Exercise Response Variations inSkeletal Muscle PCr RecoveryRate and Insulin Sensitivity Relateto Muscle Epigenomic Profiles inIndividuals With Type 2 DiabetesDiabetes Care 2018;41:2245–2254 | https://doi.org/10.2337/dc18-0296

OBJECTIVE

Some individuals with type 2 diabetes do not reap metabolic benefits fromexercise training, yet the underlying mechanisms of training response varia-tion are largely unexplored. We classified individuals with type 2 diabetes (n = 17)as nonresponders (n = 6) or responders (n = 11) based on changes in phosphocreatine(PCr) recovery rate after 10 weeks of aerobic training. We aimed to determinewhether the training response variation in PCr recovery rate was marked by distinctepigenomic profiles in muscle prior to training.

RESEARCH DESIGN AND METHODS

PCr recovery rate as an indicator of in vivo muscle mitochondrial function in vastuslateralis (31P-magnetic resonance spectroscopy), insulin sensitivity (M-value;hyperinsulinemic-euglycemic clamp), aerobic capacity (VO2peak), and blood pro-files were determined pretraining and post-training. Muscle biopsies were per-formed pretraining in vastus lateralis for the isolation of primary skeletalmuscle cells (HSkMCs) and assessments of global DNA methylation and RNAsequencing in muscle tissue and HSkMCs.

RESULTS

By design, nonresponders decreased and responders increased PCr recovery ratewith training. In nonresponders, insulin sensitivity did not improve and glycemiccontrol (HbA1c) worsened. In responders, insulin sensitivity improved. VO2peak

improved by ∼12% in both groups. Nonresponders and responders were distin-guished by distinct pretraining molecular (DNA methylation, RNA expression)patterns in muscle tissue, as well as in HSkMCs. Enrichment analyses identifiedelevations in glutathione regulation, insulin signaling, and mitochondrial metab-olism in nonresponders pretraining, which was reflected in vivo by higher pre-training PCr recovery rate and insulin sensitivity in these same individuals.

CONCLUSIONS

A training response variation for clinical risk factors in individuals with type 2diabetes is reflected by distinct basal myocellular epigenomic profiles in muscletissue, some of which are maintained in HSkMCs, suggesting a cell-autonomousunderpinning. Our data provide new evidence to potentially shift the diabetestreatment paradigm for individuals who do not benefit from training, such thatsupplemental treatment can be designed.

1Translational Research Institute forMetabolismand Diabetes, Florida Hospital, Orlando, FL2Sanford Burnham Prebys Medical DiscoveryInstitute, La Jolla, CA3Sanford Burnham Prebys Medical DiscoveryInstitute, Orlando, FL

Corresponding author: LaurenM. Sparks, lauren.sparks@flhosp.org.

Received 8 February 2018 and accepted 15 July2018.

Clinical trial reg. no. NCT01911104, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc18-0296/-/DC1.

N.A.S. and B.B. contributed equally to this work.

© 2018 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

Natalie A. Stephens,1 Bram Brouwers,1

Alexey M. Eroshkin,2 Fanchao Yi,1

Heather H. Cornnell,1 Christian Meyer,1

Bret H. Goodpaster,1,3 Richard E. Pratley,1,3

Steven R. Smith,1,3 and Lauren M. Sparks1,3

Diabetes Care Volume 41, October 2018 2245

CARDIOVASCULA

RANDMETA

BOLIC

RISK

By 2050, the number of people who havereceived a diagnosis of diabetes in theU.S. will reach 29 million (1). Skeletalmuscle insulin resistance is a hallmark oftype 2 diabetes and is linked with de-ficient skeletal muscle (muscle) mito-chondrial oxidative capacity (function)(2,3). Considering group averages, exer-cise training restores muscle mitochon-drial function in individuals with type 2diabetes to levels observed in healthyindividuals, in parallel with improve-ments in insulin sensitivity (4). Interven-tion studies (5–7) have shown thatglucose homeostasis, insulin sensitivity,or muscle mitochondrial density fails toimprove with supervised exercise train-ing in ;20–40% of individuals. Conclu-sions about the beneficial effects ofexercise training are too often drawnbased on average responses, and littleis known about those individuals whodo not respond favorably to training.Variation in the training response in

vivo has been linked with distinct basalgenetic and transcriptional profiles inmuscle tissue (8–10); however, the in-fluence of the epigenome on trainingresponse variation in individuals withtype 2 diabetes remains poorly under-stood. Individuals within a family re-spond more similarly than those ofdifferent families, suggesting that DNAsequence variation or epigenomic modifi-cations contribute to training responsevariability (6,11,12). For example, in first-degree relatives of patients with type 2diabetes, the single nucleotide polymor-phism rs540467 in theNDUF6B genewaslinked to exercise response variation inmuscle ATP synthesis rates in vivo afteracute aerobic bouts (8). In individualsat risk for the development of type 2diabetes, higher activation of transform-ing growth factor-b (TGFB1) signalingcontributed to a blunted exercise-induced improvement in insulin sensitiv-ity via the suppression of key regulatorsof muscle mitochondrial fuel oxidation(13). Moreover, DNA hypomethylationis reported to be an early event incontraction-induced gene activation (tran-scriptional upregulation) in muscle, andchronic exercise training modifies genome-wide DNA methylation patterns inhuman muscle, specifically in genes re-lated with metabolic processes andmitochondrial function (e.g., PPARGC1A,PDK4, PPARD) (14). We have previ-ously reported (10) large differences

in pretraining mRNA expressions of genesinvolved in substrate metabolism andmitochondrial biogenesis in muscle tis-sue of individuals with type 2 diabeteswho did or did not improve their met-abolic profile with exercise training,suggesting a unique baseline gene ex-pression pattern, which may underliean individual’s lack of training response.

In this study, we classified individualswith type 2 diabetes as nonrespondersor responders, based on their changes inphosphocreatine (PCr) recovery rate af-ter 10 weeks of supervised aerobic train-ing. Training responses in metabolicrisk factors and cardiorespiratory fitnesswere compared between groups. Globallevels of pretraining DNA methylationand RNA expression were analyzed inmuscle tissue and in primary humanskeletal muscle cells (HSkMCs) of non-responders and responders.Wehypoth-esized that training response variationin muscle mitochondrial function in vivowould be marked by distinct epigenomicprofiles in the muscle prior to the ex-ercise training intervention.

RESEARCH DESIGN AND METHODS

ParticipantsSeventeen individuals with type 2 dia-betes (n = 8 females; mean age 50.76 1.9years; BMI 34.96 0.9 kg/m2) completedthe study. Six individuals dropped out ofthe study for personal reasons and/orat the discretion of the Principal In-vestigator. Participants were seden-tary, defined as not being physicallyactive $3 days per week for the previous6 months. Type 2 diabetes was de-termined by self-report and/or fastingplasma glucose $7.0 mmol/L. HbA1c

levels at the time of study enrollmenthad to be #8.5% (69 mmol/mol) forindividuals with type 2 diabetes receivingglucose-lowering medication (e.g., met-formin, sulfonylurea) and between 6.0%(42 mmol/mol) and 8.5% (69 mmol/mol)for individuals with type 2 diabetesbeing treated with diet alone. Participantsceased glucose-lowering treatment 15–17 days prior to the start of the aerobictraining program and remained off med-ication for the entire duration of the study.Two individuals resumed treatment withtheir prescribed sulfonylurea during thestudy because of consistent hyperglycemia(fasting plasma glucose .22.2 mmol/L).Participants were asked to maintain die-tary behavior during the study, and a 3-day

food recall was collected pretraining andpost-training. The 2 days prior to thepretraining and post-training metabolicassessments, participants were pro-vided meals as part of a standard Amer-ican diet (35% fat, 16% protein, 49%carbohydrate) and instructed to con-sume all of the food provided. On theevening prior to the hyperinsulinemic-euglycemic clamp, participants stayedovernight in the metabolic ward andconsumed a meal as part of a standardAmerican diet. All participants providedwritten informed consent. All proce-dures were performed under a researchprotocol approved by the Florida Hos-pital Institutional Review Board.

Aerobic Training ProtocolParticipants underwent supervised aer-obic training on a treadmill for 10 weeks(4 days/week) of a ramped training pro-tocol proven to improve insulin sensitiv-ity in individuals with overweight andobesity (15,16). All exercise sessionswere supervised to ensure that the tar-geted intensity (based on target heartrate) and duration were achieved. Dur-ing weeks 1–4, participants performedexercise for at least 20 min per sessionat an intensity of 50–70% VO2peak. Duringweeks 5–8, participants increased theexercise time to 45 min per session atthe same intensity. During weeks 9–10,participants performed exercise for 45min per session at 75% VO2peak.

In VivoMuscleMitochondrial FunctionIn vivo muscle mitochondrial functionwas evaluated by PCr recovery rate invastus lateralis using phosphorus (31P)magnetic resonance spectroscopy on a3-T Achieva magnet (Philips Healthcare,Andover, MA), as previously described(2). The PCr time constant (tau) was usedto measure PCr recovery rate (1/tau). Thecoefficient of variation for this measure-ment was 4.5%. Participants were sep-arated into two groups (nonresponders,n = 6; responders, n = 11) based on thepercentage change in muscle mitochon-drial function in vivo after 10 weeks ofaerobic training, which was defined asthe PCr recovery rate post-training minusthe PCr recovery rate pretraining, dividedby the PCr recovery rate pretraining.Participants who decreased the PCr re-covery rate were identified as nonre-sponders; participants who increasedthe PCr recovery rate were identified asresponders (Fig. 1A).

2246 Exercise Response Variation in Type 2 Diabetes Diabetes Care Volume 41, October 2018

Insulin SensitivityInsulin sensitivity was measured byhyperinsulinemic-euglycemic clamp, aspreviously described (17) with modi-fications. Participants arrived at theresearch facility the evening prior tothe hyperinsulinemic-euglycemic clamp,consumed a standard American meal,and stayed overnight in the metabolicward. After an overnight fast, insulin(100 mU/m2/min) and 20% glucosewere administered for 2 h to maintainthe plasma glucose concentration at;5.0 mmol/L. Plasma levels of glucosewere measured at 5-min intervals,and steady state was reached after95 to 120 min (Supplementary Fig.1A). Whole-body insulin sensitivity(M-value) was assessed as the meanglucose infusion rate adjusted for glu-cose space correction (mg/kg fat-freemass/min) during the steady state (18).Mean plasma insulin levels were compa-rable during steady state (Supplemen-tary Fig. 1B). The M-value/insulin (M/I)ratio was calculated as the M-valuedivided by the mean insulin concentra-tions during steady state for every par-ticipant.

Maximal Aerobic CapacityVO2peak was determined by an incre-mental treadmill test on a TrackmasterTMX 425c (Full Vision, Inc., Newton, KS)as previously described (19).

Body Composition and Blood AnalysesBody composition was measured by DXAusing a Lunar iDXA Whole-Body Scanner(GE Healthcare Lunar, Madison, WI) (2).Fasting and steady-state (clamp) bloodsamples were analyzed in clinical chem-istry laboratory at either Florida Hospitalor onsite at the Translational ResearchInstitute for Metabolism and Diabetes,where appropriate (20).

Muscle BiopsyMuscle biopsies were performed invastus lateralis muscle prior to training.Muscle tissue was either snap frozen orused for the isolation of myogenic pro-genitor cells (20,21).

Myogenic Progenitor Cell Isolation and

Differentiation

Myogenic progenitor cells (HSkMCs)were isolated as described previously(22). HSkMCs (myoblasts) were immuno-purified using mouse monoclonal 5.1H11anti-CD56 antibody (23). Myoblasts were

differentiated into myotubes for 5–7days (22).

Genome-Wide DNA Methylation Analysis

Total DNA was analyzed in all nonre-sponders (n = 6) and in the responderswith the most prominent increase in PCrrecovery rate with training (n = 6). DNAwas isolated from pretraining muscletissue and myotubes using the DNeasyBlood and Tissue Extraction Kit (Qiagen,Valencia, CA). Genome-wide methyla-tion analysis was performed using theIllumina Infinium MethylationEPICBeadChip platform, which interrogates. 850,000 CpG methylation sites. Rawdata were summarized into BeadStudioIDAT files for further analysis using thePartek Genomic Suite (Partek, Inc., St.Louis, MO). Data were normalized usingthe SWAN (Subset-Quantile With ArrayNormalization) method. Differentiallymethylated CpG sites between nonre-sponders and responderswere identifiedby ANOVA using relaxed conditions todefine differentially methylated CpGsites (unadjusted P value cutoff = 0.05;estimated change in b-value cutoff$0.2, corresponding to a $20% changein methylation between groups). Cor-responding genes were used for enrich-ment analyses.

RNA Sequencing Analyses

Total RNA was analyzed in pretrainingmuscle tissue and myotubes of all non-responders (n = 6) and in those of re-sponders (n = 6; same participants as forDNA methylation) with the most prom-inent increase in PCr recovery rate withtraining, and isolated using RNeasy Fi-brous Tissue kit (Qiagen). Raw data werequality controlled by FastQC. Mappingto human genome (UCSC hg19) wasperformed using the Illumina CufflinksAssembly & DE workflow. Abundanceestimates for genes defined in a genomereference, FPKM (fragments per kilobaseof transcript per million mapped reads),were used for downstream analysis. Datawere further filtered as generally recom-mended: low expressed genes were re-moved (genes with max FPKM value forall samples #1), and FPKM signals werefurther log2 transformed. Genes wereremoved based on a cutoff for SD oflog2 of signals for all samples #0.2. Dif-ferentially expressed genes (DEGs) be-tween nonresponders and responderswere defined using ANOVA (unadjustedP value ,0.05 and fold change .1.5).

Figure 1—Pretraining values and training-induced changes in metabolic risk factors andcardiorespiratory fitness. A: The individual percentage changes in PCr recovery rate in vivoin skeletal muscle with training. Training response variation for PCr recovery rate was used toclassify individuals with type 2 diabetes as nonresponders (n = 6) or responders (n = 11) after10 weeks of aerobic training. Group averages in nonresponders and responders for PCr recoveryrate pretraining (B), training-induced change in PCr recovery rate (C), insulin sensitivity pretraining(D), training-induced change in insulin sensitivity (E), glycemic control (HbA1c) pretraining (F),training-induced change in glycemic control (HbA1c) (G), VO2peak (cardiorespiratory fitness)pretraining (H), and training-induced changes in VO2peak (cardiorespiratory fitness) (I). J:Correlation betweenpretraining PCr recovery rate andpretrainingM-value. *P,0.050 comparedwithpretrainingwithin the samegroup,∞P,0.050ANOVAmain timeeffect (pretrainingvs. post-training). Different capital letters [A and B] indicate significant differences between groups post-training. Data are mean 6 SD. FFM, fat-free mass.

care.diabetesjournals.org Stephens and Associates 2247

Enrichment analyses were subsequentlyperformed.

Statistical AnalysesValues are reported as the mean 6 SD.Statistical significance was set at P ,0.050. An unpaired Student t test wasused to analyze group differences pre-training. Repeated measures were ana-lyzed using a two-way repeated-measuresANOVA model. When normality was vi-olated, the appropriate nonparametrictest was performed. An ANOVA wasused to analyze pretraining differences inDNA methylation and RNA expression.A x2 test was used to analyze group dis-tributions for sex, race, and medicationuse. For associations between variables,Pearson’s correlations were performed.All statistical analyses were performedusing JMP version 13 (SAS Institute,Cary, NC), Partek NSG (Partek, Inc.), andPrism 6 (GraphPad, San Diego, CA).

RESULTS

Participant CharacteristicsIndividuals with type 2 diabetes (n = 17)were classified as nonresponders (n = 6)or responders (n = 11) based on whetherPCr recovery rate decreased or increasedwith training, respectively (Fig. 1A). Pre-training characteristics of nonrespondersand responders are summarized in Table1. Nonresponders and responders didnot differ in age (P = 0.245), sex (P =0.858), or race (P = 0.493). BMI wascomparable between nonrespondersand responders at the time of studyenrollment (34.3 6 4.2 and 35.2 63.7 kg/m2, P = 0.676, for nonrespondersand responders, respectively). The du-ration of type 2 diabetes was similarbetween nonresponders and respond-ers (P = 0.737). Nonresponders hadhigher PCr recovery rates (P = 0.005)(Fig. 1B) and higher insulin sensitivity(P = 0.041) (Fig. 1D) than responderspretraining. Values for HbA1c (P = 0.769)(Fig. 1F), VO2peak (P = 0.571) (Fig. 1H),fasting plasma glucose (P = 0.575), fast-ing insulin (P = 0.201), total cholesterol(P = 0.909), LDL cholesterol (LDL-C; P =0.767), HDL cholesterol (HDL-C; P = 0.618),triglycerides (P = 0.356), fasting plasmafree fatty acids (P = 0.998), body weight(P = 0.512), fat mass (P = 0.725), and fat-free mass (P = 0.640) were not differentbetween nonresponders and respond-ers pretraining. Pretraining PCr recoveryrate correlated positively with pretraining

insulin sensitivity (r = 0.558, P = 0.020)(Fig. 1J).

Compliance, Adherence, andMedication UseValues for compliance with (percentageof training sessions completed of 40 intotal) and adherence to (percentage oftime spent in the individual target heartrate zone during each training session)the training protocol were similar be-tween nonresponders and responders(P = 0.571 and P = 0.958, respectively)(Table 1). Caloric intake was not signif-icantly different between nonrespondersand responders pretraining (P = 0.762)(Table 1). Glucose-lowering medicationuse before the start of the study wassimilar between groups, with 67% ofnonresponders and 64% of respondersusing metformin alone (Table 1).

Training-Induced ChangesBy design, the PCr recovery rate innonresponders decreased (P = 0.031),whereas the PCr recovery rate in re-sponders increased (P = 0.001) after10 weeks of aerobic training, and thetraining-induced relative change in PCrrecovery rate was significantly differentbetween nonresponders and responders(P , 0.001) (Fig. 1C and Table 1). Un-published data from our laboratory dem-onstrate that there are no significantchanges in PCr recovery rate measured3–4 weeks apart without intervention(i.e., free living) in the same healthysedentary individuals. Insulin sensitivityincreased in responders (P = 0.032), butnot in nonresponders (P = 0.438) (Fig. 1Eand Table 1). Training-induced abso-lute changes in (D) insulin sensitivitypositively correlated with DPCr recoveryrate (r=0.517,P=0.034), andpercentagechanges in insulin sensitivity positivelycorrelated with percentage changes inPCr recovery rate (r = 0.498, P = 0.042).HbA1c increased in nonresponders (P =0.004), but not in responders (P = 0.999),and the training-induced relative changein HbA1c was significantly different be-tween nonresponders and responders(P = 0.002) (Fig. 1G and Table 1). DHbA1cnegatively correlated with the DPCr re-covery rate (r = 20.690, P = 0.019), andpercentage changes in HbA1c negativelycorrelated with percentage changes inPCr recovery rate (r =20.727, P = 0.011).VO2peak significantly increased in bothnonresponders (+14.21%)andresponders

(10.37%) after 10 weeks of training (P ,0.001) (Fig. 1I and Table 1). Body weight(P = 0.017) and fat-free mass (P = 0.005)increased in nonresponders, and therelative changes were different betweennonresponders and responders (Table 1).Values for fat mass (P = 0.429), plasmaglucose (P = 0.388), plasma insulin (P =0.985), total cholesterol (P = 0.558),LDL-C (P = 0.979), HDL-C (P = 0.346),triglycerides (P = 0.110), and plasma freefatty acids (P = 0.578) under fastingconditions did not significantly changewith training in either group (Table 1).Caloric intake did not significantly changewith training in either group (P = 0.695)(Table 1).

Pretraining Genome-Wide DNAMethylation and RNA SequencingAnalysesWe investigated the potential epige-nomic and transcriptomic contributionsto the training response variation inmuscle mitochondrial function in vivoand insulin sensitivity. To this end, globalDNA methylation of the inner cyto-sine within the CCGG sequence (CpGmethylation) and RNA expression pat-terns were assessed in muscle tissuepretraining. Striking differences in DNAmethylation patterns, representing 533differentially methylated CpG sites,were observed between nonrespondersand responders (Fig. 2A and Supplemen-tary Table 1). Of those 533 differen-tially methylated CpG sites, methylationlevels were significantly lower in thepromoter region (TSS200 region, P =0.024) (Fig. 2B) and in the CpG islandregion (P = 0.004) (Fig. 2C), which tendsto overlap with the promoter regions(24), of nonresponders versus respond-ers. RNA sequencing (RNAseq) analy-ses revealed a similar pattern (similarto DNA methylation) in nonresponderscompared to that in responders, repre-senting 118 DEGs (heatmap not shown)(Supplementary Table 3). Global DNAmethylation levels in these regions inmuscle tissue showed significant corre-lations with pretraining values for PCrrecovery rate and insulin sensitivity(Supplementary Fig. 2) and with DPCrrecovery rate and DHbA1c (Supplemen-tary Fig. 3). Enrichment analyses of bothDNA methylation and RNAseq data iden-tified pathways linked to glutathionemetabolism, insulin signaling, and mito-chondrial metabolism. The calculated

2248 Exercise Response Variation in Type 2 Diabetes Diabetes Care Volume 41, October 2018

P values for the 533 differentially meth-ylated CpG sites and their chromosomallocations are shown in Fig. 2D. The glu-tathione S-transferase m (GSTM) classgene family dominated both DNA meth-ylation and RNA expression patterns.Twelve CpG sites for GSTM1 and GSTM5were located within the gene promotorregions (chromosome 1) (Fig. 2D). Theinset in Fig. 2D shows that methylation inall 12 CpG sites located on chromosome1 for GSTM1 and GSTM5 were downregu-lated in nonresponders compared withresponders (fold change). Methylationlevels in nonresponders and respondersfor those 12 CpG sites are shown quan-titatively in Fig. 2E. Corresponding RNAexpressions for GSTM1, GSTM4, andGSTO1 were significantly higher in non-responders and responders (Fig. 2F).In the insulin signaling pathway, weobserved differentially methylated

CpG sites in the promotor regions ofmicrofibril-associated protein 3 like(MFAP3 L), phosphofructokinase muscle,and phosphoinositide-3-kinase regulatorysubunit 3 (Fig. 2E). RNA expression forMFAP3 L was significantly lower in non-responders compared with responders(Fig. 2F). RNA expressions for numeroussubunits of the electron transport systemcomplexes were differentially expressedbetween nonresponders and responders,being significantly higher in nonrespond-ers (Fig. 2F), as follows: NADH-ubiquinoneoxidoreductase subunit A2 (NDUFA2),NDUFA7, NDUFA9, NDUFA11, andubiquinol-cytochrome c reductase hingeprotein (UQCRHL). Pyruvate dehydroge-nase kinase 4 (PDK4) and growth arrestand DNA damage inducible a (GADD45A),genes related to mitochondrial metabo-lism, were differentially expressed in non-responders versus responders (Fig. 2F). In

general, DNA methylation in the promoterregions of genes corresponded with areciprocal RNA expression pattern. TheDNA methylation and RNA expressionpatterns in pretraining muscle tissue sug-gest higher antioxidant defense, insulinsignaling, and mitochondrial metabolismin nonresponders prior to the trainingintervention, which is reflected in vivoby means of higher pretraining PCr re-covery rate and higher pretraining insulinsensitivity in nonresponders comparedwith responders.

To determine whether these epige-nomic patterns were intrinsic to themuscle, we assessed global DNA meth-ylation and RNA expressions in myoge-nic progenitor cells (HSkMCs) derivedfrom these same individuals pretraining.A total of 491 CpG sites were differen-tially methylated in HSkMCs from non-responder compared with responders,

Table 1—Pretraining values and percentage changes from pretraining

Pretraining Change (%)

Nonresponders Responders Nonresponders Responders

Age (years) 47.2 6 9.5 52.6 6 6.6

Sex (F/M), n 3/3 5/6

Sex (F/M), % 50/50 45/55

Race (black/white/Hispanic/unknown), n 2/4/0/0 1/8/1/1

Race (black/white/Hispanic/unknown), % 33/67/0/0 9/73/9/9

Diabetes duration (years) 4.67 6 3.20 5.45 6 5.07

Caloric intake (kcal/day) 1,912 6 468 2,015 6 578 27.18 6 27.27 14.33 6 31.16

PCr recovery rate (1/s) 0.026 6 0.005A 0.017 6 0.004B 223.92 6 12.28*A 31.64 6 37.23*B

M-value (mg/kg fat-free mass/min) 6.5 6 2.9A 3.6 6 2.3B 213.52 6 41.3 102.69 6 211.41*

M/I 0.040 6 0.016A 0.019 6 0.012B 214.04 6 47.13 94.48 6 172.01*

HbA1c (%) 7.4 6 0.9 7.6 6 0.8 11.09 6 5.49*A 20.36 6 7.83B

HbA1c (mmol/mol) 57 6 9.8 60 6 8.7 11.09 6 5.49*A 20.36 6 7.83B

VO2peak (mL/kg/min) 20.3 6 4.8 22.0 6 4.6 14.21 6 5.24* 10.37 6 7.77*

Fasting plasma glucose (mmol/L) 9.7 6 2.3 10.6 6 2.9 13.81 6 14.73 22.49 6 24.93

Fasting insulin (mIU/mL) 16.7 6 5.5 13.1 6 3.6 26.51 6 56.89 6.04 6 36.17

Total cholesterol (mmol/L) 4.0 6 0.9 4.0 6 0.5 4.40 6 18.77 3.15 6 11.32

LDL-C (mmol/L) 2.3 6 0.6 2.4 6 0.5 3.80 6 29.14 23.45 6 16.03

HDL-C (mmol/L) 0.9 6 0.3 0.9 6 0.1 6.67 6 8.93 20.43 6 11.85

Triglycerides (mmol/L) 1.7 6 0.4 1.4 6 0.5 10.55 6 31.79 28.32 6 41.36

Fasting plasma free fatty acids (mmol/L) 0.50 6 0.27 0.50 6 0.14 2.63 6 22.36 5.99 6 10.96

Body weight (kg) 96.8 6 13.6 101.7 6 15.4 2.57 6 2.65*A 20.08 6 1.68B

Fat mass (kg) 39.5 6 9.3 41.2 6 8.7 2.77 6 4.49 21.13 6 3.69

Fat-free mass (kg) 57.3 6 14.5 60.5 6 10.1 2.87 6 2.43*A 0.56 6 1.48B

Compliance (%) 92.6 6 4.8 94.4 6 6.6

Adherencea (%) 76.8 6 11.2 77.2 6 12.8

Type 2 diabetes medications, n (%)Metformin 4 (67) 7 (64)Metformin + sulfonylurea 2 (33) 3 (27)Diet 0 (0) 1 (9)

Dataare themean6SD,unlessotherwise indicated. F, female;M,male. aRecordedevery5min. *P,0.050pretraining vs.post-trainingwithin the samegroup; different capital letters [A and B] indicate significant differences between nonresponders and responders (P , 0.050).

care.diabetesjournals.org Stephens and Associates 2249

with 113 CpG sites in common betweenpretraining muscle tissue and HSkMCsderived from the same tissue of the sameindividuals (Fig. 3A and SupplementaryTable 2). In accordance with muscletissue, of the 491 differentially methyl-ated CpG sites, methylation levels inHSkMCs were lower in the promoterregion (TSS1500 region, P = 0.005;TSS200 region, P = 0.062) (Fig. 3B) andthe CpG island region (P, 0.001) (Fig. 3C)of nonresponders compared with re-sponders. Global DNA methylation levelsin these regions in HSkMCs showedsignificant correlations with pretrainingand DPCr recovery rate and HbA1c

(Supplementary Figs. 4 and 5). Enrich-ment analysis identified pathways re-lated to glutathione metabolism. Thecalculated P values for the 491 differen-tially methylated CpG sites and their

chromosomal locations are shown inFig. 3D. Three CpG sites in GSTM1 andthree CpG sites in GSTM5 were locatedwithin the promotor regions (chromo-some 1) (Fig. 3D) and downregulated innonresponders compared with respond-ers (fold change; inset in Fig. 3D). Meth-ylation levels in nonresponders andresponders for those six CpG sites areshown quantitatively in Fig. 3E. DNAmethylation in the promotor regionsof protein kinase C a, phospholipase Cb 4 (PLCB4) and PLCL2 related to insulinsignaling and for NDUFA12 related tomitochondrial metabolism were alsodifferent in nonresponders comparedwith responders (Fig. 3E). RNAseq anal-yses in the pretraining HSkMCs (n = 147)(Supplementary Table 4) showed anoverlap of five DEGs in nonresponderscompared with responders with the

pretraining muscle tissue DEGs (n =118) (Supplementary Table 3). Of the147 total DEGs in pretraining HSkMCs,one gene was also differentially methyl-ated in the HSkMCs (PLCL2). RNA ex-pressions of selected genes relevant toglutathione regulation, mitochondrialmetabolism, and insulin signaling areshown quantitatively in Fig. 3F. Interest-ingly, GSTM1 is higher (P = 0.070) inpretraining HSkMCs from nonrespond-ers compared with responders (Fig. 3F),which corresponds to the findings inmuscle tissue GSTM1 RNA expression(Fig. 2F), as well as reduced DNA meth-ylation in the promoter region of GSTM1in both pretraining muscle tissue andHSkMCs (Figs. 2E and 3E). Thus, thetraining response variation observed inthe epigenomic profiles of muscle tissueappeared to be cell autonomous.

Figure 2—Pretraining DNA methylation and gene expression in muscle tissue. In total, 533 individual CpG sites were differentially methylated inpretraining muscle tissue of nonresponders vs. responders as depicted in the heatmap (A). B: Calculating the average methylation level for these533 differentially methylated CpG sites in muscle tissue based on the functional genome revealed lower DNA methylation in the promotor region forthese genes in nonresponders vs. responders. C: Calculating the average methylation level for these 533 differentially methylated CpG sites in muscletissue based on the CpG content and neighborhood content revealed lower DNAmethylation in the CpG island region, which is known to overlap withthe promotor regions. D: Manhattan plot of the P values calculated from the genome-wide CpG site methylation analysis in muscle tissue onall chromosomes (n = 533 differentially methylated CpG sites in total), and the inset shows the fold changes of the 12 CpG sites differentially meth-ylated in GSTM1 and GSTM5 on chromosome 1 for nonresponders vs. responders. E: CpG sites in the promotor region of genes related to glutathionemetabolism and insulin signaling were differentially methylated in muscle tissue of nonresponders vs. responders, with lower DNA methylation inthe promotor region of these genes in nonresponders vs. responders. F: Transcript levels (FPKM values from RNAseq data) for genes involved inglutathione metabolism, mitochondrial metabolism, and insulin signaling were different in muscle tissue of nonresponders vs. responders, withhigher expressions in nonresponders vs. responders. Data are mean 6 SD, where appropriate. #P , 0.050 compared with nonresponders.

2250 Exercise Response Variation in Type 2 Diabetes Diabetes Care Volume 41, October 2018

CONCLUSIONS

We used training response variation inmuscle mitochondrial function in vivo(measured by PCr recovery rate) after10 weeks of aerobic training to classifyindividuals with type 2 diabetes as non-responders or responders. Individualsclassified as nonresponders did not im-prove insulin sensitivity and worsenedglycemic control with training, whereasresponders improved insulin sensitivity.The training response variation in musclemitochondrial function in vivo wasmarked by a distinct pretraining molec-ular pattern in muscle tissue and inmyogenic progenitor cells (HSkMCs) ofnonresponders compared with respond-ers. DNA methylation and RNA expres-sion patterns showed elevations inantioxidant defense, insulin signaling,

and mitochondrial metabolism in non-responders, which were reflected in vivoby higher pretraining muscle mitochon-drial function and insulin sensitivity inthese same individuals.

On average, exercise-training inter-ventions improve muscle mitochondrialfunction (25,26) and even restore musclemitochondrial function in individualswith type 2 diabetes to levels observedin healthy individuals (4,27). Conclusionsabout the beneficial effects of exercisetraining are too often drawn based onaverage responses, and little is knownabout those individuals who do not re-spond favorably to exercise training.Here we identified individuals withtype 2 diabetes who did not improvemuscle mitochondrial function in vivowith training that correlated with a lack

of improvement in insulin sensitivity andworsened glycemic control. In a previousstudy, first-degree relatives of individualswith type 2 diabetes who did not improvemuscle ATP synthesis rates in vivo afterthree bouts of exercise did not improveinsulin sensitivity, whereas those whoimproved in vivo muscle ATP synthesisrates also improved insulin sensitivity (8).Evidence for a cause-and-effect rela-tionship between muscle mitochondrialfunction and insulin resistance is stilllimited (25). However, it is well docu-mented that, on average, improve-ments in muscle mitochondrial functioninduced by exercise training in individu-als with type 2 diabetes are paralleledby improvements in insulin sensitivity(4,25–27). Our data expand these pre-vious findings by demonstrating that the

Figure 3—Differences in pretraining DNAmethylation were preserved in myogenic progenitor cells. Some differences in the methylation of CpG sitesin muscle tissue were preserved in myogenic progenitor cells (HSkMCs). In total, 420 CpG sites were differentially methylated only in muscle tissue(red), 378 CpG sites were differentially methylated only in HSkMCs (yellow), and 113 differentially methylated CpG sites were in common betweenmuscle tissue and HSkMCs (orange) of nonresponders vs. responders (A). B: Calculating the average methylation level for these 491 differentiallymethylated CpG sites in HSkMCs based on the functional genome revealed lower methylation in the promotor region for these genes innonresponders vs. responders. C: Calculating the average methylation level for these 491 differentially methylated CpG sites in HSkMCs based onthe CpG content and neighborhood content revealed lower methylation in the CpG island region, which is known to overlap with the promotorregions. D: Manhattan plot of the P values calculated from the genome-wide CpG site methylation analysis in HSkMCs (n = 491 sites across allchromosomes). E: As in muscle tissue, CpG sites in the promotor regions of genes related to glutathione metabolism, mitochondrial metabolism,and insulin signaling were differentially methylated in HSkMCs in nonresponders vs. responders, with lower methylation in the promotor region ofthese genes in nonresponders vs. responders. F: Transcript levels (FPKM values from RNAseq data) for genes involved in glutathione metab-olism, mitochondrial metabolism, and insulin signaling in HSkMCs of nonresponders vs. responders. Data are mean 6 SD, where appropriate.#P , 0.050 compared with nonresponders.

care.diabetesjournals.org Stephens and Associates 2251

variation in training response in musclemitochondrial function in vivo also cor-responds to the variation in training re-sponse in insulin sensitivity in individualswith type 2 diabetes.Interest in the topic of biological in-

dividuality has a long history (28). Al-though the emphasis has been onnutrition and nutrients, its basic conceptis relevant for adaptations to exercisetraining as well. The concept of trainingresponse variation was proposed 35 yearsago (29) and has been traditionallydefined as the interindividual variationin changes in VO2max with exercise train-ing. The most comprehensive data ontraining response variation related toVO2max comes from the HERITAGE FamilyStudy (30). In this study, exercise trainingincreased VO2max for the groups, but in-dividual changes were highly variable,such that some individuals did not im-prove VO2max with training. Low VO2max isassociated with metabolic risk factorspresent in type 2 diabetes, such as ab-errant muscle mitochondrial function,insulin resistance, and hyperglycemia(2,31). The relationship between training-induced changes in VO2max and training-induced changes in metabolic risk factors,however, is not completely clear. Somestudies showed associations betweentraining-induced changes in VO2max andmetabolic risk factors (32), whereasothers failed to show this association(33,34). In the current study, we foundthat nonresponders had similar improve-ments inVO2peak comparedwith respond-ers. In a previous exercise training study(33), only one in every three individualswith type 2 diabetes showed improvedVO2max, but all showed decreased HbA1clevels. Thus, it seems that an associationbetween training responses in VO2max andmetabolic risk factors is not apparent forevery individual.Moreover, our data sug-gest that training response variation re-lated to metabolic risk factors seems tobe independent of the training responsevariation related to cardiorespiratory fit-ness. Identifying the training responsevariation related tometabolic risk factorsin individuals with type 2 diabetes couldinform future combinations of lifestyleand pharmaceutical therapeutic strate-gies aimed at resolving metabolic com-plications of type 2 diabetes (35).Nonresponders were characterized

with better in vivo muscle mitochondrialfunction and higher insulin sensitivity

pretraining than responders. Thereis some previous evidence suggestingthat individuals with a favorable meta-bolic profile benefit less from exerciseor lifestyle interventions than individualswith a higher metabolic burden (32).For example, first-degree relatives ofindividuals with type 2 diabetes whodid not improve in vivo muscle ATPsynthesis rates after three bouts of ex-ercise were characterized with highermuscle ATP synthesis rates at baselinecompared with individuals who im-proved muscle ATP synthesis rates (8).On the other hand, unfavorable glucosehomeostasis before training bluntedtraining-induced improvements in 2-hpost–oral glucose tolerance test bloodglucose levels (36) and insulin sensitivity(37), suggesting that individuals with anunfavorablemetabolic profile at baselinebenefit less from exercise interventions.These conflicting results might be ex-plained by a ceiling effect or a regressionto the mean for some variables, differentpopulations, and different study designs.These findings, together with the datapresented herein, are just the tip of theiceberg, and more research is warrantedin this exciting area of investigation tobetter understand the impact of prein-tervention metabolic profiles on the ex-ercise response in hopes of developingbiomarkers as potential predictors of theexercise response. High plasma free fattyacid levels have also been linked toimpaired muscle mitochondrial functionand insulin resistance (38,39); however,fasting plasma free fatty acid levels werenot different between groups, nor didtraining influence plasma free fatty acidlevels, suggesting no direct relationshipbetween training response variability formuscle mitochondrial function in vivoand plasma free fatty acids levels.

Individuals within a family respondmore similarly to exercise training interms of metabolic risk factors than thosebetween families, suggesting that hered-itary factors contribute to the trainingresponse variability (6,11,12). A singlemuscle contraction induces hypomethy-lation in the promotor regiondfollowedby transcriptional upregulationdofgenes involved in mitochondrial metab-olism, highlighting the potential epige-nomic control of exercise trainingresponses (14).We previously describeddistinct baseline (pretraining) transcrip-tional profiles in muscle tissue between

individuals who did (responders) or didnot (nonresponders) have improvedHbA1c levels, muscle mitochondrial con-tent, body fat, and BMI after 9 monthsof supervised exercise training (10). Inthe current study, we identified distinctdifferences in the pretraining muscletissue DNA methylation and transcrip-tional profiles of nonresponders versusresponders, and differences in DNA meth-ylation were specifically located withinthe promotor regions. Differences in DNAmethylation in muscle tissue of non-responders compared with responderslargely overlapped with differences inDNA methylation in myogenic progen-itor cells (HSkMCs) of these same in-dividuals, demonstrating that theseepigenomic differences are intrinsic tomuscle. Our study is the first to investigateDNA methylation profiles in muscle tissueand myogenic progenitor cells in non-responders and responders to exercisetraining and suggests that the trainingresponse variations observed in theepigenomic profiles of muscle tissueare cell autonomous.

The most robust differentially meth-ylated and expressed gene setwas linkedto glutathione metabolism, with lowermethylation and higher mRNA expres-sions of these genes in nonresponderscompared with responders. Differencesin glutathione metabolism in muscletissue have been described in high andlow responders to a diet intervention(40). Glutathionemetabolism is essentialfor several antioxidant defense systems(41), and thus higher expressions ofglutathione-metabolism-related genessuggests higher antioxidant activity inmuscle. Training-induced oxidative stressresponses seem to be important for theactivation of signaling pathways relatedto the beneficial effects of exercise (42).Therefore, higher baseline antioxidantactivity in the muscle of nonrespondersmight have contributed to the impairedtraining response inmusclemitochondrialfunction and insulin sensitivity. Enrich-ment analyses also revealed pathwayslinked to mitochondrial metabolism andinsulin signaling in muscle tissue of non-responders compared with respondersprior to exercise training. Recent find-ings have shown that mitochondria areregulated by and can in turn regulateepigenomic mechanisms via mitonu-clear communication (43). Changes inmitochondrial function, for example,

2252 Exercise Response Variation in Type 2 Diabetes Diabetes Care Volume 41, October 2018

influenced the activity of jmjd-1.2/PHF8and jmjd-3.1/JMJD3, two histone lysinedemethylases that are part of an epige-nomic mechanism that regulates lifespan (44). Our data suggest a connectionbetween mitochondrial function and ep-igenomic modifications in the contextof exercise training in humans. Upregu-lation of epigenomic signals involvingmitochondrial metabolism and insulinsignaling in nonresponders (vs. respond-ers) at baseline aligned with higher mi-tochondrial function in vivo and insulinsensitivity in nonresponders (vs. respond-ers) at baseline. Baseline epigenomicmechanisms in muscle may thereforeserve as a key regulator of pathways thatcan influence an individual’s metabolicexercise-training response.In conclusion, training response vari-

ation in muscle mitochondrial functionin vivo corresponds to training responsevariation for insulin sensitivity and gly-cemic control, but is independent oftraining response variation for cardiore-spiratory fitness. A distinct pretrainingmolecular pattern in muscle tissue char-acterizes the training response varia-tion in mitochondrial function in vivo,which includes differential enrichment ofpathways linked to antioxidant defense,insulin signaling, and mitochondrial me-tabolism. These differences in pretrain-ing epigenomic profiles were maintainedin myogenic progenitor cells, suggestingthat the training response variation ob-served in the epigenomic profiles of mus-cle tissue was cell autonomous. Our dataprovide new evidence to potentially shiftthe diabetes treatment paradigm by iden-tifying individuals that do not reap met-abolic benefits from exercise training,such that supplemental treatment op-tions can be designed.

Acknowledgments. The authors thank thevolunteers from the study for their participation.The authors also thank the recruitment team andall clinical personnel and all laboratory personnelat the Translational Research Institute for Me-tabolism and Diabetes.Funding. This work is supported by a grant fromthe American Diabetes Association (#7-13-JF-53).Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. N.A.S. performed ex-periments and data analyses and criticallyreviewed and edited the manuscript. B.B. per-formed experiments and data analyses andwrote the manuscript. A.M.E. and F.Y. performeddata analyses and critically reviewed and editedthe manuscript. H.H.C. performed experiments

and critically reviewed and edited the manu-script. C.M., B.H.G., and R.E.P. designed thestudy, analyzed the data, and critically reviewedand edited the manuscript. S.R.S. aided in studydesign and critically reviewed and edited themanuscript. L.M.S. designed the study, per-formed experiments, analyzed the data, andcritically reviewed and edited the manuscript.L.M.S. is the guarantor of this work and, as such,had full access to all the data in the study andtakes responsibility for the integrity of the dataand the accuracy of the data analysis.

References1. Boyle JP, Thompson TJ, Gregg EW, Barker LE,Williamson DF. Projection of the year 2050 bur-den of diabetes in the US adult population:dynamic modeling of incidence, mortality, andprediabetes prevalence. Popul Health Metr2010;8:292. Bajpeyi S, Pasarica M, Moro C, et al. Skeletalmuscle mitochondrial capacity and insulin re-sistance in type 2 diabetes. J Clin EndocrinolMetab 2011;96:1160–11683. Kelley DE, He J, Menshikova EV, Ritov VB.Dysfunction of mitochondria in human skeletalmuscle in type 2 diabetes. Diabetes 2002;51:2944–29504. Meex RC, Schrauwen-Hinderling VB, Moonen-Kornips E, et al. Restoration of muscle mitochon-drial function andmetabolic flexibility in type 2diabetes by exercise training is paralleled byincreased myocellular fat storage and im-proved insulin sensitivity. Diabetes 2010;59:572–5795. Stephens NA, Sparks LM. Resistance to thebeneficial effects of exercise in type 2 diabetes:are some individuals programmed to fail? J ClinEndocrinol Metab 2015;100:43–526. Boule NG, Weisnagel SJ, Lakka TA, et al.;HERITAGE Family Study. Effects of exercise train-ing on glucose homeostasis: the HERITAGEFamily Study. Diabetes Care 2005;28:108–1147. Hagberg JM, Jenkins NT, Spangenburg E.Exercise training, genetics and type 2 diabetes-related phenotypes. Acta Physiol (Oxf) 2012;205:456–4718. Kacerovsky-Bielesz G, Chmelik M, Ling C, et al.Short-term exercise training does not stimulateskeletal muscle ATP synthesis in relatives ofhumans with type 2 diabetes. Diabetes 2009;58:1333–13419. Kacerovsky-Bielesz G, Kacerovsky M, ChmelikM, et al. A single nucleotide polymorphismassociates with the response of muscle ATPsynthesis to long-term exercise training in rel-atives of type 2 diabetic humans. Diabetes Care2012;35:350–35710. Stephens NA, Xie H, Johannsen NM, ChurchTS, Smith SR, Sparks LM. A transcriptional sig-nature of “exercise resistance” in skeletalmuscleof individuals with type 2 diabetes mellitus.Metabolism 2015;64:999–100411. AnP,Teran-GarciaM,RiceT,etal.;HERITAGEFamily Study. Genome-wide linkage scans forprediabetes phenotypes in response to 20weeksof endurance exercise training in non-diabeticwhites and blacks: the HERITAGE Family Study.Diabetologia 2005;48:1142–114912. Lakka TA, Lakka HM, Rankinen T, et al. Effectof exercise training onplasma levels of C-reactive

protein in healthy adults: the HERITAGE FamilyStudy. Eur Heart J 2005;26:2018–202513. Bohm A, Hoffmann C, Irmler M, et al. TGF-bcontributes to impaired exercise response bysuppression of mitochondrial key regulators inskeletal muscle. Diabetes 2016;65:2849–286114. Barres R, Yan J, Egan B, et al. Acute exerciseremodels promoter methylation in human skel-etal muscle. Cell Metab 2012;15:405–41115. Toledo FG, Menshikova EV, Ritov VB, et al.Effects of physical activity and weight loss onskeletal muscle mitochondria and relationshipwith glucose control in type 2 diabetes. Diabetes2007;56:2142–214716. Toledo FG, Menshikova EV, Azuma K, et al.Mitochondrial capacity in skeletal muscle is notstimulated by weight loss despite increases ininsulin action and decreases in intramyocellularlipid content. Diabetes 2008;57:987–99417. Moro C, Galgani JE, Luu L, et al. Influence ofgender, obesity, and muscle lipase activity onintramyocellular lipids in sedentary individuals.J Clin Endocrinol Metab 2009;94:3440–344718. Brouwers B, Schrauwen-Hinderling VB,Jelenik T, et al. Metabolic disturbances of non-alcoholic fatty liver resemble the alterationstypical for type 2 diabetes. Clin Sci (Lond) 2017;131:1905–191719. Church TS, Blair SN, Cocreham S, et al. Effectsof aerobic and resistance training on hemoglo-bin A1c levels in patients with type 2 diabetes:a randomized controlled trial. JAMA 2010;304:2253–226220. Sparks LM, Johannsen NM, Church TS, et al.Nine months of combined training improvesex vivo skeletalmusclemetabolism in individualswith type 2 diabetes. J Clin Endocrinol Metab2013;98:1694–170221. Bergstrom J. Percutaneous needle biopsyof skeletal muscle in physiological and clinicalresearch. Scand J Clin Lab Invest 1975;35:609–61622. Sparks LM, Moro C, Ukropcova B, et al.Remodeling lipid metabolism and improving in-sulin responsiveness in human primary myo-tubes. PLoS One 2011;6:e2106823. Petersen KF, Shulman GI. Pathogenesis ofskeletal muscle insulin resistance in type 2 di-abetes mellitus. Am J Cardiol 2002;90(5A):11G–18G24. Ronn T, Volkov P, Davegardh C, et al. Asix months exercise intervention influencesthe genome-wide DNA methylation pattern inhuman adipose tissue. PLoS Genet 2013;9:e100357225. Hesselink MK, Schrauwen-Hinderling V,Schrauwen P. Skeletal muscle mitochondriaas a target to prevent or treat type 2 diabetesmellitus. Nat Rev Endocrinol 2016;12:633–64526. Szendroedi J, Phielix E, RodenM. The role ofmitochondria in insulin resistance and type 2diabetes mellitus. Nat Rev Endocrinol 2011;8:92–10327. Phielix E, Meex R, Moonen-Kornips E,Hesselink MK, Schrauwen P. Exercise trainingincreases mitochondrial content and ex vivomitochondrial function similarly in patientswith type 2 diabetes and in control individuals.Diabetologia 2010;53:1714–172128. Burbridge D. Francis Galton on twins, he-redity and social class. Br J Hist Sci 2001;34:323–340

care.diabetesjournals.org Stephens and Associates 2253

29. Bouchard C, Sarzynski MA, Rice TK, et al.Genomic predictors of the maximal O2 uptakeresponse to standardized exercise training pro-grams. J Appl Physiol (1985) 2011;110:1160–117030. Bouchard C, An P, Rice T, et al. Familialaggregation of VO(2max) response to exercisetraining: results from the HERITAGE FamilyStudy. J Appl Physiol (1985) 1999;87:1003–100831. Solomon TP, Malin SK, Karstoft K, et al.Association between cardiorespiratory fitnessand the determinants of glycemic control acrossthe entire glucose tolerance continuum. Diabe-tes Care 2015;38:921–92932. Bohm A, Weigert C, Staiger H, Haring HU.Exercise and diabetes: relevance and causes forresponsevariability. Endocrine2016;51:390–40133. Pandey A, Swift DL, McGuire DK, et al.Metabolic effects of exercise training amongfitness-nonresponsive patients with type 2 di-abetes: the HART-D Study. Diabetes Care 2015;38:1494–150134. Brennan AM, Lam M, Stotz P, Hudson R,Ross R. Exercise-induced improvement in insulin

sensitivity is not mediated by change in cardio-respiratory fitness. Diabetes Care 2014;37:e95–e9735. Sparks LM. Exercise training response het-erogeneity: physiological andmolecular insights.Diabetologia 2017;60:2329–233636. Solomon TP, Malin SK, Karstoft K, Haus JM,Kirwan JP. The influence of hyperglycemia on thetherapeutic effect of exercise on glycemic controlin patients with type 2 diabetes mellitus. JAMAIntern Med 2013;173:1834–183637. AbouAssi H, Slentz CA, Mikus CR, et al. Theeffects of aerobic, resistance, and combinationtraining on insulin sensitivity and secretion inoverweight adults from STRRIDE AT/RT: a ran-domized trial. J Appl Physiol (1985) 2015;118:1474–148238. Brehm A, Krssak M, Schmid AI, Nowotny P,Waldhausl W, Roden M. Acute elevation ofplasma lipids does not affect ATP synthesis inhuman skeletal muscle. Am J Physiol EndocrinolMetab 2010;299:E33–E3839. Brehm A, Krssak M, Schmid AI, NowotnyP, Waldhausl W, Roden M. Increased lipid

availability impairs insulin-stimulated ATP syn-thesis in human skeletal muscle. Diabetes 2006;55:136–14040. Thrush AB, Zhang R, Chen W, et al. Lowermitochondrial proton leak and decreased gluta-thione redox in primary muscle cells of obesediet-resistant versus diet-sensitive humans.J Clin Endocrinol Metab 2014;99:4223–423041. Mailloux RJ, McBride SL, Harper ME. Un-earthing the secrets of mitochondrial ROS andglutathione in bioenergetics. Trends Biochem Sci2013;38:592–60242. Powers SK, Jackson MJ. Exercise-inducedoxidative stress: cellular mechanisms and im-pact on muscle force production. Physiol Rev2008;88:1243–127643. Matilainen O, Quiros PM, Auwerx J. Mito-chondria and epigenetics - crosstalk in homeo-stasis and stress. Trends Cell Biol 2017;27:453–46344. Merkwirth C, Jovaisaite V, Durieux J, et al.Two conserved histone demethylases regu-late mitochondrial stress-induced longevity. Cell2016;165:1209–1223

2254 Exercise Response Variation in Type 2 Diabetes Diabetes Care Volume 41, October 2018