Protective Factors and the Pathogenesis of Complications in Diabetes

Abstract

Chronic complications of diabetes are due to myriad disorders of numerous metabolic pathways that are responsible for most of the morbidity and mortality associated with the disease. Traditionally, diabetes complications are divided into those of microvascular and macrovascular origin. We suggest revising this antiquated classification into diabetes complications of vascular, parenchymal, and hybrid (both vascular and parenchymal) tissue origin, since the profile of diabetes complications ranges from those involving only vascular tissues to those involving mostly parenchymal organs. A major paradigm shift has occurred in recent years regarding the pathogenesis of diabetes complications, in which the focus has shifted from studies on risks to those on the interplay between risk and protective factors. While risk factors are clearly important for the development of chronic complications in diabetes, recent studies have established that protective factors are equally significant in modulating the development and severity of diabetes complications. These protective responses may help explain the differential severity of complications, and even the lack of pathologies, in some tissues. Nevertheless, despite the growing number of studies on this field, comprehensive reviews on protective factors and their mechanisms of action are not available. This review thus focused on the clinical, biochemical, and molecular mechanisms that support the idea of endogenous protective factors, and their roles in the initiation and progression of chronic complications in diabetes. In addition, this review also aimed to identify the main needs of this field for future studies.

Graphical Abstract

Graphical Abstract.

Essential Points.

• Chronic complications of diabetes should be reclassified from using the microvascular and macrovascular categories to those of vascular, parenchymal, and hybrid (vascular and parenchymal) origin

• The pathogenesis of diabetes complications often involves an interplay between systemic and tissue-specific risk and protective factors

• Endogenous protective responses may help explain the differential severity of diabetes complications, and even the lack of pathologies in some tissues, organs, and individuals

• An important limitation in the study of protective factors is the definition of the “protected” phenotype, and more extensive analysis are clearly needed in order to identify new protective pathways and elucidate additional mechanisms for those already identified

Diabetes causes myriad pathologies in many organ systems, which is not unexpected since it causes significant abnormalities in almost all metabolic pathways of cells. The chronic complications of diabetes are generally due to insulin deficiency or insulin resistance, in combination with persistent hyperglycemia and dyslipidemia, and disorders of other metabolic pathways. These complications of diabetes are responsible for most of the morbidity and mortality associated with the disease. In addition, the societal and financial costs related to diabetes are also mainly associated with the onset and treatment of these complications. Traditionally, chronic complications of diabetes have been classified into either macrovascular disease, which includes cardiovascular disease (CVD), or microvascular disease, which includes retinopathy (DR), nephropathy (DN), and neuropathy. However, it is clear that this artificial classification needs to be revised since there are many complications of diabetes that cannot be easily or correctly placed in only the microvascular or macrovascular categories. For instance, neuropathy is thought to be caused by both microvascular and neuronal/axonal cell dysfunction (1). Similarly, other complications, such as cognitive dysfunction, increased risk for certain malignancies, periodontitis, poor wound healing, and outcomes of viral, bacterial, and fungal infections, as witnessed during the COVID-19 pandemic, are either mainly due to parenchymal tissue damage or parallel vascular and parenchymal pathologies (2-9).

Thus, in this review, we have separated the chronic complications of diabetes into pathological categories of vascular tissues, parenchymal tissues, and hybrid tissues (Fig. 1). Vascular complications are those affecting pure or predominantly cardiovascular or vascular tissues, such as the arteries, renal glomeruli, the myocardium, and the vasculature of the central nervous system (CNS) (strokes). Parenchymal tissue complications, on the other hand, affect mainly nonvascular organ components, such as in malignancies, Alzheimer disease, and osteoarthritides. In osteoarthritides, diabetes may decrease bone remodeling, increase synovial inflammation, promote chondrocyte apoptosis, and alter extracellular matrix (ECM) structure in bone and cartilage (10). Additionally, a large number of chronic complications in diabetes is the result of combined abnormalities in both vascular and parenchymal tissues. The rationale for having a new perspective in viewing these complications resides in the diverse pathologies of organs, which are mainly caused by metabolic changes induced by insulin deficiency, insulin resistance, or both. Vascular and neuronal systems are clearly important in the development of many complications in diabetes since they connect and supply critical signals and nutrients to all organs. However, separating all complications into only the macrovascular and microvascular categories suggests that vascular abnormalities are the initiators of all chronic complications of diabetes. This is unlikely to be true since it is well-established that metabolites from hyperglycemia, and the loss of insulin's effects associated with diabetes, can induce cellular changes in parenchymal tissues despite normal vascular function. While metabolic changes induced by insulin deficiency, insulin resistance, or both are the major risk factors for diabetes complications, some patients with these risk factors are protected from complication development. The concept proposed in this review is that the pathogenesis of diabetes complications is due to an interplay of risk and protective factors, which strongly argues for the important role of local parenchymal tissue responses to diabetes, with and without vascular pathologies.

Figure 1.

Vascular and tissue pathologies of complications in diabetes. ECM, extracellular matrix; Smooth M. Cell, smooth muscle cell; CNS, central nervous system.

Early features of diabetes-induced tissue pathologies that are associated with hyperglycemia and dyslipidemia are commonly observed in blood vessels and peripheral nerves, which is not surprising since these cells experience the greatest degree of persistent elevations and fluctuating levels of various metabolites. Moreover, dysfunctions in parenchymal cells, such as the myocardium, renal tubules, peripheral nerves, and CNS, have been reported in parallel with early changes in the vasculature (11). Cellular and tissue damage can be due to direct chemical reactions between hyperglycemia, lipids, toxic metabolites, and components of various circulating and tissue proteins, nucleic acids, and carbohydrate structures. For instance, chronic elevations of oxidants and inflammatory byproducts of hyperglycemia can cause cellular damage, even though their short-term responses are critical in maintaining homeostasis for essential biological and physiological functions. Traditional studies on the pathogenesis of complications in diabetes have largely focused on risk factors, such as hyperglycemia, oxidative stress, inflammation, and advanced glycation end products (AGEs), activation of stress kinases, protein kinase C (PKC), and others (12). However, recent studies have illustrated the equally important roles of endogenous tissue protective factors, which may help explain the differential severity of these complications and even the lack of pathologies in some tissues (13, 14). While there have been numerous reviews focusing on the mechanisms of complications in diabetes and their corresponding risk factors, comprehensive reviews on the topic of protective tissue factors and their mechanisms of action are sparse. This review will thus focus on the clinical, biochemical, and molecular mechanisms that support the idea of endogenous protective factors and their roles in the initiation and progression of complications of diabetes. Due to space constraints, this review will not address in detail the role of genetics and epigenetics in the development of complications in diabetes, which remain unclear despite having been studied and reviewed extensively. We will not discuss the complexities of dyslipidemia or the clinical treatment of the various complications of diabetes, as these have been likewise reviewed in numerous recent publications, except when they are pertinent to the discussion of protective factors.

The Interplay of Risk Factors and Protective Factors in the Development of Chronic Complications of Diabetes

Many factors contribute to the initiation and progression of chronic complications in diabetes, as they result from the responses of various tissues to systemic metabolic changes over time (Fig. 2). This highlights the importance of reclassifying diabetes complications not just into the traditional microvascular and macrovascular categories, but also into pathological categories of vascular tissues, parenchymal tissues, and hybrid tissues. It is well-known that hyperglycemia is significant for the initiation and development of DR, DN, and neuropathy, although other systemic factors such as hypertension and insulin resistance may also contribute as well. In contrast, macrovascular complications such as CVD arise from multiple factors that may be equally as important as hyperglycemia, such as insulin resistance, dyslipidemia, hypertension, thrombosis, and systemic inflammation (3, 15). In addition, hyperglycemia may impose a greater risk for CVD in people with type 1 diabetes (T1D) than in those with type 2 diabetes (T2D) (3, 16). Apolipoprotein C3 is a risk factor for incident coronary artery disease in people with T1D, with serum apolipoprotein C3 being associated with increased insulin resistance and coronary artery calcification in T1D (17). It is likely that many of the systemic risk factors responsible for the initiation of these complications are not the same as those causing the pathologies as they progress to advanced stages. For instance, hyperglycemia is clearly the initiating factor triggering early DR, which is characterized by neuronal dysfunction and capillary pericyte apoptosis with microaneurysms. However, with progression to severe DR or proliferative DR (PDR), elevation of vascular endothelial growth factor (VEGF), which causes retinal neovascularization, is likely the result of hypoxia from capillary loss in the retina, and not directly due to hyperglycemia itself (18). This example of DR and its pathogenic factors illustrates that the pathologies of complications in diabetes arise from the interplay of both systemic and tissue-specific risk factors and protective factors, with all of these subject to potential genetic and epigenetic modification and changes at different stages of complications.

Figure 2.

Interactions of systemic and tissue-specific risk and protective factors in diabetes complications. AGEs, advanced glycation end products; APC, activated protein C.

The importance of risk factors, especially systemic risk factors, has been clearly established by many observational and interventional studies. Both the Diabetes Control and Complications Trial (DCCT) and its follow-up arm, the Epidemiology of Diabetes Interventions and Complications (EDIC), for T1D; and the United Kingdom Prospective Diabetes Study (UKPDS) for T2D, as well as many other subsequent studies, have strongly demonstrated that hyperglycemia is an important risk factor for almost all complications of diabetes (2, 19, 20). Likewise, targeting hyperlipidemia, insulin resistance, and hypertension have additive beneficial actions in CVD, DN, cognitive function, and other complications (Fig. 2) (21). Recent positive results from trials on sodium-glucose transporter-2 (SGLT2) inhibitors, glucagon-like peptide (GLP)-1 receptor (GLP-1R) agonists, and GLP-1R/glucose-dependent insulinotropic polypeptide (GIP) dual agonists have indicated that improvements in metabolic parameters, weight reduction, diuresis, and blood pressure (BP) can reduce CVD mortality and renal decline not only in people with T2D, but also in those without T2D (22-35). In particular, the studies showing the benefits of SGLT2 inhibitors on people without diabetes have provided strong evidence that improving glucose and fuel metabolism in the renal parenchyma and myocardium can improve cellular function and survival, even without the toxic effects of hyperglycemia (26, 32-35). Although the mechanisms of action of SGLT2 inhibitors are incompletely understood, they could include changes in cellular fuel metabolism, reduction in oxidative stress, enhanced mitochondrial bioenergetics, or increased protein synthesis (36). In addition to systemic risk factors, tissue-specific risk factors have also been observed in studies employing transcriptomics, metabolomics, and metabolic flux analyses on glucose and lipid metabolism in the kidney, eye, and nerve (37). Interestingly, increased urinary tricarboxylic acid (TCA) metabolites that were not upregulated in plasma-predicted DN progression, with an enrichment of glycolytic, fatty acid, and amino acid pathways that were different from those found in the retina and nerves. Likewise, in another DCCT/EDIC-based cohort study, tissue-specific mechanisms leading to CVD in diabetes were also explored. The investigators proposed that in people with either T1D and T2D, chronic hyperglycemia causes subclinical myocardial injury leading to leakage and exposure of myocardial proteins, including α-myosin, to the immune system. In patients with T1D with poor glycemic control, the aberrant adaptive immune system hyperresponds to the myocardial injury, causing a rise in CD4+ T-cells and the development of autoantibodies directed to α-myosin heavy chain as well as other cardiac antigens. This proinflammatory state was associated with elevated high sensitivity C-reactive protein levels, and the investigators found that positivity for ≥2 cardiac autoantibodies significantly increased the risk of both CVD and coronary artery calcification after decades (38).

At the tissue level, local responses can produce both risk and protective factors in the development of various complications (Fig. 2). The interactions between glucose and lipids with various cellular components, including those of the vasculature, can induce oxidative and inflammatory responses that in turn may contribute to chronic organ damage. In addition, high levels of intracellular free glucose levels can induce nonenzymatic reactions with amines in proteins or lipids, thus forming Schiff bases, Amadori products, and, later, AGEs (12). Many reviews have already addressed the various mechanisms involved in the injurious effects of hyperglycemia on local tissues, as well as their interactions leading to the development of various complications (12, 13, 39-41). However, local protective responses, such as the elevated expression of antioxidant enzymes or resolvins, can mitigate the spectrum of vascular complications through their antioxidative and anti-inflammatory actions (Fig. 2). For instance, people with T1D, unlike those with T2D, were generally observed to have normal pulmonary function despite the presence of hyperglycemia, hyperlipidemia, DR, and DN (42). Similarly, only 40% of those with DR will progress to severe PDR, even though 80% to 90% of individuals with diabetes will develop early lesions of DR (43). While the general prevalence of PDR is approximately only 35% to 40%, lesions of early DR could be present in as much as 90% to 95% of those with diabetes. In the same way, less than 35% to 40% of people with diabetes will progress to severe and clinically significant DN. However, early glomerular and tubular lesions of DN may also be present in a larger percentage of people with diabetes, if only these lesions could be visualized as easily as capillary abnormalities in the retina. These differential responses to diabetes and its abnormal metabolic milieu clearly support the importance of interplay between systemic and local tissue risk and protective factors in the pathogenesis of chronic complications in diabetes (Fig. 2).

A strong supporting case for the presence of protective factors are the epidemiological findings of the Joslin Kidney Study of the Natural History of Microalbuminuria, which demonstrated that while 30% to 35% of people with T1D or T2D are at risk for developing CKD, two-thirds of patients with T1D and microalbuminuria did not develop a significant decrease in renal function even after 15 years of diabetes duration. Intriguingly, a third of these patients experienced regression of microalbuminuria to normal urinary albumin excretion rate during follow-up, with no significant change in estimated glomerular filtration rate (eGFR) in the same time span, which could not be explained by the use of renin–angiotensin–aldosterone system blockers (44). In studies on people with T1D and DN, DN resistors have been shown to possess preferential efferent arteriolar constriction, as opposed to preferential afferent arteriolar dilatation that had an amplified effect of increasing intraglomerular pressure in the individuals with DN (45). Thus, while early lesions characterizing the development of complications in diabetes are mainly due 1 or 2 risk factors or affect only 1 cell type, the progression of these complications to advanced stages require multiple systemic abnormalities overcoming myriad endogenous protective mechanisms that reside in multiple cell types in an organ.

Thus, these and other clinical observations have demonstrated that while risk factors are essential elements for the development of complications in diabetes, the role of protective factors is equally as significant in modulating the severity of these adverse metabolic changes. We propose that there are myriad systemic and tissue-specific protective factors and mechanisms that can neutralize the disruptive actions of multiple toxic metabolites encountered in diabetes (Fig. 2). Similar to the profiles of risk factors, many of these protective factors, such as antioxidant enzymes and anti-inflammatory cytokines, may be present in the circulation and in all tissues. More importantly for this discussion, recent progress in this area suggests that each parenchymal tissue may have its own profile of specific protective responses, leading to differential pathologies observed in various complications (Fig. 1). These protective factors have been demonstrated to work even in the presence of hyperglycemia, hyperlipidemia, low-grade inflammation, and oxidative stress (13). In the next sections, we will demonstrate the clinical features of a typical “protected” T1D cohort (the Joslin 50-Year Medalist Study), as well as provide a general overview regarding the search for both systemic and tissue-specific protective factors.

Case Study of a “Protected” Cohort: The Joslin 50-Year Medalists

Clinically, there are multiple examples of mechanisms that are protective against the development of diseases. Some of the strongest clinical data on protective factors for complications in diabetes come from the Joslin 50-Year Medalist Study (“Medalists”), which has characterized more than 1000 subjects with T1D for ≥50 years (and some for >80 years). Despite this long T1D duration, a significant proportion (>85%) are free from DN (eGFR < 45 mL/min/1.73 m²), and almost half are free from PDR and CVD. Additionally, 32% retained detectable C-peptide levels >0.05 ng/mL, even though >90% of the Medalists have traditional human leukocyte antigen (HLA) DR3/4 risk alleles for T1D (46). The Medalists as a group consistently exhibited excellent glycemic control (median HbA1c of 7.1%), as shown by their clinical history and follow-up studies for over 5 to 15 years (46, 47). However, no association between glycemic control (HbA1c) and either DR (via fundus photography) or DN (through either clinically determined eGFR or histologically verified whole kidney specimens) was observed in the Medalists; likewise, blood pressure was not strongly correlated with DR in this cohort (14, 47, 48). Further studies are in progress to determine in detail whether glycemic variability may contribute to the presence of various complications of diabetes. In the Medalists, DR severity has a bimodal distribution, with 40% having no–mild DR, 11% having moderate–severe none diabetic proliferative retinopathy (NPDR), and 49% having PDR. In a Medalist subset that has been followed longitudinally for over 50 years, 1 group manifested PDR at <20 years of diabetes duration, while the other half mostly progressed only to mild NPDR even after >50 years of insulin-dependent diabetes, despite the fact that both groups of Medalists had excellent HbA1c levels (14). These clinical findings on lack of DR progression strongly support the idea that local protective factors are mitigating the toxic effects of hyperglycemia and other systemic risk factors on the development of DR.

Moreover, the Medalists, particularly those without CVD, were observed to possess normal levels of circulating and endothelial progenitor cells, both thought to be markers of endothelial repair after vascular injury and associated with CVD (49-51). Circulating monocytes expressing osteocalcin, which is strongly linked to vascular calcification, were also lower in Medalists without CVD than in those with CVD (52). Additionally, the Medalists possess elevated levels of cardioprotective high-density lipoprotein (HDL) cholesterol (median: 63 mg/dL) (46), and those without vascular complications exhibited significantly higher concentrations of medium-sized HDL particles, independently of total and HDL cholesterol levels. The Medalists without vascular complications also had higher levels of HDL-associated paraoxonase 1, an enzyme that inhibits atherosclerosis in animal models (53).

In addition, given that hyperglycemia and insulin deficiency have been linked to poorer bone health in patients with T1D, a surprisingly low rate of nonvertebral fractures, plus normal femoral neck, lumbar spine, and radial bone mineral density Z-scores (as measured by dual-energy X-ray absorptiometry), was found in the Medalists, clearly indicating that the Medalists were also spared from severe skeletal disease (54, 55). Finally, only 13.5% of Medalists were found to have periodontitis, which is a rate much lower than reported levels in age-matched patients with diabetes (56). Thus, given all the available clinical evidence, the Medalists are an ideal group for studying specific endogenous protective factors for complications in diabetes. Of note, similar cohorts have been described elsewhere, such as the Pittsburgh Epidemiology of Diabetes Complications, UK Golden Years Study, and Canadian Study of Longevity cohorts (57-60). Studies from all of these cohorts with long-duration T1D have strongly supported the idea that tissue protective factors play significant clinical roles in the development of various complications of diabetes. The fact that the Medalists are protected not just from the traditional microvascular and macrovascular complications of diabetes, but also from other complications such as bone disease and periodontitis, further provides support for our proposed reclassification of these complications into vascular, parenchymal, and hybrid categories.

Categories of Protective Factors

Since complications in diabetes arise as the sum of myriad pathological events due to systemic and tissue-selective risks, it is plausible to classify protective factors as exerting either systemic or tissue-specific actions. The number of protective factors is growing, as the concept of the importance of protective factors is becoming more widespread. The list is summarized in Table 1 (13, 53, 61-113).

Table 1.

Protective factors against chronic complications of diabetes

Systemic protective factors	Reference
Growth factors: Insulin, IGF-1/2, FGF, PDGF, VEGF, TGF-β	(13, 61-67)
High-density lipoprotein	(53)
Vitamins: vitamin C, D, E	(68)
Circulating antioxidant enzymes: catalase, SOD, glutathione peroxidase	(69-72)
Anti-inflammatory mediators: flavonoids, adiponectin	(73-75)
Proresolving mediators: omega-3 polyunsaturated substances, resolvins, maresin-2	(76-79)
Anti-inflammatory cytokines: IL-4, IL-10, IL-13	(80, 81)
Anti-AGEs	(82-84)
APC	(85, 86)
Tissue-Specific Protective Factors
IGF-1/2	(87, 88)
Retinol-binding protein 3	(89-91)
PDGF	(92-96)
Nitric oxide, 12, 13-diHOME	(97, 98)
VEGF	(97-105)
Glycolytic enzymes: glyoxalase 1, PKM2	(101, 106)
KEAP/NRF2	(68, 107-110)
Antioxidants: SOD, glutathione, vitamins, α-lipoic acid, carotenoids, coenzyme Q10	(72, 111-113)

Abbreviations: AGE, advanced glycation end product; APC, activated protein C; FGF, fibroblast growth factor; IGF, insulin-like growth factor; IL, interleukin; KEAP/NRF2, Kelch-like-ECH-associated protein/nuclear factor erythroid 2-related factor; MGO, methylglyoxal; PDGF, platelet-derived growth factor; PKM2, pyruvate kinase M2; SOD, superoxide dismutase; TGF-β, transforming growth factor-beta; VEGF, vascular endothelial growth factor.

Systemic Protective Factors

Early on, even as both the DCCT/EDIC (for T1D) and UKPDS (for T2D) studies have established the significance of systemic risk factors, these same landmark studies have also shown the concept of systemic metabolic protection by strongly demonstrating that systemic interventions improving glycemic control can decrease the long-term risk for multiple complications from diabetes. The relative benefit of improved glucose control in DCCT was largest for DR, reducing the risk of development and progression by as much as 76% compared with conventional glycemic therapy. In EDIC, further long-lasting systemic protection against both macrovascular and mixed parenchymal/vascular complications was observed, a phenomenon known as “metabolic memory” (114, 115). This was similarly observed during the post-trial monitoring phase of UKPDS, where, despite the disappearance of HbA1c and BP differences between the intensive and conventionally managed groups, the intensively managed group continued to display significant risk reductions in vascular complications for up to 10 years (15, 19). While this concept of “metabolic memory” is mainly based from the results of clinical trials, it was first reported in DR studies among dogs with diabetes that developed retinal capillary lesions after an initial 2.5 years of poor glycemic control, despite being subsequently followed by 2.5 years of good glycemic control (116). The mechanism for the development of metabolic memory, which can be either protective or pathogenic, has not been elucidated, although many have suggested a role for genetic or epigenetic contributions due to the chronicity of its development and resolution. However, other biochemical and biological processes can also contribute to the clinical manifestations of chronic complications in diabetes. For instance, systemic metabolic or vascular changes in diabetes are known to induce alteration of the ECM. Hyperglycemia and diabetes have been documented to cause basement membrane (BM) thickening and changes in the composition of the vasculature and of many epithelial tissues. These BM changes can significantly alter the function and even the survival, migration, and proliferation of attached cells, such as by binding to integrins with subsequent changes in intracellular signaling (117). Since the turnover of BM components such as collagen, and the normalization of ECM components, may take several years, ECM changes and remodeling may provide a different and additional molecular pathway to explain systemic protection and “metabolic memory,” without the need to invoke genetic or epigenetic changes (118). Other systemic protective mechanisms include circulating antioxidants or anti-inflammatory factors, such as vitamins C, D, and E, carnitines, flavonoids, and omega-3 polyunsaturated fatty acids and their derivatives, which can all be important in maintaining metabolic activities. There are large volumes of data to support the association between the consumption of food substances enriched with these protective factors and decreased risks for a variety of diseases that share similarities with complications in diabetes, including cardiovascular and kidney pathologies, strokes, and cancers (111, 112). In addition, deficiencies of these essential factors are also clearly related to multiple illnesses. However, the ingestion of these specific antioxidants, without evidence of deficiency and in purified supplements, has not been shown to be efficacious in delaying the onset or progression of complications of diabetes. This lack of efficacy of using isolated antioxidants, either individually or in combination, is not only observed with complications of diabetes but also with many other chronic diseases as stated above (111, 112). More detailed discourse regarding the differential efficacy of antioxidants ingested by food, as compared with purified supplements, will be discussed further in the section on antioxidants.

Another category of systemic protective factors are peptides or proteins including hormones, growth factors, cytokines, and other factors such as insulin, insulin growth factor (IGF)-1, fibroblast growth factor, and even transforming growth factor (TGF)-β, which can deactivate immune cells in inflammatory states. Insulin is an important example of this category, as it has powerful systemic effects on the metabolism of glucose, amino acids, and fatty acids. Insulin also has important tissue-protective actions ranging from the skeletal muscles, pancreas, and the liver to other organs such as the myocardium, brain, kidney, wound healing, and periodontal tissues, which will be discussed in detail in the succeeding sections (97). Another category of systemic protective factors that have gained recent attention are small circulating molecules such as cystatin C, bilirubin, and others, which have been proposed as important physiological modulators of oxidative stress and chronic inflammation in metabolic syndrome and diabetes (119).

Tissue-Specific Protective Factors

Aside from systemic protective factors, there are myriad local protective factors that act to balance the toxic effects of hyperglycemia and insulin resistance in affected tissues (Table 1). One well-established protective system at the local and cellular levels are the antioxidative stress pathways, which are critical for countering oxidative environments and the development of chronic complications in diabetes. As cells and biological systems are constantly exposed to both exogenously and endogenously produced oxidants via oxidative phosphorylation, signaling pathways, or mediators of cellular action, the neutralization of these oxidants is mediated by multiple pathways, including those of superoxide dismutase and glutathione. One powerful and well-established system to neutralize oxidants involves the Kelch-like-ECH-associated protein (Keap)-1/nuclear factor erythroid 2-related factor (Nrf)-2 pathway, in which toxins oxidize cysteines on Keap, facilitating its dissociation from Nrf2. Nrf2 subsequently accumulates and migrates to the nucleus, where it acts as a transcription factor in inducing the expression of many phase II antioxidative and antitoxin enzymes (120). This carefully regulated system can modulate oxidant levels in the various compartments of cells and prevent excessive metabolic byproduct accumulation. In the presence of diabetes and hyperglycemia, increased glucose metabolism through multiple pathways can elevate oxidant production in several subcellular compartments, including the cytosol, mitochondria, endoplasmic reticulum, and others, which can cause dysfunction in various subcellular organelles and enhance the development of chronic complications in diabetes. Therapeutic cocktails with 1 or multiple antioxidants have not succeeded in part because they may target a limited number of oxidants in extracellular or intracellular locations that could not be relevant to a given complication.

It is also likely that tissue-specific protective responses are responsible for the diverse and sometimes contradictory vascular presentations of complications in diabetes. One clear example are the vascular pathologies of PDR and DN vs that of the heart, peripheral limbs, and wound-healing processes (11, 65). Pathologically, excessive neovascularization is the classic definitive finding of PDR, while glomerular endothelial cell proliferation and angiogenesis are also noted in early DN (13). In contrast, inadequate neovascularization and collateral vessel formation around areas of arterial thrombosis are often observed and associated with myocardial ischemia and poor lower limb perfusion and wound healing in people with diabetes. Studies at the molecular level have shown that this discordance in angiogenesis in the retina and glomeruli, vs the peripheral limbs and myocardium, is related to VEGF expression in specific tissues. Additionally, studies, including those from our laboratory, have demonstrated that VEGF expression is elevated in retinas with severe DR or PDR—a protective response to retinal hypoxia from capillary closures in the macula, hyperglycemia-induced endothelial dysfunction, and pericyte apoptosis (65). The resulting retinal neovascularization forms immature capillaries without the usual 1:1 ratios of endothelial cells and pericytes in normally avascular areas, leading to loss of vision (65). However, in the peripheral limbs and myocardium, VEGF expression in response to hypoxia from atherosclerosis and excessive occlusion of the coronary or peripheral arteries is suboptimal, causing inadequate collateral vessel formation (11, 65). The mechanisms for these discordant responses to hypoxia in the retinal and myocardial tissues are not completely clear, despite being exposed to the same systemic factors. Differential regulation by insulin has been proposed as a possible mechanism for the diverse responses to hypoxia and hypoxia-inducible factor (HIF)-1α activation (121). Nevertheless, understanding the causes of these diverse tissue responses is critical in order to facilitate tissue-specific therapeutic interventions, which can otherwise lead to detrimental effects in other vital tissues. For instance, targeting angiogenic changes in the retina using systemic anti-VEGF agents will accelerate hypoxia for other peripheral organs. We suggest that the differences in the potential role of insulin in regulating VEGF expression as a possible tissue-specific protective action could be responsible for the discordant changes in VEGF expression between the retina and peripheral vascular tissues. The protective actions of insulin on the peripheral vessels will be discussed in the succeeding sections.

The idea that protective factors are equally as important as risk factors in the development of complications in diabetes has recently gained widespread interest and momentum, which included studies employing proteomic techniques in order to better characterize the presence of tissue-specific protective factors (Fig. 1). In the Medalists, we have recently used proteomic analysis of donated retina and demonstrated the presence of a tissue-specific protective factor, retinol-binding protein (RBP)-3, in the vitreous of retinal cells that were protected from the toxic effects of hyperglycemia and PDR—which will be discussed in further detail (90). Similarly, modification of DNA damage checkpoint proteins may also protect against DR (122). Krolewski and colleagues also recently observed that the combined effect of 3 circulating proteins in the plasma of independent T1D and T2D cohorts with DN (fibroblast growth factor-20, angiopoietin-1, and tumor necrosis factor/TNF ligand superfamily member 1) was associated with a very low cumulative risk for end-stage renal disease (ESRD), as demonstrated in individuals who had baseline concentrations above median for all 3 proteins. This protective effect was found to be independent of circulating inflammatory proteins and clinical covariates and was confirmed in a third cohort of individuals with diabetes without DN (123). In addition, rare variants in gene coding regions could also potentially have a greater impact on disease-related phenotypes than common variants through disruption of their encoded protein. A gene-based exome array analyses of 15 449 genes in 5 large cohorts with T1D and DN found that protein-coding variants in the hydroxysteroid 17-β dehydrogenase 14 gene (HSD17B14) had a net protective effect against development of ESRD at exome-wide significance (n = 4196). As this encoded sex steroid enzyme is a druggable target, this opens a potential new avenue for therapeutic development (124). Similarly, the Diabetic Nephropathy Collaborative Research Initiative, which assembled nearly 20 000 samples from T1D participants with and without kidney disease, found 16 new DN-associated loci at genome-wide significance, with the rs55703767 minor allele (Asp326Tyr) being protective against DN and being significantly associated with glomerular BM width, although the mechanisms for these genetic changes and the accompanying decreased DN risks have not been studied (125). However, in general, genome-wide scans of large populations of people with diabetes, including the Medalists, have not delineated reproducible risk or protective alleles for various complications. The mechanisms of each of the individual protective factors will be discussed in greater detail in the next section.

Mechanisms of Individual Protective Factors

Protective factors mediate their actions either by neutralizing the toxic effects of risk factors directly, or through alternative mechanisms indirectly. Examples of direct neutralizing effects are the neutralizing actions of antioxidants on toxic byproducts of glycolysis and oxidative phosphorylation, as well as chemical reactions between elevated levels of free glucose with other proteins, lipids, and DNA containing active reactive sites, such as primary amines. It is possible that many of these protective factors are able to exert their actions on the parenchyma alone without necessarily involving the vasculature, further supporting our proposed reclassification of diabetes complications. As stated in the first section, we will not provide a detailed discussion regarding the various pathways that have been demonstrated to be activated or changed by hyperglycemia or insulin resistance. However, we will provide a brief description of the pathways that are pertinent to the succeeding sections, which will discuss in detail their mechanisms of action and their roles in the development of chronic complications in diabetes.

In general, the toxic effects of hyperglycemia or loss of insulin action are mediated at the cellular level (Fig. 3). Excess nutrients are transported into cells by nutrient transporters such as glucose transporter (GLUT) 1 to 4 and fatty acid binding proteins. Elevated levels of free glucose, free fatty acids (FFAs), phospholipids, and cholesterol will be metabolized by vascular and nonvascular cells via glycolysis, β-oxidation, and oxidative phosphorylation, leading to multiple biochemical abnormalities that include altered oxidants, lipids, nicotinamide adenine dinucleotide (NAD)/reduced NAD (NADH), and glucose metabolites, which can either induce cellular signals or inhibit them (Fig. 3). In addition, these metabolites can produce active products such as diacylglycerol (DAG) and methylglyoxal (MGO), or activate a variety of other pathways. Increased glycolysis, without coordinated activation of oxidative phosphorylation via mitochondrial flux, can induce elevations of other pathways and their resultant products, thereby leading to cellular dysfunction. For instance, a byproduct of glycolysis at the triose phosphate step can produce excess MGO, which are reactive compounds that can cause intracellular damage. We and many others have reported that increased DAG levels can also activate PKC and lead to changes in a variety of cell signaling pathways (92, 126). Activation of PKC β or δ isoforms is known to phosphorylate stress kinases like p38 mitogen activated protein kinases (MAPKs) and phosphatases such as SHC-1 (127). We have reported that this cascade of events can dephosphorylate tyrosine kinases and reduce their activation by insulin, platelet-derived growth factors (PDGF), and other hormones and growth factors (92). Since these hormones and growth factors are survival factors, loss of their function will accelerate apoptosis of critical cells such as pericytes, endothelial cells, and podocytes (Fig. 3).

Figure 3.

Disruption of cellular signaling by risk factors. GLUT, glucose transporter; FFA, free fatty acids; FABP, fatty acid binding protein; B-oxidation, beta-oxidation; DAG, diacylglycerol; MGO, methylglyoxal; PKC, protein kinase C; SHC-1, Src homology 2 domain-containing protein tyrosine phosphatase 1; PI3K/Akt, phosphoinositide-3 kinase/Akt pathway; NF-KB, nuclear factor kappa B; NOX, NADPH phosphate oxidase; AGE, advanced glycation end-products; TLR, toll-like receptor; RAGE, receptor for advanced glycation end-products; ECM, extracellular matrix; TGF-β, transforming growth factor-beta; ET-1, endothelin-1.

Another source of cellular stress is the activation of the nuclear factor (NF) kappa-B (KB) pathway by increased intracellular oxidant production via glycolysis, mitochondrial dysfunction, or activation of oxidases such as reduced NADPH phosphate oxidases (NOXs). In addition, circulating cytokines and AGEs are known to activate NF-KB pathways to induce cellular changes and transcriptional processes that can activate inflammatory pathways and production of inflammatory cytokines in a vicious cycle. Thus, at each step at which cells and tissues are exposed to toxic nutrient byproducts and loss of insulin action, there are likely protective processes that have evolutionarily developed to mitigate these changes and achieve homeostasis, since maintenance of fuel metabolism is critical to the survival of an organism.

Insulin and Its Dual Roles on the Vascular Tissues

Possible protective roles of insulin

Recent studies have shown that insulin can regulate metabolism in many cell types. Aside from classically described cells such as hepatocytes, adipocytes, and muscle cells, insulin has also been reported to regulate metabolic and nonmetabolic activities in many cells, including fibroblasts, vascular cells, cardiomyocytes, neurons, pericytes, glomerular podocytes, epithelial cells, and many others (97, 128). For this discussion, we will focus on insulin's actions on vascular cells, since these are involved in many of the complications of diabetes. The biological actions of insulin range from regulation of glucose and amino acid metabolism to cellular proliferation and survival. However, insulin also possesses specialized functions, including the induction of nitric oxide (NO) production via activation of endothelial nitric oxide synthase (eNOS) and regulation of extracellular matrix proteins and VEGF in glomerular podocytes. These functions of insulin, in general, are protective for tissue function and cell survival (97). However, in certain conditions with insulin resistance, insulin's mitogenic actions may also exacerbate pathological processes such as restenosis, which will be described below (87).

All vascular cells and cardiomyocytes have insulin receptors (IRs) of varying levels of density. IR have been reported and studied in endothelial cells, vascular smooth muscle cells (VSMCs), cardiomyocytes, glomerular mesangial cells, podocytes, and pericytes, as well as in immune cells such as monocytes, macrophages, T-cells and neutrophils. In these cells, IR structure and signal transduction are similar to other well-studied cell types. IR is a dimer composed of α and β subunits, which can form hybrid receptors with IGF-1 receptors. IR signaling is mediated by at least 2 pathways: First, through activation of insulin receptor substrate (IRS) 1/2 and the phosphoinositide-3 kinase (PI3K)/Akt pathway, which mediates many of insulin's rapid actions including those on glucose and amino acid transport and metabolism. The second signaling pathway is mediated by the Src/MAPK cascade, which activates many of insulin's mitogenic actions that take hours to days to manifest (97).

In the vessel wall and myocardium, insulin actions mediated by IRS/PI3K/Akt pathway activation include increased expression and activation of eNOS, VEGF, and heme oxygenase-1 (HO-1), and decreased expression of vascular cell adhesion molecule (VCAM)-1. This pathway has vasodilatory, anti-inflammatory, antioxidative, and antiatherogenic effects (63, 129-131). Insulin also exerts its protective actions by blocking apoptosis through several mechanisms, such as repressing the transcription factor FoxO, inhibiting the proapoptotic molecule caspase-9, and increasing the antioxidant activity of HO-1 (132-134).

The most well-documented protective vascular action of insulin is its activation of eNOS by Akt phosphorylation in endothelial cells. Earlier studies by Quon et al and Baron et al have demonstrated that eNOS can be activated by insulin, increasing NO production and causing vasodilation (135, 136). These findings were thought to physiologically increase blood flow in the skeletal muscles (137). In endothelial cells, IR are able to transport insulin across the endothelial barrier without degradation—a process called receptor-mediated transcytosis—which, together with eNOS activation and expression, may regulate insulin sensitivity (138). The protective effects of insulin on the vascular wall were conclusively demonstrated by studying ApoE (−/−) mice with knockout of the IR gene targeted to the vascular endothelial cells (EIRAKO mice). We reported that loss of IRs and action in the endothelium significantly exacerbated atherosclerosis in these mice by almost 3-fold compared with ApoE (−/−) controls (63). This enhanced atherosclerosis developed without changes in systemic insulin levels, lipid metabolism, insulin sensitivity, glucose tolerance, or BP, and was eventually shown to be related to the loss of insulin-mediated VCAM-1 suppression via the IRS/PI3K/Akt pathway in endothelial cells, thus elevating VCAM-1 levels in the arteries of EIRAKO mice. This in turn increased the binding and uptake of monocytes, increasing inflammatory cell numbers in the atherosclerotic plaque (133) (Fig. 3).

The anti-atherosclerotic effect of insulin was confirmed recently with the results from mouse studies overexpressing IRS1 targeted to the endothelial cells IRS1/ApoE (−/−), which greatly enhanced insulin signaling and NO production with parallel decreases in the severity of atherosclerosis via activation of the IRS1/PI3K/Akt pathway, and increasing the expression of endothelin (ET)-B receptors in endothelial cells (131). As ET-B receptors can improve eNOS activation via the calmodulin pathway, this revealed a new mechanism by which insulin can enhance eNOS activation and consequently NO production. In turn, it may affect the pathogenesis of DN and DR, given that diminished eNOS activity worsened renal function in several rodent models of DN, and that activation of ET-B receptors in retinal vessels increased retinal blood flow, an important mechanism for delaying DR progression (139-141). Recently, insulin's actions on endothelial cells in activating eNOS have been shown to regulate the differentiation of perivascular progenitor cells into brown adipose tissue (BAT). Since the amount of BAT can increase calorie consumption especially with cold exposure, this action of insulin on endothelial cells can regulate body energy expenditure via IRS/PI3K/Akt activation. Furthermore, the increase in BAT also coincided with the release of BAT adipokines, including 12, 13-diHOME, which we have reported to interact with endothelial cells in increasing NO production and decreasing inflammatory cytokines to attenuate atherosclerosis risk (98). All these observations showed that overall insulin action at physiological levels has antiatherogenic actions, supporting the notion that increasing insulin action on the endothelium to produce protective factors, such as NO, 12,13-diHOME, and others, can decrease atherosclerosis in individuals with prediabetes, diabetes, and insulin resistance.

In addition to its favorable effects on atherogenesis, insulin can also promote wound healing by stimulating the transition of macrophage phenotypes from proinflammatory (M1) to anti-inflammatory (M2), a phenomenon that was found to involve inhibition of hyperglycemia-induced activation of p38, NF-KB, and STAT1 transcriptional activity through the IRS/PI3K/Akt pathway. Furthermore, the peroxisome proliferator-activated receptor-gamma signaling pathway was activated as well (142). The IRS/PI3K/Akt pathway is also central to insulin action in the brain, where it enhances neuronal growth, catecholamine release, regulation of ligand-gated ion channels, modulation of synaptic plasticity, and both neuronal and glial survival. The plurality of insulin's actions in the brain makes insulin resistance an important pathogenic mechanism for the development of cognitive decline and subsequent Alzheimer disease (143). Additionally, insulin's physiological actions in regulating appetite in the hypothalamus have been studied extensively and supported by IR deletion studies targeted to the various regions of the CNS in rodents (144, 145). Clinical trials are ongoing to determine whether the introduction of insulin to the CNS directly can improve cognitive function and other actions regulated by CNS.

Proliferative roles of insulin

Insulin also possesses proliferative or mitogenic actions, many of which are mediated via activation of the Src/MAPK pathway and usually at higher insulin concentrations. These include induction of ET-1 and plasminogen activator inhibitor (PAI)-1 expression and the migration, prevention of apoptosis, and even proliferation of pericytes and VSMCs (13, 138). In the past, the mitogenic actions of insulin have been associated mostly with adverse consequences including atherosclerosis, fibrosis, and thrombosis, since these disorders are associated with hyperinsulinemia and insulin resistance. This finding of insulin's dual role, that both hyperinsulinemia and loss of insulin action can exacerbate many of these pathologies, has resulted in the concept of selective insulin resistance (146). This idea proposes that hyperglycemia or elevated free fatty acids can activate PKC α, β, or δ isoforms, other stress kinases, and inflammatory cytokines to phosphorylate IRS2 and p85/PI3K and inhibit their signaling via the IRS/PI3K/Akt/mTOR pathway, and consequently the beneficial vasodilatory, anti-inflammatory, antioxidative, and antiatherogenic effects of insulin are blunted. At the same time, insulin activation of the MAPK pathway (which requires high insulin levels) is not inhibited in insulin resistance and may even be potentiated by stress kinases and inflammatory cytokines, thus explaining the association between hyperinsulinemia and atherosclerosis risk (127, 147).

A clear and striking example of the pathophysiological importance of selective insulin resistance is in the endothelium, as exemplified by the endothelial dysfunction observed in insulin resistance, diabetes, and other diseases associated with metabolic syndrome. Endothelial dysfunction develops due to loss of insulin-induced eNOS expression or activation via the IRS/PI3K/Akt pathway and the phosphorylation of eNOS (138). Hyperglycemia, elevated free fatty acids, and inflammatory cytokines can increase DAG-activated stress kinases and PKC. These stress kinases can phosphorylate serine or threonine residues in IRS1/2 and PI3K and inhibit their activity. When the PKCβ isoform was overexpressed in the endothelium, using the vascular endothelial cell cadherin promoter on ApoE (−/−) mice, we directly demonstrated that PKCβ activation in the endothelium affected atherosclerosis in insulin resistance and diabetes. When transgenic PKCβ ApoE (−/−) mice were fed a high-fat diet, they did not differ from transgenic ApoE (−/−) mice in terms of systemic insulin sensitivity, plasma lipids, or BP. However, insulin action in the endothelium and femoral arteries were selectively impaired with respect to PI3K/Akt and eNOS activation, and the severity of atherosclerosis in the aorta from the transgenic PKCβ ApoE (−/−) mice was significantly increased (92).

In addition to atherosclerosis, insulin's proliferative actions may also enhance the restenosis process. Clinical studies have shown that restenosis rates after coronary stents and angioplasty are much higher in people with T1D and T2D than in those without diabetes (148). Both hyperinsulinemia and insulin resistance have been associated with the restenosis process, since both insulin and IGF-1 can induce VSMCs to proliferate via the Src/MAPK pathway. We have shown that deletion of IR specifically in VSMCs caused a reduction of the restenosis response in mice on high fat diet with insulin resistance and diabetes. In contrast, deletion of IGF-1R appeared to increase VSMC proliferation. The mechanism for the contrasting actions of IR and IGF-1R deletion in VSMCs could be due to the fact that VSMCs have significant amounts of heterozygous (hybrid) IR/IGF-1R in addition to homozygous IR and IGF-1R. This can be biologically important since IR/IGF-1R hybrid receptors have diminished signaling via the IRS/PI3K/Akt pathway compared to homozygous IR receptors, which have a much greater effect on VSMC migration and proliferation than IGF-1R. Thus, reductions in IGF-1R can actually enhance restenosis (87). These results strongly indicate that IR, and not IGF-1R, contributes to elevated risks for restenosis in people with diabetes. Thus, therapeutic approaches to reduce restenosis should be focused on reducing the effects of IR, and not IGF1R, on VSMCs. These findings also indicate that IR or IR/IGF-1R can regulate glucose and fatty acid metabolism in VSMCs (87). The distributions of these homozygous or heterozygous IR receptors may affect the differentiation of VSMCs from contractile to proliferative or even inflammatory phenotypes, and thus affect the development of pathologies such as restenosis and atherosclerosis.

Furthermore, insulin signaling has a complex relationship with left ventricular remodeling, with studies indicating that either insufficient or excessive insulin signaling can both promote pathological cardiac hypertrophy. Mice that lacked IR or IRS1/2 in cardiomyocytes were shown to have enhanced left ventricular remodeling, especially after pressure overload (149, 150). In these animal models, identified mechanisms include unrestrained myocardial autophagy and loss of connexin gap junction proteins (150). However, on the other end, pathologic overactivation of insulin signaling in cardiomyocytes may also lead to left ventricular remodeling and heart failure after conditions such as ischemia, and has been shown to be mediated chiefly by IRS1, as opposed to IRS2 (151). This deleterious pathway has been linked to activation of protein kinase A and G-protein receptor kinase 2, leading to reduced β1-adrenergic and enhanced β2-adrenergic receptor activity (152). Nevertheless, major clinical trials such as ORIGIN and DEVOTE, which evaluated CV outcomes of exogenous long-acting insulins, have not shown a significant elevation of CV risks (153, 154).

In summary, numerous studies have demonstrated the dual pro- and anti-atherogenic effects of insulin on the endothelium, VSMCs, and even perivascular stem cells. In diabetes and insulin resistance, these anti-atherogenic effects are partly or selectively lost, either by changes in IR configuration as homo- or heterozygous receptors, or selective inhibitory actions of stress kinases on the IRS/PI3K/Akt pathway. By modifying these pathways, it may be possible to enhance insulin action on the endothelium, VSMCs, and perivascular stem cells (BAT) without affecting systemic metabolism, thereby decreasing restenosis and atherosclerosis and promoting other beneficial extravascular insulin actions on whole body energetics, cognition, and wound healing. However, all of these rodent studies of atherosclerosis have shown that diabetes exacerbated the severity of atherosclerosis without exhibiting the unstable properties of plaques, such as thinning of the plaque cap and decreased VSMCs and ECM with increased inflammation, as observed in people with T1D and T2D. Future studies will be needed to elucidate the role of insulin on the stability of atherosclerotic plaques. It is important to exercise some caution in translating the results of these experimental studies on insulin to humans, as no clinical trial has yet demonstrated an unequivocal or dramatic benefit in humans. However, additional analysis of the DCCT/EDIC study has shown that virtually all (96%) of the beneficial effect of intensive vs conventional therapy on progression of vascular complications was explained by the reductions in mean HbA1c (155). Furthermore, 20-year follow-up data from the Diabetes Mellitus Insulin Glucose Infusion in Acute Myocardial Infarction (DIGAMI-1) trial showed that intensified insulin-based glycemic control after acute myocardial infarction in patients with diabetes and hyperglycemia at admission had a long-lasting effect on longevity (156). More conclusive clinical studies are needed to demonstrate insulin's protective actions on cardiovascular and other tissues, such as the CNS, in a hyperglycemic environment.

Growth Factors

Levels of many factors and cytokines change homeostatically in accordance with the functional requirements of their target cells and tissues. However, the expressions of many growth factors or cytokines are affected by the chronic stress of diabetes, and their levels can be increased or decreased depending on the duration and stages of different complications. When discussing the pathogenesis of complications in diabetes, elevated levels of specific growth factors such as VEGF, TGF-β, connective tissue growth factor, and PDGF are usually associated with the pathogenesis of DR, DN, atherosclerosis, restenosis, wound healing, and others (65-67). However, in many cases, the increased or decreased expression of these growth factors actually represents physiological tissue-specific responses to hyperglycemia-induced injury. Furthermore, many inflammatory cytokines that have been traditionally associated with the pathogenesis of diabetes complications in fact play essential roles in protecting organs against acute stress. For example, transient elevations of interleukin (IL)-6 are observed after exercise or acute hypoglycemia, which would play a protective role (157, 158). On the other hand, chronic elevations of IL-6 and TNF-α in diabetes are associated with the presence of various vascular complications, with detrimental impacts on cells and organs (159). In the succeeding sections, we will discuss a few examples of the protective functions of these growth factors in neutralizing the toxic effects of hyperglycemia.

Platelet-derived growth factor

PDGF and its corresponding tyrosine kinase receptors are produced by many cells, including retinal endothelial tip cells, pericytes, and VSMCs. PDGF-α and β act via the tyrosine kinase receptor, which is expressed by pericytes during angiogenesis. PDGF is critical for angiogenesis, wound healing, and VSMC/fibroblast migration and proliferation, and excessive PDGF levels have been associated with atherosclerosis and associated platelet activation (93). Thus, PDGF can regulate vascular cell survival as well as suppress their proliferation (160). In diabetes, both decreased and elevated levels of PDGF have been reported depending on the tissues being studied: Decreased PDGF levels and action have been linked to poor wound healing in diabetes, especially with poor glycemic control (94). In contrast, elevated PDGF levels have been found in the vitreous of people with diabetes (67). In contrast, embryonic global PDGF knockout in mice is lethal in utero, with absent capillary pericyte development, leading to inadequate capillary development and fetal mortality (161). However, an increased number of microaneurysms and acellular capillaries, as well as an increased tendency for retinal neovascularization during ischemic retinopathy, mimicking diabetic retinopathy, were reported in mice with heterozygous deletion of the PDGF gene (18). These findings strongly indicated that loss of PDGF's protective actions may play an important role in the pathogenesis of PDR (95, 96). In contrast, PDGF-β deletion in neurons did not change pericyte coverage in the brain (162). These PDGF changes in the retina and vitreous suggest that PDGF resistance exists in the retina and could be pathologically significant. The mechanistic explanation for retinal PDGF expression being elevated in diabetes, and, yet, PDGF deletion paradoxically causing capillary pathologies similar to DR, was clarified by our previous reports that hyperglycemia inhibited PDGF activation of its tyrosine kinase receptors via the sequential activation of PKC-δ and MAPK, leading to SHC-1 upregulation and PDGF receptor dephosphorylation and apoptosis in the pericytes (163). This study also showed that hyperglycemia causes PDGF resistance in the retinal vessels. In VSMCs, SHC-1 activation or overexpression reduced PDGF-stimulated VSMC proliferation and migration as well as the severity of restenosis. However, in rodent models of diabetes or insulin resistance, SHC-1 expression and activity in the arteries was decreased, which may partly explain the accelerated migration and proliferation of VSMCs (106). Thus, the differential expression of PDGF in the retina, vs the peripheral tissues, again stresses the importance of local factors in influencing the toxic effects of systemic hyperglycemia. PDGF resistance is likely to be pathologically important in poor wound healing, as observed in people with diabetes and hyperglycemia. While therapeutically introduced PDGF appears to be efficacious for improving wound healing in people without diabetes, it is not so effective in studies involving diabetes, again probably due to the inactivation and dephosphorylation of the PDGF tyrosine kinase receptor (164-166).

Vascular endothelial growth factor

VEGF is an excellent example of a growth factor that is induced when localized tissue areas experience hypoxia, leading to the activation of the HIF-1α pathway and protecting the local tissues from ischemia (167). Various isoforms of VEGF and their corresponding receptors are clearly critical for survival since their global deletions are lethal (168). VEGF has also been shown to have neuroprotective effects, in which primary dorsal root ganglion cultures lacking VEGF-β or VEGF receptor (VEGFR)-1 showed elevated neuronal stress, while mice lacking VEGF-β developed signs of sensory neuropathy. In contrast, addition of VEGF-β to dorsal root ganglion cultures counteracted neuronal stress and stimulated neuronal survival, and mice selectively overexpressing VEGF-β in neurons were protected against distal neuropathy (99). The main regulator of VEGF expression is hypoxia, in which oxygen levels regulate protein levels of HIF-1α, that in turn, can regulate the expression of many genes including VEGF. However, metabolic changes such as activation of the pyruvate kinase M2 isoform (PKM2) and inflammatory cytokines may also regulate VEGF expression and action via HIF-1α–dependent and –independent pathways, which have been termed metabolic hypoxia (100, 101).

In diabetes, the changes in VEGF expression vary with tissue type and duration of disease. Elevation of retinal VEGF level is the best documented change induced by diabetes, as well as the best documented cause of serious and pervasive ocular complications of diabetes—diabetic macular edema and PDR (65). This elevation occurs in the late stages of severe DR or PDR in response to retinal hypoxia, which itself is the result of retinal capillary closure and loss due to hyperglycemic effects on endothelial cells and pericytes. Thus, elevation of VEGF levels in the retina is an expected response by the retina to protect itself from hypoxia. However, the ensuing neovascularization process is defective, since any new capillaries formed are mostly devoid of pericytes, are very leaky, and proliferate in critically avascular areas of the retina, such as the avascular macula. This will interfere with light transmission, especially when the new capillaries hemorrhage (102). The success of clinical usage of inhibitory antibodies to VEGF have clearly demonstrated the validity of VEGF being the primary major factor in causing severe DR (103).

In peripheral tissues such as the myocardium or extremities, VEGF expression or action is reduced, resulting in insufficient neovascularization or collateral vessel formation around areas of ischemia. In these circumstances, VEGF is needed for protection against ischemia or for improving the wound healing process (11). One major reason for the reduction of VEGF as a protective factor is the loss of insulin-induced VEGF expression via the IRS/PI3K/Akt pathway, which is selectively inhibited by the activation of stress kinases such as p38 MAPK or PKC (97). Insulin does not seem to regulate VEGF expression in the retina, as peripheral insulin has limited access to the retina owing to the tight retinal–blood barrier formed by the continuous endothelium of the retinal capillaries (169). This idea of tight junctions created by the retinal endothelial cells is supported by studies showing that insulin introduced systemically does not activation or induce IR receptor phosphorylation in the retina (170).

In the kidney, VEGF is mainly produced in podocytes (101). Deletion of VEGF in the podocytes is lethal and resulted in complete lack of glomerular endothelial and mesangial cells, supporting its role in the maintenance of these cells. Heterozygous VEGF knockout in mouse podocytes caused endothelial cell necrosis, endothelial cell fenestration and mesangial cell loss, podocyte foot process effacement, and eventually proteinuria and ESRD (104). These findings strongly support the idea that VEGF acts as a protective and survival paracrine factor in the glomeruli. While glomerular VEGF expression in the glomeruli appears to be elevated early on in DN, its expression eventually decreases in association with decreasing eGFR and increasing DN severity and fibrosis. In mice with diabetes, partial VEGF knockout in podocytes likewise resulted in glomerular cell apoptosis, glomerulosclerosis, and proteinuria (105). Hyperglycemia-mediated SHC-1 activation may also impair VEGF survival signaling, causing increased podocyte apoptosis and endothelial dysfunction in renal glomeruli (171). Furthermore, although the beneficial effects of insulin on the vasculature are partly mediated via VEGF upregulation, insulin resistance may cause impaired VEGF production in podocytes and subsequently promote DN, and deletion of IRs in the podocytes has resulted in severe glomerular pathologies (130, 172). One important paracrine protective effect of VEGF appears to be on the mitochondrial actions of endothelial cells. This was suggested by findings in transgenic mice, which overexpressed PKM2 targeted to the podocytes. These transgenic mice, which overexpressed PKM2 in the podocytes, were prevented from development of diabetes-induced mitochondrial dysfunction of the whole glomeruli and were also able to maintain in parallel increased VEGF expression in the podocytes (101, 106). While the etiology of the paradoxical VEGF changes in the kidney has not yet been fully elucidated, it is likely that true metabolic hypoxia is present in the glomeruli in the early phases of DN, thus leading to elevated VEGF; while in the later stages of DN, podocyte loss contributes to decreased VEGF (173). However, these findings suggest that the promotion, and not inhibition, of VEGF in the glomeruli in early DN may prevent or protect the progression of glomerular pathology in diabetes.

In the myocardium of rodents with diabetes and insulin resistance, the expression of VEGF, VEGFR-1, and VEGFR-2 also decreased significantly and likely can be normalized with improvement in insulin sensitivity or treatment. Likewise, 2-fold reductions in VEGF and VEGFR-2 were observed in the ventricles of patient donors with T2D compared with those without diabetes (11). The decreases in VEGF expression in diabetes or insulin resistance are likely due to the loss of insulin signaling in the myocardium. In mice with IR deletion targeted to the cardiomyocytes, VEGF expression was decreased, especially when hypoxia was induced by coronary vessel ligation. The reduction of VEGF in these mice resulted in poor angiogenesis and collateral vessel formation, similar to those observed in myocardial diseases among people with insulin resistance and diabetes (11). In cultured cardiomyocytes, insulin has been shown to induce VEGF expression via the IRS/PI3K/Akt pathway, which we have shown to be selectively inhibited by either hyperglycemia or other stress kinases (97). As previously discussed, 1 example of tissue-specific protection in diabetes is the contradictory presentation of increased neovascularization in PDR compared with the decreased capillary density in response to ischemia in peripheral vascular beds (11, 65). However, again, elevated VEGF levels in these various tissues are likely an appropriate response to poor blood flow or metabolic hypoxia. Thus, a complete understanding of the roles of various growth factors or cytokines is required before therapeutic approaches are undertaken to treat the various complex pathologies of complications in diabetes.

Other Potential Protective Growth Factors

Transforming growth factor-β

TGF-β has been reported to be elevated in multiple tissues in the presence of diabetes, and has been associated with complications in which fibrosis is a prominent pathological feature, such as DN and the myocardium (66, 174). Hyperglycemia, with associated elevation of oxidative stress and inflammatory cytokines and PKC activation, can induce TGF-β expression, which in turn may promote cellular hypertrophy and ECM biosynthesis in the kidney, myocardium, and other tissues, contributing to the observed fibrosis in those tissues (66). Treating db/db mice with a neutralizing monoclonal TGF-β antibody hindered glomerulosclerosis and delayed the impairment of renal function in animals with diabetes, although this had no effect on urinary albumin excretion (175). Another study showed that while a TGF-β receptor kinase activity inhibitor reduced glomerular fibrosis and kidney mRNA levels of PAI-1 and collagen I and III, it also did not reduce albuminuria (176).

However, TGF-β is also known to be a potent regulator of macrophage and lymphocyte activation, mediating anti-inflammatory actions (177). In addition, the mechanism of action of TGF-β is complex and affects many cell types. Thus, it is likely that hyperglycemia-associated overexpression of TGF-β in various tissues could be an endogenous response to the elevation of oxidants and inflammatory cytokines in vascular cells. This paradoxical role of TGF-β in vascular tissues is a challenge in drug development, and may entail targeted tissue-specific drug delivery—allowing TGF-β renal suppression without increasing inflammatory activity in other tissues. Recently, clinical trials of TGF inhibitors for pulmonary fibrosis have led to successful approval for clinical use (178).

Activated Protein C

Protein C, a central factor in the coagulation cascade, has been shown to prevent endothelial cell and podocyte apoptosis in the renal glomeruli. Protein C that is activated by an endothelial procoagulant factor, thrombomodulin (activated protein C; APC), is highly expressed in mouse glomeruli, although downregulated in mice with diabetes. In these mice, mitochondrial apoptotic pathways were activated by diabetes and modulated by APC via the protease-activated receptor-1 and the endothelial protein C receptor pathways. Thus, the reduction in APC led to increased apoptotic signals and the release of mitochondrial cell death–inducing factors (85). Although the renoprotective mechanism of APC is yet unknown, 1 study reported that APC-mediated suppression of lipopolysaccharides (LPSs) increased levels of the vasoactive peptide adrenomedullin and infiltration of induced nitric oxide synthase–positive leukocytes into renal tissue. The anticoagulant function of APC suppressed LPS-induced stimulation of various proinflammatory mediators such as angiotensin-converting enzyme-1, IL-6 and IL-18, possibly explaining its effect on renal hemodynamics (179).

Incretins

Incretins are hormones that decrease blood glucose levels in a glucose-dependent manner by increasing insulin and decreasing glucagon secretion. In humans, the most important incretins are GLP-1 and gastric inhibitory polypeptide (GIP). GLP-1 is produced by intestinal enteroendocrine L-cells and certain neurons within the nucleus of the solitary tract in the brainstem, and secreted in response to food consumption, while GIP is synthesized by K-cells in the small intestine (180). Endogenous incretins are rapidly degraded by dipeptidyl peptidase-4, as well as by neutral endopeptidase. Therefore, pharmacological compounds have been developed to either activate the GLP-1R or inhibit dipeptidyl peptidase-4; most recently, a dual GLP-1/GIP agonist has also been developed (181).

Recent human trials focusing on the effects of GLP-1R agonists on major adverse cardiovascular event outcomes in patients with T2D have achieved widespread interest. In the LEADER trial, CVD morbidity and mortality was reduced among liraglutide-treated participants with T2D with vascular complications (81.3% with preexisting CVD) (25). Other GLP-1R agonists, such as albiglutide, dulaglutide, semaglutide, and efpeglenatide, likewise met either superiority or noninferiority safety endpoints for CVD in the HARMONY, REWIND, SUSTAIN-6, and AMPLITUDE-O trials, respectively (27-29, 182). Most recently, tirzepatide, a novel dual GLP-1/GIP agonist, also did not show increased CV risk in the SURPASS trial (183).

The exciting results of these trials raise the question about mechanisms linking GLP-1R signaling to improved CVD. While improvements in BP, weight, and glycemic control are likely to be contributory, the role of decreased inflammation and postprandial lipemia, as well as possible direct actions of GLP-1R agonists on blood vessels, immune cells, platelets, and the heart, are still to be proven, although some data are already available (184). One such example demonstrated that GLP-1R agonists were associated with improvement of ventricular function in patients with acute myocardial infarction and impaired ejection fraction (185). Concerns on the safety of these compounds have been raised, however, such as acute gallstone disease and a nonsignificant increase in some cancers, including pancreatic cancer, in the LEADER trial (25). Additionally, it is also interesting to note that the GLP-1R is not only present in the gastrointestinal tract but also in the endothelium and kidney, where it may stimulate NO production (186, 187). This led to improved endothelial function and prevention of some renal pathologies in rodents with diabetes (188, 189). In endothelial cells, GLP-1 may likewise decrease the expression of VCAM-1 and TNF-α (188, 190). We also found a new mechanism for GLP-1, showing its inhibition of angiotensin II via the c-Raf/extracellular signal-related kinase 1/2/PAI-1 pathway in glomerular endothelial cells (191). These findings may well be part of the pleiotropic effects of GLP-1R agonists.

Endogenous Antioxidant and Anti-Inflammatory Networks

The antioxidant pathways in cells and organs are the most well-demonstrated protective systems in living organisms, since oxygen is the most pervasively used fuel in biological systems. There have been numerous reviews regarding the role of oxidative stress in the development of complications of diabetes (12, 39-41). Thus, we will focus on the most recent advances and the possible reasons for the difficulties in demonstrating the therapeutic efficacy of these antioxidants for complications of diabetes.

As previously described, hyperglycemia can induce oxidant production via multiple pathways (40). These frequently arise from the mitochondria where the citric acid cycle produces NADH and reduced flavin adenine dinucleotide, which act as electron donors for the electron transport chain. This in turn generates a proton gradient through the inner mitochondrial membrane, with intracellular hyperglycemic states providing excessive reducing equivalents for this process. The proton gradient inhibits the transfer of electrons from reduced co-enzyme Q (ubiquinone) to complex III of the electron transport chain. Consequently, electrons are transferred to molecular oxygen, causing production of superoxide (41). Superoxide is also produced in vascular cells through phagocytic NOXs, which are expressed in endothelial cells, VSMCs, and many other cells, and prefers NADH as a substrate (192). In rodents with diabetes, inhibition of NOX, specifically NOX1 and NOX4, protects against DN (193). Another study showed promising results of NOX-36, a pro-inflammatory chemokine C-C motif-ligand 2 inhibitor, in lowering HbA1c and albuminuria in patients with T2D (194). Neutralization of reactive oxygen species (ROS) can also be an effect of NO, although eNOS can conversely become an ROS source if an already pro-oxidant redox state supports oxidation of the eNOS cofactor tetrahydrobiopterin (195). This in turn results in electron transport uncoupling and subsequent release of superoxide. Furthermore, glycation processes involving elevated glucose levels and primary amines of proteins can form AGEs or oxidized lipids such as oxidized LDL, which can then induce inflammatory cytokines in various vascular and immune cells. Likewise, glycolysis itself can also form adducts such as glyoxalate, thereby inducing intracellular glycation and causing cellular dysfunction (12).

As previously described, there are multiple endogenous antioxidant systems, which include the synthesis of small molecules such as glutathione; oxidant-neutralizing enzymes such as catalase, superoxide dismutase, thioredoxins, thioredoxin reductases, peroxiredoxins, methionine sulfoxide reductases, and glyoxalase; and diet- or metabolically derived antioxidants such as retinol, lycopene, Vitamins C, D, and E, and many others. As previously mentioned, 1 major cellular antioxidant and antitoxin system is the Keap1/Nrf2 transcription factor system, which will be briefly described as the prototype antioxidant system, since several potential therapies for diabetic complications were developed based on the activation of Nrf2 (68). It has been demonstrated that Nrf2 activation is important in bodily defense against environmental oxidants and toxins, as it enhances over a hundred antioxidant and antitoxin enzymes, including HO-1, glutathione synthase, and other enzymes in the glutathione biosynthesis pathway (107). Cytosolic Nrf2 complexes with Keap1 and is ubiquitinated and degraded via proteasomes. In the presence of oxidants, cysteine residues in Keap1 are oxidized, decreasing the ubiquitination and degradation of Nrf2 and leading to the accumulation of Nrf2. Elevated levels of Nrf2 then translocate to the nucleus and form heterodimers with other transcription factors, inducing transcriptional activation by binding to antioxidant-responsive elements or electrophile-responsive elements (120).

Activation of the Nrf2 pathway has been clinically studied as potential treatment for DN. One example is bardoxolone methyl, a synthetic triterpenoid that reduced oxidative stress and inflammation in experimental models of diabetes involving DR, DN and neuropathy (108-110). Bardoxolone methyl activates Nrf2, which in turn activates numerous genes for antioxidant enzymes (196). A clinical trial of patients with T2D and moderate DN treated with bardoxolone methyl, failed to show a decrease in proteinuria, although eGFR was improved (109). Another trial comprising patients with T2D with severe DN similarly did not reduce the risk of ESRD or death from CVD with the compound; additionally, the trial revealed significant safety concerns (110). In an experimental DR model, Nrf2 may also protect against capillary and neuronal decay in retinal ischemia–reperfusion injury. This caused elevated retinal levels of superoxide and proinflammatory mediators, as well as leukocyte infiltration of the retina and vitreous, in Nrf2 (+/+) mice (197). Whether newer and more specific compounds targeting Nrf2 will alter vascular complications in diabetes is a topic for future studies (198-200).

Similar to the results of clinical trials on Nrf2 pathway activators, other large controlled clinical studies using a variety of antioxidants have also been unsuccessful when hard clinical endpoints were used as primary outcomes. While many antioxidants such as vitamins C and E, folic acid, and L-carnitine have shown some positive effect in initial small clinical studies or in animal models of diabetes, they almost always fail in large randomized clinical trials (111, 112). One possibility for this persistent failure of single targeted antioxidants could be due to their inability to fully neutralize the increased oxidant production induced by hyperglycemia, which arises from multiple intracellular and extracellular pathways. In this regard, the activation of Nrf2 appears to be a desirable approach. However, as described above, studies have shown that Nrf2 activation does not always lead to decreases in oxidative and inflammatory stress (109, 110, 197). Nevertheless, the various antioxidant and anti-inflammatory networks are a prime example of endogenous protective mechanisms both in conditions with and without diabetes.

Retinol-Binding Protein-3

Recently, we have been studying the Medalist cohort to identify endogenous protective factors that are suggested to exist, due to the Medalists’ clinical history of DR and DN being not correlated to recent levels of glycemic control (47, 48). As described above, analysis of DR provided the strongest evidence that endogenous protective factors may exist, since the severity of DR in the Medalists showed a bimodal pattern and did not correlate with HbA1c levels. Furthermore, longitudinal follow-up showed a significant subset of Medalists who did not progress from no–mild DR to moderate–severe DR even after 50 years of T1D, which again was not associated with glycemic control (14). Given the high number of Medalists with only no–mild DR even with elevated HbA1c for many years, we investigated the existence of potential protective proteins in their retina and vitreous using proteomic analysis (mass spectrometry). Bioinformatic analysis of the vitreous and retinal proteomes suggested that RBP3, a retinol transport protein or interstitial RBP, was upregulated in the Medalists who did not have DR, even in those who had elevated HbA1c (89, 90). RBP3 is a secreted protein, expressed mainly in the photoreceptors of the retina, and recycles trans- and cis-retinals between photoreceptors and retinal pigmented epithelium. Since trans- to cis-retinol regeneration is essential for light transmission, mutations in RBP3 have been found in families with retinitis pigmentosa and retinal degeneration (201). Using a custom-designed enzyme-linked immunosorbent assay specifically for human RBP3, the inverse association between vitreous RBP3 levels (1-5 µg/mL) and DR severity was confirmed in people with T1D and T2D (90, 202). In a follow-up study, vitreous RBP3 levels were not only associated with less severe DR, but were also correlated with reduced risk of progression to PDR (202).

However, RBP3 has not been shown to have other direct actions on retinal cells apart from binding and transporting retinols. Structurally, it has distinct domains: The first bears similarities to other RBPs, while the second possesses lipid-binding sites for retinal transport (203). Serum levels have not been reported, but can be a 1000-fold lower than in the vitreous due to the dilution of vitreous levels in the circulation. Intervention and prevention studies in rodent models of DR have shown that intravitreous injection and overexpression of RBP3, specifically targeted to photoreceptors, can inhibit expression of VEGF and inflammatory cytokines in the retina. Furthermore, VEGF-induced retinal capillary permeability was reduced in a dose-dependent manner when intravitreous recombinant human RBP3 (rhRBP3) was administered together or 1 day after VEGF injection. After 2 months of diabetes, a 3-day intravitreous injection of rhRBP3 reduced retinal capillary permeability in rats to nondiabetic levels and improved electroretinogram amplitudes of oscillatory potential. Moreover, RBP3 overexpression prevented acellular capillary formation and thinning of both total and individual (outer nuclear layer, inner segment ellipsoid + end tip) retinal layers. These effects were replicated with the generation of transgenic mice overexpressing RBP3 targeted to the photoreceptors (RBP3Tg mice) (90).

At the biochemical and cellular levels, rhRBP3 inhibited hyperglycemia- and VEGF-induced migration of bovine retinal endothelial cells, as well as VEGFR2 tyrosine phosphorylation, by binding to VEGFR2, thus reducing the binding of VEGF to its receptors. Studies in RBP3Tg mice with diabetes showed that the levels of inflammatory cytokines and VEGF were also reduced compared with wild-type (WT) mice with diabetes. This finding was replicated using bovine retinal Muller cells, in which rhRBP3 was able to reduce the expression of IL-6, TNF-α, and VEGF when exposed to elevated levels of glucose (25 mM vs 5.6 mM). This action of RBP3 was viewed as significant since Muller cells are postulated to be a major source of VEGF and inflammatory cytokines in diabetes. The action of RBP3 on Muller cells appears to be related to its binding to GLUT1, decreasing glucose uptake as measured by both 3-O-methyl-d-glucose and 2-deoxy-d-glucose uptake. The decreased glucose uptake in Muller cells was supported by the lowering of basal and maximal extracellular acidification rates when rhRBP3 was added to cells in culture. Overall, these findings suggest that RBP3 in the retina, where it exists at high concentrations, can bind to other retinal cells such as Muller cells and protect these from the toxic effects of hyperglycemia (90).

Comparing the eyes of Medalists and other patients with diabetes in terms of DR severity, studies found a steady decline in the RBP3/VEGF ratio, a possible index of protective capacity, from those with no–mild DR to those with PDR (91). In addition, retinal studies of several DR models in rodents have also shown that RBP3 expression was decreased. The mechanism causing the reduction of RBP3 expression in the photoreceptors has not been elucidated, although some studies have suggested that inflammatory cytokines or oxidants may be able to reduce RBP3 expression in cultured retinoblastoma cells (89). These promising findings suggest that RBP3 is clearly a protective factor and may serve as a potential therapeutic tool in preventing and delaying the progression of early DR. Furthermore, if ultrasensitive assays can be developed to measure RBP3 levels down to the pg/mL levels in the circulation, its plasma or serum levels can be used as biomarkers for DR severity or progression, which is not yet available at this time.

Metabolic Compensation for Hyperglycemia

Abnormalities of glucose, lipid, and amino acid uptake and metabolism have been described in almost all cell types as a consequence of diabetes and hyperglycemia, especially of long duration. Mitochondrial disorders are among the most common metabolic disorders in the presence of diabetes and hyperglycemia, and markers of mitochondrial dysfunction have been proposed to help stratify patients at risk for developing chronic complications of diabetes (204).

Cells derived from tissues exposed to hyperglycemia, or cultured cells exposed to elevated glycose levels, exhibit mitochondrial dysfunction and produce excessive amounts of hydrogen peroxide, thus becoming the main sources of free radicals that could shorten the life expectancy of cells and tissues (205). In diabetes, mitochondrial dysfunction increases ROS production, driving a proinflammatory response and PKC activation, and leading to selective insulin resistance with downregulation and upregulation of the PI3K/Akt and Src/MAPK pathways, respectively (206). These diabetes-induced mitochondrial perturbations can affect mitochondrial fusion, which normally enables protein complementation, mitochondrial DNA repair, and equal distribution of metabolites; as well as mitochondrial fission, which normally facilitates equal segregation of mitochondria during cell division, enhancement of mitochondrial distribution along cytoskeletal tracks, and isolation of damaged mitochondrial segments (207).

Major studies have suggested that mitochondrial superoxide production was the unifying mechanism in regulating major biochemical and molecular pathways (PKC, polyol flux, hexosamine flux, and AGEs) for the development of chronic complications in diabetes (39). The elevation of glucose metabolites, as a result of elevated intracellular or extracellular levels of free glucose levels, can be toxic to cell function and survival. As discussed above, studies in cells and transgenic mice overexpressing antioxidant enzymes have provided supportive evidence that agents with the ability to neutralize the adverse effects of these pathways (such as inhibitors of AGE formation, PKC activation, antioxidants, antifibrotics, anti-inflammatory cytokines, and stress kinases) can also prevent the pathologies of DN, DR, and even atherosclerosis. However, clinical studies using these similar compounds (bardoxolone methyl, folate, L-carnitine, ruboxistaurin, pimagedine) have provided very modest effects (109-112, 208, 209). One possible reason for the lack of success among these clinical studies—vs cell-based and animal studies—could be the fact that most nonclinical studies have utilized prevention approaches rather than intervention models, which are the basis for all clinical drug studies for chronic complications of diabetes. In addition, it is important to consider that mice with diabetes do not develop the advanced stages of complications observed in humans, such as PDR, ESRD, or significant nerve fiber loss. However, Bornfeldt and colleagues have developed a transgenic mouse model that closely mimics complex atherosclerosis in humans with T1D, by breeding LDL receptor-deficient mice with transgenic mice in which T1D can be induced at will (210). Antioxidant treatment can also potentially block both the damaging effects and the critical physiologic effects of mitochondrial hydrogen peroxide production, which is essential for intracellular communication, cell differentiation, autophagy, and organized responses to insulin and growth factor stimulation. Lastly, the causes of chronic complications are likely due to multiple pathways and may vary at different stages of complications. For instance, while anti-VEGF agents are very effective for PDR, they may not be as effective for early stages of DR, which is caused by other factors apart from VEGF (67, 211). While major trials such as DCCT/EDIC and UKPDS have demonstrated major beneficial effects on DR and DN with reduction of hyperglycemia, several other large trials, including the Action to Control Cardiovascular Risk in Diabetes (ACCORD), Action in Diabetes and Vascular Disease: Preterax and Diamicron Controlled Evaluation (ADVANCE), and the Veterans Affairs Diabetes Trial (VADT), have all previously failed to show lowering of coronary artery vascular complication risk in T2D individuals via targeting hyperglycemia and hypertension (2, 16, 20, 212-214). The Cholesterol Treatment Trialists (CTT), however, have observed that the proportional reduction in major vascular events with lipid lowering was similar in people with diabetes compared to those without diabetes (215). The mixed results observed in different studies could be due to the fact that multiple metabolic abnormalities are likely to be pathologically important in causing heart disease in T2D. A number of specific pathways mediating hyperglycemia risk for DR, DN, and neuropathy have been discussed earlier in this review. All of these abnormalities are associated with fuel metabolism, in which the mitochondria plays a central role. Thus, there has been a great deal of interest as to whether the activation or regeneration of the mitochondria could prove therapeutic in stopping, delaying, or reversing the chronic complications of T2D, or even T2D itself.

Before discussing the potential mechanisms that can protect intracellular derangements from hyperglycemia and the increased flux of free glucose and FFA metabolism, there are also adaptive changes in the plasma membrane that could mitigate the toxic effects of these nutrients. For tissues involved in various complications of diabetes, glucose transport across the plasma membrane is mediated by GLUT1, or possibly GLUT3 in neuronal cells (216). In some cells, GLUT1 expression can be regulated by ambient glucose levels at either the transcriptional or post-transcriptional steps. Exposure of VSMCs to elevated glucose concentrations, from physiological levels of around 5 to 25 mM in those with poorly controlled diabetes, have been reported to downregulate GLUT1 mRNA and protein expression by more than 50%, with parallel changes in glucose uptake (217). In contrast, similar exposure of bovine aortic endothelial cells to elevated glucose levels did not alter GLUT1 expression. These findings suggested that the differential regulation of GLUT1 could be important for various cellular responses to hyperglycemia. For instance, while endothelial dysfunction is an early finding in response to diabetes, changes in arterial VSMCs are not noted until later in the course of diabetes (97). Decreasing GLUT1 and glucose uptake via inhibitors of glucose transporters has also been shown to delay the severity of complications (218). One example of an endogenous modulator of glucose transport is RBP3, as described previously (Fig. 4).

Figure 4.

Contribution of risk factors and protective mechanisms to the development of diabetes complications. RBP3, retinol-binding protein 3; PP Shunt, pentose phosphate shunt; GSH, glutathione; O⁻, oxidant; PKM2, pyruvate kinase M2; FFA, free fatty acid; B-oxidation, beta-oxidation; AGE, advanced glycation end-products; NF-KB, nuclear factor kappa B; PI3K/Akt, phosphoinositide-3 kinase/Akt pathway; KEAP/NRF2, Kelch-like-ECH-associated protein/nuclear factor erythroid 2-related factor; IGF, insulin-like growth factor; VEGF, vascular endothelial growth factor; MAPK, mitogen-activated protein kinase; ECM, extracellular matrix.

With regards to FFA elevation, which is also commonly detected in poorly controlled T2D and insulin resistance, the role of transporters is much more complicated since there are many types of FFAs and FFA transporters. One common and well-studied FFA transporter is CD36, which is present on many vascular and inflammatory cells. CD36 is a scavenger receptor that can transport not only FFA but also other molecules such as pathogen-associated molecular patterns and damage-associated molecular patterns. As hyperglycemia and diabetes have been reported to increase CD36 expression in some cells, the possibility also exists that endogenous CD36 regulators in vascular and immune cells could act to modulate its activity (219).

At the intracellular level, 1 therapeutic approach to target mitochondrial dysfunction is possibly by mitochondrial activation through 5′ adenosine monophosphate–activated protein kinase (AMPK). AMPK activation is the major sensor for cellular energy needs and is activated in response to insufficient cellular energy supply by reduction of protein synthesis, activation of glucose uptake, and stimulation of mitochondrial biogenesis via peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) (220). Decreased and dysfunctional AMPK activation is observed in high-caloric states such as diabetes, although this model does not explain increased double strand DNA breaks seen in diabetes, nor does it explain reduced sirtuin activity induced by hyperglycemia (221). One important factor affecting sirtuin reduction is decreased NAD, which can be induced by poly(ADP ribose) polymerase activation in cells exposed to high glucose, thus adversely affecting mitochondrial homeostasis (222).

Among organs of interest in the area of chronic complications, AMPK reduction has been observed in the kidney, retina, myocardium, and nerves of animal models with diabetes (211, 221, 223-226). Moreover, berberine, an AMPK activator, partially mitigated high glucose and increased FFA oxidation in the hearts of rats with diabetes, while adiponectin was similarly shown to stimulate PGC1α, AMPK, and mitochondrial biogenesis after myocardial infarction in mice with diabetes, resulting in a lowered susceptibility to further myocardial injury (227, 228). Some drugs used in patients with diabetes, such as metformin, renin–angiotensin–aldosterone system blockers, and canagliflozin (but not dapagliflozin or empagliflozin) have been also reported to stimulate AMPK (229-231).

Recent studies derived from proteomic analysis of glomeruli in individuals with long-duration T1D with and without DN have suggested that maintaining and enhancing metabolic glycolytic flux may induce mitochondrial biogenesis and thereby prevent DN. We have shown that glycolytic and mitochondrial enzymes were elevated in DN-protected Medalists from glomerular proteomic analysis, and were protective against the toxic effects of hyperglycemia, even when these people have had T1D for ≥50 years (106). This finding was unexpected since most hypotheses involving mitochondrial dysfunction in diabetes have been attributed to overnutrition or increased glycolytic flux from hyperglycemia. However, metabolic enzymes in glycolysis (triose phosphate isomerase, glyceraldehyde 3-phosphate dehydrogenase, enolase, PKM2, and lactate dehydrogenase), polyol (aldose reductase [AR] and sorbitol dehydrogenase), MGO (glyoxalase, hydroxyacyl glutathione hydrolase), and oxidative phosphorylation (enzymes from mitochondrial complexes I and III) pathways were upregulated in the glomeruli of patients without DN. Specifically, the expression and activity of PKM2, was significantly upregulated in the glomeruli of the DN-protected participants compared to those who had DN, and was strongly correlated with eGFR. Pyruvate kinase catalyzes the last step in glycolysis, the dephosphorylation of phosphoenolpyruvate to pyruvate, and is responsible for net adenosine triphosphate production in glycolysis. Pharmacokinetic enzymatic activity of PKM2 in the DN-protected Medalist group was also significantly higher than in those with DN, although not different from those without diabetes, and may reflect the ROS-induced inactivation of PKM2 activity by oxidation of critical sulfhydryl groups. Metabolomic analyses of serum from the same DN-protected Medalists supported these proteomic results, as there were lower levels of glucose-6 phosphate, fructose-6 phosphate, lactate, and sorbitol in the glycolytic and AR pathways, as well as lower levels of metabolites and DAG in the lipid synthesis and MGO pathways (thus reflecting reduction due to elevated levels of glyoxylase) than in those with DN. Using [U-¹⁴C₆] glucose labeling in mice and podocyte cell lines, we observed that PKM2 activation could increase glycolysis and decrease the DAG synthesis pathway. Reports have shown that hyperglycemia can increase the levels of DAG and activate PKC pathways in vascular and renal glomerular cells (134). To demonstrate that normalized PKM2 activity protected against DN in mouse models, PKM2 was activated with a pharmacological compound (TEPP-46) that in turn normalized renal pathology in 2 different mouse models with diabetes. Even more excitingly, PKM2 activation not only decreased toxic metabolite accumulation from the DAG and MGO pathways, but also promoted mitochondrial biogenesis gene expression, membrane potential, and DNA content (106, 232).

In a follow-up study, we sought to replicate and extend these findings by evaluating levels of PKM2 and other glycolytic enzymes in both the glomeruli and plasma from individuals with T1D of shorter duration as well as those with T2D (232). In line with our results from the earlier study, 4 glycolytic enzymes (PKM1, PKM2, enolase, and triose phosphate isomerase) and a mitochondrial enzyme (MTCO2) were likewise upregulated in the glomeruli of T2D individuals with preserved renal function. Interestingly, plasma proteomic analysis of the DN-protected individuals was consistent with upregulation of glucose metabolism pathways (pentose phosphate, glycolytic, gluconeogenic, and pyruvate pathways), with a corresponding reduction in levels of known renal damage markers such as TNF-α receptors, cystatin C, β2 microglobulin, and neutrophil gelatinase-associated lipocalin. Metabolomic studies of the DN-protected Medalists’ plasma further supported these findings, showing reduction of metabolites that were part of the glucose/hexose (glucoronate, sorbitol, and inositol), mitochondrial (fumarate, maleate, and aconitate), amino acid (adenine and thymine), and purine degradation (urate and hypoxanthine) pathways. Finally, amyloid precursor protein, known to be causally linked to Alzheimer disease, also showed up as a potential protective factor against DN (232). It is important to note, however, that among those with shorter T1D duration, we were unable to replicate our previous findings of increased glycolytic proteins in the glomeruli of DN-protected Medalists, and that PKM2 and glyoxalase 1 protein levels were also not statistically different in T2D glomeruli compared to controls (232). Additionally, levels of enzymes do not necessarily reflect their activities since many glycolytic enzymes are allosterically regulated by metabolites of glucose (106).

The extraneural functions of amyloid precursor protein have not been characterized in great detail but experimental work supported its role as a protective factor for the kidney, eye, and platelets (233-235). Moreover, as amyloid precursor protein has been reported to possess antithrombotic activities, our results found a significant inverse correlation between its levels and those of prothrombotic proteins such as thrombospondin-2 and -4 and tissue factor (232). Both the thrombospondins and tissue factor are upregulated during specific stages of wound healing and tissue remodeling, and thrombospondin-2 has even been recently shown to be closely associated with the severity and progression of metabolic syndrome and nonalcoholic fatty liver disease (236, 237). In a nutshell, these findings from both our previous and current work strongly indicate that an increase or maintenance of glucose metabolic pathways in the renal glomeruli (and probably other organs) is important for preservation of mitochondrial biogenesis, and consequently, renal function.

The mechanism by which PKM2 exerts its actions is of particular interest. PKM2 is activated by the upstream glycolytic intermediate fructose 1,6 bisphosphate, and can be modified by many post-translational modifications, including oxidation, acetylation, methylation, and phosphorylation. It was largely reported that PKM2 can mediate its actions either through its tetrameric form with high enzymatic activities, or transport to the nucleus by its monodimeric form (238-244). Moreover, its tetrameric form is also involved in the maintenance of vascular integrity, and the regulation of cellular apoptosis and proliferation, independently from its enzymatic activity (238, 239). PKM2 has also been shown to be overexpressed in different cancer types (245-247). The amount of PKM2 rises in transformed cells, leading to an increase in anaerobic glycolysis and a decrease in oxidative phosphorylation in the mitochondria—a phenomenon known as the Warburg effect (239). Meanwhile, a recent study suggested that the less active PKM2 dimeric form is highly expressed in monocytes and macrophages of patients with coronary artery disease, forming a transcriptional complex with HIF-1α to promote inflammation (248). In contrast, activation of PKM2 into the tetrameric form, with TEPP-46 and another agonist, DASA-58, inhibited LPS-induced nuclear translocation and subsequent expression of IL-1β and other HIF-1α–dependent genes in macrophages. TEPP-46 is said to stabilize the PKM2 tetramer via stoichiometric interaction with fructose 1,6 bisphosphate–bound PKM2. The results suggested that PKM2 may serve as a critical determinant of LPS-mediated macrophage activation and inflammatory responses (239). In another study, a shift was observed from PKM1 to PKM2 in patients with heart failure, which was partially reversed following mechanical unloading (249). It has been reported that PKM2 tetramerization and activity can be inhibited by its oxidation at Cys358 (240). We found lower levels of sulfenylated and oxidized PKM2 in the renal cortex of DN-protected Medalists, which suggested that hyperglycemia-induced oxidative stress may reduce PKM2 activity. Moreover, in a intervention study from our lab, TEPP-46, a PKM2 activator, was administered to mice after 3 months of diabetes. After 6 months of diabetes, TEPP-46 reversed the pathology and abnormalities of DN, along with lowering of toxic glucose metabolites and improvement of mitochondrial function to nondiabetic levels (106).

Phosphorylation of PKM2 at S-37 and T-454 induces the Warburg effect by increasing nuclear accumulation of PKM2. It was reported that digoxin, a cardiac glycoside, could protect liver inflammation and damage in alcoholic and nonalcoholic steatohepatitis by binding to PKM2, thus decreasing HIF-1α transactivation (244). PKM2 was also observed to mediate cell progression in proliferating endotheilia cells (ECs) and vascular barrier function in quiescent ECs, via the suppression of p53 and NF-KB, respectively. These functions of PKM2 were independent of its enzymatic activity (238).

In diabetes, pyruvate kinase activity was observed to be decreased in skin fibroblasts from patients with T1D with DN compared with those without DN (250). In our laboratory, we showed that the glomerular PKM tetramer and dimer/monomer ratio was decreased in mice with streptozotocin (STZ)-induced diabetes. To further confirm these findings, we observed that phospho-PKM2 Y105 was increased in the kidneys of these mice, indicating a less active oxidated PKM2 isoform in diabetes. These studies suggested that PKM2 acts as a key player in various metabolic and inflammatory conditions by modulating glucose flux and the TCA cycle (106). It has been reported that the potent antioxidant sulforaphane limits mono/dimerization and nuclear residence of PKM2 accompanied by reduced HIF-1α levels, Stat3 phosphorylation at tyrosine 705, and IL-1β expression, while preserving high levels of cytosolic PKM2 tetramer with high glycolytic enzyme activity. Sulforaphane prevented glutathionylation of PKM2 in LPS-stimulated macrophages, which may have accounted for the reduced loss of PKM2 tetramer (251). Furthermore, a study in the renal tubules also showed that protection by the S-nitrosothiol-CoA reductase system, part of NO-based cellular signaling, was mediated by inhibitory S-nitrosylation of PKM2 through a novel locus of regulation, thereby balancing fuel utilization (through glycolysis) with redox protection (through the pentose phosphate shunt). The authors noted that S-nitrosothiol-CoA reductase protection was lost when eNOS was deleted (252). In diabetes models, ROS have been reported to mediate uncoupling of eNOS dimers to superoxide-producing eNOS monomers in endothelial cells (253).

Lastly, we have most recently also demonstrated that mice with STZ-induced diabetes and PKM2 overexpression specifically in podocytes (PPKM2Tg) experienced normalization of diabetes-induced impairments in glycolytic rate and mitochondrial function in the glomeruli, in concordance with elevated PGC1α and VEGF expression. Seahorse assay initially showed that after 7 months of diabetes, the oxygen consumption rate (OCR) and extracellular acidification rate of glomeruli in WT mice with diabetes were both significantly reduced, indicating that the impairments in mitochondrial function and glycolytic flux occurred during advanced stages of DN, which were improved in PPKM2Tg mice with diabetes. It is important to remember that mouse models could not fully mimic DN in the human kidney, especially for C57 mice which are very resistant to DN. These have been discussed in the guidelines of the Animal Models of Diabetic Complications Consortium (AMDCC) regarding the mouse model of DN (254, 255). Interestingly, restoration of VEGF expression improved glomerular maximal mitochondrial function in WT and PPKM2Tg mice with diabetes, while inhibition of VEGF by anti-VEGF neutralizing antibodies suppressed the OCR improvements in OCR in the same mice. Specifically, VEGF protein was increased by 3-fold in the glomeruli of the DN-protected group vs those who had DN. PKM2 protein was also increased by 1.5-fold in the whole glomeruli, and total pyruvate kinase activity increased by 1.3-fold, in the PPKM2Tg mice. These data suggested that enhanced VEGF expression in PPKM2Tg mice can improve mitochondrial metabolism not only in the glomeruli but also in other cell types such as endothelial cells, even with exposure to diabetes of long duration. Mechanistically, the preservation of mitochondrial function and VEGF expression was dependent on tetrameric structure and enzymatic activities of PKM2 in podocytes, unlike in ECs. These findings illustrated that PKM2 structure and enzymatic activation in podocytes can preserve glomerular mitochondrial function against glucotoxicity via paracrine factors such as VEGF (101).

Thus, based on all these observations, our hypothesis is that increased expression or activation of PKM2 in the kidney elevates glucose flux via glycolytic pathways to the TCA cycle, lowering toxic metabolites. This has been confirmed by our in vitro studies, in which the restoration of mitochondrial OCR and extracelluar acidification rate (ECAR) was no longer observed when PKM2 was changed to its inactive enzymatic form. We also propose that an increase in glycolytic flux and AR activity can lower intracellular free glucose levels and reverse hyperglycemia-induced mitochondrial dysfunction and decrease oxidant production, rather than mitochondrial dysfunction as the primary defect causing DN. However, it is also important to note that AR has a high affinity and enzyme activity for other substrates, including several glycolytic intermediates such as glyceraldehyde-3-phosphate and its degradation product, MGO (256, 257). Gene profiling of ROS-related enzymes has shown that diabetes increased p47phox, Nox2, and Nox4 mRNA levels in WT mice. However, in PPKM2Tg mice, elevations of these ROS-related genes were mitigated compared to WT mice with diabetes, suggesting that diabetes generated lower ambient ROS in the glomeruli of PPKM2Tg mice. PKM2 upregulation may protect the kidney from the toxic effects of hyperglycemia and even alter hyperglycemia-driven mitochondrial dysfunction and apoptosis of podocytes. In addition, PKM2 activation may be also protective in other tissues such as the retina, where its elevation has been reported to decrease photoreceptor degeneration (258). Likewise, PKM2 activation may decrease the formation of M1 macrophages and increase endothelial tight junctions, thus decreasing inflammation and capillary permeability, both of which are exacerbated in diabetes and are related to the severity of chronic complications including DR (239). This reversal of mitochondrial dysfunction further suggests that epigenetic abnormalities, which are known to cause mitochondrial dysfunction in diabetes, could be potentially reversed (259, 260). In fact, transgenic overexpression of the long noncoding RNA taurine-upregulated 1 specifically in podocytes has been shown to ameliorate DN in mice. Taurine-upregulated 1 was noted to regulate mitochondrial function in podocytes by epigenetic targeting of expression of the transcription factor PGC1α, thus improving mitochondrial energy balance and delaying DN progression (261).

Conclusions, Limitations, and Areas of Uncertainty

In summary, there are many different systemic and tissue-specific protective factors that may act directly or indirectly in preventing and ameliorating the development and progression of vascular and parenchymal tissue complications in diabetes. While traditional pathways that protect many tissues—such as the role of insulin, growth factors, and antioxidant pathways—have already been established in the literature, recent findings from unique clinical cohorts like the Medalists have enabled a better and deeper understanding of previously unknown phenomena, such as the novel protective roles played by RBP3 and the glycolytic and TCA cycles in DR and DN, respectively. Despite hyperglycemia being the main metabolic feature in diabetes, insulin resistance, altered insulin signaling, and dyslipidemia, the availability of new drugs with multiple mechanisms of action, such as SGLT2 inhibitors and GLP-1R agonists, has increased therapeutic options for patients, and suggests that there are unknown pathways that can protect different organs from being damaged by the metabolic stresses of diabetes. However, despite the tremendous amount of scientific progress and increasing emergence of available tools for biomedical research, many challenges remain in the study of chronic vascular and parenchymal tissue complications in diabetes. While multiple recent studies have established the important role of both systemic and tissue-specific protective factors in the development of these complications in several organs, more extensive analyses are clearly needed in order to characterize new protective pathways and elucidate additional mechanisms for those already identified. An important limitation in these studies is the definition of the “protected” phenotype, as oftentimes, investigators have to settle for intermediate and/or indirect biological markers such as eGFR and albuminuria for DN or the presence of neuropathic pain for neuropathy, given that ideal and/or direct markers are sometimes impossible to obtain (eg, performing large numbers of renal biopsies for research purposes). More extensive studies are needed to further validate the novel hypotheses proposed by existing literature, and more detailed investigations are required to clarify the molecular mechanisms regulating these protective pathways, in order to explain their activation in specific subsets of people with diabetes. The differential activation of these protective pathways will be complex and are unlikely to be purely due to changes in DNA sequences As shown by this review, these new findings serve as important starting points and stepping stones in the development of future agents that may further help prevent and delay the onset of diabetes-related complications.

Abbreviations

AGE

advanced glycation end product

AMPK

5′ adenosine monophosphate–activated protein kinase

APC

activated protein C

AR

aldose reductase

BAT

brown adipose tissue

BM

basement membrane

BP

blood pressure

CNS

central nervous system

CVD

cardiovascular disease

DAG

diacylglycerol

DR

diabetic retinopathy

DN

diabetic nephropathy

ECM

extracellular matrix

eGFR

estimated glomerular filtration rate

eNOS

endothelial nitric oxide synthase

ESRD

end-stage renal disease

ET

endothelin

FFA

free fatty acid

GLP

glucagon-like peptide

GLP-1R

glucagon-like peptide-1 receptor

GLUT

glucose transporter

HDL

high-density lipoprotein

HIF

hypoxia-inducible factor

HO

heme oxygenase

IGF

insulin growth factor

IL

interleukin

IR

insulin receptor

IRS

IR substrate

LPS

lipopolysaccharide

KB

kappa-B

Keap

Kelch-like-ECH-associated protein

MAPK

mitogen activated protein kinase

MGO

methylglyoxal

NAD

nicotinamide adenine dinucleotide

NADH

reduced nicotinamide adenine dinucleotide

NF

nuclear factor

NO

nitric oxide

Nrf

nuclear factor erythroid 2-related factor

NOX

NADPH phosphate oxidase

OCR

oxygen consumption rate

PAI

plasminogen activator inhibitor

PDGF

platelet-derived growth factor

PDR

proliferative diabetic retinopathy

PGC1α

proliferator-activated receptor gamma coactivator 1-alpha

PI3K

phosphoinositide-3 kinase

PKC

protein kinase C

PKM2

pyruvate kinase M2 isoform

RBP

retinol-binding protein

rh

recombinant human

ROS

reactive oxygen species

SGLT2

sodium-glucose transporter-2

T1D

type 1 diabetes

T2D

type 2 diabetes

TCA

tricarboxylic acid

TGF

transforming growth factor

VCAM

vascular cell adhesion molecule

VEGF

vascular endothelial growth factor

VEGFR

VEGF receptor

VSMC

vascular smooth muscle cell

WT

wild type

Disclosures

M.G.Y. has been supported by the American Diabetes Association (ADA 9-18-CVD1-005) and the Mary K. Iacocca Foundation. D.G. has received lecture honoraria from AstraZeneca, Bayer, Boehringer Ingelheim, Delta Medical Communications, EASD eLearning, Finnish Nephrology Association, Kidney and Liver Foundation in Finland, all outside the submitted work; advisory board honoraria from AstraZeneca, Bayer, Boehringer Ingelheim, all outside the submitted work; and support from the Wilhelm and Else Stockmann Foundation, Liv och Hälsa Society, Medical Society of Finland (Finska Läkaresällskapet), Sigrid Juselius Foundation, Helsinki University Hospital (vtr), University of Helsinki (Clinical Researcher stint), Minerva Foundation Institute for Medical Research, and Academy of Finland (UAK1021MRI). J.F. has been supported by the Mary K. Iacocca Foundation. G.L.K. has received grants from NIDDK (P30DK036836, as part of the Joslin Diabetes Research Center Grant), NEI (R01EY026080), NHLBI (R01HL161864), Thomas J. Beatson Jr. Foundation, and the Dianne Nunnally Hoppes Fund. K.P. and Q.L. have nothing to disclose.