Changes in proteomic cargo of circulating extracellular vesicles in response to lifestyle intervention in adolescents with hepatic steatosis

Abstract

Background.

Recent studies suggest that proteomic cargo of extracellular vesicles (EVs) may play a role in metabolic improvements following lifestyle interventions. However, the relationship between changes in liver fat and circulating EV-derived protein cargo following intervention remains unexplored.

Methods.

The study cohort comprised 18 Latino adolescents with obesity and hepatic steatosis (12 males/6 females; average age 13.3 ±1.2 y) who underwent a six-month lifestyle intervention. EV size distribution and concentration were determined by light scattering intensity; EV protein composition was characterized by liquid chromatography tandem-mass spectrometry.

Results.

Average hepatic fat fraction (HFF) decreased 23% by the end of the intervention (12.5% [5.5] to 9.6% [4.9]; P = 0.0077). Mean EV size was smaller post-intervention compared to baseline (120.2 ± 16.4 nm to 128.4 ± 16.5 nm; P = 0.031), although the difference in mean EV concentration (1.1E+09 ± 4.1E+08 particles/mL to 1.1E+09 ± 1.8E+08 particles/mL; P = 0.656)) remained unchanged. A total of 462 proteins were identified by proteomic analysis of plasma-derived EVs from participants pre- and post- intervention, with 113 proteins showing differential abundance (56 higher and 57 lower) between the two timepoints (adj-p <0.05). Pathway analysis revealed enrichment in complement cascade, initial triggering of complement, creation of C4 and C2 activators, and regulation of complement cascade. Hepatocyte-specific EV affinity purification identified 40 proteins with suggestive (p < 0.05) differential abundance between pre- and post-intervention samples.

Conclusions.

Circulating EV-derived proteins, particularly those associated with the complement cascade, may contribute to improvements in liver fat in response to lifestyle intervention.

Introduction

Metabolic dysfunction-associated fatty liver disease (MASLD), formerly known as nonalcoholic fatty liver disease (NAFLD), is a chronic, progressive condition that encompasses a histologically defined spectrum of disorders ranging from simple steatosis to steatohepatitis, and potentially leading to fibrosis and hepatocellular carcinoma (HCC) [1]. In the United States, MASLD has emerged as the predominant chronic liver disease among children, impacting 13% of the overall population and 26% of those with obesity [2–4]. Risk factors for pediatric MASLD include obesity, unhealthy lifestyle, family history, and Latino ethnicity [2]. Latino youth, particularly those of Mexican ancestry [5, 6], experience the highest rates of MASLD [2, 7–9]. This trend extends into adulthood, where Latino adults experience a disproportionate burden of MASLD [10–15] and develop the disease at a younger age compared to other ethnic groups [11], which accelerates adverse outcomes such as HCC [16] and HCC-related mortality [17, 18].

The implications of pediatric MASLD are far-reaching, as it is associated with unfavorable long-term outcomes such as type 2 diabetes [19], cardiovascular disease [20], an indication for liver transplantation [21], and increased morbidity and mortality in adulthood [22]. While lifestyle interventions can reduce risk factors for pediatric MASLD [23], the mechanisms underpinning health improvements remain largely unknown. The limited understanding of the mechanisms driving MASLD risk and risk reduction further impedes the development of appropriate diagnostic and therapeutic strategies.

Extracellular vesicles (EVs) are recognized as important constituents of intercellular communication, transporting bioactive molecules such as proteins, lipids, and nucleic acids between cells and tissues. In addition, EVs are involved in immune modulation, tissue regeneration, and cancer metastasis [24]. Growing evidence also suggests a connection between EVs and MASLD [25–30]. Individuals with MASLD exhibit elevated plasma EV levels, which correlate positively with disease severity [30–32]. In children and adolescents with obesity, circulating levels of EVs are higher compared to controls without obesity and are associated with markers of liver dysfunction [33]. Circulating EVs influence intracellular signaling, tissue injury and repair, and matrix remodeling in liver cells, particularly hepatocytes and hepatic stellate cells [34]. Steatotic hepatocytes release a greater number of EVs into the extracellular space relative to normal hepatocytes [32, 35, 36] and these EVs communicate deleterious messages between cell types in the liver [35, 37–43].

Recent studies suggest that EV cargo may contribute to metabolic changes after lifestyle and surgical interventions. For example, significant associations between changes in circulating EVs and improvements in metabolic parameters following interventions like bariatric surgery and dietary modifications have been reported [44–47]. To our knowledge, however, no studies have yet explored the relationship between changes in liver fat and circulating EV-derived protein cargo following lifestyle intervention. The aim of the current study was to assess the impact of lifestyle intervention on EV characteristics in Latino youth with obesity and hepatic steatosis at baseline and perform a comprehensive proteomic analysis to identify changes in protein abundance resulting from the intervention. We also explored the proteome of hepatocyte-enriched EVs from plasma. Our findings reveal distinct effects of lifestyle intervention on cargo proteome, confirm the close relationship between MASLD and EV biology, and provide additional mechanistic insights into the disease and avenues for potential reversal of its pathophysiology.

Materials and Methods

Study participants.

The study cohort comprised 18 Latino adolescents (12 males / 6 females) with obesity and MRI-derived hepatic fat fraction ≥5.0%. Participants were selected from a cohort who completed a 6-month intensive lifestyle intervention that targeted improvements in eating behaviors (e.g., decreased sugar and saturated fat intake, portion control, and increased fiber intake) and increases in physical activity. Details of the intervention [48] and primary outcomes can be found elsewhere [49]. Criteria for study inclusion included self-reported Latino ethnicity, age 12 – 16 years, presence of obesity defined as a BMI percentile ≥ 95^th percentile for age and gender or BMI ≥ 30 kg/m², and evidence of hyperglycemia. Individuals were excluded from the intervention if they had medical conditions requiring the use of medications known to impact carbohydrate metabolism (such as polycystic ovary syndrome and corticosteroids), limited physical activity (e.g., due to conditions like cerebral palsy), or the presence of unstable emotional health issues (e.g., uncontrolled clinical depression). Written consent and assent for research were obtained by study staff from the parent/guardian and adolescent. This study was approved by the Institutional Review Board of Arizona State University (Phoenix, Arizona) under protocol # 00017738 on April 05, 2023.

Changes in cardiometabolic risk.

Before and after the intervention, participants were assessed for fasting lipids, liver enzymes (ALT and AST), glucose homeostasis (HbA1c, fasting and 2-hour post-challenge insulin and glucose concentrations), and total body composition by Dual Energy X-ray Absorptiometry (Lunar iDXA; GE Healthcare). Hepatic fat fraction (HFF) was quantified by advanced chemical shift-encoded water-fat magnetic resonance imaging (MRI) with Philips’ mDIXON pulse sequence using Philips Ingenia^® 3T MRI platform. Fasting blood specimens for molecular experiments were collected before and after the intervention in K2-EDTA tubes and stored at −80°C until analysis.

Isolation of EVs from plasma.

EVs were isolated by size exclusion chromatography (SEC) using qEVoriginal 70 nm Gen 2 columns (Izon Science; Medford, MA) from fasting plasma samples taken at baseline and at the conclusion of the intervention. Following column equilibration with 15 mL of phosphate buffered saline (PBS), 500 μL of plasma sample was loaded onto the column and 6 × 500 μL fractions (F) were collected (F7-F12), following the collection of the 3 mL void volume. Fractions F7-F12 were combined and concentrated with 50K Amicon filters (Millipore, UFC805096) to a final volume of 500 μL. The combined concentrated EVs were used for downstream analyses. The EV size distribution and concentration were determined by light scattering intensity using the NanoSight NS300 instrument (Malvern Panalytical USA; Westborough, MA).

Affinity purification of hepatocyte EVs from plasma.

Magnetic beads coated with antibodies targeting ASGR1 and CYP2E1, both recognized as markers of hepatocytes [50], were utilized to selectively capture this EV sub-population. Briefly, 1 mL of Dynabeads (Invitrogen; Waltham, MA) was coupled with 20 μg of ASGR1 (Invitrogen, PA5–80356) and CYP2E1 (Invitrogen, PA5–86704) antibodies for 1 hour at 4°C. Following this coupling step, excess antibodies were removed by washing with a buffer solution buffer (PBS supplemented with 0.1% bovine serum albumin and 2 mM EDTA, pH 7). Subsequently, 1 mL of total EVs previously isolated from plasma using SEC at a concentration of 5.0E10 particles/mL was incubated with the prepared beads overnight at 4°C to capture ASGR1- and CYP2E1-specific EVs. The captured EVs were eluted from the beads with elution buffer (0.1 M citrate, pH 2–3). The elution buffer was immediately replaced with 20 mL PBS and concentrated with 50K Amicon filters to a final volume of 200 μL.

Sample preparation for proteomics analysis.

Isolated EVs were solubilized in a 2% sodium deoxycholate-based (DOC) lysis buffer and sonicated using cup-horn shaped sonotrode (UTR2000, Hielscher Ultrasonics). Protein extracts were clarified by centrifugation and protein concentration was determined using the BCA (Pierce). Proteins from isolated EVs (80 μg/sample) were processed as previously described [51, 52]. Briefly, proteins were reduced with 10 mM dithiothreitol at room temperature for 30 minutes, alkylated with 20 mM iodoacetamide in the dark for 30 minutes, and digested by overnight incubation at 37°C with Trypsin Gold (Promega; Madison, WI) at a ratio of 1:50. Following trypsin digestion, samples were acidified and DOC was removed by centrifugation. Peptide solid phase extraction (SPE) was performed using 1cc SEP-PAK C18 cartridges (Waters; Milford,MA). Eluted peptides from SPE were vacuum-centrifuged to dryness and stored at −80 °C until analysis. ASGR1- and CYP2E1- enriched EVs were prepared as described for the total EVs, with the exception of the amount of protein digested, i.e., 60 μg/sample. Dried peptides from each sample were reconstituted in loading solvent (98% water, 2% acetonitrile, 0.1% formic acid) and quantified using BCA assay. To generate a spectral library for unbiased EV samples, equal amounts of peptide (i.e., 28 μg) were pooled from each sample and subjected to offline fractionation via high pH reverse phase chromatography on an Ultimate 3000 HPLC system (Thermo Scientific; Waltham, MA). Peptides were loaded onto a 10 cm C18 column (Waters XBridge C18, 4.6 mm ID, 3.5 μm particle size) and eluted over a 96-minute method into a 96-well plate. The resulting 96 fractions were combined to 12 peptide fractions for LC-MS/MS acquisition. To generate a spectral library for the hepatocyte-enriched EVs samples, equal amounts of peptide per sample from the enriched EV digests were combined. The pooled sample was fractionated into eight fractions using the Pierce high pH Reversed-Phase Peptide Fractionation Kit (Thermo) according to the manufacturer’s protocol. An additional ninth fraction was collected by eluting peptides with 80% acetonitrile. All samples and library fractions were spiked-in with iRT peptides (Biognosys; Switzerland) and 200 ng peptides per sample were injected on-column.

Liquid chromatography tandem-mass spectrometry (LC-MS/MS).

All mass spectrometry data were acquired on a nanoElute liquid chromatography system coupled to a timsTOF Pro 2 (Bruker Daltonics; Billerica, MA) mass spectrometer with a captive spray source (Bruker) using a 62-minute LC gradient at a flow-rate of 850 nL/min on a 25 cm C18 column (Bruker PepSep, 150 μm ID, 1.5 μm particle size). Individual library fractions were acquired in DDA-PASEF (Data Dependent Acquisition – Parallel Accumulation Serial Fragmentation) mode with MS1 scans covering a mass range of 100–1700 m/z, TIMS mobility window (1/K0) between 0.70 and 1.50 with 100 ms accumulation and ramp time. DDA scans involved 7 PASEF ramps for a total cycle time of 0.85 second and a collision energy ramp of 20 eV to 65 eV for ion mobility window (1/K0) of 0.6–1.6. Each EV sample was acquired in DIA (Data Independent Acquisition)-PASEF mode keeping the same accumulation and ramp time as DDA runs. DIA window placement was optimized by pyDiaD [53] software using Spectronaut generated sample specific spectral library as its input [53, 54]. The capillary voltage was kept at 1700 V and dry gas temperature was kept at 200 °C.

Proteomics Data Analysis.

Spectral libraries from the DDA-PASEF runs were created by Pulsar search engine within the Spectronaut 17.6 software against a human SwissProt database (downloaded May 2020). Theoretical digestion was performed using trypsin allowing for a maximum of 2 missed cleavages. Cysteine carbamidomethylation was set as a fixed modification, while methionine oxidation and protein N-terminal acetylation were set as variable modifications. PSMs, proteins, and peptides were all filtered for FDR < 1%. DIA-MS data was searched against sample-specific spectral libraries using default parameters (cross-run normalization and data imputation were disabled). Departures from normality were evaluated using the Shapiro-Wilk test.

Statistical analysis.

Values of anthropometric and clinical data were expressed as means and standard deviation, and baseline vs post-intervention values were compared using paired t-test. Values of EV size and concentration were expressed as means with standard deviation. Changes in EV concentration and size were correlated with clinical variables by linear regression both in unadjusted analyses and in those adjusted for age and sex. Protein abundances were normalized using variance stabilization normalization (vsn package) in R v4.2.0 and differences between baseline and post-intervention measures were assessed using a paired Student’s t-test with a Benjamini-Hochberg (BH) correction for multiple testing [55]. Volcano plots were generated using R-ggplot2 using log₂ fold change) (x-axis) and raw or adjusted-p-value (y-axis). Heatmaps were generated with Euclidian distance and the complete clustering method using the R-package pheatmap. We performed pathway analysis with the differentially abundant proteins with reference to the Gene Ontology (GO) database using the enrichGO function, and also to REACTOME (function: enrichPathway) as implemented in the clusterProfiler package. P-values were adjusted using the BH method, and pathways with an adjusted-p (adj-p) < 0.05 were considered statistically significant.

Results

Anthropometric and clinical characteristics of study participants

Participant characteristics before and after the intervention are presented in Table 1. With the exception of height, no statistically significant differences in anthropometric measures were observed between the two timepoints. Participants exhibited significant reductions in total fat mass and increases in total lean mass resulting in a significant decrease in percent body fat. Similarly, HFF was reduced by 23.2% compared to baseline (P = 0.0077) with heterogeneity observed within the cohort where individual changes ranged from a decrease of 13.3% to an increase of 2.7% (Fig 1A). Of the 18 participants, 13 exhibited a reduction in HFF and 5 increased HFF compared to baseline.

Table 1.

Characteristics of study participants before and after a six-month lifestyle intervention

	Pre-intervention (N=18)	Post-intervention (N=18)	P-value
Anthropometric
Weight (kg)	92.9 ±22.9	94.2 ±23.0	0.2383
Height (cm)	165.0 ±7.0	166.9 ±6.7	0.0002
Waist circumference (cm)	106.5 ±13.4	106.7 ±14.6	0.8627
BMI (kg/m2)	33.8 ±5.8	33.5 ±6.1	0.3891
BMI (z-score)	2.3 ±0.3	2.2 ±0.4	0.0640
Imaging
Hepatic fat fraction (%)	12.5 ±5.5	9.6 ±4.9	0.0077
Visceral adipose tissue (cm3)	1385.1 ±644.5	1314.7 ±524.6	0.2898
Subcutaneous adipose tissue (cm3)	6343.3 ±2493.9	5963.9 ±2488.5	0.6506
Body fat (%)	44.7 ±4.8	42.2 ± 5.3	<0.0001
Lean body mass (kg)	45.6 ±9.5	48.5 ±9.7	<0.0001
Fat (kg)	39.6 ±13.2	38.0 ±13.1	0.0434
Liver Biochemistry
ALT (IU/L)	29.8 ±12.7	23.2 ±5.6	0.0214
AST (IU/L)	24.8 ±7.4	25.9 ±6.2	0.4150
Lipids
Total cholesterol (mg/dL)	150.8 ±25.1	142.4 ±19.2	0.0644
TG (mg/dL)	159.5 ±88.1	141.9 ±108.8	0.3054
HDL-C (mg/dL)	40.1 ±8.4	39.7 ±8.0	0.7322
LDL-C (mg/dL)	79.9 ±22.0	76.4 ±17.0	0.3947
Glucose homeostasis
FPG (mg/dL)	100.7 ±6.5	99.1 ±6.0	0.3565
2 hr glucose (mg/mL)	144.8 ±33.4	130.0 ±19.8	0.0642
HbA1c (%)	5.7 ±0.3	5.6 ±0.2	0.0337
Fasting insulin (mg/dL)	25.5 ±13.8	23.0 ±14.0	0.4051
2 hr insulin (mg/dL)	206.3 ±127.3	156.1 ±108.3	0.1537

All data are given as mean ±SD

Fig. 1. Changes in hepatic fat fraction (HFF) and EV characteristics in individual participants following a six-month lifestyle intervention.

Baseline and post-intervention measures are shown for A) HFF, B) mean EV size, and C) EV particle number. Values for females and males are shown in blue and gray, respectively. HFF was quantified by advanced chemical shift-encoded water-fat MRI. The red line in 1B and 1C indicates the mean values ± standard deviation.

EV characteristics are correlated with clinical variables

We first assessed the impact of lifestyle intervention on EV characteristics, comparing EV size and concentration at baseline and post-intervention. The majority of individuals exhibited a decrease in mean EV size and concentration (Fig 1B). Specifically, there was a statistically significant reduction in mean EV size at the conclusion of the intervention (128.4 ± 16.5 nm to 120.2 ± 16.4 nm; P = 0.031), although the difference in mean EV concentration (1.1E+09 ± 4.1E+08 particles/mL to 1.1E+09 ± 1.8E+08 particles/mL; P = 0.656)) was not statistically significant (Fig 1C).

Given the variation in EV size and concentration among study participants, we explored potential correlations between alterations in EV characteristics and clinical variables. After adjusting for sex and age, we observed a significant correlation between change in EV size and changes in subcutaneous adipose tissue (SAT: β = 0.010; P = 0.0085), high-density lipoprotein cholesterol (HDL: β = 1.664; P = 0.0314), fat mass (β = 3.069; P= 0.0340), and BMI (β = 6.472; P= 0.0389) (Table S1). Additionally, there was a trend in the correlation between change in EV size and percent body fat (β = 4.133; P = 0.0664). Changes in mean EV concentration were significantly correlated with changes in HDL (β = 54294496.2; P = 0.032) (Table S2).

Proteomic profiling of plasma EVs identifies differentially abundant proteins post-intervention

A total of 462 proteins were identified by label-free LC-MS/MS proteomic analysis of plasma-derived EVs from participants pre- and post-lifestyle intervention, with 161 proteins showing suggestive differential abundance between baseline and post-intervention (p <0.05). After adjusting for multiple comparisons, 113 proteins, 56 with higher and 57 with lower levels, were differentially abundant following the intervention (adj-p <0.05; Fig 2A, Table S3). Among the proteins showing the most significantly decreased abundance post-intervention were complement C8 gamma chain (C8G), alpha-1-B glycoprotein (A1BG), immunoglobulin heavy constant gamma 4 (IGHG4), complement factor B (CFB), ceruloplasmin (CP), and complement C8 beta chain (C8B). The proteins that showed the most significant increase included LDL receptor related protein 1 (LRP1), protein S (PROS1), joining chain of multimeric IgA and IgM (JCHAIN), and complement component 4 binding protein alpha (C4BPA). In the heatmap generated using the top significant abundant proteins (adj-p < 0.01), pre- and post-intervention groups formed distinct clusters (Fig 2B).

Fig 2.

A) Volcano plot depicting differentially abundant proteins in response to the lifestyle intervention. The x-axis represents the log₂-fold change (FC), and the y-axis represents the -log10 of the adjusted p-values. Each point on the plot represents a protein, with red and blue dots indicating proteins with a significant increase or decrease in abundance, respectively. Black dots depict proteins not meeting statistical significance (adj-p <0 0.05). B) Heatmap illustrating the differential abundance of proteins pre- and post-lifestyle intervention. Rows represent individual proteins, and columns represent study participants (pre = gray; post = black). The color scale indicates the magnitude of protein abundance changes, with red indicating increased abundance and blue indicating decreased abundance. Proteins with adjusted p-values < 0.05 were considered statistically significant. C) Top results from enrichment analysis Top twenty functional classes for all differentially abundant proteins detected from the enrichment analysis using the Gene Ontology (GO) database. D) Top pathways of involvement identified using the Reactome database. For 2C and 2D, “All” refers to all differentially abundant proteins (adj-p <0.05), “Up” refers to proteins with higher abundance and “Down” refers to proteins with lower abundance.

To gain functional insights into this set of 113 differentially abundant proteins (adj-p < 0.05), we performed enrichment analysis using the Gene Ontology (GO) database. Our analysis identified 181 significantly enriched functional classes (Table S4) and the top twenty of these are shown in Fig. 2C for all differentially abundant proteins, as well as those with higher or lower abundance relative to baseline. Differentially abundant proteins were predominantly associated with blood microparticle, collagen-containing extracellular matrix, secretory granule lumen, cytoplasmic lumen, and vesicle lumen. Pathway analysis using the Reactome database was also performed to identify pathways of involvement for the differentially abundant proteins. This analysis identified 30 significantly enriched functional pathways (Table S5); top pathways are shown in Fig. 2D. The pathways most enriched for proteins showing increased abundance included complement cascade, initial triggering of complement, creation of C4 and C2 activators, binding and uptake of ligands by scavenger receptors, and regulation of complement cascade, while those showing decrease abundance included platelet degranulation, response to elevated platelet cytosolic calcium, regulation of complement cascade, complement cascade, platelet activation, signaling, and aggregation, and intrinsic pathway of fibrin clot formation.

Given the diverse responses in liver fat reduction post-intervention, we conducted a targeted analysis focusing on participants who experienced a decrease in HFF. In this subgroup, we identified 94 proteins with significantly differential abundance (adj-p < 0.05), comprising 44 proteins with increased and 50 proteins with decreased abundance relative to baseline levels (Table S6). GO analysis of these differentially abundant proteins identified 212 enriched functional classes, including many of the same major GO terms identified in the untargeted comparisons (Fig S1A; Tables S7), while the Reactome analysis revealed 30 enriched pathways (Fig S1B; Table S8). We also detected eight proteins unique to individuals who experienced an increase in liver fat displaying suggestive evidence of differential abundance (p < 0.05; Table S9).

Enrichment of hepatocyte-specific EVs from plasma reveals additional differentially abundant proteins

Many of the proteins showing differential abundance in plasma EVs between the pre- and post-intervention groups were found to either play an important functional role in the liver, such as FGL1 and RBP4 or exhibit high hepatic expression, including A1BG, F12, FGL1, GC, HRG, LRG1, MBL2, ORM1, RBP4, and SERPINA6. Based on these findings, we hypothesized that EVs released by the liver into plasma might yield insights into the biological changes resulting from alterations in HFF. To explore this possibility, we performed proteomic profiling of EVs that were affinity purified from plasma using the hepatocyte markers ASGR1 and CYP2E1 in the same set of samples. A total of 1528 proteins were detected in this enriched EV population. Forty proteins (20 lower and 20 higher) showed suggestive differential abundance (p < 0.05) between pre- and post-intervention samples, including complement C2 (C2), lipocalin 1 (LCN1), and malic enzyme 2 (ME2) (Fig S2; Table S10). Among these proteins, C2, complement factor properdin (CFP), biotinidase (BTD), beta-2-glycoprotein 1 (APOH), fibrinogen-like protein 1 (FGL1), vitamin D-binding protein (GC), hemopexin (HP), and lumican (LUM) overlapped with the differentially abundant proteins identified in unenriched plasma EVs.

Discussion

In this study, we describe variations in the proteome of circulating EVs in Latino youth with obesity and hepatic steatosis following a six-month lifestyle intervention. These findings align with previous research by Apostolopoulou et al [56], who reported significant changes in 262 proteins within circulating EVs following a 12-week high-intensity interval training intervention in adult males. We also observed changes in the proteomic cargo of EVs from participants who experienced a reduction in liver fat that were not found in those in whom HFF increased. These findings indicate that a lifestyle intervention can have a positive impact on hepatic fat levels and induce changes in the proteomic content of circulating EVs specific to HFF reduction.

The study cohort consisted of a well-defined group of Latino adolescents at increased risk for metabolic disease. Comprehensive clinical and anthropometric data were accessible for all participants at baseline and following intervention. There were no significant changes in weight, waist circumference, BMI, or the volumes of visceral and subcutaneous fat between the two time points. However, significant alterations were observed in percent body fat, fat mass, and lean body mass, indicating improvements in total body composition. Importantly, the average HFF significantly decreased from 12.5% (±5.5) to 9.6% (±4.9), and ALT levels improved from 29.8 IU/L (±12.7) to 23.2 IU/L (±5.6) in response to the intervention.

We did not observe significant changes in plasma EV concentrations following the lifestyle intervention. While several studies have reported an increase in EV levels in response to acute exercise; these changes appeared to be transient [46, 57–60]. In line with our findings, other studies have found no significant alterations in circulating EVs in humans or rats subjected to chronic exercise [61, 62]. The absence of significant changes in EV concentrations in our study could be attributed to a potential normalization during the six-month intervention period or the methodology used to measure EV concentration. For example, nanoparticle tracking analysis (NTA), as employed in our study, does not distinguish between EVs and other types of particles, such as lipoprotein particles or protein aggregates, that may be present in the sample and result in masking of EV signal [58, 63]. Therefore, the use of NTA for EV quantification has certain limitations that warrant consideration when interpreting these results.

Eguchi et al [64] observed elevated circulating EV numbers in individuals with obesity and metabolic syndrome compared to lean controls. Following a 3-month hypocaloric dietary intervention, EV concentrations decreased 35% and were associated with a reduction in measures of insulin resistance. Similar to the present findings, Eguchi and colleagues did not find evidence for a correlation between circulating EV levels and changes in BMI and fat mass. In an independent study, only individuals who experienced improved insulin sensitivity in response to 12 weeks of exercise training exhibited an elevated number of circulating EVs, while those who remained insulin resistant showed no changes [56]. With the exception of HDL, we did not find any evidence supporting a relationship between change in mean EV concentration and change in other clinical parameters measured in the study cohort.

On the other hand, we did observe a significant reduction in mean EV size as a consequence of the lifestyle intervention. Furthermore, we identified significant correlations between changes in mean EV size and alterations in SAT, HDL, fat mass, and BMI. A recent study by Kobayashi et al [33] found that children and adolescents with obesity demonstrated larger diameter EVs in circulation compared to their normal weight counterparts. Together, these findings suggest that obesity correlates with larger-sized EVs, and that improvements in metabolic function resulting from lifestyle interventions may contribute to a reduction in EV size.

Many of the proteins exhibiting significant differential abundance between the two study timepoints, including the top three proteins, complement C8 gamma chain (C8G), LDL receptor related protein 1 (LRP1), and protein S (PROS1), are linked with liver dysfunction, and circulating plasma levels of these proteins have been associated with hepatic inflammation or fibrosis [65–67]. C8G, a member of the lipocalin family binding small hydrophobic ligands, is one of the three subunits constituting C8. This protein actively participates in the formation of the membrane attack complex (MAC), which also includes C5b, C6, C7, and C9. The MAC induces complement lysis and has been associated with hepatic inflammation in individuals undergoing liver transplantation [68]. C8 has also been associated with liver stiffness in individuals with NAFLD [65]. Our analysis revealed a decrease in C8G post- intervention, suggesting a potential reduction in MAC formation, and consequently, an attenuation of hepatic inflammation. LRP1 is integral to intracellular signaling, lipoprotein metabolism, and clearance of apoptotic cells, and assumes a pivotal role in maintaining lipid homeostasis in concert with liver sinusoidal endothelial cells and hepatocytes [69]. Liver-specific LRP1 knockout mice, when subjected to a high fat diet, rapidly developed hepatic steatosis, along with obesity, insulin resistance, and hyperglycemia [70]. Lastly, protein S (PROS1), a vitamin K-dependent glycoprotein primarily synthesized in the liver, exhibits anticoagulant, anti-inflammatory, anti-apoptotic, and immunomodulatory properties. In a chronic liver injury model, PROS1 worsened liver injury, resulting in exacerbated hepatic fibrosis and elevated levels of circulating aminotransferases [71]. Human studies have shown significant associations between plasma PROS1 levels and other blood coagulation markers with the fatty liver index [72]. Our analysis also detected additional proteins with coagulant properties (Tables S7–S8), further supporting a role for hemostatic factors in hepatic steatosis.

One of the most interesting aspects of our study was the number of complement-related proteins that were altered following the intervention. Previous research has established the activation of both the complement system and the complement alternative pathway in individuals with MASLD [73, 74]. A comprehensive meta-analysis further demonstrated elevated levels of complement components in MASLD, finding a correlation between these proteins and increased risk, as well as severity, of liver dysfunction [75]. Apoptotic cells, triggered by hepatic lipid accumulation, are believed to initiate complement activation [76]. Our findings contribute to the growing body of evidence suggesting that the positive impacts of lifestyle intervention on hepatic fat content may translate to a reduction in complement activation. Importantly, these alterations are reflected in the cargo composition changes of EVs, thereby highlighting a potential mechanism through which lifestyle interventions positively influence complement-related processes.

In addition, our analysis revealed a significant reduction in the abundance of proteins associated with critical platelet functions, including platelet degranulation, response to elevated platelet cytosolic calcium, and platelet activation, signaling, and aggregation. Platelet activation is an important feature of MASLD and contributes to disease progression by enhancing the pro-thrombotic and pro-inflammatory state [77]. Platelets participate in liver homeostasis through complex mechanisms [77], and like the complement system, they are also key mediators of the innate immune response [78]. Our findings suggest that lifestyle intervention may improve hepatic steatosis through mechanisms related to attenuated platelet activation and that these changes can be detected in the proteomic profile of circulating EVs.

We also noted variations in the proteomic composition of EVs among a subset of participants who exhibited a reduction in liver fat at the conclusion of the intervention. Among the proteins that displayed the most significant decrease in individuals with reduced liver fat were coagulation factor XII (F12), alpha-1-B glycoprotein (A1BG), and complement factor B (CFB). Conversely, proteins that exhibited the most significant increase included alpha-2-macroglobulin (A2M), Collagen alpha-3(VI) chain (COL6A3), and JCHAIN. F12 activity is increased in human NAFLD, suggesting that hepatic steatosis may contribute to thrombosis [79]. A1BG was able to distinguish individuals with liver fibrosis from healthy controls [80] and a recent meta-analysis found that increased CFB was associated with MASLD severity [75]. Some studies have shown preliminary evidence linking levels of A2M [81, 82] and COL6A3 [83] with metabolic liver dysfunction. Interestingly, COL6A3 abundance was lower in humans with obesity, but increased following weight loss from either dietary or surgical interventions [84]. The relationship of many of the other differentially abundant proteins to hepatic steatosis is currently unknown, and may therefore represent novel pathways for exploration.

Several of the differentially abundant circulating EV-derived proteins identified in individuals who experienced a reduction in HFF as a result of the intervention overlapped with plasma proteins previously found to be altered in children with MASLD [82]. These overlapping proteins included members of complement components (e.g., CFB, CFI, CFHR1), immunoglobulin chains (e.g., IGHG1, IGHV3–7, IGKV3–20, IGLV1–47), and acute phase/regulatory proteins (e.g., A2M, AFM, AGT, APOE, APOM, CLEC3B, GC, GSN, HPX, ORM1, RBP4, SERPINA1, SERPINA6, SERPINA7, SERPINC1). The majority of overlapping proteins exhibited alterations in opposite directions. In other words, individuals who demonstrated reduced HFF post-intervention displayed a profile of changes in the opposite direction compared to youth with MASLD. These results indicate that lifestyle intervention may improve hepatic steatosis through pathways that are activated in MASLD and are identifiable in the EV proteome. While the EV and plasma proteomes in the two studies share similarities, the presence of additional differentially abundant EV proteins implies that EVs may provide information beyond what is accessible from plasma alone.

We also conducted a profiling of hepatocyte-specific EVs that were affinity-purified from plasma. This profiling enabled us to identify 40 proteins with suggestive evidence of differential abundance between pre- and post-intervention samples. The majority of these proteins were found to be closely linked to liver function and expression and may provide preliminary insight into potential biological changes resulting from modifications in HFF. There was limited overlap between the proteins identified in hepatocyte-enriched EVs compared to plasma-derived EVs. However, this finding was anticipated, as the purified EVs constituted only a small fraction of the total plasma EV population, which posed challenges in discerning specific proteomic cargo amid the complex background.

In a mouse model of MASLD, blood-derived EVs showed enrichment of hepatocyte-specific contents [38]. In these animals, circulating EV levels also reflected changes in liver histopathology and were associated with hepatocyte cell death and liver fibrosis. Higher levels of circulating hepatocyte-specific EVs were also observed in mice fed a diet high in saturated fat, fructose, and cholesterol compared to chow-fed animals [35]. Inhibition of EV release attenuated hepatic fibrosis and inflammation in these mice. Combined, these findings indicate that circulating EVs may serve as markers of NAFLD severity and may reflect molecular perturbations occurring during hepatocellular injury.

Strengths of the study including the comprehensive phenotyping of the study participants, the well-structured, culturally appropriate lifestyle intervention that spanned a substantial duration, and the focus on Latino youth with obesity, who represent a vulnerable population at elevated risk for the development of cardiometabolic consequences related to hepatic steatosis. Despite these strengths, we acknowledge the limitations of the study, which include the small sample size and the need for larger and more diverse cohorts. However, the number of individuals comprising the study cohort is comparable to or greater than those used in similar intervention studies conducted in adults where significant findings were observed [57, 58, 60–64]. We also recognize potential sources of bias, such as a predominantly male study sample, the potential effects of sex hormones on the EV proteome, and the role of heterogeneity in baseline metabolic characteristics in determining response to intervention. Further exploration of these factors is required to validate and extend the findings.

In summary, our findings reveal distinct effects of lifestyle intervention on cargo proteome, confirm the close relationship between MASLD and EV biology, and provide additional mechanistic insights into the disease and avenues for potential reversal of its pathophysiology. The complex nature of the observed changes emphasizes the need for further research to fully comprehend the mechanisms that contribute to improvements in liver health as a result of lifestyle intervention and to better understand and address the impact of hepatic steatosis in youth.

Supplementary Material

1

Fig S1. A) Top results from enrichment analysis B) Top pathways of involvement identified using the Reactome database in individuals who exhibited reductions in HFF as a result of the intervention. “All” refers to all differentially abundant proteins (adj-p <0.05), “Up” refers to proteins with higher abundance and “Down” refers to proteins with lower abundance.

3

Fig S2. Volcano plot depicting changes in abundance (post/pre intervention) of proteins from hepatocyte-enriched EVs from plasma. The x-axis represents the log₂-fold change (FC), and the y-axis represents the -log₁₀ of the p-values. A total of 1528 proteins were detected, and those with both a fold-change and a p-value are represented by a point on the plot. The red dots indicate proteins showing significant suggestive evidence (p <0.05) for increased abundance, while the blue dots represent proteins showing significant suggestive evidence (p <0.05) for decreased abundance. Black dots depict proteins not meeting statistical significance for differential abundance.

Abbreviations:

EVs

extracellular vesicles

HFF

Hepatic fat fraction

C2

complement 2

C4

complement 4

MASLD

metabolic dysfunction-associated fatty liver disease

NAFLD

nonalcoholic fatty liver disease

HCC

hepatocellular carcinoma

BMI

body mass index

ALT

alanine aminotransferase

AST

aspartate aminotransferase

MRI

magnetic resonance imaging

SEC

size exclusion chromatography

ASGR1

asialoglycoprotein receptor 1

CYP2E1

cytochrome P450 2E1

DOC

sodium deoxycholate

BCA

bicinchoninic acid assay

LC-MS/MS

liquid chromatography tandem-mass spectrometry

SAT

subcutaneous adipose tissue

HDL

high density lipoprotein

C8G

complement C8 gamma chain

A1BG

alpha-1-B glycoprotein

IGHG4

immunoglobulin heavy constant gamma 4

CFB

complement factor B

CP

ceruloplasmin

C8B

complement C8 beta chain

LRP1

LDL receptor related protein 1

PROS1

protein S

JCHAIN

joining chain of multimeric IgA and IgM

C4BPA

complement component 4 binding protein alpha

FGL1

fibrinogen-like protein 1

RBP4

retinol binding protein 4

F12

coagulation factor XII

GC

vitamin D binding protein

HRG

histidine rich glycoprotein

LRG1

leucine-rich alpha-2 glycoprotein

MBL2

mannose-binding protein C

ORM1

alpha-1 -acid glycoprotein 1

SERPINA6

corticosteroid-binding globulin

C2

complement 2

LCN1

lipocalin 1

ME2

malic enzyme 2

CFP

complement factor properdin

BTD

biotinidase

LRP1

LDL receptor related protein 1

MAC

membrane attack complex

A2M

alpha-2-macroglobulin

COL6A3

collagen alpha-3(VI) chain