The animal care and experimental protocols were conducted in accordance with the Animal Management Rules of China (55, 2001, Ministry of Health, China), and the study was approved by the Animal Research Committee of Shanghai Jiao Tong University.

Male Sprague-Dawley (SD) rats with an average weight of 400 g were anesthetized with isoflurane inhalation (MATRX VIP 3000, United States). The left carotid arteries were exposed, and a percutaneous transluminal angioplasty balloon dilatation catheter (2 F, 0.67 mm, Edwards Lifesciences, United States) was used to establish vascular intimal injury (Raugi et al., 1990). The arteries were then harvested after 2 weeks, and the undamaged right carotid artery served as the self-control.

The carotid arteries samples were fixed in 4% paraformaldehyde, dehydrated in 30% sucrose solution, and then cut into 6-μm sections. The frozen sections were washed three times with PBS, permeabilized with 0.3% Triton X-100 for 30 min, and immersed in a solution of 10% goat serum for 30 min at room temperature to block nonspecific binding. Subsequently, the sections were incubated with the primary antibodies against Col8a1 (1:400, Proteintech Group, United States), Col8a2 (1:400, Abcepta, United States), SMA (1:500, Invitrogen, United States), vWF (1:500, Cell Signaling Technology, United States), CD41 (1:500, Cell Signaling Technology, United States) at 4°C overnight. After incubated with secondary antibody (1:1000. Abcam, United Kingdom) for 2 h, DAPI was used for nuclei staining for 15 min at room temperature. Staining at the cellular level, paraformaldehyde was used to fix VSMCs for 30 min and then followed the above method. Photographs were taken by confocal microscopy (LV1000; Olympus).

Six-μm frozen-sections of carotid artery samples were treated with 0.3% H₂O₂ to block endogenous peroxide activity and proteinase K (5 mg⋅mL^–1) for permeabilization. The samples were then hybridized with a miR-92a-3p biotinylated probe or a NC probe (Shanghai GenePharma, China) (200 nM) overnight at 56°C. The FISH signals were amplified with Tyramide SuperBoost Kits (Thermo Fisher Scientific, United States) followed by an Alexa Fluor Tyramide Kit (Thermo Fisher Scientific, United States), and photographed under confocal microscopy (LV1000; Olympus). The sequences of the RNA oligos were listed in Supplementary Table 1.

Multi-mode Ultrasound Imaging System (Fujifilm VisualSonics, United States) was used to detect the change in arterial diameter (%) of the carotid artery in rats after intimal injury. SD rats were anesthetized with isoflurane (MATRX VIP 3000, United States) and signals were collected using 20 MHz MX Series transducer (MX250S) in “M-Mode”. Data analysis were performed on FUJIFILM VisualSonics Measurement software.

Piuma Nanoindenter (Optics11, Netherlands) was used to detect the stiffness of the carotid artery after intimal injury (Xie et al., 2018). A probe with a 0.49 N⋅m^–1 spring constant and a 31 μm spherical indentation radius was used. All measurements were performed with the carotid artery flattened onto the bottom of a dish and submerged in PBS at room temperature. The indents were depth controlled (10 μm), and the loading and unloading period was set to be 2 s. Based on the load-displacement curves, the Young’s modulus was calculated using the Hertz spherical indentation model in Piuma Software (version: V3.3.0). For vascular stiffness analysis, 5-10 measurements were made on each carotid artery.

Nanoparticle tracking analysis was used to analyze the number and diameter of pEVs (Szatanek et al., 2017). The obtained pEVs were diluted with PBS and loaded into the NanoSight module (NanoSight NS300, United Kingdom) for measurement. The module was washed with PBS after each measurement.

The intimal injury arteries harvested after 2 weeks, and the undamaged right carotid artery served as the self-control, then performed transcriptome sequencing (GEO accession numbers: GSE164050). And the miR sequencing used by activated platelets and platelets released pEVs. Differentially expressed mRNAs or miRs were analyzed utilizing DESeq with the following criteria: fold change > 2 and false discovery rate (FDR) < 0.05. Afterwards, the ClustVis. web tool¹ was used to upload raw data and create heatmaps.

The possible biological processes and functional classifications were obtained with Ingenuity Pathway Analysis (IPA) software² (Content version: 57662101, Qiagen, Germany). “Formation of extracelluar matirx” related genes in “Diseases and Functions model” were fist selected based on transcriptome sequencing data. Besides, the most abundant 30 miRs expressed in pEVs were uploaded into IPA to analyze their downstream target genes and main functions involved in. IPA integrated the available knowledge on genes, drugs, chemicals, protein families, processes, and pathways based on the interactions and functions derived from the Ingenuity Pathways Knowledge Database Literature, and understands the complex biological and chemical systems at the core of life science research based on lectures or predicated analysis (Dai et al., 2009).

LY294002 (10 μM), a highly selective inhibitor of phosphatidylinositol 3 (PI3) kinase, was preincubated with VSMCs for 1 h before adding pEVs. The same volume of DMSO was used as control.

The concentration of PIP₃ in VSMCs was determined by ELISA using a rat PIP₃ ELISA kit (Shanghai FanTai Biotechnology, China). The kit used bi-antibody sandwich method to determine the level of PIP₃. First of all, the sample added to the microwells of the coated PIP₃ monoclonal antibody, and then it is combined with HRP-labeled PIP₃ antibody to form an antibody-antigen-enzyme-labeled antibody complex. After 3 times washing by PBS, the substrate TMB and acid were added for chromogenic reaction. The absorbance (OD value) was measured with the microplate reader (Bio-Rad 680, Bio-Rad, United States) at a wavelength of 450 nm.

The 3′ untranslated regions (UTRs) of PTEN including the predicted miR-92a-3p binding sequences, and the mutation segment were all obtained by gene synthesis. The segments were inserted into the downstream of the luciferase reporter gene (psiCheck-2, Promega, United States), respectively. To determine the suppressing efficiency of miR-92a-3p, HEK-293T cells were transfected with the reporter plasmid or the mutated vectors together with miR-92a-3p mimic or NC. Twenty-four h later, firefly and renilla luciferase activities were measured consecutively using a dual luciferase reporter assay system (Promega, United States).

After carotid artery intimal injury surgery, the rats were randomly assigned to two groups: subcutaneous injections of 300 μl of anti-col8a1 siRNA (1 μM) or negative control (Pillé et al., 2005). Subcutaneous injections were repeated every 2 days for a total of 2 weeks.

Each experiment was performed at least in quadruplicate of biological replicates. Statistical analysis was performed and figures were prepared with the GraphPad Prism 6.0 (GraphPad Software, CA, United States). All values are expressed as the mean ± SD. The Gaussian distribution of values was analyzed by the Kolmogorov-Smirnov test. A paired t-test was used for paired data with a Gaussian distribution; A Wilcoxon matched-pairs signed rank test was used for paired data that lacked a Gaussian distribution or a sample size that was less than 5. In addition, the Friedman test was used for multiple comparisons with a single reference group whose sample size was less than 5. An unpaired t-test was used for unpaired data with a Gaussian distribution. Differences with values of P < 0.05 were regarded as statistically significant.

To explore the mechanical properties of the carotid artery after injury, we used ultrasound imaging to measure the deformation of the carotid artery following cardiac pulsation (Figure 1A) and a Piuma Nanoindenter to measure the stiffness of the carotid artery (Figure 1B) at 2 weeks after injury. Compared with that of the self-contralateral common carotid artery, the change in arterial diameter (%) (Safar et al., 1981) at 2 weeks after intimal injury surgery was significantly decreased (Figure 1A). Moreover, compared with that of the control, the stiffness of the carotid artery increased from an average of 13.45 kPa to an average of 20.98 kPa (Figure 1B).

To address the progression of vascular remodeling after intimal injury, elastin-van Gieson, hematoxylin-eosin (HE), and picrosirius red staining were used to examine the neointima, vascular morphology and collagen levels in the common carotid artery at 2 weeks after injury (Figure 1D). Compared with that of the self-contralateral common carotid artery, the neointima and VSMCs were significantly thickened and increased in the injured carotid artery (Figures 1C,D). Collagen accumulated in the vessel wall, especially in areas of neointimal hyperplasia (Figure 1D). These results suggested that in the intimal injury model, the stiffness of the injured vessel was increased at approximately 2 weeks. The main component in areas of hyperplasia was VSMCs, and there was a large amount of collagen in the neointima, which may affect the mechanical properties of the injured artery.

Immunofluorescence staining revealed that CD41-positive pEVs were closely adjacent to VSMCs in the injured carotid artery in vivo (Figure 2A). Then, Nanoparticle Tracking Analysis (NTA) was used to examine the concentration and size distribution of the circulating EVs after intimal injury surgery. The results showed that compared with the control group, the number and size of circulating EVs did not change significantly after intimal injury surgery (Supplementary Figure 1). This may because that the activation of platelets and the release of pEVs mainly occur in the injured area, and cannot significantly affect the number of pEVs in the entire blood.

In vitro, Electron microscopy imaging and NTA indicated that the size of most pEVs secreted by activated platelets was between 100 to 200 nm (Figures 2B,C and Supplementary Figure 2). Moreover, nanoparticle tracking analysis (NTA) indicated that the size a peak was at 164 nm. In addition, there was a small peak at 36 nm, indicating that there might be a small amount of exosomes in the extracted pEVs (Figures 2B,C).

To investigate the adhesion of pEVs to VSMCs in vitro, pEVs were labeled with the red fluorescent cell linker PKH26. The immunofluorescence staining results showed that PKH26-positive pEVs adhered to VSMCs after pEV stimulation for 1 h (Figure 2F). When pEVs and newly digested and suspended VSMCs were mixed and seeded on an uncoated glass plate, more VSMCs adhered and spread on the glass surface than in the control group (Figures 2D,E), indicating that pEVs promoted VSMC adhesion. In addition, increased bundles of long microtubules were present in VSMCs stimulated by pEVs (Figures 2G,H), which might be because VSMCs secrete large amounts of extracellular matrix (ECM), and VSMCs have improved focal adhesion on a smooth glass plate (Humphries, 2009). Subsequently, picrosirius red staining was used to analyze whether collagen secretion by VSMCs was changed after pEV stimulation. Compared with unstimulated VSMCs, VSMCs stimulated by pEVs secreted more collagen (Figure 2I). These results suggested that during vascular intimal injury, pEVs could interact with VSMCs and cause VSMCs to secrete increased ECM, thereby inducing vascular ECM remodeling. Then, we focused on the ECM, expecially the members of collagen to determine their roles in the process of endometrial injury.

To explore the ECM and ECM regulatory factors involved in neointimal hyperplasia in an intimal injury model, we used mRNA sequencing to analyze the changes in mRNA expression in the injured carotid artery at 2 weeks after surgery. Among the differentially expressed genes, 237 genes were identified as “extracellular space” by IPA (Supplementary Table 2). Subsequently, we used UniProt³ and GO “molecular function” analyses to distinguish these “extracellular space” molecules into the “extracellular matrix structural constituent” and “extracellular matrix organization and metalloendopeptidase activity” categories (Supplementary Table 3). Among them, the “extracellular matrix structural constituent” category contained 20 molecules, and the “extracellular matrix organization and metalloendopeptidase activity” category contained 43 molecules (Supplementary Table 3 and Figure 3A).

We selected the 10 molecules with the highest abundance and greater than 2-fold differential expression in the “extracellular matrix structural constituent” category for subsequent analysis. As shown in Figure 3B, real-time RT-PCR analysis validated that col8a1, col12a1, col8a2, elastin, col6a3, acan and tenascin expression levels were significantly upregulated at 2 weeks after intimal injury surgery compared with those of the self-contralateral common carotid artery. In vitro, after pEV stimulation for 12 h, qPCR analysis showed that col8a1, elastin, col16a1, col8a2 and col12a1 expression levels were significantly upregulated compared with those of the control (Figure 3C). The expression levels of col8a1, col8a2, col12a1, and elastin changed in response to both stimuli. Among these molecules, the Col VIII family has been proven to play an important role in changes in the stiffness of atherosclerotic plaques (Merei et al., 2017). Therefore, we performed immunofluorescence and western blotting analyses of Col8a1 and Col8a2.

The immunofluorescence staining results showed that in the neointima after intimal injury in vivo and in VSMCs stimulated with pEVs in vitro, the fluorescence intensities of Col8a1 and Col8a2 increased significantly (Figures 3D,E and Supplementary Figure 3). Col8a1 was mostly located in the cytoplasm, as shown immunofluorescence staining, and the qPCR and western blotting results revealed that col8a1 mRNA and protein expression also significantly increased (Figures 3F,G). However, Col8a2 was mostly located in the nucleus (Supplementary Figure 3B), which was different from the traditional understanding of VSMCs (Stephan et al., 2004; Adiguzel et al., 2006). These results suggest that pEV-induced VSMCs produce and secrete Col8a1, which may contribute to ECM remodeling during vascular intimal injury.

To investigate the molecules transported by pEVs that are involved in regulating the production of Col8a1, IPA software was used to analyze related molecules that may participate in the regulation of “formation of extracellular matrix” (Figure 4A). The results revealed that 36 molecules correlated with the formation of extracellular matrix (Figure 4A and Supplementary Table 4). Among them, we focused on Akt signaling, which is the core signaling pathway that participates in this network (Figure 4B). Western blotting was used to measure the expression of phosphorylated Akt in VSMCs after pEV stimulation, and we found that Akt was significantly activated (Figure 4C). In order to detect whether the Akt carried by pEVs contribute to the Akt changes in VSMCs, which subsequently increased the phosphorylation of Akt, the expression of total Akt in VSMCs after pEVs stimulation was analyzed (Supplementary Figure 4). The results showed that the expression of total Akt was similar between the control group and pEVs stimulation group, which indicated that the increased pAkt may be activated by pEVs but not via the direct delivery.

Since multiple molecules, such as PIP₃, PDK1 and mTORC2, have been reported to be involved in the activation of Akt (Xu et al., 2015), western blotting and ELISA were further performed (Figures 4D–F). Among these three molecules, only the expression of PIP₃ increased significantly in VSMCs stimulated with pEVs. The stimulation of pEVs was complicated, and using IPA software the potential interleaving of multiple signal networks which may lead to the activation of Akt and inactivation of PDK1 and mTOR were analyzed (Supplementary Figure 5). Therefore, Akt activation may be due to the activation of upstream PIP₃ induced by pEVs.

To further verify the effect of PIP₃/Akt signaling on the production of Col8a1 by VSMCs, cells were pretreated with LY294002, a specific inhibitor of PIP₃/Akt signaling, for 1 h and then treated with pEVs. LY294002 significantly abrogated the expression level of p-Akt induced by pEVs (Figure 4G). After treatment with pEVs for 12 h, LY294002 also had the same effect on the repression of col8a1 mRNA (Figure 4I). Immunofluorescence staining and western blotting showed that the addition of LY294002 blocked the protein expression of Col8a1 in VSMCs stimulated with pEVs (Figures 4H,J). These results suggested that PIP₃/Akt signaling participated in the production of Col8a1 by VSMCs stimulated with pEVs.

Recent studies have shown that pEVs can deliver a variety of miRs (Laffont et al., 2013). MiR sequencing was used to identify the top 30 miRs expressed in pEVs secreted from activated platelets (Figure 5A). The functions of the downstream target molecules (82 in total) of these 30 miRs were analyzed by IPA software, and we found that the downstream molecules were mainly related to “cellular movement” and “cardiovascular system development” (Figure 5B and Supplementary Table 5). A total of 41 of the 82 molecules formed a core network structure, of which 7 key molecules were involved in the Akt signaling pathway (Figure 5C and Supplementary Table 6).

Among the 30 miRs, 19 core seed sequences could bind to the 3’UTRs of different target gene mRNAs (Supplementary Table 7). Among the 30 miRs, 14 miRs targeted the same molecule, PTEN, which was the most common of all 7 targeted molecules associated with PI3K/Akt signaling (Figure 5D and Supplementary Table 8). Therefore, miR-92a-3p, which targeted PTEN and was the most highly expressed in pEVs, was selected for follow-up studies.

After pEV stimulation for 12 h in vitro, the qPCR results verified that miR-92a-3p expression in VSMCs was significantly upregulated in comparison with that of the control (Figure 5E). To detect whether the increased miR- 92a-3p was produced by VSMCs or was delivered by pEVs to VSMCs, the precursor of miR-92a-3p (pre-miR-92a-3p) in VSMCs was detected. If the miR-92a-3p was produced by VSMCs, pre-miR-92a-3p would increase in VSMCs (Supplementary Figure 6). The results showed that the expression of pre-miR-92a-3p were similar between the control group and pEVs stimulation group, which indicated that the increased miR-92a-3p was transferred by pEVs. Moreover, western blotting showed that PTEN expression in VSMCs was significantly downregulated after pEV stimulation for 24 h (Supplementary Figure 7). In vivo, qPCR and FISH results verified that miR-92a-3p expression levels were significantly upregulated in VSMCs 2 weeks after intimal injury surgery in comparison with those of the self-contralateral carotid artery (Figures 5F,G). There was a predicted binding site for miR-92a-3p at the PTEN 3’UTR site 2808-2815 (Figure 5H and Supplementary Table 9), indicating that miR-92a-3p may negatively regulate PTEN protein expression.

Subsequently, miR-92a-3p was overexpressed or knocked down in VSMCs with a specific mimic or inhibitor, respectively (Figure 5I). Western blotting indicated that the miR-92a-3p mimic significantly reduced the protein expression of PTEN, whereas the inhibitor increased PTEN levels compared with those of the respective NC (Figure 5J). These results indicate that miR-92a-3p plays an important role in the regulation of PTEN expression.

Dual luciferase reporter gene analysis was then used to assess the binding and inhibitory capacity of miR-92a-3p for the target sites of the PTEN 3′UTR. Compared with the NC, cotransfection of the miR-92a-3p mimic with the wild-type 3′UTR of PTEN significantly decreased the luciferase activity in HEK-293T cells (Figure 5K). In contrast, there was no significant change when miR-92a-3p was cotransfected with the mutant PTEN 3’UTR compared to the NC (Figures 5H,K).

Then, we evaluated the impact of miR-92a-3p changes on Col8a1. The qPCR and western blot results indicated that the miR-92a-3p mimic significantly increased the mRNA and protein expression of col8a1 (Figure 5L and Supplementary Figure 8). These results indicated that miR-92a-3p negatively regulated the expression of PTEN and subsequently promoted the production of Col8a1 in VSMCs.

To further elucidate the effects of Col8a1 on vascular stiffness in the intimal injury model, col8a1 small interfering RNA (siRNA) was injected locally to knockdown Col8a1 expression (Supplementary Figure 9). Three pairs of specific siRNAs were designed, and the most efficient siRNA, si-#2, was identified (Figure 6A and Supplementary Table 1). In intimal injury, the siRNA significantly reduced the mRNA and protein expression of col8a1 compared with that of the NC (Figures 6B,C). Vascular morphological analysis revealed that the vessel wall was thickened, but the collagen density was reduced after col8a1 siRNA injection (Figures 6D,E).

Regarding the mechanical properties of the carotid artery, compared with that of NC injection after intimal injury, siRNA injection significantly decreased the change in arterial diameter (%) at 2 weeks after intimal injury surgery from an average of 13.93% to an average of 4.82% (Figure 6F). Moreover, the stiffness of the carotid artery decreased from an average of 20.31 kPa to an average of 16.28 kPa compared with that of the NC group (Figure 6G). These results suggested that Col8a1 knockdown could alleviate the increase in vascular stiffness in the intimal injury model.