Paracrine signal emanating from stressed cardiomyocytes aggravates inflammatory microenvironment in diabetic cardiomyopathy

Summary Myocardial inflammation contributes to cardiomyopathy in diabetic patients through incompletely defined underlying mechanisms. In both human and time-course experimental samples, diabetic hearts exhibited abnormal ER, with a maladaptive shift over time in rodents. Furthermore, as a cardiac ER dysfunction model, mice with cardiac-specific p21-activated kinase 2 (PAK2) deletion exhibited heightened myocardial inflammatory response in diabetes. Mechanistically, maladaptive ER stress-induced CCAAT/enhancer-binding protein homologous protein (CHOP) is a novel transcriptional regulator of cardiac high-mobility group box-1 (HMGB1). Cardiac stress-induced release of HMGB1 facilitates M1 macrophage polarization, aggravating myocardial inflammation. Therapeutically, sequestering the extracellular HMGB1 using glycyrrhizin conferred cardioprotection through its anti-inflammatory action. Our findings also indicated that an intact cardiac ER function and protective effects of the antidiabetic drug interdependently attenuated the cardiac inflammation-induced dysfunction. Collectively, we introduce an ER stress-mediated cardiomyocyte-macrophage link, altering the macrophage response, thereby providing insight into therapeutic prospects for diabetes-associated cardiac dysfunction.


INTRODUCTION
In clinical settings, heart failure (HF) is a serious comorbidity and a highly fatal condition in patients with type 2 diabetes (T2D), predicted to account for approximately 30% of hospital admissions with cardiovascular disease (CVD) (Shah et al., 2015). T2D is now recognized to have a detrimental impact on the myocardium leading to the onset and progression of cardiac dysfunction, termed diabetic cardiomyopathy (DCM) (Isfort et al., 2014). The underlying mechanism of DCM's etiopathogenesis involves a complex interplay of intra-and intercellular signaling (Frati et al., 2017). However, the integration of these signaling processes is incompletely defined.
The incidence of cardiac dysfunction in diabetes is at least partially attributable to uncontrolled myocardial inflammation. In the diabetic heart, the first typology of cardiac inflammation is subclinical, initiated by the secretion of inflammatory cytokines, chemokines, adhesion molecules, and danger-associated molecular patterns (DAMPs) from stressed cardiomyocytes, fibroblasts, and endothelial cells (Frati et al., 2017). These inflammatory factors stimulate monocytes and lymphocytes to exaggerate the cardiac inflammatory microenvironment and contribute to cardiac pathological remodeling and dysfunction (Frieler and Mortensen, 2015;Sanders-van Wijk et al., 2020;Tan et al., 2020). Nonetheless, a translational gap remains to produce effective DCM treatments based on identifying anti-inflammatory targets.
Generally, DCM is characterized by co-manifestation of chronic ER stress and sterile inflammation in clinics (Jia et al., 2018) and pre-clinical studies (Tan et al., 2020;Wold et al., 2005). Mounting data imply that ER dysfunction drives the inflammatory milieu contributing to disease pathogenesis (Gotoh et al., 2011;Zhang and Kaufman, 2008). The molecular underpinnings of ER stress-associated inflammation are documented in We aimed to investigate the molecular link between these concurrent pathological insults, ER stress and inflammation, in the diabetic heart. Over time, experimental DCM revealed that the loss of PAK2 disrupts ER homeostasis aiding heightened myocardial inflammation and dysfunction. Mechanistically, we identified that CHOP induces direct transcriptional upregulation of high-mobility group box-1 (HMGB1), and its active release from stressed cardiomyocytes drives paracrine M1 macrophage subsets, causing an adverse inflammatory reaction in myocardium. Therapeutically, activated-PAK2 remediated ER function, and extracellular inhibition of HMGB1 represses ER stress-induced inflammation and subsequent dysfunction in diabetes. In summary, our findings establish the dysregulated ER response etiology for adverse myocardial inflammation via HMGB1, also supporting treatment strategies for preventing DCM by reducing the pro-inflammatory paracrine signal from cardiomyocytes.

Maladaptive cardiac ER stress response in DCM
To assess ER stress response in DCM, we first examined the key pathways and chaperones required for ER function in the hearts of patients with diabetes. A decrease in the ratio of spliced X-box binding protein 1 (XBP1s) versus unspliced XBP1 (XBP1u) was observed in the diabetic human heart, whereas the protein kinase R (PKR)-like ER kinase (PERK) and activating transcription factor 4 (ATF4) branch was increased (Figure 1A). CHOP, mainly regulated by PERK-ATF4 and activating transcription factor 6 (ATF6), was significantly higher in the diabetic myocardium, although ATF6 remained unchanged ( Figure 1A). Regarding the chaperones initiating the adaptive UPR, the GRP94 levels were reduced in the human heart with diabetes ( Figure 1A). Noteworthy, under diabetes, the expression and phosphorylation of PAK2, an ER stress response modulator, were both decreased in the human heart ( Figure 1A). The data elicit aberrant changes in ER stress response and its known modulator in the diabetic heart.
We then examined whether the onset and progression of DCM are associated with maladaptive cardiac ER stress in a pre-clinical model. For this, we fed 6-week-old male C57BL/6J mice with high-fat high-sucrose diet (HFHSD) for 8 to 24 weeks in a cross-sectional study. The mice displayed weight gain and higher serum cholesterol levels compared to age-matched chow-fed mice (Figures S1A and S1B). Diet-induced diabetes was verified by increased serum insulin, fasting blood glucose, and impaired glucose tolerance ( Figures  1B-1D). As a result of sustained metabolic stress, cardiac dysfunction was exemplified by a prolonged isovolumetric relaxation time (IVRT), a rising E/A ratio ( Figure 1E), and both reduced fractional shortening (FS%) and ejection fraction (EF%) ( Figure 1F), albeit unaffected cardiac conductivity (Table S1). Increased left ventricular mass (LV mass), enlarged relative wall thickness (RWT), and end-diastolic posterior wall thickness (dPW) were also observed during diabetes (Figures S1C-S1E). The transition to cardiac dilation iScience Article occurred after 24 weeks of HFHSD feeding, as indicated by reduced RWT, and enlarged left ventricular chamber (LVEsD) with an increase in Nppb levels ( Figures S1F and 1G). The deleterious heart function was associated with cardiomyocyte hypertrophy and increased fibrosis in HFHSD-fed mice (Figures S1G and S1H). The key ER-associated transcription factors were examined to establish the dysregulated ER stress response in the myocardium from diabetic mice. Interestingly, XBP1s was activated at the early stages of diabetes but was downregulated later ( Figure 1H). Cardiac ATF4 remained enriched after 12 weeks of HFHSD feeding, whereas ATF6 was increased at a later stage ( Figure 1H). Moreover, CHOP expression increased gradually in the mouse myocardium after HFHSD feeding ( Figure 1H). Consistent with the clinical findings, PAK2 activation was also enhanced in the myocardium during the early stages of diabetes, with a significant decline in expression and phosphorylation levels later ( Figure 1H). Finally, swollen ER was observed in the failing diabetic mouse heart ( Figure S2), suggesting a role for the maladaptive ER stress response in DCM development.

Myocardial inflammation in the heart from diet-induced diabetes
To investigate inflammation as a mechanism of DCM pathogenesis, the myocardial inflammatory response was assessed. Following 16 weeks of HFHSD feeding, we examined macrophages in the diabetic heart using immunofluorescence and found that Mac3 and CD68 were evident (Figures 2A and 2B). However, neutrophil (Ly6G) infiltration was not apparent in the diabetic myocardium ( Figure S3). Cytokines facilitate immune cell activation and initiation of inflammatory responses. Hence, we tested the inflammatory profiles of the myocardium at different stages of diabetes using qPCR. The major pro-inflammatory cytokines, such as Tnfa, Il1b, Ifng, Il6, and C-reactive protein (Crp), were increased after 16 weeks of HFHSD feeding, reaching a peak at a later stage ( Figure 2C). Interestingly, Hmgb1 was elevated as early as 12 weeks post HFHSD ( Figure 2C). In contrast, it was observed that the cytokines acting as anti-inflammatory factors, such as in Il10, Il15, Mmp9, and growth differentiation factor 15 (Gdf15) first peaked and eventually decreased in the diabetic mouse heart ( Figure 2C), while Tgfb and Il11 displayed a continuous and robust increase until 24 weeks ( Figure 2C). Besides, upregulation of the chemokines triggering macrophages activation, such as monocyte chemoattractant protein-1 (Mcp1/Ccl2), Ccl5, and Ccl24, were present in the myocardium upon 16-24 weeks of HFHSD ( Figure 2C), indicating that the initial rise in cytokine levels likely originates from the myocardium before macrophage appearance.
The macrophage-facilitated inflammatory response is composed of an elaborate equilibrium of pro-and anti-inflammatory signatures (Bajpai and Tilley, 2018). Furthermore, the prototypical markers distinguishing M1 and M2 subsets were measured to assess macrophage phenotypes in diabetic hearts. Notably, nitric oxide synthase 2 (Nos2), a key M1 subtype determinant, was upregulated in the mouse myocardium after 16 weeks of HFHSD feeding, with an even higher expression at 24 weeks ( Figure 2C). Conversely, the genes characterizing M2 macrophages, including mannose receptor 2 (Mrc2), arginase 1 (Arg1), resistin-like alpha (Retnla), beta-N-acetylhexosaminidase (Ym1), and galactose N-acetyl-galactosamine-specific lectin 2 (Mgl2), were enriched in the mouse myocardium at 12 weeks of HFHSD feeding, but were reduced at the later stages ( Figure 2C). Thus, we infer that the development of DCM is accompanied by notable inflammation with a predominance of M1 macrophages in the myocardium.

Malfunctioned cardiac ER prompts myocardial inflammation in diabetes
As observed in heart, under the stimulation of high fatty acid and glucose (HFHG), an in vitro model displayed an oscillating pattern in PAK2 phosphorylation at different time points, coinciding with the transition from an adaptive UPR to a maladaptive ER stress response ( Figures S4A and S4B). PAK2 activation was likely through both Cdc42 and Rac1, which are the putative activators of the PAKs family (Renkema et al., 2002), as its phosphorylation was diminished in both Cdc42-and Rac1-deficient cells ( Figures S4C and S4D). Furthermore, we utilized Pak2-deficient and -overexpressing H9C2 cells with HFHG stimulation to assess whether PAK2 is required for ER housekeeping under diabetic conditions. Pak2-knockdown cells exhibited an aberrant ER function, which was determined by increased CHOP along with dampened XBP1s levels ( Figure S5A). Conversely, the overexpression of PAK2 rescued XBP1s and suppressed CHOP under HFHG ( Figure S5B). These observations support cardiac PAK2 regulation of ER stress pathways under diabetic stress.
A negative correlation of myocardial Nppb and PAK2 in diabetic mice ( Figure 3A)  iScience Article increased comparably ( Figure S6A), and their metabolic profiles were examined by measuring serum cholesterol ( Figure S6B), insulin ( Figure S6C), glucose levels ( Figure S6D), and glucose tolerance test (Figure S6E), which demonstrated a comparable diabetic condition in both HFHSD-fed groups. This indicated that cardiac PAK2 loss does not affect the diet-impaired systemic metabolic changes. Cardiac function was assessed by echocardiography and conscious ECG (Tables S2 and S3). Diastolic dysfunction occurred in Pak2 cKO at an early stage of diabetes, as reflected by prolonged IVRT and increased E/A ratio 8-12 weeks post HFHSD feeding ( Figure 3B). JT intervals, indicating ventricular repolarization, were longer after 16 weeks of HFHSD feeding ( Figure 3C and Table S3), also supporting compromised diastolic abnormality. In other respect, systolic dysfunction was observed in Pak2 cKO mice fed with HFHSD, which was evidenced by reduced FS% and EF% ( Figure 3D). Additionally, Pak2 cKO hearts displayed ventricular dilation, as corroborated by increased end-diastolic left ventricular internal diameter (LVEdD), and reduced RWT and dPW thickness ( Figures 3E-3G). Cardiac remodeling in Pak2 cKO hearts was evident, as indicated by increased myocardial Nppb, cardiomyocyte size, and fibrosis ( Figures S7A-S7C). Furthermore, maladaptive ER stress iScience Article response was exacerbated in the Pak2 cKO myocardium after feeding with HFHSD, determined by higher CHOP expression ( Figure 3H). Collectively, these results further substantiate the apparent importance of cardiac PAK2 in regulating ER stress response under diabetic stress.
Cumulative evidence demonstrates a crosstalk between ER stress and inflammation; however, whether the cardiac ER stress response is a cause or a consequence of inflammation in the heart remains elusive. Therefore, we elucidated the connection between the maladaptive ER stress and myocardial inflammation. Pak2-cKO hearts displayed severe inflammation following 16 weeks of HFHSD feeding, exemplified by a greater number of Mac3-labelled and CD68-positive non-cardiomyocytes in the myocardium ( Figures 3I, 3J, and S8A). Notably, the abundant localization of CD68 macrophages adjacent to cardiomyocytes suggested a possible paracrine communication ( Figures 3K and S8B). More convincingly, the assessment of cytokines in Pak2 fl/fl and Pak2 cKO myocardium showed that Il1b, Crp, Mcp1, Ccl24, and Hmgb1 were higher in Pak2-cKO ; on the contrary, Il10, Mmp9, and Gdf15 were lower in Pak2 cKO post HFHSD ( Figure 3L). In addition, increased Nos2 expression indicated elevated M1 macrophages in Pak2 cKO hearts, while Retnla and Ym1 were scarce ( Figure 3L). As such, the above findings illuminate the previously overlooked effects of cardiac ER function on myocardial inflammation in metabolic disorder.

Cardiac-sourced HMGB1 contributes to myocardial inflammation in diabetes
Intracellular HMGB1 interacts with nucleosomes and transcription factors to facilitate gene regulation, whereas extracellular HMGB1 secreted from stressed cells, functions as a pro-inflammatory molecule (Chikhirzhina et al., 2020). Noticeably, myocardial Hmgb1 was positively correlated with Nppb, supporting its detrimental effects on cardiac function ( Figure 4A). In addition, lysine acetylation of HMGB1 is known to be vital for its active secretion from immune cells (Chikhirzhina et al., 2020). Of note, the elevated cardiac Hmgb1 was observed in C57BL/6J mice as early as 12 weeks after HFHSD feeding ( Figure 2C), its protein levels increased accordingly ( Figure 4B). Interestingly, acetylated HMGB1 in the myocardium ( Figure 4B) with concomitant increased HMGB1 levels in serum ( Figure 4C) were observed in diabetic mice. More importantly, consistent observations of increased HMGB1 were made in the myocardium from human diabetic subjects with CVD and serum from patients with diabetes ( Figure 4D). Further confirmation was provided by cultured human heart slices after HFHG stimulation (Ou et al., 2019), exhibiting more evident HMGB1 and its lysine-acetylated form following HFHG stress ( Figure 4E), with a consequent increase in its secretion to the conditional medium ( Figure 4E). Similar results were also observed in isolated rat cardiomyocytes following exposure to HFHG ( Figure 4F). Taken together, these results reveal that hyperglycemia and hyperlipidemia induce cardiac-sourced HMGB1 upregulation and secretion in diabetes.
Prompted by the observations that an increased HMGB1 in the myocardium occurred prior to macrophage activation (Figures 2, 4B, and 4C), we speculated that HMGB1 actively secreted from stressed cardiomyocytes is crucial in shaping macrophage phenotype in diabetes. Hence, we explored whether and how cardiac-sourced HMGB1 exerts its pro-inflammatory action in the myocardium. Macrophages have the capability to directly recognize DAMPs and cytokines to provoke chronic inflammation in metabolic disorders (McNelis and Olefsky, 2014). We therefore obtained evidence of the interplay between cardiac HMGB1 and myocardial inflammatory reactions under diabetes. BM-derived macrophages (BMM) were isolated from mice fed with chow and HFHSD, following differentiation and stimulation, F4/80 + selected macrophages were classified as M1 and M2 subsets by CD86 + and CD206 + markers, respectively (Figures 4G and S9A). First, M1/M2 ratio increased following HFHSD feeding, indicating that the macrophage proinflammatory nature timed with increasing expression of myocardial HMGB1 ( Figures 4H and S9B). Secondly, upon exposure to either recombinant HMGB1 (rHMGB1) or combination of rHMGB1 and HFHG, iScience Article BMM isolated from mice at 12 and 16 weeks post HFHSD manifested significant polarization to M1 subtype ( Figures 4I and S9C). Afterwards, we focused our efforts on the relevance of HMGB1 to the myocardial inflammatory response. Differentiated BMM from non-diabetic mice were treated with distinct stimulators of M1 and M2 macrophage subtypes. As expected, M1 macrophage activation by IFNg was heightened in combination with lipopolysaccharides (LPS). Importantly, rHMGB1 regardless of HFHG also increased M1 polarization of IFNg-primed macrophages (Figures 4J and S9D). In comparison to IL4-triggered M2 macrophage subtype, rHMGB1 impeded the M2 phenotype ( Figures 4J and S9D). The H9C2-conditioned medium from long-term HFHG stimulation induced M1 polarization; however, glycyrrhizin reduced this polarization, when used as an adjuvant to sequester secreted HMGB1 ( Figure 4K). Combined, these data suggest that the diabetic state instigates cardiac upregulation and secretion of HMGB1, further supporting the notion of HMGB1-driven macrophage polarization toward M1 phenotype.

ER dysfunction-induced cardiac HMGB1 in diabetes
Of note, the negative correlation between cardiac Hmgb1 and PAK2 ( Figure 5A) warranted further investigation on how ER dysfunction induces myocardial HMGB1 under diabetic stress. Interestingly, cardiac Pak2 ablation accelerated increase of HMGB1 at both transcript ( Figure 3L) and protein level ( Figure 5B) post HFHSD feeding. Likewise, Pak2 depletion gave rise to a corresponding increase in intracellular and extracellular-acetylated HMGB1 ( Figures 5B and 5C). We then asked how ER dysfunction induces HMGB1 gene upregulation in response to diabetic stress. CHOP plays a multifunctional role in the transcriptional regulation of genes (Nishitoh, 2012). It was increased in the HFHG-stimulated human heart slices and isolated rat cardiomyocytes (Figures S10A and S10B). Similarly, CHOP expression and nuclear localization were more pronounced in the Pak2 cKO myocardium in diabetes ( Figures 3H and 5D). In addition, higher nuclear CHOP and concurrent shuttling of HMGB1 from the nucleus to the cytosol was detected in cells following 12-h HFHG stimulation ( Figure 5E). Subsequently, we investigated whether CHOP transcriptionally regulates HMGB1. In silico screening of Hmgb1 promoter region detected multiple potential binding sites of CHOP, which are highly conserved across mammalian species ( Figure 5F). CHOP transcriptional regulation of HMGB1 was corroborated by chromatin immunoprecipitation (ChIP)-qPCR assay. CHOP overexpression led to a notable increase in the occupancy of CHOP to Hmgb1 promoter regions containing its consensus binding sequence (À127 to À202 and À4366 from TSS) ( Figure 5F). To validate cardiac stress-mediated inflammation, conditional medium from stimulated H9C2 was used to culture BMM ( Figure 5G). Reinforcing ChIP findings, CHOP overexpression in H9C2 exhibited higher HMGB1 and its acetylated form ( Figure 5H), leading to its subsequent secretion into the culture medium ( Figure 5I). Next, stimulating mouse BMM with H9C2 culturing medium from Chop-overexpressing cells promoted M1 macrophage phenotype, confirming CHOP's role in HMGB1-mediated inflammation ( Figure 5J). More supportively, mirroring the diabetic Pak2 cKO mice, Pak2 abrogation in H9C2s increased CHOP accompanied by higher HMGB1 and its lysine-acetylated form upon HFHG ( Figure 5K), with an elevated HMGB1 secretion ( Figure 5L). Consequently, the conditioned medium from Pak2-knockdown H9C2s followed by stimulation of HFHG amplified M1/M2 ratio of BMM ( Figure 5M). Consistently, M1 macrophage phenotype was repressed with conditioned medium from Pak2-overexpressing H9C2s followed by HFHG stimulation ( Figure 5N). In summary, these data unmasked that cardiac HMGB1 expression and secretion is upregulated by means of exaggerated ER stress response-increased CHOP under diabetic conditions.

PAK2-mediated functional ER is a requisite for vildagliptin's action on myocardial inflammation under diabetic conditions
Given that diet-induced diabetic Pak2 cKO myocardium exhibited impaired cardiac ER function with a severe inflammatory response, we hypothesized that manipulating ER function has therapeutic potential for preventing myocardial inflammation in diabetes. We initially examined whether the current anti-diabetic medications . Continued (I-K) (I) 3D bar plot representing mean %CD206 + and %CD86 + macrophages from HFHSD-fed mice, stimulated with rHMGB1 (500 ng/mL) with and without HFHG (boxed, mean M1/M2 ratio, AU) (N = 3/group). %CD206 + and %CD86 + macrophages from chow-fed mice following (J) IL4 (20 ng/mL), IFNg (50 ng/mL), lipopolysaccharide (LPS) (100 ng/mL), rHMGB1 (500 ng/mL) with and without HFHG stimulation, and (K) culture medium of HFHG stimulated H9C2 with and without glycyrrhizin (Gly). Boxed: Data showing mean M1/M2 ratio (AU) from each experiment (normalized to control, AU). All IP experiments (N = 1/group/cohort). Each data point represents one animal/experiment. All data are presented as mean G SEM. iScience Article regulate ER pathway, including empagliflozin (sodium-glucose co-transporter 2 inhibitor), semaglutide (glucagon-like peptide-1 receptor agonists), and vildagliptin (dipeptidyl peptidase-4 inhibitor). Among them, vildagliptin retained PAK2 activation and lessened HFHG-induced CHOP during short-term stress, while the others failed to function on managing ER stress via PAK2 ( Figure S11A). The action of vildagliptin on ER normality was further confirmed in isolated cardiomyocytes and myocardium (Figures 6A, S11B, and S11C). It has also been demonstrated that vildagliptin exhibits an anti-inflammatory potential in diabetes (Kagal et al., 2017). Moreover, vildagliptin has been reported to exert cellular protective roles by alleviating aggravated ER stress in hepatocytes and b-cells (Ma et al., 2018a;Shimizu et al., 2012). Therefore, we acquired functional evidence on whether vildagliptin restricts myocardial inflammation by mitigating ER stress in diabetes. The cardiac ER dysfunction mice (Pak2 cKO mice) were fed with HFHSD for 8 weeks, followed by vildagliptin treatment ( Figure 6B). Here, we observed that moderate vildagliptin administration did not affect the BW or glucose tolerance ( Figures 6C and 6D). Also, it had no significant impacts on cardiac contractility in Pak2 cKO hearts ( Figures 6E and 6F; Table S4). It is noteworthy that even though vildagliptin leads to an increase in left ventricular volumes in pre-clinical and clinical studies, it has no major effects on cardiac contractility (McMurray et al., 2018). Considering that cardiac PAK2 was abolished during diabetes development, we postulated that restored PAK2 serves as a molecular basis for supporting vildagliptin-induced ER regulation. Therefore, a combination of vildagliptin with PAK2 restoration using a gene delivery system was subjected to Pak2 cKO mice fed with HFHSD. Strikingly, following cardiac PAK2 restitution, vildagliptin treatment attenuated cardiac dysfunction post HFHSD, as demonstrated by reduced IVRT and reserved contractility compared to solely vildagliptin-treated mice (Figures 6E and 6F; Table S4). Pathological remodeling was also diminished by treatment of vildagliptin in the presence of PAK2, whereas standalone vildagliptin treatment gave rise to only a minor improvement ( Figure S12).
More importantly, the elevated macrophages in the myocardium, identified by CD68 and Mac3, were significantly reduced by vildagliptin only in the attendance of cardiac PAK2 ( Figure 6G). Further analyses of myocardial inflammatory profiles revealed that the pro-inflammatory cytokines, including Tnfa, Il1b, Ifng, Il6, and Crp, were significantly impeded by vildagliptin combined with cardiac PAK2 restoration (Figure 6H). Also, the dominant M1 macrophage subtype was lowered, which was evidenced by decreased Nos2 and Hmgb1 ( Figure 6H). On the contrary, the markers of M2 macrophage subtype, such as Il10, Gdf15, and Retnla, were further enhanced particularly by the combined treatment ( Figure 6H). Because macrophage polarization and inflammatory response are likely attributable to cardiac ER stress-induced HMGB1, we finally examined whether PAK2 existence is required for vildagliptin's anti-inflammatory function. Vildagliptin was able to considerably suppress the levels of cardiac CHOP and HMGB1 in the presence of PAK2 ( Figure 6I). Collectively, these results unveil that regulated ER stress in cardiomyocytes is necessary for the net beneficial effects of vildagliptin on myocardial inflammation.

Blocking extracellular HMGB1 decelerates deleterious effects of myocardial inflammation on DCM development
Extracellular HMGB1 has been considered as a potential therapeutic target in inflammatory diseases (Andersson et al., 2018). Finally, we ascertained functional evidence whether targeting HMGB1 has beneficial effects on diabetes-related cardiac dysfunction through anti-inflammatory activity. Glycyrrhizin is a natural compound that binds HMGB1 and inhibits its cytokine activities (Deutch et al., 2019). To seek its effects on DCM, C57BL/6J mice were fed with HFHSD for 16 weeks, the time point at which cardiac HMGB1 secretion

OPEN ACCESS
rose, followed by glycyrrhizin administration in drinking water for further 8 weeks ( Figure 7A). Although blockage of extracellular HMGB1 did not affect the systemic metabolic profiles, evidenced by hyperglycemia and glucose intolerance ( Figure 7B), glycyrrhizin rescued cardiac function (Figures 7C and 7D; Table S5) and ameliorated cardiac pathological remodeling with reduced Nppb levels ( Figures 7E, 7F, and S13). We further appraised whether prevention of DCM development is, at least partly, through abrogating inflammatory signatures in the myocardium. Macrophage prevalence in the myocardium was reduced by glycyrrhizin as confirmed by Mac3 and CD68 immunofluorescence staining ( Figure 7G). Most importantly, examination of the inflammatory cytokines in the myocardium confirmed that pro-inflammation was diminished, shown with a decrease in Tnfa, Il1b, Il6, Crp, and Ccl2 ( Figure 7H). In addition, the genes featuring the M1 macrophage phenotype, such as Nos2, were decreased, while M2 macrophage markers, including Il10, Gdf15, and Retnla, were maintained in the myocardium from mice with glycyrrhizin administration ( Figure 7H). In a further demonstration, macrophages were isolated from the heart and sorted as CD45 + Ly6G À F4/80 + cells ( Figure S14A). The percentage of macrophages that originated from either the M1 or M2 sub-population were labeled CD86 + or CD206 + , respectively. M1 macrophages, as the major type of macrophages observed in the diabetic heart, were constrained after antagonizing copious HMGB1 using glycyrrhizin ( Figure 7I). Finally, the whole heart extracts were examined with a Mouse cytokine array assay of 96 factors, which revealed a decrease in CD40, CTACK, CXCL1, MIP1g, and sTNFRI, also supporting that glycyrrhizin halts myocardial inflammation in diabetes ( Figure S14B). Taken together, these outcomes demonstrate that neutralizing HMGB1 hampers myocardial inflammation, which consequently slows down the progression of DCM and restores cardiac capacity to withstand chronic metabolic stress.

DISCUSSION
This study highlights the therapeutic notion of regulating cardiac ER stress response-associated inflammation in DCM. We dissect that PAK2 is a regulator of ER stress response in diabetes with an increase in CHOP because of the dysregulated ER response. Acting as a transcription factor, CHOP directly upregulates HMGB1, subsequently resulting in its active secretion from cardiomyocytes. Myocardial inflammation is further aggravated from cardiomyocyte-sourced HMGB1 activation of pro-inflammatory macrophages. In addition, as functional evidence, we showed that un-mitigated cardiac ER response due to PAK2 loss under diabetes may account as a barrier for leveraging the anti-inflammatory potential of vildagliptin. Also, sequestering extracellular HMGB1 restricts the excessive inflammatory reaction and decelerates DCM development.

Cardiac Pak2 loss induces maladaptive ER stress in response to metabolic stress
The overactivated PERK and ATF6 branches of the UPR signaling cascade have been implicated in DCM development in rodent hearts (Lakshmanan et al., 2013). On the contrary, XBP1s is reported to be protective in type 1 diabetic mice hearts (Wang et al., 2021) and metabolic stress-induced HF with preserved EF model (HFpEF). The latter model driven by high-fat diet in combination with endothelial dysfunction-based hypertensive stress (Schiattarella et al., 2019) displayed a decline in myocardial XBP1s associated with reduced diastolic cardiac performance. This is consistent with our observations that ER dysfunction is evident with decreased XBP1s and increased XBP1u in the failing heart from diabetic patients and HFHSD-fed C57BL/6J mice. Moreover, cardiomyocyte-restricted overexpression of XBP1s rescued the HFpEF phenotype (Schiattarella et al., 2021), highlighting the protective role of adaptive UPR. As illustrated here, dysregulated ER response following PAK2 ablation is associated with early diastolic dysfunction and late decreased contractility following metabolic stress. We previously discovered that cardiac PAK2-mediated XBP1s restores ER homeostasis in response to pressure overload stress (Binder et al., 2019). The increase in ATF4 and ATF6 levels along with a reduction in PAK2 in diabetic myocardium from mice and humans, suggests that inhibition of XBP1s results in an exacerbated ATF4-and ATF6-mediated CHOP upregulation. Given the plethora of DCM instigators and temporal nature of the ER stress response, there is potential therapeutic significance in re-establishing ER homeostasis.

ER stress directly triggers inflammatory response
Under pathological conditions, prolonged ER stress, as in diabetes, induces CHOP-dependent maladaptive effects (Doroudgar and Glembotski, 2013;Minamino et al., 2010). Accumulating data demonstrate a iScience Article clear connection between ER stress and inflammation in a vicious cycle contributing to disease pathogenesis (Zhang and Kaufman, 2008). For instance, IRE1 and PERK branches activate the nuclear factor kappa B upregulation of inflammatory genes (Garg et al., 2012). In particular, CHOP facilitates the expression of pro-inflammatory cytokines, such as IL1b, IL8, and IL23, in either immune cells or non-immune cells (Endo et al., 2006;Goodall et al., 2010;Park et al., 2010). Consequently, the ER stress-induced inflammatory reaction further propagates underlying pathological outcomes. The current dogma regarding CHOP in cardiomyocytes is aligned with its transcriptional action on apoptotic genes facilitating cardiac apoptosis, pathological remodeling, and heart failure (Fu et al., 2010;Okada et al., 2004). Of importance, myocardial inflammation following ischemia-reperfusion injury is reduced in CHOP-KO mice (Miyazaki et al., 2011). Additionally, in response to LPS, IL6 failed to increase in CHOP-deficient cardiomyocytes (Miyazaki et al., 2011). Our study affirms that the detrimental ER stress response gave rise to increased CHOP followed by aggravated inflammatory responses, which highlights cardiac ER stress as a contributing factor to inflammation in the stressed myocardium.

Phenotype of macrophages affects cardiac function
Macrophage phenotypes may contribute to myocardium homeostasis or disease. Myocardial macrophages have two origins: cardiac-resident macrophages and infiltrating circulatory monocytes (Ma et al., 2018b). Given the increase in macrophages in the diabetic heart, the fact that the BMMs were activated in the conditional medium from cultured cardiomyocytes following HFHG, potentially implies the activation of blood-borne macrophages in vivo. Moreover, it is reported that hyperglycemia alters the BMM phenotype and imposes a high CVD risk (Edgar et al., 2021). Interestingly, the BMMs from diabetic mice retained these pro-inflammatory and -atherogenic characteristics, even when cultured under physiological glucose levels. Similar pro-inflammatory phenotype was observed in the BMMs from HFHSD-fed mice under basal state. Therefore, it is hypothetically suggested that the increased susceptibility of BM-sourced macrophages to M1 polarization participates in the exacerbated myocardial inflammation observed in the experimental DCM models. A concept is emerging whereby manipulating macrophage polarization is key to developing macrophage-mediated therapeutics. Clinical studies have demonstrated that circulating monocytes isolated from diabetic patients are vulnerable to activation into pro-inflammatory M1 subtype (Nikiforov et al., 2017), which is in line with our observation that BMMs from diabetic mice were more susceptible to polarization to M1 macrophages. Moreover, conditional medium from CHOP-overexpressed cells drove the induction to M1 phenotype, akin to the observation that CHOP facilitates CCL11and CCL24-mediated M1 polarization in adipose tissue (Suzuki et al., 2017). M1 macrophages are known to produce cytokines, such as TNFa, IL1b, IL6, CCL2 (MCP1), and CCL5, which were upregulated in the diabetic myocardium in our study. It is speculated that some cytokines (e.g. IL1b and IL6) were originated from cardiomyocytes at an early diabetic stage, followed by an even greater increase in cytokine levels due to both stressed cardiomyocytes and activated M1 macrophages. Noticeably, signatures of M2 macrophages (Arg1, Retnla, and Ym1) and their featured cytokines (IL10 and IL15) were predominant at an early stage, indicating that the M2 macrophages-mediated anti-inflammatory response was declined upon longterm diabetic stress. On the other hand, the pro-fibrotic role of M2 macrophages was presented by increased Tgfb along with decreased Mmp9 (Braga et al., 2015), as a consequence of increased fibrosis in the diabetic heart. Therefore, these results support the previous hypotheses that an imbalance in macrophage polarization is associated with DCM development (Bajpai and Tilley, 2018).

Regulation of cardiac intracellular HMGB1
HMGB1 is a non-histone chromosome-binding protein which participates in gene regulation; however, it acts as a DAMP promoting inflammation once released from inflammatory and stressed cells (Magna and Pisetsky, 2014). Although cardiac nuclear HMGB1 has a protective role in preventing pathological remodeling (Funayama et al., 2013), exogenous HMGB1 enhances pressure overload-induced cardiac hypertrophy (Zhang et al., 2016(Zhang et al., , 2019 iScience Article is decreased in the failing heart, whereas its cytoplasmic level is enhanced in response to neurohumoral stress (Funayama et al., 2013). Consistent with this, we detected HMGB1 translocation from nuclei to cytosol upon the change from compensated to decompensated state of diabetic stress in cardiomyocytes. It is worth recognizing that the molecular mechanisms that control HMGB1 expression in cardiomyocytes under pathological stress would be first described by this study. Evidently, forced expression of CHOP upregulated HMGB1 expression in cardiomyocytes. More convincingly, ChIP assay determined the direct binding of CHOP to the proximal promoter regions of HMGB1. Likewise, CHOP-regulated HMGB1 was also demonstrated in renal tubular cells (Zhang et al., 2015). Hence, HMGB1 is a novel target transcriptionally regulated by CHOP, which could bridge the knowledge gap regarding ER dysfunction-mediated cardiac inflammation; however, whether CHOP is a feasible target for preventing inflammation in the myocardium needs to be evaluated.

HMGB1 is secreted from stressed myocardium
With regards to HMGB1 secretion, it occurs via ER-independent secretory pathways and depends on its post-translational modifications such as acetylation and its redox state, specifically disulfide bonds (Andersson et al., 2021;Kwak et al., 2020). In immune cells, its cytoplasmic translocation and subsequent release require acetylation in the nuclear localization sequence (Magna and Pisetsky, 2014;Ramadan et al., 2017) via Janus kinases/signal transducer and activator of transcription-1 or poly(ADP-ribose) polymerase-1 pathway (Lu et al., 2014;Yang et al., 2014). Of note, we found that myocardial HMGB1 was upregulated prior to macrophage appearance, indicating its possible source from non-immune cells. Its increase was also detected earlier than marked cardiomyocyte death. HMGB1 levels in the serum and the myocardium are higher in patients with T2D than in non-diabetic subjects (Volz et al., 2010;Zhao et al., 2013). In agreement with this fact, we demonstrate that the secretory form of HMGB1 was rising in the diabetic heart from both humans and mice. Such an increase was also observed due to CHOP overexpression, suggesting that HMGB1 can be secreted actively from stressed cardiomyocytes. However, the molecular basis of how CHOP mediates cardiac HMGB1 acetylation and translocation will be further investigated.

Extracellular HMGB1 can be targeted for dampening of myocardial inflammation
Because secreted HMGB1 recruits and activates pro-inflammatory macrophages through TLR4 or the receptor for advanced glycation end products (RAGE) (Volz et al., 2010), we found an increased ratio of M1/M2 induced by rHMGB1 under diabetic stress. Moreover, induction of phenotypic switching to M1 macrophage polarization via HMGB1-TLR2/TLR4 signaling cascade has been previously associated with age-related cardiac dysfunction (Karuppagounder et al., 2016) and autoimmune myocarditis development (Su et al., 2016). Therefore, repression of extracellular HMGB1 appears to be an attractive treatment approach for decelerating myocardial inflammation without any negative effects on cardiomyocytes (Palomer et al., 2018). Previous studies illustrated that glycyrrhizin displays cardioprotective actions (Bangert et al., 2016;Cai et al., 2017). More importantly, glycyrrhizin also ameliorates insulin resistance, hyperglycemia, and hyperlipidemia in obesity and diabetes (Alizadeh et al., 2018;Sen et al., 2011). Here, we show that diabetes-associated myocardial inflammation, as assessed by cytokine levels and markers of macrophage subtypes, was attenuated by sustained administration of glycyrrhizin. These findings also explicate the previous finding that glycyrrhizin alleviates cardiac remodeling in the diabetic rat (Thakur et al., 2021). However, time will tell whether its beneficial effects can persist with meticulous investigation.

Synergistic effects of antidiabetic drug and cardiac Pak2 maintenance in DCM
Finally, vildagliptin, a dipeptidyl peptidase-4 inhibitor, maintains incretin action in producing more insulin.
In clinical use, a combination of vildagliptin and metformin has been approved as the first-line treatment for T2D (Das et al., 2020). Although pre-clinical research demonstrates that vildagliptin attenuates cardiac dysfunction (Li et al., 2021;Rankovic et al., 2021), clinical studies did not show compelling evidence of any beneficial cardiac effects (Das et al., 2020). Moreover, vildagliptin is capable of inhibiting free-fatty acidinduced HMGB1 release from endothelial cells (Qi et al., 2019). Here, we show that vildagliptin only participated in PAK2 phosphorylation with no effect on its expression. In consequence, vildagliptin minimized the ER dysfunction-induced CHOP expression solely with preserved PAK2 levels in the heart. Pre-eminently, we used a moderate dose of vildagliptin to obtain functional evidence for its impacts on cardiac inflammation. We observed that the cardiac PAK2 deficiency-associated maladaptive ER response restricted vildagliptin's inhibitory effect on CHOP-upregulated HMGB1 expression and thereby myocardial inflammation. In contrast, PAK2 maintenance was required to restore an intact ER response following vildagliptin treatment, in turn suppressing CHOP. As a consequence, cardiac HMGB1-amplified inflammation was diminished ll OPEN ACCESS iScience 25, 103973, March 18, 2022 iScience Article under diabetic stress. Prompted by the decline in cardiac PAK2 in long-term diabetes, our data provide a hypothetical explanation for clinical outcomes where vildagliptin has a limited effect on cardiac function due to its futile effect on PAK2 activation. Furthermore, our study suggests that ER function regulators as an adjunct to clinical antidiabetic interventions may have synergistic effects on both myocardial inflammation and DCM progression.

CONCLUSION
In summary, we identify a new cardiomyocyte-macrophage communication in a cardiac ER dysfunctiondependent manner, altering the macrophage response in the myocardium. Specifically, PAK2 loss disrupted ER homeostasis and led to upregulation of cardiac-sourced HMGB1 via CHOP, profoundly affecting myocardial inflammation under metabolic stress. According to the results reported here, maintaining cardiac ER function could be a potential option, particularly in conjunction with antidiabetic medications, for limiting cardiomyocyte-related paracrine inflammation. Hence, future strategies aimed at the maintenance of cardiac ER function might be a feasible therapeutic approach to mitigate inflammation, thus alleviating cardiac dysfunction in patients with diabetes.

Limitations of study
Our findings indicate that HMGB1 derived from stressed-cardiomyocytes induces pro-inflammatory effects on polarization of macrophages. However, at present, we lack direct evidence of the source of HMGB1 in serum during different time durations of metabolic stress, although in vitro findings established that cardiomyocytes are one source. Thus, the possible effects of cardiomyocyte-derived HMGB1 on non-myocytes during DCM might also be explored. Moreover, it can be useful to distinguish whether the activated macrophages in the diabetic myocardium are systemic or resident. Further investigations are required to understand the mechanisms of the interactions between cardiomyocytes and macrophages. In addition, although data from cells have clearly shown that CHOP overexpression upregulates HMGB1 expression, hyperacetylation, and subsequent secretion, genetically modified models with CHOP overexpression or knockdown would be better to affirm our findings. Additionally, we should also investigate the in-depth molecular mechanisms of HMGB1 intracellular translocation, specific acetylation sites, and secretion pathways from cardiomyocytes. Also, our study quantified hyperacetylated isoform of HMGB1 in human and mouse serum samples; it is worthwhile investigating the other isoforms of HMGB1, such as redox state, for a better understanding on the role of extracellular HMGB1. Metabolic stress induces HF with reduced and/or preserved EF. In this study, cardiomyocyte-specific PAK2 knockout model exhibited HF with reduced EF under such a stress. As a result, other ER dysfunction models and large animal models with metabolic disorders, including HF with preserved EF, could be utilized to prove the molecular basis for the paracrine effects of stressed cardiomyocytes on myocardial inflammation, thereby enhancing the therapeutic potential of the current findings.
Suzuki, T., Gao, J., Ishigaki, Y., Kondo, K., Sawada, S., Izumi, T., Uno, K., Kaneko, K., Tsukita, S., Takahashi, K., et al. (2017) (Radu et al., 2015). Cardiomyocyte-specific Pak2 knockout mice (Pak2 cKO ) were bred by mating Pak2 fl/fl mice with mice expressing Cre under a-myosin heavy chain (aMHC) promoter. Pak2 deletion in cardiomyocytes was verified by PCR of genomic DNA using (5 0 -TGAAGCTGCATCAATCTATTCTG-3 0 ) and (5 0 -TGAAGCTGCATCAATCTATTCTG-3 0 ) primers. Mice were backcrossed into C57BL/6J background for 8 generations. Pak2 cKO and their littermates (Pak2 fl/fl ) were used in this study. When weaned, mice were randomly split into housing units containing three to six mice, blinded by the genotype. Each was considered an experimental unit and assigned feeding and treatment combinations. Each mouse was counted as a biological replicate. All in vivo studies were blinded for genotype and treatment during measurement and analysis. No exclusions were made. Mice were euthanized by cervical dislocation while sedated.

Human heart slices and culture
The cultured heart protein extracts were a kind gift from Dr. Tamer M.A. Mohamed (University of Louisville). Fresh human heart was provided from a consented and deidentified donor through the USA transplantation network 'Novabiosis'. The protocol was approved by the IRB committee at the University of Louisville as a non-human subject research. The donor was aged 37 years old with no CVD history. The harvest and slicing of the heart are described previously (Ou et al., 2019). Briefly, freshly sliced heart was maintained in the culture medium overnight, followed by HFHG (500 mM PA in 0.5% BSA; 6 g/L glucose) stimulation for 16 hs.
Murine BM-derived macrophages culture BM progenitors were isolated from male chow-fed 6-week-old C57BL/6J mice or mice fed HFHSD for 12 and 16 weeks. Following cervical dislocation, both hind legs were dissected and harvested. Tibia and femur were cleaned for excess tissue and rinsed in PBS. BM isolation was carried out with DMEM (11966-025, Gibco) supplemented with 1 mM sodium pyruvate using a 27G needle and 10 cc syringe. The flushedout bone marrow was passed through a 19G needle to break down cell aggregates, filtered through a 70 mm strainer and centrifuged at 400 g for 10 mins. BM progenitors were resuspended in DMEM supplemented with 10% FBS, 1 mM sodium pyruvate and 20% L929 (CCL-1, ATCC, Middlesex, UK) conditioned media, containing macrophage-CSF. The resuspension was plated onto 100 mm sterile bacterial dishes at 14 million cells/dish and differentiated at 37 C and 5% (v/v) CO 2 for 6 days, with additional feeding ll OPEN ACCESS iScience Article QUANTIFICATION AND STATISTICAL ANALYSIS

Histological imaging and analysis
Masson's, H&E-stained and DAB-developed sample images were acquired on a 3D-Histech Pannoramic-250 microscope slide-scanner using a Zeiss 20x/0.80 Plan Apochromat objective. Snapshots of the slidescans were taken using the Case Viewer software (3D-Histech) with whole-heart images for Masson's and section images at 340 magnification for H&E and DAB developed samples. Heart images were analyzed using color thresholding for Masson's, and cross-sectional area measurement (mm 2 ) for H&E on ll OPEN ACCESS iScience 25, 103973, March 18, 2022 iScience Article ImageJ software. Average cross-sectional area of 100 cardiomyocytes was measured for each sample. To determine the nuclear H-score, 10 images per heart were analyzed using IHC profiler plugin (Varghese et al., 2014). The nuclear pixel intensity was assigned a score as high positive (3+), positive (2+), low positive (1+) and negative (0).
Fluorescent-labelled slides were imaged using a Zeiss Axioimager.D2 upright microscope using a 40x/0.2 Plan-neofluar objective and captured using a Coolsnap HQ2 camera (Photometrics) through Micromanager software v1.4.23. Specific band pass filter sets for DAPI (nuclei), FITC and Texas Red were used to prevent bleeding through from one channel to the next. Images were then processed and analyzed using Image J software. For 3D reconstruction, images were collected on a Leica TCS SP8 AOBS inverted confocal using a HC PL APO CS2 40x/1.30 OIL objective and 0.75 confocal zoom. The confocal settings were as follows, pinhole 1 airy unit, scan speed 400 Hz unidirectional, format 1024 3 1024. Images were collected using hybrid detectors with the following detection mirror settings; DAPI 415-472 nm; Alexa-Flour488 497-572 nm; AlexaFlour594 604-752 nm using the white light laser with 405 nm (20%), 488 nm (16%) and 594 nm (12%) laser lines, respectively. The images were collected sequentially and for acquiring 3D optical stacks the confocal software was used to determine the optimal number of Z sections. The 3D images were reconstructed using Imaris Microscopy Image Analysis software (Bitplane AG, Switzerland).
For immunocytofluorescence analysis, percentage CHOP total pixel intensity in nuclei area was quantified by Image J macro: Intensity Ratio Nuclei Cytoplasm Tool (RRID:SCR_018573).

Data analysis
Data are presented as bar or dot plots showing mean G SEM. The animal sample size (N) required to detect 5.20 difference of IVRT mean with SD of 2.58 between mice fed with chow or HFHSD was determined based on a power analysis (statistical power R0.8 and a < 0.05), suggesting a minimum of N = 4 per group. In figure legends, indicated N and n represent biological and technical replicates, respectively. Data were analyzed using one-way or two-way ANOVA followed by post hoc tests where appropriate. Comparisons between two groups were performed using Student's t test. For data with skewed distribution, non-parametric tests were used. Correlations were determined using Pearson's coefficient (R). Statistical analysis was performed using the GraphPad Prism 9 software and p values < 0.05 were considered statistically significant.