Extracellular microRNAs and endothelial hyperglycaemic memory: a therapeutic opportunity?

Abstract

Type 2 diabetes mellitus (T2DM) is a major cause of cardiovascular (CV) disease. Several large clinical trials have shown that the risk for patients with diabetes of developing CV complications is only partially reduced by early, intensive glycaemic control and lifestyle interventions, and that such complications result from changes in complex, not fully explored networks that contribute to the maintenance of endothelial function. The accumulation of senescent cells and the low‐grade, systemic, inflammatory status that accompanies aging (inflammaging) are involved in the development of endothelial dysfunction. Such phenomena are modulated by epigenetic mechanisms, including microRNAs (miRNAs). MiRNAs can modulate virtually all gene transcripts. They can be secreted by living cells and taken up in active form by recipient cells, providing a new communication tool between tissues and organs. MiRNA deregulation has been associated with the development and progression of a number of age‐related diseases, including the enduring gene expression changes seen in patients with diabetes. We review recent evidence on miRNA changes in T2DM, focusing on the ability of diabetes‐associated miRNAs to modulate endothelial function, inflammaging and cellular senescence. We also discuss the hypothesis that miRNA‐containing extracellular vesicles (i.e. exosomes and microvesicles) could be harnessed to restore a ‘physiological’ signature capable of preventing or delaying the harmful systemic effects of T2DM.

Introduction

Type 2 diabetes mellitus (T2DM) is a chronic, multifactorial, metabolic disease caused by a complex interplay among environmental and genetic factors 1. The number of patients with diabetes is increasing relentlessly in western countries, and is expected to reach 552 million by 2030 2. T2DM is a source of disability and morbidity, especially as a result of its vascular complications, which eventually lead to retinopathy, nephropathy, neuropathy, ischaemic heart disease and peripheral vasculopathy 2. Endothelial dysfunction (ED), chronic, low‐grade systemic inflammation and (probably) cellular senescence contribute to the development of severe vascular complications, and have been proposed as key therapeutic targets for T2DM 3, 4, 5. Large clinical trials 6, 7 have found that early hyperglycaemia can promote disease progression and late complications, perpetuating ED and vascular damage despite the achievement of improved glycaemic control, a phenomenon that has been called ‘metabolic memory’ 1. The term indicates the vascular damage that persists after glucose normalization, whereas the general, long‐term, harmful effects of diabetes (i.e. complications other than vascular) have been referred to as the ‘legacy effect’ 8. Different cell types are affected by metabolic memory, including endothelial cells (ECs), immune cells, smooth muscle cells and fibroblasts 1. The lasting molecular changes involving the endothelium could be termed ‘endothelial hyperglycaemic memory’. According to recent evidence, oscillating glucose levels may actually be more harmful, and induce more enduring effects on endothelial health, than hyperglycaemia itself 9.

A variety of mechanisms are involved in metabolic memory, including increased production of advanced glycation end products (AGEs), AGE receptor overexpression, increased anion superoxide formation, mitochondrial protein glycation, mitochondrial DNA damage, protein kinase C activation, and polyol pathway and hexosamine flux alterations 1. However, targeting these changes with new therapies has had limited success in slowing down disease progression and the development of complications 10, indicating that not all the imbalances experienced by people with diabetes can be addressed by therapies addressing single targets. Moreover, even though combined treatment with glucose‐lowering and lipid‐lowering drugs and antihypertensive medications has greatly improved diabetes management, it cannot prevent the eventual development of vascular complications, especially in patients with long‐standing disease 10, 11.

Genome‐wide association studies, linkage studies, candidate gene association studies and meta‐analyses have identified a number of genes involved in susceptibility to both T2DM and its complications 12. However, genetic testing cannot predict the clinical risk of vascular complications in T2DM with accuracy, suggesting that the cardiovascular (CV) complications of diabetes are only partially attributable to genetic predisposition 12, 13.

Recently, epigenetic mechanisms have been hypothesized to be a crucial interface between genetic and environmental factors to explain metabolic memory 11, 14, 15, 16. Hyperglycaemia can induce a variety of epigenetic changes that persist for days after normalization of glucose levels 11, 14, 15, 16, 17, 18, 19, 20, mainly through the involvement of inflammatory genes 17, 18. DNA methylation and post‐translational histone modifications (PTHMs) are the most extensively investigated epigenetic mechanisms involved in metabolic memory. Hyperglycaemia can affect the activity of PTHMs and DNA methyltransferases, and changes may become irreversible over time, explaining the long‐term harmful effects of metabolic memory 16, 17, 18, 19, 20.

Recently, further epigenetic mechanisms have been identified. Non‐coding RNAs, including microRNAs (miRNAs), have emerged as key factors in gene expression regulation and are likely to participate in metabolic memory modulation. More than 2000 human miRNAs have been identified to date, making them one of the most abundant classes of epigenetic regulatory molecules 21. MiRNAs were previously thought to act mainly as negative regulators of gene expression, by binding to the three‐untranslated regions of their target protein‐coding mRNAs in a sequence‐dependent manner 21; however, a growing body of evidence supports the notion that they are not only post‐transcriptional regulators of gene expression, but can directly repress or stimulate target gene transcription by directly binding to promoter regions, a phenomenon that has been called RNA activation 22. Moreover, miRNAs can target enzymes involved in DNA methylation and miRNA genes which, in turn, are closely regulated at the level of promoter methylation, transcription and processing 23. Although miRNA modulation in the bloodstream and tissues has been extensively studied in patients with diabetes 24, their involvement in the CV complications of diabetes has only recently been established conclusively 25, 26.

In the present paper, we review the latest data on miRNA changes in diabetes, address their potential relevance to the development of CV complications, and highlight the possible relationships among some affected pathways, altered molecular data and the major pathogenic factors (i.e. low‐grade inflammation and ED) that are involved in the vascular complications of T2DM. The possibility of erasing metabolic memory by restoring physiological miRNA levels using innovative therapies harnessing the miRNAs contained in microvesicles (MVs) or exosomes is also discussed.

Chronic, Low‐grade Inflammatory Phenotype and Type 2 Diabetes

Several age‐related conditions, including T2DM and CV diseases, share a chronic, low‐grade inflammatory state 3, 4, 5, 27. According to a recent, brilliant hypothesis, the build up of cells with a senescence‐associated secretory phenotype (SASP) could promote the development of diabetes and its vascular complications 5. Senescent cells are believed to accumulate during physiological aging, driving the development of age‐related diseases through chronic secretion of a variety of (SASP‐related) factors that contribute to inflammaging (i.e. the chronic, low‐grade, systemic inflammation that accompanies aging) 28. The SASP is capable of transmitting senescence (via the ‘bystander effect’) and of exerting harmful effects in a paracrine as well as systemic way. The inflammatory phenotype is characterized by persistent activation of the nuclear factor kappa B (NF‐kB) pathway, which induces transcription of a number of genes involved in inflammatory response modulation, including adhesion molecules such as VCAM‐1 and cytokines such as interleukin (IL)‐6 and tumour necrosis factor (TNF)α 29, 30. These genes are chronically activated in cells from patients with diabetes 14, 15, 16, 17, 18, 19, 20. The role of senescence in the development of the vascular complications of T2DM, and whether their establishment precedes or follows low‐grade inflammation and vascular complications are being extensively investigated 5, 31. In vivo senescence probably encompasses a spectrum of states ranging from a low to a high secretory phenotype, depending on its inducers (i.e. replication or hyperglycaemia) and cell types, among other factors. Epigenetic modifications leading to chronic inflammation have been described in ECs and immune cells of patients with diabetes even in the absence of replicative senescence biomarkers 14, 15, 16, 17, 18, 19, 20; however, most of the inflammatory mediators involved in the vascular complications of diabetes, which are induced in vitro by hyperglycaemia in ECs and immune cells, are the molecules released by cells bearing the SASP (i.e. NF‐kB, IL‐1, IL‐6, TNFα, VCAM‐1) 14, 15, 16, 17, 18, 19, 20, suggesting a causal role for them in the maintenance of the chronic, systemic inflammation that accompanies diabetes. A comparative analysis of gene (and pseudogene) expression in replicative and hyperglycaemia‐induced senescence could shed some light. Hyperglycaemia clearly promotes the acquisition of a proinflammatory cellular phenotype that may be defined as diabetes‐ (DASP) or hyperglycaemia‐associated secretory phenotype (HASP). Moreover, mounting evidence suggests an important role for the inflammasome platform in both T2DM and atherosclerotic disease 32, 33. The NOD‐like receptor (NLR)‐caspase 1‐IL‐1β cascade can be activated by endogenous metabolism or injury‐derived byproducts called damage‐associated pattern molecules, resulting in chronic secretion of inflammatory cytokines 34, 35. Strikingly, the inflammasome controls the transmission of the SASP senescence signal 30. Besides NLR activation, toll‐like receptor (TLR) activation has also been proposed to be involved in T2DM and its complications, supporting a role for innate immunity, and probably for microbiota, in the diabetic inflammatory milieu 36, 37. Remarkably, all lines of evidence point to the chronic, low‐grade inflammation typical of T2DM as a key therapeutic target 3, 38, 39.

Our group has recently published a pioneering study suggesting that some miRNAs may be part of the secretome of cells bearing the SASP 40, 41, 42. MiRNAs are expressed by all living cells and can actively be released or shed in the bloodstream and taken up in active form by receiving cells, acting as highly efficient systemic communication tools. Easy detection in serum and plasma makes miRNAs emerging, minimally invasive biomarkers of complex processes like age‐related diseases, including T2DM and CV diseases 43, 44. MiRNAs can be secreted or released by cells within small membranous vesicles (e.g. exosomes, MVs and apoptotic bodies), or packaged in HDLs or RNA‐binding proteins (e.g. Argonaute) 44. MiRNAs have been shown to be functional mediators capable of coordinating multiple pathways and of modulating virtually all cellular responses to environmental stimuli, according to each individual's genetic make‐up. Factors associated with diabetic complications, such as hyperglycaemia, ED, inflammation and senescence, can induce deregulation of epigenetic mechanisms, thus affecting circulating miRNA profiles. As a consequence, the expression of specific genes in receiving cells, especially ECs, fibroblasts, vascular smooth muscle and immune cells, may exhibit extensive changes even in the absence of other adverse stimuli (i.e. return to normoglycaemia), disseminating and possibly amplifying a pathological signature.

MiRNAs Involved in the Pathogenesis of Diabetic Complications and Metabolic Memory

After the initial metabolic insult, the pathways involved in diabetic vascular complications are complex, interlinked and self‐perpetuating 10. It is unlikely that a single druggable pathway can prevent their onset. Targeting a number of pathways to slow down the development of diabetic complications seems to hold greater promise, but has not proved effective so far, probably because the intricate connections among the mechanisms giving rise to such complications create redundancy 10. Since a single miRNA can target several genes, and multiple miRNAs share common targets, miRNAs are particularly suited to target processes and pathways at the ‘network’ level 45, and could prove effective in eradicating metabolic memory as well as multifactorial age‐related diseases and metabolism‐related diseases 46, 47. Moreover, recent findings indicate that circulating miRNAs, either carried by exosomes or bound to HDL/proteins, can provide efficient communication between different tissues and organs, suggesting that they can exert a remote action to regulate gene expression in target cells 44.

A miRNA array approach involving human aortic ECs has found that the expression of some miRNAs (miR‐125b, miR‐146a‐5p and miR‐29a‐3p) is associated with metabolic memory. Interestingly, these miRNAs are involved in the modulation of proinflammatory pathways and EC dysfunction 48. The demonstration that direct inhibition of miR‐125b expression, or miR‐146a‐5p upregulation, improves endothelial function, blunting NF‐kB signals, suggests that glucose‐induced changes in miR‐125b and miR‐146a‐5p are related to long‐standing activation of the NF‐kB pathway, and help perpetuate metabolic memory. These data strongly suggest that miRNA modifications could have a significant role in metabolic memory. They also provide a miRNA‐based explanation for the constitutive activation of the NF‐kB pathway, which is considered to be one of the main causes of ED in metabolic memory 14, 15, 16, 17, 18, 19, 20. Importantly, miR‐146a is the most extensively investigated inflammation‐related miRNA (inflamma‐miRNA) and senescence‐associated miRNA 49, 50. Under chronic stimulation, it is overexpressed in several cell types, including ECs and white blood cells, restraining inflammation and switching off acute inflammation after removal of the harmful stimulus 49, 50. Altered (increased or decreased) miR‐146a expression has been detected in several diabetic tissue and cell types exposed to hyperglycaemia 51, 52, 53. Moreover, it plays an important role in mitochondrial homeostasis 54, 55, possibly connecting aging‐ or hyperglycaemia‐induced low‐grade inflammation to mitochondrial alterations, which are a hallmark of the complications of diabetes and of cellular senescence 56, 57. Because a single miRNA can influence multiple features (i.e. low‐grade inflammation and ED) that are modulated by different pathways in different tissues, some circulating miRNA signatures during aging and in patients with the major age‐related diseases suggest that they participate in a complex cross‐talk among tissues and organs. Interestingly, miR‐146a is found in exosomes, and its content increases after bacterial stimulation, suggesting its systemic spread in some conditions 58, 59.

A recent meta‐analysis has found 40 circulating miRNAs, including miR‐21, miR‐29a, miR‐34a, miR‐103, miR‐107, miR‐126, miR‐132, miR‐142‐3p, miR‐144 and miR‐375, that are significantly deregulated in T2DM 43.

MiR‐126 is the most extensively studied miRNA in T2DM. Its best characterized biological function is to maintain vascular integrity, which promotes the mobilization of haematopoietic stem/progenitor cells and vascular cell survival 60. A number of reports have documented that miR‐126 is downregulated in plasma/serum, ECs and endothelial progenitor cells of patients with diabetes 42, 61, 62, 63, 64. Our group has reported that circulating miR‐126 increases both during aging and EC senescence, and that diabetes/hyperglycaemia abolish this trend 42. The recent demonstration that miR‐126 targets insulin receptor substrate (IRS)‐1 expression via PI3K/AKT signalling pathways suggests that it is involved in IR modulation 65. The key tissues regulating glucose homeostasis in response to insulin are liver, skeletal muscle and adipose tissue. IRS‐1 is a key insulin signalling protein, whose adipose tissue expression is reduced in humans and animals with T2DM, impairing downstream insulin signalling through the PI3K and AKT pathways, and resulting in reduced insulin‐stimulated glucose uptake 66, 67. Moreover, IRS‐1 is a target of miR‐126 in adipose tissue and hepatocytes 68. miR‐126 downregulation in the diabetic milieu could therefore be a compensatory mechanism counteracting the loss of insulin sensitivity 69. The reduced intracellular miR‐126 levels seen in cells cultured in hyperglycaemic conditions as well as in plasma from patients with T2DM could have a beneficial role in adipose cells and hepatocytes, reducing IR. However, because miR‐126 has pleiotropic effects, being also involved in the maintenance of endothelial function, its downregulation in ECs exposed to hyperglycaemic conditions promotes ED (Figure 1). In this scenario, the extracellular exchange of miR‐126 could have clinical relevance in T2DM progression. Interestingly, miR‐126 has also been detected in circulating exosomes/MVs of endothelial and adipose origin. MiR‐126 downregulation has been documented in exosomes from patients with diabetes 42, 61, 62, 63, 64; however, circulating miR‐126 does not seem to have predictive value for the development of CV complications in subjects with diabetes 51, 70, possibly as a result of the large number of factors and tissues contributing to its circulating levels (e.g. age, gender, medications and intrinsic response differences between cell types) 42, 51. Exosomes/MVs of endothelial origin might provide a more accurate source of information about EC health. A large prospective study of patients with stable coronary artery disease has shown that only MVs containing miR‐126 predicted a CV event over the following 6 years, whereas circulating levels were uninformative 71. Similarly, ECs exposed to hyperglycaemia release less miR‐126 into the culture medium 42; this is especially evident in miR‐126 found in vesicles, both exosomes and MVs (unpublished data from our laboratory). Convincing data show the possibility of microparticle exchange among ECs, where miR‐126 interchange can regulate SPRED‐1 expression, and consequently the proliferation status of receiving cells, a mechanism that is blunted in the diabetic environment 62.

Figure 1.

Pleiotropic effects of hyperglycaemia‐induced miR‐126 underexpression and insulin resistance. Hyperglycaemia‐induced underexpression of miR‐126 may have favourable effects on adipose cells and hepatocytes in patients with insulin resistance, because miR‐126 downregulates insulin receptor substrate (IRS)‐1, the key gene signalling insulin activation (a). Reduced miR‐126 levels in such a setting could thus increase cell survival chances; however, hyperglycaemia‐induced miR‐126 underexpression can exert harmful effects on EC, inducing upregulation of SPRED1 and PIK3R2, two of the most effective angiogenic pathway inhibitors (b). In turn, endothelial dysfunction promotes development of diabetic complications.

Recently, new molecular changes have been described in association with altered miR‐126 levels. Unacylated ghrelin protects diabetic mice from peripheral artery disease by restoring miR‐126 levels and consequently VCAM‐1, SIRT1 and SOD‐2 regulation, suggesting that miR‐126 could have anti‐inflammatory and anti‐senescence activity 72. A huge amount of data stresses the central role of miR‐126 in IR and endothelial homeostasis. The development of cell type‐specific delivery strategies could turn miR‐126 mimics into a therapeutic opportunity.

MiR‐21 is extensively studied in cancer, but recently a role for it in aging‐induced inflammation 73 and endothelial senescence has also been disclosed 74. Several reports have described its downregulation in serum and endothelial progenitor cells from patients with diabetes 61, 63, 70 and still lower serum levels in patients with diabetic complications 70. In contrast, a tissue‐specific increase in miR‐21 has been reported in different hyperglycaemic environments 75, 76, 77, including patients with diabetes with proliferative diabetic retinopathy 75, while tissue upregulation has been seen to promote renal fibrosis in diabetic nephropathy 76. Interestingly, the levels of circulating miR‐21 can predict the development of end‐stage renal disease 78. Moreover, it plays a crucial role in cardiac fibrosis and related heart failure in a mouse pressure‐overload‐induced model 79, an effect that seems to be mediated by miR‐21* contained in fibroblast‐derived exosomes 80. Pharmacological inhibition of miR‐21* in a mouse model of angiotensin II‐induced cardiac hypertrophy has attenuated the disorder 80; similar findings are emerging for diabetic cardiomyopathy 81. A list of circulating miRNAs differentially expressed in plasma, serum and blood‐derived microparticles from patients with T2DM compared with healthy subjects is reported in Table 1.

Table 1.

Circulating microRNAs differentially expressed in patients with type 2 diabetes and control subjects, and sample type.

miRNA	Expression in patients with T2DM versus control subjects	Sample type	Proteins targeted in recipient cells by miRNA transfer
let‐7a	Down	Plasma 82
let‐7f	Down	Plasma 82
let‐7i	Down	Serum 83
miR‐100	Down	Whole blood 84
miR‐124a	Up	Serum 85
miR‐125b	Down	Plasma 86
miR‐126	Down Down Down Down Down Down Down	Plasma 86 Plasma 87 Plasma61 Microparticles 88 Circulating microparticle, plasma 62 Serum 89 Plasma 70	SPRED1 62 IRS‐1 90 FGF2 90
miR‐1303	Up	Serum 91
miR‐130b	Down	Plasma 86
miR‐140‐5p	Up	Plasma 86
miR‐142‐3p	Up	Plasma 86
miR‐144	Up	Peripheral blood 92
miR‐146a	Up Up Down Down Down	Serum 85 Plasma 93 Serum 83 Peripheral blood 92 Serum 94	IRAK1 59 TRAF6 59 NFkB pathway 59
miR‐150	Up	Peripheral blood 92
miR‐15a	Down	Plasma 61
miR‐182	Down	Peripheral blood 92
mir‐186	Down	Serum 83
mir‐191	Down Down	Serum 83 Plasma 61
miR‐192	Down Down Up	Plasma 86 Serum 83 Peripheral blood 92
miR‐195	Down	Plasma 86
miR‐197	Down	Plasma 61
miR‐199a	Up	Plasma 94
miR‐20b	Down	Plasma 61
miR‐21	Down Down	Plasma 61 Plasma 70
miR‐222	Up	Plasma 86	ICAM‐1 95
miR‐223	Down	Plasma 61
miR‐23a	Down	Serum 83
miR‐23b	Down	Peripheral blood 96
miR‐24	Down	Plasma 61
miR‐26a	Down	Microparticles 62
miR‐27a	Up	Whole blood 97
miR‐28‐3p	Up	Plasma 61
miR‐29a	Up Up	Serum 85 Peripheral blood 92
miR‐29b	Down	Plasma 61
miR‐30d	Up	Serum 85
miR‐320a	Down Up Up	Plasma 61 Peripheral blood 92 Serum exosomes 97	IGF1 98 Hsp20 98 Est2 98
miR‐326	Up	Plasma 82
miR‐34a	Up	Serum 85
miR‐375	Up Up	Serum 85 Plasma 99
miR‐423‐5p	Down	Plasma 86
mir‐486	Down Down	Serum 83 Plasma 61
miR‐503	Down Up	Serum 100 Plasma 101	EFNB2 102 VEGFA 102
miR‐532‐5p	Down	Plasma 86
miR‐571	Up	Serum 91
miR‐661	Up	Serum 91
miR‐770‐5p	Up	Serum 91
miR‐892‐5p	Up	Serum 91
miR‐9	Up	Serum 85
mir‐96	Down	Serum 83

IRS, insulin receptor substrate; miRNA, microRNA; NF‐kB, nuclear factor kappa B; T2DM, type 2 diabetes.

Target proteins are reported only for those microRNAs whose transfer has been shown to regulate protein expression levels in recipient cells.

MiR‐21, miR‐126 and miR‐146a are three extensively studied miRNAs in relation to T2DM and its vascular complications, as they display altered circulating as well as tissue levels 11, 43, 51, 103; however, several other miRNAs, including miR‐1, miR‐16, miR‐125b, miR‐133, miR‐155, miR‐206, miR‐221, miR‐223 and miR‐503, have been associated with the vascular complications of diabetes 11, 14, 51, 104 through inflammatory pathway alterations and impairment of endothelial function 11, 14, 51, 104.

According to a recent interesting paper, intensive glycaemic control in streptozotocin‐treated mice is unable to reverse the deregulation of a large miRNA panel in the diabetic heart; in particular, 268 of 316 miRNAs remained dysregulated after intensive glycaemic control with insulin for 3 weeks 105, suggesting a strong role for miRNAs in diabetic cardiomyopathy and metabolic memory; informatics analysis then disclosed that the majority of dysregulated miRNAs were involved in inflammation, fibrosis, apoptosis and hypertrophy. An ex vivo study of left ventricle biopsies from patients with heart failure has found several deregulated miRNAs (miR‐34b/c, miR‐199b, miR‐210, miR‐223 and miR‐650) in individuals with diabetes compared with individuals without 106.

Epigenetic therapy has finally moved from the workbench to the clinic 107. A variety of pharmacological tools have been developed to target miRNA pathways 46 or exploit miRNAs for selective gene therapy 108. Promising in vivo results have been achieved in patients with CV disease, and progress is continuous 46, 47, 108. Several experimental strategies have been tested to deliver miRNA mimics or antagonists. Synthetic miRNA or pre‐miRNA duplexes, chemically modified to enhance stability and cellular uptake, have been loaded onto different delivery systems, including lipid nanoparticles with surface receptor ligands to improve tissue specificity. Adeno‐associated viruses and other viral‐based vectors are further well‐studied delivery methods 46. Antisense oligonucleotides complementary to the mature miRNA sequence, or ‘antagomiRNAs’, were the first miRNA inhibitors to be used in mammals 109. AntagomiRNAs were subjected to a number of chemical adjustments. Cholesterol conjugation via a 20′‐O‐methyl linkage in the 30 end, phosphorothioate linkage, 20′‐O‐methyl‐modified ribose sugar, 20‐′,40′‐constrained 20′‐O‐ethyl‐modified nucleotides, 20′‐O‐methoxyethyl and 20′‐fluoro and 20′‐fluoro/methoxyethyl are all modifications introduced to improve their pharmacokinetic and pharmacodynamic properties 46. Finally, locked nucleic acid (LNA)‐antagomiRNA technology has successfully been tested in an in vivo trial 107. The ribose moiety of an LNA nucleotide has been modified with an extra bridge connecting the 2′ oxygen and 4′ carbon, conferring higher stability, binding affinity and increased selectivity to complementary RNA 46.

Both miRNA antagonists and miRNA mimics, however, still have some technical, pharmacological and pharmacokinetic problems 46. MiRNA shuttling by exosomes or MVs is expected to overcome technical difficulties, providing a valuable, practical strategy for efficient delivery of corrective or protective miRNA signatures to target cells. Extracellular vesicles (EVs) are physiological cell‐derived nanocarriers that are immunologically inert if purified from a compatible cell source 110. Moreover, it has been shown that polymeric nanoparticles can be engineered to target certain tissues selectively 111. In particular, any nanoparticles designed to target the vascular endothelium could provide an attractive drug delivery tool. In this context, EV integrin expression patterns appear as the main determinants of vesicles tropism 112.

MiRNAs and Off‐target Effects of Diabetes Medications

Among the medications currently used to treat patients with T2DM, some molecules have shown better results in terms of protection against CV complications. For example, a number of clinical studies have shown that metformin, a hypoglycaemic agent, reduces the risk of myocardial infarction and all‐cause mortality compared with other medications 113, 114. The drug's off‐target molecular effects are not yet clear, but some interesting data suggest that it can mitigate endothelial senescence both in vitro 115 and in vivo 116, and that it exerts similar effects on the SASP through NF‐kB inhibition in oncogene‐induced senescence 117. Moreover, metformin has proven molecular in vitro efficacy against metabolic memory in ECs through SIRT‐1 activation 118 and an in vivo anti‐inflammatory effect in patients with diabetes and atherosclerosis 33, 119, 120. These data suggest the existence of shared molecular and epigenetic alterations in diabetes and aging that are probably related to low‐grade inflammation; such changes could provide other possible exploitable targets for T2DM treatment (i.e. the sirtuin family) 121, 122. In a recent study, circulating levels of miR‐140‐5p and miR‐222, two of the most extensively studied inflamma‐miRNAs that are altered in patients with diabetes, were reduced by 3‐month metformin treatment 86. Interestingly, miR‐222 has recently been shown in endothelial microparticles; its transfer can regulate ICAM‐1, which is impaired in a hyperglycaemic environment 95. MiR‐222 also has an important role in ED and atherosclerosis progression 123. Circulating miRNA profiling after human administration could offer key information on the off‐target effects of metformin. If its anti‐inflammatory/secretory activity is confirmed 33, it can be harnessed to design new drugs or miRNA‐based strategies.

At present, antihypertensive medications are among the most effective treatments for ED prescribed to subjects with diabetes 124. Blood pressure reduction confers the strongest protection against CV events in such patients 124. Angiotensin‐receptor blockers (ARBs) seem to be able to reverse metabolic memory in diabetes 125, 126.

Beyond the extensive molecular imbalances induced by high blood pressure on endothelial function and the inflammatory profile 127, a possible role for senescence and the associated epigenetic changes should also be considered. Hyperglycaemia and hypertension are strong individual inducers of senescence 127, 128, 129, 130. It is conceivable that high blood pressure and hyperglycaemia, combined with the characteristic, low‐grade chronic inflammation of diabetes, can accelerate the onset of senescence, which would otherwise develop later in life. Remarkably, there seems to be a partial overlap between miRNAs that are deregulated in patients with diabetes and in hypertension because the levels of miR‐21 131, miR‐126 132, miR‐146a 133, miR‐155 134, and a long‐coding RNA, which functions as a host transcript for miR‐221 and miR‐222 135, are affected in either condition, both in the circulation and in tissue. In particular, a disturbed flow can negatively regulate miR‐126‐5p and abrogate EC proliferation at predilection sites in response to hyperlipidaemic stress through upregulation of Dlk1 expression 136. Administration of miR‐126‐5p rescued EC proliferation at predilection sites and limited atherosclerosis; moreover, miR‐126 downregulation and the subsequent SPRED‐1 increase contribute to right ventricle failure in pulmonary arterial hypertension 137. Furthermore, combined treatment with an ARB and a statin has been shown to counteract the effects of both acute hyperglycaemia and acute hyperlipidaemia 138, and to reduce circulating miR‐146a/b and TLR4 signalling in patients with coronary artery disease 133. ARBs appear to be more effective than angiotensin‐converting enzyme inhibitors in modulating a panel of TLR4‐responsive miRNAs 139. Statins have a known pleiotropic anti‐inflammatory effect 140. Recently, a role for them has been proposed in telomerase and senescence regulation 140. Because miR‐146a increases during senescence 50, 141, attenuating IL‐6 release in both fibroblasts and ECs acquiring the SASP, it is conceivable that the anti‐inflammatory effect of statins is partly mediated by miR‐146a. Moreover, oscillatory shear stress is capable of upregulating miR‐21, which in turn targets peroxisome proliferator‐activated receptor‐α (PPAR‐α) in an autoregulatory loop, modulating flow‐induced endothelial inflammation 142. Fenofibrate, the only PPAR‐α agonist approved for human use, has shown great potential in diabetic retinopathy 143. Various mechanisms have been proposed to explain this off‐target effect 143. Fenofibrate also modulates miR‐199a and miR‐214 144, which play an important role in retinal neovascularization 145.

It is difficult to establish whether miRNA modulation after drug treatment is direct or mediated by other medication‐modified factors; however, since existing anti‐diabetic drugs can modulate miRNA levels, the topic deserves further investigation. The literature describing miRNA expression changes after administration of antidiabetic treatment is summarized in Table 2.

Table 2.

MicroRNAs expression changes in the bloodstream after treatment with currently used diabetes medications.

Treatment	Experimental procedure	Modulated miRNAs	Sample type	References
Metformin	Three‐month metformin treatment in patients with T2DM	↓ miR‐140‐5p ↓ miR‐222 ↓ miR‐192	Plasma	86
Angiotensin‐receptor blocker or angiotensin‐converting enzyme inhibitor + statin	Twelve‐month combined treatment with atorvastatin and telmisartan or atorvastatin and enalapril in patients with coronary artery disease	↓ miR‐146a/b ↓ miR‐31 ↓ miR‐181a ↓ miR‐16 ↓ miR‐145	PBMCs Plasma	133 139
Metformin + anti‐diabetic agents (dipeptidyl peptidase 4 inhibitors and glynides)	Glucose‐lowering treatment followed by clinical re‐evaluation at 12 months	↑ let‐7a ↑ let‐7f	Plasma exosomes	82

miRNA, microRNA; PBMC, peripheral blood mononuclear cells; T2DM, type 2 diabetes.

Therapeutic Potential of Exosome/ Microvesicle‐contained miRNAs and Metabolic Memory

One approach to overcome some technical problems of miRNA‐based treatment is to use physiological, human‐derived, and ready‐packaged EVs, either exosomes or MVs. EVs released from donor cells by shedding from the plasma membrane are commonly referred to as MVs, whereas those secreted by multivesicular endosomes are called exosomes 146. EVs contain mRNA, miRNAs, other non‐coding RNAs, and a variety of protein types. EVs can be transferred to recipient cells, where shuttled RNA can be functional 147, 148. The functional relevance of miRNA‐containing EV transfer has been described both in vitro and in vivo 147, 148. EVs, particularly exosomes, have attracted considerable interest for their potential use both as biomarkers and as vehicles for gene (or pseudogene) therapy 147, 148. They are found in the circulation in healthy individuals, and their number rises in several CV conditions associated with inflammation; a growing number of reports have been documenting a role for them in endothelial function regulation 149.

MiR‐146a and miR‐155 are involved in the vascular complications of diabetes 51, 150 and are probably the two best explored inflamma‐miRNAs 41, 49. A recent and innovative study has found them in exosomes released from dendritic cells after LPS (a TLR4 ligand) stimulation. In particular, exogenous miRNAs can reprogramme the cellular response to endotoxin, where miR‐155 enhances and miR‐146a reduces inflammatory gene expression 59. NLR and TLR pathways play a significant role in the pathogenesis of inflammaging and inflammation‐mediated ED 36. Endogenous TLR ligands activate the TLR pathway, inducing NF‐kB activation and promoting inflammation‐mediated ED 36, 39. It is conceivable that diabetic adipose tissue is a source of inflammatory exosomes with altered miRNA content. For instance, stimulation of human‐isolated adipocytes with LPS induces release of specific miRNAs into the culture medium 151. Interestingly, secreted miR‐155, miR‐221 and miR‐222, which play a role in the CV complications of diabetes, are shared between inflamed adipocytes and M1 macrophages 151. Evidence of EVs containing inflamma‐miRNAs that can modulate ED is already being published 64, 152, 153, 154, 155.

It is still unclear whether the EV content closely reflects the cell of origin or whether it may be altered also in the absence of major imbalances in parent tissue 147, 148. Evidence for both options has been provided 156, 157. Some discrepancies probably depend on the size of the EVs examined, as different molecular mechanisms regulate the sorting of molecules into exosomes or MVs 147. In any case, hyperglycaemia itself can induce epigenetic damage 14, 15, 16, 17, 18, 19, 20; the resulting EVs may thus have an altered content capable of propagating an ‘incorrect’ signature that modifies the epigenetic set‐up in receiving cells even after stimulus removal. This would perpetuate the insult despite glucose normalization (Figure 2), a phenomenon that could be defined as ‘epigenetic damage transmission’.

Figure 2.

Epigenetic damage transmission. Postulated mechanism. Extracellular vesicles (EVs) contain mRNAs, microRNAs (miRNAs) and other non‐coding RNAs, as well as a number of proteins. EVs can be transferred to recipient cells, where shuttled RNA can be functional. The endothelium uses EVs for physiological cell–cell communication (a). Hyperglycaemia can exert semi‐permanent epigenetic damage in endothelial cells (b). The resulting EVs may have an altered content capable of propagating an ‘incorrect’ signature that modifies the epigenetic set‐up in receiving cells even after stimulus removal; this would perpetuate the insult despite glucose normalization (c), which can be achieved through hypoglycaemic medications and/or lifestyle interventions.

Reports of altered EV content in diabetic humans or mice and of hyperglycaemia‐challenged ECs, fibroblasts, adipose, immune and pancreatic cells are increasingly frequent 62, 82, 95, 98, 102, 153, 155, 158, 159, 160, 161. Intra‐ and inter‐tissue horizontal miRNA transfer through exosomes or MVs appears to be an important phenomenon, especially for the vascular complications of diabetes. Imbalances in the content of pro‐ or anti‐inflammatory miRNAs (i.e. miR‐21, miR‐146a and miR‐155) and pro‐ or anti‐angiogenic miRNAs (i.e. miR‐126, miR‐320 and miR‐503) currently seem to be the most promising exploitable differences 62, 95, 98, 102, 153, 155, 160, 161, 162.

Exosomes derived from cardiomyocytes of Goto‐Kakizaki rats, a widely used T2DM model, have been seen to increase miR‐320 and reduce miR‐126 content. Their transfer achieved functional downregulation of target genes (e.g. IGF‐1, Hsp20 and Ets2) in recipient ECs, and miR‐320 overexpression inhibited endothelial migration and tube formation 98. Engineered exosomes, enriched with miR‐320 antagonists, have already been proposed as a therapeutic option to increase angiogenesis in the diabetic heart 98, 160. Moreover, high glucose induces NF‐kB binding to the miR‐503 promoter region and upregulates miR‐503 expression in ECs. NF‐kB further induces shedding of endothelial microparticles carrying miR‐503, inducing its transfer from ECs to vascular pericytes; integrin‐mediated uptake of miR‐503 in recipient pericytes reduces EFNB2 and VEGFA expression, resulting in impaired migration and proliferation 102.

Finally, proof‐of‐concept showing that miRNA‐rich exosomes secreted from fibrocytes can accelerate wound healing in diabetic mice has been provided 162.

Conclusions and Future Prospects

A range of interventions, including lifestyle modification and/or pharmacological treatment, can be harnessed to improve outcomes in patients with diabetes; however, they are not sufficient, alone, to prevent the onset of the long‐term disease complications. Epigenetic mechanisms, including DNA methylation, histone modifications and non‐coding RNA expression modulation, have tremendously expanded our knowledge of some basic mechanisms of metabolic memory. EVs containing miRNAs are emerging as ideal candidates to provide diagnostic and prognostic information about diabetes and its CV complications. Moreover, exosome/MV‐shuttling of miRNAs might provide a novel therapeutic approach to mitigate ED and inflammation in T2DM by trying to avoid or delay the harmful effects of diabetes on CV complications.

How do we go on from here? Further progress requires provision of two sorts of experimental data: (i) extensive comparative characterization of the nucleic acid (mRNA, miRNAs and other non‐coding RNAs) and protein content of exosomes/MVs from diabetic and healthy subjects; and (ii) the demonstration that chronic EV administration (chronic parabiosis) from a diabetic to a healthy mouse and vice versa is sufficient to induce and mitigate the CV complications of diabetes, to confirm the feasibility of ‘small balls’ therapy.

Conflict of Interest

None of the authors have a conflict of interest to declare.

F.P., F.O. and A.C. conceived the idea and have been involved in manuscript conception and drafting. A.G., A.C., V.D.N., G.P. and L.L.S. collected the literature, drew the figure and supervised the parts devoted to circulating miRNAs in T2DM and its complications. A.C. and A.D.P. supervised the parts addressing the pharmacological and clinical aspects of T2DM and its complications. S.G. and R.T. supervised the paragraph related to metabolic memory and revised the manuscript critically. All authors have given their final approval of this version to be published. All authors have read and approved the final manuscript.