Delivery of muscle-derived exosomal miRNAs induced by HIIT improves insulin sensitivity through down-regulation of hepatic FoxO1 in mice

Significance

The beneficial metabolic effects of exercise are mediated at least in part by the release of soluble factors by the muscles. Exosomes, small vesicles that facilitate the exchange of biological components among cells and tissues, may constitute one of these factors. Here, we show that exercise triggers the release of exosomes by the trained muscle, carrying a specific miRNA signature that induces gene expression changes in the liver, finally contributing to increased insulin sensitivity. Molecular characterization of exercise-induced exosomal miRNAs and their effects may drive the design of novel therapeutic strategies to alleviate insulin resistance and other aging-related ailments in an increasingly older society.

Abstract

Implementation of regular physical activity helps in the maintenance of a healthy metabolic profile both in humans and mice through molecular mechanisms not yet completely defined. Here, we show that high-intensity interval training (HIIT) modifies the microRNA (miRNA) profile of circulating exosomes in mice, including significant increases in miR-133a and miR-133b. Importantly, treatment of sedentary mice with exosomes isolated from the plasma of trained mice improves glucose tolerance, insulin sensitivity, and decreases plasma levels of triglycerides. Moreover, exosomes isolated from the muscle of trained mice display similar changes in miRNA content, and their administration to sedentary mice reproduces the improvement of glucose tolerance. Exosomal miRNAs up-regulated by HIIT target insulin-regulated transcription factor forkhead box O1 (FoxO1) and, accordingly, expression of FoxO1 is decreased in the liver of trained and exosome-treated mice. Treatment with exosomes transfected with a miR-133b mimic or with a specific siRNA targeting FoxO1 recapitulates the metabolic effects observed in trained mice. Overall, our data suggest that circulating exosomes released by the muscle carry a specific miRNA signature that is modified by exercise and induce expression changes in the liver that impact whole-body metabolic profile.

MicroRNAs (miRNAs) are small noncoding RNA molecules that modulate gene expression in the face of physiological and pathological stresses (1). As such, they are perfectly positioned to participate in the development of diseased states (2). Importantly, aside from their canonical function, miRNAs can be released by many cells and the profile of extracellular miRNAs is modified in association with diverse pathological conditions (3). Therefore, they can be used as biomarkers to improve diagnosis and/or prognosis of metabolic diseases (4). In this regard, we have shown that the abundance of specific extracellular miRNAs is altered in prediabetic subjects and obese, glucose-intolerant mice alike (5, 6).

Extracellular miRNAs can be found bound to protein complexes, associated with high-density lipoproteins or located inside a subtype of vesicles called exosomes (7). Exosomes are released by most cell types and can be captured by other cells, where the miRNAs that they carry may induce transcriptomic changes, in essence acting as novel intercellular messengers (8). Exosomal miRNAs from cancer cells participate in tumor progression, angiogenesis, and metastasis (9). Importantly, recent studies evidence that exosomal miRNAs, principally those released by the adipose tissue, regulate insulin sensitivity in other tissues (8, 10). We have recently described alterations in exosomal miRNAs in obesity, which mediate the development of metabolic disease in mice by disturbing the communication between the liver and the adipose tissue (6). Hence, if tissue cross-talk mediated by exosomal miRNAs is central to the maintenance of glucose homeostasis, they may also participate in the establishment of beneficial metabolic patterns, for instance by mediating the systemic adaptations to exercise.

The implementation of regular physical activity improves cardiorespiratory fitness (CRF) and helps maintain metabolic homeostasis (11). High CRF is a predictor of lower type 2 diabetes incidence and increased longevity (12). An array of circulating factors collectively denominated “exerkines” are secreted by the muscles during exercise and have been proposed as candidates to modulate the function of other cell types, thus improving the metabolic profile in both humans and mice (13). Interestingly, exercise also increases the number of exosomes in blood and modifies the profile of circulating miRNAs (14, 15). These changes may contribute to the systemic signaling associated with the adaptation to exercise (16).

Recently, several studies have found that high-intensity interval training (HIIT), an exercise modality involving brief bouts of intense activity followed by periods of recovery, can elicit similar metabolic adaptations to classical endurance exercise training in humans (17). HIIT can improve CRF (18), glycemic control (19), and insulin sensitivity (17), all of them potential risk factors for the development of metabolic disease. The fact that exosomal miRNAs have been shown to respond differently as a function of the exercise modality and intensity, coupled with the lack of characterization of the underlying mechanisms of the beneficial effects of HIIT prompted us to investigate the role of exosomes in the context of the metabolic response to HIIT.

Hence, we wanted to determine if exosomal miRNAs may convey part of the beneficial effects of HIIT by modifying the function of target cells to enhance insulin sensitivity in peripheral tissues. Here, we show that HIIT triggers the release of circulating exosomes by the muscle, carrying a specific miRNA signature that induces gene expression changes in the liver, finally improving metabolic homeostasis.

Results

HIIT Increases Exercise Capacity and Improves the Metabolic Profile of Mice.

To explore the effects of exercise upon CRF and metabolism, male C57BL/6J mice were subjected to a HIIT regime for 5 wk. As expected (20), trained mice (RUN) increased the distance traversed (Fig. 1A) and the peak oxygen uptake (Fig. 1B) during a capacity test, as compared to their sedentary control littermates (CT).

Fig. 1.

HIIT increases exercise capacity and improves the metabolic profile of mice. (A and B) Distance run (A) and peak oxygen uptake (B) of sedentary mice (CT) and mice trained under a HIIT regime for 5 wk (RUN) during a capacity test. (C–F) GTT (C), plasma TGs (D), ITT (E), and FFA concentration (F) of the same mice at the end of the training period. Data are presented as mean ± SEM n = 4 (A and B); n = 6 (C and E); n = 19 CT and n = 24 RUN (D); n = 11 CT and n = 14 RUN (F) *P < 0.05, **P < 0.01, ****P < 0.001, Student’s t test.

Moreover, RUN mice displayed greater glucose tolerance (Fig. 1C) demonstrated by a lower glycemia area under the curve (AUC) obtained from a glucose tolerance test (GTT) (SI Appendix, Fig. S1A). RUN mice also presented lower plasma levels of triglycerides (TGs) (Fig. 1D) and greater lipid tolerance (SI Appendix, Fig. S1B) as measured by a TG tolerance test (TTT). Altogether, this new metabolic profile leads to greater insulin sensitivity (Fig. 1E) according to an insulin tolerance test (ITT), with a significant decrease of basal insulin levels during the GTT (SI Appendix, Fig. S1C).

RUN mice also presented marginally increased circulating free-fatty acids (FFAs) (Fig. 1F) which may be due to the enhanced responsiveness of the adipose tissue to lipolytic cues resulting from training (21). However, the adipose tissue of RUN mice was more sensitive to the suppressive action of insulin (22), and FFA plasma levels of RUN mice reached similar values as those of CT mice 15 min after insulin injection during the ITT (SI Appendix, Fig. S1D). Accordingly, we also observed a significant decrease in body weight (SI Appendix, Fig. S1E), which was mostly explained by a smaller epididymal white adipose tissue (eWAT) (SI Appendix, Table S1). Overall, these findings are suggestive of improved whole-body insulin sensitivity following exercise training and, accordingly, the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) index for the trained mice was significantly lower (SI Appendix, Fig. S1F).

Circulating Exosomes from RUN Mice Improve Glucose and Lipid Metabolism in Sedentary Mice.

Release of soluble factors including exosomes from skeletal muscle has been proposed to mediate the systemic adaptations to exercise (16). Hence, we isolated exosomes from equal amounts of plasma from CT and RUN mice 48 h after the last exercise session. Vesicle morphology was characterized by transmission electron microscopy (Fig. 2A), size distribution was determined by nanoparticle tracking analysis (Fig. 2B), and presence of well-known exosome marker proteins was detected by Western blotting (Fig. 2C). We did not observe differences in exosome number between CT and RUN mice as estimated from esterase activity levels (SI Appendix, Fig. S2A).

Fig. 2.

Circulating exosomes from RUN mice improve glucose and lipid metabolism in sedentary mice. (A–C) Representative electron micrograph (A), size distribution determined by NTA (B), and Western blots of CD63, CD9, and HSP90 (C) of exosomes isolated from plasma of CT and RUN mice. (D) Liver (Upper) and muscle (Lower) sections of control mice 6 h after i.v. injection with PBS (Left) or PKH26-labeled exosomes (Right). (E–H) GTT of sedentary mice after 4 wk of biweekly injections with exosomes isolated from either CT (EXO-CT) or RUN (EXO-RUN) mice plasma (E), plasma TGs (F), FFA concentration (G), and ITT (H) in the same mice. (I–K) RER (I), VO₂ uptake (J), and EE (K) of CT, RUN, and EXO-RUN mice obtained by indirect calorimetry. Data are presented as mean ± SEM n = 2 (A–D); n = 5 (E and H); n = 12 (F and G); n = 4 (I–K); *P < 0.05, ****P < 0.001 with respect to their respective control groups, Student’s t test.

We first studied if exogenously administered exosomes could be captured by acceptor cells by injecting control mice with PKH26-labeled exosomes intravenously (i.v.) through the tail vein. As we and others have shown before (6, 10), fluorescent signals were evident in different tissues, including liver and skeletal muscle, 6 h after the injection (Fig. 2D).

To assess the influence of exosomes upon energy metabolism, we next injected sedentary mice with exosomes isolated from plasma of either sedentary (EXO-CT) or trained mice (EXO-RUN). We found a positive correlation between total exosome protein and exosome number for both CT and RUN mice (SI Appendix, Fig. S2B). Hence, i.v. injections of 15 µg exosomes, corresponding to ∼5 × 10⁹ vesicles, took place biweekly during 4 wk. Surprisingly, EXO-RUN mice showed improved glucose tolerance (Fig. 2E and SI Appendix, Fig. S2C) and decreased plasma levels of TGs (Fig. 2F) with improved lipid tolerance (SI Appendix, Fig. S2D). However, we did not observe differences in the levels of FFAs (Fig. 2G). Moreover, EXO-RUN mice were significantly leaner than their EXO-CT counterparts with smaller eWAT fat pads (SI Appendix, Table S1). Finally, this treatment led to greater insulin sensitivity (Fig. 2H) as in the case of RUN mice.

Activation of fatty acid oxidation (FAO), for instance, by increasing physical activity, may constitute a therapeutic target to treat insulin resistance (23). Substrate use can be assessed by measuring the respiratory exchange ratio (RER). We assessed the influence of both exercise and the exosome treatment upon energy balance through indirect calorimetry. Access to food was restricted to the night period in order to decrease the variability inside the groups. No changes were observed in either cumulative food intake (SI Appendix, Fig. S2E) or total activity (SI Appendix, Fig. S2F) in any group of mice. Moreover, we did not detect changes in the RER in either the RUN or the EXO-RUN mice (Fig. 2I), indicating unaltered substrate preference. However, significant elevations in VO₂ uptake (Fig. 2J), CO₂ production (SI Appendix, Fig. S2G), and, consequently, energy expenditure (EE) (Fig. 2K) were detected only in RUN mice, thus suggesting increases in the total rate of both FAO and carbohydrate oxidation (CHOx).

Overall, these data suggest that HIIT induces a better glycemic control by improving insulin sensitivity without changes in the RER but with a significant increase in VO₂ uptake. Importantly, the treatment with exercise-induced exosomes recapitulates the metabolic benefits without increasing EE.

HIIT Modifies the miRNA Content of Circulating Exosomes in Mice.

To identify the bioactive molecules responsible for the metabolic effects of exercise-induced exosomes, we focused on the miRNA content because exosomal miRNAs have been shown by us and others to participate in the establishment of metabolic phenotypes (6, 10). We isolated total RNA from plasma exosomes of RUN mice and transfected it into exosomes isolated from control mice. Hence, we generated a pool of control exosomes enriched in exercise-associated miRNAs. With this strategy, we aimed to isolate the effect of the miRNAs, discarding other sources of variation such as the protein content or lipid cover of the exosomes. As a control, we also transfected total RNA isolated from plasma exosomes of CT mice (SI Appendix, Fig. S3A). Sedentary mice were then i.v. injected biweekly during 2 wk with these transfected exosome preparations. Importantly, mice injected with control exosomes enriched with miRNAs derived from trained mice (RUNmiR) recapitulate the improvement of glucose tolerance observed in RUN and EXO-RUN mice (Fig. 3A and SI Appendix, Fig. S3B). These data suggest that the exosomal miRNA content is modified by exercise and participates in the establishment of a beneficial metabolic profile.

Fig. 3.

HIIT modifies the miRNA content of circulating exosomes in mice. (A) GTT in sedentary mice after 2 wk of injections of exosomes transfected with miRNAs isolated from plasma exosomes of CT (CTmiR) or RUN mice (RUNmiR). (B) Volcano plot of RT-PCR profiling of the miRNA content of plasma exosomes of CT and RUN mice. (C) Heat map showing clustering of exosomes from CT and RUN mice according to the most up- and down-regulated miRNAs in exosomes of RUN mice. (D) Heat map showing miRNAs with significant positive (red) or negative (green) correlations to miR-133b. Data are presented as mean ± SEM n = 3 (A); n = 4 (B–D); *P < 0.05, Student’s t test.

RT-PCR profiling of exosomes showed up-regulation of six miRNAs in exosomes from RUN mice, whereas only one was significantly down-regulated (Fig. 3B and SI Appendix, Fig. S3C and Table S2). Interestingly, among the up-regulated miRNAs we identified miR-133a, miR-133b, and miR-206, three highly muscle-specific miRNAs (myomiRs). On the other hand, we observed down-regulation of miR-192 (SI Appendix, Table S2), an exosomal miRNA that we recently described as up-regulated in obesity (6). Heat map representation and principal component analysis of the most up-regulated and down-regulated miRNAs showed separated clustering of exosomes from CT and RUN mice (Fig. 3C and SI Appendix, Fig. S3D). In addition, we performed a correlation analysis between each pair of miRNAs across all samples and selected those miRNAs with significant correlation to miR-133b (P < 0.05). Heat map representation of those miRNAs suggests that exercise drives the release of a novel population of exosomes into the circulation (Fig. 3D).

Muscle-Derived Exosomes Are the Main Contributors to the Exercise-Induced Circulating Exosomal miRNA Profile.

An important question is to identify which tissue is the main source of these circulating exosomes. To address this point, we isolated interstitial exosomes from liver, muscle, and eWAT of CT and RUN mice by adapting a recently described strategy (24). Representative electron micrographs showed successful isolation of interstitial exosomes (Fig. 4A). We used RT-PCR to measure the expression of candidate miRNAs in these exosomes as well as the total tissues. Heat map representation showed that exosomes from muscle were enriched in miR-133a and miR-133b as compared to exosomes from liver or eWAT (Fig. 4B). Moreover, the level of these miRNAs was even higher in the case of exosomes from the muscle of RUN mice. This differential expression in exosomes followed a pattern similar to that observed in total tissues (Fig. 4C). Interestingly, we also measured the levels of two of the miRNAs most up-regulated in obesity (6), miR-122 and miR-192, both of which had a high liver expression, in both the tissue and exosomal fractions, and were decreased with exercise.

Fig. 4.

Muscle-derived exosomes are the main contributors to the exercise-induced circulating exosomal miRNA profile. (A) Electron micrographs of exosomes isolated from eWAT, liver, and muscle from RUN mice. Arrows indicate representative exosomes. (B and C) Heat maps showing differential expression of selected miRNAs among exosomes isolated from muscle, liver, and eWAT (B) or the total tissues (C) from CT and RUN mice. (D and E) GTT in sedentary mice after 4 wk of injections of liver-derived exosomes (liver exoCT and liver exoRUN) (D) and muscle-derived exosomes (liver exoRUN and muscle exoRUN) from the same mice (E). Data are presented as mean ± SEM n = 2 (A); n = 6 (B); n = 4 (C); n = 3 (D and E); *P < 0.05 with respect to their respective control groups, Student’s t test.

To determine which population of exosomes is involved in the observed metabolic effects, sedentary mice were injected biweekly during 4 wk with either liver- or muscle-derived exosomes from CT and RUN mice. After the treatment, mice injected with exosomes derived from trained muscle (muscle exoRUN) were the only ones showing a significant effect, mimicking the improvement of glucose tolerance observed in RUN and EXO-RUN mice (Fig. 4 D and E and SI Appendix, Fig. S4A). The total number of interstitial exosomes is the same in CT and RUN mice for each tissue (SI Appendix, Fig. S4B). Overall, these data suggest that the exosomes released by the trained muscle may be the principal agents involved in the establishment of a better response to glucose.

Treatment with Exercise-Induced Exosomal miRNAs Results in Down-Regulation of Hepatic FoxO1.

Our data suggest that part of the beneficial effects of HIIT upon metabolism may be mediated by the up-regulated exosomal miRNAs modifying the expression of target genes in peripheral tissues. Therefore, we used bioinformatics tools to identify the main pathways affected by them (SI Appendix, Table S3). Network analysis provided different pathways converging on insulin-regulated transcription factor forkhead box O1 (FoxO1) (Fig. 5A). Notably, a comprehensive prediction of miRNA target sites using miRmap reported binding sites for most of the up-regulated miRNAs in the 3′ UTR of FoxO1 (25, 26) (SI Appendix, Fig. S5A). In particular, we identified two target sites for the miR-133 family. There is no prior evidence of a direct regulation of FoxO1 by miR-133 in the literature. Hence, we transfected 3T3-L1 cells with a miR-133b mimic. RT-PCR analysis evidenced significant down-regulation of FoxO1 expression (SI Appendix, Fig. S5B), thus suggesting that FoxO1 is a direct target of miR-133 family miRNAs.

Fig. 5.

Treatment with exercise-induced exosomal miRNAs results in down-regulation of hepatic FoxO1. (A) Network showing the interaction of some of the candidate target genes of exosomal miRNAs up-regulated after HIIT. (B and C) FoxO1 expression in the eWAT (B) and gastrocnemius (C) of CT, RUN, EXO-CT, and EXO-RUN mice. (D) Representative immunostaining of GLUT4 (green) in muscle sections of the same mice. (E and F) Expression of FoxO1 (E) and a selection of its targets in the liver of the same mice (F). (G and H) Representative Oil Red staining (G) and FFA content in liver of the same mice (H). Data are presented as mean ± SEM n = 4 in RUN model and n = 6 in EXO model (B and C); n = 4 (D and G); n = 8 (E and F); n = 16 in RUN model and n = 6 in EXO model (H); *P < 0.05, **P < 0.01, ***P < 0.005 with respect to their respective control groups, Student’s t test.

Intriguingly, modulation of FoxO1 is a potential therapy to improve insulin sensitivity (27). Hence, we analyzed FoxO1 gene expression in both RUN and EXO-RUN mice. In the eWAT, we did not observe any changes (Fig. 5B). Moreover, hematoxilin and eosin (H&E) staining evidenced that adipocyte size did not differ between RUN or EXO-RUN mice and their respective controls (SI Appendix, Fig. S5C).

Interestingly, FoxO1 expression remained unchanged in gastrocnemius muscle of EXO-RUN mice, whereas it was significantly decreased in RUN mice (Fig. 5C). In addition, we observed an important increase in glycolytic capacity as evidenced by glycerophosphate dehydrogenase (GPD) staining at the expense of decreased oxidative capacity measured by succinate dehydrogenase (SDH) staining only in the muscle of RUN mice (SI Appendix, Fig. S5D). These results are in line with previous work showing that HIIT increases the glycolytic capacity in gastrocnemius muscles (28) and induces a nearly complete slow-to-fast fiber type transformation (29). Accordingly, we also observed increased membrane staining for the insulin-sensitive glucose transporter GLUT4 in muscle sections of RUN but not EXO-RUN mice (Fig. 5D). Altogether, these data suggest that HIIT improves muscle glucose uptake, but this feature is not transmitted by the treatment with exercise-induced exosomes.

Importantly, the analysis of hepatic gene expression showed down-regulation of FoxO1 (Fig. 5E) and several of its target genes (Fig. 5F) in both RUN and EXO-RUN mice. Moreover, decreased FoxO1 expression was associated with enhanced accumulation of hepatic lipid droplets (Fig. 5G) as demonstrated by Oil Red staining and increased FFA content in both animal models (Fig. 5H). Hence, exosomes derived from exercise do not affect muscle FoxO1 expression or its metabolic features but induce robust gene expression changes in liver.

miR-133b Modulates Hepatic Gluconeogenesis Both In Vitro and In Vivo.

Exercise is generally associated with increased glucose uptake and use by the muscles (30). In order to test if this is also the case for HIIT, we decided to measure changes in RER during a GTT in CT and RUN mice. Interestingly, RUN mice displayed significantly increased RER values during the test (Fig. 6A), indicative of a higher level of glucose oxidation. On the other hand, according to a pyruvate tolerance test (PTT), RUN mice achieved lower levels of gluconeogenesis (GNG) than CT mice (Fig. 6B), suggestive of increased hepatic insulin sensitivity. However, we did not observe changes in AKT phosphorylation in RUN mice after insulin injection (SI Appendix, Fig. S6A). This would be in accordance with the sensitizing point being downstream of AKT, as would be expected for the observed decrease in FoxO1 expression (31).

Fig. 6.

miR-133b modulates hepatic GNG both in vitro and in vivo. (A and B) RER values during a GTT (A) and PTT (B) performed in CT and RUN mice. (C) Glucose production and FoxO1 expression in primary hepatocytes transfected with a nontargeting siRNA (siC), a siRNA against FoxO1 (siFoxO1), a miR-133b mimic (miR-133b), or incubated with exosomes isolated from plasma of CT and RUN mice (EXO-CT and EXO-RUN). (D) Glucose production and FoxO1 expression in primary hepatocytes transfected with siC or siFoxO1, and subsequently with a cel-miR-39-3p mimic as a negative control or miR-133b. (E–G) PTT (E), GTT (F), and ITT (G) in sedentary mice after 4 wk of injections of control exosomes loaded with negative control cel-miR-39-3p or miR-133b mimic. Data are presented as mean ± SEM n = 6 (A and B); n = 5 (C and E–G); n = 3 (D); *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001, Student’s t test.

To further explore the effects of exercise-induced exosomal miRNAs upon hepatic metabolism, we developed an in vitro approach (SI Appendix, Fig. S6B). Primary hepatocytes were either incubated in the presence of exosomes isolated from the plasma of CT or RUN mice, or directly transfected with a miR-133b mimic or a siRNA against FoxO1 as a positive control. Twenty-four hours afterward, media were changed to serum-free medium overnight, and then the cells were challenged for 6 h with a medium containing the GNG substrates lactate and pyruvate, and the cAMP analog pCT-cAMP (32). Glucose levels measured in the media after the incubation reflect the GNG potential of the hepatocytes exposed to the different treatments. Interestingly, both the incubation with RUN exosomes and the transfection of miR-133b resulted in significantly decreased glucose production similar to the effects of knocking down FoxO1 (Fig. 6 C, Left). Importantly, this effect was accompanied by significantly decreased FoxO1 expression (Fig. 6 C, Right). Still, miR-133b and FoxO1 could be exerting their effects through parallel pathways. To provide further evidence that mir-133b is acting through the FoxO1 pathway, we repeated the experiment but by overexpressing mir-133b in FoxO1 knocked-down hepatocytes (SI Appendix, Fig. S6C). Importantly, the combination of miR-133b with siFoxO1 had a similar effect upon glucose production and FoxO1 expression as each factor separately (Fig. 6D).

Overall, these data suggest that exercise-induced exosomal miRNAs decrease hepatic FoxO1 expression, resulting in decreased GNG and improved glucose tolerance. To test this hypothesis, we transfected control exosomes with a miR-133b mimic and treated sedentary mice with this preparation with biweekly injections for 4 wk. Similar to RUN mice, a PTT showed decreased GNG levels in the mice treated with the miR-133b mimic (Fig. 6E), and, as expected, the treatment resulted in improved glucose tolerance (Fig. 6F and SI Appendix, Fig. S6D) and increased insulin sensitivity (Fig. 6G). Again, as observed for the EXO-RUN mice, no differences were detected in GLUT4 staining in muscle (SI Appendix, Fig. S6E). These data further support the hypothesis that, different to the situation in the RUN mice, where improved glucose tolerance is the result of a combination of increased glucose use and decreased glucose production, the improved metabolic response in mice treated with exosomal miRNAs is mainly due to their hepatic effects.

Experimental Down-Regulation of Hepatic FoxO1 Recapitulates the Effects of Exosomal miRNAs upon Metabolism.

We hypothesized that decreased hepatic FoxO1 expression could explain part of the phenotype of RUN and EXO-RUN mice. Hence, we used an alternative strategy by injecting sedentary mice with control exosomes transfected with siFoxO1 or a nontargeting siRNA as a control. After the treatment, we detected decreased expression of both FoxO1 (SI Appendix, Fig. S7A) and its target gene G6pc (SI Appendix, Fig. S7B) and a reduction of FOXO1 immunostaining (SI Appendix, Fig. S7C) in liver. Remarkably, siFoxO1 mice presented improved glucose tolerance (Fig. 7A and SI Appendix, Fig. S7D) and insulin sensitivity (Fig. 7B), with significantly decreased circulating levels of TGs (Fig. 7C). Similar to RUN and EXO-RUN mice, we observed a concurrent increase in hepatic FFAs (Fig. 7D) and lipid content (Fig. 7E) akin to what has been described in genetic models of FoxO1 silencing (31, 33) which are also more insulin sensitive than their controls. Altogether, these data show that treatment with siFoxO1-transfected exosomes decreases FoxO1 expression in liver, and this is enough to recapitulate part of the metabolic profile of RUN mice and EXO-RUN mice (Fig. 7F).

Fig. 7.

Experimental down-regulation of hepatic FoxO1 recapitulates the effects of exosomal miRNAs upon metabolism. (A–C) GTT (A), ITT (B), and plasma TG (C) in sedentary mice after 4 wk of injections of control exosomes loaded with a nontargeting siRNA (siC) or a siRNA against FoxO1 (siFoxO1). (D and E) Representative Oil Red staining (D) and FFA content (E) in liver of the same mice. (F) Proposed model. Data are presented as mean ± SEM n = 5 (A, C, and E) n = 4 (B and D); *P < 0.05, **P < 0.01, ****P < 0.001, Student’s t test.

Discussion

Our results showing improved CRF and insulin sensitivity in mice trained under a HIIT protocol are in accordance with previous evidence in humans (17–19). Surprisingly, here we show that treatment with exosomes isolated from trained mice is able to improve glucose tolerance and insulin sensitivity in sedentary mice to a level similar to that of exercise itself.

Age and dietary excess are both characterized by insulin resistance derived from excessive fat storage due to either decreased mobilization or increased availability, respectively (34). Activation of FAO, for instance by increasing physical activity, may constitute a therapeutic target to treat insulin resistance (23). However, there is controversy as to whether HIIT increases FAO. A number of studies describe significant increases in FAO, concurrent or not with increases in CHOx, in response to HIIT (35, 36) but other reports contradict these findings (37). We did not detect changes in substrate preference after HIIT. However, EE was significantly elevated in RUN mice, thus indicating increases in the total rate of both FAO and CHOx. Our treatment with exosomes from trained mice was able to recapitulate the metabolic effects of exercise but not the increase in EE. In line with these results there is controversy as to whether EE can be increased in the absence of increased physical activity (38). If this were the case, the attainment of biochemical means to mimic the effects of exercise without actually increasing physical activity would be strongly curtailed. Still, finding the pathways mediating some of the positive responses to exercise may protect patients from metabolic decline even when they are unable to exercise (38), an important objective in a progressively aging society.

Different reports have demonstrated that exercise induces the release of exosomes enriched in specific miRNAs following different time courses, depending on the exercise type (39). Although some studies have shown increased exosome number with exercise (39), we did not observe this effect. This discrepancy may be due to either the exercise modality or the timing of sample collection. Our data show that the circulating exosome pool in trained mice is significantly enriched in vesicles released by the skeletal muscle, in agreement with previous reports (40). In particular, we observed up-regulation of miR-133a, miR-133b, and miR-206, three myomiRs whose circulating levels in vesicles have been shown by others to be increased in correlation with VO₂ max in humans (41). Moreover, several of these miRNAs have already been linked with improved insulin sensitivity in different animal models (25, 26).

We identified binding sites for most of the up-regulated miRNAs in the 3′ UTR of FoxO1. Conditional FoxO1 ablation in skeletal muscle results in increased formation of MyoD-containing fast fibers and altered fiber type distribution at the expense of myogenin-containing slow fibers (42). Decreased expression of FoxO1 in the gastrocnemius muscle after HIIT has already been shown in rats (43). Remarkably, we detected muscle FoxO1 down-regulation only in trained mice, which may indicate that this adaptation is dependent on the direct effect of exercise and not mediated by exosomal miRNAs. Conversely, lack of effect of the exogenous exosomes upon muscle FoxO1 expression could be due to the specific route of administration. However, exogenous PKH26-labeled exosomes reach the muscle. Additionally, in a previous experiment we injected exosomes transfected with nonmammalian miRNA, cel-miR-39-3p (6) and were able to detect its presence by RT-PCR in muscle (Ct 30.5 ± 0.1), liver (Ct 26.7 ± 0.2), and eWAT (Ct 34.3 ± 2.9). Overall, these data would argue that the exosomes, albeit with a preference for the liver, reach the muscle, but the exercise-induced miRNAs do not affect the expression of, at least, our identified target FoxO1. It is important to note that, in contrast to the pharmacological-like effects of siRNAs, the effects of miRNAs are largely context dependent (44), and the miR-133 family is expressed at high levels in the muscle, while being almost absent in liver (Fig. 4C). Hence, it is not surprising that we observe a higher effect in the latter tissue. In addition, we detected other changes in the muscle of RUN mice that are not mirrored in the EXO-RUN mice, such as increased GLUT4 expression and GPD activity. We postulate that these changes, as well as the enhanced EE of RUN mice, are independent of the exercise-induced exosomal miRNAs and dependent on exercise itself.

Exercise-derived exosomes have been shown to display liver tropism when i.v. administered (45). Accordingly, we observed down-regulation of hepatic FoxO1 and several of its target genes in both RUN and EXO-RUN mice. Hepatic glucose production is finely controlled by insulin signaling that dampens FoxO1 and reins in gluconeogenesis (46), although at the expense of increased de novo lipogenesis leading to hepatic lipid storage (33). This same phenotype of enhanced insulin sensitivity and hepatic fat accumulation is reproduced in mice injected with exosomes transfected with siFoxO1, in accordance with previous reports showing that FoxO1 knockdown is a possible therapeutic strategy for improving insulin resistance (47). However, due its bivalent effects, alternative approaches, such as selective inhibition of specific FoxO1 target genes have been proposed to lower glycemia without concurrent steatosis (48).

Altogether, our data indicate that HIIT improves glucose tolerance both by increasing glucose use, as measured by the higher RER values during a GTT, and by decreasing endogenous glucose production, as evidenced by a PTT. In this sense, HIIT has been recently shown to improve hepatic insulin sensitivity and decrease endogenous glucose production in insulin-resistant humans (49). For this reason, HIIT is becoming the preferred exercise modality to improve glucose levels in prediabetic patients with impaired fasting glucose, which are mainly characterized by reduced hepatic insulin sensitivity (50–52). Importantly, treatment with exercise-induced exosomes or direct transfection of miR-133b decrease FoxO1 expression and hepatocyte glucose production in vitro. Moreover, transfection of miR-133b in FoxO1 knocked-down hepatocytes did not further decrease glucose production than FoxO1 knockdown alone. This, together with the fact that miR-133b is able to decrease FoxO1 expression in 3T3-L1 cells and in primary hepatocytes suggest that the miRNA is exerting at least part of its effects through FoxO1. However, we cannot discard that additional miR-133b targets are also weighing in, as miRNAs regulate a wide range of genes. Finally, mice treated with miR-133b transfected exosomes also show improved glucose tolerance associated with decreased GNG.

Overall, we show that HIIT triggers the release of circulating exosomes by the trained muscle, carrying a specific miRNA signature that induces gene expression changes in the liver, leading to an improved metabolic profile. Our data suggest that exosomes from athletes or control exosomes bioengineered to incorporate specific miRNAs or short interfering RNAs (siRNAs) may be useful in the treatment of metabolic disorders (38).

Materials and Methods

Experimental Animal Models.

C57BL/6J male mice were used in all experiments. RUN mice were subjected to a HIIT protocol for 5 wk (53). EXO-RUN mice were injected i.v. through the tail vein with a 100 μL phosphate-buffered saline (PBS) suspension containing 15 μg exosomes isolated from plasma of either sedentary or trained mice. EXOmiR mice were injected with 25 μg exosomes isolated from plasma of control mice and transfected with total RNA isolated from 7 μg of either sedentary or trained plasma exosomes. iEXO mice were injected with 7 μg interstitial exosomes isolated from liver or muscle of either sedentary or trained mice. miR-133b mice were injected with 25 µg exosomes transfected with 125 pmol miR-133b mimic. siFoxO1 mice were injected with 25 μg exosomes transfected with 190 pmol of siFoxO1 or a nontargeting siRNA. For the biodistribution studies, mice were injected with 100 μg exosomes labeled with the fluorescent dye PKH26 and killed 6 h afterward. Metabolic tests were performed as described (6). Indirect calorimetry was carried out using a 16-chamber TSE PhenoMaster monitoring system (54). Studies were conducted in accordance with principles of laboratory animal care (register: 46/18).

Exosome Characterization.

Exosomes were isolated from mouse plasma as described (6). Interstitial exosomes were isolated by adapting a recently described protocol (24). Exosome transfections were performed with the Exo-Fect exosome transfection reagent as described (6) using the miRNA mimics and siRNAs described in SI Appendix, Table S4.

Cell Culture and In Vitro Experiments.

The 3T3-L1 cells and primary hepatocytes were transfected with 60 nM siFoxO1 and 30 nM miR-133b-3p either in parallel in separate wells or sequentially in the same well (SI Appendix, Table S4) by using Metafectene Pro (6). Primary hepatocytes were incubated with 7 μg/mL exosomes isolated from CT or RUN mice. Hepatocyte glucose production was determined by measuring glucose concentration in the media as described (32).

RNA Analysis.

RNA was extracted from 10 μL plasma exosomes or 30 μL interstitial exosomes with miRNeasy Mini Kit and retrotranscribed with miRCURY LNA Universal RT Kit. Profiling of 378 miRNAs was performed by RT-PCR. For tissue samples, RNA was extracted with TRI reagent and analyzed by RT-PCR using the High-Capacity RT Kit and homemade primers (SI Appendix, Table S4).

Protein Analysis.

Protein lysates were resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS/PAGE) and visualized by blotting with horseradish peroxidase (HRP)-conjugated secondary antibodies. For histochemistry, liver and muscle were frozen in liquid N₂-cooled isopentane and embedded in optimal cutting temperature (OCT) compound. eWAT was fixed with 4% paraformaldehyde (PFA) and embedded in paraffin. H&E and Oil Red staining were performed by following the protocols at IHCWorld. Gastrocnemius sections were stained as described (55). For GLUT4 immunohistochemistry, muscle cryostat sections were stained with 1:100 anti-GLUT4 (1F8) antibody. For FOXO1 immunohistochemistry, liver sections were stained with 1:100 FOXO1 (C29H4) antibody.

Statistical Analyses.

Differences between groups were determined by either t test analysis when only two groups were compared or by one-way ANOVA with t test analysis for the pairwise comparison of three or more groups with different number of values. Asterisks indicate significance with respect to control group, unless otherwise specified. Correlation analyses were performed by Pearson regression.

Data Availability.

All study data are included in the article and SI Appendix.