CaM Kinase II-δ Is Required for Diabetic Hyperglycemia and Retinopathy but Not Nephropathy

Since 1980, the global prevalence of type 2 diabetes (T2D) has doubled. In 2014, >420 million people were suffering from T2D (1). Late diabetic complications in organs such as the kidney and eye can lead to kidney failure and blindness. Hyperglycemia has long been suggested to cause diabetic late complications (2). Based on this assumption, numerous therapeutic strategies were developed to reduce blood glucose levels. However, some patients with well-adjusted glycemic control still develop late diabetic complications, whereas others with less adequate control do not (3). Thus, it must be debated whether hyperglycemia is only a symptom of diabetes whereas other diabetes-related metabolic signaling cause late diabetic complications.

Here, we focused on CaM kinase II (CaMKII), which has been shown to be activated in diabetes (4). CaMKII is a serine/threonine protein kinase with a broad spectrum of substrates, consisting of four isoforms. The α- and β-isoforms are almost exclusively expressed in the brain, whereas the δ- and γ-isoforms are ubiquitously expressed, with higher expression levels of CaMKIIδ in most organs (5). Besides Ca²⁺/calmodulin-binding, diabetes-related posttranslational modifications, including oxidation and O-GlcNAcylation, lead to CaMKII activation (6–8). CaMKII in turn is known to regulate hepatic glucose production (HGP), thereby potentially elevating plasma glucose levels (9–11). However, the role of CaMKII for late diabetic complications is unclear.

We crossed leptin receptor-mutant mice (Lepr^db/db) as a model of T2D to mice with a global CaMKIIδ deletion (knockout [KO]). We show for the first time that CaMKIIδ plays a crucial role in regulating diabetic hyperglycemia. To our great surprise, CaMKIIδ was required for diabetic hyperglycemia and retinopathy but not nephropathy.

CaMKIIδ-KO mice (12) were crossed to Lepr^db/db mice (The Jackson Laboratory) to obtain CaMKIIδ-deficient Lepr^db/db (Lepr^db/db/KO) and Lepr^+/+ (Lepr^+/+/KO) compared with Lepr^db/db/wild-type (WT) and Lepr^+/+/KO. These mice were maintained in a C57BL/6J genetic background, were fed a standard diet, and were kept on a 12-h light and dark cycle at 22 ± 2°C and room humidity of 55%. All experimental procedures were reviewed and approved by the Regierungspräsidium Institutional Animal Care and Use Committee, Karlsruhe, Germany.

Blood glucose was measured with Accu-Check Aviva III (Roche). Glycated hemoglobin (HbA_1c) was determined by cation-exchange chromatography on a PolyCAT A column (PolyLC). The relative amount of HbA_1c is expressed as a percentage of hemoglobin, based upon the area under the respective peaks.

Mice (12–16 weeks old) were starved for 12 or 4 h, respectively, then glucose (2 g/g body wt dissolved in 0.9% NaCl) or insulin (0.0005 units/g body wt dissolved in 0.9% NaCl) were injected intraperitoneally. Blood glucose was measured from tail vein blood at baseline and at 15 min, 30 min, 60 min, and 120 min after the intraperitoneal injection.

Acellular capillaries and pericyte loss were evaluated in retinal digest preparations as described previously (13). For visualization of the retinal morphology and morphometry, periodic acid Schiff (PAS) staining was performed on digested retinas. Quantification of acellular capillaries was done using the Cell^F software (Olympus Opticals, Hamburg, Germany). The number of acellular capillaries was calculated to the total retinal area. The number of pericytes was quantified in the 10 randomly selected fields in the central retina under magnification ×400 by a blinded observer.

Kidneys were fixed in 4% formalin and embedded in paraffin. The kidney paraffin sections were stained with PAS and scanned at magnification ×40 using the Aperio AT2 Digital Pathology Slide Scanner (Leica, Wetzlar, Germany) for digital microscopy. The mean mesangial expansion was quantified blinded in 60 randomly selected glomeruli per kidney with ImageJ software.

The method of estimating the thickening of the basal membrane was described previously (14). In brief, specimens were fixed with Karnovsky fixative, dehydrated with an alcohol series, and embedded in araldite (SERVA). Sections of 70-nm thickness were cut using an Ultracut E microtome (Leica). Sections were analyzed with a Zeiss EM 900 or 910 transmission electron microscope (Zeiss, Oberkochen, Germany) and macrographs were taken with a CCDK2 camera (TRS-Tröndle, Dünzelbach, Germany). Basal membrane thickness was estimated with ImageSP software (TRS-Tröndle).

Proteinuria was measured with collected spot urine using the Mouse Albumin ELISA Quantification Kit (E90-134; Bethyl Laboratories). In brief, a 96-well plate was coated (1 h), washed (five times), and blocked (30 min) with corresponding antibody and solutions. Diluted urine samples and a prepared standard were added in duplicates and incubated for 1 h. Subsequently, the plate was washed again, and the horseradish peroxidase detection antibody was added to each well and incubated for another hour. Afterward, the plate was developed protected from light for 15 min with 3,3′,5,5′-tetramethylbenzidine (TMB) substrate solution, and the reaction was stopped with stopping dilution. The absorbance was measured in a microplate reader at 450 nm (EnSpire Multimode Plate Reader; Perkin Elmer).

Paraffin kidney sections were dewaxed, rehydrated, and then incubated with citrate-based antigen retrieval solution (H-3300; Vector Laboratories). Endogenous peroxidase activity was blocked by incubating the sections with 3% hydrogen peroxide solution. Sections were then blocked with PBS containing 3% donkey serum and 0.5% Triton X-100 for 1 h at room temperature. Sections were incubated with primary antibody against WT-1 (ab89901; Abcam) in a dilution of 1:300 overnight at 4°C. The rest of the staining protocol and analysis was performed as previously described (15).

Paraffin kidney sections were dewaxed, rehydrated, and then stained using the trichrome staining kit (HT15; Sigma-Aldrich) according to the manufacturer’s instructions. The slides were dehydrated and mounted using anhydrous mounting medium (6638.1; Carl Roth). Ten visual fields per mouse kidney were randomly selected, and the percentage of fibrotic area was determined using ImageJ software.

Proteins were isolated from homogenized tissue or from transfected M1 and MES13 cells with radioimmunoprecipitation assay buffer. Protein concentrations were determined with bicinchoninic acid. Western blot analysis was performed with 10% SDS gels, and proteins were blotted on nitrocellulose membranes. Membranes were blocked with 5% skim milk or 5% BSA and incubated with primary antibodies overnight and secondary antibodies for 1 h the next day. Detection was performed with Luminol Reagent (Santa Cruz Biotechnology). Antibodies used for immunoblotting were anti-GLUT1 (1:1,000; Abcam), anti-GLUT4 (1:1,000; LifeSpan BioSciences or 1:1,1000; Santa Cruz Biotechnology), anti-CaMKII (1:1,000; BD Bioscience), anti-tubulin (1:2,000; Sigma-Aldrich), anti-GAPDH (1:10,000; Chemicon), anti-p38 mitogen-activated protein kinase (MAPK) (1:1,000; Cell Signaling), anti–phosphorylated (p)-p38MAPK (1:1,000; Cell Signaling), and β-actin (1:2,000; 4967S; rabbit; Cell Signaling).

RNA was isolated using TRIzol (TRI Reagent; Sigma-Aldrich). cDNA synthesis was performed with the First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). Quantitative PCR was performed with the Universal Probe Library System (Roche) using the TaqMan Universal PCR Master Mix (Applied Biosystems) and detection on a 7,500 fast cycler (Applied Biosystems). Primers and probes used were 5′-aacccgttttctgggttga-3′ and atgtgtgggcgatgacatt (probe #105) for Pck-1, 5′-gaaagtttcagccacagcaa-3′ and tctgtcccggatctaccttg (probe #19) for G6pc, 5′-acaccatgctccgtcctg-3′ and 5′-gtcattggtgttggcttgtg-3′ (probe #98) for SGLT2, 5′-cctggccatagaggtgga-3′ and ggggagaggtatccaggtgt (#32) for CaMK2a, 5′-agccccaaaggatctctcc-3′ and gggttatggataacggtggtt (#74) for CaMK2b, 5′-gatcaaagctggagcctacg-3′ and gcttcaggagtgactgtgtcc (#9) for CaMK2g, 5′-agttcacagggacctgaagc-3′ and cgccttgaacttctatggcta (#68) for CaMK2d, and 5′-ctggcaggccgaagtatg-3′ and ttccaatgttactggcaaagag (#49) for SGLT1. Relative changes in gene expression measured with real-time quantitative PCR were analyzed using the 2^−ΔΔCT method (16).

Murine kidney epithelial (M1) and mesangial (MES13) cells, immortalized with SV40 large T antigen, were obtained from ATCC (Mannassas, VA). A mixture of four siRNAs targeting murine CaMKIIδ were purchased from Horizon Discovery Ltd (siGENOME, M-040821-01). Then, 1 × 10⁶ cells were transfected with 500 pmol siRNA in 100 μL resuspension buffer by electroporation using the Neon Transfection System (Thermo Fisher Scientific, Karlsruhe, Germany). Cells were also transfected with a nontargeting pool (siGENOME, D-001206-14) as a negative control.

Glycogen and sorbitol content of muscle tissue(s) were measured with an end point, colorimetric assay (BioVision) according to the manufacturer’s instructions and normalized to total protein content, as determined by the Bradford assay.

Plasma insulin was measured by sandwich immunoassay (ALPCO), according to the manufacturer’s instructions.

Methylglyoxal (MG) levels were determined by liquid chromatography, followed by tandem mass spectrometric detection, as described previously (17).

Most of the procedure was done with R and bioconductor using the Next-Generation Sequencing (NGS) analysis package systempipeR (18). Quality control of raw sequencing reads was performed using FastQC (Babraham Bioinformatics). Low-quality reads were removed using trim_galore (version 0.6.4). The resulting reads were aligned to mouse genome version GRCm38.p6 from Genecode and counted using kallisto version 0.46.1 (19). The count data were transformed to log₂-counts per million using the voom function from the limma package (20). Differential expression analysis was performed using the limma package in R. A false-positive rate of α = 0.05 with false discovery rate correction was taken as the level of significance. Volcano plots and heat maps were created using ggplot2 package (version 2.2.1) and the ComplexHeatmap package (version 2.0.0) (21). Pathway analysis was made with the fgsea package (22) and the enrichmentbrowser package (23) in R using the pathway information from Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.genome.jp/kegg/pathway.html).

Samples (∼2–4 nmol of total lipid) underwent an acidic Bligh and Dyer liquid-liquid extraction (24). Lipid standards were added before extractions, using a master mix as described (25). Samples were analyzed on a QTRAP 6500+ mass spectrometer (Sciex) with chip-based (HD-D ESI Chip; Advion Biosciences) electrospray infusion and ionization via a Triversa Nanomate (Advion Biosciences), as previously described (26). Data evaluation was done using LipidView (1.3 β; Sciex) and in-house–developed software (ShinyLipids).

Results are expressed as the mean ± SD. Statistical analysis was performed with GraphPad Prism (version 6.0; GraphPad Software, San Diego, CA). Statistical analyses were performed using one-way ANOVA, where appropriate. P < 0.05 was considered statistically significant.

The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Lepr^db/db/KO were viable and had comparable body weights as Lepr^db/db/WT littermates (Supplementary Fig. 1). Apart from mildly elevated glucose levels at an early age, Lepr^db/db/KO did not develop hyperglycemia at age >16 weeks compared with Lepr^db/db/WT (Fig. 1A and B). Likewise, HbA_1c levels in Lepr^db/db/KO were equal to WT and KO (Fig. 1C). Insulin plasma levels were equally elevated in Lepr^db/db/KO as in Lepr^db/db/WT (Fig. 1D).

We next conducted glucose tolerance tests and insulin tolerance tests. Glucose tolerance was impaired in both Lepr^db/db/WT and Lepr^db/db/KO with prolonged hyperglycemia (Fig. 2A). Upon insulin injection, blood glucose levels of Lepr^db/db paradoxically increased, which we attribute to an endogenous stress response and/or dysregulated insulin sensing. Notably, this effect was reduced in Lepr^db/db/KO compared with Lepr^db/db/WT, indicating that CaMKIIδ mutation restores whole-body insulin responsiveness but not glucose tolerance in leptin-mutant mice (Fig. 2B).

In skeletal muscle—a major glucose store of the body—protein levels of the insulin-independent GLUT, GLUT1, and the insulin-dependent GLUT, GLUT4, (27) were increased in both Lepr^db/db/WT and Lepr^db/db/KO (Fig. 3A and B). However, GLUT1 and GLUT4 were also upregulated in nondiabetic KO mice, pointing to a general CaMKIIδ-dependent mode of GLUT expression. The additional increase in expression of the insulin-dependent transporter GLUT4 in skeletal muscle of Lepr^db/db/KO is suggestive to contribute to the prevention of hyperglycemia. In support, the increase in skeletal muscle glycogen content was relatively higher in Lepr^db/db/KO (Fig. 3C). The glucose intermediate sorbitol of the polyol pathway was not significantly enhanced in Lepr^db/db mice, but we observed a trend toward elevation in Lepr^db/db/KO (Supplementary Fig. 2).

Hepatic glycogen content was not significantly decreased (Supplementary Fig. 3A), while sorbitol levels were increased in both Lepr^db/db groups (Supplementary Fig. 3B), indicating that glucose enters the polyol pathway, which might result in higher oxidative stress and the formation of advanced glycation end products.

To investigate whether CaMKIIδ regulates HGP, we examined the levels of p-p38MAPK, which regulates the localization of FOXO1, the transcriptional regulator of HPG production (28), and found it was attenuated in Lepr^db/db/KO (Supplementary Fig. 4A and B). Moreover, gene expression of PEPCK (Pck-1) and glucose-6-phosphatase (G6pc)—two key enzymes of gluconeogenesis—are downregulated in CaMKIIδ-deficient db/db mice (Supplementary Fig. 4C and D), pointing to a contribution of CaMKIIδ to the activation of HGP. In the liver of Lepr^db/db/WT, we did not observe an upregulation of CaMKIIγ, which was implied to regulate HGP, pointing to a specific yet unknown effect of CaMKIIδ on HGP (Supplementary Fig. 4E).

Whereas mRNA levels of the sodium–glucose cotransporter 2 (SGLT2) were slightly but not significantly increased in kidneys of Lepr^db/db/KO, SGLT1 mRNA was increased in Lepr^db/db/WT but not in Lepr^db/db/KO (Supplementary Fig. 5A and B). But, because urinary glucose concentrations were decreased in Lepr^db/db/KO compared with Lepr^db/db/WT littermates (Supplementary Fig. 5C), it seems rather unlikely that a decrease in renal glucose excretion contributes to low blood glucose levels.

More studies are needed to determine the relative contributions of the observed changes to the here described phenoptye.

To our big surprise, we observed more pronounced mesangial expansion in Lepr^db/db/KO despite normoglycemia compared with hyperglycemic Lepr^db/db/WT kidneys (Fig. 4A). Likewise, thickening of the basal membrane (Fig. 4B) and loss of podocytes (Fig. 4C) were more pronounced. Furthermore, we observed fibrosis (Fig. 4D), tubular dilatation (Fig. 4E), and proteinuria (Fig. 4F) in Lepr^db/db/KO as in Lepr^db/db/WT. Kidney weight was slightly elevated in 16-week-old Lepr^db/db/KO, and glomerular size increased over time in both Lepr^db/db groups (Supplementary Fig. 6A–D). These data show that diabetic nephropathy in Lepr^db/db can even occur under normoglycemia. This is a central finding of this study, highlighting that the development of diabetic nephropathy is a complex process that is not determined by glucose levels in the blood.

Because CaMKIIα and CaMKIIβ were undetectable and CaMKIIγ was not increased in Lepr^db/db/KO, it is unlikely that other isoforms compensated to induce nephropathy (Supplementary Fig. 7A). Indeed, p-calcineurin, a marker for CaMKII activity (29), was increased in renal tissue of Lepr^db/db/WT but not of Lepr^db/db/KO (Fig. 4G), confirming that CaMKIIδ contributed entirely to elevated CaMKII activity in the diabetic kidneys.

GLUT1 expression was elevated in Lepr^db/db/WT and also in Lepr^+/+/KO (Fig. 5A), and GLUT4 was slightly but not significantly increased in both KO groups, suggestive for increased renal glucose uptake in KO. In Seahorse experiments, siRNA-mediated CaMKIIδ deletion in immortalized tubular cells (M1) and mesangial cells (MES13) did not reveal differences in the glycolysis stress tests, suggesting that substrate utilization for energy production in mesangial and tubular cells is not regulated by CaMKIIδ (Supplementary Fig. 8A–D). We therefore speculated that the higher glucose uptake in Lepr^db/db/KO does not contribute to energy production via glycolysis but “fuels” glycolysis side pathways, which may result in accumulation of reactive glucose metabolites. Thus, we measured total O-GlcNAcylation in renal tissue (30), and we found elevated protein O-GlcNAcylation in Lepr^db/db/WT and Lepr^db/db/KO and also in Lepr^+/+/KO (Fig. 5B), suggesting that protein O-GlcNAcylation is governed by CaMKIIδ. MG has been suggested as another diabetes-associated reactive glucose metabolite (31). We measured levels of MG in the kidney and found increased concentrations in both Lepr^db/db/WT and Lepr^db/db/KO mice compared with their respective nondiabetic littermates (Fig. 5C), suggesting that MG is CaMKIIδ-independently produced and may be additive to renal protein O-GlcNAcylation as a cause for diabetic nephropathy.

As markers for diabetic retinopathy we measured the number of acellular capillaries, pericytes, and migrating pericytes (32) (Fig. 6A). Lepr^db/db/WT showed a significant increase in the number of acellular capillaries (Fig. 6B), a loss of pericytes (Fig. 6C), and a higher number of migrating pericytes (Fig. 6D) compared with nondiabetic mice. Remarkably, Lepr^db/db/KO did not show pathological alterations in the aforementioned parameters, indicating that these mice are protected from diabetic retinopathy. Other than in the kidney, we found a general upregulation of all CaMKII genes in Lepr^db/db/WT but surprisingly not in Lepr^db/db/KO (Supplementary Fig. 7B), pointing to a yet unknown positive feedback loop of CaMKIIδ- or diabetes-induced CaMKIIα, CaMKIIβ, or CaMKIIγ gene expression. The profound changes in CaMKII gene expression prompted us to systematically analyze diabetes-related gene expression in different diabetic organs by RNA sequencing (GSE157739 provides access to all data). In analogy to the clear CaMKIIδ-dependent phenotypic data, we found a clear pattern of CaMKII-dependent differential gene expression in the retina compared with the kidney (Supplementary Fig. 9A and B). We were particularly interested in the genes that were only regulated in WT retina under diabetic conditions but not in KO (CaMKII-dependent specific diabetic retinopathy genes) (Fig. 6E and F) and in genes that were regulated in WT retina and kidney under diabetic conditions but not in KO retina (Fig. 6E and Supplementary Fig. 9C and D). A few of the top-regulated genes in the retina (Supplementary Fig. 9A and B) were retina specific (marked in red, Fig. 6C). The signature of retina-specific and CaMKIIδ-dependent diabetes-induced genes needs further functional analysis. Many genes (e.g., Gpat2) with “lipid” as a keyword in Gene Ontology were differentially regulated in the retina of Lepr^db/db/WT versus Lepr^+/+/WT but not Lepr^db/db/KO versus Lepr^+/+/KO mice (Supplementary Fig. 10A), whereas CaMKIIδ-dependent changes in the kidney were less clear (Supplementary Fig. 10B). In an attempt to validate the significance of retinal gene expression changes concerning lipid mediators, we failed to identify differences by lipidomics (Supplementary Fig. 10C). Specific diabetes-induced phosphatidylethanolamine (PE) and cholesterol reduction in the retina appeared to be CaMKIIδ dependent (Supplementary Fig. 10D–F), but the detectable quantitative differences questionably contribute to the striking phenotype rescue (Supplementary Fig. 10G).

In this study we show 1) that CaMKIIδ induces diabetic hyperglycemia possibly by reducing insulin sensing with decreased glucose uptake into stores and an increase in HGP, 2) that hyperglycemia per se does not lead to diabetic nephropathy, and 3) that hyperglycemia is a good measure to assess the risk of diabetic retinopathy (Fig. 7).

To date, little is known about the role of CaMKII in insulin signaling and the regulation of GLUTs. In a rat skeletal muscle cell line, CaMKII mediates phosphorylation of the insulin receptor substrate 1 (IRS1) at serine 612, and this in turn blocks downstream signaling of the insulin receptor (33). As a consequence, expression of GLUT4 is inhibited, indicating that CaMKII regulates GLUT4 expression via the insulin signaling pathway in their model system. Another report describes the regulation of GLUT4 through CaMKII on gene expression level via the transcription factor myocyte enhancer factor 2 (MEF2) (34), a prominent downstream target of CaMKII signaling (35).

Regarding the influence of CaMKIIδ on HGP, it has been shown that CaMKIIγ is a regulator of HGP (9–11,36). Compared with the aforementioned study, we observed a similar but more modest impact of CaMKIIδ on HGP. However, the seemingly lower impact of CaMKIIδ deficiency on baseline glucose production in the liver may be due to a compensatory activation of CaMKIIγ and suggests that CaMKIIγ is the more predominant isoform involved in HGP. However, CaMKIIγ is not upregulated in the liver in a CaMKIIδ-dependent manner as in the diabetic retina, suggesting that either CaMKIIδ contributes to HGP in the diabetic liver more than CaMKIIγ or that CaMKIIγ compensates at the functional levels as was described in pathological cardiac remodeling (29).

Another striking finding is that the CaMKIIδ-dependent reduction of blood glucose in Lepr^db/db/KO mice does not affect the integrity of all organ systems to the same extent (i.e., no effect on diabetic nephropathy but on retinopathy), indicating that diabetes-related defects occur in a tissue- and cell type–specific manner. This observation challenges the prevailing idea that reducing hyperglycemia is a unifying target to limit late complications in patients with diabetes. We propose that not the clinical presentation of hyperglycemia but rather specific not well-defined alterations of glucose handling and the production of glucose metabolites such as MG or O-GlcNAcylation contribute more directly to late diabetic complications.

Accordingly, the opposite phenotypes observed in the kidneys and the retinas of Lepr^db/db/KO mice can be explained by specific differences in glucose metabolism in these organs. A recent study described tissue-specific alterations in kidneys and retinas in diabetic versus nondiabetic mice using transcriptomics, metabolomics, and metabolic flux analyses (37). However, whereas in this study the activity of glucose and fatty acid metabolism was more enhanced in the kidney than the retina, we found more specific gene expression signatures in the retina, which depended largely on CaMKII. However, we closely checked the possible contribution of lipid metabolites but could not provide evidence for a significant contribution. While renal glucose metabolites such as MG and O-GlcNAcylation persisted in CaMKIIδ-deficient mice, the retina showed, for example, a significant attenuation of the rate-limiting enzyme of the hexosamine biosynthesis pathway in a CaMKII-dependent manner.

Based on the collected data, we cannot distinguish whether the observed retinoprotection depends directly on normoglycemia or retinal CaMKIIδ itself. The latter is supported by an in vitro study with primary cultures of retinal Müller cells prepared from Sprague-Dawley rats. In this model system, high glucose induced expression of hypoxia-inducible factor-α and vascular endothelial growth factor in a CaMKII-CREB–dependent manner. Activation of this pathway was suggested to be the underlying pathogenesis of retinopathy (38,39). The newly obtained list of diabetes-induced and CaMKII-dependent retinal genes will be helpful to further study this issue. Conditional deletion of CaMKIIδ in retina versus other glucose storage tissues is also warranted. Our quantitative PCR data demonstrating relatively high expression of all four CaMKII isoforms in retinal tissue in a CaMKIIδ-dependent manner also point to a specific role of CaMKII in the retina, although the induction of the α, β, and γ CaMKII isoforms can be secondarily triggered by hyperglycemia. Further promoter analysis of these CaMKII genes may lead to a better understanding of this interesting finding.

Putting our observation in a clinical context, we can conclude that monitoring glucose and HbA_1c levels is not sufficient to predict late complications in patients with diabetes, in particular diabetic nephropathy. This is supported by a 10-year follow-up study of patients with T2D receiving conventional therapy with dietary restriction or an intensive therapy with sulfonylurea, insulin, or metformin for glucose control (40). The investigators reported that levels of HbA_1c in both conventionally and intensively treated patients were reduced to the same degree after a 1-year therapy and that the reduction in vascular risk and death from any cause was continued. However, patients receiving an intensified therapy show fewer late complications compared with patients with simple dietary restriction despite equally adjusted blood glucose. Moreover, it was suggested that normalizing blood glucose is not sufficient to prevent diabetic late complications until the acquired metabolic alterations are restored (41). However, to date, therapeutic approaches that target reactive metabolites are not shown to combat neither diabetic nephropathy nor retinopathy (42,43). Based on these observations, recent therapeutic approaches focus on the identification of protective factors that potentially neutralize hyperglycemia-associated damage rather than assessing risk factors. Individuals monitored in the Joslin Medalist Study with long-term type 1 diabetes (>50 years) without presenting the typical late complications and their glycemic control did not correlate with damage neither in kidneys nor in retinas (44,45). Only 12% of the observed patients with diabetes had diabetic nephropathy. Individuals with or without diabetic nephropathy from the same study underwent proteomic analysis of glomeruli isolated postmortem to identify potential protective targets (46). They observed a strongly clustered proteomic signature of enzymes related to glycolysis, the sorbitol pathway, MG production, and mitochondrial function, suggesting that reduction of free glucose and its reactive metabolites can preserve the kidney function.

Thus, there is an urgent need for innovative treatment strategies to combat diabetic late complications. Our present work suggests that adjusting blood glucose levels in clinical management alone is not sufficient to prevent organ damage in patients with diabetes. It is rather necessary to gain a mechanistic understanding leading to failure of individual organ systems in patients with diabetes.

Acknowledgments. The authors thank Jutta Krebs-Haupenthal, Sabine Kuss, Michaela Oestringer, and Silvia Harrack (Institute of Experimental Cardiology, Heidelberg University, Heidelberg, Germany) for technical help. Furthermore, the authors would like to thank the Center for Model System and Comparative Pathology (Institute of Pathology, University Hospital Heidelberg, Heidelberg, Germany) technical support. The authors gratefully acknowledge the data storage service SDS@hd supported by the Ministry of Science, Research and the Arts Baden-Württemberg.

Funding. J.C., T.F., J.M., H.-J.G., H.-P.H., P.P.N., M.H., and J.B. were supported by Deutsche Forschungsgemeinschaft, SFB 1118, project number 236360313. J.C. received the Otto-Hess-Scholarship from the German Society for Cardiology. H.A.K. and J.B. were supported by the Deutsches Zentrum für Herz-Kreislauf-Forschung–German Centre for Cardiovascular Research and by the Bundesministerium für Bildung und Forschung (German Ministry of Education and Research). This work was also supported by the German Research Foundation (grants INST 35/1314-1 FUGG and INST 35/1503-1 FUGG).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.C., T.F., S.K., M.D., K.H., A.S., M.K., M.P.W., B.P., F.S., J.L., J.M., and A.E. performed experiments. J.C., D.K., P.S., H.-J.G., H.-P.H., C.S., B.B., M.H., and J.B. analyzed and interpreted the data. J.C., M.H., and J.B. designed the study. J.C., M.H., and J.B. wrote the paper. H.-J.G., H.-P.H., P.P.N., B.I., B.B., and H.A.K. provided research support and conceptual advice. J.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 84th Annual Conference of the Deutsche Gesellschaft für Kardiologie (DGK), Mannheim, Germany, 4–7 April 2018.