high glucose (HG) (Fig. 1M–P). As periostin is a prototypical myofibroblast marker, it is interesting to know whether TGF-β1 or angiotensin II had an impact on the expression of periostin in CF. Results showed that transforming growth factor β1 (TGF-β1) had the strongest ability to upregulate the mRNA level of periostin in CF, followed by HG and angiotensin II (Ang II) (Additional file 2: Fig. S1). These results support the close relationship between cardiac periostin levels and DCM.

Fig. 1

Expression of periostin in patients with DCM, diabetic mice and HG-exposed CF. A A mouse cytokine antibody array showing the cytokine protein changes in heart tissues from control or diabetic mice. B Serum periostin levels in healthy participates, diabetic patients with normal cardiac function, and diabetic patients with heart failure. C Negative correlations between periostin levels and EF. D ROC curve for the diagnostic performance of serum priostin between healthy participants and DCM patients. E ROC curve for the diagnostic performance of serum priostin between diabetic patients with normal cardiac function and DCM patients. F Serum periostin levels in control and diabetic mice. G Relative mRNA level of periostin in the hearts from control and diabetic mice. H Cardiac periostin levels in control and diabetic mice using ELISA. I Representative blots and quantitation of periostin protein in control and diabetic mice. J Immunofluorescence showing the periostin expression in control and diabetic mice. K Cardiac periostin mRNA level in isolated cardiomyocytes (CM), cardiac endothelial cells (EC), and cardiac fibroblasts (CF) from control and diabetic mice. L Tabula Muris database showing the distribution of periostin different cells within the hearts. M Relative mRNA level of periostin in CF exposed to NG or HG. N The protein level of periostin in CF exposed to NG or HG using ELISA. O The protein level of periostin in CF exposed to NG or HG using western blot. P Immunofluorescence showing the periostin expression in CF exposed to NG or HG. n = 4–6. *P < 0.05 versus Control (Con) or normal glucose (NG). Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test (B). The statistical significance of correlations was assessed by Pearson’s correlation coefficient analysis (C). The P-value was calculated by unpaired two-tailed Student’s t-test (F-K)

Periostin deficiency attenuates cardiac injury and dysfunction in diabetic mice

Next, we evaluated the effect of high-fat diet (HFD) plus low dose of STZ on the biometric, morphological, and echocardiographic characteristics in both control and periostin shRNA mice. Consumption of a HFD intake and STZ injection resulted in elevated fasting blood glucose (FBG), enhanced homeostatic model assessment for insulin resistance (HOMA-IR) index, and increased serum cholesterol and triacylglycerols in both control and periostin shRNA mice, effects that were similar in all mouse groups (Additional file 1: Table S7). Noninvasive transthoracic echocardiography disclosed that ablation of periostin led to a significant improvement of diabetes-elicited cardiac diastolic and systolic dysfunction (Fig. 2A–D, Additional file 1: Table S7). Consistent with the improvement of cardiac function, serum levels of lactate dehydrogenase (LDH) and creatine kinase-cardiac (CK-MB) were lower in periostin-deficient mice response to HFD and STZ (Fig. 2E, F), suggesting that periostin knockout mice were resistant to diabetes-induced heart damage. Under normal conditions, no significant difference was found in cardiac injury between control and periostin shRNA mice (Fig. 2G–I). However, HFD and STZ promoted cardiac hypertrophy, fibrosis and oxidative stress in control mice only, but not in the periostin-deficient mice, as evidenced by H&E staining, sirius red staining and DHE staining (Fig. 2G–I). Likewise, loss of periostin suppressed diabetes-triggered upregulations of the pro-hypertrophic genes (ANP, BNP, β-MHC), pro-fibrogenic genes (Col I, Col III, α-SMA), pro-inflammtory genes (IL-1β, IL-6, TNF-α) and pro-oxidative genes (NOX2, NOX4, 12-LOX) at the mRNA levels (Fig. 2J, K).

Fig. 2

Periostin-deficient mice were resistant to diabetes-indcued cardiac dysfunction. A Representative blots and quantitation of periostin protein. B, C Left ventricle EF and FS were quantified. D Representative echocardiographic images showing the effects of perisotin knockdown on cardiac function in control and diabetic mice. E Serum LDH levels in mice. F Serum CK-MB levels in mice. G Representative photographs of the myocardium with H&E staining (Scale bar = 100 μm). H Representative photographs of the myocardium with Sirus red staining (Scale bar = 100 μm). I Representative images of the myocardium with DHE staining (Scale bar = 200 μm). J Heatmap showing the mRNA levels of pro-hypertrophic genes (ANP, BNP, β-MHC) and pro-fibrogenic genes (Col I, Col III, α-SMA). K Heatmap showing the mRNA levels of pro-inflammtory genes (IL-1β, IL-6, TNF-α) and pro-oxidative genes (NOX2, NOX4, 12-LOX). n = 4–6. *P < 0.05 versus Control (Con), †P < 0.05 versus Diabetes. Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test

Periostin overexpression exacerbated diabetes-induced cardiomyopathy

To further explore the role of periostin in DCM, diabetic hearts were infected with AAV vectors encoding mouse periostin. As expected, AAV-mediated periostin overexpression significantly elevated the periostin protein levels in heart tissues (Fig. 3A). Although periostin overexpression itself did not exert any effect on body, nor did it affected global metabolic profiles (Additional file 1: Table S8), its ectopic expression induced cardiac dysfunction, and aggravated diabetes-evoked changes in echocardiographic indices (Fig. 3B–D, Additional file 1: Table S8). Consistent with these results, overexpression of periostin caused increases in serum markers of cellular damage, LDH and CK-MB, with a more pronounced effect under diabetic state (Fig. 3E, F). Histological analysis by H&E staining, sirius red staining and DHE staining also showed that upregulation of periostin accelerated cardiomyocyte hypertrophy, fibrosis and oxidative injury in STZ/HFD-induced diabetic mice as compared with control mice (Fig. 3G–I). Transcriptional analysis of heart revealed that cardiac hypertrophy, fibrosis, inflammation and oxidative burst were intensified in diabetic mice with periostin overexpression (Fig. 3J, K). These findings indicated that periostin might be a driving factor for the development of DCM.

Fig. 3

Periostin overexpression worsened cardiac injury and dysfunction in diabetic mice. A Representative blots and quantitation of periostin protein. B, C Left ventricle EF and FS were quantified. D Representative echocardiographic images showing the effects of perisotin knockdown on cardiac function in control and diabetic mice. E Serum LDH levels in mice. F Serum CK-MB levels in mice. G Representative photographs of the myocardium with H&E staining (Scale bar = 100 μm). H Representative photographs of the myocardium with Sirus red staining (Scale bar = 100 μm). I Representative images of the myocardium with DHE staining (Scale bar = 200 μm). J Heatmap showing the mRNA levels of pro-hypertrophic genes (ANP, BNP, β-MHC) and pro-fibrogenic genes (Col I, Col III, α-SMA). K Heatmap showing the mRNA levels of pro-inflammtory genes (IL-1β, IL-6, TNF-α) and pro-oxidative genes (NOX2, NOX4, 12-LOX). n = 4–6. *P < 0.05 versus Control (Con), †P < 0.05 versus Diabetes. Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test

Periostin knockdown blunted, while periostin overexpression aggravated HG-induced CF activation

To further ascertain the protective role of periostin against DCM, we next determined the effect of periostin knockdown and overexpression on CF activation. As shown in Fig. 4A–C and Additional file 2: Fig. S2A, silencing periostin mitigated HG-elicited CF proliferation, as demonstrated by EdU staining and CCK-8 assay. HG stimulation significantly promoted the formation of MDA and superoxide anions in CF, while this was normalized by depletion of periostin (Fig. 4D–F). Immunoblotting assay revealed lower protein expressions of α-SMA and Col I in periostin-deficient CF when compared with that in vehicle-treated cells (Fig. 4G), indicating that periostin might affect myofibroblast transformation and subsequent cardiac fibrosis. In addition, HG environment lifted the mRNA levels of IL-1β and TNF-α in CF, whereas this inflammatory phenotype was not observed in periostin-deficient cells (Fig. 4H). In contrast to periostin knockdown studies, the proliferation, oxidative burst, inflammation and myofibroblast transformation of CF were further deteriorated in response to HG after overexpression of periostin (Fig. 4I–P, Additional file 2: Fig. S2B). Although hyperglycemia did not change the level of periostin in cardiomyocytes, global knockout of periostin can inhibit cardiomyocyte hypertrophy caused by diabetes. This drives us to investigate whether CF-released periostin could affect cardiomyocyte functions. CF-derived periostin level could be detectable in CF under NG condition, whereas this secreted protein was significantly enhanced in response to HG (Additional file 2: Fig. S3A). Thus, we further studied whether CF-secreted periostin played a role in cardiomyocyte behaviors by transferring the conditioned medium from CF to CM. The conditioned medium containing periostin from CF inhibited the CM viability and caused cellular LDH release in CM (Additional file 2: Fig. S3B-D). The conditioned medium containing periostin from CF resulted in CM hypertrophy, along with higher mRNA levels of pro-hypertrophic genes, including ANP, BNP, β-MHC (Additional file 2: Fig. S3E, F). Additionally, TUNEL staining showed that CF-derived periostin enhanced CM apoptosis (Additional file 2: Fig. S3G), accompanied by the upregulation of pro-apoptotic gene Bax and downregulation of anti-apoptotic gene Bcl-2 (Additional file 2: Fig. S3H).

Fig. 4

Periostin knockdown attenuated, while periostin overexpression intensified HG-induced CF activation. A, B Representative photographs and quantitative analysis of EdU-positive cells. C Cell viability. D MDA contents. E, F Representative photographs and quantitative analysis of DHE staining. G Representative blots and quantitation of α-SMA and collagen I. H Representative mRNA levels of TNF-α and IL-1β. I, J Representative photographs and quantitative analysis of EdU-positive cells. K Cell viability. L MDA contents. M, N Representative photographs and quantitative analysis of DHE staining. O Representative blots and quantitation of α-SMA and collagen I. P Representative mRNA levels of TNF-α and IL-1β. n = 4–6. *P < 0.05 versus Control (Con), †P < 0.05 versus HG. Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test

Periostin is induced by HG in a TGF-β/Smad dependent manner

We further identified the potential mechanisms that underlie HG-induced overexpression of periostin in CF. The luciferase activity from a series of deletion mutants of periostin promoter constructs were measured to discern the potential promoter regions that contributing to periostin upregulation in HG-exposed CF. The luciferase activity in a full-length promoter region of periostin tended to be higher in CF after treatment with HG (Additional file 2: Fig. S4A, B). Nevertheless, the luciferase activity was not further increased when a small region (− 1495 to − 1195) of periostin promoter was deleted under HG conditions (Additional file 2: Fig. S4A, B), suggesting that this promoter region (− 1495 to − 1195) might mediate the upregulation of periostin transcription levels. Previously, the protein expression of periostin was reported to be upregulated in neu-positive breast cancer cells through the PI3K/Akt signaling [32]. TGF-β acts on the receptor-activated Smad2 and Smad3 to regulate homeostatic functions of fibroblasts by regulating the expression of periostin [33, 34]. Therefore, we investigated whether HG upregulated the transcriptional and translational levels of perisotin by regulating the PI3K/Akt and/or TGF-β/Smad pathways. Western blots and RT-PCR results showed that induction of perisotin by HG in CF was suppressed by TGF-β neutralizing antibody, Smad2/3 knockdown (Additional file 2: Fig. S2C, D), rather than PI3K or Akt inhibitors (Additional file 2: Fig. S4C, D). Besides, incubation of human TGF-β1 is sufficient to promote the expression of perisotin in CF (Additional file 2: Fig. S4E,G). Similar results were obtained in CF with Smad2/3 overexpression (Additional file 2: Fig. S4F, H and Additional file 2: Fig. S2E, F). Predictions using Jasper online databases identified three highly conserved sequences in the periostin promoter that was predicted to bind Smad2, and one highly conserved sequence in the periostin promoter that was predicted to bind Smad3 (Additional file 2: Fig. S5A–D). Interestingly, these prediction sequences were located in the activated region (− 1495 to − 1195) of periostin promoter. TGF-β1 stimulation, Smad2 overexpression, and Smad3 overexpression heightened the activity of periostin luciferase reporter gene (Additional file 2: Fig. S4I). Chromatin immunoprecipitation (ChIP) assay showed higher bindings of Smad2/3 to the periostin promoter in HG-challenged CF (Additional file 2: Fig. S4J, K), furthering confirming that the TGF-β/Smad pathway was required for HG to facilitate perisotin expression in CF.

RNA sequencing revealed that perisotin was a diver of NAP1L2 in CF

Next, we performed RNA sequencing to investigate the transcriptomic changes caused by periostin overexpression. The cluster analysis of differential gene expression level showed that 189 genes were upregulated and 146 genes were downregulated in CF after periostin overexpression (Fig. 5A). Volcano dots were used to display dysregulated genes between two groups, and all identified genes were clearly classified into two different cohorts (Fig. 5B). Radar map of differential gene expression level displayed the 30 upregulated or downregulated genes with the lowest Q or P value (Fig. 5C). We then conducted RT-PCR to examine the mRNA levels of the top 10 elevated genes, and found that periostin overexpression markedly upregulated the mRNA level of the nucleosome assembly protein 1-like 2 (NAP1L2) (Fig. 5D). In keeping with this, HG incubation boosted the protein expression of NAP1L2 (Fig. 5E). Western blotting analysis demonstrated a decreased protein level of NAP1L2 in periostin-knockdown CF under both NG and HG conditions (Fig. 5F). By contrast, upregulation of periostin not only raised the protein of NAP1L2, but also worsened HG-induced expression of NAP1L2, indicating that periostin is an upstream mediator for NAP1L2 (Fig. 5G). To further assess the functional role of NAP1L2 in CF activation, we used NAP1L2 siRNA to downregulate this protein. As expected, NAP1L2 downregulation rendered CF more invulnerable to HG-induced CF activation, as seen by lower fibrosis, oxidative stress, and inflammation (Fig. 5H–M), indicating that periostin is a positive regulator of NAP1L2, leading to CF activation in diabetes. As a class of small noncoding RNA molecules, miRNAs are established to induce the translation repression or mRNA degradation by binding to the 3’-untranslated regions (3’-UTRs) of a target mRNA sequence (9). It is interesting to know whether periostin positively regulated NAP1L2 expression in a miRNA-dependent manner. To this, end, the potential miRNAs that regulated NAP1L2 were explored using TargetScan, miRWalk, and miRDB databases. Venn diagrams are utilized to explore overlapping miRNAs, and results showed that these four databases share only one miRNA, miR-27b-3p (Additional file 2: Fig. S6A). The conserved sites for miRNA families broadly conserved among vertebrates were shown in Additional file 2: Fig. S6B. Further, we used dual-luciferase reporter gene assay to determine the interaction between miR-27b-3p and NAP1L2. Ectopic expression of miR-27b-3p significantly suppressed the luciferase activity of wild-type (WT) NAP1L2 reported gene, this was disappeared in mutant NAP1L2 reported gene (Additional file 2: Fig. S6C, D), suggesting that miR-30a-5p negatively regulated the expression of NAP1L2 in CFs. Importantly, inhibition of miR-30a-5p further potentiated, whereas upregulation of miR-30a-5p significantly reversed the actions of periostin overexpression on the protein of NAP1L2 (Additional file 2: Fig. S6E, F). These results indicated that periostin positively modulated the expression of NAP1L2 by dissociating the binding of miR-30a-5p to the 3’-UTRs of NAP1L2.

Fig. 5

RNA sequencing revealed that perisotin was a diver of NAP1L2 in CF. A Heatmap showing the differential gene cluster analysis between control heart and periostin OE hearts. B Volcano map showing the overall distribution of differentially expressed genes between control heart and periostin OE hearts. C The differential gene expression level radar graph showed the 30 up/down genes with the lowest Q or P values. D The mRNA levels of top 10 upregulated genes in mouse hearts with periostin OE. E The protein expression of in CF exposed to NG or HG using western blot. F Effects of periostin siRNA on the protein expression of NAP1L2. G Effects of periostin OE on the protein expression of NAP1L2. H–J Representative blots and quantitation of α-SMA and collagen I. K Cell viability. L MDA contents. M Representative mRNA levels of TNF-α and IL-1β. n = 4–6. *P < 0.05 versus Vehicle (Veh), Con or NG, †P < 0.05 versus HG. The P-value was calculated by unpaired two-tailed Student’s t-test (D, E). Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test (F–M)

NAP1L2 regulated SIRT3 in CFs

As is known, NAP1L2 is a histone chaperone that is involved in the dynamics of histone acetylation at its target loci during cellular differentiation [35]. The sirtuin (SIRT) deacetylases are important members that govern histone acetylation in both health and diseases [36]. We thus investigated whether NAP1L2 exerted its function through a similar epigenetic mechanism via regulating the SIRT family. The potential interactions between NAP1L2 and SIRT1-7 were detected by the co-IP, and results demonstrated that NAP1L2 could bind to SIRT1 and SIRT3, but not SIRT2, SIRT4, SIRT5, SIRT6 and SIRT7 (Additional file 2: Fig. S7A). Intriguingly, the decreased SIRT3 protein was markedly restored in HG-incubated CF upon NAP1L2 silencing (Additional file 2: Fig. S7B). On the contrary, deletion of NAP1L2 had no effect on HG-induced inhibition of SIRT1 in CF (Additional file 2: Fig. S7B). These observations indicated that perisotin upregulated NAP1L2 to recruit SIRT3 to affect CF functions in diabetes.

Perisotin/NAP1L2 blunted BCAA catabolism in CF

Branched chain amino acids (BCAAs), consisting of leucine, isoleucine, and valine, are a group of essential amino acids that can be catabolized in the heart [37]. Alterations of BCAA metabolism have been found to lead to numerous prevalent diseases, including cardiometabolic diseases [38]. Aberrant accumulation of BCAAs in the heart is a driving force for cardiac pathological remodeling and heart failure [39]. Enrichment analysis chord diagram from RNA sequencing showed that periostin overexpression affected numerous signaling pathways, including amino acid metabolism (Fig. 6A). Gene Set Enrichment Analysis (GSEA) showed that overexpression of periostin affected BCAA catabolism (Fig. 6B). Bioinformatics analysis showed that periostin significantly influenced the levels of BCAAs catabolism-related genes (Fig. 6C). Targeted metabolomic analysis further identified that serum levels of leucine, valine, isoleucine were significantly higher in diabetic mice when compared with control mice (Additional file 2: Fig. S8). To further confirm whether periostin exerted the metabolism of BCAAs, we examined the levels of serum and cardiac BCAAs in mice. In accordance with transcriptomics, serum and cardiac levels of BCAAs were dramatically elevated in diabetic mice, an observation that was counteracted by the absence of periostin (Fig. 6D). Conversely, periostin overexpression mice had higher levels of serum and cardiac BCAAs, this phenomenon was further intensified in diabetic mice (Fig. 6E). The levels of BCAAs are mediated by tissue-specific inactivation of BCAA catabolizing enzymes, including branched-chain amino transferase 2 (BCAT2), branched-chain α–keto acid dehydrogenase (BCKDHA), branched-chain α–keto acid dehydrogenase kinase (BCKDK) and mitochondrial phosphatase 2C (PP2Cm) [40]. To examine the effects of periostin on BCAA catabolism, we examined the levels of the catabolic enzymes BCAT2, BCKDHA, BCKDK, and PP2Cm in the hearts. RT-PCR revealed lower mRNA levels of BCAT2 and PP2Cm, but higher mRNA level of BCKDK in diabetic hearts than that found in normal heart, such effects were reversed by deficiency of periostin and aggravated by overexpression of periostin (Fig. 6F, G), indicating that periostin might facilitate the impairment of BCAA catabolism in diabetic hearts through regulating BCAA catabolizing enzymes. Consistent with this, overexpression of BCAT2 or PP2Cm (Additional file 2: Fig. S9A, C, D, F) obviously prevent myofibroblast differentiation of CF after periostin overexpression, as indicated by measurement of Col I (Additional file 2: Fig. S9G). Conversely, knockdown of BCAT2 or PP2Cm (Additional file 2: Fig. S9B, C, E, F) markedly prevent myofibroblast differentiation of CF in the presence of periostin overexpression (Additional file 2: Fig. S9H). These findings revealed that impaired BCAA metabolism was required for the myofibroblast phenotype, leading to DCM-induced cardiac dysfunction.

Fig. 6

Perisotin/NAP1L2 blunted BCAA catabolism in CF. A Enrichment analysis chord diagram showing the BCAA catabolism was impaired in hearts after periostin OE. B Branched-chain amino acid catabolism (R-MMU-70895).gsea. C Branched-chain amino acid catabolism (R-MMU-70895) Heatmap of the Analyzed GeneSet. D Serum and cardiac BCAA levels in mice without periostin. E Serum and cardiac BCAA levels in mice with periostin OE. F Representative mRNA levels of BCAT2, BCKDHA, BCKDK, and PP2Cm in hearts from mice without periostin. G Representative mRNA levels of BCAT2, BCKDHA, BCKDK, and PP2Cm in hearts from mice with periostin OE. H Representative blots and quantitation of BCAT2 and PP2Cm in CF after silencing NAP1L2. I Representative blots and quantitation of BCAT2 and PP2Cm in CF after treatment with ADTL-SA1215 (5 μM), an agonist of SIRT3. J Representative blots and quantitation of H3K27me2, H3K36me2, H3K79me3, H3K27ac, H3K14ac, and H3K9ac in CF upon NAP1L2 overexpression. K H3K27acoccupancy at the promoters of BCAT2 and PP2Cm in CF transfected with NAP1L2 OE plasmid by ChIP. n = 3–4. *P < 0.05 versus Con, Con siRNA or Vector, †P < 0.05 versus HG or Diabetes. The P-value was calculated by unpaired two-tailed Student’s t-test (K). Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test (D–I)

Next, we determined whether the NAF1L2/SIRT3 axis had an impact on BCAA catabolism in diabetes. As expected, the protein expression of BCAT2 and PP2Cm were higher in NAF1L2-deficient or SIRT3-activated CF than those of control cells in response to HG stimuli (Fig. 6H, I and Additional file 2: Fig. S2G), indicating that the NAF1L2/SIRT3 pathway was involved in hyperglycemia-induced dysfunction of BCAA catabolism. As mentioned earlier, NAP1L2 and SIRT3 regulated cellular functions through histone acetylation modification, we hypothesized that NAP1L2 may recruit SIRT3 to induce deacetylation of promoters around the BCAT2 and PP2Cm. We screened the global status of the most common histone modifications upon NAP1L2 overexpression, including H3K27me2, H3K36me2, H3K79me3, H3K27ac, H3K14ac, and H3K9ac. We found only H3K27ac level was remarkably diminished in comparison with the non-target control (Fig. 6J). Moreover, the results of ChIP-qPCR demonstrated that the acetylation modification of H3K27 on the promoter regions of BCAT2 and PP2Cm was strikingly inhibited by overexpression of NAP1L2 (Fig. 6K), which would induce their downregulation and subsequent BCAA deposition in the heart. Intriguingly, overexpression of NAP1L2 prominently enhanced the acetylation modification of H3K27 on the promoter regions of known fibrotic genes, such as Col I and α-SMA (Additional file 2: Fig. S10). Totally, NAP1L2 might regulate BCAA catabolism enzymes and fibrotic proteins through H3K27 acetylation chromatin remodeling in the development of DCM.

Chemical screening and identification of glucosyringic acid (GA), which directly targeted and inhibited perisotin expression in diabetic hearts

Because of the critical role of periostin in DCM, we aimed to determine whether pharmacologic suppression of periostin would be a therapeutic approach for DCM. To screen the potential inhibitors of periostin, we established the CF expressing a luciferase reporter driven by a periostin-containing promoter. From a small molecule pool containing about 349 natural products, GA was found to significantly decrease the periostin-luciferase activity (Additional file 2: Fig. S11A–C). Molecular docking analysis revealed a direct interaction between GA and periostin with binding energy of − 6.6 kJ/mol (Additional file 2: Fig. S11D). In support, surface plasmon resonance (SPR) also confirmed the potential interaction of GA with periostin (Additional file 2: Fig. S11E). GA treatment had negligible cytotoxicity in CF, even at the higher concentrations (Additional file 2: Fig. S11F), and inhibited periostin protein and mRNA levels in HG-incubate CF (Additional file 2: Fig. S11G, H). Importantly, GA treatment increases CF resistance to HG-induced CF proliferation, oxidative stress, inflammation, and fibrosis (Additional file 2: Fig. S11I–L). Next, we assessed the pharmacological potential of GA in vivo in DCM mice. Results showed that cardiac function and injury from GA-treated mice were notably restored than those from control mice (Fig. 7A–E), while GA treatment had no effect on body weight and blood biochemical indicators (Additional file 1: Table S9). Likewise, histological analysis results indicated that heart from GA-treated mice exhibited lower cardiomyocyte size, fibrosis, oxidative stress (Fig. 7F–K), which was in line with decreased mRNA levels of hypertrophic and fibrogenic genes (Fig. 7L). Moreover, GA treatment repressed the protein expression of periostin, NAP1L2, α-SMA and Col (Fig. 7M). In addition, the protein expression of BCAT2, PP2Cm and SIRT3 was unsurprisingly restored by GA treatment in diabetic mice (Additional file 2: Fig. S12A). Accordingly, treatment of GA noticeably decreased circulating and cardiac BCAA contents in DCM mice (Additional file 2: Fig. S12B-C). Both in vivo and in vitro studies suggest that the mitigation of DCM in mice by GA may be dependent on periostin suppression.

Fig. 7

Chemical screening showing that GA improved DCM by directly targeted and inhibited perisotin expression in mice. A, B Left ventricle EF and FS were quantified. C Representative echocardiographic images showing the effects of perisotin knockdown on cardiac function in control and diabetic mice. D Serum LDH levels in mice. E Serum CK-MB levels in mice. F, I Representative photographs of the myocardium with H&E staining (Scale bar = 100 μm). G, J Representative photographs of the myocardium with Sirus red staining (Scale bar = 100 μm). H, K Representative images of the myocardium with DHE staining (Scale bar = 200 μm). L The mRNA levels of pro-hypertrophic genes (ANP, BNP, β-MHC) and pro-fibrogenic genes (Col I, Col III, α-SMA). M Representative blots and quantitation of α-SMA, collagen I, and NAP1L2. n = 4–6. *P < 0.05 versus Con, †P < 0.05 versus Diabetes. Differences between groups were assessed with ANOVA followed by Bonferroni post-hoc test (D–I)