Basic Mechanisms of Diabetic Heart Disease

Abstract

Diabetes predisposes affected individuals to a significant spectrum of cardiovascular complications, one of the most debilitating in terms of prognosis is heart failure. Indeed, the increasing global prevalence of diabetes and an aging population has given rise to an epidemic of diabetes-induced heart failure. Despite the significant research attention this phenomenon, termed diabetic cardiomyopathy, has received over several decades, understanding of the full spectrum of potential contributing mechanisms, and their relative contribution to this heart failure phenotype in the specific context of diabetes, has not yet been fully resolved. Key recent preclinical discoveries that comprise the current state-of-the-art understanding of the basic mechanisms of the complex phenotype that is the diabetic heart, form the basis of this review. Abnormalities in each of cardiac metabolism, physiological and pathophysiological signaling and the mitochondrial compartment, in addition to oxidative stress, inflammation, myocardial cell death pathways and neurohumoral mechanisms are addressed. Further, the interactions between each of these contributing mechanisms and how they align to the functional, morphological and structural impairments that characterize the diabetic heart, are considered in light of the clinical context: from the disease burden, its current management in the clinic, and where the knowledge gaps remain. The need for continued interrogation of these mechanisms (both known and those yet to be identified) is essential to not only decipher the how and why of diabetes-induced heart failure, but also to facilitate improved inroads into the clinical management of this pervasive clinical challenge.

INTRODUCTION

The global prevalence of diabetes (particularly type 2 diabetes, T2D) has been progressively increasing over many years in a troubling manner^1,2. The most recent estimates from the International Diabetes Federation Diabetes Atlas considered in excess of 450 million adults with diabetes across the globe, a number that was projected to reach 693 million by 2045². This diabetes pandemic imposes a significant burden on society, both in terms of the substantial healthcare costs and the poor health outcomes for affected patients². Almost five decades ago, the Framingham Heart study established the epidemiological links between diabetes and increased risk of heart failure (HF)³. Indeed, diabetes has long been recognized to not only escalate the risk of heart failure, but also to increase its incidence approximately 2.5-fold (and double this in females), independent of age or concomitant comorbidities such as coronary artery disease and dyslipidemia^4–8. Diabetic patients account for up to one third of patients in clinical HF trials, with diabetes persisting as an independent predictor of poor outcome^9–11 (and as recently reviewed^12–15). Together with an aging population, this has now given rise to a worldwide epidemic of heart failure⁵. Indeed, the existence of ‘diabetic cardiomyopathy” as a distinct entity became increasingly apparent over the almost five decades since its first description in a small cohort of 4 patients in 1972¹⁶.

The phenomenon defined as diabetic cardiomyopathy comprises impairments in cardiac structure and function independent of the macrovascular complications of diabetes (including hypertension, coronary artery disease and atherosclerosis)^17,18, described in a plethora of clinical and experimental studies (as reviewed^19–21). Despite these statistics, there has been a dearth of effective treatments for diabetes-induced cardiac dysfunction, (and even traditional means of intensive blood glucose control often have failed to improve cardiac function or reduce risk of heart failure)^22–24. Conventional therapy for established heart failure is the same, whether or not the patient has diabetes. Mechanism-specific therapy for the diabetic-associated heart failure is currently not available (although the new era of sodium glucose cotransporter-2 inhibition [SGLT2i] offers some promise in this direction). Thus, therapies that target diabetes-induced cardiac dysfunction and subsequent heart failure are urgently needed. Further, patients with concomitant diabetes and heart failure in the absence of other comorbidities such as obesity, hypertension and atherosclerosis are likely the exception rather than the rule in clinical medicine. It is thus now timely in this compendium article to both reconsider the nomenclature and definition of the diabetic heart, and to update the contributing mechanisms to the complex phenotype, identified from preclinical discoveries reported in the last five years.

FUNCTIONAL IMPAIRMENTS THAT CHARACTERIZE THE DIABETIC HEART

Cardiac dysfunction is often clinically silent in diabetes and frequently is not detected until later stages of disease. Even in asymptomatic, normotensive patients with well-controlled diabetes, ~50% are considered to exhibit some degree of cardiac dysfunction^12,25–27. It is widely accepted that one of the hallmarks of the diabetic heart is left ventricular (LV) diastolic dysfunction, which is one of the first signs of diabetic cardiomyopathy, often detectable earlier than clinically-significant LV systolic dysfunction^{17,19–21,28}. Systolic and diastolic dysfunction have been attributed to differences in gross cardiac morphology²⁹. Notably, the existence of isolated diastolic dysfunction as an indicator of diabetic cardiomyopathy was until recently challenged at times, as patients in the early stages of diabetes were not routinely subject to careful assessment of diastolic function. The cardiac complications of diabetes were usually only investigated after overt heart failure symptoms became evident. The era of more sensitive imaging techniques, and the recognition that heart failure is becomingly increasingly common in people affected by diabetes, may overcome this. Regardless of whether this is assessed in the clinical or preclinical/laboratory context, cardiac magnetic resonance imaging (CMRI) is largely considered the gold standard approach for assessment of cardiac function^4,15. However, although diabetes is associated with impaired CMRI-derived parameters of LV diastolic function in the clinic^28,30, CMRI is not always readily available (with limited access in regional, remote and/or low socioeconomic circumstances). Likewise, synchrotron-based cardiac imaging in experimental laboratory animals enables considerable high-resolution imaging information³¹ but does not lend itself to high-throughput, serial assessment. Advancement in ultrasound-based echocardiography imaging approaches in recent years, in both clinical and experimental contexts, now permits relatively high-resolution, detailed, non-invasive, serial assessment of cardiac function without the time-consuming or availability/access restraints often posed by CMRI and synchrotron-based imaging and as such is the imaging approach of choice for the majority of clinical and preclinical cardiac routine and interventional research studies^32,33.

Traditionally, the initial impairments in LV diastolic function that are considered characteristic of diabetes-induced cardiomyopathy include prolonged and delayed LV filling and LV relaxation³², often in the absence of concomitant impairments in LV systolic function^15,27. As detailed by the most recent guidelines published by the American Society of Echocardiography and the European Association of Cardiovascular Imaging, a range of Doppler-based signals can be used for the assessment of diastolic function, with several parameters (usually at least 3-4) of diastolic function assessed. These parameters include those determined at the end of diastole (which correlate with LV end-diastolic pressure, EDP). These are obtained by measuring blood flow velocity across the mitral valve (via Doppler flow echocardiography), or the analogous velocity of motion of the adjacent myocardial tissue (via tissue Doppler imaging). These approaches yield two major phases of movement, an early phase (denoted as the E wave on Doppler flow, and E’ on tissue Doppler) and a later, atrial phase (denoted as the A wave on Doppler flow, and A’ on tissue Doppler)¹⁹. Mitral peak A wave velocity and tissue Doppler–derived mitral annular a′ wave velocity, as well as those that are measured earlier in diastole, such as mitral E/A ratio, E wave deceleration time and E/e′ ratio, are common parameters derived from these approaches. E/e’ is one of the most reproducible and reliable of these examples³². Isovolumetric ventricular relaxation time (IVRT) has also been widely used to assess diastolic function, but is highly-dependent on heart rate and afterload³². Interested readers are referred to the guidelines for a comprehensive viewpoint on the strengths and potential limitations in the presence and absence of comorbidities such as age, hypertension, rhythm disturbances, etc^12,32. There is considerable evidence supporting the presence of impairments in LV diastolic function in the diabetic population, often manifest in a significant proportion of otherwise healthy, complication-free diabetic individuals^34,35. Prolonged deceleration time, reduced e’ and e’/a’, and increased E/e’ are reported in diabetic patients in the absence of clinically-significant impairments in LV systolic function^{30,33,35–38}, in addition to impairments in other (less reliable) measures of diastolic function, such as reduced E/A and prolonged IVRT^30,35,38. Diastolic dysfunction can also be reliably determined in the diabetic population by other echocardiography-derived parameters, including LV global longitudinal strain, speckle-tracking echocardiography and strain rate imaging^18,39,40. Again, these approaches confirm that LV systolic dysfunction is only evident later than diastolic dysfunction in the diabetic human myocardium, and/or is milder, and often in the context of heart failure with preserved ejection fraction (HFpEF, discussed later)^15,32,41,42. These more sensitive means of detection can reveal mild (subclinical) impairments in LV systolic function in diabetic patients, even when LV ejection fraction (EF) is preserved^{15,28,37,43,44}. Regardless of whether concomitant LV systolic function is impaired, evidence from epidemiological studies suggests that LV diastolic dysfunction imparts significant prognostic implications, with a four-fold risk of death relative to healthy individuals⁴⁵. These functional impairments in the diabetic heart are likely independent of concomitant atherosclerosis in both the human and preclinical contexts. Indeed, they are clearly manifest in diabetic rodents which are resistant to coronary atherosclerosis²⁰.

AVAILABLE PRE-CLINICAL MODELS OF DIABETES

A selection of animal models of diabetes have been described over the years, predominantly in rodents, for the study of diabetes and its complications. As has been comprehensively reviewed^14,19,46, these each mimic different aspects of disease pathology, and the choice of model should reflect the clinical aspects relevant to the study goals. For example, streptozotocin (STZ)-induced mouse models of diabetes are amongst the most common models, and background mouse strain can affect the cardiac phenotype observed. In male FVB/N mice, LV diastolic dysfunction (evident from 4 weeks) precedes LV systolic dysfunction by at least several weeks^47,48, and in some studies, the timeframe may not have been sufficiently long for this later systolic dysfunction to become manifest (in both C57/Bl6 and FVB/N strains)^49–51. STZ-induced diabetes superimposed on the FVB/N strain is often associated with negligible differences in body weight at study end compared to non-diabetic sham animals, in contrast to STZ-induced diabetes in C57/Bl6 mice, where a smaller bodyweight at endpoint may be evident in diabetic mice^52,53. Increased levels of reactive oxygen species (ROS), inflammation, fibrosis and apoptosis are routinely seen in the myocardium following STZ-induced diabetes in mice. A clear strength of the STZ-induced mouse model of diabetes is the functional phenotype of LV diastolic dysfunction, evident earlier (and perhaps even in the absence of) LV systolic dysfunction, which thus mimics the common functional phenotype of the diabetic human heart. The weakness of this model is the need to use a toxin to induce diabetes and severe untreated hyperglycemia. In contrast to the STZ mouse model, in male STZ rats, both systolic and diastolic dysfunction are often present, mimicking the later functional phenotype of the diabetic human heart. In addition, the STZ rat retains the cardiac phenotype of increased ROS, inflammation, fibrosis and apoptosis, but weaknesses of this model in rats are the significantly retarded weight gain and bradycardia that are also evident^54–56, which can confound interpretation of results obtained. Other models of type 1 diabetes (T1D) have also been used to study the cardiac complications of the disease, including the Akita (on the C57/Bl6 background) and OVE26 mouse models (on the FVB/N background). The strength of both of these models is that they avoid use of the toxin STZ to limit insulin availability and hence induce T1D^57–62. Both of these genetic models of T1D however do not always reproduce all aspects of the human disease. For example, Akita mice also exhibit retarded weight gain, the cardiac dysfunction is not as reproducible as the STZ mouse and the animals often lack cardiac oxidative stress and cardiac remodeling that are considered common features of the diabetic human heart (as detailed below)^58,59. The OVE26 mouse model of T1D reproduces a cardiac functional phenotype similar to that seen in the STZ rat, whereby both systolic and diastolic dysfunction are evident^60–62. Weaknesses of the OVE26 model is the very early T1D onset (during first week of postnatal development), it is not the most readily-available mouse line, and the disease results from overexpression of pancreatic β-islet cell calmodulin, not a common cause of T1D in humans^60–62. For these reasons, the STZ-induced mouse model of T1D remains a preferred model for many preclinical scientists.

A widely-used model of T2D is the spontaneously-diabetic db/db mouse^63–66. The strength of this model is that it reliably reproduces LV diastolic dysfunction characteristic of the diabetic human heart, together with cardiac oxidative stress and inflammation. A weakness of the db/db mouse is however the morbid obesity and a systemic metabolic phenotype including hyperleptinemia and severe insulin resistance that is quite distinct to the human condition. Moreover, homozygotes are sterile. In contrast, rodent models that combine a milder STZ regime than that used to induce T1D combined with a high fat diet may be a more appropriate choice in some instances, where a milder systemic phenotype combined with a robust cardiomyopathy phenotype of cardiac remodeling and dysfunction is observed⁶⁷. Strengths of these diet-based models is the extent to which they can be widely-available, are not limited by breeding constraints (particularly useful when crossing with another mouse line) and that their systemic phenotype more closely parallels what is observed in human diabetes. Their weakness is the longer amount of time required to produce a cardiac phenotype (perhaps six months), and the milder cardiac dysfunction that may result, compared to the db/db mouse. Rat models of T2D include spontaneously diabetic models such as the Zucker diabetic fatty rat^68,69 and Goto Kakizaki rat (difficult to source outside of Japan)⁷⁰ as well as dietary-based models (high fructose, sucrose or fat)^71–73 and combined STZ-high fat diet models^71,74. Of these models, the spontaneously-diabetic db/db mouse model of T2D, followed by a diet-based mouse model (such as STZ-high fat) have been the preferred model by many preclinical scientists. For a more detailed analysis of the preclinical models of diabetes available, and their strengths and weaknesses, readers are referred to several comprehensive reviews devoted to this topic^14,19,46.

In general, rodent models of diabetic cardiomyopathy do not recapitulate all of the aspects of diabetic cardiomyopathy and although they have provided important mechanistic insights, a strong imperative exists for the development of additional models in other species. Human-derived cells such as inducible pluripotent stem (iPS) stem cells and/or three-dimensional human cardiac organoids to complement animal studies are also being developed, including some with a more mature phenotype^75–80.

MORPHOLOGICAL AND STRUCTURAL PHENOTYPES THAT CHARACTERIZE THE DIABETIC HEART

There has been considerable assessment of the structural abnormalities that are characteristic of the diabetic heart, in clinical and experimental settings, ranging from indices of cardiac remodeling to impairment in coronary microvascular perfusion^19,20,81. These morphological defects likely contribute to the functional impairments evident in the diabetic human heart, but evaluation of whether these alterations precede deficits in cardiac function, relies predominantly from observations in animal models of the disease, where they have been attributed to the pathophysiological mechanisms that lead to diabetic heart disease, as discussed in the next section.

• Cardiac fibrosis:

There is robust evidence of cardiac fibrosis in the diabetic human heart. Early evidence of diabetes-associated cardiac fibrosis in the human heart was largely derived from post-mortem or biopsy analyses^82,83. Increased deposition of interstitial collagen types I and III (detected using immunohistochemistry) is evident in diabetic compared to non-diabetic human myocardial biopsies⁸², confirming earlier reports of interstitial and perivascular fibrosis detected by standard histological and electron microscopic ultrastructural analyses in post-mortem samples ^83,84. However, recent approaches using either CMRI-derived T1 mapping and late gadolinium enhancement (LGE) imaging, or echocardiography-derived integrated backscatter, have proved useful in the non-invasive detection of cardiac fibrosis in humans^85–88. These approaches reveal that diabetes is associated with increased cardiac fibrosis, even in the absence of prior ischemic injury^{15,30,84–86,89,90}, but some observations suggest fibrosis may emerge later in disease progression (after deficits in cardiac function have manifest)^12,16. It has been suggested that a component of the diabetes-induced increase in cardiac fibrosis may be replacement fibrosis⁸³. Further, diabetic patients with LGE-derived evidence of cardiac fibrosis, exhibit poor prognosis across major adverse cardiovascular events (including survival), highlighting a potential role for these non-invasive cardiac imaging approaches in clinical management or risk stratification of the diabetic patient⁸⁵. This clinical evidence of cardiac fibrosis in the diabetic heart is supported by a plethora of preclinical studies in experimental rodent models of diabetes reporting diabetes-associated cardiac fibrosis^{47,49,56,63,65,81,91–96}.

• Cardiac hypertrophy:

An association between diabetes and increased LV mass is well-substantiated^97–101 and is regarded as a structural hallmark of the disease in humans^19,20. CMRI revealed increased LV mass in the human heart in patients with T2D, even in relatively younger patients (e.g. at 40 years of age)^4,102. Echocardiography-derived increases in LV mass are also evident^{18,37,38,103,104}, and may even be exaggerated in diabetic females relative to males^15,36,99,101. It is however pertinent to note that these increases in LV mass in humans affected by diabetes reflects the net effects at the level of cardiac fibrosis (outlined above), cardiomyocyte hypertrophy and cardiomyocyte loss via myocardial cell death pathways (discussed later). Thus, increased LV mass in diabetic patients might not exclusively reflect the impact of cardiomyocyte hypertrophy per se in the diabetic human heart. Observations from diverse animal models, including STZ-induced T1D mice and rats^{51,105, 6}, high-fat-fed mice^107,108, Goto-Kakizaki T2D rats¹⁰⁹ and spontaneously-T2D db/db mice^63,65 suggest that diabetes is commonly associated with cardiomyocyte hypertrophy. However cardiomyocyte hypertrophy has not been observed in other animal models such as in some studies utilizing the Akita and STZ-induced mouse models of T1D^15,110. Although the reason for this controversy regarding whether diabetes increases cardiomyocyte size remains to be resolved, concomitant obesity and/or insulin resistance are potential contributing mechanisms to diabetes-induced cardiomyocyte hypertrophy^111–113, that may exacerbate mechanisms by which high glucose alone may promote cardiomyocyte hypertrophy, at least in experimental settings^106,114.

• Impaired coronary microvascular perfusion:

There are several reports in the literature in which impairments in each, of coronary flow reserve, myocardial blood volume and myocardial blood flow, with increased coronary resistance, are observed in humans affected by diabetes. Myocardial specimens studied at autopsy in diabetic patients reveal reduced numbers of capillaries and arterioles with increased thickness of arteriolar walls¹¹⁵. Diabetic patients also exhibit reduced sensitivity to dipyridamole-induced increases in coronary blood flow, with increased coronary resistance and reduced coronary reserve during cardiac catheterization in vivo¹¹⁶. Similar findings of increased coronary resistance and reduced coronary reserve have been obtained using whole-body positron emission tomography (PET)¹¹⁷. Evidence of microvascular disease, including reduced myocardial blood flow reserve, has also been obtained in the diabetic human heart using contrast echocardiography^36,103. Whilst much of the clinical evidence has been obtained in T2D patients, similar observations have been reported in human subjects with T1D in vivo. For example, reduced sensitivity to cold pressure test and adenosine-induced increases in myocardial blood flow, with increased coronary resistance and reduced coronary reserve (via PET)¹¹⁷, as well as reduced myocardial blood volume and blood flow (via contrast echocardiography)¹¹⁸, have been reported in T1D patients. These diabetes-induced microvascular impairments would reduce the delivery of oxygen and other essential nutrients to the myocardium. Indeed, coronary rarefaction in the human heart has been identified as a predictor of cardiac fibrosis¹¹⁹. This evidence of microcirculation impairment in humans with diabetes is recapitulated in experimental models of diabetes^120–122. Diabetic rodents exhibit significantly decreased myocardial capillary density; this impairment precedes diabetes-induced cardiac fibrosis and in turn, reduced density of cardiomyocytes (at least in STZ-induced diabetic rats)^94,123. Similarly, T2D rats exhibit reduced coronary flow reserve and increased coronary vascular resistance^109,120. Indirect evidence for impaired capillary density in the diabetic myocardium also comes from reduced cardiac expression of pro-angiogenic vascular endothelial growth factor (VEGF), and both of its receptors, VEGF-R1 and VEGF-R2, in both diabetic rats and humans^94,123,124. These microvascular impairments are likely further exacerbated by impaired coronary endothelial function and increased microvascular stiffness as a result of sustained diabetes^{19,117,123,125–129}.

MECHANISMS CONTRIBUTING TO DIABETIC HEART DISEASE

With each passing decade, the list of potential contributing mechanisms to diabetes-induced cardiomyopathy continues to expand. As illustrated in Figure 1, well-established mediators identified for a number of years have included oxidative stress, inflammation and impaired Ca²⁺ handling, as well as alterations in substrate metabolism/utilization, insulin signaling, gene regulation, mitochondrial dysfunction, endoplasmic reticulum (ER) stress, neurohumoral activation and cardiac cell death^19,20. An overview of each of these contributing mechanisms is provided below, with consideration of recent insights to have now emerged. Figure 2 illustrates these contributing mechanisms, their complexity and sites of likely overlap in more comprehensive detail.

Figure 1.

Oxidative stress, inflammation, alterations in metabolic pathways (including abnormalities in substrate utilization, mitochondrial function, advanced glycation end-product [AGE] formation and O-GlcNAcylation), as well as changes at the level of insulin signaling, gene regulation, endoplasmic reticulum (ER) stress, neurohumoral activation and cardiac cell death, have all been widely accepted as mediators of diabetes-induced myocardial remodeling and dysfunction (see text for references).

Figure 2.

The contributing mechanisms to the cardiac phenotype of diabetes-induced cardiomyopathy are complex and multifactorial, as are their consequences. Indeed, there is significant intersection between many of these, driving the diabetic myocardium towards failure (see text for references). Illustration Credit: Ben Smith.

• Oxidative stress:

Oxidative stress is defined as an imbalance between inappropriate (i.e. excessive) ROS generation and the capacity for these to be degraded. Major sources of ROS such as superoxide in the heart include NADPH oxidase, mitochondrial respiration (normally a minor byproduct but increased in diabetes) and uncoupled nitric oxide synthases; the latter two becoming particularly problematic in pathological contexts such as diabetes^{19,81,130–132}. In addition to increased capacity for ROS generation in settings of diabetes (as a result of induction of all three major cardiac sources of ROS), endogenous antioxidant mechanisms are often also impaired in diabetes (as reviewed^19,133,134). Intervention with appropriate antioxidants such as superoxide dismutase (SOD) mimetics or coenzyme Q₁₀, in animal models of diabetes, either early as a preventative measure to limit progression to diabetic cardiomyopathy, or later in the disease process (once cardiac impairment is evident) have established the causal contributing role of oxidative stress to diabetic heart disease (in both T1D and T2D)^{65,113,135–141}. Similar evidence for a causal role in clinical settings of diabetic cardiomyopathy is less abundant, largely hampered by use of relatively poor antioxidant approaches (e.g., vitamin E)^{19,132,142,143}, although increased ROS levels are apparent in human diabetic myocardium and vasculature^144–146. In addition to the direct oxidative damage (to proteins, lipids, DNA) as a result of inappropriate cardiac ROS levels, ROS is also a key trigger of inflammasome activation (discussed below) ^18,142,147. ROS also activate multiple other pathological signaling cascades, such as protein kinase C (PKC), apoptosis signal-regulating kinase-1 (Ask1), p38 mitogen-activated protein kinase (p38-MAPK), NH₂-terminal Jun kinases (JNK) and JAK-STAT pathways^{19,20,105,130,148–155}, which themselves have been implicated in diabetes-induced cardiac complications^156–158. Moreover, several of these (including PKC, p38-MAPK and JAK-STAT)^{130,152,157,159} can also induce ROS generation, creating a detrimental feedforward loop. Two of the most promising approaches for limiting diabetes-induced oxidative damage in the heart in recent years is through pharmacological activation of Nrf2 (regarded as the master antioxidant regulator)^147,160,161 and inhibition of NADPH oxidase with the joint NOX1/NOX4 inhibitor GKT137831, as demonstrated in several preclinical models of diabetes^{129,162–165}. Their efficacy with respect to limiting diabetic cardiomyopathy (particularly in humans) remains to be confirmed.

• Inflammation:

Diabetes is an inflammatory disorder, with increased levels of ROS playing a key contributing role in this context^{19,131,136,147,166}. Indeed, a detrimental feedforward loop in which increased ROS induces inflammatory cytokines and vice versa is thought to provide the momentum for this highly pro-inflammatory environment^{21,147,166,167}. Systemic inflammation is clearly present in both T1D and T2D patients, and is considered a contributing mechanism to peripheral disease progression in several organ systems (e.g. liver, pancreas, kidney and vasculature). Increased circulating cytokines, chemokines, immune cells and other inflammatory biomarkers are clearly evident^168–170. Moreover, diabetes promotes tissue infiltration of macrophages and their polarization towards an M1-like ‘pro-inflammatory’ phenotype in affected patients, with upregulated leukocyte inflammatory cytokine signaling¹⁷¹. Mouse and rat models of T1D and T2D also display systemic inflammation relatively early in disease progression, as indicated by increased circulating monocytes and neutrophils¹⁷⁰, as well as increased tumor necrosis factor-α (TNFα), interleukins (IL-1β, IL-6), transforming growth factor-β (TGF-β), interferon-γ (IFNγ), the inflammatory transcriptional regulator NFκB, and various chemokines and vascular adhesion molecules^{155,172–177}. A pathophysiological role for inflammatory signaling is thus clearly implicated in diabetic complications. The influx of infiltrating immune cells has emerged as a significant contributor to the development and progression of myocardial injury and LV dysfunction in a range of cardiac pathologies, including myocardial ischemia, HF with reduced ejection fraction (HFrEF) and HFpEF^178–185. Inflammatory processes are also activated in human diabetic myocardium, as evidenced by increased cytokines, chemokines, and various leukocyte populations. This implicates cardiac inflammation in the pathogenesis of cardiomyopathy in diabetic patients¹⁶⁹. Induction of pro-inflammatory genes and proteins (TNFα, IL-1β, IL-6, TGF-β, IFNγ and NFκB) have also been reported (at timepoints later than systemic inflammatory markers) in the LV of T1D and T2D rodent models of diabetes^{54,131,136, 141,153–155,158,166,167,186–188}. Damage-associated molecular pattern molecules (DAMPs, e.g. S100A8/S100A9), infiltration of other inflammatory cells (e.g. T cells, B cells, neutrophils, dendritic cells and mast cells) and/or TH1/TH17 immune responses could also contribute to cardiomyopathy in this context^{170,172,182,187}. Much of this cardiac inflammation in the diabetic heart may be triggered or amplified by increased ROS levels, as revealed by the ability of ROS-lowering strategies to ameliorate the pro-inflammatory state in T1D and T2D rodent models of diabetes^136,167,189. This pro-inflammatory state is thought to be kickstarted to a large extent by ROS-triggered activation of the NLRP3 inflammasome in the diabetic heart¹⁹⁰, as has been observed in multiple preclinical models^{141,161,189,191}. NLRP3 activation triggers a cascade of events, recruiting pro-caspase-1 to facilitate caspase-1 activation, and cleavage of IL-1β and IL-18 precursors to generate their active products^21,147. As a result, a plethora of the above pro-inflammatory mechanisms are enhanced¹⁸⁹, favoring an altered macrophage polarization towards an M1-like phenotype in diabetes^21,192. The net effect of infiltrating inflammatory cells such as macrophages can be implicated in diabetes-induced LV fibrosis and dysfunction.

In addition to the pro-inflammatory state evident in diabetes, how the body attempts to recover from this continuous insult may also be impaired. Infiltrating macrophages exhibit defective phagocytosis of apoptotic neutrophils at the site of vascular inflammatory injury (efferocytosis)¹⁸². It is thus possible that defective resolution of inflammation may hence represent an additional mechanism of diabetes-induced cardiac damage secondary to macrophage infiltration, as is evident in other tissues^192–194. In support of this, approaches including targeting of the chemokine receptor CCR2, or enhancement of heme-oxygenase-1 (downstream of Nrf2), which favor macrophage polarization towards an M2-like state, suppress diabetes-induced oxidative stress, remodeling and dysfunction in the myocardium in both T1D and T2D settings^155,186. There is thus considerable interest in the development of novel pro-resolving therapies to target a range of inflammatory diseases^195,196 and diabetes and the associated cardiomyopathy could be one such indication for such an approach.

• Changes at the level of insulin sensitivity and signaling:

Impaired systemic insulin sensitivity is evident in, and characteristic of, people with diabetes^197–199. In the heart, sensitivity to insulin as a result of diabetes is impaired at multiple levels. This includes reduced insulin-stimulated glucose uptake into cardiomyocytes and other cardiac cell types, with decreases in both LV GLUT4 expression as well as reduced translocation of GLUT4 from the cytoplasm to the cell membrane¹⁹. Responsiveness to insulin is also evident at the level of myocardial contractility and regulation of vascular tone, in which phosphoinositide 3-kinase[p110α], PI3Kα, and its mediator Akt play contributing roles²⁰⁰. Moreover, the relationship between altered insulin signaling and diabetic cardiomyopathy is complex, reflecting consequences of increased or imbalanced activation of certain insulin signaling pathways that could promote cardiac remodeling on the one hand versus inhibition of cardioprotective pathways on the other²⁰¹. Activation of pathways that may promote LV remodeling include signaling pathways downstream of insulin receptor substrate 1 (IRS1) or crosstalk between insulin signaling and beta adrenergic signaling via the G-coupled receptor kinase (GRK2)^202,203. Altered effectiveness of endogenous physiological mechanisms in the heart, including insulin sensitivity and downstream signaling, through both insulin and insulin-like growth factor-1 (IGF1) receptors (via activation of PI3Kα, and its mediator Akt), have also been observed in the diabetic myocardium^70,204,205. Indeed, defects in PI3Kα/Akt signaling exaggerate diabetic cardiomyopathy^105,135 and indeed may be sufficient to impair insulin responsiveness, at least in the liver and adipose^206–208. On the other hand, increased cardiac activity of PI3Kα in the context of diabetic cardiomyopathy is cardioprotective; these actions are closely associated with suppression of diabetes-induced cardiac ROS and inflammation^47,105,209. This evidence of impairments in insulin sensitivity and signaling in the diabetic myocardium (and their likely mechanistic role in the diabetic heart) is largely reliant on observations in diabetic rodent models^{47,65,70,109,210–212}, as corresponding observations in myocardial biopsies are logistically challenging in humans, although a small number of studies have provided evidence to support imbalanced insulin signaling^201,213. Therapeutic targeting of these changes thus represents innovative means of limiting the cardiac consequences of diabetes, particularly if selective for the myocardium^47,49,105 (a feat not yet met by pharmacological means).

• Altered cardiac metabolic pathways

Diabetes alters myocardial substrate utilization^{20, 21}, such that there is reduced glucose oxidation, increased fatty acid oxidation and reduced glycolysis. In addition, ROS-induced mitochondrial uncoupling in the diabetic myocardium markedly reduces cardiac efficiency^64,204. Formation of advanced glycation end-products (AGEs) and a maladaptive hexosamine biosynthesis pathway (HBP) also contribute to glucotoxicity^{19, 20}.

Advanced glycation end-products (AGEs):

As described by Brownlee and others, AGEs are long-lasting molecules that result from advanced, non-enzymatic glycosylation of excess sugars which can induce an irreversible form of posttranslational protein modification^20,214,215. AGE-modification of extracellular matrix (ECM) and other structural proteins are thought to provide a physical impediment to myocardial compliance. Preclinical studies targeting AGEs and their receptors, RAGE, reveal that hyperglycemia-induced increases in the AGE-RAGE axis may contribute to diabetic heart disease^20,216. Moreover, a detrimental feed-forward cycle has been proposed, whereby ROS trigger increased AGE formation and RAGE activation^20,217, and the AGE-RAGE axis also drives ROS formation and concomitant inflammation (particularly in the context of diabetes and its complications)^218,219. The crosslink breaker Alagebrium (ALT-711, which disrupts crosslinks between proteins and AGEs) attenuates the morphological characteristics of both the T1D rat heart⁹⁵ and the high fat fed mouse heart²¹⁹. These cardioprotective effects of Alagebrium mirror those seen in mice lacking RAGE^219,220. Indeed, enhancing levels of AGEs even in the absence of diabetes is sufficient to drive endothelial inflammation and atherogenesis²²¹. More recently, the mitochondrial-targeted methylglyoxyl scavenger, mitoGamide, has been shown to also protect LV diastolic function in Akita T1D mice⁵⁷. AGEs thus likely contribute to diabetes-induced cardiac remodeling and dysfunction. AGE modifications of cardiac proteins is not restricted to ECM proteins but occurs more broadly in the diabetic myocardium, as observed for example with the Ca²⁺-handling protein, sarco(endo)plasmic reticulum Ca²⁺-ATPase (SERCA2a), and myofilament proteins including troponin, actin and α-myosin heavy chain^{220,222–224}. Although Alagebrium has been taken into clinical trials in humans with heart failure in the absence and presence of diabetes²¹⁵, these have however largely failed to be completed due to financial reasons. As such, therapies specifically targeting AGEs have not crossed over into routine clinical care.

Hexosamine biosynthesis pathway (HBP):

Under normal physiological conditions, the majority of glucose metabolized in the heart is attributed to glycolysis, the pentose phosphate pathway and glycogen synthesis. Hexosamine biosynthesis represents a relatively minor consequence of glucose metabolism (crudely estimated as <5% of all glucose metabolism)^225–227, with UDP-N-acetylglucosamine (UDP-GlcNAc) its ultimate product. UDP-GlcNAc provides O-GlcNAc (an effector of glucose) for post-translational modification of more than 3000 nuclear, mitochondrial, and cytosolic proteins²²⁸. Analogous to phosphorylation, O-GlcNAcylation occurs specifically at Ser/Thr residues, changing activity of affected proteins. In contrast to phosphorylation, only two functionally-opposing enzymes control all protein modification by O-GlcNAcylation. The O-GlcNAc-transferase (OGT) and O-GlcNAcase (OGA) are responsible for the addition and removal of O-GlcNAc from susceptible proteins, respectively^225–228. Whilst it has been known for some time that glutamine fructose-6-phosphate aminotransferase (GFAT) is the rate-limiting enzyme of the HBP^225–227, recent work now reveals that whereas HBP flux increases in direct proportion with glucosamine levels, glucose metabolism via the HBP only represents ~0.006% of that which is consumed by glycolysis²²⁹. This is much lower than previous estimates. The O-GlcNAc modification (‘O-GlcNAcylation’) is distinct from other forms of glycosylation that occur in the ER and Golgi apparatus (N-linked, mucin-like, O-mannosylation), as well as advanced glycation²²⁷. In diabetes, protein O-GlcNAcylation becomes maladaptive, and the sustained induction of O-GlcNAcylation contributes to cardiac glucotoxicity^{198,225–227,230}. Both sustained hyperglycemia and increased cardiac ROS generation increase the flux of glucose metabolism through hexosamine biosynthesis, enhancing UDP-GlcNAc levels and subsequent O-GlcNAcylation^{225–227,231}. In contrast, an acute increase in glucose load (over minutes rather than hours) has minimal impact on cardiac UDP-GlcNAc levels²²⁹. In addition to the known ability of ROS to drive protein O-GlcNAcylation, hyper O-GlcNAcylation in return drives NADPH oxidase, illustrating the intimate relationship between O-GlcNAcylation and ROS/oxidative stress^81,232–234. Further, both O-GlcNAcylation and ROS are regulators of inflammation, autophagy and ER stress^81,235.

Activity of the rate-limiting enzyme of the HBP, GFAT and of its downstream product, UDP-GlcNAc, are increased in human T2D at the level of skeletal muscle^198,230 and the heart^233,234. Gene expression and activity of GFAT were also reported to be increased in human T2D myocardium, in conjunction with increased total protein O-GlcNAcylation²³⁶. Similar correlations have been described in leucocytes^237,238. This might be exacerbated by reduced OGA expression in human T2D (both systemically and in the myocardium), impairing removal of O-GlcNAc from affected proteins and impairing O-GlcNAc balance^233,239. Interestingly, genetic mutations resulting in a truncated, less-effective form of OGA carries a 3-fold increased risk of developing T2D^238,240. O-GlcNAcylation has also emerged as a novel regulator of inflammation, of both pro- and anti-inflammatory pathways, as recently reviewed²⁴¹. For example, O-GlcNAc-modified TGF‑β-activated kinase-1 triggers pro-inflammatory signals such as NFκB, which enhances levels of cytokines and promotes macrophage polarization towards an M1-like phenotype^241,242. O-GlcNAcylation has also been implicated in the activation of both T-cells and B-cells²⁴³. There are still many unanswered questions as to functional correlates of the many (>3000) proteins susceptible to this posttranslational modification. O-GlcNAc-modified Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) under hyperglycemic conditions is directly linked to enhanced CaMKII activity and cardiac Ca²⁺ sparks²³⁶. The cardioprotective signals PI3Kα and Akt are O-GlcNAc-modified in diabetes, an effect that is ablated by cardiac gene delivery of O-GlcNAc-ase in parallel with blunting of the structural and functional characteristics of the diabetic mouse heart^233,234. Further studies interrogating the functional consequences of O-GlcNAc-modifications of specific proteins is thus warranted. On the other hand, O-GlcNAc modification can favor protective anti-inflammatory actions, such as in the atherosclerotic vasculature, particularly in an acute context^244.245, but this is unlikely to apply in the context of diabetes, and the diabetic myocardium, where sustained O-GlcNAcylation is evident, and detrimental^233,234. Together, these observations demonstrate clinical relevance of maladaptive O-GlcNAcylation in human T2D myocardium.

Other metabolic alterations in the diabetic heart:

In addition to these glucose-driven abnormalities in cardiac metabolism, other impairments in substrate utilization, particularly of fatty acids and the resultant lipotoxicity, are also evident. Circulating levels of free fatty acids, and other lipids (triglycerides, etc.) are elevated in diabetes, particularly T2D^{63,64,246,247}, likely as a result of impaired insulin action in adipose tissue and liver^201,248. Importantly, diabetes-associated cardiac lipotoxicity is not unique to the T2D heart, but is also observed in T1D^247,248. As recently reviewed^21,248,249, the diabetic heart exhibits altered substrate utilization for energy (i.e. ATP) generation, with an enhanced reliance on fatty acids (at the expense of glucose utilization)^64,250. This altered utilization is facilitated in part by the increased expression of sarcolemmal transport proteins on cardiomyocytes that mediate free fatty acid uptake by the heart (particularly CD36), whilst expression and sarcolemmal localization of glucose transport proteins on the heart (i.e. GLUT4) is reduced^{20,54,205,247,249–251}. Myocardial lipid content increases as a consequence, as observed in both humans and animal models of diabetes^64,247,252. Increased cardiac lipid accumulation has been observed as a relatively early change in the diabetic heart (preceding onset of impaired LV function), and hence represents another potential contributor to diabetic heart disease^20,246–248. Concomitant with this switch in substrate preference, abnormalities at the level of the mitochondria (across oxidative phosphorylation, dynamics, morphology and ROS generation) in the diabetic myocardium are not only manifest but have been convincingly implicated as contributing causal mechanisms of diabetes-induced impairments in mitochondrial function^21,249, as detailed below. ER stress is also evident in the diabetic myocardium, likely resulting from multiple abnormalities, including increased ROS/oxidative stress, altered substrate utilization (and subsequent lipotoxicity), activation of cell death pathways (apoptosis, autophagy, etc.), amongst other contributing mechanisms^20,253.

• Diabetes-induced impairments in the mitochondrial structure and function:

Mitochondria are essential integrators of redox signals and metabolic flux^14,254. In T2D, in both patients and preclinical models, mitochondrial dynamics are impaired and mitochondrial ROS generation exceeds the endogenous scavenging capacity, leading to cardiac oxidative stress and inflammation^{19,144,249,255,256}. Mitochondrial number may actually be increased, but mitochondria are smaller in size, and often fragmented^256,257, suggestive of an imbalance in fission/fusion quality control, as supported by cardiac mitochondrial fragmentation and reduced expression of the mitochondrial fusion protein, mitofusin-1, in the diabetic human myocardium²⁵⁵. This mitochondrial dysfunction plays an essential part in the progression of diabetic cardiomyopathy^{19,144,255,256}. Multiple mechanisms impair mitochondrial function in diabetic hearts, much of which directly results from the altered myocardial substrate utilization outlined above. Impaired insulin signaling itself also has a profound effect on mitochondrial morphology and function^258,259. Further, disruption of mitochondrial physiology following impaired glycemic control is a major contributing mechanism to diabetic heart disease, with many of the above cytoplasmic perturbations reciprocated within the mitochondrial environment.

Mitochondrial ROS:

Metabolic perturbations such as hyperglycemia and increased cardiac lipid load increase mitochondrial ROS generation^131,260,261. This augmented mitochondrial ROS generation has been observed in human myocardium¹⁴⁴. Evidence from preclinical models suggests this is sufficient to impair mitochondrial dynamics, and trigger mitochondrial fragmentation and dysfunction, with dysregulation of mitochondrial fission/fusion proteins such as DRP1 and OPA1 implicated in this process^253,260. The net effect of these diabetes-induced impairments in mitochondrial function correlate with worsening diabetes¹⁴⁴.

Mitochondrial O-GlcNAcylation:

A functional O-GlcNAc cycle, including OGT and OGA, is housed within mitochondria^231,262,263, and the pyrimidine nucleotide carrier can transport UDP-GlcNAc into mitochondria from the cytoplasm²⁶². Overexpression of cytosolic OGT is sufficient to increase O‑GlcNAcylation of mitochondrial proteins²⁶⁴, but whether his is also true for mitochondrial OGT is yet to be determined. Hyperglycemia increases O-GlcNAcylation of several proteins key to mitochondrial respiration (e.g. Complexes I, III, IV and their subunits, e.g. NDUFA9). Evidence from short-term rodent models of T1D (over 2–4 weeks, prior to impairment in cardiac function) suggests that O-GlcNAcylation of numerous proteins key to mitochondrial dynamics is elevated^{226,262,263,265}. Affected proteins include DRP1, OPA1, pyruvate dehydrogenase, mitochondrial transcription factor A (TFAM) and mitofusin-2. Whether O-GlcNAcylation of these mitochondrial components is increased in the heart over the longer-term (when onset of cardiac dysfunction is evident) and/or in the more prevalent T2D, is currently not known. How sustained O-GlcNAc-modifications of specific mitochondrial proteins affects both their activity and net cardiac function also remains unresolved, but we predict that maladaptive O-GlcNAcylation of key mitochondrial proteins may provide another mechanistic link between T2D and cardiac dysfunction.

• Myocardial cell death pathways:

As outlined above, altered homeostatic pathways such as oxidative and ER stress have been implicated as drivers of cell death pathways in metabolic disease, with input from impaired mitochondrial dynamics likely contributing to this. Key forms of cell death evident in the diabetic heart, in both T1D and T2D models, include apoptosis, autophagy and necrosis^{14,19,20,47,48,53,94,211,266–271}. Under physiological conditions, a low-level of apoptosis and autophagy are essential for tissue homeostasis, permitting timely removal of unwanted cells and organelles/proteins respectively whose time has come^20,272. Whilst the increased apoptosis evident in the diabetic heart (reflected to a considerable extent as cardiomyocyte death) and subsequent replacement fibrosis are conventionally accepted as a detrimental phenomenon^19,20 whether this also applies to autophagy remains less clear¹⁴. Insulin signaling is an important regulator of myocardial autophagy^201,273, with reduced insulin action being associated with increased myocardial autophagy, whereas excessive insulin signaling suppressing autophagy. Thus the relationship between insulin action and autophagy in diabetes, might depend on the relationship between activation or repression of insulin signaling pathways. Additionally, lipotoxicity has also been implicated in autophagy dysfunction in the context of diabetes^274,275. Increased autophagy observed in high-fructose-fed insulin-resistant mice for example is associated with modest increases in both cardiac ROS and collagen levels²⁶⁸, but the impact on LV function was not determined. In contrast, blunted cardiac autophagic flux has been observed in the obese ob/ob mouse heart²⁷⁶, and the T1D heart, in at least 2 different mouse models^277,278, hence the impact of diabetes on cardiac autophagy, and its consequences, still remain to be resolved¹⁴. However, Mst1 has been implicated as a possible master switch regulating whether the cell progresses towards apoptotic or autophagic mechanisms in the diabetic heart²⁷⁹. In addition to activation of cell death pathways in diabetic hearts, it is also apparent that a number of endogenous protective mechanisms that would otherwise defend against cell death are also negatively affected by diabetes²⁰⁵. Taken together, these observations suggest that the diabetic myocardium appears destined to suffer increased cardiac cell death. Indeed subclinical myocardial injury has long been described in epidemiological studies as evidenced by increased circulating concentrations of troponin (detected by highly sensitive assays) in individuals with T2D and prediabetes²⁸⁰.

• Impaired calcium handling:

A slower decay of the Ca²⁺ transient has repeatedly been reported in studies of isolated diabetic cardiomyocytes, even before any evidence of LV systolic dysfunction becomes manifest^15,281–286. Putative mechanisms of these impairments in decay of the Ca²⁺ transient include reduced gene and protein expression, and/or reduced activity, of SERCA2a¹⁵. Post-translational modifications of this Ca²⁺-handling protein as a result of advanced glycation end-products and/or O-GlcNAcylation have also been implicated in this deficit, as well as oxidative modifications^{15,19,20,224,233,234,287–290}. Multiple other aspects of Ca²⁺-handling, as a result of impaired Ca²⁺ influx and efflux to/from the cytoplasm from the extracellular space, altered Ca²⁺levels in intracellular organelles (including both the SR and mitochondria), altered expression and/or activity of other Ca²⁺-handling proteins (e.g. phospholamban, ryanodine receptors, etc.), as well as myofibrillar responsiveness to Ca²⁺, are also defective in the diabetic heart^{19–21,220,282,286,288,289,291–295}. Indeed Ca²⁺-handling defects were one of the earliest defective mechanisms identified in the diabetic heart^{291,292,296,297}. Moreover, decreased PI3Kα signaling has been associated with blunted inward Ca²⁺ currents (and thus contractility), and defects in both PI3Kα and inward Ca²⁺ currents have been reported in the T1D Akita mouse^298–300. This spectrum of diabetes-induced defects in Ca²⁺-handling and responsiveness to Ca²⁺ underlie delays in cardiomyocyte relengthening in the diabetic heart, and thus have been attributed to account for a significant component of diabetes-induced LV diastolic dysfunction in vivo.

• Neurohumoral mechanisms:

It has been appreciated for several decades now that the renin-angiotensin-aldosterone-system (RAAS)^{19,21,301–306} and the endothelin-1 system^19,307 are activated in diabetes. This is evident both systemically and within the heart, and across both clinical and preclinical contexts. As the RAAS is a critical regulator of blood pressure^308,309, both the resultant increased afterload secondary to diabetes-induced systemic RAAS upregulation, as well as the direct actions of the cardiac RAAS on the myocardium, contribute to diabetes-induced remodeling. Indeed pharmacological and gene knockout approaches to targeting these systems is effective at limiting diabetic cardiomyopathy in animal models^{65,136,302,303,310}, implicating them as contributing mechanisms to diabetic heart disease. For this reason, in heart failure patients with concomitant diabetes the use of ACE inhibitors or angiotensin receptor antagonists^81,143,311 has been advocated in patients with adequate renal function. Diabetes is also associated with autonomic dysfunction, including at the level of the heart (such as abnormalities in heart rhythm regulation), observed in patients and animals with diabetes^312–315. Diabetes-induced autonomic dysfunction likely further exacerbates inflammation in this context, with a resulting increase in levels of immune cells, cytokines and DAMPs³¹⁶. Demonstrating a potential role for cardiac autonomic dysfunction in diabetic heart disease however is precluded by lack of available approaches with which to manage this dysfunction. More recently, fibroblast growth factor 21 (FGF21) has emerged as a novel skeletal hormone that is important in the regulation of glucose and lipid metabolism in diabetes³¹⁷. Interestingly, its expression is induced in this context, with increased levels detected in the circulation in T2D humans^318,319. Given that FGF21 has been shown to both enhance fatty acid oxidation and glucose utilization, whilst concomitantly blunting lipogenesis^320,321, it likely attempts to serve as a compensatory brake against further metabolic disturbances in diabetes, as recently reviewed³²². A direct role of FGF21 in ameliorating cardiac dysfunction in diabetes remains to be determined.

• Changes at the level of cardiac gene regulation, such as epigenetic and microRNA-mediated mechanisms, have also been reported as regulators of the phenotype evident in preclinical models of the diabetic heart, as previously reviewed^{19–21,81,323,324}. Increased cardiac expression and/or circulating levels of several small, non-coding single-stranded miRNAs have been described in both experimental settings of diabetes, including miR-199a, miR-21^{19,81,325–327}, as well as in the diabetic human myocardium (including miR-199b, miR-210, miR-223, miR-34b, miR-34c and miR-650 amongst multiple others)^19,328. Other miRNAs are downregulated in the diabetic heart, including miR-1, miR-133a, miR-30c, miR-181a and miR-203^{19,324,325,327,329,330}. Whether each of these alterations are a consequence versus a causal contributing mechanism to diabetic cardiomyopathy is yet to be resolved for many of these³²³. miR-1, miR-133a and MiR-203 may normally serve protective roles against glucose-driven impairments in cardiomyocyte death and/or size, and induction of fibrosis^20,327,331, whilst rescue of myocardial miR-92a expression in the T2D mouse heart preserves LV systolic and diastolic function³³². An imbalance of detrimental miRs increased in the myocardium (and/or loss of protective miRs) as a result of sustained diabetes may thus contribute to the net cardiac phenotype in this context. Other non-coding RNA molecules including long non-coding RNAs (lncRNAs) and circRNAs have been shown to be regulated by high glucose in cardiac fibroblasts and cardiomyocytes in in vitro studies³²³. Likewise, small non-coding RNAs (snoRNAs), including the family U32a, U33 and U35a, have been shown to negatively regulate both lipotoxicity and systemic glucose metabolism, blunting oxidative stress^333,334. These non-coding RNA molecules may potentially also emerge as contributing to components of the diabetic myocardial phenotype. Lastly, DNA methylation and histone modifications (acetylation and deacetylation) are altered as a result of diabetes in multiple organs, and have been suggested as potential co-conspirators in the progression of diabetic cardiomyopathy, at least an animal models of the disease⁸¹. Consensus on their causal role in this process with respect to the diabetic heart is however yet to be reached. Indeed, a small number of recent studies have been exploring the area of epigenetic changes in diabetic cardiomyopathy^271,335. For example, diabetes is associated with abnormalities in DNA methylation in both human (cardiac mesenchymal cells and cardiac biopsies) and rodent myocardium (mouse and rat models of T1D and T2D)^271,335. We anticipate that these might yield additional insights in the future into the causal factors underlying diabetes-associated heart failure over time.

As the significant plethora of above studies indicates, both the consequences of, and contributing mechanisms to, the cardiac phenotype is complex and multifactorial. As illustrated in Figure 2, many of the drivers of this phenomenon interact, creating a web of multiple, detrimental feed-forward pathways that, without an effective intervention, push the diabetic myocardium towards a fate of poor prognosis. The status of current understanding of the time of onset of each of the major players in the phenotype of the diabetic myocardium is stylized in Figure 3; this timing is considered in the context of common key comorbidities.

Figure 3.

A stylized overview of current understanding of the time-course of diabetes progression, from the initial onset of hyperglycemia, through the mid and late stages of the disease, and their consequences for the phenotype of the diabetic myocardium (see text for references).

THE DIABETIC HEART AND HFpEF

HFpEF describes a diagnosis of heart failure in symptomatic patients whose LV EF is >40–50% (and in whom non-cardiac causes of symptoms have been excluded)^{119,336–338}. This HF phenotype is now slightly more common than HFrEF in hospital admissions for HF^336,339, an increased prevalence secondary to the increased global prevalence of diabetes, obesity and hypertension in an aging population^337,340. HFpEF likely represents a spectrum of several etiologies depending on which comorbidities and risk factors are also present. These can include aging, obesity, and hypertension amongst others. Both female subjects (particularly elderly females) and diabetic subjects are over-represented in the HFpEF population (compared to HFrEF)^{336,340–342}. Hypertension is a common feature of HFpEF. patients³³⁸. Considering HFpEF as a single phenotype is however a misconception, as the disorder is particular heterogeneous, with at least three distinct clusters described^338,340. This includes the obesity-cardiometabolic phenotype that has been classically associated with HFpEF (patients with a high body mass index, marked diastolic dysfunction, and commonly accompanied by diabetes and/or hypertension). Other subgroups include the natriuretic peptide deficiency syndrome (obese patients with low BNP who are often hypertensive as a consequence), as well as the high cardiorenal phenotype with right ventricular (RV) failure^340,343,344. Of these three clusters, event-free survival is highest for those in the natriuretic peptide deficiency cluster and poorest in the cardiorenal-RV failure cluster³⁴⁰. Intersection between these phenotype clusters is also evident³³⁷.

The increasing emergence and awareness of HFpEF in the HF research field might tempt some to (falsely) consider diabetic cardiomyopathy simply the manifestation of HFpEF coexistent with diabetes, given that both share a number of common characteristics. Indeed, diabetic patients may be affected by either HFpEF or HFrEF^341,345,346. If these diabetic patients also exhibit microvascular complications, then a HFpEF phenotype (rather than HFrEF) is however more prevalent, as is a poorer prognosis^341,345. Post-mortem characteristics of the human HFpEF myocardium include clear evidence of increases in both cardiac mass and fibrosis, with reduced microvascular density¹¹⁹. These observations are also evident via echocardiography^336,347. It is likely that a large component of the increased cardiac mass is attributable to fibrosis rather than hypertrophy per se, as clinically-significant hypertrophy may only be evident in approximately 50% of HFpEF patients¹¹⁹. In addition to cardiac fibrosis and microvascular dysfunction, inflammation and cardiac stiffness are also evident³³⁸. In contrast to HFpEF cohorts (where females tend to be over-represented)³³⁶, diabetic cardiomyopathy does not appear to “favor” one gender over the other³⁴⁸. Further, diabetes-associated HF also includes patients with HFrEF and those with HFpEF³⁴⁶. Thus although diabetic cardiomyopathy is more complex than simply the manifestation of HFpEF coexistent with diabetes, it is clear that the HFpEF phenotype is enriched in diabetes. Future study is required to further understand the mechanistic contribution of diabetes to HFpEF.

BRIEF COMMENTARY ON CURRENT THERAPY FOR THE DIABETIC HEART

Poor glycemic control contributes to increased risk of HF in diabetes, with risk of HF correlated to HbA1c³⁴⁹. Prior to the current era, intense glucose-lowering therapies have largely failed to reduce the risk of incident HF, and may even exacerbate HF risk (particularly thiazolidinediones, sulfonylureas, certain dipeptidyl peptidase-4 (DPP-4) inhibitors and potentially insulin and glucagon-like peptide [GLP-1] agonists)^350–355. Although the GLP-1 agonist was shown to be safe and tolerable in T2D patients in the REWIND study, with evidence of sustained glucose-lowering (reduction in HbA1c of 0.61 percentage units over 5 months) together with modest but significant reductions in bodyweight (by 1.5kg) and systolic blood pressure (BP, by 1.7mmHg) compared to placebo over 5 years, no differences in cumulative incidence of cardiovascular outcomes was seen³⁵⁶. The arrival of sodium-glucose co-transporter-2 inhibitors (SGLT2i) has brought a welcome change^355,357,358. The EMPA-REG OUTCOME study provided the first observation that SGLT2i represented robust cardioprotective actions for an effective glucose-lowering therapy. Risk of HF hospitalization and cardiovascular mortality were reduced by 35–50% with empagliflozin compared to placebo in T2D patients³⁵⁸. Similar findings with respect to benefits at the level of both reduced HF hospitalization and cardiovascular mortality have subsequently been observed in clinical studies in T2D with both canagliflozin (CANVAS, CREDENCE, CVD-REAL)^359–361 and dapagliflozin (DECLARE-TIMI 58)^362,363. The reduction in HF hospitalizations has thus now emerged as a class effect^364,365. Moreover, SGLT2i appears equally beneficial in T2D patients affected by either HFpEF or HFrEF^346,355.

Contributing mechanisms for this SGLT2i cardioprotection particularly with respect to beneficial impact on HF in the context of diabetes remains to be resolved, with a plethora of putative mechanisms reported in the literature^366–369. The benefit with respect to HF outcomes is surprisingly early, at least at the level of hospitalizations for HF (with the event curves separating within weeks of treatment commencement). Whether beneficial effects at the level of cardiac remodeling and/or dysfunction are evident at this point in the diabetic human heart remains to be determined. Whilst it has been suggested that the benefits of SGLT2i simply reflect mechanisms such as enhanced glycosuria, natriuresis and reduced BP³⁶⁵, consensus on the mechanism of SGLT2i mechanism of cardioprotection is yet to be reached. SGLT2 is not detected in heart^370,371, hence the mechanism of cardioprotection is likely extracardiac in nature; moreover, more than one mechanism may contribute. Although enhanced glycosuria and natriuresis are likely to aid renoprotection, they are not likely to be sufficient to completely account for the cardiac benefit, particularly with respect to HF. Glucose-lowering therapies historically have largely been ineffective with respect to cardiovascular endpoints (prior to the current era), potentially arguing against glucosuria as the sole, standalone mechanism of benefit of SGLT2i. Diuretics do not match the efficacy seen with SGLT2i to date, hence volume unloading on its own is similarly insufficient to fully account for cardioprotection in this context. BP-lowering SGLT2i actions have also come under the spotlight as a contributing mechanism of benefit^359,372,373, but the antihypertensive effects of SLGT2i are generally more modest than seen for example with ACE inhibitors³⁶⁷, although of note patients in the EMPA-REG study were well-matched for BP control^358,365. Preclinical studies have implicated the impact of empagliflozin on sodium/hydrogen exchanger-1 in the myocardium³⁶⁶, and effects on sympathetic tone (which in turn regulates BP). Most recently, it was revealed that increased hematocrit and enhanced renal erythropoietin synthesis, boosting the systemic oxygen-carrying capacity of erythrocytes, could contribute a significant component of SGLT2i cardioprotection³⁷⁴. Whether these latter putative mechanisms are unique to empagliflozin or a class effect also remains to be resolved. A multifactorial mechanism of SGLT2i benefit, combining glucose-lowering with other, diuretic, natriuretic, antihypertensive and metabolic benefits (analogous to a poly-pill) may indeed explain the cardioprotective consequences of SGLT2i therapy. Despite this lack of consensus regarding the mechanism responsible for SGLT2i benefit, this benefit has been accepted with certainty and excitement. Indeed clinical trials are currently underway for SGLT2i in HF even in the absence of diabetes, and evidence has already emerged that SGLT2i are efficacious in the context of non-diabetic HF^375–378. For example, the DAPA-HF phase 3 randomized clinical trial of dapagliflozin or placebo (on top of standard care) in >4700 HF patients with an LVEF <40% revealed a markedly reduced risk of worsening HF or cardiovascular death, in both the diabetic and non-diabetic cohorts³⁷⁸. The smaller DEFINE-HF trial in HFrEF patients reported a greater proportion of patients with improved health status in the dapagliflozin treatment arm than placebo, an effect that was comparable between the diabetic and non-diabetic HFrEF patients³⁷⁹. Readers are referred to the position statement from the European Society of Cardiology/Heart Failure Association for a comprehensive review³⁵⁵. Even with the promising signs that SGLT2i may offer new hope for the debilitating combination of diabetes and HF, the problem of the diabetic heart (and its effective clinical management) is not yet solved. It is likely that more than one component of the complex and multifaceted pathophysiology of diabetic heart disease (and hence more than one treatment/target) are required to optimally manage this growing clinical burden.

CONCLUDING REMARKS:

Given that other comorbidities (atherosclerosis, hypertension, coronary artery disease, etc.) commonly co-manifest with HF in patients with diabetes (Figure 3), it is thus now timely in 2020 to reconsider the definition of diabetic cardiomyopathy. Originally defined as impairments in cardiac structure and function independent of the macrovascular complications of diabetes, perhaps the phenomenon commonly referred to as “diabetic cardiomyopathy” should instead be known as “diabetic heart disease” or “diabetes-associated HF”. While the knowledge base around the mechanisms of diabetic heart disease has continued to expand, there remains several unanswered questions, by no means the least of which can lead us towards development of better therapeutic approaches for tackling this major health burden (and how these might be distinct from conventional approaches for systolic HF in the absence of diabetes). We in the diabetic heart disease research community await with excitement the elucidation of remaining unknown pathways, novel contributors and their interactions that will inform the basic mechanisms of diabetic disease, and new treatments likely to emerge in the coming decade that will arise from the intensive ongoing research activity.

NON-STANDARD ABBREVIATIONS AND ACRONYMS

AGEs

advanced glycation end-products

Ask1

apoptosis signal-regulating kinase-1

BP

blood pressure

CaMKII

Ca²⁺/calmodulin-dependent protein kinase II

CMRI

cardiac magnetic resonance imaging

DAMPs

Damage-associated molecular pattern molecules

DPP-4

dipeptidyl peptidase-4

GLP-1

glucagon-like peptide

EDP

end-diastolic pressure

ECM

extracellular matrix

EF

ejection fraction

ER

endoplasmic reticulum

FGF21

fibroblast growth factor 21

GFAT

glutamine fructose-6-phosphate aminotransferase

GRK2

G-coupled receptor kinase

HBP

hexosamine biosynthesis pathway

HF

heart failure

HFpEF

heart failure with preserved ejection fraction

HFrEF

heart failure with reduced ejection fraction

IFNγ

interferon-γ

IGF1

insulin-like growth factor-1

IL

interleukin

IRS1

insulin receptor substrate 1

IVRT

isovolumetric ventricular relaxation time

JNK

NH₂-terminal Jun kinase

LGE

late gadolinium enhancement

lncRNAs

long non-coding RNAs

LV

left ventricular

miR

microRNA

OGA

O-GlcNAcase

OGT

O-GlcNAc-transferase

p38-MAPK

p38 mitogen-activated protein kinase

PET

positron emission tomography

PI3Kα

phosphoinositide 3-kinase[p110α]

PKC

protein kinase C

RAAS

renin-angiotensin-aldosterone-system

RAGE

AGE receptors

ROS

reactive oxygen species

SERCA2a

sarco(endo)plasmic reticulum Ca²⁺-ATPase

SGLT2i

sodium glucose cotransporter-2 inhibition

snoRNAs

small, non-coding RNAs

STZ

streptozotocin

T1D

type 1 diabetes

T2D

type 2 diabetes

TFAM

mitochondrial transcription factor A

TGF-β

transforming growth factor-β

TNFα

tumor necrosis factor-α

UDP-GlcNAc

UDP-N-acetylglucosamine

VEGF

vascular endothelial growth factor