Characterization of GlycolyticEnzymes and Pyruvate Kinase M2in Type 1 and 2 DiabeticNephropathyDiabetes Care 2019;42:1263–1273 | https://doi.org/10.2337/dc18-2585

OBJECTIVE

Elevated glycolytic enzymes in renal glomeruli correlatedwith preservation of renalfunction in the Medalist Study, individuals with ‡50 years of type 1 diabetes.Specifically, pyruvate kinase M2 (PKM2) activation protected insulin-deficientdiabetic mice from hyperglycemia-induced glomerular pathology. This study aimsto extend these findings in a separate cohort of individuals with type 1 and type 2diabetes and discover new circulatory biomarkers for renal protection throughproteomics and metabolomics of Medalists’ plasma. We hypothesize that increasedglycolytic flux and improved mitochondrial biogenesis will halt the progression ofdiabetic nephropathy.

RESEARCH DESIGN AND METHODS

Immunoblots analyzed selected glycolytic and mitochondrial enzymes in post-mortemglomeruli ofnon-Medalistswith type1diabetes (n=15), type2diabetes (n=19), and no diabetes (n = 5). Plasma proteomic (SOMAscan) (n = 180) andmetabolomic screens (n = 214) of Medalists with and without stage 3b chronickidney disease (CKD) were conducted and significant markers validated by ELISA.

RESULTS

Glycolytic (PKM1, PKM2, and ENO1) and mitochondrial (MTCO2) enzymes weresignificantly elevated in glomeruli of CKD2 versus CKD+ individuals with type 2diabetes. Medalists’ plasma PKM2 correlated with estimated glomerular filtrationrate (r2 = 0.077; P = 0.0002). Several glucose and mitochondrial enzymes incirculation were upregulated with corresponding downregulation of toxic metab-olites in CKD-protectedMedalists. Amyloid precursor protein was also significantlyupregulated, tumor necrosis factor receptors downregulated, and both confirmedby ELISA.

CONCLUSIONS

Elevationof enzymes involved in themetabolismof intracellular free glucose and itsmetabolites in renal glomeruli is connected to preserving kidney function in bothtype 1 and type 2 diabetes. The renal profile of elevated glycolytic enzymes andreducedtoxic glucosemetabolites is reflected in thecirculation, supporting theiruseas biomarkers for endogenous renal protective factors in people with diabetes.

1Joslin Diabetes Center, Boston, MA2Harvard Medical School, Boston, MA3Folkhalsan Research Center, University of Hel-sinki, Helsinki, Finland4Abdominal Center Nephrology, Helsinki Univer-sity Hospital, Helsinki, Finland5Translational Research and Early Clinical De-velopment, Cardiovascular and Metabolic Re-search, AstraZeneca, Molndal, Sweden6Sanofi Deutschland GmbH, Frankfurt am Main,Germany7Medical Faculty Mannheim, University of Hei-delberg, Mannheim, Germany8Sanofi-Genzyme, Cambridge, MA

Corresponding author: George L. King, george.king@joslin.harvard.edu

Received 18 December 2018 and accepted 11April 2019

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc18-2585/-/DC1.

D.G. and H.S. are co–first authors.

© 2019 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

Daniel Gordin,1,2,3,4 Hetal Shah,1,2

Takanori Shinjo,1,2 Ronald St-Louis,1,2

Weier Qi,1,2,5 Kyoungmin Park,1,2

Samantha M. Paniagua,1

David M. Pober,1,2 I-Hsien Wu,1

Vanessa Bahnam,1 Megan J. Brissett,1

Liane J. Tinsley,1 Jonathan M. Dreyfuss,1

Hui Pan,1 Yutong Dong,1

Monika A. Niewczas,1,2 Peter Amenta,1,2

Thorsten Sadowski,6 Aimo Kannt,6,7

Hillary A. Keenan,1,2,8 and George L. King1,2

Diabetes Care Volume 42, July 2019 1263

PATH

OPHYSIO

LOGY/CO

MPLIC

ATIO

NS

Chronic kidney disease (CKD) is a majorcause of mortality and morbidity in peo-ple with diabetes (1,2). Multiple mech-anisms have been proposed to mediatehyperglycemic adverse effects, includingincreased production of sorbitol, meth-ylglyoxal, diacylglycerol, reactive oxygenspecies, and advanced glycation endproducts (3–7). Although experimentaldata support the role of these pathwaysin CKD pathogenesis, clinical trials usingspecific inhibitors of these pathwaysyielded only modest results (8–10). Find-ings from the Joslin Medalist Study on1,008 individuals with $50 years oftype 1 diabetes suggest inherent pro-tective factors against the developmentof CKD and other diabetes-related com-plications (11,12). We recently identifieda cluster of enzymes of glucose metab-olism, including pyruvate kinase M2(PKM2), which may protect against hy-perglycemia-induced CKD (13). PKM2, anenzyme active at the juncture of glycol-ysis and the Krebs cycle, was upregulatedin glomeruli of CKD-protected Medalistsand strongly correlated with renal func-tion (estimated glomerular filtration rate[eGFR]). Furthermore, we observed sig-nificant upregulation of several enzymesand reduction of their correspondingmetabolites in the glycolytic, aldose re-ductase (AR), glyoxalase, and mitochon-drial pathways among the CKD-protectedMedalists (13). We demonstrated thatan increased glucose metabolic flux couldneutralize or even lower levels of hyper-glycemia-induced toxic metabolites. More-over, PKM2-selective activator TEPP-46reversed hyperglycemia-induced meta-bolic abnormalities, mitochondrial dys-function, and renal glomerular pathologyin type 1 diabetic mouse models (13).The hypothesis of this study was that

the upregulation of PKM2 and enzymesof glucose metabolism and tricarboxylicacid (TCA) cycle protecting from diabetickidney disease as found in individualswith extreme duration of type 1 diabetes(Medalists) could be replicated and ex-tended in two tissues, glomeruli andplasma, of people with type 1 diabetesof shorter duration and type 2 diabetes.Hence, we aimed to extend our previousfindings on the elevations of glomerularPKM2 and other glycolytic enzymes inboth type 1 and type 2 diabetes. Addi-tionally, we sought to replicate the glo-merular metabolomic and proteomicfindings from CKD-protected Medalists

by conducting similar analyses of plasmain a larger sample of Medalists and dis-cover new markers of renal protectionthrough these unbiased omic screens.

RESEARCH DESIGN AND METHODS

The various subsets used in the variousaspects of this investigation are outlinedin Supplementary Fig. 1.

The MedalistsThe 50-year Medalist Study recruited par-ticipants across the U.S. (n = 1,008) withwell-documented $50 years of type 1diabetes. Detailed clinical descriptions ofthe Medalist Study, including HLA geno-typing and autoantibody titers, have beenpublished previously (12–14). The JoslinCommittee on Human Subjects approvedthe study protocol. Each participant un-derwent written informed consent, med-ical history questionnaires, and physicalexaminations at the Joslin Diabetes Center.

The Medalist study design and bio-specimen collection have been previ-ously described (12,13). For the plasmaomic studies, CKD was defined by ChronicKidney Disease Epidemiology Collabo-ration (CKD-EPI) eGFR ,45 mL/min/1.73 m2 (stage 3b CKD). For immuno-blotting experiments, “low eGFR cate-gory” was designated as ,30 mL/min/1.73 m2 to evaluate a sufficient numberof samples per group, as it was challeng-ing to procure postmortem kidneysamong those with preserved renal func-tion because they are prioritized fororgan transplantation.

Postmortem Glomeruli andImmunoblotsPostmortem glomeruli from whole kidneysof non-Medalist individuals with type 2diabetes (n = 19), type 1 diabetes (n = 15),and no diabetes (control subjects) (n = 5)were retrieved after death to study en-zymes of glucose metabolism and TCAcycle via immunoblotting (SupplementaryTable 2).

Collection of kidneys was approved byJoslin’s Institutional Review Board andcoordinated by the National Disease Re-search Interchange or the InternationalInstitute for the Advancement of Med-icine; both use a human tissue collectionprotocol approved by a managerial com-mittee and subject to oversight by theNational Institutes of Health. Kidneyswere shipped on ice and saline gauzewithin 10 h of death.

Each kidney was decapsulated andbisected axially, and the cortex collected,minced on ice, and passed through twodifferent-sized (253 and 89 mm) sieveswith cold PBS. Glomeruli were extractedusing the 89-mm sieve, washedwith PBS,and pelleted by centrifugation. The tissuefor light microscopy was formalin fixedand paraffin embedded. Sections werecut (2 mm), stained (hematoxylin andeosin, periodic acid Schiff, and Massontrichrome), and graded according to theRenal Pathology Society classification ofCKD by pathologists blinded to clinicaldata (13).

Bio-Rad Mini-PROTEAN TGX (Bio-RadLaboratories, Hercules, CA) precast gelswere used for all Western blots (anti-bodies detailed in Supplementary Table1) and ImageJ software applied to quan-tify proteins.

ProteomicsMedalists (N=180)with andwithout CKDunderwent plasma proteomic analysistargeting 1,129 proteins by SOMAscan(SomaLogic, Boulder, CO), a multiplexedDNA aptamer-based assay, using specificaffinity-binding reagents (SOMAmers)quantified on a custom Agilent hybrid-ization chip (15). Aptamers were immo-bilized to streptavidin-coated beads andincubated with samples to assay targetsin a multiplexed manner.

MetabolomicsMedalists (N=214)with andwithout CKDunderwent plasma metabolomics by liq-uid chromatography mass spectrometry(Broad Institute, Cambridge, MA) (16,17),targeting 58 metabolites (sugars, lipids,fatty acids, and amino acids) from keypathways relevant to our recent study (13).

Both the metabolomic and proteomicsubsets were chosen to include individ-uals with extreme phenotypes, includingprogressors (eGFR ,45) with low HbA1cand resistors (eGFR .45) with highHbA1c. Sample availability in storageand cost considerations resulted in theslightly lowernumber sent for SomaLogicproteomic studies.

Replication and Validation ofProteomic FindingsTumor necrosis factor receptor super-family members 1A and 1B (TNFRSF1Aand TNFRSF1B), well-known markers ofCKD (18), were significantly downregu-lated among CKD-protected Medalists in

1264 PKM2 and New Markers of Renal Protection in T1D Diabetes Care Volume 42, July 2019

our proteomic analysis. We validatedthese SOMAscan proteins by ELISA ina subset of Medalists (N = 30) to dem-onstrate the robustness of our findings.Novel marker amyloid precursor protein(APP), discovered from our interactomeanalysis, was further validated and rep-licated in Medalist samples (N = 159)through ELISA as described below.

Plasma Measurement of TNF-sRI and -sRII

Soluble TNFRSF1A and TNFRSF1B wereimmunoassayed inplasmabyELISA (EMDMillipore, Billerica, MA) according to themanufacturer’s protocols (catalog num-ber DRT100, DRT200; R&D Systems, Min-neapolis, MN). All measurements wereperformed in duplicate. Intra-assay co-efficient of variation was ,5% for both.Assay protocols have been previouslydescribed (18).

Urinary Assessment of KIM-1

KIM-1, a marker of tubular injury, wasassessed in urine samples of the samesubset of Medalists (N = 30) describedabove. Measurements were conductedon the Luminex platformwith amagneticbead-enhanced, sandwich-type immu-noassay with Human Kidney ToxicityPanel 3 (catalog number HKTX3MAG-38 K; EMD Millipore). All measurementswere performed in duplicate. Intra-assaycoefficient of variation was ,5%. Assayprotocols have been previously de-scribed (18).

Assessment of Plasma APP

Plasma APP was measured by humanELISA Kit (catalog number DY850; R&DSystems) according to protocol. Briefly,an ELISA 96-well plate was coated with a100 mL of polyclonal APP capture anti-body in filtered PBS (4 mg/mL/well) andincubated overnight at room tempera-ture (RT). The plate was washed with400 mL/well of washing buffer (0.05%Tween-20) and blockedwith 300mL/wellof dilution buffer (1% BSA in PBS) for30 min at RT. Duplicates of 100 mL ofserially diluted recombinant human APPpeptides from 50 ng to 0.2 ng/mL as astandard, plasma samples from Medal-ists, IgG (0.1 mg/mL), and albumin(0.1 mg/mL) as negative controls wereall loaded on the same plate. After a 2-hincubation at RT, the plate was washedand incubated with 100 mL of biotiny-lated APP detection antibody (1 mg/mL/well) for 2 h at RT, then washed, in-cubated with 100mL of streptavidin-HRP(1:200; R&D Systems) for 30 min at RT,

and finally incubated with 100 mL/well oftetramethylbenzidine for 10–20 min atRT. The reaction was stopped with 50 mLof 1 mol/L H2SO4, and absorbance wasread at 450 or 560 nm using the PlateReader (Promega, Madison, WI). Assaysensitivity and specificity were validatedusing human IgG, BSA, and mouse/ratplasma (Supplementary Fig. 2).

Statistical AnalysisDistributions of variables were examinedto determine appropriate statisticalmethods. Differences between groupswere analyzed with Student t, Mann-Whitney U, Kruskal-Wallis, or ANOVAtests, as appropriate. Proportions withinand between groups were comparedusing x2 tests. For continuous outcomes,linear regression models were run withappropriate adjustments. Analyses wereperformed using SAS v9.4 (SAS Institute,Cary, NC) with significance threshold Pvalue ,0.05 for assessing group differ-ences in baseline characteristics. Forassessing group differences in the im-munoblotting experiments, while a sig-nificance threshold of P , 0.05 was set,we also applied a suggestive significancethreshold of P, 0.10 due to the smallernumber of samples and challenge ofprocuring postmortem kidney specimens.

Proteomics and Metabolomics

Bioinformatics

The SOMAscan data were log2-trans-formed to normalize distributions. Formetabolomics, missing data were im-puted using half of the minimum acrossall samples for that metabolite. Metab-olites not present in at least 20% of thesamples were filtered out, and data werelog2-transformed. Differential abundan-ces of proteins among CKD-negativeversus CKD-positive individuals werecalculated using the linear modelingR-package limma (19). Multiple testingcorrections were applied using Benjaminand Hochberg false discovery rate(FDR) ,0.05 for significance (20). Vol-cano plots were plotted by R-packageggplot2. Pathway analysis was done vialimma ROAST method (21).

The integrated SOMAscan and metab-olite interaction network analysis in-cluded proteins related to PKM2 andanalytes with between-group FDR #5 31025. Isolated nodes in the networkwereremoved. The global network was de-rived from Pathway Commons and plot-ted by R-package igraph (22,23).

Out of the 14 enzymes of glucose andTCA pathways earlier identified by massspectrometry in glomeruli of CKD-protectedMedalists (13), six were detectedby the SOMAscan, letting us validate andreplicate our previous findings, now bymeans of different individuals, differenttissues (plasma), and different methods(SOMAscan). Simultaneously, the unbi-ased proteomic screen allowed for un-covering new markers of renal protectionin type 1 diabetes.

RESULTS

Key Enzymes of Glucose Metabolismin Postmortem Renal GlomeruliSelected enzymes of glucosemetabolism(glycolytic, AR, and glyoxalase pathways)and TCA cycle were studied in postmor-tem glomeruli of individuals with type 2diabetes (n = 19), type 1 diabetes (n = 15),and no diabetes (control) (n = 5) viaimmunoblotting (Supplementary Table2). Individuals with type 2 diabeteshad shorter diabetes duration comparedwith those with type 1 (P , 0.001).Average eGFR was 40.0, 24.6, and69.6 mL/min/1.73 m2 among thosewith type 2 diabetes, type 1 diabetes,and no diabetes, respectively (P = 0.23).No significant differences were found forsex, age, orBMI (Supplementary Table2).Glomerular enzyme levels were com-pared in individuals with high eGFR($30 mL/min/1.73 m2) versus low(,30 mL/min/1.73 m2). In type 2 andtype 1 diabetes, respectively, median(interquartile range) eGFRs were 57.5(42.7–77.6) mL/min/1.73 m2 and 76.2(48.7–91.6) mL/min/1.73 m2 in the highgroups and 14.9 (8.8–28.1) and 13.8 (11.9–19.6) mL/min/1.73 m2 in the low groups.

Histological CKD classifications includedno-to-mild [0–IIA] versus moderate-to-severe [IIB–IV] CKD (SupplementaryTable 2). All control subjects withoutdiabetes were scored as no-to-mildCKD in their renal specimens. In type2 diabetes, 14 individuals had no-to-mildand 5 had moderate-to-severe CKD. Intype 1 diabetes, six had no-to-mild andeight hadmoderate-to-severe CKD. A goodstructural–functional association betweeneGFR and histological CKD grading wasobserved, as shown by median (interquar-tile range) eGFR of 57.0 (28.1–76.2) mL/min/1.73 m2 in the no-to-mild group and13.2 (9.3–32.2) mL/min/1.73 m2 in themoderate-to-severe group (P , 0.01).

care.diabetesjournals.org Gordin and Associates 1265

PKM2 expression was lower in indi-viduals with type 2 diabetes versus con-trol subjects, with a trend towardsignificance (P = 0.08) (Fig. 1A). Com-pared with control subjects, other en-zymes in glycolytic (PKM1, TPI1, andENO1), TCA (MTCO2), AR, and glyoxa-lase pathways were decreased in bothtype2and type1diabetes (P,0.05) (Fig.1B–H).In type 2 diabetes, PKM2was trending

toward significantly lower levels in thosewith severe CKD versus those with per-sisting renal function (P = 0.07) (Fig. 1A).Similarly, levels of PKM1 and ENO1 werealso significantly decreased in those withsevere CKD (P , 0.05 for PKM1 andENO1), whereas GLO1 reached sugges-tive significance (P = 0.07) (Fig. 1B–H).Expressions of TPI1 and GLO1 were

higher in type 2 individuals with histolog-ically graded no-to-mild CKD comparedwith those with moderate-to-severe CKD(P = 0.06 and P = 0.03, respectively). Noother differences were observed amongthe histological classes.

Plasma Proteomics and MetabolomicsAmong CKD-Protected Individuals

Proteomics (SOMAscan)

Compared with those without proteo-mics data, Medalists in the SOMAscanset were older and had longer durationof diabetes; higher BMI, HbA1c, triglycer-ides, and albumin-to-creatinine ratio(ACR); lower HDL and eGFR; and greaterprevalence of cardiovascular disease(CVD), proliferative diabetic retinopathy(PDR), and detectable C-peptide levels,with lower exercise rates (SupplementaryTable 3).Compared with the CKD-negative

group, CKD-positive individuals includedin the SOMAscan had higher rates ofCVD, PDR, and antihypertensive treat-ment and had lower exercise rates,HbA1c, HDL, and detectable C-peptidelevels (Table 1). By definition, eGFRwas lower (P , 0.001) and ACR higher(P , 0.001) in the CKD-positive group.Of 1,129 investigated plasma proteins,

162 were significantly downregulatedand 140 upregulated in the CKD-protected Medalists compared with thosewith CKD (FDR ,0.05) (Fig. 2A andSupplementary Tables 5 and 6). PlasmaSOMAscan detected 6 of the 14 differ-entially regulated enzymes of the glu-cose and TCA pathways observed in ourprevious glomerular mass-spectrometry

studies (13). Confirming the results of theglomerular mass spectrometry and im-munoblot analysis described above in aseparate group of individuals with type 1and type 2 diabetes, PKM2, TPI1, LDHB,and PGM1 were significantly upregulatedin CKD-protected Medalists (FDR,0.05)(Supplementary Table 4), while GAPDHand cytochrome c showed a similar trendbut did not reach statistical significance.As expected, many downregulated pro-teins in the unbiased screen were knownmarkers of renal damage including cys-tatin C, b2-microglobulin, neutrophil ge-latinase-associated lipocalin (NGAL),and TNFRs (FDR ,0.05) (Supplemen-tary Table 6).

Plasma PKM2 in Type 1 Diabetes and CKD.

Previously, a 2.7-fold upregulation ofPKM2 in glomeruli of CKD-protectedMedalists was observed (13). Aforemen-tioned, we confirmed the associationbetween PKM2 and CKD protection bytwo proteomic methods in both hu-man renal tissue (immunoblotting) andplasma (SOMAscan; FDR ,0.05) intype 1 and type 2 diabetes. Additionally,plasma PKM2 correlated with ACR (r2 =0.031; P = 0.03) and eGFR (r2 = 0.077; P =0.0002) at baseline. Adjustments by co-variates (BMI, daily insulin dose per kg,HbA1c, HDL, triglycerides, ACR, antihy-pertensivemedication,neuropathy,CVD,PDR, and detectable C-peptide levels) ina multivariate regression analysis did notattenuate the PKM2-eGFR association (b[95% CI] 4.97 [1.57–8.36]; P = 0.01), butACR no longer correlated significantly(data not shown). Among the CKD-neg-ative and -positive groups, PKM2 andthe other glycolytic enzymes did notcorrelate with ACR (SupplementaryTable 7).

Pathway analysis indicated that fourout of eight upregulated pathways (FDR,0.05) in the CKD-protected Medalistswere involved in processing free intra-cellular glucose (pentose phosphate, gly-colysis, gluconeogenesis, and pyruvate)(Fig. 2B). Axon guidance, notch signaling,and cytokine–cytokine receptor interac-tion pathways were downregulated.

Metabolomics

Althoughmost clinical characteristics didnot differ from the larger Medalist co-hort, individuals (n = 214) included inmetabolomic analysis had a higher pro-portion of males, lower HDL, diastolicblood pressure, and eGFR levels, and

higher C-peptide levels than those ex-cluded (Supplementary Table 8).

Within the metabolomics study, thosewith CKD were slightly younger, hadlower HbA1c and HDL levels, and hadhigher BMI, triglycerides, and frequencyof antihypertensive treatment comparedwith those without CKD (SupplementaryTable 9). Again, by definition, the CKD-positive group had lower eGFR (P ,0.001), higher ACR (P , 0.001), andhigher prevalence of CVD and PDR.

Thirty-two metabolites were downre-gulated, whereas two were upregulatedin the CKD-protected Medalists versusthose with CKD (FDR ,0.05) (Fig. 2C).Further, pyruvate, a PKM2 metabolite,was higher in Medalists without CKD(fold change 1.29; P = 0.0358). Metab-olites from the glycolytic (glucoronateand inositol), AR (sorbitol), and TCA(fumarate/maleate and aconitate) path-ways were downregulated (FDR ,0.05)in the protected individuals, validatingfindings from our previous study, butnow in a larger cohort (n = 214 vs. 29)(Fig. 2D) (13). Metabolites from addi-tional pathways, including tryptophan(xanthurenate and quinolinate), glucose(sebacate, lactose, and sucrose), fattyacid (suberate and hippurate), and pu-rine (hypoxhantine and urate), weredownregulated (FDR ,0.05) in CKD-protected individuals, whereas metabo-lites of energy-generating pathways,including pyruvate, ADP, GDP, anduridine, were upregulated (FDR ,0.05)(Supplementary Table 10).

As mentioned above, three (sorbitol,aconitate, and fumarate/maleate) of ninemetabolites included in the metabolomicplatform replicated from our previousstudy (13). The remaining six did not reachsignificance but had similar trends as ourprevious study: a-ketoglutarate, lactate,and phosphoenolpyruvate were upregu-lated in CKD-protected Medalists, whilecitrate, succinate, and fructose/glucose/galactose were downregulated.

In the CKD-protected Medalists, fourof the significantly (FDR ,0.05) down-regulated pathways were involved in theprocessing of glucose (tryptophan, ino-sitol, galactose, and tyrosine), whereasupregulated pathways were alanine, am-monia, cysteine, andpropanoate (Fig. 2D).

Proteomic-Metabolomic Interactome

In a network analysis for significant pro-teins and metabolites, glucose and TCA

1266 PKM2 and New Markers of Renal Protection in T1D Diabetes Care Volume 42, July 2019

pathways were activated in CKD-protected Medalists (Fig. 2E), as wellas AR and pentose phosphate pathways.Three nodes that connected to several

proteins and metabolites were upregu-lated in CKD-negative versus CKD-positive Medalists. These included APP(fold change 1.4; P = 3.81 3 1026),

epidermal growth factor receptor (EGFR)(fold change 1.15; P = 1.38 3 1026), andsmall glutamine-rich tetratricopeptide-a(SGTA) (fold change1.69;P=1.3931026).

Figure 1—Western blots of glomeruli of individuals with andwithout diabetes. PKM2 (A), PKM1 (B), GAPDH (C), TPI1 (D), ENO1 (E),MTCO2 (F), AR (G), andGLO1 (H) are shown. Data are mean6 SD. Differences by Student t test. Human glomeruli samples were prepared and Western blotting was performed aspreviouslydescribed(13).Atotalof30mgproteinfromthelysatewasresolvedby10%SDS-PAGEusingBio-RadMini-PROTEANTGX(Bio-RadLaboratories)precastgels. Control, participantwithout diabetes; high, eGFR$30mL/min/1.73m2; low, eGFR,30mL/min/1.73m2; T1, type 1 diabetes; T2, type 2 diabetes.

care.diabetesjournals.org Gordin and Associates 1267

Replication and Validation ofProteomic Findings

TNFRSF1A and TNFRSF1B

There were no significant differences inbaseline characteristics between SOMA-scan Medalists (N = 180) and Medalists(N = 30) whose samples were assayed forTNFRSF1A and 1B (Supplementary Table11). Among the 12 overlapping samples,there was significant correlation betweenthe SOMAscan and ELISA measurements[TNFRSF1A: R2 = 0.77, P = 1.63 1024; andTNFRSF1B: R2 = 0.88, P = 7.83 1026] (Fig.3A and B). ELISA confirmed significantlyhigher levels of these markers among CKD-positive (eGFR ,45 mL/min/1.73 m2)compared with CKD-negative individuals(TNFRSF1A: P = 0.0003; and TNFRSF1B:P = 0.003) (Fig. 3C and D). These TNFRsinversely correlated with PKM2, butnot with the other glycolytic enzymes(Supplementary Table 12).

Markers of Tubular Damage

We assessed tubular injury marker KIM-1in urine samples of the same subsetof Medalists (N = 30) described above.

KIM-1 did not correlate with eGFR (r =0.17; P = 0.44).

APP

Because plasma APP was a major centralnode in our interactome analysis andsignificantly elevated in CKD-protectedMedalists, we validated the finding viaELISA. Out of 159 samples assayed,40 were nonoverlapping with SOMAscansamples. There were no significant dif-ferences in baseline characteristics be-tween SOMAscan Medalists (N = 180)and the pure replication set (N = 40)(Supplementary Table 13). Among the119 overlapping samples (validation set),there was a significant, albeit weak,correlation between the APP measure-ments by SOMAscan and ELISA (R2 = 0.03;P = 0.05) (Fig. 3E). In the replication set,APP ELISA levels were significantly higheramong CKD-negative (eGFR .45 mL/min/1.73 m2) versus CKD-positive indi-viduals (P = 0.01) (Fig. 3F). The validationset revealed similarfindings (Supplemen-tary Fig. 3). At baseline, APP was signif-icantly associated with BMI, eGFR, CVD,

PDR, and CKD (Supplementary Table 14).When controlling for possible confound-ers, PDR and CKD remained stronglyassociated with APP (SupplementaryTable 15). Even after excluding individ-uals with PDR from the analysis, APPremained significantly elevated amongCKD-negative versus CKD-positive indi-viduals (estimated b = 3.12; P = 0.006).APP had a nominal inverse correlationwith TNFRs (Supplementary Fig. 4). Amongenzymes of glycolytic and TCA path-ways, APP correlated with TPI1, LDHB,and MDH1 (Supplementary Table 16).

APP and Thrombotic Proteins

As APP has been reported to have an-tithrombotic activities, we examined itsrelationship to prothrombotic proteins inthe SOMAscan. Three proteins signifi-cantly correlated inversely with APP, in-cluding thrombospondin 2 (THBS2),THBS4, and tissue factor, whereas com-plement C5 and platelet glycoprotein VIwere positively correlated with APP(Supplementary Table 16). Of these,THBS2, THBS4, and tissue factor weresignificantly downregulated among CKD-protectedMedalists (P, 83 1027), andglycoprotein VI was significantly upregu-lated (P = 3 3 1025) (SupplementaryTable 17). THBS4, prominently con-nected to APP in the network analysis(Fig. 2E), was also significantly down-regulated among Medalists withoutPDR compared with those with PDR (foldchange 21.15; P = 0.0019) (Supplemen-tary Table 17).

CONCLUSIONS

Our current and previous reportsstrongly suggest that elevated expres-sions of multiple enzymatic pathways ofintracellular glucose metabolism are as-sociated with CKD protection in individ-uals with long-standing type 1 diabetes(13). Similar to those reported in type 1diabetes, four glycolytic enzymes (PKM1,PKM2, ENO1, and TPI1) were also upre-gulated in the renal glomeruli of individ-uals with type 2 diabetes with preservedrenal function. Elevated mitochondrialenzyme (MTCO2) in individuals withtype 2 diabetes without CKD suggeststhat preservation of mitochondrial func-tion is important to retain kidney functionand prevent glomerular pathology (24).This study also showed that expressionsof both AR and glyoxalase were higherin people with type 2 diabetes with

Table 1—Baseline clinical characteristics of participants in the SOMAscan

Characteristic No CKD (N = 96) CKD (N = 84) P value

Male, N (%) 51 (52.58) 38 (42.70) 0.29

Duration (years) 54.66 (5.43) 56.26 (6.92) 0.10

Age (years) 66.2 (7.49) 67.35 (7.61) 0.34

Insulin dose (units/kg) 0.46 (0.14) 0.46 (0.19) 0.90

BMI (kg/m2) 26.29 (4.16) 28.22 (6.06) 0.02

HbA1c (%) 7.63 (0.69) 7.29 (1.02) 0.01

HbA1c (mmol/mol) 59.92 (7.57) 56.15 (11.10) 0.01

LDLc (mg/dL) 84.71 (23.99) 80.5 (27.05) 0.32

LDLc (mmol/L) 2.19 (0.62) 2.1 (0.7) 0.32

Total cholesterol (mg/dL) 164.46 (32.87) 158.38 (37.16) 0.28

Total cholesterol (mmol/L) 4.26 (0.85) 4.1 (0.96) 0.28

Triglycerides (mg/dL) 68.82 (29.15) 98.52 (52.47) ,0.001

Triglycerides (mmol/L) 0.78 (0.33) 1.11 (0.59) ,0.001

HDLc (mg/dL) 66.25 (21.14) 57.89 (21.29) 0.01

HDLc (mmol/L) 1.72 (0.55) 1.5 (0.55) 0.01

eGFR (mL/min/1.73 m2) 76.63 (15.97) 35.84 (7.88) ,0.001

ACR (mg/mg) 39.23 (103.90) 208.79 (650.71) ,0.001

ACR (mg/mmol) 4.44 (11.75) 23.62 (73.61) ,0.001

Systolic blood pressure (mmHg) 132.2 (16.95) 132.44 (18.08) 0.93

Diastolic blood pressure (mmHg) 62.87 (7.58) 63.7 (9.34) 0.53

Lipid-lowering medication, N (%) 68 (72.32) 60 (74.07) 0.87

Hypertensive medication, N (%) 56 (59.57) 66 (81.48) 0.002

Detectable C-peptide, N (%) 49 (50.52) 27 (33.33) 0.01

CVD, N (%) 37 (39.78) 50 (60.98) 0.03

PDR, N (%) 41 (52.56) 47 (73.44) 0.01

Neuropathy, N (%) 73 (80.22) 54 (68.35) 0.08

Data aremean (SD) for continuous variables unless otherwise noted. CKD is stage 3b, as defined byCKD-EPI eGFR ,45 mL/min/1.73 m2. HDLc, HDL cholesterol; LDLc, LDL cholesterol.

1268 PKM2 and New Markers of Renal Protection in T1D Diabetes Care Volume 42, July 2019

Figure 2—Plasma proteomics and metabolomics among CKD-protected Medalists vs. those with CKD. A: Plasma proteomics (volcano plot): folddifferences between the protected (n = 96) and nonprotected (n = 84) groups (x-axis) are plotted against 2log10 FDRs. Proteins of interest (glucosemetabolism, glycolysis, pathways, and renalmarkers) are presentedasblue (downregulated) and red (upregulated) circles. All otherproteins are presentedaswhite circles. Proteomic profiling of plasma samples from180Medalists evaluated a total of 1,129 proteins. B: Proteomic pathway analysis:most significantpathways shown (FDR ,0.05). Red bars indicate upregulated pathways in CKD-protected Medalists, and blue bars represent downregulated pathways. C:Plasma metabolomics (volcano plot): fold differences between the protected (n = 157) and nonprotected (n = 57) groups (x-axis) are plotted against2log10FDRs.Metabolites of interest (glucosemetabolism, glycolysis, TCA, and purine pathways) are presented as blue (downregulated) and red (upregulated) circles.All other metabolites are presented as white circles. Metabolomic profiling of plasma samples from 214 Medalists evaluated a total of 59 metabolites. D:Metabolomic pathway analysis:most significant pathways shown (FDR,0.05). Red bars indicate upregulated pathways in CKD-protectedMedalists, and bluebars represent downregulated pathways. E: Network analysis: subnetworks connecting proteins and metabolites significantly up- and downregulatedin CKD-protected Medalists. Network shows proteins (circles) and metabolites (squares) presented as blue (downregulated) and red (upregulated). Nodeswith FDR #5 3 1025 in no CKD vs. CKD and hexoses are included in the subnetworks. Isolated nodes were removed. KEGG, Kyoto Encyclopedia ofGenes and Genomes. CKD is stage 3b, as defined by CKD-EPI eGFR ,45 mL/min/1.73 m2.

care.diabetesjournals.org Gordin and Associates 1269

preserved renal function, confirming theprevious report among CKD-protectedindividuals with type 1 diabetes of ex-treme duration (13). These findings ad-vocate that glomerular cells, underprolonged hyperglycemic conditions,decrease or maintain intracellular glu-cose levels in physiological ranges by

increasing flux via glycolysis and AR path-ways and promote degradation of toxicglucose metabolites like methylglyoxal.

Interestingly, plasma proteomic anal-ysis also indicated upregulation of glu-cose metabolism pathways (pentosephosphate, glycolytic, gluconeogenic, andpyruvate) in CKD-protected Medalists.

Four of the enzymes (PKM2, TPI, LDH,and PGM1) that were elevated in CKD-protected glomeruli in our previous studywere also elevated in the plasma (13).These findings of similar enzymatic changesin plasma and glomeruli in CKD protec-tion were surprising because the contri-butions of these cytosolic enzymes from

Figure 3—Replication and validation of SOMAscan findings: TNF-a receptors and APP ELISAs. A and B: Correlation plots of TNFRSF1A and TNFRSF1BSOMAscan vs. ELISA measurements. Markers were log2-transformed for normalization. C and D: Levels of TNF markers among Medalists with andwithout CKD, as measured by ELISA. Markers were log2-transformed for normalization. Student t test was used to test for significant differencesbetween the groups. E: Correlation plot of APP SOMAscan vs. ELISAmeasurements.Markers were log2-transformed for normalization. F: Levels of APPamong Medalists with and without CKD, as measured by ELISA. APP was log2-transformed for normalization. Student t test was used to test for significantdifferences between the groups. CKD is stage 3b, as defined by CKD-EPI eGFR ,45 mL/min/1.73 m2.

1270 PKM2 and New Markers of Renal Protection in T1D Diabetes Care Volume 42, July 2019

the glomerular cells were unlikely to besignificant in the systemic circulation,suggesting that preserving physiologicalregulation of glucose and mitochondrialmetabolism may be critical in protectingother organs in hyperglycemic condi-tions. Additionally, in our proteomicstudy, though it did not meet the signif-icant threshold of FDR ,0.05, we alsofound FGF23 significantly associated withCKD (fold change 1.11; P = 0.0185; FDR =0.06). FGF23 is an exciting new biomarkershown to be part of the pathology ofrenal and general vascular disease andeventually mortality in these individuals(25).Metabolomics of CKD-protectedMed-

alists’ plasma showed reduction of me-tabolites that were part of the glucose/hexose (glucoronate, sorbitol, and ino-sitol), mitochondrial (fumarate, maleate,and aconitate), amino acid (adenine andthymine), and purine degradation (urateand hypoxanthine) pathways (26). Ele-vations of pyruvate and ADP withoutparallel increases of lactate and ATPsuggested preservation of oxidativephosphorylation and mitochondrial me-tabolism (27–30). Similarly, elevatedplasma GDP suggested preserved mito-chondrial functions in CKD-protectedMedalists due to its actions on AMPK(31). Metabolomic pathway analysis sup-ported the individual metabolite findingsby showing downregulation of hexose,mitochondrial, amino acid, and purinepathways, which is expected in peo-ple with preserved renal function.Again, metabolomics were conductedin plasma, reinforcing the idea that path-ways regulating intracellular free glucoseinvolve multiple organs other than thekidney.The upregulation of glucose metabo-

lism and mitochondrial biogenesis sug-gest a higher ATP yield per O2 as theamount of pyruvate presented to themitochondrial TCA is presumably ele-vated. This is important, especially ina hyperglycemic milieu, in which thisprocess results in 36 ATP moleculescompared with only 2 when energy isgenerated through aerobic glycolysis(the Warburg effect) (32–34). This maybe very critical in the context of theextremely high-energy requirement ofthe kidney as well as considering thatthe kidney is extremely vulnerable tohypoxia, which has been increasinglyimplicated in the pathogenesis of diabetic

kidney disease (35). In diabetic nephrop-athy, mitochondrial energetics are al-tered to increase reactive oxygenspecies, eventually causing a decreasein ATP production and an increase inapoptosis (36).

Network analysis identified severalnodes of interest, such as enhancementof APP, EGFR, SGTA, and PKM2 and re-duction of vascular endothelial growthfactor (VEGF) in CKD-protected Medal-ists. Inflammatory pathways, includingTNF-a receptors, were downregulated inthese individuals. Some of these findingswere expected, including reduced in-flammatory cytokines and VEGF clusters,because elevations of these pathwayshave been associated with increased CKDrisk (37). However, the connection ofincreased APP levels to decreased levelsof inflammatory cytokine receptors(TNFR), as well as proliferation of relatedsignaling (mitogen-activated protein ki-nase and PKCG) and VEGFA, is new andindicates that APP elevation may be acontributing factor in protection fromCKD (38). Importantly, the elevatedAPP in CKD-negative individuals was val-idated through plasma ELISA assay ofanother distinct subset of Medalists.

The discovery of elevated APP in peo-ple without CKD is novel and interesting.Results of the network analysis seem toindicate that APP may have effects onmany pathways including glucose me-tabolism (PKM, TPI, and GAPDH) and cellgrowth (EGFR and mitogen-activatedprotein kinase). APP, together with threeamyloid precursor-like proteins, builds afamily of transmembrane glycoproteins,highly expressed in the human brain,kidney, eye, and platelets (39,40). Inthe brain, it is thought to have a cau-sal relationship to Alzheimer diseasethrough amyloid plaques. The extraneu-ral functions of APP are less well known.APP is also expressed in the kidney, al-though its function is unknown. APP2/2

mice show impaired differentiation ofboth amacrine and horizontal cells, sug-gesting APP modulates normal neuronaldevelopment in the retina (40). In linewith these experimental findings, weobserved APP not only to be upregulatedin CKD-protected individuals but also inthose without PDR. Finally, in platelets,APP inhibits thrombin, regulating hemo-stasis and eventually serving as a co-agulation inhibitor (41), which may partlyexplain the beneficial effects of APP on

the vessels. In our study, APP inverselycorrelated with a few prothromboticproteins, of which THBS4 was the mostremarkable, as it is known to promotemyocardial fibrosis and retinal angiogen-esis partly through the TGF-b pathway(42,43).

Vascular endothelial APP expressionincreases during hyperglycemia (44) andis further enhanced by ischemia, stress,and inflammation, suggesting that APPelevations may be a response to cellularstress (39). It is possible that elevatedAPP in renal tissue and systemically mayexert protection for vascular damage;however, APP has also been reportedto promote oxidative stress and inflam-mation (39). Further, a previous reportsuggests the opposite findings that theexpression of APP may be elevated inrodents with experimental diabetes (45).Clearly, more studies are needed toclarify APP’s role on renal metabolismand function.

There are multiple limitations of thisstudy. First, not all proteins that wereupregulated in our previous study inMedalists’ glomeruli were found in theplasma SOMAscan, but 50% of thoseincluded could be replicated. This isexpected because the contribution ofglomerular proteins to plasma proteinswill be fairly low. Second, lower glycolyticproteins in glomeruli of CKD-positiveindividuals with type 2 diabetes is sur-prising, given the short duration of di-abetes. It is premature to assume thatonly some of these individuals will befully protected from CKD, and otherfactors such as good glycemic controlneed to be considered, but our databaselacks information on HbA1c levels at thetime of death of these individuals. Aswell, given that the postmortem resultsshowed no difference between the con-trol subjects and the group with type 2diabetes with high GFR, it may suggestthat in the general population, glycolyticenzymes are decreased in late stages oftype 2 diabetes, rather than elevated inearly stages as seen in the protectedcohort of the Medalists (13). Levels ofenzymes do not also necessarily reflecttheir activities in different stages of di-abetes. Overall, however, these datasupport the beneficial role of increasedglycolytic flux even in a general popula-tion with diabetes. Among those with ashorter duration of type 1 diabetes, wecould not replicate our previous findings

care.diabetesjournals.org Gordin and Associates 1271

of increased glycolytic proteins in glo-meruli of CKD-protected Medalists, per-haps because we have included a widerange of diabetes duration among theparticipants studied, and due to thechallenge of procuring postmortem kid-ney specimens, we had a limited powerto reach P , 0.05 for the fold changesobserved in our previous publication.Based on the numbers of samplesfrom postmortem whole kidneys thecutoff for severe renal disease was setto eGFR,30mL/min/1.73m2 in order toinclude a sufficient amount of individualsin the groups. This matter weakens theargument for comparability of the post-mortem study with the omics study. It is,however, improbable that the observedresults would not be valid due to thiscategorization, also considering that thefocus was on CKD-protective proteins inthose with less severe kidney disease,which would only dilute the results ifanything. Third, the plasma proteomicanalyses were only performed and rep-licated in people with long duration oftype 1 diabetes. Further studies will beneeded in people with shorter durationof type 1 and type 2 diabetes. Our datasuggesting increased glycolytic flux andimproved mitochondrial biogenesis aswell as upregulation of a novel protein,APP, to protect from diabetic kidneydisease in type 1 and type 2 diabetesneeds to be replicated in other cohortsof diabetes. Ideally, these cohorts wouldcomprise longitudinal blood samples andrenal specimen both pre- and postmor-tem. Only then would the true role ofthese enzymes and pathways be clarifiedand the best protein and pathway iden-tified for pharmacological interventions.In summary, the combined findings

from our current and past reportsstrongly indicate that elevation or main-tenance of glucose metabolic pathwaysin renal glomeruli and probably otherorgans is important for preservation ofrenal function. Both proteomics and me-tabolomics analyses in renal glomeruliand plasma imply that the mechanism forprotection against CKD appears to in-volve controlling intracellular free glucoseand neutralizing toxic glucose metabo-lites. Surprisingly, elevated glycolytic en-zymes, which may increase glycolyticflux, as shown previously, were associatedwith maintaining mitochondrial func-tions, opposite to previous postulations.Bioinformatic analysis of plasma proteins

validated known biomarkers of CKD suchas TNFRs and also suggested new poten-tial targets including APP, which may berelated to multiple glycolytic enzymes andgrowth-related pathways important formaintaining renal glomerular and tubularstructures and functions. Further clinicalstudies will be needed to detect whichactivities of glucose pathways can be help-ful in preserving renal function amongindividuals with diabetes.

Acknowledgments. The Medalist Study thanksthe Clinical Research Center, Joslin DiabetesCenter, Boston, MA, for assistance as well as theMedalists for participating in this study. The authorsalso thank Jialin Fu, Joslin Diabetes Center,Boston, MA, for technical help.Funding. The50-YearMedalist Study is fundedbythe National Institute of Diabetes and Digestive andKidney Diseases (P30-DK-036836, UL1-RR-025758-03,R24-283-DK-083957-01, DP3-DK-094333-01, and T32-DK-007260), JDRF (17-2013-310), the TomBeatson, Jr 284 Foundation, andmanyMedalists.D.G. was supported by a Mary K. Iacocca Fellow-ship provided by the Iacocca Foundation andgrants from the Wilhelm and Else StockmannFoundation, The Medical Society of Finland (FinskaLakaresallskapet), the Finnish Medical Founda-tion, and the Biomedicum Helsinki Foundation.Duality of Interest. This study was also sup-ported by a basic research grant from SanofiDeutschland GmbH (Frankfurt am Main, Ger-many). W.Q. is an employee of AstraZeneca. T.Sa.and A.K. are employees of Sanofi. H.A.K. is anemployee of Sanofi-Genzyme. No other poten-tial conflicts of interest relevant to this articlewere reported.AuthorContributions. D.G. andH.S. researcheddata, performed statistical analyses, and wrote themanuscript. T.Sh., R.S-L., and K.P. researched data,performed laboratory experiments, and contrib-uted to themanuscript.W.Q., I-H.W., V.B.,M.J.B.,and L.J.T. recruited and managed the participant-level clinical data as well as the blood and tissuesamples. S.M.P., D.M.P., J.M.D., andH.P. researcheddata, performed statistical analyses, and contributedto the manuscript. Y.D., T.Sa., A.K., and H.A.K. con-tributed to discussion and reviewed and edited themanuscript. M.A.N. supervised the TNF markermeasurements and analyses. P.A. assessed all ofthe human kidney pathology. D.G., H.S., T.Sh.,R.S-L., W.Q., K.P., S.M.P., D.M.P., I-H.W., V.B.,M.J.B., L.J.T., J.M.D.,H.P., Y.D.,M.A.N.,P.A., T.Sa.,A.K., H.A.K., and G.L.K. reviewed themanuscript.G.L.K. researched data and wrote the manu-script. G.L.K. is the guarantor of this work and, assuch, had full access to all of the data in the studyand takes responsibility for the integrity of thedata and the accuracy of the data analysis.Prior Presentation. Parts of this study werepresented at the 78th Scientific Sessions of theAmerican Diabetes Association, Orlando, FL, 22–26 June 2018.

References1. Gregg EW, Li Y, Wang J, et al. Changes in

diabetes-related complications in the United

States, 1990-2010. N Engl JMed 2014;370:1514–15232. Groop P-H, Thomas MC, Moran JL, et al.;FinnDiane Study Group. The presence and se-verity of chronic kidney disease predicts all-causemortality in type 1 diabetes. Diabetes 2009;58:1651–16583. OrchardTJ, SecrestAM,MillerRG,CostacouT.In the absence of renal disease, 20 year mortal-ity risk in type 1 diabetes is comparable to thatof the general population: a report from thePittsburgh Epidemiology of Diabetes Complica-tions Study. Diabetologia 2010;53:2312–23194. Lee TS, Saltsman KA, Ohashi H, King GL. Acti-vation of protein kinase C by elevation of glucoseconcentration: proposal for a mechanism in thedevelopment of diabetic vascular complications.Proc Natl Acad Sci U S A 1989;86:5141–51455. Dunlop M. Aldose reductase and the roleof the polyol pathway in diabetic nephropathy.Kidney Int Suppl 2000;77:S3–S126. Greene DA, Lattimer SA, Sima AAF. Sorbitol,phosphoinositides, and sodium-potassium-ATPase in the pathogenesis of diabetic compli-cations. N Engl J Med 1987;316:599–6067. Giacco F, Du X, D’Agati VD, et al. Knockdownof glyoxalase 1 mimics diabetic nephropathy innondiabetic mice. Diabetes 2014;63:291–2998. Mann JF, Green D, Jamerson K, et al.; ASCENDStudy Group. Avosentan for overt diabetic ne-phropathy. J Am Soc Nephrol 2010;21:527–5359. Frank RN. The aldose reductase controversy.Diabetes 1994;43:169–17210. de Zeeuw D, Akizawa T, Audhya P, et al.;BEACON Trial Investigators. Bardoxolone methylin type 2 diabetes and stage 4 chronic kidneydisease. N Engl J Med 2013;369:2492–250311. Keenan HA, Costacou T, Sun JK, et al. Clinicalfactors associated with resistance to microvas-cular complications in diabetic patients of ex-treme disease duration: the 50-year MedalistStudy. Diabetes Care 2007;30:1995–199712. Gordin D, Harjutsalo V, Tinsley L, et al.Differential association of microvascular attribu-tions with cardiovascular disease in patientswith long duration of type 1 diabetes. DiabetesCare 2018;41:815–82213. Qi W, Keenan HA, Li Q, et al. Pyruvate kinaseM2 activation may protect against the progressionof diabetic glomerular pathology and mitochon-drial dysfunction. Nat Med 2017;23:753–76214. Tinsley LJ, Kupelian V, D’Eon SA, et al. As-sociation of glycemic control with reduced riskfor large-vessel disease after more than 50 yearsof type 1 diabetes. J Clin EndocrinolMetab 2017;102:3704–371115. Ngo D, Sinha S, ShenD, et al. Aptamer-basedproteomic profiling reveals novel candidate bio-markers and pathways in cardiovascular disease.Circulation 2016;134:270–28516. Finley LWS, Lee J, Souza A, et al. Skeletalmuscle transcriptional coactivator PGC-1a me-diates mitochondrial, but not metabolic, changesduring calorie restriction. Proc Natl Acad Sci U S A2012;109:2931–293617. JainM,NilssonR, Sharma S, et al.Metaboliteprofiling identifies a key role for glycine in rapidcancer cell proliferation. Science 2012;336:1040–104418. Niewczas MA, Gohda T, Skupien J, et al. Cir-culating TNF receptors 1 and 2 predict ESRD in type2 diabetes. J Am Soc Nephrol 2012;23:507–515

1272 PKM2 and New Markers of Renal Protection in T1D Diabetes Care Volume 42, July 2019

19. Ritchie ME, Phipson B, Wu D, et al. limmapowers differential expression analyses forRNA-sequencing and microarray studies. NucleicAcids Res 2015;43:e4720. Benjamini Y, Hochberg Y. Controlling thefalse discovery rate: a practical and powerfulapproach tomultiple testing. J R Stat Soc B 1995;57:289–30021. Wu D, Lim E, Vaillant F, Asselin-Labat M-L,Visvader JE, Smyth GK. ROAST: rotation geneset tests for complex microarray experiments.Bioinformatics 2010;26:2176–218222. Cerami EG, Gross BE, Demir E, et al. PathwayCommons, aweb resource forbiological pathwaydata. Nucleic Acids Res 2011;39:D685–D69023. Csardi G, Nepusz T. The igraph softwarepackage for complex network research. Inter-Journal Complex Syst 2006;1695:1–924. Reidy K, Kang HM, Hostetter T, Susztak K.Molecular mechanisms of diabetic kidney dis-ease. J Clin Invest 2014;124:2333–234025. Kuro-O M. The Klotho proteins in healthand disease. Nat Rev Nephrol 2019;15:27–4426. Darshi M, Van Espen B, Sharma K. Metab-olomics in diabetic kidney disease: unravelingthe biochemistry of a silent killer. Am J Nephrol2016;44:92–10327. Avogaro A, Toffolo G, Miola M, et al. In-tracellular lactate- and pyruvate-interconversionrates are increased in muscle tissue of non-insulin-dependent diabetic individuals. J ClinInvest 1996;98:108–11528. Persson MF, Franzen S, Catrina S-B, et al.Coenzyme Q10 prevents GDP-sensitive mito-chondrial uncoupling, glomerular hyperfiltrationand proteinuria in kidneys from db/db mice as a

model of type 2 diabetes. Diabetologia 2012;55:1535–154329. Kiviniemi TO, Yegutkin GG, Toikka JO, et al.Pravastatin-induced improvement in coronaryreactivity and circulating ATP and ADP levelsin young adults with type 1 diabetes. FrontPhysiol 2012;3:33830. Nakatsu D, Horiuchi Y, Kano F, et al. L-cysteine reversibly inhibits glucose-induced bi-phasic insulin secretion and ATP production byinactivating PKM2. Proc Natl Acad Sci U S A 2015;112:E1067–E107631. Hawley SA, Ross FA, Chevtzoff C, et al. Useof cells expressing gamma subunit variants toidentify diverse mechanisms of AMPK activation.Cell Metab 2010;11:554–56532. Vander Heiden MG, Cantley LC, ThompsonCB. Understanding the Warburg effect: themetabolic requirements of cell proliferation.Science 2009;324:1029–103333. Yang W, Lu Z. Nuclear PKM2 regulates theWarburg effect. Cell Cycle 2013;12:3154–315834. Palsson-McDermott EM, Curtis AM, Goel G,et al. Pyruvate kinase M2 regulates hif-1a ac-tivity and IL-1b induction and is a critical de-terminant of the Warburg effect in LPS-activatedmacrophages. Cell Metab 2015;21:34735. Wang Z, Ying Z, Bosy-Westphal A, et al.Specific metabolic rates of major organs andtissues across adulthood: evaluation by mech-anistic model of resting energy expenditure.Am J Clin Nutr 2010;92:1369–137736. Flemming NB, Gallo LA, Ward MS, ForbesJM. Tapping into mitochondria to find noveltargets for diabetes complications. Curr DrugTargets 2016;17:1341–1349

37. Dei Cas A, Gnudi L. VEGF and angiopoie-tins in diabetic glomerulopathy: how far for anew treatment? Metabolism 2012;61:1666–167338. Muller UC, Deller T, Korte M. Not justamyloid: physiological functions of the amyloidprecursor protein family. Nat Rev Neurosci 2017;18:281–29839. d’Uscio LV, He T, Katusic ZS. Expression andprocessing of amyloid precursor protein in vas-cular endothelium. Physiology (Bethesda) 2017;32:20–3240. Dinet V, An N, Ciccotosto GD, et al. APPinvolvement in retinogenesis of mice. Acta Neu-ropathol 2011;121:351–36341. Smith RP, Higuchi DA, Broze GJ Jr. Plateletcoagulation factor XIa-inhibitor, a form of Alz-heimer amyloid precursor protein. Science 1990;248:1126–112842. Muppala S, Xiao R, Krukovets I, et al. Throm-bospondin-4mediates TGF-b-induced angiogen-esis. Oncogene 2017;36:5189–519843. Stenina OI, Desai SY, Krukovets I, et al.Thrombospondin-4 and its variants: expressionand differential effects on endothelial cells.Circulation 2003;108:1514–151944. Chao AC, Lee TC, Juo SH, Yang DI. Hyper-glycemia increases the production of amyloidbeta-peptide leading to decreased endothelialtight junction. CNS Neurosci Ther 2016;22:291–29745. Gilbert RE, Cox A, Dziadek M, Cooper ME,Jerums G. Extracellular matrix and its interac-tions in the diabetic kidney: a molecular bi-ological approach. J Diabetes Complications 1995;9:252–254

care.diabetesjournals.org Gordin and Associates 1273