Metabolic Memory and Chronic Diabetes Complications: Potential Role for Epigenetic Mechanisms

Abstract

Recent estimates indicate that diabetes mellitus currently affects more than ten percent of the world’s population. Evidence from both the laboratory and large scale clinical trials has revealed that prolonged hyperglycemia induces chronic complications which persist and progress unimpeded even when glycemic control is pharmaceutically achieved via the phenomenon of metabolic memory. The epigenome is comprised of all chromatin modifications including post translational histone modification, expression control via miRNAs and the methylation of cytosine within DNA. Modifications of these epigenetic marks not only allow cells and organisms to quickly respond to changing environmental stimuli but also confer the ability of the cell to “memorize” these encounters. As such, these processes have gained much attention as potential molecular mechanisms underlying metabolic memory and chronic diabetic complications. Here we present a review of the very recent literature published pertaining to this subject.

Introduction

The World Health Organization estimates that diabetes mellitus (DM) affects approximately 346 million people worldwide, and these numbers are projected to increase above 400 million by 2030 [1]. DM is a disease of metabolic dysregulation that results in reduced life expectancy due to disease specific microvascular (retinopathy, nephropathy, neuropathy, impaired wound healing) and macrovascular (heart disease and stroke) complications [2]. A unifying mechanism for the induction of complications due to hyperglycemia has been proposed by Brownlee. Central to this mechanism is the increased production of reactive oxygen species (ROS) which in turn promotes flux through the polyol, hexosamine, protein kinase C, and advanced glycation end-product formation pathways leading to altered gene expression profiles of affected cells [2;3]. The results of several large scale clinical trials indicate that once initiated, these complications persist and continue to progress unimpeded even when glycemic control is achieved through pharmaceutical intervention [4-9]. The Diabetes Control and Complications Trial (DCCT), initiated in 1983, studied a cohort of patients with type 1 diabetes over a period of 6.5 years and divided the patients into either conventional therapy (normal glycemic control) or intensive therapy groups (greater degree of glycemic control). As the study progressed it became clear that the conventional therapy group developed retinopathy, neuropathy, microalbuminuria, and albuminuria at a much increased rate when compared to their intensive therapy counterparts. [4]. The results of this study were so dramatic that at the end of the DCCT, all patients were placed on intensive therapy and followed up long term in the Epidemiology of Diabetic Complications and Interventions Trial (EDIC) [10]. The EDIC trial documented that patients who received intensive treatment from the onset of the DCCT benefitted from the early glycemic control indicated by a slower incidence and progression of diabetic micro-vascular complications relative to those who were initially in the conventional DCCT treatment group [8]. In 2009, another DCCT follow-up study reported that the cumulative incidence of diabetic complications was still dramatically increased in the original conventional treatment group as compared to the intensive treatment group almost 30 years later [11].These studies indicate that a prolonged period of poor glycemic control causes a persistence of diabetic complications even after the establishment of good control and this has been termed, the legacy effect or the metabolic memory phenomenon.

Four clinical trials involving type II DM patients examined the risk of developing cardiovascular disease as a function of glycemic control level (as determined by HbA1C levels). The United Kingdom Prospective Diabetes Study [5], The Action to Control Cardiovascular Risk in Diabetes study (ACCORD) [12], The Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation trial (ADVANCE) [13], ) and the Veterans Affairs Diabetes Trial (VADT) [14] all reported that increased glycemic control did not significantly reduce cardiovascular events. [13-16]. However, the ADVANCE trial did report a reduction in the incidence nephropathy seen in the DCCT trials. In contrast to the above 4 trials the United Kingdom Prospective Diabetes Study Post Trial (UKPDS post-trial study) which followed newly diagnosed type 2 diabetic patients on standard and intensive therapy for 10 years reported that the metabolic memory effect emerged. Patients that received either standard or intensive treatment following their diagnosis showed a marked decrease in CVD endpoints indicating that early glycemic intervention has the greatest impact on preventing this diabetic complication (UK Prospective Diabetes Study (UKPDS) [7] The metabolic memory effect was also seen in the Steno-2 trial [6;17]. Collectively, these trials illustrate the harmful nature of the metabolic memory phenomena and have lead to an increase in studies aimed at deciphering the molecular mechanisms that support it.

Animal models of DM have been critical for discovery related to the pathophysiology of diabetic complications and the metabolic memory phenomenon. In fact, the persistence of diabetic complications was first documented in a canine model of diabetic retinopathy which has since been supported by multiple lines of experimental evidence using a variety of animal models and in vitro culture systems [18-24]. Collectively, these studies clearly show that the initial hyperglycemic period results in permanent abnormalities (including aberrant gene expression) of the target organs/cells and this harmful phenomenon has been termed, metabolic memory [25;26]. The ability to sustain these complications in the absence of hyperglycemia (hereafter referred to as the metabolic memory state) invokes a role for the epigenome in perpetuating diabetic complications.

Epigenomes consist of all the chromatin modifications for a given cell type and are responsible for a cell’s unique gene expression pattern. These chromosome modifications are dynamic during development by supporting cell differentiation, are responsive to external stimuli, can be altered in disease [27-29] and are mitotically stably inherited [30;31]. Epigenetic mechanisms include: post translational histone modifications, non-canonical histone variant inclusion in octomers, chromatin access changes through DNA methylation, and gene expression control through non-coding micro RNAs [32-35]. Altogether, these processes allow cells/organisms to quickly respond to changing environmental stimuli, [36-38] and also confer the ability for the cell to “memorize” these encounters once the stimulus is removed [30;31]. Therefore, as gene expression control changes resulting from epigenetic processes are stable in the absence of the signal that initiated them and are heritable through cell division, they have gained great interest as underlying molecular mechanisms of metabolic memory. The results that are emerging in the context of DM and epigenetics parallel advancements in other diseases in that a plethora of epigenetic changes induced by hyperglycemia cause remarkable persistent changes in transcriptional networks of cells. Here we present a review of the literature published between January 2011 and April 2012 regarding the epigenetic control of gene expression leading to chronic diabetic complications.

Histone Modifications

The basic building block of chromatin is the nucleosome which consists of 147 bp of DNA wound around an octomer of histone proteins (2 copies of H2A, H2B, H3 and H4). Over 100 post-translational modifications of histone proteins have been identified with the methylation and/or acetylation of lysine residues being most extensively studied [39]. In general, acetylation neutralizes the positive charge of the lysine residue, has profound effects on chromatin structure, and is associated with transcriptionally active genes. Methylation, on the other hand, can stimulate or repress transcription depending on which lysine residues within the tails of H3 and H4 are being modified [40]. Initial reports relating epigenetic modifications and persistent DM induced expression changes examined specific acetylation or methylation events. These primarily focused on the sustained activation of ROS induced inflammatory genes such as nuclear factor κB p65 (NF-κB p65) [41-47]. Histone modifications (marks) are not present singly and thus the transcriptional activity is more likely dependent upon the sum of the histone marks at any given locus. Therefore, in order to understand the epigenetic control of persistent gene expression changes induced by hyperglycemia and persistent in metabolic memory it will likely become necessary to identify the complete chromatin profile at each locus. This is illustrated by a recent report from the lab of Renu Kowluru [48] where the authors examined histone modifications at the retinal MnSOD (super oxide dismutase locus), and Sod2, in an attempt to account for its decreased expression in response to hyperglycemia. In this study diabetic rats [streptozotocin (STZ) or high galactose induced] were maintained in poor glycemic control for 4 months, good glycemic control for 4 months, or poor control for 2 months, followed by good control for 2 months (metabolic memory state). Through the use of chromatin immuno-precipitation experiments they provide data showing that in both the poor control and metabolic memory groups there was an increase in acetylated H3K9, and the recruitment of the NF-κB p65 transcription factor at its promoter/enhancer. Both of these are usually associated with enhanced transcriptional activity and would appear to stand in apparent contrast to the documented decrease in MnSOD. However, increased trimethyl H4K20me³ (a repressive histone mark) was also found at the Sod2 locus which can presumably override the presence of activating elements. Additionally, inhibiting the enzyme responsible for H4K20 trimethylation, (SUV420H2), resulted in restored MnSOD expression supporting the importance of this modification in sustaining the MnSOD expression impairment. Although we are just beginning to understand the chromatin landscape at a single locus (in the context of hyperglycemia), the advent of genome wide association technologies such as those performed by Pirola et al. (where they compared histone acetylation and DNA methylation changes in the acute diabetic state) will undoubtedly yield further insight into epigenetic control of metabolic memory [49]. While identifying all of the hyperglycemia induced histone modifications is essential it must be mentioned that there is growing evidence that these marks may not be instructive as was previously thought and require careful consideration when drawing conclusions regarding their status and gene expression [40;50]..

Defining the chromatin status at each locus is of obvious importance; however, it is equally vital to examine the molecular machinery that performs these modifications as they could serve as targets for future therapies. The Class III histone deacetylase (HDAC) sirtuin 1 (SIRT1), is a multifunctional protein that not only deacetylates histone tails but also has many non-histone substrates including transcription factors and co-regulators [51]. Recently it was documented that SIRT1 can regulate many cell processes including inflammatory responses and ROS levels, leading Zheng et al. [52] to hypothesize that that SIRT1 may play a major role in the metabolic memory phenomenon. These investigators examined this hypothesis by incubating bovine retinal capillary endothelial cells (BRECs) in normal glucose for 3 weeks, high glucose for 3 weeks (acute diabetic), or high glucose for 1 week followed by 2 weeks of normal glucose (metabolic memory state). They also examined retinas from normal rats, STZ-treated rats that didn’t receive insulin for 6 weeks (poor control) and a metabolic memory group of STZ-treated rats that had an initial 2-week period of no insulin followed by 4 weeks of hyperglycemic control. SIRT1 expression and activity was reduced in response to high glucose which could not be reversed in the metabolic memory state. In both experimental systems, NF-κB p65 and the pro-apoptotic gene, Bax were increased by hyperglycemia induced ROS, and remained elevated after normoglycemia was restored. Evidence was provided that SIRT1 deacetylates and stimulates the AMP-activated protein kinase which, in turn, activates MnSOD, uncoupling protein 2, and catalase collectively reducing ROS. This was supported by data where BRECs treated with siRNA against SIRT1 had increased sensitivity to hyperglycemic stress, whereas when SIRT1 was over-expressed or activated by metformin ROS, Nf-κb and Bax expression were all decreased. From this study, it would appear that increasing SIRT1 levels might be therapeutically valuable, however this benefit may be tissue limited. Vahtola et al. [53] have documented that in cardiomyocytes of GK diabetic rats, SIRT1 levels are increased by the hyperglycemic environment. Additionally, they reported that the SIRT1-p53 pathway is key in pronounced hypertrophy, interstitial fibrosis, and cardiomyocyte apoptosis seen with diabetic rats leading to post-infarct heart failure. Discrepancies like these may suggest that the role of epigenetic mechanisms might be tissue specific however they do provide evidence that they could become potential targets for future therapies.

The manipulation of the epigenetic machinery has also been recently examined in the context of diabetic nephropathy (DN). Although increased urine albumin secretion is clinically the first sign of kidney disease, podocyte loss, glomerular hypertrophy and mesangial matrix expansion, are thought to precede it and likely underlie this increase. Epidermal growth factor (EGF) and its receptor have been implicated in the hyperglycemia induced expansion and this pathway can be suppressed in human breast cancer cells by the inhibition of histone deacetylases [54-56]. Combining these facts led Gilbert et al. [55] to investigate if the HDAC inhibitor, vorinostat (HDAC classes I, II and IV, not sirtuins), could inhibit hyperglycemia induced kidney enlargement. When they cultured proximal tubule cells in the presence of vorinostat, EGFR protein and mRNA were reduced and cellular proliferation was attenuated. Moreover, daily vorinostat administration for 4 weeks to STZ-treated rats greatly reduced renal growth and glomerular hypertrophy. Like the above retinal study it appears the modulation of the epigenetic machinery may provide a valuable therapeutic avenue.

MicroRNAs

MicroRNAs (miRNAs) are a family of short (19–24 nucleotides in length), noncoding single-strand RNA molecules that regulate gene expression via binding to specific sites on the 3′-untranslated region (3′-UTR) of the target mRNAs, causing translational repression or mRNA degradation [57]. Computational predictions estimate that each miRNA can regulate up to 200 mRNA species and collectively, these molecules may control the expression of as much as 60 percent of protein coding genes [58]. As these genes are integral to a vast array of cellular processes, it is not surprising, that mis-expression of miRNAs has been examined in a wide variety of diseases including DM [57;59;60]. MiRNA expression regulation has been implicated in most aspects of DM including; incidence [61-63], insulin sensitivity [64-66] and the development of persistent complications (below). Several recent reports have suggested that miRNAs contribute to the persistence of DN. Long et al. identified nine miRNAs that were differentially up-regulated in glomeruli from db/db mice, high glucose-treated kidney microvascular endothelial cells, and high glucose-treated podocytes [67]. They provided convincing evidence that one of these, miR-29c, directly targets the Sprouty homolog 1 (Spry1) and inhibits its synthesis in a hyperglycemia responsive manner. This is critical for DN as SPRY1 negatively regulates Rho kinase whose activity is linked to DN through inducing podocyte apoptosis and mesangial fibronectin synthesis. Importantly, in this study they used a chemically modified antisense oligo technique to knockdown miRNA29c expression and this was correlated with a marked improvement in albuminuria in db/db mice. Similarly, Dey et al. have recently documented that expression of miR-21 is induced by hyperglycemia which in turn inhibits PTEN (tensin homolog deleted in chromosome 10) expression causing an increase in mesangial hypertrophy and increased fibronectin abundance [68].

In the diabetic kidney, tubulointerstitial fibrosis due to increased expression of extracellular matrix proteins such as collagens and fibronectins is initiated and sustained by a number of different factors including the transforming growth factor-beta (TGF ) family. This family of inflammation mediators is previously documented to be aberrantly expressed in metabolic memory. More specifically, TGF-β2 appears to support the fibrotic pathway by increasing the expression of matrix proteins and triggering the epithelial-to-mesenchymal transition (tubular EMT) in proximal tubular cells [69]. Wang et al. [70] treated rat proximal-tubular epithelial cells (NRK52E) with TGF-β1 and TGF-β2 for 3 days, and reported a morphological and phenotypic transition characteristic of EMT. This change was accompanied by the reduced expression of the epithelial marker, E-cadherin; and by increased expression of the mesenchymal marker, vimentin, the ECM proteins, fibronectin, collagen I, and collagen IV, and importantly miR200a. Interestingly, the 3′ UTR of TGF-β2 includes a target site for the miR-141/200a family and in this study ectopic over-expression of miR-141 or miR-200a decreased the expression of TGF-β2 possibly reversing the increased synthesis of ECM genes in vitro. As additional evidence for the involvement of miR141/200a family in DN, the authors reported that the levels of these are reduced in the cortex of kidneys from diabetic apoE mice. Collectively this data suggests that miR-200a plays a central role in ECM accumulation and DN.

Rama Natarajan and colleagues have taken discovery of miRNA in the context of DN to higher level, providing evidence that modulating miRNA expression involved in diabetic kidney disease may be possible [71]. Previous work from this lab documented that miRNA-192 may be a master regulator in DN as it presides over a cascade of events leading to glomerular hypertrophy. This miRNA downregulates E-box repressors such as Zeb1 and Zeb2 in mouse mesangial cells and glomeruli of diabetic mice [72]. As several E-boxes are found in the upstream promoter regions of ECM proteins (collagen type I a2 and collagen type IV a1), TGFβs, connective tissue growth factor, fibronectin and other miRNAs, (miR-216a/217, and the miR-200 family) modulating the levels of miRNA-192 may have pleiotropic effects. In their latest work Putta et al. examined the efficacy of using a locked nucleic acid (LNA)–modified inhibitor of miR-192 in mouse models of diabetic nephropathy. When they treated diabetic mice with LNA–anti-miR-192 (from the onset of diabetes up to 17 weeks), the levels of miR-192 in the kidneys of both normal and STZ-induced diabetic were significantly reduced. Moreover, a concomitant increase in Zeb1/2 was achieved which led to decreased gene expression of collagen, TGF-β, and fibronectin. From a DN viewpoint, treatment of the diabetic mice with LNA–anti-miR-192 attenuated proteinuria. This study therefore illustrates that pharmaceutical intervention of this epigenetic mechanism may provide a means for future success in treating DN.

A role for miRNAs in the pathogenesis of diabetic retinopathy (DR) has been recently proposed [73;74].The first of these manuscripts examined the control of VEGF (significant in both the early and late stages of DR) by miR-200b (reviewed in [57]). In the second report, Feng et al. exposed human umbilical vein endothelial cells to high glucose and reported an increase in fibronectin (FN) expression (observed with several diabetic complications) with an accompanying decrease in miR146a expression [74]. The 3′UTR of the FN mRNA houses a target site for miR146a and both the binding of miR146a and control of FN expression by this miRNA was established through transfection experiments. The decrease in miR146a expression was then confirmed to occur in vivo using STZ induced diabetic mice. Additionally, the authors provided evidence that the glucose-induced miR-146a down-regulation is mediated through the HDAC p300 and that the FN/p300/miR146a triad is seen in heart and kidney tissue as well. Therefore, this study not only establishes the control of FN expression by miRNA, but also provides a functional link for the control of gene expression for miRNA and histone modification that hasn’t previously been documented in the context of DM.

DNA Methylation

DNA methylation occurs as 5-methyl-cytosine; mostly in the context of CpG dinucleotides. In vertebrates, genomic DNA methylation is found throughout the genome with the exception of short unmethylated regions termed CpG islands which are approximately 1 kb CpG rich regions typically found in gene promoters [34;75]. Multiple roles for DNA methylation including: gene silencing, silencing of transposable elements, developmental regulation of transcription, cell cycle control, and differentiation have been documented [76-80]. Historically, the central mechanism that DNA methylation was thought to function by was in maintaining chromatin stably in a repressed state and thereby inhibiting promoter activity. However, recent genome wide DNA methylation analysis has indicated that methylation in the “bodies” of active genes is significantly higher than those of inactive genes [81;82]. This appears to be evolutionarily conserved and may function to suppress inappropriate transcription, regulate mRNA splicing, modulate elongation, and regulate tissue specific alternative promoter usage [83;84;84-88]. Not unexpectedly, due to the critical role in gene expression, altered DNA methylation is associated with several human diseases including multiple sclerosis, Alzheimer’s disease, and many cancers [89-93]. Variations in “normal” DNA methylation have been correlated with many aspects of DM including, susceptibility [94-97], insulin resistance [98], complication development [99], and early detection [100-102]. Very recently, a comprehensive genomic DNA methylation profiling of type 2diabetic islets revealed that 276 CpG loci displayed a significant hypomethylation phenotype and may provide insight on the dysregulation of diabetic islets and disease pathogenesis [103].

The first report demonstrating a cause and effect relationship between hyperglycemia and altered DNA methylation documented that genomic hypomethylation was induced within the liver of type 1 diabetic rats as early as 2 weeks post hyperglycemia onset [104]. In contrast, the same group reported that hepatic DNA hypermethylation was evident at 12 weeks of age in the Zucker diabetic fatty rat (a type II model) [105]. Pirola et al. examined primary aortic endothelial cells exposed to high glucose in vitro and performed a more comprehensive analysis of both histone acetylation and DNA methylation. They used a methyl-capture followed by sequencing technique to assay DNA methylation changes. In this study they observed significant alterations in DNA methylation patterns and showed that induced hypermethylation localized to regions within five kilobases of transcriptional start sites. They also observed broad changes to H3K9/K14 acetylation and reported that regionalized hyperacetylation correlated very well with DNA hypomethylation and hyperglycemia induced gene induction. Unfortunately, none of the above studies have looked at the prolonged hyperglycemic or the metabolic memory state.

Hyperglycemia-induced DNA methylation changes that persist into the metabolic memory state have recently been examined in a STZ-induced zebrafish type 1 diabetes model [24;106]. As the zebrafish is capable of fully repairing/regenerating its damaged pancreas, this model provides a unique opportunity to examine the metabolic memory in a euglycemic environment similar to what would be seen in a post islet or pancreatic transplant state in other organisms. A second advantage for this system is that mitotically transmissible components (epigenetic changes) can be examined cleanly as they can be separated from the potential complicating factors (AGE, ROS, etc) of the previous diabetic state [24]. In this model, the complications of impaired limb (caudal fin) regeneration and impaired wound healing persist following restoration of glucose control. Caudal fin tissue that was examined via a methylated DNA immunoprecipitation followed by sequencing approach and genome-wide CpG island analyses revealed that hyperglycemia induced global demethylation was largely maintained in the metabolic memory state. Additionally, the data clearly revealed an equal distribution of DNA hypomethylation present at all genomic loci including promoter, intergenic, and intragenic sites. When these data were viewed within the context of global gene expression changes, a correlation for a subset of genes was observed. Most interesting from an epigenetic perspective, is that at least 3 members of the Epigenetic Code Replication Machinery (ECREM) complex, which is proposed to duplicate the epigenetic code during DNA replication [107], were altered in their expression. As such, the ECREM complex could provide a mechanism to coordinate DNA methylation and histone modification alterations induced by hyperglycemia and maintained through cell division.

Retinal Mitochondria

Damage to retinal mitochondria is thought to play a significant role in the pathogenesis of DR. The first stage of DR is characterized by the loss of capillary and other retinal cells through apoptosis [108]. Two studies revealed that the activity of the matrix metalloproteinases MMP2 and MMP9 cause mitochondria DNA (mtDNA) damage and degradation of mitochondrial membranes in retinal capillary cells which in turn induces apoptosis of the same [109;110]. In another study, the same group reported that DM increased mtDNA damage and decreased the quantity of mitochondria in mice retinas, retinal endothelial cells and retinas from human donors. The mechanism by which this occurs appears to involve inhibition or prevention of transcription factor A translocation into the mitochondria (critical for biogenesis) [111] From a metabolic memory perspective, two studies were published utilizing STZ-treated mice including mice that were maintained under good, poor or metabolic memory (poor then good) glycemic control. Zhong and Kowluru [112] reported that the poor control group had enlarged mitochondria with partial cristolysis in the retinal microvasculature and decreased mitochondrial abundance (ie import) of proteins required for mitochondrial function. Furthermore, these deleterious changes were maintained in the metabolic memory group. Similar to this study, Santos and Koluru [113] also reported that diabetes-induced mitochondrial dysfunction was maintained in the metabolic memory group; however, supplementation of the metabolic memory group with lipoic acid (reduces ROS) had beneficial effects on mitochondrial biogenesis and DR progression. As the changes in both of these studies were also observed in human donor samples it is expected that therapies that target mitochondrial may help in DR

Conclusions

To date, many in vitro systems and model organism have been used to reveal the molecular mechanisms underlying diabetic complications and recently metabolic memory. A feature common to many of these studies is that hyperglycemia induces gene expression changes in thousands of genes which are maintained in the metabolic memory state for at least a subset of the altered loci. In this review we have described the latest discoveries that document the influence epigenetic mechanisms have on chronic diabetic complications and metabolic memory. Within this last year new insight regarding epigenetic control has been provided in the areas of histone modification, modulation of the chromatin modifying machinery, control of gene expression through miRNAs, and hyperglycemia induced persistent DNA methylation changes. With the advent of more affordable genome-profiling technologies and more powerful bioinformatics analytical techniques, it is expected that the amount of information regarding the epigenome and diabetes mellitus will grow exponentially within the next few years. As such, the potential to identify all hyperglycemia induced changes (gene expression and epigenetic) within cells and how they intersect may become a reality in the not so distant future. However, we feel that the main challenge will not be in the collection of this data but instead will lie in the ability to interpret that data to yield biologically meaningful conclusions and most importantly, use this information to prevent the progression of chronic diabetic complications.