The Warburg Effect in Diabetic Kidney Disease

Summary

Diabetic kidney disease (DKD) is the leading cause of morbidity and mortality in diabetic patients. Defining risk factors for DKD using a reductionist approach has proven challenging. Integrative omics-based systems biology tools have shed new insights in our understanding of DKD and have provided several key breakthroughs for identifying novel predictive and diagnostic biomarkers. In this review, we highlight the role of the Warburg effect in DKD and potential regulating factors such as sphingomyelin, fumarate, and pyruvate kinase muscle isozyme M2 in shifting glucose flux from complete oxidation in mitochondria to the glycolytic pathway and its principal branches. With the development of highly sensitive instruments and more advanced automatic bioinformatics tools, we believe that omics analyses and imaging techniques will focus more on singular-cell-level studies, which will allow in-depth understanding of DKD and pave the path for personalized kidney precision medicine.

Diabetic kidney disease (DKD) develops in approximately 40% of patients with diabetes and is the leading cause of chronic kidney disease worldwide.¹ Metabolic alterations associated with diabetes lead to renal pathologic changes including tubulointerstitial inflammation and fibrosis, glomerular hypertrophy, and glomerulosclerosis.¹ However, because only less than 10% of patients with diabetes ultimately reach end-stage renal disease, there must be something missing in our understanding of the pathophysiology of DKD. Genetic studies have not identified a major genetic contribution,^2,3 although it is likely that there are important genetic determinants.^4,5 As part of the International Society of Nephrology Forefronts Symposium of Systems Biology, the application of systems biology tools to understanding DKD was a major topic.

Detailed roles of mitochondria in DKD has been reviewed previously.^5–7 In the presence of hyperglycemia, the metabolic flux of glucose can be shifted from complete oxidation in mitochondria to the glycolytic pathway, and the deleterious effects of hyperglycemia are considered to be owing mainly to increased activity of five different pathways: pentose phosphate pathway,⁸ sorbitol/polyol pathway,⁹ advanced glycation end-products pathway,^10,11 protein kinase C pathway,¹² and hexosamine pathway (Fig. 1).^13,14 Activation of the earlier-mentioned pathways was reported to be associated with diabetic complications, either by causing oxidative stress, inflammation, fibrosis, DNA damage, and vascular changes, or by resulting in pathologic gene expression.⁵ Overproduction of superoxide in mitochondria might play a unifying role for the activation of the aforementioned metabolic pathways.¹⁵ New evidence indicates that alternative pathways such as enhanced fatty acid oxidation in mitochondria are involved in diabetic complications.^15,16 However, it still is unclear whether mitochondrial dysfunction in DKD results in altered glycolysis flux, or if glycolysis flux/accumulation of glycolytic intermediates in DKD drives the metabolic flux into the five pathways discussed earlier, leading to impaired mitochondrial function, or if a bidirectional causality exists. The current review high-lights the diabetes-induced metabolic switch from oxidative phosphorylation to glycolysis and the potential link to downstream effects on kidney function and disease progression.

Figure 1.

Hyperglycemia enriches metabolic flux to glycolysis and five principal branches including the polyol pathway, pentose phosphate pathway (PPP), hexosamine pathway, protein kinase C (PKC) pathway, and advanced glycation end-products (AGE) pathway. Accumulation of four toxic glucose metabolites such as lactate, sorbitol, diacylglycerol (DAG), and methylglyoxal (MG) might contribute to the development of diabetic kidney disease. Abbreviations: 5-LGST, 5-lactoylglutathione; AR, aldose reductase; DHAP, dihydroxyacetone phosphate; ENO1, alpha-enolase; ETC, electron transport chain; F-6-P, fructose 6-phosphate; G-6-P, glucose 6-phosphate; G-3-P, glyceraldehyde 3-phosphate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFAT, glutamine:fructose-6-phosphate amidotransferase; Gln, glutamine; Glu, glutamate; GluA-6-P, glucosamine-6-phosphate; GLO1, glyoxalase 1; HAGH, hydroxyacyl glutathione hydrolase; LDH, lactate dehydrogenase; OXPHOS, oxidative phosphorylation; PGM1, phosphoglucomutase-1; R-6-P, ribulose-5-phosphate; SORD, sorbitol dehydrogenase; TPI1, triosephosphate isomerase 1; UDP-GlcNAc, uridine diphosphate N-acetylglucosamine.

WARBURG EFFECT/AEROBIC GLYCOLYSIS AND ITS POTENTIAL ROLE IN THE DEVELOPMENT OF DKD

The Warburg effect or aerobic glycolysis, first observed in 1924 by Otto Warburg, has had a profound influence on cancer metabolism.¹⁷ Warburg proposed that, independent of cellular oxygen, tumor cells synthesize adenosine triphosphate (ATP) through glycolysis and a metabolic state involving enhanced glucose uptake leads to local acidification through enhanced lactate production. Although Warburg’s¹⁸ studies suggested that mitochondrial dysfunction is the root of aerobic glycolysis, subsequent studies and technical advances have since shown that mitochondria in cancer are indeed functional and that genetic or environmental cues may contribute to this metabolic shift to the glycolysis in tumor cells.¹⁹ Although enhanced glucose entry and glycolysis has been observed in diabetic tissues, the Warburg effect has not been proposed as a major feature in diabetic complications.

Recent omic studies in diabetes and diabetes kidney disease have provided unbiased evidence that both mitochondrial dysfunction and the Warburg effect play pivotal roles in the development of DKD.^16,20,21 Several groups now have found that reduced mitochondrial function plays a key role in DKD both in mouse models and in patients with DKD.^20,22,23 By using transcriptomic, metabolomics, and flux approaches, Sas et al¹⁶ reported significantly increased glycolytic intermediates and enzymes in kidney cortex, along with a significant reduction in mitochondrial function in a type 2 diabetic mouse model. Interestingly, in a very recent study, Qi et al²¹ reported that enhanced pyruvate kinase II (PKM2) activity may preserve mitochondrial function by increasing glucose flux through glycolysis in podocytes and alleviate the progression of DKD in patients with diabetes.

To understand how the Warburg effect may play a role in DKD, here we review the major metabolic pathways and associated enzymes and intermediates that are involved in DKD pathometabolism. Compared with mitochondrial respiration, aerobic glycolysis is an inefficient pathway of generating ATP per unit of glucose.²⁴ Although complete oxidation of one molecule of glucose via pyruvate within the tricarboxylic acid (TCA) cycle in the presence of oxygen generates 38 molecules of ATP, the glycolysis, the first step of glucose oxidation process that occurs in cytosol, generates only two molecules of ATP, leading to pyruvate as a final product. However, the rate of glucose metabolism through aerobic glycolysis is much higher (10–100 times faster) than the complete oxidation of glucose through the TCA cycle and oxidative phosphorylation in the mitochondria. Therefore, in diabetic patients, activation of aerobic glycolysis might help metabolize glucose rapidly from systemic circulation. In addition, with the evidence of mitochondrial dysfunction including low peroxisome proliferator-activated receptor γ coactivator 1α (PGC-1α) levels, abnormalities in electron transport chain complex assembly/activity, alterations in pyruvate dehydrogenase complex (PDH) phosphorylation, and altered TCA cycle intermediates in DKD,^20,25 it is not surprising that the glucose oxidation through glycolysis is a more favorable process toward ATP synthesis.

METABOLOMICS IDENTIFY MITOCHONDRIAL DYSFUNCTION AND WARBURG MECHANISMS IN DKD

Recent omics studies in diabetes and DKD have provided unbiased evidence that both mitochondrial dysfunction and the Warburg effect play pivotal roles in the development of DKD.^16,20,21 By using targeted metabolomics and systems biology tools we reported that human DKD is associated with mitochondrial dysfunction. Gas chromatography-mass spectrometry–based targeted metabolomics were performed and 94 urinary metabolites were quantified in patients with established DKD and compared with healthy controls.²⁰ Results indicated a marked reduction in organic anions, the TCA cycle, and amino acid metabolites. We established a 13–urinary metabolite biomarker signature of DKD, all of which were decreased significantly. Correlation analysis between each of 13 metabolites and common biomarkers of DKD showed that five of the metabolites (ie, 3-hydroxy isovalerate, aconitic acid, citric acid, glycolic acid, and uracil) were correlated significantly with estimated glomerular filtration rate, and another three metabolites (ie, 2-methyl acetoacetate, 3-methyl crotonyl glycine, and 3-methyl adipic acid) were associated with albuminuria.²⁰ In particular, 12 of the 13 metabolites are associated with mitochondrial metabolism, the metabolites are either generated within the TCA cycle or regulated by the mitochondrial enzymes (Table 1), suggesting suppressed mitochondrial activity in DKD.²⁰ Further analysis of kidney biopsy samples and urinary exosomes showed a significant reduction in PGC-1α, mitochondrial proteins, and mitochondrial DNA levels in diabetic nephropathy patients as compared with healthy controls. These findings suggest an overall reduction in mitochondrial biogenesis in the kidney of DKD patients, contributing to a reduction in the TCA cycle metabolites and a potential shift to glycolysis.

Table 1.

Thirteen Urinary Metabolite Biomarkers for DKD and Their Associated Enzymes and Pathways

Metabolite Decreased in DKD	HMDB ID of Metabolite	Function in Intermediary Metabolism	Enzyme(s) Producing the Metabolite	Subcellular location of Enzymes
3-Hydroxyisovaleric acid	HMDB00754	Leu metabolite	3-methylglutaconyl CoA hydratase	Mitochondria
Glycolic acid	HMDB00115	Gly (peroxisomes) and 4-hydroxyproline (mitochondria)	NADPH-glyoxylate reductase	Peroxisomes, mitochondria
Citric acid	HMDB00094	The TCA cycle and lipid synthesis	Citrate synthase	Mitochondria
2-Ethylhydracrylic acid	HMDB00396	Ile metabolite	From R-pathway of Ile metabolism (with 2MBDH deficiency)	Mitochondria
Uracil	HMDB00300	Pyrimidine synthesis	Coenzyme Q10: dihydroorotate dehydrogenase, UMPS	Mitochondria
3-Hydroxyisobutyric acid	HMDB00023	Val metabolite	3HIBCH	Mitochondria
Aconitic acid	HMDB00072	The TCA cycle	Aconitase	Mitochondria
3-Methyladipic acid	HMDB00555	Indicates incomplete BCFA oxidation	From decreased intake of phytanic acid or increased α-oxidation of BAFA	Mitochondria
Tiglylglycine	HMDB00959	Ile metabolite	FAD and 2MBDH	Mitochondria
3-Methylcrotonylglycine	HMDB00459	Leu metabolite	FAD and IVD	Mitochondria
2-Methylacetoacetic acid	HMDB03771	Ile metabolite	NAD+ and MHBD	Mitochondria
Homovanillic acid	HMDB00118	Dopamine metabolite	COMT and MAO	Cytosol, Mitochondria
Hydroxypropionic acid	HMDB00700	Ile, Val, Thr, and Met metabolite	2MAACT (Ile), NAD+ and MMSDH (Val), NAD+ and 2KBDH (Thr and Met)	Mitochondria

Abbreviations: 2MAACT, 2-methylacetoacetyl Coa thiolase; 2MBDH, 2-methylbutyryl-CoA dehydrogenase; 2KBDH, 2-ketobutyrate dehydrogenase; 3HIBCH, 3-hydroxyisobutryl-CoA hydrolase; BAFA, branched-chain fatty acids; COMT, catechol-O-methyl transferase; FAD, flavin adenine dinucleotide; Gly, glycine; Ile, isoleucine; IVD, isovaleryl-CoA dehydrogenase; Leu, leucine; MAO, monoamine oxidase; Met, methionine; MHBD, 2-methyl-3-hydroxybutyryl CoA dehydrogenase; MMSDH, methylmalonate semialdehyde dehydrogenase; NAD+, nicotinamide adenine dinucleotide; Thr, threonine; UMPS, uridine monophosphate synthetase; Val, valine.

Adapted with permission from Sharma et al.²⁰

The mechanisms of mitochondrial dysfunction and reduction in TCA metabolites in DKD are not fully understood. It is unclear if low TCA cycle metabolite flux is the causal or consequence of the mitochondrial dysfunction. Recent studies by Sas et al¹⁶ using a systems-based approach combining transcriptomic, metabolomic, and metabolomic flux in a mouse model of type 2 diabetic (db/db) kidney cortex and in type 2 diabetic human patients identified enhanced metabolic flux into glycolysis, fatty acid oxidation, and TCA cycle pathways. Several of the glycolytic enzyme transcripts including hexokinase, phosphofructokinase, and pyruvate kinase were increased significantly in glomeruli-depleted kidney cortex of the db/db mice as compared with the control db/m mice. No significant changes were observed in the expression of TCA cycle pathway-related genes, suggesting a diabetes-induced metabolic shift to glycolysis. This was confirmed further by metabolomic analysis of kidney cortex, urine, plasma, and isolated mitochondria. Several glycolytic metabolites were up-regulated in the diabetic mouse kidney and urine at both 12 and 24 weeks. Likewise, several of the TCA cycle metabolites and acylcarnitines were increased in kidney cortex, isolated mitochondria, and urine samples of the diabetic mice, suggesting that mitochondrial metabolic alterations in the kidney cortex are reflective of whole-body metabolism. The TCA cycle metabolites and acylcarnitines were increased significantly in the mitochondria from kidney cortex of 24-week-old compared with 12-week-old mice, indicating a progressive increase in metabolism in the mitochondria with disease progression in DKD. Metabolomic flux analysis using [U-¹³C₆]glucose, [2,3-¹³C₂]pyruvate, and [¹³C₁₆]palmitate were performed to assess flux through glycolysis, the TCA cycle, and fatty acid oxidation, respectively. Results from these studies showed that db/db kidney tissues had significantly increased flux of all three metabolic pathways, although there were no changes in corresponding ATP. Analysis of oxygen consumption of isolated mitochondria from kidney cortex further identified that the db/db kidney mitochondria have reduced state 3 respiration (ADP-stimulated respiration) and enhanced proton leaking, suggesting uncoupling of electron transport chain from ATP synthesisoccurs. Therefore, a compensatory increase in metabolic flux and a switch to alternate pathways including glycolysis for ATP synthesis occurs. Interestingly, the increased metabolism also was associated with enhanced protein acetylation. Sirtuin 1 (SIRT1) dependent deacetylation of PGC-1α is one of the major regulatory mechanisms that PGC-1α activity and PGC-1α levels and activity have been shown to be reduced in several animal models and human studies of DKD.^26–28

SPHINGOMYELIN: A NOVEL MEDIATOR FOR ENHANCED ATP PRODUCTION VIA GLYCOLYSIS IN DIABETES

Adenosine monophosphate (AMP)-activated protein kinase (AMPK) is a major cellular energy sensor and is activated in the state of caloric depletion (eg, low ATP/high AMP).²⁹ Reduced glucose oxidation in diabetes is expected to decrease ATP and increase AMP and AMPK activity. However, a reduction in renal AMPK activity was reported in human DKD and mouse models of DKD and in high-fat-diet–fed mice.^23,30 PGC-1α is a key regulator of mitochondrial biogenesis.³¹ Suppressed AMPK and PGC-1α path-ways have been recognized in diabetic kidneys from both human and animal models, however, the regulating mechanisms were not fully illustrated.^23,32

One of the hypotheses for energy regulation was that high levels of circulating glucose may increase the glycolytic flux and increase glycolytic ATP and ATP/AMP, which suppresses AMPK activity and PGC-1α expression. The hypothesis of an enhanced glycolytic pathway in diabetes was supported by recent studies from our laboratory using mass spectrometry imaging (MSI)-based spatial metabolomics in DKD animal models and human kidney tissue.³³ We applied matrix-assisted laser desorption/ionization (MALDI)-MSI to localize and quantify renal nucleotides.³³ MALDI-MSI at 70-µmol/L resolution identified ATP, adenosine diphosphate, and AMP widely distributed throughout the kidney and showed a significant increase in ATP and decrease in AMP, leading to significantly increased ATP/AMP in glomeruli of diabetic mice when compared with nondiabetic ones. Furthermore, biochemical analysis of freshly extracted nucleotides from total kidney cortex showed a similar increase in total ATP levels in db/db diabetic mice compared with the db/m controls.

To identify analytes that may regulate ATP production in the diabetic kidney, we used untargeted high spatial resolution (25-µmol/L resolution) MALDI-MSI to analyze both human and mouse kidney sections. Particularly, three peaks (ie, m/z values) of analytes were distributed mainly in mouse and human glomeruli compared with the nearby regions. Based on high accuracy measurements of parent m/z’s using high resolving power Fourier transform ion cyclotron resonance mass spectrometry and their subsequent tandem mass spectrometry analysis, three m/z values, 703.578 (+H⁺), 725.558 (+Na⁺), and 741.534 (+K⁺), were annotated as a specific sphingomyelin (ie, SM [d18:1/16:0]) in METLIN and LIPID MAPS databases.^34,35 High spatial resolution MALDI-MSI in kidney tissue samples from type 1 diabetic and high-fat diet–fed mice tissue samples identified a significant accumulation of glomerular SM(d18:1/16:0) compared with the controls.

Sphingolipid metabolism is a critical cellular process and has been characterized extensively in the mammalian system.^36,37 During the de novo biosynthesis ceramide is converted to SM by SM synthase (SMS), whereas SM levels are maintained by sphingomyelinase (SMase), which regulates lipid homeostasis by degrading SM to ceramide.³⁷ Accumulation of ceramide and other sphingolipid signaling intermediates have been reported in type 1 and type 2 diabetic animal models and human studies.^38,39 By using immunohistochemical staining of SMS and SMase, we found increased expression of SMS1 and SMS2, and decreased neutral SMase2 in the glomeruli of diabetic mice compared with WT mice. In addition, in murine mesangial cells, we found that SM (d18:1/16:0) suppressed AMPK activity and reduced PGC-1α protein expression.³³ The liposomal SM (d18:1/16:0) treatment significantly increased lactate secretion in murine mesangial cells, suggesting that accumulation of SM and an increase in ATP in the diabetic kidney may be owing to the result of an increase in glycolysis. This was proved further by the treatment of glycolytic inhibitor 2-deoxy-D-glucose, which significantly inhibited ATP production, suggesting the potential role of SM (d18:1/16:0) in enhancing the glycolytic pathway to enhance ATP production and thus contribute to a Warburg type effect in glomeruli of the diabetic kidney (Fig. 2). These findings suggest alterations to sphing-homyelin signaling in DKD may contribute to enhanced ATP synthesis via glycolysis leading to metabolic imbalance and oxidative stress resulting in tissue damage.

Figure 2.

Accumulation of sphingomyelin (SM), a potential mediator in enhancing glycolysis-derived ATP, leads to reduced phosphorylation of AMP-activated protein kinase (pAMPK) and mitochondrial function. Abbreviations: ADP, adenosine diphosphate; ATPase, ATP synthase; ETC, electron transport chain; G-6-P, glucose 6-phosphate; Mito, mitochondrial.

PKM2 IN DKD: NOVEL MECHANISMS LINKING GLYCOLYSIS TO MITOCHONDRIAL FUNCTION

Although current understanding is that hyperglycemia enhances glycolysis and other toxic glucose-metabolite pathways as outlined in Figure 1, a recent study from the Joslin Diabetes Center has found that higher enzyme levels corresponding to glycolytic pathways can protect against diabetic nephropathy by reducing toxic glucose metabolites and increasing mitochondrial function. The study analyzed proteins from postmortem glomeruli of the kidneys of patients with long-standing diabetes with (nonprotected group) and with-out (protected) histologic signs of DKD,²¹ and showed that the protected group had increased levels of proteins related to glucose metabolism compared with nonprotected individuals. In particular, enzymes from the glycolytic (PKM2, PKM1, and enolase 1), methylglyoxal (eg, glyoxalase 1, and hydroxyacyl glutathione hydrolase), polyol (aldose reductase and sorbitol dehydrogenase), and mitochondrial nicotinamide adenine dinucleotide (NADH):ubiquinone oxidoreductase (complex I) and cytochrome c oxidase (complex pathways were up-regulated in the protected group. Up-regulation of glycolytic, methyl glyoxal, and polyol pathways have been considered to contribute to induce deleterious effects in diabetes. Therefore, the finding of an increase in these pathways in protective individuals was rather unexpected. Plasma metabolomics analysis further identified that an increase in glomerular enzymes may be associated with a decrease in metabolites from sorbitol and methyl glyoxal pathways in protected individuals, suggesting an increased ability to process the toxic metabolites by up-regulating the enzymes. In addition to the glycolytic enzymes, the study also found an increase in mitochondrial complex I– and complex IV–related proteins in the protected group, suggesting improved mitochondrial function. In addition, proteomic studies indicated that the protected group had significantly increased glycolytic enzyme pyruvate kinase M2 isoform. Pyruvate kinase, a rate-limiting enzyme in the glycolytic pathway, catalyzes the irreversible conversion of phosphoenolpyruvate to pyruvate.²¹ There are four PK isoforms in mammals including PKL (in liver), PKR (in red blood cell), PKM1 (predominantly in adult muscle, brain, bladder, and fibroblasts), and PKM2 (in most cells except for adult muscle cells).^40,41 The role of PKM2 in glycolysis and cancer progression (eg, the Warburg effect) has been reviewed by several investigators.^41–44 PKM2 is overexpressed in the majority of cancer cells, such as renal cell carcinoma.^45–48 Increased PKM2 is associated with increased glucose uptake and lactate production, and reduced oxygen consumption, therefore PKM2 has been well studied for its role in cancer and the Warburg effect.^41,46 The study by Qi et al²¹ using animal models as well as PKM2-specific activators in mice reported that PKM2 activation enhances glucose flux through glycolysis, reduces toxic metabolites such as sorbitol and methylglyoxal, and also enhances mitochondrial metabolism by improving PGC-1α, mitochondrial biogenesis, and mitochondrial function. Their findings suggest that in podocytes, activation of PKM2 may be beneficial in protecting against diabetic nephropathy, however, the regulation of PKM2 is complex and whether sole regulation of PKM2 will be sufficient in human disease remains to be determined. Nevertheless, the proteomic, metabolomic, and transcriptomic studies from human samples all indicate a key role for glycolysis and mitochondrial metabolism to be critical in human diabetic nephropathy.

REACTIVE OXYGEN SPECIES, NOX4, AND FUMARATE IN DKD

Oxidative Stress and NOX4 in DKD

Oxidative stress resulting from enhanced reactive oxygen species (ROS) is recognized as a key factor in the development of diabetic complications.^8,49 The major sources of endogenous ROS in the diabetic kidney include mitochondria, nicotinamide adenine dinucleotide phosphate (NADPH) oxidases, nitric oxide synthases, xanthine oxidase, and lipoxygenase.^8,49,50 ROS include the free radical superoxide anion $({\mathrm{O}}_2^{.-})$, the nonradical hydrogen peroxide (H₂O₂), highly reactive hydroxyl free radical (.OH), peroxynitrite (ONOO⁻), and singlet oxygen (¹O₂). Specifically, ${\mathrm{O}}_2^{-}$, H₂O₂, and ONOO⁻ have been investigated mostly in the diabetic kidney.

Oxidative stress occurs in tissues when excessive ROS or oxidant production exceeds local antioxidant capacity. Studies have shown that accumulation of ROS in diabetes induces β-cell dysfunction, increases insulin resistance, and enhances diabetes-related complications in both type 1 and type 2 diabetes.⁵¹ Almost all published studies have shown a uniform over-production of ROS in the tissues of animals and patients with diabetes.^52–54 In particular, a common belief is that mitochondrial superoxide is a major contributor for diabetic complications, such as DKD.^49,55,56 However, anti-oxidant–based clinical trials have proven to be largely negative.⁵⁷ Based on the new evidence, the theory of "mitochondrial hormesis" has been proposed wherein optimal levels of ROS are considered a marker for functional mitochondria.⁸⁵ In steptozotocin-induced and Akita type 1 diabetic mouse models, using a combination of in vivo real-time imaging and electron paramagnetic resonance approaches, we reported that diabetic kidneys have reduced overall ${\mathrm{O}}_2^{-}$ along with a reduction in mitochondrial function, PDH activity, and mitochondrial biogenesis, which all were rescued with activation of AMPK using an AMPK activator.²³ A persistent reduction in mitochondrial oxidative phosphorylation can lead to a reduction in mitochondrial superoxide and a simultaneous activation of cytosolic oxidative pathways triggering cytosolic ROS, resulting in increased oxidative stress. Indeed, the PDH complex, a major gateway for pyruvate carbons to enter the TCA cycle in mitochondria, was hyperphosphorylated and inhibited in diabetic kidneys.⁵⁸ Therefore, although convincing data support increases in oxidative stress in diabetes, the source of oxidative stress and ROS needs further investigation.

NADPH oxidase (NOX) family enzymes are one of the major sources of cytosolic ROS. NOX enzymes catalyze the reduction of molecular oxygen to super-oxide and hydrogen peroxide using NADPH as an electron donor.⁵⁹ Among the seven members of the NOX family, NOX1, NOX2, and NOX4 are expressed in kidneys of human beings and mice.^60,61 In the NOX family, NOX4 in particular has been reported to be upregulated in DKD by different researchers, including our laboratory.^61–64 Several studies have shown the beneficial effects of NOX inhibitors in reducing oxidative stress in DKD.^22,65 However, the mechanisms underlying NOX4 in DKD and the key down-stream targets that mediate the progression of DKD still remain unclear. Recently, in our laboratory, by using podocyte-specific NOX4 transgenic mice, we showed that podocyte-specific induction of NOX4 alone was sufficient to induce characteristic glomerular changes associated with DKD.²² By using urinary metabolomics in type 1 Akita diabetic mice after intervention with a NOX1/4 inhibitor, we identified a reduction in the specific TCA cycle metabolite fumarate.

Fumarate in DKD

In the TCA cycle fumarate is generated from the precursor adenylsuccinate by succinate dehydrogenase and is converted to malate by fumarate hydratase (FH).²² The accumulation of fumarate can be attributed to the inhibition of FH or stimulation of succinate dehydrogenase. Our further studies showed that FH protein is reduced in the F1 Akita and in human diabetic kidney tissue. In addition, FH levels were reduced in the podocyte-specific Nox4 transgenic mice and in the cells expressing NOX4, suggesting FH is a direct target of NOX4. Administration of the NOX1/4 inhibitor increased glomerular FH levels in diabetic mice, suggesting NOX4 have a regulatory role in FH expression. In addition to NOX4, H₂O₂ treatment also reduced the expression of FH and increased fumarate levels in HEK293 cells, suggesting that the down-stream effects of NOX4 can be attributed to accumulation of H₂O₂. FH gene mutations have been reported in leiomyomatosis and renal cell cancer. Fumarate, known as an oncometabolite, accumulates in FH-deficient cells and has been shown to inhibit α-ketoglutarate–dependent dioxygenases including hypoxia inducible factor (HIF)-prolyl hydroxylases, thereby increasing HIF-1α levels. In our study, we found that both HIF-1α messenger RNA and protein levels were increased upon treatment with fumarate or with H₂O₂. Integrative data from three tiers including gene (FH) expression, protein (FH) analysis, and metabolomic (fumarate) analyses indicate that FH is a target of NOX4 in diabetic kidney.²² A reduction in FH and the subsequent accumulation of fumarate in diabetic kidney may result in renal pathology, potentially through the inhibition of HIF prolyl hydroxylase and increasing HIF-1α (Fig. 3).^66–68

Figure 3.

Fumarate accumulation, HIF-1α regulation, and glycolysis in renal cells with increased intracellular glucose. The reduced pAMPK enhances NOX4 leading to increased H₂O₂, which reduces fumarate hydratase (FH) levels and then leads to fumarate accumulation. Both increased H2O2 and fumarate contribute to ER stress and thereafter apoptosis. The increased fumarate suppresses HIF-PHD activity resulting in HIF-1α accumulation, which drives the metabolic flux to the glycolytic pathway (eg, through increased LDH) while suppresses oxidative phosphorylation (eg, through reduced PDK and increased PDH) and reduces mitochondrial ${\mathrm{O}}_{2·}^{-}$ HIF-1α also increases levels of glucose transport proteins (eg, GLUT1 and GLUT4) and growth factors (eg, VEGF and TGF-β). Abbreviations: ADP, adenosine diphosphate; G-6-P, glucose 6-phosphate; GLUT1/GLUT4, glucose transporters; LDH, lactate dehydrogenase; PDK, pyruvate dehydrogenase kinase; PHD, prolyl hydroxylase; TGF-β, transforming growth factor β; VEGF, vascular endothelial growth factor.

HIF-1α regulation in diabetic kidney and its role in stimulating glycolysis for lactate production via enhanced lactate dehydrogenase and inhibiting PDH and oxidative phosphorylation via increased pyruvate dehydrogenase kinase has been elaborated.⁴⁹ HIF-1α and HIF-2α are transcription factors that up-regulate aerobic glycolysis-associated genes (eg, glucose transporter and vascular endothelial growth factor) in cancer. Both glucose transporter and vascular endothelial growth factor enhance glucose transport and tumor vascularity.⁶⁶ In both in vitro and in vivo studies, NOX4 activation sufficiently induces HIF-1α and the NOX1/NOX4 inhibitor reduces HIF-1α in DKD, therefore NOX4 might be an important mediator for HIF-1α in DKD.²² Moreover, it was reported that NOX4 is a novel target gene of HIFs, whose levels also are regulated by NOX4.^69,70 Results from MSI, gene expression, protein, and metabolomic analyses suggest a potential feedback loop mechanism (ie, increased ATP/AMP, reduced AMPK, activated NOX4, increased H₂O₂, suppressed FH, accumulated fumarate, increased HIF-1α, and enhanced glycolysis) in the development of DKD.⁴⁹ More sophisticated approaches are needed to accurately determine the sources of ATP and ROS that can provide further insights into the pathogenesis of DKD.

UNDERSTANDING DKD USING A SYSTEMS BIOLOGY APPROACH

Genomic, transcriptomic, proteomic, and metabolomic/lipidomic biomarkers of DKD have been reviewed previously.^71–75 In particular, new noninvasive urinary and serum/plasma biomarkers for DKD provide valuable insights in interpreting the disease pathogenesis and show potential roles in clinical medicine. To better understand the role of DKD biomarkers in the development of disease, Saito et al⁷⁶ integrated 13 previously reported DKD urinary metabolite bio-markers with publicly available protein–protein interaction networks.²⁰ Multi-omics studies and bioinformatics tools (eg, Cytoscape [UC San Diego, La Jolla, CA] and MetBridge Generator [UC San Diego, La Jolla, CA]) have shown that mouse double minute 2 homolog in glomeruli and tubules of DKD had the highest number of protein–protein interaction connections, and its reduction was associated with two decreased metabolites biomarkers (ie, 3-methylcroto-nylglycine and uracil) in DKD.^76,77

Experimental evidence supports that an altered glycolytic pathway is associated with the development of DKD.²¹ However, only a limited number of glycolysis-related enzymes, intermediates, or genes were studied. What is the degree of aerobic glycolysis in DKD? Whether it benefits or impairs the kidney still is not fully known. Beyond ATP production, aerobic glycolysis also support macromolecular synthesis (eg, DNA, RNA, proteins, and lipids) and connections between the glycolytic pathway and biosynthetic precursors were elaborated.⁷⁸ For example, the glycolytic intermediate dihydroxyacetone phosphate is the precursor to glycerol-3-phosphate, which is vital for the biosynthesis of triacylglycerols and phospholipids that serve as primary structural lipids in cell membranes. Dihydroxyacetone phosphate is also the precursor for other lipid species such as cardiolipin, which is an important component of mitochondrial membranes and plays a pivotal role in mitochondrial function.⁷⁹ Glycolytic intermediates (eg, 3-phosphoglycerate and pyruvate) also are direct precursors for the biosynthesis of amino acids (eg, alanine, cysteine, glycine, and serine).⁷⁸

FUTURE PERSPECTIVES

As a complex multifactor-derived disease, the patho-mechanism of DKD is still not fully clarified. Omics-based systems biology, which incorporate different tiers of biological data, has shed new insights into our understanding of DKD and has provided several key breakthroughs for novel predictive and diagnostic biomarkers. Recent omics-based investigations coupled with systems biology tools have shown that mitochondrial dysfunction,²⁰ oxidative stress (eg, ROS),⁴⁹ alterations of multitier information (eg, genes, proteins, and metabolites), and major metabolic pathways (eg, the TCA cycle, glycolysis, fatty acid metabolism, and amino acid metabolism)^16,21,22 all are altered in a manner consistent with a Warburg-type effect in diabetic nephropathy.

The end products of glycolysis are multidirectional. To understand how alterations of the glycolytic path-way contribute to the development of DKD, further comprehensive metabolic flux studies are required and integrate different tiers of information. By comparing different biological samples (eg, blood, urine, renal biopsy, and nephrostomy) obtained from nonprotected (ie, diabetes and DKD) and protected (ie, diabetes without DKD), we will know which metabolic path-ways are more activated or suppressed in DKD patients. Mouse models of DKD and in vitro studies also are warranted to confirm and validate the human data. Furthermore, with the development of high sensitive technologies such as global DNA methylations,⁸⁰ singular-cell RNA sequencing,⁸¹ RNA bar-coding based on single-molecule fluorescence in situ hybridization,⁸² MALDI-MSI,³³ laser capture micro-dissection -omics,⁸³ and cytometry by time of flight,⁸⁴ it becomes feasible to measure multiomics (ie, genomics, transcriptomics, proteomics, and metabolomics/lipidomics) data in a single-cell or cell type. In the near future, we could link pathologic changes with multi-omics in glomeruli, podocytes, and tubules of DKD, which will allow in-depth understanding of DKD and pave the path for personalized kidney precision medicine.