Innovative biomarkers are needed to improve the management of patients with type 2 diabetes mellitus (T2DM). Blood circulating miRNAs have been proposed as a potential tool to detect T2DM complications, but the lack of tissue specificity, among other reasons, has hampered their translation to clinical settings. Extracellular vesicle (EV)-shuttled miRNAs have been proposed as an alternative approach. Here, we adapted an immunomagnetic bead–based method to isolate plasma CD31+ EVs to harvest vesicles deriving from tissues relevant for T2DM complications. Surface marker characterization showed that CD31+ EVs were also positive for a range of markers typical of both platelets and activated endothelial cells. After characterization, we quantified 11 candidate miRNAs associated with vascular performance and shuttled by CD31+ EVs in a large (n = 218) cross-sectional cohort of patients categorized as having T2DM without complications, having T2DM with complications, and control subjects. We found that 10 of the tested miRNAs are affected by T2DM, while the signature composed by miR-146a, -320a, -422a, and -451a efficiently identified T2DM patients with complications. Furthermore, another CD31+ EV-shuttled miRNA signature, i.e., miR-155, -320a, -342-3p, -376, and -422a, detected T2DM patients with a previous major adverse cardiovascular event. Many of these miRNAs significantly correlate with clinical variables held to play a key role in the development of complications. In addition, we show that CD31+ EVs from patients with T2DM are able to promote the expression of selected inflammatory mRNAs, i.e., CCL2, IL-1α, and TNFα, when administered to endothelial cells in vitro. Overall, these data suggest that the miRNA cargo of plasma CD31+ EVs is largely affected by T2DM and related complications, encouraging further research to explore the diagnostic potential and the functional role of these alterations.

Type 2 diabetes mellitus (T2DM) is a chronic, heterogeneous disease caused by a multi-layer interaction between the genetic makeup and environmental/lifestyle-dependent factors (1). T2DM is a cause of morbidity and mortality mainly due to its cardiovascular (CV) complications, e.g., ischemic heart disease and peripheral vascular disease (2). A number of additional alterations occur in the development of CV complications in patients with diabetes (1,3). Given the complexity and heterogeneity of the disease, many potential biomarkers for the development of CV diseases in T2DM patients have been explored (4).

Blood circulating miRNAs have been proposed as a potential tool to improve CV risk assessment among patients with T2DM and other conditions (57). Given plasma miRNA stability and their ability to sense environmental stressors and modulate multiple pathways accordingly, miRNAs seem to be the ideal candidates to provide useful information in complex and heterogeneous diseases like T2DM complications and CV diseases in general (8,9). However, at present, neither single miRNAs nor miRNA signatures have been translated into clinical settings for diagnostic purposes. Many factors have hampered their use as biomarkers, including the lack of disease specificity and the confounding effect provided by the relative contributions of different tissues to the plasma miRNA pool (10,11). For overcoming these issues, the isolation of specific miRNA cargos, and in particular of extracellular vesicles (EVs), has been proposed (10,12).

EVs are membrane-coated nanoparticles actively and/or passively released by almost all cell types. EVs can be categorized according to a specific characteristic or by their size. Small EVs (diameter <100 nm) derive from either multivesicular bodies or the plasma membrane, while larger vesicles (between 100 nm and 1 μm) mostly derive from the plasma membrane (13). Both EV types are able to shuttle and deliver functional nucleic acids, including miRNAs. Blood contains a heterogeneous mixture of EVs of different origin, which are currently being characterized for therapeutic and diagnostic purposes (1416). In particular, recent evidence suggests a key role for EV-shuttled miRNAs in the etiopathogenesis of both T2DM and its vascular complications (1720). However, few studies have quantified the miRNA payload of circulating EVs in relation to human T2DM (21,22), while only one study assessed EV-shuttled miRNAs in a setting of T2DM complications (23). A recent article showed that the majority of circulating EV-shuttled miRNAs derives from the adipose tissue, a key organ for the development of T2DM (24). On the other side, platelets, immune cells, and endothelial cells play a prominent role in the development of T2DM-related complications (25,26). Given the marked and specific expression of CD31, i.e., platelet endothelial cell adhesion molecule 1 (PECAM1), in these three cell types, we adapted a previously published immunomagnetic method (27) using commercially available beads to selectively capture CD31+ EVs from plasma. After Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines–driven characterization, we measured a panel of 11 miRNAs previously associated with vascular performance in a cross-sectional cohort of 218 subjects, categorized as healthy control subjects (Ctrl), patients with T2DM but without complications (T2DM-NC), and T2DM patients with complications (T2DM-C). Here, we provide the first evidence of a specific miRNA signature shuttled by CD31+ EVs able to efficiently discriminate between T2DM patients with complications and those without complications and show that CD31+ EVs from patients with T2DM are able to promote proinflammatory pathways when administered to endothelial cells in vitro.

Cohort Description, Plasma Isolation, and Sample Size

Samples derive from a previously published cohort composed of 501 patients with T2DM and 400 healthy control subjects (28). Informed consent was obtained from each subject, and the study protocol was approved by the local research ethics committee. T2DM was diagnosed according to the American Diabetes Association criteria (29). Inclusion criteria for patients with diabetes were BMI <40 kg/m2, age 35–85 years, and ability and willingness to give written informed consent. Information collected included data on vital signs, anthropometric factors, medical history, and behaviors. The presence/absence of diabetes complications was evidenced as follows: diabetic retinopathy by fundoscopy through dilated pupils and/or fluorescence angiography; nephropathy, defined as a urinary albumin excretion rate >30 mg/24 h or an estimated glomerular filtration rate (eGFR) <60 mL/min per 1.73 m2; neuropathy established by electromyography; ischemic heart disease defined by clinical history or ischemic electrocardiographic alterations; and peripheral vascular disease, including atherosclerosis obliterans and cerebrovascular disease based on history, physical examinations, and Doppler velocimetry. Among the 101 patients analyzed with T2DM and at least one complication, 28 had neuropathy, 22 peripheral vascular disease, 20 nephropathy, 48 retinopathy, and 52 major adverse CV events (MACE). Healthy subjects were selected from a larger population of subjects belonging to a prevention program (30). Concentrations of common analytes were measured by standard procedures. As suggested by the MISEV initiative (31), fasting blood samples (collected in 4-mL plasma EDTA tubes) of all subjects were centrifuged at 753g at 4°C to obtain platelet-poor EDTA plasma and stored at −80°C within 3 h from blood collection. Sample size was calculated based on our previous publications (9,32,33).

Isolation of CD31+ Extracellular Vesicles

Plasma EDTA samples were allowed to thaw at room temperature. Appropriate volume (100 µL when used for singular miRNA dosage, as indicated for the other experiments) was diluted with an equal amount of PBS. For removal of apoptotic bodies and residual cellularity, sample were precleared by two subsequent centrifugations at 4°C: one at 2,000g for 30 min and the following at 10,000g for 45 min. Supernatant was diluted with PBS to reach 500 µL volume and then mixed with FcR Blocking Reagent (20 μL) provided in the commercially available kit for endothelial cells isolation (130-091-935; Miltenyi Biotec). After vortexing, 20 µL CD31 MicroBeads (130-091-935; Miltenyi Biotec) were added to the suspension and incubated at 4° C in the dark for 30 min. Appropriate (according to reaction volume) columns were mounted on the magnetic field and activated with PBS. After incubation, the mixture was loaded onto the column to allow separation. After three washes with 500 µL PBS, the column was removed from the magnetic support and CD31+ EVs were eluted in 500 µL PBS with the help of a plunger (Fig. 1A). As a negative control, isotype control beads (DynaBeads M-280) and no beads (equal amount of PBS) were used for parallel isolation of EVs for testing for eventual nonspecific bindings of EVs and subjected to MACSPlex comparison (methods below). We have submitted all relevant data of our experiments to the EV-TRACK knowledgebase (EV-TRACK identifier EV200038) (34).

Figure 1

Isolation and characterization of CD31+ EVs. A: Schematic representation of the isolation method. B: NTA of one representative sample of isolated CD31+ EVs, along with the observed mean size and number (from 1 mL pooled control plasma). C: Representative TEM image of EVs isolated with CD31 beads along with the relative magnification. D: Western blot showing the expression of CD31, Alix, TSG101, CD63, apoB100, and apoA1 in CD31+ EVs and UC-collected EVs isolated from the same amount of plasma, along with the relative densitometric analysis. Whole plasma was run as positive control for apoA1 and apoB100. E: Ratio between the expression of CD31 and CD9, CD63, or CD81 in CD31+ EVs and that in EVs isolated through UC, as measured with a specific kit allowing cytofluorimetric detection (n = 6 from pooled plasma split for performance of comparative isolation starting from the same volume). F: Comparative cytofluorimetric detection of CD49e, CD9, CD63, CD62P, CD81, CD41b, CD42a, CD29, and CD69 in EVs isolated with no beads, scramble IgG beads, and CD31 beads (n = 3, from equal amount of control plasma samples). G: Concentration of collected CD31+ EVs vs. the CD31-depleted fraction of EVs subjected to UC, measured with standard NTA (n = 3). H: RT-PCR dosage of miR-126-3p, miR-146a-5p, miR-155, and miR-21-5p in whole plasma vs. total EVs isolated with UC vs. CD31+ EVs, with division of the same control samples in different aliquots (same volume, 100 μL) for comparison of the relative abundance in the various compartments (n = 8). Errors bar are ±SD. *P < 0.05, **P < 0.01, ***P < 0.001, Student t test for panels D, E, and G and one-way ANOVA for panels F and H. #P < 0.05 vs. UC for panel H. APC, allophycocyanin; a.u., arbitrary units.

Figure 1

Isolation and characterization of CD31+ EVs. A: Schematic representation of the isolation method. B: NTA of one representative sample of isolated CD31+ EVs, along with the observed mean size and number (from 1 mL pooled control plasma). C: Representative TEM image of EVs isolated with CD31 beads along with the relative magnification. D: Western blot showing the expression of CD31, Alix, TSG101, CD63, apoB100, and apoA1 in CD31+ EVs and UC-collected EVs isolated from the same amount of plasma, along with the relative densitometric analysis. Whole plasma was run as positive control for apoA1 and apoB100. E: Ratio between the expression of CD31 and CD9, CD63, or CD81 in CD31+ EVs and that in EVs isolated through UC, as measured with a specific kit allowing cytofluorimetric detection (n = 6 from pooled plasma split for performance of comparative isolation starting from the same volume). F: Comparative cytofluorimetric detection of CD49e, CD9, CD63, CD62P, CD81, CD41b, CD42a, CD29, and CD69 in EVs isolated with no beads, scramble IgG beads, and CD31 beads (n = 3, from equal amount of control plasma samples). G: Concentration of collected CD31+ EVs vs. the CD31-depleted fraction of EVs subjected to UC, measured with standard NTA (n = 3). H: RT-PCR dosage of miR-126-3p, miR-146a-5p, miR-155, and miR-21-5p in whole plasma vs. total EVs isolated with UC vs. CD31+ EVs, with division of the same control samples in different aliquots (same volume, 100 μL) for comparison of the relative abundance in the various compartments (n = 8). Errors bar are ±SD. *P < 0.05, **P < 0.01, ***P < 0.001, Student t test for panels D, E, and G and one-way ANOVA for panels F and H. #P < 0.05 vs. UC for panel H. APC, allophycocyanin; a.u., arbitrary units.

Isolation of Extracellular Vesicles Through Ultracentrifugation

For comparative experiments, an aliquot of 1 mL plasma was precleared as indicated above and then the supernatant was diluted with PBS and subjected to ultracentrifugation (UC) at 120,000g (4ºC) in a Thermo Scientific S110AT rotor in a Sorvall MX 150 ultracentrifuge for 1.5 h. Pellets were resuspended in PBS and ultracentrifuged again at 120,000g for an additional 1.5 h. The final pellets were resuspended in 500 μL PBS. In one case, the depleted fraction of the CD31 isolation method was collected and subjected to the same steps of UC to undergo nanoparticle tracking analysis (NTA).

NTA

CD31+ EVs were isolated starting from 1 mL plasma as described and resuspended in PBS. Size and concentration of vesicles were determined with NanoSight LM10 equipment (Malvern Panalytical) using different dilutions (35) and with the following parameters: camera at 30 frames/s, camera level at 16, temperature between 21 and 25°C, and video recording time 60 s. Three videos were recorded for each sample and analyzed with NanoSight NTA 3.1 software. Data were expressed as mean ± SD of the three videos.

Transmission Electron Microscopy

For exploration of vesicles morphology with transmission electron microscopy (TEM), 30 µL CD31+ EVs samples were diluted with PBS, allowed to dry on top of Formvar carbon-coated grids for 25 min, and contrasted with 2% uranyl acetate for 2 min. Preparations were observed in a JEOL 1010 100 kV Electron Microscope.

Western Blot

For Western blot experiments, CD31+ EVs were ultracentrifuged to allow PBS discharge and direct application of lysis buffer to the EV pellet. The same was done with EVs isolated through ultracentrifuge. EV lysates were prepared in radioimmunoprecipitation assay buffer containing a protease inhibitor cocktail and quantified using the Bradford method. Next, the lysates were subjected to SDS-PAGE and transferred to nitrocellulose membranes (Whatman). Membranes were then incubated with the primary antibodies overnight at 4°C. The following primary antibodies were used: CD31 (cat. no. 3528; Cell Signaling Technology), Alix (92880; Cell Signaling Technology), TSG101 (ab125011; Abcam), CD63 (ab59479; Abcam), apolipoprotein (apo)B100 (ab20737; Abcam), and apoA1 (sc-30089; Santa Cruz Biotechnology). Whole plasma proteins (50 μg) were also run as positive control for apoA1. After incubation with the specific horseradish peroxidase–conjugated antibody (Vector), proteins were detected by enhanced chemiluminescence (GE Healthcare) and band densities were quantified by densitometry with ImageJ software.

Cytofluorimetric Detection of EV Markers

A commercially available (cat. no. 130-108-813, MACSPlex Exosome Kit; Miltenyi Biotec) and previously validated (36) kit was used for cytofluorimetric detection of a large range of markers in isolated EVs. Briefly, EVs isolated starting from the same amount of plasma were prepared as described in the manufacturer protocol. The multiplex bead-based platform was analyzed by flow cytometry with use of a BD FACSCanto II flow cytometer with the corresponding software (Becton, Dickinson and Company, Franklin Lakes, NJ) equipped with a 488-nm and a 640-nm laser. Fluorescence emission was collected by 530/30 nm, 585/42 nm, and 660/20 nm bandpass filters. At least 1,000 beads per sample were examined, and mean fluorescence intensity was determined with use of BD FACSDiva 6.1 software. Background signals were determined by analysis of beads incubated only with the respective staining antibodies and subtracted from the signals obtained for beads incubated with EVs and stained with the corresponding antibody. The multiplex bead-based platform includes setup beads for flow cytometer setup.

RNA Extraction and miRNA Profiling

Plasma samples from four subjects were pooled to reach 1 mL. CD31+ EVs were isolated from five Ctrl preparations and five T2DM preparations. RNA was extracted with a commercial kit known to enrich small RNA species (Norgen Biotek Corporation). The same amount of RNA was converted to cDNA by priming with a mixture of looped primers according to the manufacturer’s instructions (Megaplex kit; Applied Biosystems). cDNA (9 µL) was used for mature miRNA profiling by a real-time PCR instrument equipped with a 384-well reaction plate and human miRNA array pool A containing 367 different human miRNA assays in addition to selected small nucleolar RNAs and negative controls (non-human miRNAs). Only miRNAs expressed in more than one sample were included in the analysis. 2−Ct of the average values of each miRNA were used to build the heat map comparing Ctrl and T2DM with the ClustVis web tool (https://biit.cs.ut.ee/clustvis/) (37). Profiling raw data were deposited in Gene Expression Omnibus (GEO) and are accessible with the accession number GSE142553.

Single miRNA Quantitation

For single miRNA quantification, CD31+ EVs were isolated from 100 µL plasma. After mixing with lysis buffer and before loading to the RNA separation column (Norgen Biotek Corporation), the synthetic Caenorhabditis elegans miRNA, cel-miR-39, was spiked into plasma before RNA extraction. Only samples with cel-miR-39 recovery >95% were used in subsequent analyses. Reverse transcription and miRNA amplification were performed as previously described (9). Relative expression corresponded to the 2−∆Ct value. Given the lack of an adequate endogenous control for plasma circulating miRNAs (11), miRNA expression levels were generally normalized by cel-miR-39 levels, unless indicated otherwise. For validation of the four-miRNA signature as a predictor of MACE in T2DM patients, global mean normalization was performed for each miRNA by subtraction of the mean of the Ct values of all the miRNAs assessed in sample i from each individual Ct value from sample i. To compare the diagnostic performance of CD31+ EV-shuttled miRNAs with their whole plasma counterparts, we used previously published data by our group for both miR-146a-5p (33) and miR-21-5p (9), extracting the miRNA quantitation data for the same patients of this study. Previous data were generated using the same amount of plasma, the same quantitation technology, and the same standardization method, thus allowing data comparison through the relative receiver operating characteristic (ROC) curves as detailed below.

EV Fluorescent Labeling, Loading With cel-miR-39 or Fluorescent Small RNA, and Treatment of Endothelial Cells

For use of CD31+ EVs for in vitro experiments, these were first detached from beads to avoid nonspecific toxicity and allow proper EV delivery to recipient cells. Briefly, EV/beads complexes (in 1.5-mL tubes) were placed on the magnetic stand for 2 min. Then, PBS is removed and 300 μL Exo-Flow buffer (System Biosciences) was added. After incubation on a shaker at 25°C for 2 h, samples are placed on the magnetic stand for removal of the supernatant containing eluted EVs, without disturbing the bead pellet. Collected EVs were quantified using NTA (data not shown).

As in vitro model of endothelial cells, human umbilical vein endothelial cells (HUVECs) were used. HUVECs were cultivated as previously described (38,39). EV-depleted FBS (through overnight centrifugation) was used for all the experiments. We fluorescently labeled 1 × 109 EVs using PKH67 membrane dye (Sigma-Aldrich). Labeled EVs were washed in 10 mL PBS, collected by UC, resuspended in PBS, and then incubated with 50,000 recipient HUVECs for 24 h. HUVEC nuclei were counterstained with PBS-diluted DAPI dihydrochloride (Sigma-Aldrich) for 15 min, and cells were imaged at a widefield microscopy (ZEISS Axio Observer A1).

HUVECs were also treated for 24 h with EVs transfected with the nonhuman cel-miR-39 or with a small fluorescent RNA. EVs were transfected with the Exo-Fect kit (System Biosciences) according to the manufacturer’s instructions. Briefly, EVs were prepared for transfection by combining of Exo-Fect solution, 20 pmol cel-miR-39 (or the small fluorescent RNA), PBS, and 1 × 108 EVs. The transfection solution was incubated at 37°C for 10 min and then put on ice. The ExoQuick-TC reagent supplied in the kit was added to stop the reaction. After centrifuging for 3 min at 140,000 rpm, the supernatant was removed. The transfected EV pellet was suspended in 300 μL PBS, and 150 μL transfected EVs was added to 50,000 HUVECs cultivated in six wells. The same amounts of cel-miR-39 and Exo-Fect reagent were used as negative control.

For assessment of the proinflammatory effect of EVs, CD31+ EVs were isolated from 1 mL plasma from Ctrl, T2DM-NC, or T2DM-C subjects; detached from beads; and used to treat 50,000 HUVECs cultivated in six wells for 24 h. mRNA measurement by RT-PCR was performed as previously described (38). The primers used were as previously reported (39).

Statistical Analysis

Continuous variables were tested for normality with the Shapiro-Wilk test and reported as mean ± SD. For comparison of the expression of CD31+ EV-shuttled miRNAs in the three different groups, Kruskal-Wallis followed by Dunn post hoc test was used, while for comparison of data from control subjects and T2DM patients Mann-Whitney U test was applied. Categorical variables were compared with use of the χ2 test. Pearson correlation was used to assess correlations between continuous variables. One-way ANCOVA was used for evaluation of differences in continuous variables between groups with controlling for selected clinical and biochemical variables.

Multinomial logistic regression models were constructed to identify factors associated with the diagnosis of T2DM and its complications. A parsimonious backward-stepwise elimination of nonsignificant variables was deemed to be appropriate in our setting. Model fit was assessed with the Hosmer-Lemeshow goodness-of-fit test. The proportion of variance explained by the final model was determined with use of the Nagelkerke R2 statistic.

ROC curves were constructed for the single miRNAs and for the predicted probabilities derived from the logistic regression models. The Youden index was used to calculate the best cutoff values, where appropriate. Multiple ROC curves were compared with the DeLong method (40).

We carried out the analyses using IBM SPSS Statistics, version 26 (IBM, Armonk, NY) and R, version 3.6.1. Statistical significance was defined as a two-tailed P value <0.05.

Data and Resource Availability

Profiling raw data were deposited in GEO and are available with the accession number GSE142553. Data relative to EV isolation and characterization have been submitted to the EV-TRACK knowledgebase (EV-TRACK identifier EV200038). All of the other data generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Isolation and Characterization of CD31+ Extracellular Vesicles

After immunomagnetic capture from 1 mL pooled plasma samples from control subjects (Fig. 1A [details in research design and methods]), standard NTA was used to quantify our EV isolate. NTA revealed that our method is able to mostly isolate vesicles compatible with the size of small EVs (14) (Fig. 1B). Isolated EVs were also characterized by TEM, which showed round-shaped vesicles with a minimal presence of contaminants (Fig. 1C). To further characterize the collected EV population and to control their positivity for CD31, we compared this method with a standard approach to isolate EVs, i.e., UC (at 120,000g). We subjected EVs isolated from 1 mL plasma for comparative analysis (from the same pooled control samples, n = 6) for both Western blot and surface markers expression with a specific kit to detect EV proteins through cytofluorimeter (36). As suggested by MISEV guidelines (14), EVs collected with CD31 beads were positive for EV-associated transmembrane, i.e., CD63 and CD31, and cytosolic, i.e., Alix and TSG101, proteins (Fig. 1D). In addition, the CD31 beads method enriched the population of CD31+ EVs compared with UC, as demonstrated by a higher ratio of CD31 to conventional EV markers. In addition, since lipoproteins are often isolated as contaminants with various EV isolation methods, we tested the expression of apoA1 and apoB100, two major components of lipoproteins. Albeit positive with both CD31 beads and UC, the immunomagnetic method was accompanied by a lower presence of these contaminants (Fig. 1D). For substantiation of these findings, the same comparison was subjected to cytofluorimetric detection of the EV markers CD9, CD63, and CD81, as well as of CD31. As expected, the CD31 beads method enriched the population of CD31+ EVs compared with UC, as demonstrated by an increased ratio of CD31 to EV surface markers in CD31+ EVs compared with EVs isolated with UC (Fig. 1E). To explore whether the use of beads for isolation is associated with a nonspecific binding of EVs and to gain preliminary insights into the putative cell source of CD31+ EVs, we compared EVs isolated through CD31 beads with those eventually collected using beads with isotype control or using no beads, starting from the same amount of volume (1 mL) (n = 3) and using the same procedure. Cytofluorimetric detection of a large range of surface markers revealed a significantly higher expression of EV markers but also of a large range of epitopes typical of platelets, i.e., CD41b and CD42a, and activated endothelial cells, i.e., CD62P. However, CD49e, CD29, and CD69 (beyond tetraspanins) also were significantly increased with CD31 beads (Fig. 1F). Indeed, all of the other tested markers (36) were not expressed in our EVs (data not shown). To explore which fraction of total plasma EVs is represented by the isolated EVs, we collected the depleted fraction of the CD31 beads method and, after UC, subjected it to NTA comparison (control samples, n = 3), which revealed that the concentration of collected CD31+ EVs is significantly lower than its depleted counterpart (Fig. 1G). In addition, we compared the surface markers expression of CD31+ EVs versus the CD31-depleted fraction. Results evidenced a higher abundance of the platelet marker CD41b and a lower positivity for common EV markers, i.e., CD9, CD63, and CD81 in CD31+ EVs, compared with the depleted fraction (Supplementary Fig. 1A). To explore the yield efficiency of our isolation technique, we compared the abundance of CD31 in UC-isolated EVs, in isolated CD31+ EVs, and in the CD31-depleted fraction starting from the same samples processed in succession. We found no significant differences in the abundance of CD31 between the initial sample and CD31+ EVs, while a significant decrease was observed in the CD31-depleted fraction compared with the initial sample (Supplementary Fig. 1B). For testing of whether isolating CD31+ EVs yields quantitatively different results in terms of miRNA abundance compared with UC and with whole plasma (41,42), the same amount of plasma was used for the isolation of CD31+ EVs or EVs with UC or left untreated (except for preclearance) and then these three samples were subjected to RNA extraction to quantify four miRNAs commonly studied in settings of T2DM and cardiovascular disease, i.e., miR-126-3p, miR-146a-5p, miR-155, and miR-21-5p (10). All miRNAs were consistently expressed with the three tested approaches, being higher in total plasma, followed by EVs isolated through UC, and finally CD31+ EVs (Fig. 1H), in line with the observation that CD31+ EVs represent only a fraction of total plasma EVs. Overall, these data suggest that our isolation method harvests CD31+ EVs, representing a fraction of plasma EVs with a heterogenous origin but compatible with the hypothesis that platelets and endothelial cells contribute to this specific EVs pool.

Comparative Concentration, Size, and miRNA Profiling of CD31+ EVs From Control Subjects and Patients With T2DM

To explore whether T2DM affects the number and size of CD31+ EVs, we compared four pooled samples from healthy Ctrl subjects with four pooled samples from T2DM-NC subjects (1 mL each). NTA showed that the concentration of CD31+ EVs was not significantly affected by T2DM (Fig. 2A), while a slight but significant decrease in their modal size was also observed (Fig. 2B). Cytofluorimetric comparison of surface markers supported NTA data, since the expression of CD31, CD9, CD63, and CD81 was also slightly but not significantly increased in T2DM samples (n = 4) (Fig. 2C).

Figure 2

Comparative concentration, modal size, and miRNA profiling of CD31+ EVs from control subjects and patients with T2DM. NTA measurement of the concentration (A) and modal size (B) of CD31+ EVs isolated from healthy control subjects and patients with T2DM (n = 4). C: Comparative cytofluorimetric detection of CD31, CD9, CD63, and CD81 of CD31+ EVs isolated from control subjects and patients with T2DM (n = 4). D: Heat map showing miRNA profiling in CD31+ EVs from control subjects and patients with T2DM (n = 5 vs. 5, pooled samples). *P < 0.05, Student t test. APC, allophycocyanin; a.u., arbitrary units.

Figure 2

Comparative concentration, modal size, and miRNA profiling of CD31+ EVs from control subjects and patients with T2DM. NTA measurement of the concentration (A) and modal size (B) of CD31+ EVs isolated from healthy control subjects and patients with T2DM (n = 4). C: Comparative cytofluorimetric detection of CD31, CD9, CD63, and CD81 of CD31+ EVs isolated from control subjects and patients with T2DM (n = 4). D: Heat map showing miRNA profiling in CD31+ EVs from control subjects and patients with T2DM (n = 5 vs. 5, pooled samples). *P < 0.05, Student t test. APC, allophycocyanin; a.u., arbitrary units.

Then, we performed a comparative profiling of miRNA content within CD31+ EVs comparing Ctrl and T2DM patients (five vs. five samples, with each sample prepared from pooled plasma). Of the 367 profiled miRNAs, 103 were detectable in at least one sample (Fig. 2D) and 39 were expressed in four out of four (80%) of the tested samples. Comparison of Ctrl and T2DM evidenced a different relative quantity of miRNAs in CD31+ EVs (Fig. 2D).

Diagnostic Performance of a Selected miRNA Signature for T2DM Status and Complications

To explore the possible association of CD31+ EV-shuttled miRNAs with T2DM status and T2DM complications, we selected 11 miRNAs for two characteristics: 1) being proposed to play a role in the development of CV complications of T2DM or previously suggested to have diagnostic potential in CV studies (Supplementary Table 1 for supporting literature) and 2) being robustly expressed in our setting of isolated CD31+ EVs (Fig. 2B). This selected panel was composed of miR-126-3p, miR-146a-5p, miR-155, miR-195-5p, miR-21-5p, miR-24-3p, miR-320a, miR-342-3p, miR-376a, miR-422, and miR-451a. We quantified single miRNAs by quantitative PCR in CD31+ EVs isolated from plasma of a cross-sectional cohort of 218 individuals, including 60 healthy Ctrl, 57 T2DM-NC, and 101 T2DM-C subjects. The clinical, anthropometrical, and biochemical variables of the subjects are reported in Table 1.

Table 1

Demographic, clinical, and biochemical characteristics of the 218 enrolled subjects

VariablesCtrl (N = 60)T2DM-NC (N = 57)T2DM-C (N = 101)P
Age (years) 66.4 (9.9) 64.4 (9.3) 67.5 (8.0) 0.112 
Male sex, n 35 32 71 0.134 
BMI (kg/m226.0 (4.1) 29.0 (5.1)* 28.5 (4.4)* <0.001 
Waist-to-hip ratio 0.86 (0.08) 0.93 (0.07)* 0.95 (0.065)* <0.001 
Total cholesterol (mg/dL) 220.6 (41.4) 216.2 (40.1) 197.1 (40.1)*# 0.001 
LDL-C (mg/dL) 126.5 (34.8) 129.0 (34.7) 107.9 (30.8)*# <0.001 
HDL-C (mg/dL) 62.8 (16.2) 51.9 (16.0)* 49.5 (12.8)* <0.001 
Triglycerides (mg/dL) 107.98 (82.29) 162.39 (123.90)* 139.65 (95.05) 0.014 
Glucose (mg/dL) 95.58 (10.021) 154.72 (50.292)* 176.64 (51.768)* <0.001 
HbA1c (%) 5.913 (0.396) 7.358 (1.149)* 7.768 (1.281)* <0.001 
Insulin (UI/mL) 5.108 (2.941) 10.704 (18.396)* 6.273 (4.789)# 0.006 
HOMA index 1.22 (0.71) 4.43 (7.24)* 2.77 (2.32)*# <0.001 
WBC (n/mm35.99 (1.36) 6.55 (1.81) 6.70 (1.58)* 0.023 
Platelets (n/mm3225.1 (50.0) 222.2 (54.0) 214.1 (62.0) 0.447 
hs-CRP (mg/L) 2.25 (2.48) 2.70 (2.26) 2.58 (8.81) 0.539 
Creatinine (mg/dL) 0.81 (0.17) 0.84 (0.22) 1.04 (0.39)*# <0.001 
Azotemia (mg/dL) 37.3 (8.7) 37.7 (11.5) 44.0 (18.5)*# 0.006 
eGFR (mL/min) 83.2 (16.1) 82.9 (22.6) 72.3 (20.7)*# 0.001 
Uric acid (mg/dL) 4.60 (1.26) 4.95 (1.46) 4.85 (1.19) 0.305 
VariablesCtrl (N = 60)T2DM-NC (N = 57)T2DM-C (N = 101)P
Age (years) 66.4 (9.9) 64.4 (9.3) 67.5 (8.0) 0.112 
Male sex, n 35 32 71 0.134 
BMI (kg/m226.0 (4.1) 29.0 (5.1)* 28.5 (4.4)* <0.001 
Waist-to-hip ratio 0.86 (0.08) 0.93 (0.07)* 0.95 (0.065)* <0.001 
Total cholesterol (mg/dL) 220.6 (41.4) 216.2 (40.1) 197.1 (40.1)*# 0.001 
LDL-C (mg/dL) 126.5 (34.8) 129.0 (34.7) 107.9 (30.8)*# <0.001 
HDL-C (mg/dL) 62.8 (16.2) 51.9 (16.0)* 49.5 (12.8)* <0.001 
Triglycerides (mg/dL) 107.98 (82.29) 162.39 (123.90)* 139.65 (95.05) 0.014 
Glucose (mg/dL) 95.58 (10.021) 154.72 (50.292)* 176.64 (51.768)* <0.001 
HbA1c (%) 5.913 (0.396) 7.358 (1.149)* 7.768 (1.281)* <0.001 
Insulin (UI/mL) 5.108 (2.941) 10.704 (18.396)* 6.273 (4.789)# 0.006 
HOMA index 1.22 (0.71) 4.43 (7.24)* 2.77 (2.32)*# <0.001 
WBC (n/mm35.99 (1.36) 6.55 (1.81) 6.70 (1.58)* 0.023 
Platelets (n/mm3225.1 (50.0) 222.2 (54.0) 214.1 (62.0) 0.447 
hs-CRP (mg/L) 2.25 (2.48) 2.70 (2.26) 2.58 (8.81) 0.539 
Creatinine (mg/dL) 0.81 (0.17) 0.84 (0.22) 1.04 (0.39)*# <0.001 
Azotemia (mg/dL) 37.3 (8.7) 37.7 (11.5) 44.0 (18.5)*# 0.006 
eGFR (mL/min) 83.2 (16.1) 82.9 (22.6) 72.3 (20.7)*# 0.001 
Uric acid (mg/dL) 4.60 (1.26) 4.95 (1.46) 4.85 (1.19) 0.305 

Variables are expressed as mean (SD) unless otherwise indicated. P values from ANOVA for continuous variables and from χ2 tests of association for categorical variables.

*P < 0.05 vs. Ctrl.

#P < 0.05 vs. T2DM-NC. HDL-C, HDL cholesterol; WBC, white blood cells.

Analysis of the expression profiles among groups revealed significant differences for all the evaluated miRNAs except miR-376a. Supplementary Table 2 reports the relative expressions of the 11 miRNAs in each group, along with the results of the Student t (Ctrl vs. T2DM) and one-way ANOVA (Ctrl vs. T2DM-NC vs. T2DM-C) tests, while Fig. 3 shows box plots of the miRNA expression values across groups. Specifically, the circulating levels of five miRNAs, i.e., miR-21-5p, -146a, -342-3p, -422a, and -451a, are increased in T2DM patients, while five miRNAs, i.e., miR-24-3p, -126-3p, -155, -195-5p, and -320a, show decreased levels in T2DM. The post hoc comparisons between the T2DM-NC and T2DM-C groups revealed significant differences for miR-21-5p, -146a, -342-3p, -422a, and -451a (increased in T2DM-C) and for miR-320a (decreased in T2DM-C).

Figure 3

Expression levels of 11 miRNAs in CD31+ EVs from healthy control (CTR), T2DM-NC, and T2DM-C subjects. Violin plots with individual points showing the expression of miR-126-3p, miR-146a-5p, miR-155, miR-195-5p, miR-21-5p, miR-24-3p, miR-320a, miR-342-3p, miR-376a, miR-422, and miR-451a in a cohort of 218 individuals, 60 healthy CTR, 57 T2DM-NC, and 101 T2DM-C subjects. **P < 0.05, ***P < 0.01, Kruskal-Wallis followed by Dunn post hoc test. a.u., arbitrary units.

Figure 3

Expression levels of 11 miRNAs in CD31+ EVs from healthy control (CTR), T2DM-NC, and T2DM-C subjects. Violin plots with individual points showing the expression of miR-126-3p, miR-146a-5p, miR-155, miR-195-5p, miR-21-5p, miR-24-3p, miR-320a, miR-342-3p, miR-376a, miR-422, and miR-451a in a cohort of 218 individuals, 60 healthy CTR, 57 T2DM-NC, and 101 T2DM-C subjects. **P < 0.05, ***P < 0.01, Kruskal-Wallis followed by Dunn post hoc test. a.u., arbitrary units.

Therefore, ROC curves were generated to evaluate the diagnostic potential of these 10 miRNAs in detecting T2DM. Analysis of the ROC curves, shown in Fig. 4A, revealed an outstanding diagnostic accuracy (area under the curve [AUC] ≥0.90) for five miRNAs (miR-146a-5p, miR-155, miR-195-5p, miR-24-3p, and miR-422a), and an excellent accuracy (0.80 ≤ AUC < 0.90) for three miRNAs (miR-21-5p, miR-342-3p, and miR-451a). A second set of ROC curves was generated to assess the ability of the six miRNAs differentially regulated in T2DM-C vs. T2DM-NC to discriminate between the two groups. The diagnostic accuracy was acceptable for all of the six miRNAs, with AUCs ranging from 0.67 (miR-146a) to 0.80 (miR-342-3p) (Fig. 4B).

Figure 4

Diagnostic performance of the differentially expressed miRNAs in CD31+ EVs. ROC curves and the relative AUC for differentially expressed miRNAs showing the diagnostic performance for T2DM vs. Ctrl (A) and T2DM-C vs. T2DM-NC (B). ROC curves for miR-146a-5p and miR-21-5p shuttled in CD31+ EVs compared with those of the same miRNAs measured in the same amount of whole plasma with the relative diagnostic performance for detection of T2DM vs. Ctrl (C) and T2DM-C vs. T2DM-NC (D).

Figure 4

Diagnostic performance of the differentially expressed miRNAs in CD31+ EVs. ROC curves and the relative AUC for differentially expressed miRNAs showing the diagnostic performance for T2DM vs. Ctrl (A) and T2DM-C vs. T2DM-NC (B). ROC curves for miR-146a-5p and miR-21-5p shuttled in CD31+ EVs compared with those of the same miRNAs measured in the same amount of whole plasma with the relative diagnostic performance for detection of T2DM vs. Ctrl (C) and T2DM-C vs. T2DM-NC (D).

To test whether harvesting CD31+ EVs increases the ability of selected plasma miRNAs to detect T2DM and its complications, we compared the diagnostic performance of miR-146a-5p and miR-21-5p shuttled in CD31+ EVs with that of the same miRNAs quantified in the same amount of whole plasma. ROC curves indicate that CD31+ EV-shuttled miR-146a-5p and miR-21-5p have a higher performance to detect both T2DM (AUC 0.911 vs. 0.562, P < 0.0001, and 0.859 vs. 0.595, P < 0.0001, respectively) (Fig. 4C) and T2DM complications (AUC 0.673 vs. 0.533, P = 0.028, and 0.744 vs. 0.511, P < 0.001) compared with their whole plasma counterparts (Fig. 4D), suggesting that the isolation of CD31+ EVs improves the diagnostic potential of plasma miR-146a-5p and miR-21-5p.

Diagnostic Performance of the Minimum miRNA Signature for T2DM Complications and MACE

To obtain the smallest possible signature with the highest discriminatory power for T2DM complications, we built a binary logistic regression to ascertain the effects of the 11 miRNAs, expressed as Z scores, on the likelihood of complications in T2DM patients, with a backward stepwise procedure to achieve the most parsimonious model. The logistic regression model was statistically significant (χ2(4) = 58.611, P < 0.001) and explained 42.5% (Nagelkerke R2) of the variance. Four miRNAs were retained into the model as significant predictors, i.e., miR-146a, -320a, -422a, and -451a (Supplementary Table 3). A similar model was built including BMI and LDL cholesterol (LDL-C) as covariates, since these variables were not balanced between groups. As shown in Supplementary Table 4, inclusion of these covariates marginally affected the results.

Next, we tested the association between the four-miRNA signature and the risk of MACE in T2DM patients. After inclusion in the model of HbA1c and the common risk factors age, sex, LDL-C, and hypertension as covariates, we still observed a strong association (P < 0.001) between our signature and history of MACE. This logistic regression model was statistically significant (χ2(6) = 102.960, P < 0.001) and explained 66.7% of the variance (Table 2). For resting of whether this signature remains significant using a different normalization method (11), an additional logistic regression model was built after recalculation of the signature using the global mean (derived from all the quantified miRNAs)–normalized expression of four miRNAs. The regression model proved statistically significant (χ2(6) = 86.572, P < 0.001, R2 = 0.587) and included the four-miRNA signature (P < 0.001), increasing HbA1c (P = 0.031), and male sex (P = 0.044) as significant predictors of MACE (Table 2).

Table 2

Predictive value of miRNAs to detect MACE

βSEPOR (95% CI)
a) Four-miRNA model + risk factors (enter method)     
 Four-miRNA signature 9.601 1.753 <0.001  
 Age 0.072 0.040 0.073  
 Sex (reference category: female) 0.828 0.625 0.185  
 Hypertension 0.042 0.683 0.951  
 HbA1c 0.354 0.216 0.100  
 LDL 0.001 0.008 0.878  
b) Four-miRNA model (global mean normalization) + risk factors (enter method)     
 Four-miRNA signature 7.267 1.270 <0.001  
 Age 0.060 0.035 0.082  
 Sex (reference category: female) 1.135 0.563 0.044  
 Hypertension −0.358 0.597 0.549  
 HbA1c 0.435 0.202 0.031  
 LDL −0.003 0.008 0.665  
c) 11-miRNA model (backward method)     
 miR-155 −12.749 5.202 0.014 3 × 10−6 (1.084 × 10−10 to 0.078) 
 miR-195-5p 3.032 1.999 0.129 20.737 (0.412–1,043.745) 
 miR-24-3p −4.039 3.641 0.267 0.018 (1.4 × 10−5 to 22.126) 
 miR-320a 0.997 0.410 0.015 2.709 (1.213–6.054) 
 miR-342-3p 0.704 0.304 0.021 2.021 (1.113–3.670) 
 miR-376a 0.898 0.371 0.015 2.454 (1.186–5.076) 
 miR-451a −1.010 0.354 0.004 0.364 (0.182–0.728) 
βSEPOR (95% CI)
a) Four-miRNA model + risk factors (enter method)     
 Four-miRNA signature 9.601 1.753 <0.001  
 Age 0.072 0.040 0.073  
 Sex (reference category: female) 0.828 0.625 0.185  
 Hypertension 0.042 0.683 0.951  
 HbA1c 0.354 0.216 0.100  
 LDL 0.001 0.008 0.878  
b) Four-miRNA model (global mean normalization) + risk factors (enter method)     
 Four-miRNA signature 7.267 1.270 <0.001  
 Age 0.060 0.035 0.082  
 Sex (reference category: female) 1.135 0.563 0.044  
 Hypertension −0.358 0.597 0.549  
 HbA1c 0.435 0.202 0.031  
 LDL −0.003 0.008 0.665  
c) 11-miRNA model (backward method)     
 miR-155 −12.749 5.202 0.014 3 × 10−6 (1.084 × 10−10 to 0.078) 
 miR-195-5p 3.032 1.999 0.129 20.737 (0.412–1,043.745) 
 miR-24-3p −4.039 3.641 0.267 0.018 (1.4 × 10−5 to 22.126) 
 miR-320a 0.997 0.410 0.015 2.709 (1.213–6.054) 
 miR-342-3p 0.704 0.304 0.021 2.021 (1.113–3.670) 
 miR-376a 0.898 0.371 0.015 2.454 (1.186–5.076) 
 miR-451a −1.010 0.354 0.004 0.364 (0.182–0.728) 

Binary logistic regression analyses of miRNAs associated with MACE in T2DM patients (a) and evaluation of the predictive value for MACE of the four-miRNA model (b) and of the four-miRNA model after global mean normalization (c) with adjustment for the conventional risk factors. Where applicable, ORs (95% CI) are expressed per 0.5-SD increase of each miRNA.

To assess whether other CD31+ EV-shuttled miRNA signatures are associated with the development of MACE in T2DM patients, we built an additional logistic regression model with backward selection on the 11 miRNAs, including age, sex, LDL-C, HbA1c, and hypertension as covariates. Again, the logistic regression model was statistically significant (χ2(7) = 155.777, P < 0.001) and explained 87.3% of the variance. Of the seven miRNAs that were retained into the model as predictors, five were statistically significant, i.e., miR-155, -320a, -342-3p, -376, and -422a. Table 2 summarizes the model and shows the adjusted odds ratios (ORs) for each miRNA.

For assessment of whether CD31+ EV-shuttled miRNAs associate with other complications of T2DM, multiple one-way ANCOVAs were computed to explore the relationship between the 11 miRNAs and T2DM complications after adjustment for age and sex. Supplementary Table 5 reports the adjusted means for each miRNA in subjects with or without a specific complication. We observed a significant differential regulation of all the CD31+ EV-shuttled miRNAs, except miR-126, in macrovascular complications, i.e., peripheral artery disease and MACE. The association between the levels of nine miRNAs and MACE remained significant even after adjustment for HbA1c and the presence of other concomitant T2DM complications. On the contrary, no significant association was found between miRNAs and any of the microvascular complications. Since blood miRNAs and especially platelet-derived miRNAs have been shown to be affected by antiplatelet therapies (43), we explored whether antiplatelet medications affected our results. However, we did not observe any significant effect of antiplatelet therapy on miRNA expression (data not shown).

Correlations With Clinical Variables and Between miRNAs

We then explored the association between the 11 CD31+ EV-shuttled miRNAs and a large range of relevant biochemical and clinical variables. The resulting color-coded correlation plot is shown in Fig. 5A. The correlation coefficient ranged from −0.51 to 0.40. A similar correlation plot was drawn to highlight reciprocal correlations between the levels of the 11 miRNAs under investigation (Fig. 5B). The complete correlation matrix is reported in Supplementary Table 6. Notably, the levels of all the T2DM-associated miRNAs, except miR-320a, are linearly related to fasting glucose and, among these, seven miRNAs correlate also with HbA1c, whereas only two miRNAs correlate with HOMA of insulin resistance. While our miRNA signature showed no significant association with the presence of diabetic nephropathy, nine miRNAs are linearly related to serum creatinine, azotemia, or eGFR, with the direction of the correlation being concordant with the expected deterioration of the renal function with worsening glycemic control. Eight miRNAs showed an association with age, with six miRNAs also being inversely correlated with peripheral blood mononuclear cells telomere length. In addition, we observed a remarkable pattern of correlations between eight miRNAs and the lipid profile. Of note, seven out of eight of these miRNAs also showed a correlation with waist-to-hip ratio, fasting glucose, and HbA1c with an inverse trend compared with their correlation with lipid profile, suggesting a divergent effect of common CV risk factors on these variables. Overall, these data indicate that CD31+ EV-shuttled miRNAs may sense a wide range of common risk factors known to be key drivers of T2DM complications development. Finally, many of the miRNAs were significantly associated with one another (Fig. 5B), extending the knowledge that circulating miRNAs are highly correlated also to those shuttled by CD31+ EVs (11).

Figure 5

Correlations between tested miRNAs and clinical variables and reciprocal correlations among miRNAs. Color-coded correlogram showing the significant Pearson correlations between tested miRNAs and clinical variables (A) and the reciprocal correlations between miRNAs (B). The intensity of the color and the dimension of the points depend on the magnitude of the correlation. γ GT, γ-glutamyltransferase; HDL-C, HDL cholesterol; PAI-1, plasminogen activator inhibitor-1; RBC, red blood cells; WBC, white blood cells.

Figure 5

Correlations between tested miRNAs and clinical variables and reciprocal correlations among miRNAs. Color-coded correlogram showing the significant Pearson correlations between tested miRNAs and clinical variables (A) and the reciprocal correlations between miRNAs (B). The intensity of the color and the dimension of the points depend on the magnitude of the correlation. γ GT, γ-glutamyltransferase; HDL-C, HDL cholesterol; PAI-1, plasminogen activator inhibitor-1; RBC, red blood cells; WBC, white blood cells.

CD31+ EVs From Patients With T2DM Promote Inflammation in Endothelial Cells In Vitro

To explore whether the altered miRNA cargo of CD31+ EVs derived from patients with T2DM affects the functional properties of these EVs when administered in vitro to endothelial cells, we unbounded collected EVs from beads with a commercially available buffer and the help of a magnet (Fig. 6A [details in research design and methods]). First, we verified the ability of endothelial cells to uptake EVs stained with a fluorescent dye after a 24-h incubation (Fig. 6B [representative image of n = 3 EVs, preparation from control samples]). Since fluorescent dyes might be accompanied by nonspecific binding to the lipidic components of the cells (44), we loaded a small RNA with a red fluorophore into EVs and then used them to treat endothelial cells. We detected a consistent red fluorescence in recipient cells, suggesting that collected EVs were able to deliver small RNAs to recipient endothelial cells (Fig. 6C). To support this observation, we administered EVs transfected with a nonhuman miRNA, i.e., cel-miR-39, to endothelial cells (n = 3, EVs isolated from control samples). The same amount of EV-free cel-miR-39 (along with the transfecting reagent), PBS, and nontransfected EVs were used as negative controls. As shown in Fig. 6D, cel-miR-39 expression was higher in endothelial cells treated with transfected EVs compared with the same amount of this miRNA not loaded onto EVs, while cel-miR-39 was undetectable in the other two negative controls (data not shown). Given that many of the CD31+ EV miRNAs found to be altered by T2DM and its complications (Fig. 3) have previously been associated with the alteration of inflammatory pathways (16,25), we treated endothelial cells with CD31+ EVs derived from Ctrl, T2DM-NC, and T2DM-C (n = 3 each, from 1 mL plasma) and measured the expression of a panel of proinflammatory genes at the mRNA level. EVs from both T2DM-NC and T2DM-C significantly increased the expression of chemokine (C-C motif) ligand 2 (CCL2) (also referred to as MCP-1, i.e., MCP-1) and interleukin-1α (IL-1α) compared with EVs from Ctrl subjects, while only EVs from T2DM-C induced the expression of TNFα in recipient endothelial cells compared with both T2DM-NC and Ctrl. Finally, the expression of IL-6, chemokine (C-X-C motif) ligand CXCL-1, and CXCL-8 was not affected by any of the treatments (Fig. 6E), possibly suggesting a peculiar proinflammatory effect of EVs rather than a nonspecific inflammatory response.

Figure 6

In vitro treatment of endothelial cells with CD31+ EVs. A: Schematic representation of the method used to detach EVs from beads. B: Representative image of endothelial cells treated for 24 h with EVs previously stained with a fluorescent green, lipophilic dye (PKH67) and stained with DAPI to evidence cell nuclei. The relative merge is also shown. C: Representative image of endothelial cells treated for 24 h with EVs previously loaded with a fluorescent (Texas Red), small RNA and stained with DAPI. D: Relative expression of the nonhuman cel-miR-39 in endothelial cells treated with EVs transfected with cel-miR-39 or with the same of amount of the miRNA without EVs (n = 3) ***P < 0.01, Student t test. E: mRNA expression of CCL2, IL-1α, TNFα, IL-6, CXCL-1, and CXCL-8 in endothelial cells treated with EVs derived from Ctrl, T2DM-NC, and T2DM-C subjects (n = 3). *P < 0.05, **P < 0.01, one-way ANOVA. a.u., arbitrary units; RT, room temperature; w/o, without.

Figure 6

In vitro treatment of endothelial cells with CD31+ EVs. A: Schematic representation of the method used to detach EVs from beads. B: Representative image of endothelial cells treated for 24 h with EVs previously stained with a fluorescent green, lipophilic dye (PKH67) and stained with DAPI to evidence cell nuclei. The relative merge is also shown. C: Representative image of endothelial cells treated for 24 h with EVs previously loaded with a fluorescent (Texas Red), small RNA and stained with DAPI. D: Relative expression of the nonhuman cel-miR-39 in endothelial cells treated with EVs transfected with cel-miR-39 or with the same of amount of the miRNA without EVs (n = 3) ***P < 0.01, Student t test. E: mRNA expression of CCL2, IL-1α, TNFα, IL-6, CXCL-1, and CXCL-8 in endothelial cells treated with EVs derived from Ctrl, T2DM-NC, and T2DM-C subjects (n = 3). *P < 0.05, **P < 0.01, one-way ANOVA. a.u., arbitrary units; RT, room temperature; w/o, without.

Circulating miRNA quantification has already been proposed as a potential approach to evaluation of CV risk (43,45,46). However, while data on circulating miRNAs in the setting of CV diagnostic for the general population are promising (43), none of the miRNAs has been translated into the clinic for diagnostic purposes, including in the context of T2DM-related complications. Lack of standardization methods and the complex contribution of different tissues and pathological processes to circulating miRNA pool are among the reasons that limit their use (10,11). For overcoming these issues, the quantification in specific miRNA cargos has been proposed. Indeed, microvesicle (MV)-shuttled miR-126 and miR-199a but not freely circulating miRNA expression predict the occurrence of CV events in patients with stable coronary artery disease (18), while T2DM patients with prevalent cardiovascular disease show low miR-26a and miR-126 levels within large MVs (47). In particular, it was suggested that MV-shuttled miRNAs derive from endothelial cells (18,48), while the most abundant miRNAs in whole plasma are among those highly expressed in platelets (7,49,50). However, since a seminal paper indicated that the majority of exosomal (i.e., small EVs) miRNAs derive from adipose tissue (24), we decided to harvest EVs expressing CD31 in order to enrich EVs derived from platelets, endothelial cells, and immune cells, i.e., the most relevant cellular components in the etiopathogenesis of T2DM complications. We showed here that the isolation of CD31+ EVs results in an EV pool compatible with a platelet and endothelial cell origin, as shown by analysis of surface markers expression. However, since also other markers were positive, the origin of collected EVs is likely heterogenous.

A recent study has characterized the abundance and functional alterations of circulating EVs from patients with T2DM, showing a higher plasma EV concentration in individuals with diabetes, an observation obtained by isolation of EVs with both a commercial kit and UC (51). However, in comparison of surface marker expression, erythrocyte-derived EVs were significantly increased by T2DM, while a nonsignificant trend was observed for platelet/endothelial cell–derived particles (51), an observation compatible with our results. In addition, the same study showed that EVs from patients with T2DM are able to induce an inflammatory response in recipient monocytes in vitro. Another study found that EVs from patients with gestational diabetes mellitus also promote inflammation when administered to endothelial cells (52). Here, we extend these findings by showing that CD31+ EVs from patients with T2DM are also able to foster low-grade inflammation in recipient endothelial cells, with a variable effect in considering T2DM with or without complications. Given that EVs can shuttle a large repertoire of molecules, research aimed at studying the specific components fostering inflammatory pathways might provide useful regarding the etiopathogenesis of the disease, especially considering that low-grade inflammation is associated with the presence of T2DM complications (53). To our knowledge, only one study used the CD31 beads approach to isolate EVs from patients with T2DM (54). In that study, CD31+ EVs were shown to boost apoptosis resistance of vascular smooth muscle cells cultured in hyperglycemic conditions, an effect possibly mediated by membrane-bound platelet-derived growth factor-BB.

Few studies have quantified miRNA abundance within EVs in cohorts with diabetes (21,22,45,47,55). Interestingly, two studies found an increased abundance of miR-126-3p in EVs from T2DM patients (22,23), while here a slightly decreasing trend was observed for CD31+ EV-shuttled miR-126-3p. This miRNA was previously suggested as one of the most downregulated miRNAs in analysis of whole plasma or large endothelial MVs (9,18,56). On the contrary, miR-21-5p in CD31+ EVs showed an opposite trend compared with the previously observed decrease in plasma of T2DM patients, but not in total EVs (9,56). Assuming that CD31+ EVs derive mainly from platelets and endothelial cells, our results might appear consistent with the observation that hyperglycemia induced a downregulation of miR-126-3p in endothelial EVs (18) and in platelets (57). The two studies finding an increased expression of miR-126-3p (22,23) were performed with use of different isolation methods that collect a broader EV population, possibly suggesting that the effect of T2DM on miR-126-3p expression is divergent when different tissues are considered. Indeed, two different mouse models of insulin resistance showed an increased expression of miR-126-3p in the liver (58) and in the adipose tissue (59). Considering also that the majority of plasma EV miRNAs is held to derive from adipose tissue (24), it is conceivable that a broad, non-tissue-specific EV collection method provides opposite results if compared with an immunomagnetic method likely enriching for tissue-specific EVs. On the other side, miR-21-5p shuttled in total EVs is increased in T2DM patients with diabetic nephropathy compared with patients without complications (23,60), which is similar to our observation when considering CD31+ EVs-shuttled miR-21-5p and T2DM complications as a whole. Notably, even though miR-21-5p shuttled in CD31+ EVs was not associated with prevalent nephropathy in our cohort, its levels were correlated with multiple measures of kidney function. All these observations suggest that harvesting rare EV fractions or subpopulations might hold an increased potential for miRNA-biomarker discovery compared with broader EV collections or with whole plasma, possibly limiting the heterogeneous, pleiotropic effect of T2DM on the expression of miRNAs at the tissue level. The results showing an increased performance of CD31+ EV-shuttled miR-21-5p and miR-146a-5p compared with the whole plasma levels of the same miRNAs in detection of both T2DM and its complications might support this hypothesis.

Limitations of the Study

The main limitation of this study is that we used a cross-sectional cohort; thus, we cannot determine whether the obtained signature is able to longitudinally identify patients at risk of T2DM complications or specifically MACE. Moreover, given this study design, we cannot perform a direct comparison with already available tools. In addition, the BMI of patients with T2DM was significantly higher compared with control subjects. However, since we isolated a specific fraction of EVs, the concentration of which was not affected by T2DM, it is unlikely that the observed differences in miRNA abundance are solely ascribable to the diverse quantity of adipose tissue between T2DM patients and control subjects. In addition, plasma isolation was performed with a low centrifugation speed, which might have left residual platelets in the samples that could have then been activated by thawing. However, before magnetic isolation, samples were precleared with two subsequent centrifugations, thus minimizing the risk of a consistent contamination by platelet granules or fragments. Finally, we did not perform functional experiments to explore which components of the EV cargo are responsible for the observed proinflammatory effect.

Conclusion

In summary, we have here isolated CD31+ EVs in a large cohort of T2DM patients, showing that specific miRNA signatures associate with T2DM complications as a whole or individually with MACE. We also show that harvesting CD31+ EVs, compared with whole plasma, improves the ability of miR-21-5p and miR-146a-5p to detect T2DM and its complications. Finally, we also demonstrated that CD31+ EVs from T2DM patients are endowed with proinflammatory properties when administered in vitro to endothelial cells, overall encouraging further research to explore both the diagnostic potential and the functional role of T2DM-driven EV alterations.

This article contains supplementary material online at https://doi.org/10.2337/figshare.13102910.

F.P. and V.D.N. contributed equally to this work.

Funding. G.M. is funded by Fondazione Italiana Sclerosi Multipla (grant 2016/R/18 and 2018/S/5) and Progetti di Rilevante Interesse Nazionale (PRIN) 2017 K55HLC 001. P.d.C. is funded by Fondazione Italiana Sclerosi Multipla (FISM n. 2018/R/4). This work was also supported by grants from La Maratò de TV3 to A.C. and by the Italian Ministry of Health (Ricerca Corrente) to IRCCS MultiMedica.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.P., F.O., and A.C. conceived the idea and designed the study. V.D.N. performed the majority of experiments. J.S. and A.Gi. performed statistical analysis and prepared figures and tables. R.S. and E.M. performed miRNA profiling and Western blot experiments. C.C., M.P., and I.C. performed cytofluorimeter and TEM experiments. A.Gr., N.B., M.R.R., A.R.B., L.L.S., A.D.P., S.G., G.M., A.N., and P.d.C. revised the manuscript for intellectual content and provided additional expertise. F.P., J.S., F.O., and A.C. wrote the manuscript. The final version of the manuscript was approved by all authors. F.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
DeFronzo
RA
,
Ferrannini
E
,
Groop
L
, et al
.
Type 2 diabetes mellitus
.
Nat Rev Dis Primers
2015
;
1
:
15019
2.
Dal Canto
E
,
Ceriello
A
,
Rydén
L
, et al
.
Diabetes as a cardiovascular risk factor: an overview of global trends of macro and micro vascular complications
.
Eur J Prev Cardiol
2019
;
26
:
25
32
3.
Gaede
P
,
Lund-Andersen
H
,
Parving
HH
,
Pedersen
O
.
Effect of a multifactorial intervention on mortality in type 2 diabetes
.
N Engl J Med
2008
;
358
:
580
591
4.
Bachmann
KN
,
Wang
TJ
.
Biomarkers of cardiovascular disease: contributions to risk prediction in individuals with diabetes
.
Diabetologia
2018
;
61
:
987
995
5.
Willeit
P
,
Skroblin
P
,
Moschen
AR
, et al
.
Circulating microRNA-122 is associated with the risk of new-onset metabolic syndrome and type 2 diabetes
.
Diabetes
2017
;
66
:
347
357
6.
Zampetaki
A
,
Mayr
M
.
Sweet dicer: impairment of micro-RNA processing by diabetes
.
Circ Res
2015
;
117
:
116
118
7.
Mayr
M
,
Zampetaki
A
,
Kiechl
S
.
MicroRNA biomarkers for failing hearts
?
Eur Heart J
2013
;
34
:
2782
2783
8.
Zampetaki
A
,
Willeit
P
,
Drozdov
I
,
Kiechl
S
,
Mayr
M
.
Profiling of circulating microRNAs: from single biomarkers to re-wired networks
.
Cardiovasc Res
2012
;
93
:
555
562
9.
Olivieri
F
,
Spazzafumo
L
,
Bonafè
M
, et al
.
MiR-21-5p and miR-126a-3p levels in plasma and circulating angiogenic cells: relationship with type 2 diabetes complications
.
Oncotarget
2015
;
6
:
35372
35382
10.
Prattichizzo
F
,
Giuliani
A
,
De Nigris
V
, et al
.
Extracellular microRNAs and endothelial hyperglycaemic memory: a therapeutic opportunity
?
Diabetes Obes Metab
2016
;
18
:
855
867
11.
Sunderland
N
,
Skroblin
P
,
Barwari
T
, et al
.
MicroRNA biomarkers and platelet reactivity: the clot thickens
.
Circ Res
2017
;
120
:
418
435
12.
Coumans
FAW
,
Brisson
AR
,
Buzas
EI
, et al
.
Methodological guidelines to study extracellular vesicles
.
Circ Res
2017
;
120
:
1632
1648
13.
Witwer
KW
,
Théry
C
.
Extracellular vesicles or exosomes? On primacy, precision, and popularity influencing a choice of nomenclature
.
J Extracell Vesicles
2019
;
8
:
1648167
14.
Théry
C
,
Witwer
KW
,
Aikawa
E
, et al
.
Minimal Information for Studies of Extracellular Vesicles 2018 (MISEV2018): a position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines
.
J Extracell Vesicles
2018
;
7
:
1535750
15.
Prattichizzo
F
,
Giuliani
A
,
Sabbatinelli
J
, et al
.
Extracellular vesicles circulating in young organisms promote healthy longevity
.
J Extracell Vesicles
2019
;
8
:
1656044
16.
Prattichizzo
F
,
Micolucci
L
,
Cricca
M
, et al
.
Exosome-based immunomodulation during aging: a nano-perspective on inflamm-aging
.
Mech Ageing Dev
2017
;
168
:
44
53
17.
Xiao
Y
,
Zheng
L
,
Zou
X
,
Wang
J
,
Zhong
J
,
Zhong
T
.
Extracellular vesicles in type 2 diabetes mellitus: key roles in pathogenesis, complications, and therapy
.
J Extracell Vesicles
2019
;
8
:
1625677
18.
Jansen
F
,
Yang
X
,
Hoelscher
M
, et al
.
Endothelial microparticle-mediated transfer of microRNA-126 promotes vascular endothelial cell repair via SPRED1 and is abrogated in glucose-damaged endothelial microparticles
.
Circulation
2013
;
128
:
2026
2038
19.
Sun
Y
,
Shi
H
,
Yin
S
, et al
.
Human mesenchymal stem cell derived exosomes alleviate type 2 diabetes mellitus by reversing peripheral insulin resistance and relieving β-cell destruction
.
ACS Nano
2018
;
12
:
7613
7628
20.
Castaño
C
,
Novials
A
,
Párrizas
M
.
Exosomes and diabetes
.
Diabetes Metab Res Rev
2019
;
35
:
e3107
21.
Katayama
M
,
Wiklander
OPB
,
Fritz
T
, et al
.
Circulating exosomal miR-20b-5p is elevated in type 2 diabetes and could impair insulin action in human skeletal muscle
.
Diabetes
2019
;
68
:
515
526
22.
Ghai
V
,
Kim
TK
,
Etheridge
A
, et al
.
Extracellular vesicle encapsulated microRNAs in patients with type 2 diabetes are affected by metformin treatment
.
J Clin Med
2019
;
8
:
617
23.
Florijn
BW
,
Duijs
JMGJ
,
Levels
JH
, et al
.
Diabetic nephropathy alters the distribution of circulating angiogenic microRNAs among extracellular vesicles, HDL, and Ago-2
.
Diabetes
2019
;
68
:
2287
2300
24.
Thomou
T
,
Mori
MA
,
Dreyfuss
JM
, et al
.
Adipose-derived circulating miRNAs regulate gene expression in other tissues
.
Nature
2017
;
542
:
450
455
25.
Prattichizzo
F
,
De Nigris
V
,
Spiga
R
, et al
.
Inflammageing and metaflammation: the yin and yang of type 2 diabetes
.
Ageing Res Rev
2018
;
41
:
1
17
26.
Kaur
R
,
Kaur
M
,
Singh
J
.
Endothelial dysfunction and platelet hyperactivity in type 2 diabetes mellitus: molecular insights and therapeutic strategies
.
Cardiovasc Diabetol
2018
;
17
:
121
27.
Clayton
A
,
Court
J
,
Navabi
H
, et al
.
Analysis of antigen presenting cell derived exosomes, based on immuno-magnetic isolation and flow cytometry
.
J Immunol Methods
2001
;
247
:
163
174
28.
Testa
R
,
Olivieri
F
,
Sirolla
C
, et al
.
Leukocyte telomere length is associated with complications of type 2 diabetes mellitus
.
Diabet Med
2011
;
28
:
1388
1394
29.
American Diabetes Association
.
2. Classification and diagnosis of diabetes: Standards of Medical Care in Diabetes—2019
.
Diabetes Care
2019
;
42
(
Suppl. 1
):
S13
S28
30.
Testa
R
,
Bonfigli
AR
,
Salvioli
S
, et al
.
The Pro/Pro genotype of the p53 codon 72 polymorphism modulates PAI-1 plasma levels in ageing
.
Mech Ageing Dev
2009
;
130
:
497
500
31.
Witwer
KW
,
Buzás
EI
,
Bemis
LT
, et al
. Standardization
of sample collection, isolation and analysis methods in extracellular vesicle research
.
J Extracell Vesicles
2013
;
2
32.
Olivieri
F
,
Bonafè
M
,
Spazzafumo
L
, et al
.
Age- and glycemia-related miR-126-3p levels in plasma and endothelial cells
.
Aging (Albany NY)
2014
;
6
:
771
787
33.
Mensà
E
,
Giuliani
A
,
Matacchione
G
, et al
.
Circulating miR-146a in healthy aging and type 2 diabetes: age- and gender-specific trajectories
.
Mech Ageing Dev
2019
;
180
:
1
10
34.
Van Deun
J
,
Mestdagh
P
,
Agostinis
P
, et al.;
EV-TRACK Consortium
.
EV-TRACK: transparent reporting and centralizing knowledge in extracellular vesicle research
.
Nat Methods
2017
;
14
:
228
232
35.
Gardiner
C
,
Ferreira
YJ
,
Dragovic
RA
,
Redman
CW
,
Sargent
IL
. Extracellular
vesicle sizing and enumeration by nanoparticle tracking analysis
.
J Extracell Vesicles
2013
;
2
36.
Koliha
N
,
Wiencek
Y
,
Heider
U
, et al
.
A novel multiplex bead-based platform highlights the diversity of extracellular vesicles
.
J Extracell Vesicles
2016
;
5
:
29975
37.
Metsalu
T
,
Vilo
J
.
ClustVis: a web tool for visualizing clustering of multivariate data using principal component analysis and heatmap
.
Nucleic Acids Res
2015
;
43
:
W566
W570
38.
Mensà
E
,
Guescini
M
,
Giuliani
A
, et al
.
Small extracellular vesicles deliver miR-21 and miR-217 as pro-senescence effectors to endothelial cells
.
J Extracell Vesicles
2020
;
9
:
1725285
39.
Prattichizzo
F
,
De Nigris
V
,
Mancuso
E
, et al
.
Short-term sustained hyperglycaemia fosters an archetypal senescence-associated secretory phenotype in endothelial cells and macrophages
.
Redox Biol
2018
;
15
:
170
181
40.
DeLong
ER
,
DeLong
DM
,
Clarke-Pearson
DL
.
Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach
.
Biometrics
1988
;
44
:
837
845
41.
Turchinovich
A
,
Samatov
TR
,
Tonevitsky
AG
,
Burwinkel
B
.
Circulating miRNAs: cell-cell communication function
?
Front Genet
2013
;
4
:
119
42.
Turchinovich
A
,
Weiz
L
,
Langheinz
A
,
Burwinkel
B
.
Characterization of extracellular circulating microRNA
.
Nucleic Acids Res
2011
;
39
:
7223
7233
43.
Willeit
P
,
Zampetaki
A
,
Dudek
K
, et al
.
Circulating microRNAs as novel biomarkers for platelet activation
.
Circ Res
2013
;
112
:
595
600
44.
Simonsen
JB
.
Pitfalls associated with lipophilic fluorophore staining of extracellular vesicles for uptake studies
.
J Extracell Vesicles
2019
;
8
:
1582237
45.
Barwari
T
,
Joshi
A
,
Mayr
M
.
MicroRNAs in cardiovascular disease
.
J Am Coll Cardiol
2016
;
68
:
2577
2584
46.
Willeit
P
,
Skroblin
P
,
Kiechl
S
,
Fernández-Hernando
C
,
Mayr
M
.
Liver microRNAs: potential mediators and biomarkers for metabolic and cardiovascular disease
?
Eur Heart J
2016
;
37
:
3260
3266
47.
Jansen
F
,
Wang
H
,
Przybilla
D
, et al
.
Vascular endothelial microparticles-incorporated microRNAs are altered in patients with diabetes mellitus
.
Cardiovasc Diabetol
2016
;
15
:
49
48.
Jansen
F
,
Li
Q
,
Pfeifer
A
,
Werner
N
.
Endothelial- and immune cell-derived extracellular vesicles in the regulation of cardiovascular health and disease
.
JACC Basic Transl Sci
2017
;
2
:
790
807
49.
Kaudewitz
D
,
Skroblin
P
,
Bender
LH
, et al
.
Association of microRNAs and YRNAs with platelet function
.
Circ Res
2016
;
118
:
420
432
50.
Zampetaki
A
,
Willeit
P
,
Tilling
L
, et al
.
Prospective study on circulating microRNAs and risk of myocardial infarction
.
J Am Coll Cardiol
2012
;
60
:
290
299
51.
Freeman
DW
,
Noren Hooten
N
,
Eitan
E
, et al
.
Altered extracellular vesicle concentration, cargo, and function in diabetes
.
Diabetes
2018
;
67
:
2377
2388
52.
Salomon
C
,
Scholz-Romero
K
,
Sarker
S
, et al
.
Gestational diabetes mellitus is associated with changes in the concentration and bioactivity of placenta-derived exosomes in maternal circulation across gestation
.
Diabetes
2016
;
65
:
598
609
53.
Prattichizzo
F
,
Giuliani
A
,
Sabbatinelli
J
, et al
.
Prevalence of residual inflammatory risk and associated clinical variables in patients with type 2 diabetes
.
Diabetes Obes Metab
2020
;
22
:
1696
1700
54.
Togliatto
G
,
Dentelli
P
,
Rosso
A
, et al
.
PDGF-BB carried by endothelial cell-derived extracellular vesicles reduces vascular smooth muscle cell apoptosis in diabetes
.
Diabetes
2018
;
67
:
704
716
55.
Xiong
Y
,
Chen
L
,
Yan
C
, et al
.
Circulating exosomal miR-20b-5p inhibition restores Wnt9b signaling and reverses diabetes-associated impaired wound healing
.
Small
2020
;
16
:
e1904044
56.
Zampetaki
A
,
Kiechl
S
,
Drozdov
I
, et al
.
Plasma microRNA profiling reveals loss of endothelial miR-126 and other microRNAs in type 2 diabetes
.
Circ Res
2010
;
107
:
810
817
57.
Fejes
Z
,
Póliska
S
,
Czimmerer
Z
, et al
.
Hyperglycaemia suppresses microRNA expression in platelets to increase P2RY12 and SELP levels in type 2 diabetes mellitus
.
Thromb Haemost
2017
;
117
:
529
542
58.
Ryu
HS
,
Park
SY
,
Ma
D
,
Zhang
J
,
Lee
W
.
The induction of microRNA targeting IRS-1 is involved in the development of insulin resistance under conditions of mitochondrial dysfunction in hepatocytes
.
PLoS One
2011
;
6
:
e17343
59.
Fernandez-Twinn
DS
,
Alfaradhi
MZ
,
Martin-Gronert
MS
, et al
.
Downregulation of IRS-1 in adipose tissue of offspring of obese mice is programmed cell-autonomously through post-transcriptional mechanisms
.
Mol Metab
2014
;
3
:
325
333
60.
Zang
J
,
Maxwell
AP
,
Simpson
DA
,
McKay
GJ
.
Differential expression of urinary exosomal MicroRNAs miR-21-5p and miR-30b-5p in individuals with diabetic kidney disease
.
Sci Rep
2019
;
9
:
10900
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.