Umbilical Cord-Mesenchymal Stromal Cell-Derived Extracellular Vesicles Target the Liver to Improve Neurovascular Health in Type 2 Diabetes With Non-Alcoholic Fatty Liver Disease

Abstract

Type 2 diabetes mellitus (T2DM) combined with non-alcoholic fatty liver disease (NAFLD) exacerbates metabolic dysregulation and neurovascular complications, presenting significant therapeutic challenges. We demonstrate, using SPECT/CT imaging, that extracellular vesicles (EVs) from mesenchymal stromal cells (MSCs) predominantly accumulate in the liver, where they deliver miR-31-5p to suppress platelet-derived growth factor B (PDGFB) produced by hepatic macrophages. This intervention impedes NAFLD progression and establishes a mechanistic link between liver repair and neurovascular improvement. Specifically, single-nucleus RNA sequencing reveals that PDGFB suppression enhances hippocampal pericyte recovery via the PDGFB-PDGFRβ axis and orchestrates the activation of growth differentiation factor 11 (GDF11), thus promoting neuroplasticity. Furthermore, AAV injections indicate that hepatic PDGFB modulation recalibrates transthyretin (TTR) dynamics, thereby restoring its neuroprotective functions and preventing its pathological deposition in the brain. These findings position MSC-EVs as a transformative therapeutic platform that leverages the liver-brain axis to address the intertwined metabolic and neurovascular complications of T2DM, offering a promising avenue for clinical translation.

Keywords: NAFLD; PDGFB; cross‐organ; extracellular vesicles; neurovascular complications; type 2 diabetes mellitus.

FIGURE 1

EV distribution and pathological assessments in NAFLD mice. (A) Schematic procedures of EVs labelling with ^99mTc, created using Biorender.com. (B) SPECT/CT imaging of ^99mTc‐labeled EVs in db/db mice at 6‐, 24‐, and 48‐h post‐injection (planar and tomographic views). Images show predominant liver uptake over time. Display fields of view (DFOVs): 10.7, 10.6 and 10.3 cm. (C) Organ biodistribution of EVs at 1‐, 6‐, 24‐, and 48‐h post‐injection, demonstrating predominant liver and bladder accumulation (n = 8). (D) EV clearance kinetics from the liver and bladder at different time points, showing distinct fast and slow elimination phases (n = 8). (E) Total EV decay modelled using a metabolism equation, illustrating biphasic elimination kinetics (n = 8). (F) Representative liver histology in db/db mice at 2‐, 4‐, 6‐, and 8‐weeks post‐treatment. Staining includes H&E (morphology), Oil Red O (lipid deposition), Sirius Red (fibrosis) and α‐SMA (activated hepatic stellate cells, HSCs). WT mice serve as controls. Scale bars: 100 µm (H&E, Oil Red O, Sirius Red) and 50 µm (α‐SMA). (G–I) Quantification of Oil Red O, Sirius Red and α‐SMA staining in db/db mice, showing significant reductions in lipid accumulation, fibrosis, and HSC activation in EV‐treated mice (db/db‐EVs) compared to vehicle‐treated controls (db/db‐Veh) (n = 6). (J) NAFLD activity scores, demonstrating attenuated progression to NASH in db/db‐EVs mice compared to db/db‐Veh controls (n = 6). (K) Representative liver histology in MCD mice at 2‐ and 4‐weeks post‐treatment. Staining includes H&E, Oil Red O, and Sirius Red, with naïve (healthy) mice as controls. Scale bars: 100 µm. (L, M) Quantification of Oil Red O (n = 6) and Sirius Red (n = 6) staining in MCD mice, showing significantly reduced lipid deposition and fibrosis in EV‐treated groups (MCD‐EVs) compared to vehicle‐treated controls (MCD‐Veh). All data are presented as mean ± SEM. “n” represents biological replicates. Statistical analysis was performed using one‐way ANOVA with Bonferroni post hoc tests. Significance levels: ns (p > 0.05); *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

FIGURE 2

Assessments of neurological outcomes in NAFLD mice. (A–D) Neurobehavioral assessments including percentage of alternation triplets in the Y‐maze (A), speed (B) and time spent in the central area (C) in the open field test (OFT), and number of entries into the open arms (D) in the elevated plus maze (EPM) of db/db mice following EV treatment at 2, 4, and 6 weeks. Sample sizes for each time point: 2 weeks (n = 10); 4 weeks (WT, n = 10; db/db‐Veh and db/db‐EVs, n = 8); 6 weeks (n = 8). (E–G) Neurobehavioral tests in MCD mice 4 weeks post‐treatment, including the Y maze (A), OFT (B), and EPM (C) (n = 8). (H–M) ELISA analysis of serum biomarkers across groups (n = 6): VCAM‐1 (H), ET‐1 (I), IL‐1β (J), (H) TNF‐α (K), NSE (L) and BDNF(M). (N–Q) Western blots (N) and quantification of hippocampal neural damage markers for S100β (O), Aqp4 (P), and HMGB1 (Q) in tissues from WT, db/db‐Veh, and db/db‐EVs groups. β‐actin was used as a loading control (n = 8). Data are presented as mean ± SEM. “n” represents biological replicates. Statistical analysis was performed using one‐way ANOVA with Bonferroni post hoc tests. Significance levels: ns (p > 0.05); *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

FIGURE 3

Impact of EVs on cell populations and TTR changes in db/db hippocampus. (A) Bar graph illustrating the proportions of major hippocampal cell types across WT, db/db‐Veh, and db/db‐EVs groups 4 weeks post‐treatment (n = 3, pooled hippocampi from two mice to generate one biological replicate). (B) Pie charts showing the distribution of hippocampal cell clusters within each group. (C) UMAP plots identifying Cluster 24, characterized as a pericyte and endothelial cell complex. Cluster 24 co‐expresses pericyte markers (Kcnj8, RGS5, Mcam) and endothelial markers (Vwf, Cldn5, Flt1, Ctla2a). (D) Representative TEM images of hippocampal microvasculature showing structural improvements in db/db mice following EV therapy (n = 4). Features include pericytes (P), basement membrane (BM), cell gaps (CG), lumen (L), endothelial cells (EC), and myelin sheath (MS). Red dashed boxes highlight tight junctions, and red arrows indicate perivascular oedema. Scale bars: upper images = 2 µm; lower images = 1 µm. (E) Heatmap showing gene expression of TTR and other differentially expressed genes (DEGs) in Cluster 24. TTR was significantly upregulated in the db/db‐EVs group compared to db/db‐Veh (absolute fold change ≥1.2, adjusted p ≤ 0.05). (F) Biological process (BP) enrichment (g: Profiler2) associating TTR upregulation with BPs such as thyroid hormone transport and retinol metabolism processes in EV‐treated db/db hippocampus. (G) HALLMARK Gene Set enrichment (g: Profiler2) associating TTR upregulation with pathways such as Bile acid metabolism in db/db hippocampus following EV therapy. (H, I) ELISA of TTR concertation in both serum (H) and hippocampal tissue (I) across WT, db/db‐Veh, and db/db‐EVs groups (n = 5–7). (J, K) ELISA results show significantly increased levels of free thyroxine (FT4) (K) and all‐trans retinoic acid (ATRA) (L) in hippocampal tissues of EV‐treated db/db mice, compared to vehicle controls (n = 7). Data are presented as mean ± SEM. “n” represents biological replicates. Statistical significance was determined by one‐way ANOVA followed by Bonferroni's post hoc test (A, H‐K). Significance levels: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

FIGURE 4

Pericyte recovery and GDF11 activation in EV‐treated db/db hippocampus. (A) Network diagram showing the interactions among sub cluster C0, C1, and other cell types. The thickness and colour of the lines represent the strength and type of interactions. CellPhoneDB analysis highlights pericyte (C1)–excitatory neuron interactions in the experimental groups. (B–D) Western blot analysis (B) and quantification of pericyte markers RGS5 (C) and the pPDGFRβ/PDGFRβ ratio (D) demonstrate significant restoration of pericyte function in EVs‐treated db/db mice (n = 8). (E) 3D imaging shows PDGFRβ+ pericytes (red) co‐localized with CD34+ endothelial cells (green). White arrows indicate detached pericytes in db/db‐Veh mice, repaired after EV treatment. Scale bar: 10 µm. (F) GDF11‐TGFβ receptor interactions are significantly upregulated in db/db‐EVs‐treated mice, based on statistically enriched ligand‐receptor analysis. (G) ELISA analysis of GDF11 levels in hippocampal tissues of WT, db/db‐Veh, and db/db‐EVs groups (n = 5). (H–J) Western blot analysis (H) and quantification of GDF11 (I, n = 7) and pSMAD2/SMAD2 (J, n = 6) expression in hippocampal tissues of each group. (K, L) Confocal imaging (K) and Sholl analysis (L, n = 3 mice, with seven technical replicates) show recovery of dendritic complexity in CA1 pyramidal neurons (excitatory neurons) of db/db‐EVs‐treated mice. Scale bars: 10 µm (overview), 5 µm (magnified). (M–P) Quantification of spine morphology shows increased total spine density (M), stubby spine density (N), long thin spine density (O), and spine length (P) in db/db‐EVs‐treated mice (n = 3 mice, with nine technical replicates). (Q–S) mEPSC recordings (Q) and quantification of CA1 pyramidal neurons show significantly increased mEPSC frequency (R), but not amplitude (S), in db/db‐EVs‐treated mice (n = 3 mice, with three technical replicates). (T, U) Western blot (T) and quantification of Synapsin I (SYN1, U) show enhanced presynaptic connectivity in db/db‐EVs‐treated mice (n = 8). Data are presented as mean ± SEM. “n” refers to biological replicates for C, D, G, I, J, U, and to pooled biological and technical replicates for L–P, R, and S. Statistical significance was determined using one‐way ANOVA with Bonferroni post hoc test for (C, D, G, I, J, M–P, R, S, U) and repeated measures ANOVA for (L). Significance levels: ns, not significant (p > 0.05); *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

FIGURE 5

Effects of PDGFB modulation in the liver and brain of db/db mice. (A) Serum PDGFB levels measured by ELISA show significant increases in db/db‐EVs mice compared to db/db‐Veh (n = 8). (B) qPCR analysis reveals significantly decreased PDGFB mRNA in the liver of db/db‐EVs mice, with no changes in other tissues, including the hippocampus, heart and kidneys (n = 7). (C, D) Western blot (C) and quantification (D) show selective reduction of PDGFB protein in the liver of db/db‐EVs group, with no changes in the hippocampus (n = 8). (E) Representative immunohistochemistry images of liver showing marked changes of PDGFB expression among the experimental groups (n = 4). Scale bar: 50 µm. (F, G) Fluorescence in situ hybridization (FISH) images (F) reveal a significant reduction in PDGFB mRNA (pink) in liver macrophages (F4/80+, green) of db/db‐EVs mice, as quantified in (G) (n = 10 F4/80+ cells from three mice per group). Scale bars: 5 µm (magnified) and 10 µm. (H) ELISA analysis of serum PDGFB levels in db/db mice treated with AAV8‐NC (db/db‐NC) or AAV8‐GP‐1‐PDGFB knockdown (db/db‐PDGFB^AAV−), confirming successful PDGFB suppression (n = 5). (I–L) ELISA measurements of liver TNF‐α (I), IL‐10 (J), TGF‐β (K) and TTR (L) in db/db‐NC and db/db‐PDGFB^AAV− groups (n = 5). (M) H&E staining of liver section from db/db‐PDGFB^AAV− group showing reduced ballooning degeneration and inflammation in the liver, marked portal area, compared to db/db‐NC group. Scale bars: 20 µm (down) and 50 µm (up). (N–P) Neurobehavioral tests reveal improved Y‐maze alternation (N), increased central time in OFT (O), and more open‐arm entries in EPM (P) in db/db‐PDGFB^AAV− mice compared to NC mice (n = 8). (Q–T) ELISA analysis reveals the changes in hippocampal TNF‐α (Q), ET‐1 (R), TTR (S), and GDF11 (T) levels in db/db‐PDGFB^AAV− mice compared to NC groups (n = 5). Data are mean ± SEM. “n” refers to biological replicates for A, B, D, H‐L, N‐T, and to pooled biological and technical replicates for G. Statistical significance was determined by one‐way ANOVA with Bonferroni post hoc test for (A, B, D, G), unpaired two‐tailed Student's t‐tests (I‐L, Q‐T), and Mann‐Whitney U test for (N–P). Significance levels: ns, not significant (p > 0.05); *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

FIGURE 6

NGS analysis of EVs and the regulatory role of miR‐31‐5p in severe NAFLD. (A) Heatmap displaying the top 50 differentially expressed miRNAs identified in EVs from three independent samples (EVs‐1, EVs‐2, EVs‐3). Differential expression analysis using edgeR (FDR < 0.05; |log2(FC)| > 1). (B) KEGG pathway enrichment analysis of these miRNAs revealed significant involvement in immune‐related processes, including mitogen‐activated protein kinase (MAPK) and phosphoinositide 3‐kinase/protein kinase B (PI3K/Akt) signalling. (C–F) ELISA quantification of PDGFB (C), IL‐1β (D), IL‐10 (E) and TGFβ (F) in the supernatant of palmitic acid (PA)‐stimulated bone marrow‐derived macrophages (BMDMs, NASH model) treated with HEK‐EV NC, HEK‐EV mimic, or EV inhibitor (n = 5–6). (G) Quantification of Oil Red O‐positive areas (%) in experimental groups (Figure S10G) shows reduced lipid accumulation in the HEK‐EV mimic group, while inhibition of miR‐31‐5p in MSC‐EVs (EV inhibitor) abrogated this effect (n = 6). (H–J) Western blot analysis (H) and quantification of PDGFB effectors, including p‐AKT/AKT (I) and p‐p65/p65 (J), in PA‐stimulated BMDMs following treatment with miR‐31‐5p NC, mimic, or inhibitor. Untreated BMDMs served as the negative control (n = 6). (K) ELISA quantification of liver PDGFB levels in MCD mice treated with miR‐31‐5p NC or agonist (miR‐31‐5p^agomir) (n = 6). (L‐O) Immunohistochemistry (L) and quantification of PDGFB (M), α‐SMA (N), and CD68 (O) in liver sections from MCD mice treated with NC or miR‐31‐5p^agomir (n = 5–6). Scale bars: 50 µm. (P–R) Histological analysis of liver sections from MCD mice treated with NC or miR‐31‐5p^agomir. H&E, Oil Red O, and Sirius Red staining (P) reveal reduced lipid accumulation and fibrosis in the miR‐31‐5p^agomir group. Quantification of Oil Red O‐positive (Q) and Sirius Red‐positive (R) areas confirms these improvements (n = 6). Scale bars: 100 µm. (S‐U) Western blot analysis (S) and quantification of p‐AKT/AKT (I) and p‐p65/p65 (J) in MCD mice followed by miR‐31‐5p NC or agomir administration (n = 6). (V‐Y) Western blot analysis of hippocampal vascular markers (S) and quantification of VE‐Cadherin, Occludin and Connexin 43 in the NC‐ or agomir‐treated mice (n = 6). Data are presented as mean ± SEM. “n” indicates the number of experimental replicates (C–G, I, J) or biological replicates (K, M–O, Q, R, T, U, W–Y). Statistical significance was determined using one‐way ANOVA with Bonferroni post hoc tests (C–G) or unpaired two‐tailed Student's t‐tests (I–K, M–O, Q, R, T, U, W–Y). Significance levels: ns (p > 0.05); *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.