Despite decades of scientific effort, diabetes continues to represent an incredibly complex and difficult disease to treat. This is due in large part to the multifactorial nature of disease onset and progression and the multiple organ systems affected. An increasing body of scientific evidence indicates that a key mediator of diabetes progression is NRF2, a critical transcription factor that regulates redox, protein, and metabolic homeostasis. Importantly, while experimental studies have confirmed the critical nature of proper NRF2 function in preventing the onset of diabetic outcomes, we have only just begun to scratch the surface of understanding the mechanisms by which NRF2 modulates diabetes progression, particularly across different causative contexts. One reason for this is the contradictory nature of the current literature, which can often be accredited to model discrepancies, as well as whether NRF2 is activated in an acute or chronic manner. Furthermore, despite therapeutic promise, there are no current NRF2 activators in clinical trials for the treatment of patients with diabetes. In this review, we briefly introduce the transcriptional programs regulated by NRF2 as well as how NRF2 itself is regulated. We also review the current literature regarding NRF2 modulation of diabetic phenotypes across the different diabetes subtypes, including a brief discussion of contradictory results, as well as what is needed to progress the NRF2 diabetes field forward.
Introduction
Diabetes continues to represent a challenging global health crisis that has a significant societal and economic impact. In the U.S. alone, the Centers for Disease Control and Prevention estimates that in 2020, ∼10% of the population had some form of diabetes, which in turn was associated with a total medical cost upwards of $300 billion (1). Furthermore, an estimated 34.5% of the adult population in the U.S. is thought to have prediabetes (2), indicating that this crisis only promises to get worse in the coming decades. As one might expect, treating patients with diabetes represents a substantial medical challenge, especially considering the multitude of causative factors, diverse organ systems affected, and differential timing of disease onset and progression. As such, the identification of propathogenic changes that influence the development of diabetic outcomes, particularly those that might span multiple environmental exposures or genetic contexts, remains critically important for developing effective therapies to prevent or mitigate disease.
While the identification of the pathogenic mechanisms that drive diabetes progression has continued to increase at an exponential rate, there are certain fundamental factors throughout a person’s life that influence their risk of developing diabetic outcomes. Among these, age, diet, genetic susceptibilities, and environmental exposures all have been shown to significantly influence the onset and progression of every form of diabetes, from gestational to type 2 (3). Regardless of the causative factors, the key driver of diabetic phenotypes, including the development of insulin resistance, dyslipidemia, and others, is a propathogenic shift in cellular metabolism. One critical regulator of cellular health, particularly during stress, is the transcription factor nuclear factor erythroid 2-related factor 2 (NRF2). Initially identified as a master regulator of the antioxidant response and drug or xenobiotic detoxification, NRF2 is now known to regulate target genes involved in proteasomal and autophagic function, iron, lipid, and carbohydrate metabolism, and DNA repair (4). Importantly, as the number of target pathways influenced by NRF2 continues to expand and evolve, so too does our understanding of its role in mediating disease progression. This is highlighted by evidence in the literature indicating that NRF2 can play both a protective and pathogenic role depending on the timing and duration of activation (4–8). As such, targeting NRF2 in any given disease, including diabetes, requires careful consideration regarding how activation or inhibition of NRF2 might influence pathogenesis.
As one might expect from a multiorgan disease where the onset of diabetogenic phenotypes can range from in utero to old age, the role of any pathway of interest is complicated. This also holds true for NRF2, where a host of genetic and pharmacological studies have demonstrated that activation or inhibition of NRF2 signaling has significant consequences across a wide variety of diabetic contexts. In this review, we will give a brief overview of NRF2 transcriptional programs as well as how NRF2 itself is regulated. Finally, the role of NRF2 activation in different diabetes subtypes will be explored.
The NRF2 Signaling Pathway
Activation of NRF2 signaling can be initiated via a variety of mechanisms, ranging from oxidative stress to autophagic dysfunction to mutations in NFE2L2/NRF2 or its negative regulators that result in its constitutive activation. Once activated, NRF2 binds to the promoter of antioxidant response element (ARE)-containing genes to initiate their transcription. While NRF2 has now been shown to regulate the expression of a wide variety of downstream transcriptional programs, regulation of its own expression can occur at the DNA, RNA, and protein levels. This includes epigenetic modifications that dictate the promoter accessibility of NFE2L2/NRF2 or its negative regulators, microRNA-dependent suppression of NRF2 translation, or regulation of NRF2 protein stability, either through interaction with E3 ligase complexes that target it for degradation (i.e., Kelch-like ECH-associated protein 1/Cullin 3/Ring box 1 [KEAP1-CUL3-RBX1], HRD1, and S-phase kinase-associated protein 1-Cullin 1-RBX1/β-transducin repeat-containing protein [SCF/β-TRCP]) or other binding partners (discussed in more detail below) that stabilize and prevent its degradation (Fig. 1). Here, we will briefly discuss the primary mechanisms that control NRF2 expression at the mRNA and protein levels as well as the transcriptional programs it mediates, particularly those that are most relevant in a diabetes context.
Regulators of NRF2 Expression
As mentioned above, NFE2L2/NRF2 expression is regulated at multiple levels. The primary mode of NRF2 regulation is at the protein level, with three primary E3 ubiquitin ligase complexes mediating the ubiquitylation and subsequent degradation of NRF2 by the 26S proteasome: 1) KEAP1-CUL3-RBX1, 2) HRD1 (also known as synoviolin), and 3) SCF/β-TRCP. Importantly, disruption of NRF2 degradation by any of these three complexes plays an important role in mediating NRF2 activation in specific pathological contexts. For example, oxidative or electrophilic modification of specific cysteine residues (i.e., C151) on KEAP1 causes a conformational change in the KEAP1-CUL3-RBX1-NRF2 complex that prevents NRF2 ubiquitylation and degradation (9–12). In the case of HRD1, studies from our group (13) have shown in a liver cirrhosis model that endoplasmic reticulum stress increases SYVN1/HRD1 expression, which in turn degrades NRF2, preventing it from mounting a protective response. Finally, GSK3-β–dependent phosphorylation of β-TRCP governs its control of NRF2 degradation, with removal of insulin, growth factors, or other critical metabolites stimulating activation of the NRF2 cascade (14–16). As such, the ability of cells to respond to different external or internal stressors relies heavily on the functionality of these degradation complexes. It is also important to note that several other proteins (CRIF1 and SIAH2) have also been shown to directly bind NRF2 and promote its ubiquitylation (17,18), although the extent to which these interactors are involved across relative (patho)physiological contexts, as well as identification of the E3 ligase responsible for degradation, requires further investigation.
In contrast to the binding partners that trigger its degradation, there are also several proteins that interact directly with NRF2, or its negative regulator, KEAP1, to prevent KEAP1-mediated NRF2 degradation and thus activate NRF2 signaling. The key cell cycle regulator p21, the redox-sensitive chaperone DJ1, and the tumor suppressor BRCA1 have all been shown to stabilize NRF2 in different pathological contexts (19–21). Several proteins that outcompete NRF2 for KEAP1 binding, resulting in its stabilization and increased translocation to the nucleus, have also been reported, including p62, FAM129B, PALB2, and WTX (22–26). Posttranslational modification of NRF2 or its degradation machinery also plays an important role in determining NRF2 stability and function, as phosphorylation, SUMOylation, and acetylation of NRF2, or O-GlcNAcylation of KEAP1, have all been shown to affect NRF2 protein stability and activity (27–31). Finally, there are also mechanisms that control transcription and translation of NRF2, including epigenetic modifications and microRNA (32,33). Thus, NRF2 expression is largely determined by its interacting partners or posttranslational modifications that dictate its stability and function. The relevance of NRF2 regulation in a diabetes context will be discussed in detail below.
NRF2-Mediated Transcriptional Programs
As alluded to briefly above, NRF2 target genes have been implicated in mediating all aspects of cellular function. NRF2 is now speculated to regulate upwards of 300 target genes. However, it is important to note that many of these were identified via DNA microarray or assessment of transcript/protein level changes following pharmacological upregulation with a single NRF2 inducer or the observation that their expression was decreased or increased in an NRF2 null or KEAP1 null setting, respectively. As such, many putative target genes of NRF2 still require full validation, including chromatin immunoprecipitation (ChIP)-based identification of NRF2-ARE binding and mutation of putative AREs to ensure functionality as well as determination of their (patho)physiological relevance. This is of particular importance considering that many NRF2 target genes may only be relevant in certain tissue, genetic, or pharmacological contexts. Below, we will briefly introduce some of the key NRF2 target genes involved in regulating protein, metabolic, and redox homeostasis, focusing primarily on those that have been well validated. (Fig. 1).
Proteostasis
The synthesis, folding, modification, and eventual degradation of proteins is tightly regulated. As highly dynamic molecules, proteins undergo frequent conformational changes that are dictated by partial unfolding/refolding, posttranslational modifications, and interactions with their binding partners or substrates. Proteotoxic stress, which results from a breakdown in proper protein processing, is driven by the accumulation of misfolded, damaged, or aberrantly modified proteins that can lead to dysregulated cell function. Importantly, proteotoxic stress has been shown to be a key mediator of numerous diabetic outcomes (34–36), thus playing a critical role in the pathogenesis of diabetes. There are several defense mechanisms that are crucial in maintaining protein homeostasis and limiting or preventing the accumulation of damaged or dysfunctional proteins. Among these, NRF2 has been shown to play a critical role in regulating intracellular proteolytic degradation systems (i.e., the proteasome and autophagy-lysosome pathway) as well as molecular chaperones designed to maintain proper protein conformation and stability. Specific examples of NRF2 regulation of these pathways are discussed below.
Proteasome.
The 26S proteasome is comprised of a 20S core particle bound to one or two 19S regulatory caps, which together mediate the degradation of ubiquitinated proteins. While fellow Cap’n’collar family member NRF1 is the primary transcriptional regulator of proteasomal subunits under conditions that inhibit the proteasome or increase ER stress, NRF2 also regulates several proteasomal subunits, particularly during oxidative stress. For example, Kensler and associates (37) showed that NRF2 induction in the mouse liver following exposure to 1,2-dithiole-3-thione increased expression of several structural and catalytic proteasomal subunits, including Psma1, Psma7, Psmb3, Psmb5, and Psmb6. In the same study, they verified that Psmb5 had a functional ARE, which was also later validated in H2O2-treated mouse embryonic fibroblasts (38). NRF2 has also been shown to bind to the promoter of proteasome maturation protein (POMP), which is critical for proteasome assembly, with NRF2-dependent regulation of POMP being crucial for human embryonic stem cell self-renewal (39,40). Thus, NRF2 activation clearly plays a role in proteasome assembly under certain oxidative conditions.
Autophagy.
Another indispensable regulator of protein homeostasis is the autophagy pathway. Autophagy is a highly coordinated degradation pathway that relies upon a tightly regulated subset of protein–protein and protein–lipid interactions that mediate the formation of autophagic vesicles and eventual cargo degradation. Much like the proteasome, NRF2 regulates several key autophagy pathway components. One example, SQSTM1/p62, a cargo-specific receptor that eliminates ubiquitylated protein aggregates and organelles via autophagy, is a known NRF2 target gene (41,42). Similarly, several autophagy initiation proteins (i.e., ULK1/2, ATG5, and ATG7) were identified to contain putative AREs, which were subsequently validated via ChIP pulldown and quantitative reverse transcriptase-polymerase chain reaction in human and mouse cell lines. Conversely, Nrf2−/− mice and mouse embryonic fibroblasts exhibited decreased expression of these key autophagy genes, further indicating NRF2 controls the autophagy pathway (42). Finally, chaperone-mediated autophagy also relies on NRF2, as Lamp2A was recently shown to have two functional AREs, with loss of NRF2 resulting in impaired chaperone-mediated autophagy in the mouse liver (43). As such, numerous components of the autophagic machinery can be regulated by NRF2 across various contexts, indicating an important connection between these two critical cell survival pathways.
Protein Folding and Stability.
Hsf1 (heat shock transcription factor 1) plays an important role in regulating the expression of several heat shock proteins (HSPs) that repair misfolded proteins under proteotoxic stress conditions. NRF2 binds to a functional ARE in the HSF1 promoter in arsenic-treated fibrosarcoma and breast cancer cell lines (44). Furthermore, it has been proposed that NRF2 and HSF1 downstream cascades rely upon one another at various levels, indicating that these pathways act in concert to maintain not only proper protein folding but also redox balance (45).
Metabolism
Along with its role in regulating protein stability and turnover, NRF2 has also been shown to regulate multiple aspects of key metabolic pathways, including lipid, carbohydrate, and amino acid metabolism, as well as iron transport and storage. Importantly, dysregulation of lipid, carbohydrate, and iron metabolism are key drivers of prodiabetogenic changes that promote diabetes progression (46,47). As such, NRF2 regulation of these metabolic cascades can have significant implications on cellular metabolism, function, and survival.
Lipid Synthesis and Catabolism.
Peroxisome proliferator-activated receptor γ (PPARG) and retinoid X receptor α (RXRA), both of which are transcription factors, play a central role in regulating lipid metabolism, particularly in adipose and liver tissues. Importantly, NRF2 transcriptionally regulates both following either genetic or pharmacological manipulation of NRF2 levels. For example, liver-specific Keap1−/− mice exhibit increased Pparg expression, as indicated via DNA microarray (48). NRF2-dependent regulation of Pparg was further confirmed in a model of hypoxia-induced acute lung injury, including validation of a functional ARE in the Pparg promoter, and evidence that Pparg expression was suppressed in Nrf2−/− mice (49). Pharmacologically, induction of NRF2 using sulforaphane (SF) or bardoxolone imidazolide (CDDO-Im) positively regulates Pparg and Rxra expression both in vitro and in vivo (39,49,50). Conversely, it has also been shown that Fasn and Scd1 (desaturase enzymes that catalyze addition of double bonds in fatty acids using NADPH as a cofactor) are upregulated in Nrf2-null mice and suppressed in an NRF2-activated setting (51,52). Several carboxylesterase enzymes (Ces1d, Ces1f, and Ces1g), lipases (LIPH and PLA2G7), and acyl CoA-oxidase 2 (ACOX2), all of which hydrolyze lipids, are also positively regulated by NRF2 (53). Thus, both lipogenesis and lipid catabolism can be regulated by NRF2, inferring that it acts as a key regulator of lipid homeostasis in the cell.
Carbohydrate Metabolism and the Pentose Phosphate Pathway.
Glucose-6-phosphate dehydrogenase (G6pdx), the rate-limiting enzyme of the pentose phosphate pathway and critical generator of NAPDH, was identified by ChIP sequencing to be an NRF2 target gene in diethyl maleate-treated mouse liver and hepatoma cells (54). The same study also found that phosphogluconate dehydrogenase (Pgd) and isocitrate dehydrogenase 1 (Idh1), other key NAPDH-generating enzymes, were transcriptionally regulated by NRF2 (54). All three were later verified to be NRF2 targets, along with malic enzyme 1 (ME1), transketolase (TKT), and transaldolase 1 (TALDO1), in A549 cells via ChIP-PCR (55). Several glycolytic enzymes, including solute carrier family 2 (facilitated glucose transporter), member 1 (Glut1), hexokinase 1 and 2 (Hk1/2), glucose phosphate isomerase 1 (Gpi1), 6-phospho-fructo-kinase/fructose-2,6-bisphosphatase 2 and 4 (Pfkfb2/4), aldolase a (Aldoa), enolase 2 and 4 (Eno1/4), and pyruvate kinase M2 (Pkm2), were shown via ChIP sequencing to be differentially regulated by NRF2 in the esophagus of postnatal day 7 Nrf2−/− and Keap1−/− mouse pups compared with the wild type (56). Recently, using liver tissue from Nrf2+/+ versus Nrf2−/− mice and ex vivo liver slices, our group identified three novel target genes, sorbitol dehydrogenase (Sord), triokinase/FM cyclase (Tkfc), and ketohexokinase (Khk), all of which are key regulators of the polyol pathway (57). As such, NRF2 clearly regulates carbohydrate intermediate flux through their respective metabolic pathways in a tissue-specific manner.
Iron Metabolism and Storage.
The transport, import, export, and storage of iron is tightly regulated. Proper iron metabolism is critical for cell survival, as ferric iron (Fe3+) plays a vital role in hemoglobin-dependent oxygen transport, as a cofactor for antioxidant and metabolic enzymes, and in electron transport in the mitochondrial respiratory chain. In particular, the transition between Fe2+ and Fe3+ is especially important, as Fe2+-driven Fenton reactions can result in the production of harmful free-radical species. It is not surprising then that NRF2 has been shown to regulate numerous aspects of iron metabolism and storage. For example, the genes that encode the light-chain (FTL) and heavy-chain (FTH1) subunits of the critical iron storage protein ferritin have been shown to be regulated by NRF2 (58). Heme oxygenase 1 (HMOX1) (an enzyme that catalyzes conversion of heme into biliverdin), ferrochelatase (FECH) (an enzyme that catalyzes heme biosynthesis), and biliverdin reductase A and B (BLVRA/B) (an enzyme that catalyzes conversion of biliverdin to bilirubin) are also transcriptional targets of NRF2 (39,50,58). Several heme transporters, including ATP-binding cassette transporter 6 (ABCB6) and solute carrier family 48, member 1 (SLC48A1), were also shown via ChIP sequencing in blood cells to be regulated by NRF2 (59). Overall, these studies indicate NRF2 is a crucial mediator of iron homeostasis in the cell, particularly in preventing the Fe2+-mediated production of free radicals.
Redox Homeostasis
Perhaps the best-known role of NRF2 is maintenance of cellular redox homeostasis, particularly during increased oxidative stress. From glutathione (GSH) synthesis to reduction of lipid peroxides, NRF2 regulates many of the enzymes responsible for restoring redox balance to the cell following oxidative insult. While the role of oxidative stress and antioxidant status in the context of disease has been reviewed extensively elsewhere (60,61), we will briefly introduce the key branches of redox homeostasis mediated by NRF2, as they are particularly relevant to the progression of diabetes.
GSH Synthesis.
GSH is a tripeptide antioxidant present in millimolar quantities in the cell. Importantly, GSH is utilized by a variety of antioxidant pathways to reduce oxidants and facilitate their removal, thus preventing damage to DNA, lipids, carbohydrates, and proteins. One of the main enzymes responsible for synthesizing GSH is glutamate cysteine ligase (GCL) (also known as γ-glutamylcysteine synthetase), which, in conjunction with GSH synthetase (GSS), is responsible for GSH synthesis. Both the catalytic (GCLC) and modifier (GCLM) subunits of GCL, as well as GSS, contain functional AREs (39,54,62). SLC7A11, a key component of the system xCT glutamate-cystine antiporter that imports the cystine needed to generate cysteine in exchange for glutamate, is also transcriptionally regulated by NRF2 (63). Accordingly, GSH levels have been shown to be markedly lower in Nrf2−/− mice, enhancing pathogenesis across a variety of lung and liver injury models (64–66). Thus, NRF2 is clearly required for both basal and stress-induced GSH synthesis in the cell.
Peroxide Reduction.
Along with mediating GSH synthesis, NRF2 also regulates the expression of the antioxidant enzymes that utilize GSH to reduce reactive oxygen and lipid species. In the case of lipid peroxides, GSH peroxidase 2 (GPX2), which reduces lipid peroxides to their alcohol form, is an NRF2 target gene (67,68). NRF2 also controls the expression of several thiol-based antioxidant enzymes, including thioredoxin 1 and its respective reductase (TXN1/TXNRD1), sulfiredoxin 1 (SRN1), as well as peroxiredoxins 1 and 6 (PRDX1/6) (39,50,54). Therefore, NRF2 mediates not only the transcription of the enzymes responsible for GSH synthesis but also the expression of those proteins that utilize GSH to reduce their respective reactive substrates.
Xenobiotic Metabolism.
Finally, NRF2 has also been shown to be a key mediator of phase I, II, and III xenobiotic/drug-metabolizing enzymes and membrane transporters. Phase I detoxification enzymes, including several aldo-keto reductase family members (AKR1C1, AKR1B1, and AKR1B10, which catalyze the conversion of aldehydes and ketones to their corresponding alcohols) and NADPH dehydrogenase quinone 1 (NQO1) (reduces reactive quinone species capable of generating reactive oxygen species [ROS]), are classical NRF2 target genes (54,69–71). NRF2 is also involved in increasing the expression of phase II enzymes responsible for conjugating reactive intermediates to GSH and glucuronate, such as GSH S-transferase α1 (GSTA1) and UDP-glucosyltransferase A1 (UGTA1) (72,73). Along with phase I and phase II enzyme induction, NRF2 is also involved in regulating phase III transporters, i.e., ATP-binding cassette subfamily C members 1–4 (ABCC1–4), to enhance transport and excretion of xenobiotics/drugs from the cell (74). Overall, NRF2 is a critical player in the metabolism, conjugation, and removal of harmful xenobiotics that could decrease cell function and survival.
In the following section, we will outline how NRF2, and the transcriptional programs discussed above, have been implicated in the prevention or promotion of the different forms of diabetes as well as the genetic (Fig. 2) and pharmacological (Fig. 3) evidence supporting the importance of this pathway in mediating the onset of diabetic complications. Finally, we will discuss promising therapeutic interventions related to this pathway that may have translational value for treating patients with diabetes.
The Role of NRF2 in Mediating Different Forms of Diabetes
Gestational Diabetes
As the name suggests, gestational diabetes mellitus (GDM) occurs during pregnancy and can have significant implications for the health of both the mother and offspring. Metabolically, high glucose has been shown to cause an ROS-dependent increase in NRF2 and its downstream target, NQO1, in both mouse placental tissue and human placenta cells (75). Similarly, NRF2, NQO1, and HO1 levels have all been reported to be higher in GDM than in normal placental tissue (76). Pharmacologically, administration of the NRF2 inducer tertiary butylhydroquinone during gestation restored insulin sensitivity and glucose tolerance and increased the number of surviving offspring in a pregnant leptin-deficient mouse model (db/+) of GDM (77). Catalase overexpression has also been shown to alleviate the increased ROS, renal hyperfiltration, and kidney injury observed in GDM pups via NRF2-dependent activation of HO1, although the mechanism of catalase activation of NRF2 was not reported (78). Finally, metformin reduced GDM-dependent increases in adipogenesis, which was associated with upregulation of NRF2 and downregulation of the NF-ΚB pathway (79). However, conflicting reports exist regarding whether NRF2 is up- or downregulated in GDM. For example, human umbilical vein endothelial cells derived from the umbilical cords of patients with GDM exhibited impaired NRF2 activation in the presence of the reactive lipid 4-hydroxynonenal, which was associated with decreased NQO1, xCT, and GCLM levels, depleted GSH, and a marked increase in reactive species production and subsequent protein oxidation (80). Despite the increase in ROS, the authors also observed a decrease in DJ1 and increase in p-GSK3β, both of which could decrease NRF2 levels via decreased stabilization. The well-established NRF2 target genes GSH–S-transferase pi 1 (GSTP1) and peroxiredoxin 6 (PRDX6) were also shown to be lower in adipose tissue from GDM patients than non-GDM control patients (81). Another interesting study found that maternal 25-hydroxy vitamin D (calcifediol) deficiency led to decreased NRF2 and increased ROS in the placenta of pregnant rats as well as in the liver and pancreas of the pups, leading to metabolic syndrome in both mother and offspring (82). Although these reports regarding NRF2 levels in GDM are contradictory, which may be due to the tissue tested and different GDM models used, it is clear that NRF2 reduction has detrimental consequences for both parent and offspring and that NRF2 induction could act as a possible means for therapeutic intervention.
Type 1 Diabetes
Type 1 diabetes is characterized by the destruction of pancreatic β-islet cells by autoreactive T-cells, resulting in a lack of the ability to produce insulin. Importantly, onset of type 1 diabetes most often occurs in juveniles, typically occurring around 4–7 to 10–14 years of age. Much like gestational diabetes, alterations to NRF2 function have been implicated in the onset and progression of type 1 phenotypes. For example, nonobese diabetic (NOD) mice crossed with Keap1 knockdown mice (NOD:Keap1FA/−) exhibited decreased pancreatic T-cell infiltration and increased insulin secretion compared with their NOD:Keap1+/+ counterparts (83). Furthermore, several pharmacological interventions have been shown to induce NRF2 to protect against type 1 phenotypes. For example, administration of the NRF2 inducer SF or cinnamaldehyde significantly attenuated both diabetic wound healing and nephropathy in Nrf2+/+ mice injected with streptozotocin (STZ) (84,85). Fenofibrate and minocycline have also been shown to ameliorate diabetic nephropathy (DN) in an NRF2-dependent manner in an STZ model of type 1 diabetes (86,87). Oral administration of naringenin to STZ-injected mice reduced pancreatic ROS production and β-cell apoptosis, restoring insulin production and glucose tolerance, in part through increased activity of superoxide dismutase (Sod), catalase (Cat), Gpx, G6pd, and Gst (88). Another NRF2 inducer, dh404, decreased ROS and increased the viability of islets transplanted into STZ-injected mice (89). In the heart, NRF2 activation via administration of SF significantly ameliorated diabetes-induced high blood pressure and cardiac hypertrophy in STZ-treated mice via upregulation of Ho1, Nqo1, metallothionein 1 (Mt1), Cat, and Sod1/2 (90).
Supporting the protective role of NRF2 in type 1 diabetes, Nrf2−/− mice exhibited more pancreatic β-cell damage in an alloxan-induced model of type 1 diabetes (91) as well as increased serum triglycerides, higher blood glucose levels, and increased hepatic gluconeogenesis compared with Nrf2+/+ mice following STZ injection (92). Interestingly, STZ-injected Nrf2−/− mice also exhibit more severe cognitive deficits due to hippocampal oxidative damage, an effect associated with chronic hypoglycemia (93). Our group has shown that Nrf2−/− mice injected with STZ exhibited increased oxidative damage in kidney glomeruli and decreased renal function (94). STZ-injected Nrf2−/− mice also exhibited delayed wound healing due to increased oxidative damage and keratinocyte cell death (85). However, a contradictory study found that cardiac-specific autophagy deficiency led to increased NRF2 levels and an enhanced cardiomyopathic phenotype, which could be reversed by crossing with cardiac-restricted Nrf2−/− mice (95). However, this is most likely attributed to the fact that acute/controlled NRF2 activation is protective, whereas chronic activation, which occurs in an autophagy-deficient setting, is detrimental, which we have documented extensively elsewhere (96). Overall, acute/controlled NRF2 activation prevents negative outcomes associated with type 1 diabetes progression.
Diabetes Insipidus
While much rarer, and not specifically related to diabetes mellitus, diabetes insipidus (DI), which is characterized by extreme thirst (polydipsia) and frequent urination (polyuria) due to an inability of the kidneys to properly concentrate urine, has also been shown to involve dysregulation of the NRF2 pathway. For example, in mice, whole-body knockout of aldose reductase (AKR1B1), a well-established NRF2 target gene, resulted in polydipsia and polyuria as early as 4 weeks of age (97). Keap1−/− mice, which are known to die by weaning age due to hyperkeratosis of the esophagus and stomach, can survive to adulthood by subsequently knocking out Nfe2l2/NRF2 in the esophagus (NEKO mice). However, these mice still exhibit defects in the kidney because of NRF2 hyperactivation, which results in polyuria and hydronephrosis, key phenotypes associated with DI (98). This same study found that administration of desmopressin, a synthetic antidiuretic hormone used as the primary treatment for central DI, was able to increase urine osmolality in control mice but not NEKO mice, inferring that NRF2 status could influence the efficacy of desmopressin as a treatment. A similar study found that conditional deletion of Keap1 in renal tubular epithelial cells in mice resulted in hydronephrosis and the onset of DI by 3 months of age (99). Conversely, Keap1 knockdown mice (Keap1FA/−) were resistant to lithium-induced DI compared with their wild-type counterparts, with pharmacological activation with bardoxolone methyl also showing protection (100). This infers that acute/controlled activation of NRF2 protect against DI, whereas chronic activation, particularly due to complete genetic ablation of Keap1, could be harmful.
Type 2 Diabetes
By far the most common form of diabetes, type 2 diabetes (T2D), is a complex, multiorgan disease that continues to affect a larger percentage of the population each year. Importantly, while T2D was once a disease considered to only affect older individuals, it can now be divided into early and late onset, with individuals showing early signs of diabetes progression starting in their teens. NRF2 has been shown to play a role in mediating all aspects of T2D diabetic complications across every diabetes-relevant organ. Below, we highlight the evidence supporting NRF2’s involvement in modulating diabetic phenotypes in some of the main organs involved.
Pancreas
Perhaps the most prevalent example of NRF2 mitigation of diabetic outcomes is in the pancreas, which fits based on the well-established role of ROS in inducing β-cell dysfunction and death. For example, both genetic (Keap1FA/−) and pharmacological (CDDO-Im) activation of NRF2 restored pancreatic insulin production and prevented loss of insulin sensitivity and glucose tolerance in db/db mice (101). A follow-up study using mice with β-cell–specific overexpression of inducible nitric oxide synthase (Nos2), which exhibit insulin resistance due to oxidative stress-driven β-cell dysfunction and death, showed that subsequent β-cell–specific ablation of Keap1 could rescue insulin secretion and restore euglycemia (102). SF-dependent activation of glucose-stimulated insulin secretion (GSIS) in INS-1 rat insulinoma cells was also associated with increased levels of NRF2 and its downstream targets, Srxn1, Hmox1, Gclc, and Nqo1 (103). The NRF2 activators dimethyl fumarate, CDDO-Im, and tertiary butylhydroquinone all protected mouse insulinoma cells from H2O2-induced damage and death, with silencing of Nrf2 worsening the phenotype (104). Supporting a protective role for NRF2 in maintaining β-cell function, islets isolated from Nrf2−/− mice were smaller and produced and secreted less insulin, which was accompanied by lower expression of several key NRF2-regulated antioxidant genes, including Nqo1, Gstp1, Hmox1, Gpx2, and Txnrd1 (101). Finally, NRF2 activation is required for GSIS-stimulated β-cell proliferation and growth, as Keap1 overexpression or Nrf2 knockdown inhibited GSIS, inferring that NRF2 is a regulator of both β-cell mass and function (105). At the proteostasis level, NRF2 activation has also been shown to increase autophagic clearance of ubiquitinated proteins in H2O2-treated islets, although the specific NRF2 target genes involved were not shown (106). Furthermore, shRNA-mediated silencing of Nrf2 in βTC6 mouse insulinoma cells suppressed expression of proteasomal subunit genes Psmb5 and Psmb6, enhancing toxicity following tunicamycin treatment, which could be prevented by pretreatment with the NRF2 activator 1,2-dithiole-3-thione (107). These studies indicate that mitigation of ROS through NRF2 activation is an important aspect of maintaining pancreatic function. In addition, we believe that the protective effect afforded by NRF2 activation also extends beyond just antioxidant defense mechanisms.
Liver and Adipose
Similar to the studies done looking at pancreatic insulin production, both CDDO-Im administration and Keap1 knockdown in diabetic mice (db/db:Keap1FA/−) resulted in an NRF2-dependent increase in hepatic production of fibroblast growth factor 21 (Fgf21), which in turn influenced carbohydrate metabolism (i.e., expression of Glut1 and Hk2) in adipose tissue and decreased plasma triglyceride levels (108). Similarly, Keap1FA/− mice fed a high-fat diet showed better glucose tolerance, less liver steatosis, and more repressed gluconeogenic and lipogenic pathway activity than wild-type mice (109). As mentioned above, while controlled NRF2 activation is beneficial, prolonged activation of NRF2 has been shown to enhance T2D phenotypes. For example, in a recent study from our group, we showed that chronic arsenic treatment caused insulin resistance and glucose intolerance in mice via p62-dependent activation of NRF2. Specifically, prolonged activation of NRF2 resulted in excess hepatic glucose production via the polyol pathway, which was not observed in arsenic-treated p62−/−, Nrf2−/−, or p62−/−;Nrf2−/− mice (57). In a genetic model of obesity, leptin-deficient ob/ob mice crossed with Keap1 knockdown mice (ob/ob:Keap1FA/−), while leaner than their ob/ob:Keap1+/+ counterparts due to decreased adipogenesis, exhibited impaired insulin signaling and a significant increase in hepatic lipid accumulation and the development of a fatty liver (110). Interestingly, this same phenotype was also observed upon global or adipose-specific ablation of Nrf2 in ob/ob mice (111). Adding to the complexity, another study showed that Nrf2−/− mice have higher hepatic expression of Fgf21 than the wild type, resulting in improved glucose tolerance when fed a high-fat diet, which is contrary to the studies reported above on Keap1FA/− mice (109,112). This discrepancy can most likely be accredited to the diet composition and duration of the experiment, as the study by Furusawa et al. (108) study used a diet of 60% calories from fat for 3 weeks followed by 20% sucrose in the drinking water for 5 weeks, whereas the Slocum et al. (109) and Zhang et al. (112) studies used a diet of 60% and 40% of calories from fat, respectively, for 12 weeks with normal drinking water. Another intriguing series of studies using l-buthionine sulfoximine indicated that depletion of GSH, whose synthesis is controlled by NRF2, decreased hepatic lipogenesis, plasma fatty acid levels, and adiposity in mice fed a high-fat diet. Conversely, loss of GSH enhanced insulin resistance in obese Zucker rats in the presence of the oxidant hydroquinone, indicating that GSH depletion can have differential effects on diabetes and obesity onset (113–116). The discrepancies between the above studies highlight the convoluted nature of studies investigating the role of NRF2 in type 2 diabetes and indicate that the role of NRF2 in diabetes progression, particularly when high-fat diet or genetic modifications that induce obesity are involved, is complex and context dependent, requiring further clarification.
Kidney
DN is a critical late-stage component of type 2 diabetes progression. Much like other aspects of the disease, one of the biggest risk factors for developing DN is chronic hyperglycemia, which results in increased oxidative stress and chronic inflammation, both of which play a key role in the progression from healthy kidney to the nephropathic state. As such, it is not surprising that NRF2 activation has been shown to play a protective role in preventing onset of DN. Similar to type 1 diabetes models of DN, a wide variety of natural and synthetic electrophilic/oxidative activators of NRF2 that ameliorate DN phenotypes have also been identified, including epigallocatechin-3-gallate (117), SF (118), and curcumin analog 66 (119). Independent of induction via Keap1 modification, a recently identified compound, Ab38, as well as resveratrol were shown to decrease extracellular matrix deposition and lower oxidative stress in a high-fat/STZ–induced model of DN; interestingly, A38b’s mechanism of increasing NRF2 was via decreased expression of Keap1, which differs from the normal oxidative/electrophilic modification of Keap1 cysteines (120). Omentin-1 was also shown to alleviate ROS-induced damage in a DN model via suppressing miR27-dependent inhibition of NRF2 expression (121). In another study, the glomeruli of db/db mouse kidneys were shown to have increased oxidative stress, decreased NRF2 expression, and impaired Parkin-PINK1-dependent mitophagy, all of which could be reversed by administration of the mitochondrial antioxidant mitoquinone (122). In HK-2 proximal tubular cells, hyperglycemia suppressed NRF2 and downstream expression of a newly identified target gene encoding mitochondrial ribosomal protein (MRPL12), resulting in defective oxidative phosphorylation and increased production of ROS (123).
Much like the other organs discussed above, activation of NRF2 could also be a driving force behind some of the negative outcomes associated with DN progression. For example, Keap1FA/FA mice, or mice treated with CDDO-Im, which were then administered adriamycin or angiotensin, exhibited worse proteinuria, with Nrf2−/− mice being protected from the phenotype (124). It is important to note in this model that CDDO-Im was administered after injury, which, along with the chronic activation induced in the Keap1FA/FA model, supports the notion that NRF2 can promote diabetes progression following onset of the pathology. Regardless, this idea infers that care should be taken when administering NRF2 activators, particularly in the context of chronic kidney disease, a concept that has been reviewed in full regarding the effects of bardoxolone in clinical trials (125). Overall, these studies indicate that NRF2 activation is required to protect against DN and that the development of more NRF2-specific inducers with fewer off-target toxic effects is needed to delay diabetes progression and the onset of diabetic outcomes.
Heart and Skeletal Muscle
In the heart, SF has been shown to ameliorate diabetic cardiomyopathy via NRF2-dependent activation of multiple downstream mediators, including metallothionein and AMPKα, in a high-fat diet/STZ model of T2D (126,127). Similarly, the hearts of db/db mice were shown to have decreased cardiac function because of lower NRF2 levels and Gclc/Gclm expression in the aorta and arterioles, leading to decreased GSH levels, which could be reversed by treatment with SF (128). Overexpression of Nrf2, or pharmacological activation using dh404, recovered insulin sensitivity and rescued glucose uptake in insulin-resistant cardiomyocytes (129). Hydrogen sulfide preconditioning has also been shown to reduce myocardial injury in db/db mice via upregulation of NRF2-mediated transcription of Nqo1 and Hmox1, which was associated with decreased Bach1 expression (130,131). Finally, muscle-specific overexpression of Nrf2 in db/db mice (db/db:Keap1MuKo) lowered blood glucose levels via increased glucose uptake into the skeletal muscle (132), and mice fed a high-fat diet supplemented with curcumin exhibited decreased muscular oxidative stress via upregulation of NRF2 (133). These studies indicate that NRF2 activation in the heart and skeletal muscle reduces oxidative stress and improves heart and muscle function in various genetic and dietary models of diabetes.
Conclusions and Future Perspectives
As expected, our understanding of the role of NRF2 in the complex progression of diabetes is continuing to evolve. Perhaps one of the biggest issues in the NRF2–diabetes field is the contradictory reports, which may be the result of differing genetic backgrounds and dietary conditions as well as acute/controlled versus chronic/uncontrolled activation of NRF2. This is evidenced by the fact that a key feature of pathological involvement of NRF2 in the progression of diabetic phenotypes is constitutive activation via genetic ablation or knockdown of Keap1 or prolonged NRF2 activation during autophagy inhibition by chronic arsenic exposure. Interestingly, an increasing number of NRF2 target genes that regulate metabolism and proteostasis continue to be discovered, and it will be interesting to see if NRF2 regulation of these targets plays a role in diabetes onset and progression as well. Notably, a strong body of literature supports the protective benefit of acute induction of NRF2 prior to the development of diabetic outcomes in most diabetes-relevant tissues. However, a big limitation to the therapeutic possibilities available up to this point is the risk of off-target liabilities, as electrophilic NRF2 activators can modify cysteines in any protein. This infers that a more targeted approach, such as using protein–protein interaction inhibitors that specifically disrupt the NRF2–KEAP1 interaction, represents a more viable approach to harnessing the protective power of NRF2 in diabetes prevention. Furthermore, the antidiabetic effects of several diabetes medications, including metformin, glucagon-like peptide-1 receptor agonists, liraglutide, and dipeptidyl peptidase-4 inhibitors, have all been shown to exert their effects through modulation of NRF2 signaling (134–140). Thus, the NRF2 pathway continues to represent a viable target for mediating diabetes progression. Overall, the continued identification of pathways regulated by NRF2 in a diabetes context, as well as the development of second-generation NRF2-selective inducers with fewer off-target toxic effects, should facilitate the use of NRF2-based therapeutic regimens to delay the onset of diabetes or reduce the severity of diabetic complications.
Article Information
Funding. D.D.Z. is supported by the following grants from the National Institutes of Health: R35ES031575 and P42ES004940.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. M.D., A.S., A.A., and J.C. wrote the manuscript. A.S. and M.D. made the figures. J.G.G.N. and D.D.Z. edited the manuscript.