Miro1 mediates skeletal muscle insulin resistance in type 2 diabetes

Abstract

Imbalanced skeletal muscle mitochondrial dynamics contribute to the onset and progression of type 2 diabetes (T2D) by mechanisms that remain incompletely understood. Here, we examined the role of Mitochondrial Rho GTPase 1 (Miro1), an outer mitochondrial membrane enzyme, in regulating skeletal muscle insulin action and glucose homeostasis in type 2 diabetes (T2D). In a cross-sectional clinical study, we show that Miro1 accumulates in the skeletal muscle of patients with obesity and T2D, which was further replicated in mouse models of T2D. This accumulation is explained by inadequate insulin-mediated interaction between AKT and Miro1 at the outer mitochondrial membrane. In a randomized controlled trial (NCT02977442), we show that exercise training relieves skeletal muscle Miro1 accumulation while improving insulin sensitivity in patients with T2D. Muscle-specific Miro1 deletion improves oxidative capacity and insulin sensitivity in mice and cells. Taken together, these data support a key regulatory role for skeletal muscle Miro1 in the pathophysiology of T2D. ClinicalTrials.gov Identifier: NCT02977442

Insulin resistance in skeletal muscle contributes to type 2 diabetes (T2D), but its causes remain unclear. We identify the mitochondrial protein Miro1 as a key regulator of muscle insulin sensitivity, providing new insights into T2D development.

Introduction

Type 2 diabetes (T2D) has emerged as one of the major public health and economic burdens of the 21st century, affecting more than 37 million people in the United States of America and more than 500 million people globally (~10.5%)¹. As these shocking numbers continue to rise, the cost of caring for patients with diabetes and its complications is placing enormous strain on the global economy, totaling nearly $1 trillion dollars in economic impact², despite the development of nearly a dozen new medications for disease management. As the scale of the problem grows exponentially, our fundamental understanding of the cellular and molecular mechanisms of T2D remains incomplete, limiting our ability to develop effective, targeted medications that address the underlying pathophysiology rather than the disease’s symptoms. T2D arises when insulin secretion is insufficient to compensate for progressive, sustained peripheral insulin resistance, leading to a systemic rise in glucose concentrations. Skeletal muscle insulin resistance is a primary and rate-limiting defect in the onset and progression of T2D, as it accounts for the majority (~85%) of whole-body glucose disposal at rest³. Obesity is the primary cause of insulin resistance and is present in >90% of patients with T2D^4,5. As such, there is a critical need to identify the cellular and molecular factors mediating the progression of obesity-related skeletal muscle insulin resistance to overt T2D.

Mitochondria are dynamic organelles that maintain networking through processes of fission, fusion, biogenesis, and mitophagy⁶. Maintenance of these networks is required to sustain mitochondrial function on multiple levels, including energy production and stress responsiveness⁷. Mitochondrial Rho GTPase 1 (Miro1) is an evolutionarily conserved atypical GTPase residing on the outer mitochondrial membrane with calcium-binding motifs and a cytoplasmic-facing C-terminal transmembrane domain⁸. Early investigations of the Miro proteins revealed influential roles in mitochondrial morphology and movement in cells with compartmentalized transport requirements, such as neurons^9–11. In contrast, the mitochondrial network configuration in muscle-specific Miro1-deficient flies was entirely intact despite a complete loss of climbing ability¹², indicating previously unidentified roles in cell signaling and metabolism. We and others have shown that skeletal muscle lipid accumulation directly contributes to the onset and progression of insulin resistance by early and sustained activation of mitochondrial fission mediated by dynamin-related protein 1 (DRP1)^13–15. DRP1 is critical to normal mitochondrial function, structure, and turnover. Patients with T2D exhibit numerous mitochondrial defects, including decreased mitochondrial content and size and altered cristae structure^16–18. Miro1 is required to anchor DRP1, promoting mitochondrial fission and restricting mitochondrial transport¹⁹. However, the role and regulation of skeletal muscle Miro1 and its relationship to obesity-related type 2 diabetes remain entirely unknown.

Here, we examined the role and regulation of skeletal muscle Miro1 in patients and mice with obesity and T2D. We discovered that Miro1 accumulates in the skeletal muscle of mice and patients with obesity and hyperglycemia, and this accumulation is caused, in part, by defective mitochondrial protein kinase B (AKT) signaling and, subsequently, by inadequate AKT-Miro1 protein interaction. Improving skeletal muscle insulin sensitivity through exercise training lowered Miro1 expression while synergistically enhancing skeletal muscle oxidative capacity and restricting lipid accrual in patients with T2D. Congruently, skeletal muscle-specific deletion of Miro1 in mice or C2C12 cells enhanced mitochondrial oxidative capacity, glucose transporter 4 (GLUT4) exocytosis, and glucose disposal. Taken together, these data support a  role for Miro1 in the regulation of glucose homeostasis and T2D.

Results

Miro1 accumulation is associated with insulin resistance in patients and mice with obesity and type 2 diabetes

DRP1- and PINK1-mediated regulation of mitochondrial quality control has been shown to regulate Miro1 stability²⁰. In the context of Parkinson’s disease, defects in mitophagy lead to the accumulation of Miro1 in skin fibroblasts²¹. We previously showed that PINK1 activation contributes to lipid-induced insulin resistance in humans¹³. Based upon these findings, we sought to examine skeletal muscle Miro1 expression across the glucoregulatory spectrum. To achieve this, we conducted a retrospective analysis of patients previously assigned to one of three groups based upon the following: healthy weight (HW: BMI 18–25 kg/m² with no prior history of T2D or use of glucose-lowering medications), obesity without T2D (Ob: BMI 25–50 kg/m² without T2D), or obesity with T2D (T2D: BMI 25–50 kg/m² with T2D)²². Following screening, all patients underwent a 3-day inpatient stay with a fasted skeletal muscle biopsy followed by a single-stage, 180 min hyperinsulinemic-euglycemic clamp (Fig. 1A). Patients with T2D had elevated HbA_1c, and impaired glucose disposal and insulin sensitivity compared to HW and Ob (Fig. 1B, C, and Table 1). Furthermore, skeletal muscle Miro1 protein accumulation was observed only in patients with T2D (Fig. 1D, E). To further investigate the relationship between Miro1 and insulin sensitivity, we performed a correlational analysis of Miro1 expression with insulin sensitivity measured during hyperinsulinemic-euglycemic clamps in this cohort (Fig. S1). Miro1 has a modest but significant negative correlation with insulin sensitivity, which is largely driven by the T2D group.

Fig. 1. Miro1 accumulation is associated with insulin resistance in patients and mice with obesity and type 2 diabetes.

A Overview of human cross-sectional study design. B, C Fasting HbA_1c and glucose disposal rate relative to insulin during hyperinsulinemic-euglycemic clamps. D, E Western blot and quantification of fold change relative to HW for Miro1 expression in mitochondria isolated from human skeletal muscle tissue mitochondria from HW, Ob, and patients with T2D (n = 12/group). F Study design of diet-induced hyperglycemia in mice. G, H Endpoint body weight and fasting glucose of mice maintained on a low (FFC_low) or high fat, fructose, and cholesterol diet (FFC_high) for 28 weeks (n = 8/group). I, J Quantification of fold change relative to FFC_low for Miro1 expression in skeletal muscle from mice fed a FFC_low or FFC_high diet (n = 4/group) and western blot. K Study design of genetically-induced hyperglycemia in mice. L, M Endpoint body weight and fasting glucose of wild-type C57BL/6 (WT) or lethal yellow (A^y) mice (n = 8/group). N, O Quantification of fold change relative to WT for Miro1 expression in skeletal muscle from WT and A^y mice (n = 4/group) and western blot. See also Fig. S1. Data are represented as mean ± SEM, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001, and were statistically compared by Kruskal–Wallis test with Dunn’s multiple comparisons (B, E), one-way ANOVA (C), or two-tailed Student’s t test (G–I and L–N). Exact p-values are as follows: B HW vs. Ob = 0.8567, HW vs. T2D < 0.0001, Ob vs. T2D = 0.0008; C HW vs. Ob <0.0001, HW vs. T2D < 0.0001, Ob vs. T2D = 0.1170; E HW vs. Ob = 0.7117, HW vs. T2D = 0.0002, Ob vs. T2D = 0.0163; G 0.0002; H 0.0002; I 0.0011; L 0.0013; M 0.0004; N 0.0024. Source data are provided as a Source Data file.

Table 1.

Cross-sectional cohort participant characteristics

				P value
Characteristics	HW (N = 12)	Ob (N = 12)	T2D (N = 12)	HW vs. Ob	HW vs. T2D	Ob vs. T2D
Age (y)	28.0 ± 8.3	34.9 ± 7.5	50.8 ± 8.5	0.1066	<0.0001	<0.0001
Sex (M, %)	5 (41.7)	2 (16.7)	7 (58.3)	0.4137	0.6711	0.097
Body weight (kg)	66.4 ± 10.7	96.8 ± 14.4	104.3 ± 17.3	<0.0001	<0.0001	0.4261
BMI (kg/m2)	22.4 ± 1.7	31.6 ± 3.8	36.5 ± 6.5	<0.0001	<0.0001	0.0297
Triglyceride (mg/dL)	94.6 ± 89.4	132.6 ± 83.2	142.2 ± 73.7	0.5026	0.3448	0.9563
Total cholesterol (mg/dL)	178.1 ± 42.4	184.7 ± 35.1	178.1 ± 39.8	0.9112	>0.9999	0.9112
HDL cholesterol (mg/dL)	62.7 ± 16.2	43.9 ± 8.2	45.5 ± 10.4	0.0016	0.0039	0.944
LDL cholesterol (mg/dL)	96.7 ± 40.6	114.2 ± 19.0	104.1 ± 30.7	0.3687	0.8308	0.7137
VO2PEAK (mL/kg/min)	40.0 ± 10.4	30.5 ± 7.2	20.2 ± 3.6	0.0116	<0.0001	0.0062
VO2PEAK (mL/kg FFM/min)	58.1 ± 7.1	51.6 ± 8.1	34.9 ± 4.6	0.0881	<0.0001	<0.0001
Fasting glucose (mg/dL)	87.6 ± 5.6	86.5 ± 7.4	167.4 ± 51.9	0.9956	<0.0001	<0.0001
Fasting insulin (µU/mL)	6.7 ± 2.4	14.0 ± 8.3	17.6 ± 8.8	0.0444	0.0018	0.4209

Patient characteristics in individuals with a healthy weight (HW), obesity (Ob), or type 2 diabetes (T2D). Data are mean ± SD. BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; FFM, fat-free mass. Comparisons where p < 0.05 are represented in bold text.

To confirm these findings, we evaluated skeletal muscle Miro1 expression in various mouse models of T2D. First, we compared mice on a high-fat, high-fructose, and high-cholesterol (FFC_high) diet (Fig. 1F), which caused obesity, hyperglycemia, and, subsequently, skeletal muscle Miro1 accumulation, compared to mice on a low-fat, low-fructose, low-cholesterol (FFC_low) diet (Fig. 1G–J). To determine whether Miro1 accumulation in mice with obesity and hyperglycemia was solely attributable to diet, we compared Miro1 expression in two monogenic mouse models of T2D. We observed that A^y and Lepr^db/db mice maintained on a low-fat diet exhibited skeletal muscle Miro1 accumulation similar to that of FFC_high mice (Figs. 1K–O and S2A–C). Taken together, these data support skeletal muscle Miro1 accumulation in humans and mice with obesity-related T2D.

Defective mitochondrial AKT translocation mediates Miro1 accumulation

We next sought to determine the etiology of skeletal muscle Miro1 accumulation in the context of hyperglycemia. Given our observations that Miro1 selectively accumulates in the background of T2D, we hypothesized that altered skeletal muscle glucoregulatory function may cause Miro1 accumulation. Since insulin is the primary endocrine mediator of intracellular glucose disposal in skeletal muscle, we first evaluated the extent to which exogenous insulin exposure affected Miro1 expression. Acute exposure of C2C12 cells to insulin decreased Miro1 expression, indicating a potential link between insulin receptor signaling and regulation of Miro1 status on the outer mitochondrial membrane (Fig. 2A, B). In the canonical insulin signaling cascade, AKT phosphorylation in the cytosol promotes glucose uptake by facilitating GLUT4 translocation to the plasma membrane²³. However, it has been previously shown that AKT phosphorylation at the insulin-dependent threonine 308 residue (AKT^Thr308) rapidly accumulates in mitochondria²⁴. We evaluated these observations by measuring AKT^Thr308 phosphorylation in isolated skeletal muscle mitochondria following in vivo insulin infusion. Patients with T2D exhibited a near-complete absence of mitochondrial AKT^Thr308 phosphorylation compared with patients with obesity alone (Fig. 2C, D). Based on our observations that mitochondrial Miro1 accumulation is associated with skeletal muscle insulin resistance, and that impaired AKT translocation in T2D promotes mitochondrial Miro1 retention, we hypothesized that aberrant mitochondrial confinement of AKT in T2D contributes to Miro1 accumulation. To test this, we performed immunofluorescence analysis of C2C12 cells, in which co-localization of AKT with Miro1 was relatively low under basal conditions, increasing to ~60% in the presence of insulin (Figs. 2E, F and S3). To further support this, we performed co-immunoprecipitation with mitochondrial AKT and Miro1 in mouse gastrocnemius to determine whether the proteins interact directly and whether this interaction is altered by insulin. We observed that under basal conditions, small but measurable amounts of Miro1 were bound to AKT (Fig. 2G–I), presumably reflecting the lower mitochondrial AKT levels observed in the absence of insulin. In the presence of insulin, there was a significant increase in the interaction between Miro1 and mitochondrial AKT, largely attributable to increased AKT expression. We next sought to determine if mitochondrial AKT translocation was required to restrict Miro1 accumulation. To address this, C2C12 cells were exposed to MK-2206, a highly selective inhibitor of pan-Akt activity²⁵, in the presence or absence of insulin. We observed that in the absence of insulin, MK-2206 did not alter Miro1 expression, whereas in the presence of insulin, MK-2206 inhibition of AKT caused rapid Miro1 accumulation (Fig. 2J–L). Taken together, these data indicate that Miro1 accumulation in T2D is mediated by defective insulin-stimulated mitochondrial AKT translocation.

Fig. 2. Defective mitochondrial AKT translocation mediates Miro1 accumulation.

A, B Western blot and quantification of fold change relative to vehicle for AKT and Miro1 from C2C12 myoblasts with and without 1 μM insulin treatment (n = 3 biological replicates/condition). C, D Western blot and quantification of fold change relative to healthy control patients for mitochondrial AKT from human skeletal muscle mitochondria (n = 4/group). E Immunofluorescence staining of AKT (red), Miro1 (yellow), DAPI (blue), and the resulting merged stack in C2C12 myoblasts treated with and without insulin. Scale = 20μm. See also Fig. S3. F Quantification of Mander’s coefficient for co-localization of Miro1 to AKT (n = 10 biological replicates/condition). G Study design of coimmunoprecipitation in ex vivo insulin-treated gastrocnemius muscle mitochondria from mice fed a high-fat diet (HFD). Created in BioRender. Heintz, E. (2026) (https://BioRender.com/7pxx79n) is licensed under CC BY 4.0. H, I Western blot and quantification of fold change relative to vehicle for gastrocnemius mitochondrial Miro1 co-expressed with AKT with and without ex vivo insulin treatment in mice fed a HFD (n = 7 biological replicates/condition). J–L Western blot and quantification of fold change relative to a non-treatment control for AKT and Miro1 in C2C12 myoblasts with and without insulin and MK-2206 treatment (n = 3 biological replicates/condition). Data are represented as mean ± SEM, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001, and were statistically compared by two-tailed Student’s t test (B, D, and I), two-tailed Mann–Whitney test (F), or one-way ANOVA with Tukey’s multiple comparisons test (K and L). Exact p-values are as follows: B pAKT = 0.0151, Miro1 = 0.0182; D 0.0200; F 0.0005; I 0.0123; K Veh vs. +Ins = 0.1163, Veh vs. +MK2206 = 0.0007, Veh vs. +Ins/MK2206 = 0.0006, +Ins vs. +MK2206 < 0.0001, +Ins vs. +Ins/MK2206 < 0.0001, +MK2206 vs. +Ins/MK2206 = 0.9983; L Veh vs. +Ins = 0.5562, Veh vs. +MK2206 = 0.9802, Veh vs. +Ins/MK2206 = 0.0001, +Ins vs. +MK2206 = 0.7742, +Ins vs. +Ins/MK2206 < 0.0001, +MK2206 vs. +Ins/MK2206 < 0.0001. For D, the p-value was adjusted for unequal variance by Welch’s correction. Source data are provided as a Source Data file.

Exercise training relieves skeletal muscle mitochondrial hyper-fragmentation and Miro1 aggregation in patients with obesity and type 2 diabetes

Given our observations that Miro1 selectively accumulates in skeletal muscle of patients with T2D, which is at least in part due to insufficient interaction between Miro1 and the insulin signaling pathway, we hypothesized that treatments that improve peripheral insulin sensitivity may lower Miro1 expression. Exercise training is a first-line lifestyle therapy for managing T2D²⁶ and is well known to resensitize skeletal muscle glucoregulatory and mitochondrial function²⁷. To address this, we conducted a parallel-arm, open-label, randomized controlled trial comparing exercise training to standard-of-care in patients with established obesity and T2D (Figs. 3 and 4A). Patients randomized to standard-of-care (Ctrl) continued with routine medical management for T2D for 12 weeks, while also receiving counseling on lifestyle management of the disease. Patients randomized to exercise training (Ex) underwent a 12-week aerobic training program consisting of moderate-to-vigorous intensity, steady-state treadmill walking/jogging for 5 days per week, 60 min per session. All patients underwent an inpatient stay consisting of a two-stage hyperinsulinemic-euglycemic clamp and skeletal muscle biopsies before and after the intervention period (Fig. 4A). Age, sex, body weight, diabetes duration, number of glucose-lowering medications, and key demographic variables were comparable between groups at baseline (Table S1). Ex increased VO_2peak and lean mass while reducing fat mass relative to Ctrl (Table 2). At the muscular level, ultrastructural analyses of the contractile apparatus revealed that exercise increased myofibrillar connectivity as evidenced by the greater number of sarcomere branches after 12 weeks (Fig. 4B, C). Additionally, the lipid droplet volume per cell significantly decreased following exercise training (Figs. 4D, E and S4A–F). Peripheral insulin sensitivity improved in Ex but not in Ctrl following the intervention (Fig. 4F), as did GLUT4 translocation to the skeletal muscle membrane (Fig. 4G, H). Ex lowered Miro1 expression, which remained unchanged in Ctrl (Fig. 4G, I). Ex reduced skeletal muscle DRP1 phosphorylation (Fig. 4G, I), as well as mediators responsible for recruiting DRP1 to the mitochondrial surface, such as mitochondrial outer membrane protein 1 (Mid51) and 2 (Mid49) (Fig. 4J, K). Since DRP1 is required for mitochondrial fission and we have previously observed improvements in mitochondrial elongation in response to exercise training²⁷, we assessed mitochondrial ultrastructure and morphology by transmission electron microscopy. At baseline, Ex and Ctrl skeletal muscle mitochondria were relatively small and punctate, with consistent distribution across the subsarcolemmal and intermyofibrillar regions (Fig. 4L). Following the intervention period, mitochondria remained largely punctate and fragmented in Ctrl, whereas Ex improved the sphericity and elongation of individual and networked mitochondria, resulting in a significant improvement in the fragmentation index (Fig. 4M). Based on our findings that exercise training improved mitochondrial dynamics, we explored changes in other protein mediators of skeletal muscle biogenesis and quality control. We did not observe increases in PGC1α, MFN1, MFN2, OPA1, PINK1, or Parkin expression in Ex compared to Ctrl following intervention (Fig. S5A, B). We observed decreased expression of LC3, which acts as a signaling protein to remove damaged mitochondria²⁸, in patients following Ex compared to Ctrl (Fig. 4J, K).

Fig. 3.

CONSORT diagram depicting participant flow through the randomized controlled trial.

Fig. 4. Exercise training relieves skeletal muscle mitochondrial hyper-fragmentation and Miro1 aggregation in patients with obesity and type 2 diabetes.

A Overview of randomized controlled trial study design. B, C Ultrastructural analyses of skeletal muscle contractile apparatus and quantification of myofibrillar connectivity based on percent of sarcomeres with a branch before and after exercise training (n = 5 biological replicates). Different colors represent individual sarcomeres. Scale bars represent 2 μm. D, E Ultrastructural analyses of skeletal muscle lipid droplets and quantification of lipid droplet content before and after exercise training (n = 4 biological replicates). Different colors represent individual lipid droplets. Scale bars represent 300 μm. See also Fig. S4. F Relative change from baseline in insulin sensitivity during hyperinsulinemic-euglycemic clamp in Ctrl and Ex groups (n = 8 Ctrl, 10 Ex). G–I Western blots and quantification from human skeletal muscle samples in Ctrl and Ex groups before and after 12-week intervention. Relative changes in GLUT4 expression in response to insulin relative to basal condition from skeletal muscle membrane fractionation before and after Ctrl or Ex (n = 4/group/condition). Relative changes from baseline in mitochondrial Miro1 (n = 4/group) and whole lysate phospho-DRP1 (n = 12 Ctrl, 8 Ex) expression in human skeletal muscle samples before and after 12-week intervention. J, K Western blots and quantification from human skeletal muscle samples in Ctrl and Ex groups before and after 12-week intervention. Relative changes from baseline in Mid49 (n = 12 Ctrl, 8 Ex), Mid51(n = 12 Ctrl, 8 Ex), and LC3 (n = 12 Ctrl, 7 Ex) expression before and after 12-week intervention. See also Fig. S5. L, M Electron microscopy images of human skeletal muscle samples from Ctrl and Ex patients before and after 12-week intervention (n = 4/timepoint/group) and mitochondrial fragmentation index quantified from all distinguishable mitochondria per field. Data are represented as mean and individual change (C and E), mean ± SEM (F, H, I, and K), or mean and quartiles (M), *p < 0.05, **p < 0.01, ****p < 0.0001, and were statistically compared by two-tailed Student’s t test (C, E, F, and I), two-way repeated measures ANOVA with Holm-Šídák’s multiple comparisons test (H), or Kruskal–Wallis ANOVA with Dunn’s multiple comparisons test (M), or two-tailed Welch’s t-test (K). Exact p-values are as follows: C 0.0250, E 0.0034, F 0.0077, H Time x Group effect <0.0001, Time effect <0.0001, Group effect <0.0001, Subject effect = 0.5784; I Miro1 = 0.0007, pDRP1Ser616 = 0.0405; K Mid49 = 0.0193, Mid51 = 0.0450, LC3 = 0.0128; M Ctrl Pre vs. Post <0.0001, Ctrl Pre vs. Ex Pre > 0.9999, Ctrl Pre vs. Ex Post <0.0001, Ctrl Post vs. Ex Pre <0.0001, Ctrl Post vs. Ex Post <0.0001, Ex Pre vs. Post <0.0001. For F, the p-value was adjusted for unequal variance by Welch’s correction. Source data are provided as a Source Data file.

Table 2.

Changes from baseline in control (Ctrl) and exercise (Ex) patients

	Ctrl	Ex	P value
Body weight (kg)	−0.77 ± 1.10	−3.26 ± 0.97	0.1060
BMI (kg/m2)	0.10 ± 0.49	−1.03 ± 0.44	0.1004
Fat mass (kg)	0.15 ± 0.35	−1.29 ± 0.32	0.0063
Lean mass (kg)	−0.41 ± 0.44	0.04 ± 0.40	0.4610
VO2PEAK (mL/kg/min)	−0.27 ± 0.67	2.28 ± 0.56	0.0081
Fasting glucose (mg/dL)	−13.3 ± 32.3	−12.5 ± 29.1	0.9552
Fasting insulin (µU/mL)	0.7 ± 4.4	−3.8 ± 6.0	0.0389

Data are shown as relative change from baseline (post-pre) ± standard error. P-value is (post-pre Ex) vs (post-pre Ctrl). BMI, body mass index. Comparisons where p < 0.05 are represented in bold text.

In line with differences in skeletal muscle mitochondrial structure and networking following exercise, we then investigated whether these improvements were associated with enhanced mitochondrial function. Using high-resolution respirometry to assess maximal oxidative phosphorylation (OXPHOS) and electron transfer (ET) capacity in isolated mitochondria from skeletal muscle of patients with T2D (Fig. 5A), we observed increases in maximal NADH (N)-, succinate (S)-, palmitoyl carnitine (F)-, and complex III (CIII)-linked OXPHOS and complex IV (CIV)-linked ET in Ex compared to Ctrl (Fig. 5B). The acceptor control ratio also trended higher following Ex (Fig. 5C), collectively indicating increases in maximal skeletal muscle OXPHOS and improved coupling of substrate oxidation to ATP phosphorylation in Ex compared to Ctrl. N-linked ET was also increased following Ex, despite similar S- and CIII-linked ET capacities between groups (Fig. S6). Based on the observed improvements in OXPHOS capacity following Ex, we wanted to determine if the changes were consistent with physiological improvements in respiratory control. To accomplish this, we employed the creatine kinase clamp²⁹, which enables evaluation of the conductivity of the respiratory system over a wide range of free energy states through external control of the extramitochondrial ATP:ADP ratios (Fig. 5D)²⁹. Consistent with our maximal OXPHOS assessments, Ex increased N- and S-linked conductance and V_max compared to Ctrl (Fig. 5E, F). Taken together, these data suggest that exercise training relieves skeletal muscle hyper-fragmentation and Miro1 accumulation while enhancing respiratory capacity and conductance in patients with T2D.

Fig. 5. Exercise training enhances skeletal muscle mitochondrial capacity and conductance in patients with obesity and type 2 diabetes.

A Overview of the experimental approach to evaluate mitochondrial respiratory capacity. Created in BioRender. Heintz, E. (2026) (https://BioRender.com/bnsgb9m) is licensed under CC BY 4.0. B, C Relative changes from baseline in maximal NADH-linked (N-linked), succinate-linked (S-linked), fatty acid-linked (F-linked), complex III-linked (CIII-linked), and complex IV-linked (CIV-linked) mitochondrial oxidation and acceptor control ratio (ACR) in Ctrl and Ex patients (n = 6 Ctrl, 7 Ex). See also Fig. S6. D Representative plot illustrating the change in free energy of ATP production (ΔG’ ATP) in response to phosphocreatine (PCr) titration using the bioenergetic creatine kinase clamp technique. E, F Relative changes from baseline in N-linked and S-linked mitochondrial conductance (n = 8 Ctrl, 7 Ex) and V_max (n = 4 Ctrl, 5 Ex). Data are represented as mean ± SEM. ^*p < 0.05, ^**p < 0.01 by two-tailed Student’s t test. Exact p-values are as follows: B N-linked = 0.0297, S-linked = 0.0052, F-linked = 0.0125, CIII-linked = 0.0068, CIV-linked = 0.0145; C 0.0853; E N-linked = 0.0059, S-linked = 0.0325; F N-linked = 0.0042, S-linked = 0.0402. Source data are provided as a Source Data file.

Miro1 negatively regulates skeletal muscle insulin action and oxidative function in mice with obesity and hyperglycemia

Since exercise promotes multi-organ, pleotropic effects on physiological function, we sought to mechanistically tease apart the contribution of skeletal muscle Miro1 to maintaining glucoregulatory function in vivo. To achieve this, we generated a skeletal-muscle-specific Miro1-deficient mouse model (Fig. S7A) by crossing homozygous Miro1-floxed mice with ACTA1-Cre mice, resulting in spatially controlled Cre-mediated deletion of Miro1 (Miro1^SkM−/−). Miro1^SkM−/− mice exhibited ~70% reduction in Miro1 protein and >90% reduction in gene expression, which was restricted to skeletal muscle tissues (Figs. 6A, B and S7B). Miro1^SkM−/− mice were fertile, viable, and did not display overt developmental defects or deficiencies (Fig. S7C). To determine effects on insulin sensitivity, male Miro1^SkM−/− and Miro1^fl/fl were placed on a high-fat diet (HFD) for 16 weeks, followed by evaluation of peripheral glucose disposal by intraperitoneal insulin tolerance testing, hyperinsulinemic-euglycemic clamps, and harvesting of muscle to evaluate insulin signaling and tissue glucose uptake (Fig. 6C). Body weight and food intake were similar in Miro1^SkM−/− mice relative to littermate controls (Fig. S8A–D), Fasting glucose, fasting insulin, and HOMA-IR were lower in Miro1^SkM−/− mice (Fig. S8E–G), and glucose clearance rates in response to intraperitoneal insulin injection were improved in Miro1^SkM−/− mice (Fig. 6D, E). Hyperinsulinemic-euglycemic clamps were performed on conscious, unrestrained mice to determine the effects of skeletal muscle Miro1 deletion on insulin sensitivity. Miro1^SkM−/− mice had higher glucose disposal rates during the clamp period and increased skeletal muscle uptake of ¹⁴C-2-deoxyglucose (Figs. 6F, G and S9). To evaluate the effects of skeletal muscle-specific Miro1 deletion on insulin signaling, gastrocnemius muscles were exposed to rapid-acting insulin ex vivo. Miro1^SkM−/− mice had increased AKT and AS160 signaling in skeletal muscle compared to Miro1^fl/fl mice (Fig. 6H–J). Membrane-bound GLUT4 in response to insulin was also higher in Miro1^SkM−/− (Fig. 6H, K). OXPHOS capacity and conductance were also assessed in the quadriceps muscle from Miro1^SkM−/− and Miro1^fl/fl mice to evaluate the direct effects on respiratory function. Miro1^SkM−/− mice exhibited increased N- and S-linked OXPHOS (Fig. 6L), as well as increased N- and S-linked maximal respiration and conductance (Fig. 6M, N).

Fig. 6. Miro1 negatively regulates skeletal muscle insulin action and oxidative capacity in mice with obesity and hyperglycemia.

A, B Western blot and quantification of skeletal muscle Miro1 expression in Miro1 floxed (Miro1^fl/fl) and skeletal muscle Miro1 knockout (Miro1^SkM−/−) mice (n = 4 Miro1^fl/fl, 7 Miro1^SkM−/−). See also Fig. S7. C Study design of Miro1^fl/fl and Miro1^SkM−/− mice on a high-fat diet (HFD). D, E Responses to an intraperitoneal insulin tolerance test and resulting area over the curve (AOC) in Miro1^fl/fl and Miro1^SkM−/− mice (n = 5 Miro1^fl/fl, 4 Miro1^SkM−/−). See also Fig. S8. F, G Steady state glucose disposal rates and skeletal muscle ¹⁴C-2-deoxyglucose uptake in Miro1^fl/fl and Miro1^SkM−/− mice resulting from hyperinsulinemic-euglycemic clamps (n = 3/group). See also Fig. S9. H–K Western blot and quantification of fold change relative to Miro1^fl/fl vehicle for gastrocnemius with and without ex vivo insulin treatment in Miro1^fl/fl and Miro1^SkM−/− mice (n = 4 Miro1^fl/fl, 7 Miro1^SkM−/−) for insulin signaling proteins AKT, AS160, and membrane-bound GLUT4. L Maximal NADH-linked (N-linked), succinate-linked (S-linked), convergent NADH- and succinate-linked (NS-linked), complex III-linked (CIII-linked), and complex IV-linked (CIV-linked) mitochondrial oxidation in Miro1^fl/fl and Miro1^SkM−/− mice (n = 4 Miro1^fl/fl, 6 Miro1^SkM−/−). M, N N-linked and S-linked mitochondrial V_max and conductance measured using creatine kinase clamps in Miro1^fl/fl and Miro1^SkM−/− mice (n = 4 Miro1^fl/fl, 6 Miro1^SkM−/−). Data are represented as mean ± SEM, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, and were statistically compared by two-tailed Student’s t test (B, E, F, G, L, M, and N) or two-way repeated measures ANOVA with Holm-Šídák’s multiple comparisons test (I, J, and K). Exact p-values are as follows: B 0.0001; E 0.0295; F 0.0298; G Gastroc = 0.0474, Quad = 0.0167, Bicep = 0.0469; I Veh Miro1^fl/fl vs. Miro1^SkM−/− = 0.0110, Ins Miro1^fl/fl vs. Miro1^SkM−/− = 0.0003, Miro1^fl/fl Veh vs. Ins = 0.6019, Miro1^SkM−/− Veh vs. Ins = 0.0043; J Veh Miro1^fl/fl vs. Miro1^SkM−/− = 0.0175, Ins Miro1^fl/fl vs. Miro1^SkM−/− = 0.0006, Miro1^fl/fl Veh vs. Ins = 0.1511, Miro1^SkM−/− Veh vs. Ins = 0.0041; K Veh Miro1^fl/fl vs. Miro1^SkM−/− = 0.0006, Ins Miro1^fl/fl vs. Miro1^SkM−/− = 0.0183, Miro1^fl/fl Veh vs. Ins = 0.0028; Miro1^SkM−/− Veh vs. Ins = 0.0167; L N-linked = 0.0129, S-linked = 0.0232, NS-linked = 0.6496, CIII-linked = 0.6852, CIV-linked = 0.7069; M N-linked = 0.0408, S-linked = 0.0015; N N-linked = 0.0151, S-linked = 0.0091. Source data are provided as a Source Data file.

Given the possibility of paracrine or endocrine adaptations to Miro1 loss-of-function in mice, we validated these findings in vitro by inserting a Miro1-targeting (shMiro1) or empty vector (shEV) lentiviral construct into C2C12 skeletal muscle cells. Using this approach, we achieved loss of gene and protein expression in shMiro1 cells that was comparable to that in our animal model (Figs. 7A–C and S10A), with no overt complications related to proliferation or cell structure (Fig. S10B). To directly assess glucose uptake, cells were labeled with tritiated glucose in the presence or absence of insulin. shMiro1 cells exhibited a marked increase in glucose uptake in the presence or absence of insulin relative to shEV cells (Fig. 7D). GLUT4 translocation to the cell membrane in response to insulin was enhanced in shMiro1 cells (Fig. 7E, F). Congruent with our animal studies, shMiro1 cells exhibited increased N-linked and convergent NS-linked OXPHOS relative to shEV (Fig. 7G). Taken together, these data indicate that Miro1 negatively regulates skeletal muscle glucose uptake and bioenergetic function in vivo and in vitro.

Fig. 7. Miro1 negatively regulates intracellular glucose uptake and oxidative capacity in skeletal muscle cells.

A Overview of the design for lentiviral empty vector control (shEV) and Miro1 silencing (shMiro1) in C2C12 myoblasts. B, C Western blot and fold change quantification relative to shEV of Miro1 protein expression in shEV and shMiro1 C2C12 myoblasts (n = 8 biological replicates shEV, 7 biological replicates shMiro1). See also Fig. S10. D [3-³H]glucose uptake in shEV and shMiro1 C2C12 cells treated with or without insulin (n = 4 biological replicates/group/treatment). E, F Immunofluorescence staining for DAPI (blue), α-tubulin (red), and GLUT4 (green) and quantification of fluorescence intensity in shEV and shMiro1 cells treated with and without insulin (n = 5 biological replicates/group/treatment). Scale = 100 μm. G Maximal NADH-linked (N-linked), succinate-linked (S-linked), convergent NADH- and succinate-linked (NS-linked), and complex IV-linked (CIV-linked) mitochondrial oxidation in shEV and shMiro1 C2C12 myoblasts (n = 5 biological replicates/group/treatment). Data are represented as mean ± SEM, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001, and were statistically compared by two-tailed Student’s t test (C and G) or two-way repeated measures ANOVA with Holm–Šídák’s multiple comparisons test (D and F). Exact p-values are as follows: C 0.0002; D Veh shEV vs. shMiro1 < 0.0001, Ins shEV vs. shMiro1 < 0.0001, shEV Veh vs. Ins = 0.0009, shMiro1 Veh vs. Ins = 0.0042; F Veh shEV vs. shMiro1 < 0.0001, Ins shEV vs. shMiro1 < 0.0001, shEV Veh vs. Ins = 0.0023, shMiro1 Veh vs. Ins <0.0001; G N-linked <0.0001, S-linked = 0.0142, NS-linked = 0.0020, CIV-linked = 0.2998. Source data are provided as a Source Data file.

Discussion

In this study, we investigated the role of Miro1 in skeletal muscle insulin resistance and glucose homeostasis in type 2 diabetes (T2D). Our findings reveal that Miro1 accumulates in the skeletal muscle of human patients and mouse models with obesity and hyperglycemia, driven in part by defective mitochondrial AKT signaling and impaired AKT-Miro1 interaction. Notably, exercise training alleviated Miro1 accumulation, improved mitochondrial oxidative capacity, and enhanced insulin sensitivity in individuals with T2D. Furthermore, skeletal muscle-specific deletion of Miro1 in mice and C2C12 cells enhanced mitochondrial function, insulin signaling, and glucose uptake, suggesting that Miro1 serves as a negative regulator of skeletal muscle insulin action (Fig. 8). These findings support a previously unknown mechanism linking mitochondrial dynamics to insulin resistance and suggest that targeting Miro1 could enhance insulin sensitivity in T2D.

Fig. 8. Graphical summary of the role of Miro1 in skeletal muscle.

Insulin lowers skeletal muscle Miro1 expression by stimulating mitochondrial AKT translocation and protein-protein interactions, ultimately facilitating cellular glucose uptake, a process that becomes defective with increasing insulin resistance and progression of type 2 diabetes.

One of the principal observations of this study was the accumulation of skeletal muscle Miro1 in the context of obesity and T2D. Miro1 belongs to the Rho GTPase family, which encompasses a broad class of proteins that act as molecular switches by cycling between GTP-bound and GDP-bound states³⁰. Miro1 structure, function, and localization are atypical for the Rho GTPase family due to its calcium-binding motifs as well as its mitochondrial outer membrane-anchored GTPase domain in the C-terminus region of the protein^8,31. Early functional studies of the protein revealed direct effects on mitochondrial dynamics, trafficking, calcium sensing, and inter-organelle connectivity^20,32,33. To this end, it was observed that loss of Miro1 function in neurons and astrocytes increased the rate of mitochondrial trafficking^33,34. In the context of Parkinson’s disease, mutations in PINK1 or Parkin prevent Miro1 degradation, leading to its accumulation and ultimately impairing axonal retrograde signaling and motility²⁰. However, unlike neuronal, skeletal muscle mitochondria reside in highly structured, reticulated networks without the persistent need for transport along axons³⁵. In Drosophila, which expresses a highly conserved orthologous protein, skeletal muscle loss of Miro1 function renders the fly flightless without altering the mitochondrial network configuration¹², indicative of a cell-autonomous effect with discrete regulatory properties. Similar to what is observed in Parkinson’s disease, Miro1 selectively accumulates in patients and mice with overt hyperglycemia, a phenomenon not explained solely by obesity, suggesting that, in skeletal muscle, this phenomenon may be glucose-dependent. Though we observed a significant negative correlation of Miro1 expression with insulin sensitivity in patients, we believe this is largely explained by patients with T2D, as Miro1 expression was only increased in patients with T2D relative to patients of healthy weight or with obesity. As such, the relationship of Miro1 is curvilinear with respect to diabetes status, and thus, based on our findings, skeletal muscle Miro1 accumulation is driven by insulin resistance rather than obesity alone.

AKT is a central mediator of insulin signaling, primarily known for facilitating glucose uptake through GLUT4 translocation via phosphorylation of key substrates such as AS160³⁶. However, emerging evidence suggests that AKT translocates to mitochondria in response to insulin³⁷, which influences mitochondrial function and dynamics. Miro1, an outer mitochondrial membrane GTPase, has been implicated in mitochondrial trafficking and fission and interacts with key mitochondrial quality-control proteins such as DRP1 and PINK1. Given that AKT phosphorylation in mitochondria has been linked to metabolic control, it is plausible that its interaction with Miro1 modulates mitochondrial dynamics in response to insulin, potentially by altering fission-fusion balance or mitochondrial trafficking to meet cellular energy demands. The defective AKT-Miro1 interaction we observed may contribute to longer-term mitochondrial defects observed in patients with T2D by impairing mitochondrial remodeling. Furthermore, inadequate sensing of insulin action at the organelle level could reduce metabolic flexibility, compromise ATP production, and impede mitochondrial turnover. Intriguingly, this effect is opposite to what is observed in the pancreas, where Miro1 is depleted in T2D, leading to defective insulin secretion via inhibition of the IRS-AKT-Foxo1 pathway³⁸. Similar to our findings, loss of Miro1 was linked to AKT status in the pancreas, further supporting the causal role of Miro1 in the onset and progression of type 2 diabetes, albeit via tissue-specific mechanisms. The opposite patterns of Miro1 observed in the pancreas and skeletal muscle in the context of T2D may be largely explained by differences in mitochondrial demand between these tissues under hyperinsulinemia. In the pancreas, mitochondria are responsible for signaling the secretion of insulin in beta cells, whereas skeletal muscle mitochondria are the primary source of ATP generation to meet metabolic demands³⁹. Furthermore, given the differences in mitochondrial network architecture between skeletal muscle and beta cells, it is highly plausible that Miro1 plays distinct yet interconnected roles in these tissues. As such, spatial-temporal regulation of Miro1 and its relationship to Akt remains an important area for future research in the context of obesity and type 2 diabetes.

Our findings demonstrate that exercise training alleviates skeletal muscle mitochondrial hyper-fragmentation and Miro1 accumulation in patients with obesity and T2D, which is associated with improved mitochondrial function and insulin sensitivity. These results align with previous studies showing that exercise reduces excessive mitochondrial fission through the AMPK/Irisin pathway, improving glucose metabolism and inflammation⁴⁰. The observed reductions in DRP1 phosphorylation, Mid49, and Mid51 expression suggest that exercise promotes mitochondrial elongation, consistent with reports that endurance training enhances mitochondrial connectivity and efficiency^27,41–45. Notably, despite significant changes in mitochondrial morphology, we found no differences in PGC1α, MFN1, MFN2, OPA1, PINK1, or Parkin expression, suggesting that exercise-induced improvements in mitochondrial function may be mediated more by network remodeling rather than biogenesis. This highlights the importance of mitochondrial dynamics and quality control in metabolic adaptation to exercise. The increase in OXPHOS capacity and respiratory conductance further supports the role of mitochondrial remodeling in enhancing energy efficiency, as previously reported in both human and animal models^46–48. Additionally, the reduction in LC3 expression suggests a decrease in mitophagy signaling, potentially indicating enhanced mitochondrial quality that reduces the need for turnover. Given that mitochondrial dysfunction and impaired trafficking contribute to insulin resistance, the exercise-induced reduction of Miro1 accumulation suggests a previously underappreciated role for Miro1 in skeletal muscle insulin signaling. These findings reinforce the therapeutic potential of exercise in restoring mitochondrial function and metabolic health in T2D, providing further evidence that targeting mitochondrial dynamics may be an effective strategy for improving insulin resistance. However, further research is required to determine whether modulating Miro1 directly alters muscle contractility or responsiveness to exercise training.

Our study demonstrates that skeletal muscle-specific Miro1 deletion enhances insulin sensitivity and mitochondrial oxidative phosphorylation, supporting a direct, non-canonical role for Miro1 in regulating glucose metabolism. These findings align with prior work linking mitochondrial dynamics to insulin action, in which excessive mitochondrial fragmentation has been implicated in insulin resistance¹⁴. The observed increase in AKT and AS160 phosphorylation, along with enhanced GLUT4 translocation in Miro1^SkM−/− mice, suggests that Miro1 deletion potentiates insulin signaling, thereby improving glucose uptake. This contrasts with previous reports where defects in mitochondrial trafficking and turnover were associated with metabolic dysfunction⁴⁹, indicating that the impact of Miro1 on metabolism may be tissue- and context-dependent. Interestingly, deletion of Miro1 was associated with higher basal rates of GLUT4 translocation in mouse skeletal muscle and C2C12 myotubes, as well as increased basal AKT and AS160 phosphorylation in mouse skeletal muscle, further supporting that Miro1 may directly inhibit insulin responsiveness in skeletal muscle. Furthermore, the increase in OXPHOS capacity and electron transport chain (ETC) efficiency in Miro1-deficient muscle suggests that Miro1 constrains mitochondrial respiratory capacity. Previous studies have shown that mitochondrial quality control mechanisms, including mitophagy, are essential for maintaining oxidative capacity and insulin sensitivity⁵⁰. Our findings extend this notion by suggesting that excessive Miro1 accumulation may disrupt these processes, leading to metabolic impairments. The congruence between our in vivo and in vitro results further supports a muscle-autonomous role for Miro1 in glucoregulatory function, independent of systemic adaptations. While this work provides a mechanistic and inferential basis for the link between skeletal muscle Miro1 accumulation and glucose homeostasis in cellular and mouse models, prospective human studies are required to see if Miro1 accumulation is observed as patients with obesity transition from pre- to type 2 diabetes.

Taken together, our findings establish skeletal muscle Miro1 accumulation as a hallmark of insulin resistance in T2D. We demonstrate that impaired mitochondrial AKT translocation in response to insulin underlies Miro1 accumulation, contributing to disrupted glucoregulatory function in skeletal muscle. Notably, exercise training reverses these deficits by promoting mitochondrial network remodeling, improving oxidative phosphorylation, and restoring insulin sensitivity. Furthermore, skeletal muscle-specific deletion of Miro1 enhances glucose uptake and bioenergetic function, confirming its role as a negative regulator of insulin action. Collectively, these data position Miro1 as a novel link between mitochondrial quality control and insulin resistance, highlighting its potential as a therapeutic target for T2D.

Methods

All human studies were conducted according to protocols approved by the Institutional Review Boards of the Cleveland Clinic and Pennington Biomedical Research Center. Patients provided written informed consent prior to initiation of each study, and patients were compensated upon completion of a trial. All human studies were conducted in accordance with the principles in the Declaration of Helsinki.

All animal protocols were approved by the Institutional Animal Care and Use Committee at Pennington Biomedical Research Center and are reported according to ARRIVE guidelines. All animals were housed at 28–30 °C (40–60% humidity) with 12:12 light:dark cycles unless otherwise described. Animal welfare was monitored daily by the study research team, animal facility staff, and when necessary, veterinary staff. Upon study completion, mice were humanely euthanized as described for each study.

Human study participants

Human Study 1: Cross-sectional assessment of healthy, overweight/obesity, or type 2 diabetes

In a cross-sectional design, adults between the age of 18–60 were prospectively assigned to one of three groups: healthy weight (HW; BMI 18–24.9 kg/m² without T2D), obesity (Ob; BMI: 25–50 kg/m² without T2D), or obesity with type 2 diabetes (T2D; BMI 25–50 kg/m² with T2D) (Methods S1). Participants assigned to T2D had a BMI ≥ 25 kg/m² and either had a prior diagnosis of T2D or met American Diabetes Association criteria for diagnosis at physician screening⁵¹. Following screening, participants were admitted for a 3-day inpatient period during which they were provided standardized meals (55% carbohydrate, 35% fat, and 10% protein) based on individual energy needs. On day 2, participants were assessed for body composition, followed by an incremental aerobic capacity test. Participants were discharged after testing and readmitted at 1700 h for an additional overnight stay and a dietary control period. At 2000 h, patients were placed on NPO status, except for water, for the remainder of the study. On day 3, patients were awakened at 0530 h for catheterization, followed by initiation of a glucose tracer at 0600 h, a muscle biopsy at 0700 h, and a single-stage hyperinsulinemic-euglycemic clamp study at 0800 h. A second muscle biopsy was collected at 1100 h under steady-state insulin-stimulated conditions. The study was approved by the Institutional Review Boards of the Cleveland Clinic and Pennington Biomedical Research Center, and all participants provided informed consent. Data were collected between April 2016 and July 2023. All data and samples were de-identified at the time of collection. A subgroup of patients from this prospective observational study also participated in trials registered on clinicaltrials.gov (NCT02697201 and NCT02977442, respectively).

Human Study #2: a randomized controlled trial of exercise training in patients with obesity and type 2 diabetes

The Dynamics of Muscle Mitochondria in Type 2 Diabetes (DYNAMMO-T2D) study was a parallel group randomized controlled trial to assess the effects of exercise training on mitochondrial dynamics in adults with obesity and type 2 diabetes (Methods S2). Sedentary adults ages 18–60 y with obesity (BMI ≥ 25 kg/m²) and T2D (HbA_1c ≥ 6.5%) were randomized in blocks of 4 (1:1) to 12 weeks of American Diabetes Association (ADA) Standards of Care in Diabetes⁵² (Ctrl) or aerobic exercise training (Ex). Patients randomized to Ctrl continued physician-guided medical therapy for 12 weeks, in conjunction with adherence to American College of Sports Medicine guidelines for exercise and nutritional therapy for T2D⁵³. Patients randomized to Ex underwent 12 weeks of supervised aerobic exercise training, 5 days per week, 1 h per day, at 60–85% of HR_MAX (~70% of VO_2PEAK)⁵⁴. Immediately prior to and upon completion of the 12-week study, participants were admitted for a 3-day inpatient stay consisting of body composition, aerobic capacity, and insulin sensitivity testing, as well as undergoing muscle biopsy procedures. During the 3-day inpatient period, they were provided standardized meals (55% carbohydrate, 35% fat, and 10% protein) based on individual energy needs. On day 2, participants were assessed for body composition, followed by an incremental aerobic capacity test. Participants were discharged after testing and readmitted at 1700 h for an additional overnight stay and a dietary control period. At 2000 h, patients were placed on NPO status, except for water, for the remainder of the study. On day 3, patients were awakened at 0530 h for catheterization, followed by initiation of a glucose tracer at 0600 h, a muscle biopsy at 0700 h, and initiation of the two-stage hyperinsulinemic-euglycemic clamp study at 0800 h. A second muscle biopsy was collected at 1100 under insulin-stimulated conditions. All data and samples were de-identified at the time of collection. The study was approved by the Institutional Review Boards at the Cleveland Clinic and Pennington Biomedical Research Center, and patients provided written informed consent before participation. Data were collected between April 2016 and November 2023. The trial was registered on clinicaltrials.gov (NCT02977442) prior to the enrollment of study participants in 2016.

Animal models

Mouse study #1: diet-induced obesity and hyperglycemia (FFC_low vs FFC_high)

18-week-old male C57BL/6J mice (n = 8/diet) were randomized to either a low-fat (10%), fructose (0%), and cholesterol (0%) (FFC_low; Research Diets D09100304) or a high-fat (40%), high-fructose (20%), high-cholesterol (2%) diet (FFC_high; Research Diets D09100310) for 28 weeks. Animals were given ad libitum access to food and water. Body weight and food intake were measured weekly, and body composition was measured every 4 weeks by NMR (Bruker LF110). After 28 weeks of treatment, mice were euthanized by inhalation of isoflurane and cervical dislocation, and blood and tissues were collected.

Mouse study #2: genetically induced hyperglycemia (wild-type vs A^y; Lepr^db/+ vs Lepr^db/db)

Male wild-type C57BL/6 mice (WT) and mice with the Lethal yellow mutation (A^y) (n = 8/genotype) were housed under standard housing conditions (21–23 °C) and fed a chow diet (Purina 5015). Animals were given ad libitum access to food and water. Body weight and food intake were measured weekly, and intraperitoneal glucose tolerance was assessed at 14 weeks of age. At 16 weeks of age, mice were euthanized by carbon dioxide inhalation and cervical dislocation, and blood and tissues were collected.

Additionally, quadriceps muscles from 10-week-old male Lepr^db heterozygous (db/+) and knockout (db/db) mice (n = 8/genotype) were purchased from the Jackson Laboratory.

Mouse study #3: ex vivo insulin signaling in HFD

Male C57BL/6 mice (n = 7) were bred in-house and assigned to a high-fat (60%) diet (HFD; Research Diets D12492) at 14 weeks of age. Animals were given ad libitum access to food and water. At 30 weeks of age, mice were euthanized by inhalation of isoflurane and cervical dislocation, and blood and tissues were collected. Gastrocnemius muscles were exposed to insulin ex vivo as described in detail below.

Mouse study #4: Miro1^fl/fl vs Miro1^SkM−/−

14-week-old male mice with a skeletal muscle-specific deletion of Miro1 (Miro1^SkM−/−; n = 10) and littermate controls (Miro1^fl/fl; n = 8) were assigned to HFD for 16 weeks. Animals were given ad libitum access to food and water unless otherwise described. Body weight and food intake were measured weekly, and body composition was assessed monthly. Intraperitoneal insulin tolerance tests (ITT) were performed at 28 weeks of age (14 weeks of treatment). At 30 weeks of age (16 weeks of treatment), mice were euthanized by inhalation of isoflurane and cervical dislocation, and blood and tissues were collected.

Cell culture and lines

C2C12 cells (CRL-1772, passage 2) were purchased from American Type Culture Collection (ATCC) and expanded in Dulbecco’s modified Eagle medium (DMEM) supplemented with 4.5 g/l D-glucose (25 mM), 50 μ/mL penicillin/streptomycin, and 10% fetal bovine serum. Cells were incubated at 37 °C, 5% CO₂ in a humidified incubator. After reaching 100% confluence, cells were induced to differentiate by replacing the growth medium with high-glucose DMEM supplemented with 2% horse serum and incubating for 48 h, unless otherwise indicated. In both growth and differentiation stages, the medium was replenished daily.

Method details

Human studies

3-day inpatient stay

Participants arrived in the evening on day 1 of the inpatient stay and were provided a standardized meal consisting of 55% carbohydrate, 35% fat, and 10% protein, with caloric composition tailored to individual needs. Body composition and incremental aerobic capacity tests were performed on day 2 in the morning. After testing was completed, participants were discharged and returned at 1700 for another overnight stay and a dietary control period identical to day 1. Patients were placed on NPO status, except for water, at 2000 h for the remainder of the inpatient stay. Patients were awakened at 0530 h on day 3 for catheterization. Glucose tracer was initiated at 0600 h, and a muscle biopsy was collected at 0700 h. The hyperinsulinemic-euglycemic clamp was initiated at 0800 h. A second muscle biopsy was collected at 1100 h under insulin-stimulated conditions. Baseline and insulin-stimulated whole-body respiratory exchange ratios (RER) and substrate metabolism were measured by indirect calorimetry.

Body composition

Height and weight were measured in a pre-weighed hospital gown. Dual-energy X-ray absorptiometry (Lunar iDXA, Madison, WI, USA) was used to determine whole body fat and lean mass. iDXA software was used to estimate fat and lean mass according to the manufacturer’s instructions.

OGTT

Glucose tolerance and insulin secretion were determined by an oral glucose tolerance test (OGTT)⁵⁵. After an overnight fast, patients were given a 75 g glucose challenge. Blood samples were obtained from an indwelling intravenous line at baseline (0) and at 30, 60, 90, 120, and 180 min. Samples were immediately assayed for glucose (YSI 2300; STAT Plus, Yellow Springs, OH, USA), and plasma was isolated by centrifugation at 2000 × g for 10 min at 4 °C and stored at −80 °C until insulin and C-peptide assays. Calculations for total area of the curve of glucose (tAUC glucose), total area of the curve of insulin (tAUC insulin), total area of the curve of C-peptide (tAUC C-peptide), homeostatic model of insulin resistance (HOMA-IR), disposition index, and Matsuda index were made from equations described previously⁵⁶.

Blood chemistry

Metabolic profiles were analyzed using an automated platform⁵⁴. Cell counts and lipid concentrations were included in the automated blood chemistry platform. Insulin was determined by a commercially available ELISA.

Aerobic capacity

Peak oxygen consumption (VO_2peak) was determined using a modified Bruce protocol⁴⁵. The following criteria were used to determine a maximal test: (1) oxygen consumption plateau (<150 mL/min), (2) heart rate within 15 beats of age-predicted maximum, (3) RER > 1.15, and/or (4) volitional fatigue. Three of the four criteria were required for the test to be considered maximal.

Insulin sensitivity

Insulin sensitivity was determined using a two-stage hyperinsulinemic-euglycemic clamp (90 mg/dL, 40 mU·m⁻²·min⁻¹). Medications for T2D were withheld for at least 48 h prior to the hyperinsulinemic-euglycemic clamp. A constant insulin infusion and a variable infusion of 20% dextrose began at 0 min. Venous blood was sampled every 5 min from an arterialized heated hand to measure glucose (YSI 2900 Biochemistry Analyzer, YSI, Inc., Yellow Springs, OH, USA). The glucose infusion rate (GIR) was adjusted accordingly to maintain plasma glucose at 90 mg/dL. Insulin sensitivity was calculated as insulin-stimulated glucose metabolism (M; mg·kg⁻¹·min⁻¹) divided by plasma insulin (I; μU/ml) over a 30-min steady-state period. Plasma for assessing glucose kinetics was deproteinized and extracted, and the resulting samples were derivatized before analysis by gas chromatography-mass spectrometry. Isotopic enrichment was determined by fitting the fractional abundances (M+2; m/z 330)/(M0; m/z 328) against a calibration curve. The rate of glucose appearance was then derived using the Steele equation⁵⁷. Whole-body RERs and substrate metabolism were determined under basal and insulin-stimulated conditions via indirect calorimetry (Vmax Encore, Viasys, Yorba Linda, CA, USA).

Exercise training RCT: standard care group

Patients randomized to Ctrl received standard-of-care medical therapy throughout the study. Patients continued all prescribed medications and received lifestyle and dietary recommendations based on ADA guidelines, as directed by the patient’s healthcare provider.

Exercise training RCT: exercise training group

Exercise consisted of walking and running on a treadmill or stationary cycling on a cycle ergometer⁵⁴. Participants completed exercise sessions 5 days/week for 50–60 min/session for 12 weeks. The initial exercise prescription was 55–60% of heart rate max (HR_max) and gradually increased, so that after 2–3 weeks participants were exercising at 80–85% HR_max (~70% VO_2max). All exercise sessions were supervised by an Exercise Physiologist. Heart rate monitors (Polar Electro Inc., Woodbury, NY, USA) were worn during exercise to provide visual feedback of individual target heart rates. Training compliance was documented through regular attendance at training sessions, heart rate recordings, and the use of training logs. Dietary intake was monitored by having each subject complete 3-day food records at baseline and at 4-week intervals during the intervention. Calorie and macronutrient intake data were used as covariates.

Skeletal muscle biopsy

Skeletal muscle specimens were collected from the vastus lateralis using a modified Bergström biopsy technique following an overnight fast⁵⁸, prior to the start of the hyperinsulinemic-euglycemic clamp and under steady-state insulin-stimulated conditions. Upon collection, fat and connective tissue were dissected. Samples were immediately placed into preservation media or frozen in liquid nitrogen. Flash-frozen samples were stored at −140 °C until analysis.

Mitochondrial isolation

Mitochondria were isolated from fresh skeletal muscle tissue based on previously published protocols with minor modifications^59,60. Skeletal muscle tissue sectioned at time of biopsy was immediately placed on ice-cold BIOPS preservation buffer (50 mM K+-MES, 20 mM taurine, 0.5 mM dithiothreitol, 6.56 mM MgCl₂, 5.77 mM ATP, 15 mM phosphocreatine, 20 mM imidazole, pH 7.1, adjusted with 5 M KOH at 0 °C, 10 mM Ca–EGTA buffer, 2.77 mM CaK₂EGTA + 7.23 mM K₂EGTA; 0.1 mM free calcium) until processing. Fresh tissue was transferred to mitochondrial isolation buffer (200 mM sucrose, 10 mM Tris, 0.1 M EGTA/Tris) containing protease and phosphatase inhibitors and minced with scissors on ice. Tissue pieces were transferred to a 7 mL glass-on-glass Dounce homogenizer with 1.5 mL of mitochondrial isolation buffer and homogenized for 12 strokes. Homogenate was transferred to a 1.7 mL Eppendorf tube and centrifuged at 600 × g for 10 min at 4 °C. The supernatant was collected and centrifuged at 7000 × g for 10 min at 4 °C. The resulting supernatant (cytosolic fraction) was collected and stored at −80 °C until further analysis. The remaining mitochondrially enriched pellet was washed with 1.5 mL of mitochondrial isolation buffer and centrifuged again at 7000 × g for 10 min at 4 °C. The supernatant was removed, and the mitochondrial pellet was incubated on ice until the respiratory function analysis. Following respiratory assays, the remaining mitochondria were re-pelleted and stored at −80 °C until further analysis.

Skeletal muscle OXPHOS and ET capacity

Skeletal muscle OXPHOS and ET capacity were determined ex vivo in isolated mitochondria as described previously with minor modifications¹³. Freshly prepared mitochondrial pellets were resuspended in Buffer D (105 mM KMES, 30 mM KCl, 10 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, with 2.5 g/L of defatted BSA, pH 7.2)²⁹ and injected into an oxygraphy chamber (~150 mg) containing 2 mL of Buffer D. OXPHOS and ET capacity were determined using two independent protocols. The first protocol was designed to measure N- and S-linked OXPHOS and ET capacity, and complex IV activity using the following substrates, uncouplers, and inhibitors: malate (2 mM), pyruvate (2.5 mM), ADP (2.5 mM), glutamate (10 mM), cytochrome c (7.5 μM) succinate (10 mM), carbonylcyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 0.5 μM increment), rotenone (75 nM), antimycin A (125 nM), tetramethyl-p-phenylenediamine (TMPD, 0.5 μM), ascorbate (2 mM), and sodium azide (200 mM). The second protocol tested F- and complex III-linked OXPHOS and CIII-linked ET capacity using the following substrates, uncouplers, and inhibitors: malate (2 mM), palmitoylcarnitine (10 μM), ADP (2.5 mM), rotenone (75 nM), duroquinol (0.5 mM), carbonylcyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 0.5 μM increment), and antimycin A (125 nM). Oxygen flux was normalized to mitochondrial protein content measured by BCA.

Skeletal muscle OXPHOS conductance

Skeletal muscle OXPHOS conductance was assessed in isolated mitochondria by the creatine kinase clamp technique as previously described with minor modifications²⁹. Isolated mitochondria were resuspended in Buffer D and injected into an oxygraph chamber with Buffer D, ATP (5 mM), creatine (5 mM), phosphocreatine (1 mM), and creatine kinase (200 units/mL). Substrate oxidation was supported by either glutamate and malate to derive NADH-linked conductivity or succinate and rotenone to derive succinate-linked conductivity. The extramitochondrial ATP:ADP ratio was then adjusted by sequential titration of phosphocreatine (3–30 mM). OXPHOS capacity was estimated in the presence of saturating substrate. To determine conductance, the ionic strength of Buffer D was calculated, and the ΔG of ATP hydrolysis at each addition of phosphocreatine was determined. Steady-state mitochondrial respiration at each stage was plotted against the ΔG of ATP hydrolysis, and a linear regression was used to determine the slope, which is ~1/R, the conductance.

Western blotting

Protein was extracted from skeletal muscle tissue, and Western blotting was performed as previously described with minor modifications⁴⁵. Frozen skeletal muscle specimens were homogenized in ice-cold cell extraction buffer with added protease inhibitor cocktail, 5 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and PhosSTOP using a Polytron. Homogenates were incubated on ice for 30 min and centrifuged for 10 min at 21,000 × g at 4 °C. The supernatant was collected, and protein content was measured using a BCA protein assay kit. Lysate (2 µg/µL) was solubilized in Laemmli buffer containing 5% β-mercaptoethanol and boiled for 5 min. Samples were loaded onto Tris-glycine gels and separated by SDS electrophoresis. Gels were transferred to PVDF membranes and blocked with 3% non-fat dry milk in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 h. Membranes were incubated overnight with primary antibodies (Table S3), washed with TBST, and incubated with species-specific horseradish peroxidase-conjugated secondary antibodies (Table S3). Immunoreactive proteins were visualized with an enhanced chemiluminescence reagent and quantified by densitometric analysis in ImageJ. Visible bands reactive with an internal control were quantified. Gel-to-gel variation and equal protein loading were controlled using a standardized sample on each gel, and values were normalized to loading control (β-actin) and expressed as relative change from baseline, unless otherwise stated.

Skeletal muscle lipid droplet and contractile structure

Intramuscular lipid and myofibril density were spatially resolved and quantified by Focused ion beam scanning electron microscopy (FIB-SEM) and segmentation analysis as described previously (Methods S3)⁶¹. ~25 mg of muscle tissue was collected in 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.2. Samples were cut into 1 mm sections and fixed in a solution containing 2.5% glutaraldehyde, 1% paraformaldehyde, and 0.12 M sodium cacodylate, pH 7.2, then transferred to a 4% osmium solution. Samples were stained with 2% osmium and incubated overnight in a 1% uranyl acetate solution at 4 °C. Samples were washed with water and dehydrated with increasing concentrations of ethanol, then incubated in a 50% Epon:50% ethanol solution for 4 h in a vacuum-sealed container, followed by 75% Epon:25% ethanol overnight. The last day of sample preparation consisted of three 100% Epon resin incubations. Samples were placed onto aluminum SEM mounts (Carl Zeiss Microscopy GmbH, Jena, Germany), excess resin was removed, and the mounts were placed in a 60 °C oven for 48 h to polymerize. Sample stubs were then mounted in a UCT Ultramicrotome and faced with a Trimtool 45 diamond knife at a thickness feed of 100 nm and a rate of 80 mm/s. Images were acquired with a Crossbeam 540 using Atlas 5 software. Lipid droplets were segmented and quantified using the deep learning model called MitoNet⁶². Myofibrils were segmented and quantified as described previously⁶³.

Mitochondrial ultrastructure and content

Two-dimensional mitochondrial structures were resolved using transmission electron microscopy⁶⁴. 15–20 mg of muscle tissue was fixed by immersion in a triple aldehyde-DMSO mixture⁶⁵. Tissue blocks were post-fixed in ferrocyanide-reduced osmium tetroxide, soaked in acidified uranyl acetate, dehydrated in ascending ethanol concentrations, passed through propylene oxide, and embedded in Poly/Bed resin. Thin sections were stained with acidified uranyl acetate⁶⁶, followed by modified Sato’s triple lead stain⁶⁷. Mitochondrial elongation was determined by deriving the form factor from manually traced, clearly discernible mitochondrial outlines using the mitochondrial shape descriptors plugin in ImageJ 4, as described previously⁶⁸.

Animal studies

Animal acquisition, housing, and husbandry

Mouse study 1: diet-induced hyperglycemia (FFC_low vs FFC_high)

Male C57BL/6 mice were bred in-house and weaned at 3 weeks of age. Mice were multi-housed until the start of the experimental diet, at which time they were separated into single-housed cages. Mice were housed at a murine thermoneutral temperature (28–30 °C) from weaning through the entire duration of the study. At 18 weeks of age, mice were randomized to an FFC_low or an FFC_high diet. Throughout the diet study, body weight and food intake were measured weekly, and body composition was assessed by NMR every 4 weeks. Mice were euthanized after 28 weeks of an experimental diet.

Mouse study 2: genetically induced hyperglycemia (wild-type vs A^y; Lepr^db/+ vs Lepr^db/db)

Male C57BL/6 and A^y mice were purchased from Jackson Laboratory. Upon arrival, mice were single-housed under standard housing conditions (21–23 °C). Mice were fed a standard chow diet. Throughout the study, body weight was assessed weekly. Mice were euthanized at 16 weeks of age.

Quadriceps muscles from 10-week-old db/+ and db/db mice were purchased from Jackson Laboratory. Male heterozygous and Lepr^db mice were bred and maintained by Jackson Laboratory under standard housing conditions and diet. Mice were euthanized by carbon dioxide inhalation at 10 weeks of age, and quadriceps tissue was collected and snap-frozen in liquid nitrogen. Tissues were shipped on dry ice and stored at −80 °C upon receipt until analysis.

Mouse study 3: ex vivo insulin signaling in HFD

Male C57BL/6 mice were bred in-house and maintained at thermoneutrality (28–30 °C) on a standard chow diet. At 12 weeks of age, mice were placed on an HFD for 16 weeks. After 16 weeks of the diet, mice were euthanized. The gastrocnemius muscle was collected into ice-cold BIOPS at the time of necropsy and held on ice until the time of assay (4–6 h). Individual gastrocnemius muscles were cleaned of fat and connective tissue and were split into two pieces at the Achilles tendon. Muscle samples were incubated in DMEM high-glucose media containing 3 µM insulin (Novolin) or DMEM vehicle control for 15 min. After incubation, the media was removed, and the samples were washed with ice-cold PBS containing 0.1 M Na₃VO₄. Samples in ice-cold PBS were transferred to Eppendorf tubes and centrifuged at 2000 × g for 1 min at 4 °C. Following centrifugation, PBS was removed, and the samples were immediately snap-frozen and stored at −80 °C until further analysis.

Mouse study 4: Miro1^fl/fl vs Miro1^SkM−/−

Mice with skeletal muscle-specific deletion of Miro1 were generated and bred in-house. Male Miro1^SkM−/− mice and floxed littermate controls were weaned at 3 weeks of age. Mice were multi-housed until the start of the experimental diet, at which time they were separated into single-housed cages. Mice were housed at a murine thermoneutral temperature (28–30 °C) from weaning through the entire duration of the study. At 14 weeks of age, mice were placed on HFD. Throughout the diet study, body weight and food intake were measured weekly, and body composition was assessed by NMR every 4 weeks. Mice were euthanized after 16 weeks of an experimental diet.

Generation and validation of mice with skeletal muscle-specific deletion of Miro1

Mice with loxP sites flanking Miro1 were purchased from Jackson Laboratory. Additionally, transgenic mice carrying Cre recombinase under the human Acta1 promoter were purchased from Jackson Laboratory (strain #006149) and bred with Miro1-floxed mice to generate a skeletal muscle-specific deletion of Miro1. Tail snips were collected from pups at 2 weeks of age for DNA extraction, PCR, and gel electrophoresis to determine genotype. Skeletal muscle-specific deletion was confirmed by qPCR and Western blot analysis.

DNA extraction, PCR, and Southern blotting

Tail snips collected from 2-week-old pups were immediately digested in 500 µL lysis buffer (1 M Tris, 5 M NaCl, 0.5 M EDTA, 10% SDS) in the presence of proteinase K at 55 °C overnight. Once digested, 500 µL of phenol-chloroform was added, and samples were vortexed and centrifuged for 10 min at 21,000 × g. The resulting supernatant was centrifuged for 10 min at 21,000 × g to pellet DNA. Pellets were cleaned with 1 mL of ice-cold 100% ethanol and centrifuged for 10 min at 21,000 ×  g. The supernatant was discarded, and the pellets were dried at room temperature. Pellets were resuspended in TE buffer (1 M Tris, 0.5 M EDTA) and stored at 4 °C until further analysis. PCR was performed using GoTaq® Flexi DNA Polymerase kit using 50–100 ng of DNA in the presence of primers for either the Acta1 Cre or the Miro1 loxP sites based on sequences provided by Jackson Laboratory (Table S4). Amplification of Cre and loxP DNA was visualized using 1.5–2.5% agarose gels.

Intraperitoneal insulin tolerance test

Insulin tolerance tests (ITT) were performed as described previously with minor modifications⁶⁹. Mice were fasted for 3 h prior to the start of the ITT, and body weight was recorded. The tail was cleaned with a 70% isopropanol wipe before creating a tail nick with a lancet. Baseline blood glucose was measured beginning at 30 min prior to the insulin injection using a glucometer. A 1 IU/kg dose of insulin in saline was injected into the peritoneum at time point 0. Blood glucose was measured by reopening the scab from the lancet nick at 10, 20, 30, 45, 60, 75, 90, and 120 min after insulin injection, and the area under the curve was calculated for each mouse.

Hyperinsulinemic-euglycemic clamps and ¹⁴C-2-deoxyglucose uptake

Hyperinsulinemic-euglycemic clamps and ¹⁴C-2-deoxyglucose uptake were performed in male Miro1^fl/fl and Miro1^SkM−/− mice maintained on a high-fat diet as previously described with minor modifications⁷⁰. Mice underwent surgery for dual jugular vein and carotid artery catheterization. Mice were anesthetized with inhaled isoflurane (3–5% induction, 1–3% maintenance) and received 5 mg/kg ketoprofen as an analgesic prior to surgical incisions. Catheters were inserted in the carotid artery and jugular vein and secured with a silk suture. Catheters were threaded to the dorsal side of the mouse and exteriorized through a single access point near the shoulder blades and flushed with sterile heparinized saline solution. Mice were administered additional ketoprofen at 24 and 48 h postoperatively, and catheters were flushed on day 5 postoperatively.

Mice with indwelling dual jugular vein and carotid arterial catheters underwent unrestrained terminal hyperinsulinemic-euglycemic clamps 5–8 days postoperatively. Mice were fasted for 5 h prior to the start of the clamp. A 120-min clamp was performed with a continuous infusion of human insulin (Humulin, Eli Lilly) at 4 mU/kg/min and varying infusion rates of 20% dextrose. Red blood cells collected from a donor mouse were infused at a constant rate of 5 µL/min to maintain blood volume. All infusions were done using mouse microinfusion pumps (Instech Model 400 Syringe Pump, Instech Laboratories). Arterial blood glucose was sampled every 10 min during the clamp, and the glucose infusion rate was adjusted to maintain glucose at 125–135 mg/dL. Blood samples were obtained at time points 0, 90, 100, 110, and 120 min for plasma insulin measurements. At 120 min, a 13 μCi bolus of ¹⁴C-2-deoxyglucose was administered via injection through the jugular line. Blood samples were collected at 2, 5, 10, 15, and 25 min following the ¹⁴C-2-deoxyglucose bolus to measure plasma ¹⁴C. Mice were euthanized 25 min after the ¹⁴C-2-deoxyglucose bolus, and tissues were collected, immediately snap-frozen, and stored at −80 °C until measurement of ¹⁴C uptake.

[¹⁴C(U)]2-deoxyglucose plasma and tissue sample preparation

Radioactively labeled standards, plasma, and tissue samples were processed as previously described with minor modifications⁷⁰. 0.25 µCi in 4 µL was saved from each preparation of [¹⁴C(U)]2-deoxyglucose bolus for use as a standard. 196 µL of water was directly added to each [¹⁴C(U)]2-deoxyglucose standard. 50 µL of diluted standard was transferred to a 1.7 mL Eppendorf tube and 950 µL of water was added. 100 µL of this dilution was transferred to a scintillation vial and 900 µL of water was added. To prepare the plasma samples, 10 µL of plasma collected at each time point was transferred to a 1.7 mL Eppendorf tube. An equal volume of saline was added to each tube, followed by 100 µL of 0.075 N Ba(OH)₂ and 100 µL of 0.075 ZnSO₄. Samples were mixed and centrifuged at 13,000 × g for 5 min at RT. After centrifugation, 100 µL of supernatant was transferred to a scintillation vial, and 900 µL of water was added. 10 mL of scintillation fluid was dispensed into each scintillation vial of standard and plasma samples.

~50 mg of frozen skeletal muscle tissue was homogenized in 750 µL of 0.5% perchloric acid diluted in water. The tissue samples were pulverized with a bead homogenizer at 6 m/s for 20 × 20-s cycles, with 10-s pauses at 4 °C. Following homogenization, an additional 750 µL of the 0.5% perchloric acid solution was added, and the samples were vortexed to mix. The samples were then centrifuged at 20,000 × g for 5 min at 4 °C. After centrifugation, 1 mL of supernatant was transferred to a 1.7 mL Eppendorf tube, adjusted to a pH of 6.5–7.5, and re-centrifuged at 2000 × g for 10 min at 4 °C. Following centrifugation, 125 µL of supernatant from each sample was transferred to a scintillation vial, then 875 µL of water was added. 250 µL of supernatant was transferred to a new 1.7 mL Eppendorf tube, followed by 125 µL of 0.3 N Ba(OH)₂ and 125 µL of 0.3 N ZnSO₄. The samples were vortexed and incubated at RT for 5 min, then centrifuged at 20,000 × g for 5 min at RT. After centrifugation, 125 µL of supernatant from each sample was transferred to a scintillation vial, followed by 875 µL of water. 10 mL of scintillation fluid was dispensed into each scintillation vial. The samples were gently vortexed to homogenize the lysate into the scintillation cocktail, and ¹⁴C radioactivity was determined on a scintillation counter. Tissue uptake rates were calculated as tissue disintegrations per minute per gram, normalized to the time-integrated plasma tracer activity, and converted to molar units using the tracer-specific activity.

Skeletal muscle ex vivo insulin signaling

The gastrocnemius muscle was collected into ice-cold BIOPS at the time of necropsy and incubated in DMEM high-glucose media containing 3 µM insulin or DMEM vehicle control for 15 min as described above.

Cellular fractionation

Mitochondria and cytosol were fractionated from frozen gastrocnemius tissue as described above. The cellular membrane was fractionated from the resulting pellet containing the nuclear and membrane fractions from the first centrifugation of the isolation using a commercially available cell membrane extraction kit (Thermo Scientific Subcellular Fractionation Kit). Nuclear pellets were resuspended in 1 mL cytoplasmic extraction buffer containing protease and phosphatase inhibitors, then briefly vortexed and sonicated. Samples were centrifuged for 5 min at 500 × g at 4 °C. The resulting supernatant was discarded, and membrane extraction buffer plus protease and phosphatase inhibitors were added to the pellets. Samples were briefly vortexed and sonicated again, then incubated in a rotary tube at 4 °C for 30 min. Samples were centrifuged at 3000 × g for 10 min at 4 °C. The resulting supernatant containing the membrane fraction was collected. Protein content of each subcellular fraction was measured using BCA.

Western blotting

1–3 µg/µL of mitochondrial, cytosolic, and nuclear lysate were used for Western blotting as described above. Membranes were incubated overnight with primary antibodies (Table S3) and visualized using chemiluminescence as described above.

Co-immunoprecipitation

AKT antibody was precipitated by incubating 10 µg of mitochondrial protein with protein A/G agarose beads at 1:10 dilution overnight at 4 °C according to a previously published protocol with minor modifications²². Protein was pulled down by centrifuging at 1000 × g for 1 min at 4 °C. Pellets were washed and then denatured by boiling for 5 min in 5× SDS sample buffer. Samples were mixed and centrifuged at 10,000 × g for 3 min at 4 °C. The resulting supernatant was loaded onto Tris-Glycine gels for electrophoresis and Western blotting, as described, against Miro1. Horseradish peroxidase–tagged secondary antibody was used. Input protein, negative beads-only control, and IgG controls were also blotted.

Skeletal muscle OXPHOS and ET capacity

Skeletal muscle OXPHOS and ET capacity were measured as described above using isolated mitochondria from the quadriceps muscle. Mitochondria were isolated from fresh quadriceps muscle collected on ice-cold BIOPS at the time of necropsy, as described above. Mitochondrial protein concentrations were determined by BCA. Mitochondrial pellets were resuspended in ice-cold Buffer D, and 20 µg of mitochondrial protein was loaded into 2 mL oxygraphy chambers containing Buffer D.

Skeletal muscle mitochondrial conductance

Mitochondrial conductance was determined by creatine kinase clamp using isolated mitochondria from the quadriceps muscle as described above. Mitochondrial pellets were resuspended in ice-cold Buffer D, and 20 µg of mitochondrial protein were loaded into 2 mL oxygraphy chambers containing Buffer D supplemented with ATP, creatine, phosphocreatine, and creatine kinase.

Cell culture and lines

Miro1 deficient C2C12 cells

Transgenic C2C12 cells were generated by expressing a lentiviral construct targeting pLKO.1 or shMiro1 at a multiplicity of infection of 50. Lentiviral particles were added to cell culture medium enriched with 0.5 μg/mL polybrene, added to cells dropwise, and incubated for 24 h. This process was repeated for 2 days, followed by puromycin selection (2 μg/mL) for 48 h. Transfection efficiency was confirmed by qPCR and Western blot. Cell experimentation was conducted on passages 3-7.

AKT inhibition

C2C12 cells at passage 4 were grown to 90% confluency in high glucose DMEM supplemented with 10% FBS and 1% Penicillin-Streptomycin. At 90% confluency, cells were washed with warm PBS and incubated with 5 µM MK2206 dihydrochloride in high-glucose DMEM for 2 h. 15 min before cells were collected, 1 μM of insulin was added, and cells were incubated at 37 °C for 15 min. Following incubation, cells were washed with ice-cold PBS and collected for Western blot analysis of pAKT, total AKT, and Miro1 protein expression.

Immunofluorescence

Passage 4 C2C12 were grown on 35 mm glass-bottom dishes until 50-60% confluency. Cells were washed 1× with warm PBS and incubated in high-glucose DMEM containing 1 μM insulin for 15 min at 37 °C. Cells were fixed with 4% paraformaldehyde for 30 min at room temperature, followed by a 20-min permeabilization with 0.1% Triton X-100 in PBS containing 100 nM glycine. Cells were blocked for 1 h in 5% goat serum in PBS, followed by overnight incubation with primary antibody at 1:200 in PBS at 4 °C. Cells were washed 3× with PBS, then incubated with secondary antibody 1:500 for 2 h at 4 °C. Cells were washed 3× with PBS, and 2 mL of PBS was added to the dish with 2 drops of NucBlue Fixed Cell ReadyProbes Reagent (DAPI). Cells were imaged using the Leica DMI6000 B microscope using 40× dry objective lens.

Glucose uptake

Intracellular glucose uptake was determined using a radiolabeled glucose uptake assay as described previously, with minor modifications⁷¹. Media was aspirated, and cells were gently washed twice with warm PBS, then incubated in serum-free low-glucose DMEM supplemented with 2% bovine serum albumin for 4 h prior to insulin stimulation. The media was aspirated, and the cells were gently washed twice with warm PBS. Cells were stimulated with vehicle media (Krebs-Ringer-Hepes (KRH) buffer (20 mM Hepes, 136 mM NaCl, 4.7 mM KCl, 1.25 mM MgSO₄·7H₂O, 1.25 mM CaCl₂, pH 7.40) with 2% bovine serum albumin, 2 mM Sodium Pyruvate, 32 mM D-Mannitol) or 1 μM insulin in vehicle media for 15 min at 37 °. After 15 min, 25 mM glucose was added to all wells, and cells were incubated for an additional 45 min. The stimulation medium was then aspirated, and the cells were gently washed twice with warm PBS. Cells were incubated with KRH containing 1 μCi/mL [3H]−2-deoxy-d-glucose and 2 mM 2-deoxy-d-glucose for 10 min at 37°. After 10 min of radiolabeled glucose uptake, cells were placed on ice, washed three times with ice-cold PBS, and then lysed with 0.1% sodium dodecyl sulfate (SDS) in ddH₂O for 30 min with gentle shaking. Lysates were then used for protein quantification and scintillation counting in Ultima Gold scintillation fluid.

Cellular fractionation

C2C12 myotubes were collected in ice-cold mitochondrial buffer plus protease and phosphatase inhibitors immediately following the insulin stimulation assay. Mitochondrial, cytosolic, and cellular membrane fractions were obtained as described above. Protein concentrations of each fraction were measured using BCA.

Western blotting

1–3 µg/µL of mitochondrial, cytosolic, and nuclear lysate were used for Western blotting as described above. Membranes were incubated overnight with primary antibodies (Table S3) and visualized using chemiluminescence as described above.

Quantitative RT-PCR

Cells were washed in ice-cold PBS prior to collection in ice-cold Trizol. RNA was isolated using the RNeasy Mini Kit according to the manufacturer’s instructions. CDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit with RNAse Inhibitor according to the manufacturer’s instructions. Knockdown efficiency was validated by quantitative PCR using targeted primers for Miro1:

Forward: CACAAGCTGACCTCAAGAGC

Reverse: GCAAAGACCGTAGCACCAAA

OXPHOS and ET capacity

OXPHOS and ET capacity were measured in C2C12 myoblasts as described above. Cells were cultured for 3–4 days until 90–100% confluency was reached as previously described. Cells were briefly washed with PBS, detached with 0.05% trypsin, and centrifuged at 600 × g for 6.5 min. Cells were resuspended in 2.5 mL of pre-warmed mitochondrial respiration buffer MiR05 (110 mM sucrose, 60 mM K+-lactobionate, 0.5 mM EGTA, 3 mM MgCl₂, 20 mM taurine, 10 mM KH₂PO₄, 20 mM HEPES adjusted to pH 7.1 with KOH at 37 °C; and 1 g/L de-fatted BSA) at a final concentration of 5 million cells/mL. 2 mL of cell suspension was added to oxygraphy chambers. Cells were permeabilized using 20 µg/mL digitonin, and maximal OXPHOS and ET capacity were determined as described above.

Quantification and statistical analyses

Statistical procedures and corresponding contrasts are detailed in each figure legend. Statistical significance was set at p = 0.05 unless otherwise stated. Two- or three-way comparisons of continuous variables were conducted via Student’s t-test or by analysis of variance, as appropriate. Abnormally distributed data were compared by non-parametric measures, including Mann–Whitney, Tukey’s HSD, and Welch’s ANOVA test, as appropriate based on the number of groups. Circumstances requiring more individualized data-handling procedures are detailed below. Human Study #1: Between-group differences were evaluated by a one-way ANOVA with 3 group fixed effects. Pairwise comparisons of significant main effects were assessed using Tukey’s HSD test. Distributional assumptions were evaluated using the Brown-Forsythe test. In the case of abnormally distributed data, Welch’s ANOVA test was employed with Dunnett’s T3 test for multiple comparisons. Data were analyzed using GraphPad Prism 10. Human Study #2: Differences in baseline characteristics were compared by linear mixed modeling. The influence of covariates on the clinical outcomes of interest was evaluated using a type III fixed effects test. Between-group differences in changes in outcomes over time following the intervention were evaluated using linear mixed modeling. Linear mixed modeling was chosen to account for the hierarchical structure of the data, with repeated measurements nested within individuals, while accommodating missing data under the assumption of missing-at-random. The primary model included fixed effects for group (intervention vs. control), time (as a categorical variable representing measurement occasions), and the group × time interaction. The interaction term was the primary parameter of interest, testing whether the intervention led to differential changes in the outcome over time relative to the control group. A random intercept for each participant was included to account for individual baseline differences, and a random slope for time was tested to assess individual variation in longitudinal trajectories. Model assumptions, including normality of residuals and homogeneity of variance, were assessed using graphical diagnostics and sensitivity analyses. Robustness checks included alternative covariance structures, the addition of baseline covariates, and multiple imputation for missing data where appropriate. Statistical analysis was conducted using SAS version 9.4. Lipid Droplet Architecture and Myofibril Assembly: Linear regression models were used to predict lipid droplet and myofibril characteristics. Individual differences were accounted for in a covariate model that included essential baseline characteristics (HbA_1c, sex, and age). For individual lipid droplet characteristics, statistical significance was set at p = 0.0001 due to the hundreds of lipid droplets per sample. Statistical analysis of the lipid droplet segmentation was performed in RStudio, version 4.1.1. For Fig. 3K, we employed a model-estimated percent change to account for missing data arising from limited muscle sample availability. To achieve this, we employed a generalized linear model to estimate the marginal mean at baseline and post, then derived the percent change from those estimates. This approach was chosen to maximize the available sample, which had fewer pairs than in the other analyses.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: C.L.A. and J.P.K. Methodology: A.L.T., J.M.S., B.V., H.A.P., B.G., and C.L.A. Software: R.A.B. and C.L.A. Validation: E.C.H., W.S.D., A.L.T., B.V., R.A.B., H.A.P., B.G., and C.L.A. Formal Analysis: E.C.H., W.S.D., A.L.T., H.A.P., B.G., and C.L.A. Investigation: E.C.H., W.S.D., E.R.M.Z., A.L.T., B.V., E.R.M.Z., K.P.B., J.M.S., J.T.M., R.A.B., H.A.P., P.A., B.G., C.L.A., and J.P.K. Resources: R.A.B., B.G., C.L.A., and J.P.K. Data Curation: E.C.H. and C.L.A. Writing - Original Draft: E.C.H. and C.L.A. Writing – Review & Editing: E.C.H., W.S.D., E.R.M.Z., A.L.T., J.M.S., B.V., E.R.M.Z., K.P.B., J.M.S., J.T.M., R.A.B., H.A.P., P.A., B.G., C.L.A., and J.P.K. Visualization: E.C.H., A.L.T., K.P.B., H.A.P., B.G., and C.L.A. Supervision: B.G., C.L.A., and J.P.K. Project Administration: C.L.A. and J.P.K. Funding Acquisition: J.P.K.

Peer review

Peer review information

Nature Communications thanks Cedric Moro and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This research was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (F31 DK139728 to ECH; R01 DK108089 to JPK), the National Institute on Aging (K99/R00 AG083239 to WSD), the National Center for Complementary and Integrative Health (T32 AT004094 to ERMZ), the National Heart, Lung, and Blood Institute (P01 HL094307 to PA; ZIA HL006221 to BG), and the National Institute of General Medical Sciences (U54 GM104940 to JPK).

Data availability

All data are available from the corresponding authors on request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77278-6.