Plasma Extracellular Vesicle Signatures of Metabolic Health and Exercise Response in a Pilot Study of Older Adults

Abstract

Extracellular vesicles (EVs) are key mediators of intercellular communication and regulators of cellular function, yet their roles in metabolic health and exercise response are poorly understood. This pilot study analyzed plasma from older adults (n=20) in subgroups of the well-characterized STRRIDE study to evaluate plasma EV biomarkers as minimally invasive biomarkers of metabolic health and exercise responsiveness. Plasma EVs comprised highly heterogeneous subpopulations defined by diverse surface markers reflecting complex cellular origins. At baseline, multiple EV biomarkers related to immune cells, skeletal muscle, and mesenchymal stem cells were associated with better indices of insulin action, including nine EV subpopulations with lower fasting insulin concentration and eight with lower Homeostatic Model Assessment for Insulin Resistance (HOMA-IR). Low-amount (~1300 kcal/week), vigorous-intensity (65–80% peak VO₂) aerobic exercise increased the FABP4⁺ EV subpopulation in older adults (n=12). High-amount (~2200 kcal/week), vigorous-intensity exercise increased fifteen EV subpopulations in older adults (n=8). These subpopulations arise from a variety of cell sources, including immune cells (primarily lymphoid cells), skeletal and cardiac muscle, erythroid cells, mesenchymal and hematopoietic stem cells. Notably, eight out of fifteen high-amount exercise-induced EV subpopulations were insulin action-related (CD29⁺, CD8⁺, CD56⁺, CD19⁺, MCAD⁺, CD73⁺, CD105⁺ CD235a⁺). The EV-based profiling platform established here is ready for validation in larger human exercise cohorts, including the full STRRIDE cohort.

NEW & NOTEWORTHY

Specific plasma EV biomarkers related to immune subsets, skeletal muscle, and mesenchymal stem cells were associated with better indices of insulin action at baseline. High-volume, vigorous-intensity aerobic exercise increased many of these insulin action-related EV subpopulations. We developed a novel, minimally invasive platform that uses plasma EV surface markers to assess metabolic health and exercise responsiveness. This platform is ready for validation in larger human cohorts, including the full STRRIDE cohort.

Graphical Abstract

INTRODUCTION

Extracellular vesicles (EVs) are nanosized particles that carry surface markers reflecting their parent cell origins and the physiological state of their parent cells(1, 2). EVs also carry cargo effectors such as proteins, peptides, DNA, messenger RNAs, small RNAs (smRNAs), and mitochondria from their cells of origin(2–8). EVs represent a key class of messengers that mediate intercellular communication and regulate cellular function. They carry out these roles by delivering bioactive cargo effectors to recipient cells or through surface ligand–receptor interactions(3, 4, 7, 9–11). For example, EVs contribute to the regulation of metabolic homeostasis by mediating crosstalk between pancreatic beta cells (β-cells) and other tissue types, such as skeletal muscle, through the transfer of their smRNA cargo(10, 12). Several plasma-derived EV subpopulations, particularly those originating from immune and muscle cells, are associated with longevity in older adults(2). Exercise interventions boosting the production of these health-promoting EVs may promote metabolic health and extend lifespan in older adults.

EVs are released by all cell types through multiple biogenesis pathways(13, 14), resulting in a heterogeneous mix of EV subpopulations in peripheral blood(2). Some plasma EV subpopulations are linked to healthy aging and longevity(2, 9), while others are associated with pathological aging processes such as osteoarthritis(15, 16). Although these associations are intriguing, the biological roles of exercise-induced EVs are poorly understood, especially with respect to metabolic health. Given the inherent heterogeneity in the phenotype and effects of plasma EVs, to harness EVs for therapeutic benefit, it is critical to identify the specific EV biomarkers involved in promoting health, or underlying diseases, and to understand their mechanistic roles in human metabolic diseases.

Skeletal muscle aging is associated with insulin resistance, which may be mediated by intermuscular mitochondrial dysfunction and elevated inflammation(17–19). In contrast, exercise improves insulin action and promotes skeletal muscle health(20, 21). This is a complex, coordinated multi-organ adaptive process, the details of which are still being investigated. It is unknown to what extent the clinical effects of exercise may be mediated, at least in part, by EVs. To address this important knowledge gap, we integrated advanced EV analyses with existing biospecimens, longitudinal clinical data, and rich biological datasets from the well-characterized exercise cohort of the Studies Targeting Risk Reduction Interventions through Defined Exercise (STRRIDE) program. STRRIDE has focused on identifying the key mediators of the differential effects of exercise on health outcomes, including insulin sensitivity(20). Our previous causal analyses in 1,507 older adults aged >71 years identified lymphocytes, physical function, activity levels, and regular exercise as key potentiators of longevity(22), underscoring the essential roles of immune and muscle function in healthy aging. Consistently, we recently identified several longevity-associated large EV (LEV) subpopulations, primarily originating from immune cells, muscle cells, and hematopoietic stem cells (HSCs). Notably, these LEV subpopulations were positively correlated with peptides enriched in long-lived adults and negatively correlated with peptides reduced in long-lived compared with short-lived older adults(2). We hypothesize that plasma EV biomarkers predominantly related to immune and skeletal muscle cells are increased with exercise training which may lead to beneficial effects on insulin action. To test this hypothesis, we adapted the 25-surface marker panel from our recent study(2) and comprehensively analyzed EVs isolated from plasma samples of older STRRIDE participants at baseline and after two vigorous intensity exercise interventions, both of 6–8 months' duration, but differing in amount. In this pilot study, we intentionally focused on an in-depth, intensive analysis of a limited sample set for EV biomarker discovery and panel development, conserving valuable STRRIDE specimens for future hypothesis-driven validation. We present a novel, minimally invasive platform utilizing plasma EV surface markers to assess metabolic health and responses to exercise, which is now ready for validation in large human cohorts, including the full STRRIDE cohort.

MATERIALS AND METHODS

Study participants

This study utilized plasma biospecimens from participants in the well described STRRIDE cohort(20, 23). Participants were stratified by race, age, and sex, and then randomly assigned to either one of the exercise training groups or a non-exercising control group. The study protocol was approved by the appropriate institutional review boards, and all participants provided written informed consent.

Inclusion criteria required participants to be sedentary, overweight or mildly obese (body mass index [BMI] 25–35 kg/m²), and to have mild to moderate lipid abnormalities (defined as low-density lipoprotein [LDL] cholesterol levels between 130–190 mg/dL, or high-density lipoprotein [HDL] cholesterol levels below 40 mg/dL for men or below 45 mg/dL for women). All female participants were postmenopausal; half of the women were taking hormone replacement therapy. Exclusion criteria included the use of other medications that might affect carbohydrate metabolism, a diagnosis of diabetes, orthopedic conditions that could limit exercise participation, hypertension, or known heart disease.

To develop a platform for analyzing EV alterations in response to exercise, we profiled EVs from 40 plasma specimens collected from 20 participants at baseline and after 6–8-month intervention in the STRRIDE study. These included: (1) twelve older adults (mean age 63.6 ± 1.9 years) assigned to the low-amount (~1300 kcal/week), vigorous-intensity (65–80% peak relative oxygen consumption [RVO₂]) aerobic exercise group (MOD group, 14 kcal/kg of body mass/week; approximately 19 km/week at 65–80% peak RVO₂); and (2) eight older adults (mean age 56.9 ± 3.0 years) assigned to the high-amount (~2200 kcal/week), vigorous-intensity (65–80% peak RVO₂) aerobic exercise group (HIGH group, 23 kcal/kg of body mass/week; approximately 32 km/week at 65–80% peak VO₂). These specimens enabled us to evaluate EV changes in older adults across two different exercise amounts. Baseline characteristics of the study participants are summarized in Table 1. We selected samples based on availability from the STRRIDE MOD and HIGH groups. Because this pilot study focused on older adults, we included only participants aged >50 years and ensured a balanced age distribution between males and females. Aside from age and sex, we were blinded to all other clinical variables until completion of all EV analyses.

Table 1.

Baseline characteristics of study participants.

Groups	MOD	HIGH	P Value (Fold – HIGH vs. MOD)
Exercise amount	1300 kcal/week	2200 kcal/week
Exercise intensity	65–80% peak RVO2	65–80% peak RVO2
Participants	N=12	N=8
Age (years, mean ± SD, range)	63.6±1.9, range 61–68	56.9±3.0, range 53–61	p <0.01 (−1.12-fold)
Sex (number)	Female (6), Male (6)	Female (4), Male (4)	p >0.99
BMI (kg/m2, mean ± SD)	29.9±3.1	29.0±2.4	p =0.64 (−1.03-fold)
Peak AVO2 (L/min)	2.2±0.6	2.4±0.8	p =0.75 (1.10-fold)
Peak RVO2 (ml/kg/min)	24.3±5.8	27.6±5.5	p =0.27 (1.14-fold)
Completed aerobic exercise (days/week, mean ± SD)	3.1±0.6	3.8±0.9	p =0.06 (1.22-fold)

Fisher’s exact test was used to compare sex and Mann Whitney test was used to compare other baseline characteristics of study subjects; a p <0.05 was considered as statistically significant. vs.: versus; SD: standard deviation; BMI: Body Mass Index; VO₂: oxygen consumption; Absolute VO₂; AVO₂: Absolute VO₂; RVO₂: Relative VO₂.

Exercise equipment included cycle ergometers, treadmills, and elliptical trainers. Participants underwent an initial ramp phase of two to three months, during which the exercise amount and intensity gradually increased. This was followed by six months of training at the prescribed exercise dose. All exercise sessions were supervised to ensure adherence to the prescribed frequency and duration to achieve their weekly dose. Compliance was verified through direct supervision and/or recorded data from heart rate monitors (Polar Electro). Participants were instructed to maintain baseline body weight and avoid reducing dietary intake throughout the intervention.

Metabolic health measures

Metabolic health data from the completed STRRIDE study included fasting glucose (mg/dL), fasting insulin (μU/mL), Homeostatic Model Assessment for Insulin Resistance (HOMA-IR)—an indicator of hepatic insulin resistance(24, 25), and an insulin sensitivity index (mU/L/min) derived from a 3-hour intravenous glucose tolerance test (IVGTT), conducted approximately 24 hours after the final exercise session(26)—an indicator of the efficiency with which insulin promotes glucose uptake and suppresses glucose production in peripheral tissues like skeletal muscle and adipose tissue. After collection of fasting blood samples, glucose (50%, 0.3 g/kg body mass) was administered at time 0, followed by insulin (0.025 U/kg body mass) at minute 20. Additional blood samples were collected at the following points: 2, 3, 4, 5, 6, 8, 10, 12, 14, 16, 19, 22, 25, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160, and 180 minutes¹⁶.

Samples were centrifuged, and plasma was separated and stored at −80 °C. Insulin concentrations were measured using an immunoassay (Access Immunoassay System, Beckman Coulter), and glucose concentrations were assessed via an oxidation reaction (YSI Model 2300 Stat Plus, Yellow Springs Instruments). Insulin sensitivity was calculated using Bergman’s minimal model(26).

EV isolation and characterization

STRRIDE blood samples collected at baseline and 16–24 hours after the final bout of exercise were analyzed to characterize EV profiles. As previously described(2, 5), cell and platelet-depleted plasma samples were isolated from EDTA whole blood and stored at −80 °C. To minimize co-isolation of contaminants, we pre-cleared biospecimens before EV isolation. Frozen plasma samples were thawed and centrifuged at 2,000 × g for 10 minutes at 4°C to remove cellular debris. EVs were then isolated from the cleared plasma using a polymer-based precipitation method. Each staining panel used 20 μL of plasma; with four panels included in this study, a total of 80 μL was required. Specifically, 80 μL of cleared plasma were mixed with 19.2 μL of precipitation solution (ExoQuick, System Biosciences), vortexed immediately, and incubated on ice for 30 minutes. Samples were then centrifuged at 3,000 × g for 10 minutes at 4 °C to separate EV pellets from the EV-depleted supernatants. The EV pellets were gently and rapidly washed once with double-filtered PBS (df-PBS) that were filtered through 100 nm filters (EMD Millipore) and resuspended in df-PBS for downstream analyses. This method yields a high amount of EVs, preserves EV integrity, and recovers EVs across the size spectrum, including small (SEVs), medium (MEVs), and large EVs (LEVs), thus enabling comprehensive profiling. However, it may not be suitable for studies specifically focused on SEVs (exosomes). In addition, it can co-isolate some non-EV particles. Therefore, it is essential to characterize the isolated EVs following MISEV recommendations (27), as demonstrated in our previous work(2). In this study, isolated EVs were characterized by multiple methods: morphology was assessed using transmission electron microscopy (TEM); size and concentration were determined by high-resolution flow cytometry; and EV surface markers and cargo effectors were analyzed using flow cytometric profiling.

Negative staining and TEM imaging

Formvar/carbon-coated 400-mesh copper grids (Ted Pella) were glow-discharged for two minutes prior to sample application. The grids were then floated on 20 μL droplets of sample for 10 minutes. Following incubation, grids were briefly washed twice with deionized water, stained with 2% aqueous uranyl acetate for 1 minute, and gently dry using filter paper. Samples were imaged using a JEOL JEM-1230 transmission electron microscope operated at 80 kV (JEOL USA). Images were captured with a Gatan Orius SC1000 CCD camera and processed using Gatan Microscopy Suite software version 3.10.1002.0 (Gatan).

High-resolution multicolor flow cytometry

We analyzed four EV marker panels comprising a total of 28 markers, designed using FluoroFinder (RRID:SCR_015485). As previously described(2), EV pellets were resuspended in df-PBS and stained overnight at 4°C in the dark with shaking (300 rpm) with fluorescence-conjugated antibodies against human markers including Cluster of Differentiation (CD)9 (#743048), CD81 (#740079), CD63 (#565426), CD29 (#743785), CD8 (#555634), CD56 (#563041), CD15 (#560827), CD14 (#555398), CD68 (#565594), CD235a (#563810), CD31 (#740777), CD41a (#559768), CD34 (#560710), HLA-ABC (#555552), HLA-DRDPDQ (#740302), CD90 (#563070), CD73 (#562430), CD105 (#563466) (BD Biosciences), CD4 (#47-0048-42), CD19 (#15-0199-42), HLA-G (#MA1-19643) (ThermoFisher Scientific), Fatty Acid-Binding Protein 4 (FABP4; #sc-271529 PE), Muscle Cadherin (MCAD; #sc-398107 FITC) (Santa Cruz Biotechnology), ryanodine receptor (RYR; #NB300-543PECY55), and RYR2 (#NBP2-80143APCCY7) (Novus Biologicals) in the presence of Brilliant Stain Buffer Plus (BD Biosciences). RyR1⁺ population was defined as RyR⁺RyR2⁻, and RyR2⁺ population was defined as RyR⁺RyR2⁺ as we previously reported(2). Following a previously reported protocol(9), EVs were also stained with PKH26 Red Fluorescent Cell Linker Midi Kit (MilliporeSigma) to label lipid bilayer of cytoplasmic membranes, along with MitoTracker^™ Green and Deep Red (ThermoFisher Scientific) to stain total and respiring mitochondria, respectively. EVs were subsequently re-pelleted using polymer-based precipitation to remove unbound dyes remaining in the supernatant.

The final volume for each staining panel was adjusted to 260 μL with df-PBS, yielding a final dilution factor of 1:13 relative to the original 20 μL of plasma used for EV isolation. The acquisition parameters of the high-resolution multicolor BD LSR Fortessa X-20 flow cytometer (BD Biosciences) were configured as previously reported(2), including forward scatter (FSC) voltage of 410, side scatter (SSC) voltage of 210, and an SSC threshold of 200 to exclude small debris particles, ensuring that acquisition of the df-PBS controls generated fewer than 10 events per second. Samples were acquired for 2 minutes at a flow rate of approximately 30 μL per minute. EV size distribution was estimated using non-fluorescent reference beads of 100 nm, 1,000 nm, and 6,000 nm (Thermo Fisher), defining large EVs (LEVs, ~1,000–6,000 nm), medium EVs (MEVs, ~100–1,000 nm), and small EVs (SEVs, ≤100 nm). Fluorescence compensation, background, and positive signals were established using unstained and single-stained df-PBS controls, EV samples, and UltraComp^™ eBeads (ThermoFisher Scientific). For each marker, percentage positive (%) and geometric mean fluorescence intensity (MFI) were quantified. Data analysis was performed using FCS Express 5 software (De Novo Software). To enhance quantitative interpretation while reducing data complexity, we calculated the integrated mean fluorescence intensity (iMFI) for each marker, as described previously(28, 29). iMFI combines the percentage of positive EVs (frequency of marker expression) and MFI (mean expression intensity among positive EVs), providing a comprehensive measure of marker abundance.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 10 (RRID:SCR_002798, GraphPad Software) and JMP 19 (RRID:SCR_014242, JMP Statistical Discovery). The statistical tests applied in this study were the following. (1) D’Agostino & Pearson omnibus test was used to assess data distribution and guide the choice of parametric or non-parametric tests. (2) Fisher's exact test was used to compare sex of study participants across study groups. (3) Mann Whitney test was used to compare other baseline characteristics across study groups. (4) Friedman test with Benjamini-Hochberg correction was used to compare the frequencies of LEVs, MEVs and SEVs. (5) Spearman correlation was performed to examine associations among baseline EV biomarkers and clinical variables. (6) Multivariable linear regression modeling, adjusting for age, sex, and BMI, was used to evaluated associations among baseline EV biomarkers and variables related to metabolic health. (7) Wilcoxon matched-pairs signed-rank test was used to compare the iMFI of each EV marker between baseline and post-exercise. All tests were two-sided. Because this was a pilot study, a formal power calculation was not performed. Statistical significance was defined as p <0.05 or a false discovery rate (FDR) q <0.05.

RESULTS

Plasma EV characterization

Isolated plasma EVs from older adults exhibited the characteristic cup-shaped morphology, with heterogeneous shapes and a broad size distribution (Fig. 1A). Consistent with our prior observations(2, 5, 9, 15, 28–31), we identified three major EV subsets (LEVs, MEVs and SEVs) in the plasma of STRRIDE participants. LEVs represented a minor population (mean 1.3% ± 1.0%), presenting at significantly lower frequencies compared to MEVs (mean 27.0% ± 5.8%) and SEVs (mean 67.2% ± 5.2%) (Fig. 1B). The majority of these EVs displayed a lipid bilayer membrane (confirmed by PKH26⁺ staining, Fig. 1C), while the PKH⁻ population may be comprised of EVs that fail to incorporate the dye, as well as non-EV particles. We detected mitochondrial signals in these EVs, including both total and respiring mitochondrial components, as indicated by MitoTracker^™ Green⁺ (mean 76.1% ± 11.3%) and MitoTracker^™ Deep Red⁺ (mean 56.5% ± 28.4%) signals, respectively (Fig. 1D). Surface marker profiling confirmed the presence of canonical EV markers (CD81, CD9, CD63; Fig. 1E), as well as a broad array of cell- and tissue-related markers: CD29, CD4, CD8, CD56, CD15, CD14, CD68, CD19, CD235a, CD31, CD41a, CD34, HLA-ABC, HLA-G, HLA-DRDPDQ, CD90, CD73, CD105, FABP4, MCAD, RYR1, and RYR2 (Fig. 1E,F). The frequencies of CD41a⁺ populations related to platelets and HSCs were low (Fig. 1E), as we previously reported(5). These findings demonstrate the high degree of heterogeneity among plasma EV subpopulations and suggest a diverse range of cellular and tissue origins, including stem cells, immune and hematopoietic cells, platelets, skeletal (RYR1, MCAD, FABP4) and cardiac (RYR2) muscle, and adipocytes (FABP4)(2, 32–35) (Fig. 1G). Notably, while the surface markers examined are enriched in certain cell types, they are not exclusively expressed by them. In this study, our discussion will focus on the predominant cellular origins of these markers (Fig. 1G). Additionally, our recent study showed that EVs isolated from plasma of older adults using the same isolation method contain low levels of APOA1 (a major component of high-density lipoprotein [HDL] particles).

Figure 1. Plasma EV characterization.

A. Transmission electron microscopy (TEM) images display heterogeneous morphology of plasma EVs. The bars in all images represent 200 nm. B. The scatter plot displays percentages of large (LEVs), medium (MEVs) and small (SEVs) EV subsets within isolated baseline plasma EVs from STRRIDE participants (n=20). Friedman test with the Benjamini-Hochberg FDR correction were used for comparison; statistical significance defined as * q<0.05. C-F. The representative plots of flow cytometry display the frequencies of the indicated markers in plasma EVs. The RyR1⁺ population was defined as RyR⁺RyR2⁻; the RyR2⁺ population was defined as RyR⁺RyR2⁺. G. The graph displays the predominant cell origins of the tested surface markers.

Several EV subpopulations at baseline were associated with better metabolic health.

To explore their potential roles, we examined correlations between plasma EV biomarker abundance and demographic and metabolic data related to insulin action in older adults. The abundance of multiple baseline EV subpopulations was positively correlated with the insulin sensitivity index and negatively correlated with fasting insulin, fasting glucose, and HOMA-IR (Supplementary Fig. 1), suggesting their association with better indices of insulin action and metabolic health. In addition, CD4⁺, CD56⁺, and CD34⁺ EVs were positively correlated and CD90⁺ EVs were negatively correlated with age (Supplementary Fig. 1).

In multivariable linear regression adjusted for age, sex, and BMI, eight EV subpopulations, negatively associated with both fasting insulin (Fig. 2) and HOMA-IR (Fig. 3), carried surface markers that are highly enriched in immune cells (CD81⁺, CD29⁺, CD8⁺, CD56⁺, CD19⁺), red blood cells (RBCs) and/or pluripotent stem cells (PSCs) (CD235a⁺), skeletal muscle (MCAD⁺), and MSCs (CD73⁺, CD105⁺). Additionally, B cell related CD19⁺ EVs were negatively associated with fasting insulin concentration (Fig. 2) but not HOMA-IR (Fig. 3). In contrast, the other tested EV subpopulations were not statistically significantly associated with fasting insulin concentration (Supplementary Fig. 2) and HOMA-IR (Supplementary Fig. 3).

Figure 2. Nine EV biomarkers at baseline were negatively associated with fasting insulin concentration.

Multivariable linear regression modeling, adjusting for age, sex, and BMI, was used to evaluated associations between baseline EV biomarkers and fasting insulin concentration. Leverage plots illustrate the relationship between fasting insulin concentration (x-axis) and integrated mean fluorescence intensity (iMFI) of the indicated EV biomarkers (y-axis). Statistical significance was defined as * p<0.05. This figure presents the nine EV biomarkers that were negatively associated with fasting insulin concentration.

Figure 3. Eight EV biomarkers at baseline were negatively associated with HOMA-IR.

Multivariable linear regression modeling, adjusting for age, sex, and BMI, was used to evaluated associations between baseline EV biomarkers and HOMA-IR. Leverage plots illustrate the relationship between HOMA-IR (x-axis) and integrated mean fluorescence intensity (iMFI) of the indicated EV biomarkers (y-axis). Statistical significance was defined as * p<0.05. This figure presents the eight EV biomarkers that were negatively associated with HOMA-IR.

Notably, EVs related to major lymphoid cell subsets involved in adaptive immunity, including CD8⁺ T cells, CD19⁺ B cells, and CD56⁺ natural killer (NK) cells, were associated with lower fasting insulin concentration (Fig. 2) and HOMA-IR (Fig. 3), suggesting better insulin action and metabolic health. In contrast, CD15⁺ and CD41a⁺ EVs, primarily derived from myeloid neutrophils and platelets, along with HLA-ABC⁺ EVs, were positively correlated with fasting glucose (Supplementary Fig. 4) that reflects poorer insulin action.

These observations suggest that specific plasma EV subpopulations, putatively originating from immune cell subsets, skeletal muscle, and MSCs, may play beneficial roles in regulating metabolic health and insulin action in older adults, although the underlying mechanisms require further investigation.

Low-amount, vigorous-intensity aerobic exercise increased FABP4⁺ EVs in older adults.

We next investigated whether exercise could regulate the production of EV subpopulations in older adults in response to the STRRIDE MOD (the low-amount, vigorous-intensity) aerobic exercise intervention. MOD aerobic exercise of 6–8 months resulted in a significant increase in FABP4⁺ EVs (1.3-fold) that putatively originate from adipocytes, skeletal muscle cells, and other cell resources (34, 35), but not the other twenty-four tested EV subpopulations (Fig. 4). This finding suggests that older adults may require a different exercise mode, modality and/or dose to elicit comparable EV responses.

Figure 4. Low-amount, vigorous-intensity aerobic exercise increased iMFI of FABP4⁺ EVs in older adults.

Plasma EVs were isolated from older adults (n=12) at baseline and after low-amount, vigorous-intensity aerobic exercise (post-exercise), followed by profiling for the indicated surface markers using high-resolution flow cytometry. The scatterplots represent integrated mean fluorescence intensity (iMFI) of the indicated EV biomarkers. Comparisons were performed using Wilcoxon matched-pairs signed rank test; significant results were defined as ∗ p < 0.05 (red). Red font indicates p<0.05; green font indicates p=0.05–0.1; grey font indicates p>0.1.

High-amount, vigorous-intensity aerobic exercise upregulated plasma EV subpopulations associated with better indices of insulin action in older adults.

We then evaluated whether increasing the amount (weekly energy expenditure) of exercise of older adults would enhance their production of insulin action-associated EV subpopulations in response to the STRRIDE HIGH (high-amount, vigorous-intensity) aerobic exercise intervention. Participants in the HIGH group were slightly younger (1.12-fold) than those in the MOD group; all other baseline characteristics did not differ significantly between the two groups (Table 1).

The 6–8 months of high-amount, vigorous-intensity aerobic exercise increased fifteen plasma EV subpopulations compared to baseline (Fig. 5), including EVs bearing surface markers predominantly expressed by the following cell types: immune cells—CD63⁺ (2.4-fold), CD29⁺ (2.0-fold), CD8⁺ (2.5-fold), CD56⁺ (2.3-fold), CD68⁺ (2.0-fold), CD19⁺ (2.2-fold), and HLA-G⁺ (2.3-fold); skeletal muscle—RYR1⁺ (3.1-fold) and MCAD⁺ (2.0-fold); cardiac muscle—RYR2⁺ (1.9-fold), hematopoietic stem cells (HSCs, CD34⁺, 1.4-fold); MSCs—CD90⁺ (1.4-fold), CD73⁺ (1.9-fold), and CD105⁺ (1.8-fold); and RBCs/PSCs (CD235a⁺, 2.0-fold) (Fig. 5). Among these, eight EV subpopulations (CD29⁺, CD8⁺, CD56⁺, CD19⁺, MCAD⁺, CD73⁺, CD105⁺ CD235a⁺) were associated with better indices of insulin action at baseline (Fig. 2, Fig. 3). These findings show that older adults upregulate only FABP4⁺ EVs in response to the MOD (low-amount, vigorous-intensity exercise) intervention but exhibit a robust EV response to the HIGH (high-amount, vigorous-intensity exercise) intervention, with multiple insulin action-associated EV subpopulations increased. This suggests that older adults are likely to require a higher exercise dose to elicit metabolically beneficial EV responses.

Figure 5. In older adults, high-amount, vigorous-intensity aerobic exercise upregulated multiple plasma EV subpopulations associated with better indices of insulin action.

Plasma EVs were isolated from older adults (n=8) at baseline and after high-amount, vigorous-intensity aerobic exercise (post-exercise), followed by profiling for the indicated surface markers using high-resolution flow cytometry. The scatterplots represent integrated mean fluorescence intensity (iMFI) of the indicated EV biomarkers. Comparisons were performed using Wilcoxon matched-pairs signed rank test; significant results were defined as ∗ p<0.05. Red font indicates p<0.05; green font indicates p=0.05–0.1; grey font indicates p>0.1.

Interestingly, in our chronic exercise modes, we found that total plasma particle concentrations in older adults were unchanged after low-amount exercise but decreased following high-amount exercise (Supplementary Fig. 5). These results suggest that the exercise-induced changes in our EV biomarkers reflect alterations in EV composition and shifts in specific EV subpopulations.

DISCUSSION

While exercise-modified smRNAs have been extensively studied(36), human research on exercise-induced changes in EVs, particularly their comprehensive surface marker profiles, remains limited. Most exercise-related EV studies have focused on samples collected immediately after acute exercise, where circulating EV characteristics are largely influenced by mechanical stimuli(37, 38). This may contribute to the inconsistent findings reported in the literature(39, 40). In the present study, we examined circulating EV profiles in plasma collected at rest 16–24 hours after the final session of the 6–8-month exercise intervention, a timing that is more likely to capture chronic exercise-induced alterations in EV phenotype rather than acute, transient EV release. Our study established a novel platform using plasma, a minimally invasive specimen, to analyze EV surface marker abundance and composition in adults from the well-characterized STRRIDE study(20, 23).

Plasma EVs were highly heterogeneous, reflecting diverse cellular and tissue origins. At baseline, several plasma EV subpopulations, putatively originating from immune cells, skeletal muscle, and MSCs, were associated with enhanced insulin action, as indicated by their negative associations with fasting insulin and HOMA-IR(20, 25, 41). Notably, lower fasting insulin and HOMA-IR have been linked to exercise-mediated geroprotective effects and improved metabolic homeostasis(20, 42–46). Consistent with our observations, previous work reported reductions in circulating MSC (CD105⁺) and immune cell (CD45⁺) derived EVs, which negatively correlated with metabolic flexibility and levels of CD105⁺, CD45⁺, and CD31⁺ EVs(47). These EVs may influence insulin sensitivity by delivering bioactive cargo that mediates crosstalk among pancreatic β-cells, MSCs, skeletal muscle, and immune cells(10, 12). For example, human MSC-derived SEVs enriched in micro (mi)RNA-21 protect β-cells from hypoxia-induced apoptosis(48), while muscle-derived plasma SEVs from exercise-trained mice, enriched in miR-133a and miR-133b, improve insulin sensitivity of sedentary mice(49).

Low-amount, vigorous-intensity exercise increased only FABP4⁺ EVs in older adults, suggesting that older adults may require a more robust exercise stimulus to enhance the production of metabolically beneficial EVs. Indeed, high-amount, vigorous-intensity exercise upregulated fifteen plasma EV subpopulations in older adults, including those putatively originating from immune cells (CD63⁺, CD29⁺, CD8⁺, CD56⁺, CD68⁺, CD19⁺, HLA-G⁺), skeletal muscle (RYR1⁺, MCAD⁺), cardiac muscle (RYR2⁺), MSCs (CD90⁺, CD73⁺, CD105⁺), HSCs (CD34⁺), and RBCs and/or PSCs (CD235a⁺). Overall, eight of these EV subpopulations were associated with better insulin action at baseline. These findings support the role of EVs in mediating tissue crosstalk between pancreatic β-cells and other tissue/cells, thereby contributing to the regulation of insulin action (10, 12).

Plasma EV subpopulations carrying our exercise-responsive surface markers, especially CD63, HLA-G, CD105, FABP4, and MCAD, have correlated positively with many longevity-associated EV peptides in older adults(2), reinforcing their relevance to metabolic health, exercise response, and healthy aging. Collectively, these results highlight specific plasma EV subpopulations, defined by readily profiled surface markers, as promising next-generation biomarkers, modified by exercise, linked to geroprotective metabolic outcomes, and potentially suitable for evaluation of physiological adaptations. Blood samples were collected 24 hours after the final exercise session, indicating that exercise-induced changes in EV concentrations are relatively resilient.

Human studies on exercise-mediated EV alterations remain limited(39, 50), and inconsistent findings persist due to incomplete understanding of EV profiles and their cellular/tissue origins(51). Our data reveal that exercise effects on plasma EVs depend on exercise amount and the EV cellular origins. The same vigorous intensity, but with a greater amount of exercise in older adults enhanced the abundance of plasma EV subpopulations linked to insulin action, potentially contributing to the significant improvements in insulin sensitivity observed with this exercise protocol(20).

This pilot study aimed to establish a platform for EV biomarker discovery and panel development. Despite inter-individual variability and small sample size, several EV biomarkers showed statistically significant changes following exercise, demonstrating robustness and potential for validation in larger cohorts. Using effect size estimates from this pilot study of 20 individuals, we performed an a priori power analysis to determine whether our available sample size of older adults (>50 years old) remaining in the STRRIDE study—STRRIDE-HIGH (n=111) and STRRIDE-MOD (n=87)—would be sufficient to detect statistically significant differences with adequate power (power=0.9, α=0.01, two-sided test) to validate our pilot study results. Power analysis (performed in JMP 19) indicated that we have adequate sample to confirm the exercise-induced changes in the following identified EV biomarkers (n=sample size needed): for the high-amount, vigorous-intensity exercise group CD63 (n=20), CD29 (n=15), CD8 (n=17), CD56 (n=12), CD68 (n=15), CD19 (n=12), CD235a (n=19), CD34 (n=18), HLA-G (n=20), CD90 (n=17), CD73 (n=18), CD105 (n=17), MCAD (n=21), RyR1 (n=41), RyR2 (n=16), FABP4 (n=26); for the low-amount, vigorous-intensity exercise, FABP4 (n=50), RyR1 (n=64).

In summary, this pilot study demonstrates that plasma EVs comprise highly heterogeneous subpopulations from diverse cell and tissue sources. Several EV subpopulations putatively originating from immune cells, skeletal muscle, and MSCs were associated with better insulin action in older adults. Notably, high-amount, vigorous-intensity aerobic exercise induced multiple insulin action-associated EV subpopulations in older adults. The study focused on a targeted panel of 25 EV surface markers, though over 800 longevity-associated EV peptides have been previously identified(2), suggesting that additional EV biomarkers likely contribute to metabolic health and exercise responsiveness. These findings support plasma EV abundance and composition, defined by surface markers, as a promising platform for monitoring metabolic health and exercise responsiveness. Validation of this platform and its associated EV signatures in larger cohorts with varying exercise regimens, such as the full STRRIDE cohort, will be essential to confirm these preliminary findings.

Supplementary Material

Supplemental Figs. 1–5: https://doi.org/10.6084/m9.figshare.30929276

DATA AVAILABILITY

Data will be made available upon reasonable request.