Silencing LncRNA AK139328 protects vascular endothelial cells from ischemia-reperfusion injury by increasing PI3K/Akt signaling

We investigated the effects of lncRNA AK139328 and PI3K/Akt signaling during rat hindlimb ischemia-reperfusion (I/R) injury. Rat vascular endothelial cells (VECs) subjected to oxygen-glucose deprivation showed high lncRNA AK139328 expression and decreased PI3K/Akt signaling, whereas lncRNA AK139328 knockdown increased PI3K/Akt signaling in VECs. Correspondingly, gastrocnemius muscles from rats subjected to hindlimb I/R showed high levels of lncRNA AK139328 and low PI3K/Akt/eNOS signaling. I/R also increased serum levels of TNF-α, VCAM-1 and malondialdehyde, serum creatine kinase activity and the number of circulating endothelial cells (CECs). Gastrocnemius tissue from rats subjected to I/R exhibited inflammation, apoptosis and trauma with loosely attached, swollen VECs associated with inflammatory immune cells. LncRNA AK139328 knockdown increased PI3K/ Akt/eNOS signaling in gastrocnemius muscles subjected to I/R, and decreased serum levels of TNF-α, VCAM-1 and malondialdehyde, serum creatine kinase activity and numbers of CECs. LncRNA AK139328 knockdown also decreased inflammation, apoptosis and trauma in gastrocnemius muscles, which showed tightly attached VECs that were not associated with inflammatory immune cells. Inhibition of PI3K/ Akt signaling by wortmannin or LY294002 reversed the effects of lncRNA AK139328 knockdown. In conclusion, I/R induces expression of lncRNA AK139328, which suppresses PI3K/Akt signaling required to prevent I/R-related pathology in VECs.


INTRODUCTION
Ischemia-reperfusion (I/R) results in remote organ injury during treatment of patients with acute limb ischemia [1,2]. The extent of I/R injury in myocardial infarction and peripheral vascular disease is determined by the magnitude of interruption in blood supply [3]. Remote organ dysfunction develops after reperfusion and is regulated by local humoral and/or cellular factors [4]. I/R injury, which is induced by clamping the femoral artery shows slow recovery despite rapid restoration of blood flow as well as successful surgery [5,6]. Reperfusion results in oxidative stress that elicits inflammatory response and induces kidney cell death [7]. Vascular endothelial cells (VECs) secrete various factors that modulate vascular permeability barrier and inflammation [8]. They produce periostin and inflammatory chemokines associated with allergic reactions [9]. Autophagy protects VECs during I/R injury [10]. Non-coding RNAs (ncRNAs) regulate growth and proliferation of human umbilical vein endothelial cells [11]. LncRNAs regulate embryonic stem cell maintenance and lineage-specific differentiation [12]. They play critical roles in mammalian neuronal development and activity [13]. LncRNAs influence diverse cellular functions like cell cycle, apoptosis, differentiation and morphogenesis [14]. LncRNA CGS21680 decreases hepatocyte death by activating the PKB/Akt signaling pathway [15]. PI3K/Akt signaling pathway

Research Paper
Oncotarget s181 www.oncotarget.com regulates tumor necrosis factor-alpha (TNF-α) expression [16]. Silencing lncRNA AK139328 ameliorated liver I/R injury by activating Akt signaling [17]. In the mouse liver I/R injury model, we observed multiple dysregulated lncRNAs including lncRNA AK139328 in the plasma with potential as novel diagnostic and prognostic biomarkers for ischemic liver injury [18]. In this study, we investigated the role of lncRNA AK139328 and PI3K/Akt signaling in rat hindlimb I/R injury model.

Identification and characterization of rat lncRNA AK139328
We searched the rat genome database with BLAST using the mouse lncRNA AK139328 as the query sequence and found the evolutionarily conserved rat lncRNA AK139328 (Figure 1). Analysis of RNA-seq data in the SRA and ArrayExpress databases with TopHat and Cufflinks softwares localized the lncRNA AK139328 to rat chromosome 3 on the rat genome (Version: RGSC Rnor_6.0/rn6) of UCSC genome database (http://genome. ucsc.edu/) ( Figure 2). We performed 3′-and 5′ RACE to obtain the full length lncRNA AK139328 transcript (Figures 3 and 4). In vitro transcription & translation assay showed that lncRNA AK139328 sequence was noncoding ( Figure 5). RNA FISH analysis demonstrated high lncRNA AK139328 expression in the nucleus of rat VECs and slight expression in the cytoplasm ( Figure 6).

LncRNA AK139328 is highly expressed in oxygen-glucose deprived rat VECs
We investigated the status of lncRNA AK139328 in rat VEC's induced by oxygen-glucose deprivation to mimic I/R. We observed that oxygen-glucose deprivation increased lncRNA AK139328 levels in rat VECs (P < 0.05; Figure 7). Moreover, VECs transfected with si-AK139328 showed low lncRNA AK139328 levels than control VECs after oxygen-glucose deprivation (P < 0.05; Figure 7). Status of PI3K/Akt signaling pathway in VEC model of I/R injury Next, we analyzed the status of PI3K/Akt signaling pathway and ROS levels in the VEC cells upon oxygenglucose deprivation. PI3K/Akt signaling (total and phosphorylated PI3K and AKT) and ROS levels decreased when rat VECs were oxygen-glucose deprived ( Figure 8 and Table 1). Moreover, rat VECs transfected with si-AK139328 showed higher PI3K/Akt signaling and ROS levels than controls (P < 0.05; Figure 8, Table 1). But, PI3K/AKT inhibitor LY294002 reversed the effects of lncRNA AK139328 silencing. The changes in PI3K and Akt mRNA and protein levels are shown in Figure 8A-8D.

Status of lncRNA AK139328 expression in rat hindlimb I/R model
Next, we analyzed the in vivo role of lncRNAAK139328 using the rat hindlimb I/R model. We observed increased lncRNA AK139328 expression in the gastrocnemius muscle of rats subjected to I/R (P < 0.05; Figure 9). However, lncRNA AK139328 knockdown reduced its levels by 60% in the gastrocnemius muscle of rats subjected to I/R (P < 0.05; Figure 9).

Status of PI3K/Akt signaling pathway, eNOS, ROS and caspase-3 in rat hindlimb I/R model
Next, we analyzed levels of ROS as well as total and phosphorylated PI3K, Akt, eNOS protein levels in the rat hindlimb I/R model. Gastrocnemius muscle from rats subjected to I/R showed reduced total and phosphorylated PI3K, Akt, eNOS as well as ROS levels (P < 0.05; Figure 10; Table 2). However, rats subjected to I/R in presence of si-AK139328 demonstrated higher total and phosphorylated PI3K, Akt, eNOS as well as ROS levels than controls (P < 0.05; Figure 10; Table 2). The effect of lncRNA AK139328 knockdown was reversed by wortmannin, a PI3K/Akt inhibitor. This demonstrated that I/R induced lncRNA AK139328 that subsequently downregulated PI3K signaling pathway, eNOS and ROS. Moreover, IHC analysis demonstrated that I/R induced caspase-3 expression in rat gastrocnemius muscles, but were decreased by lncRNA AK139328 silencing (P < 0.05; Figure 11). Wortmannin treatment reversed the effects of lncRNA AK139328 silencing and increased caspase-3 expression ( Figure 11).
Effect of lncRNA AK139328 on morphological and ultrastructural changes in skeletal muscle in rat hindlimb I/R model H&E stained skeletal muscle sections showed swollen, uneven and irregularly arranged muscle fibers upon I/R injury ( Figure 12). Muscle cells showed severe swelling, fracture, enlargement, lighter color, narrow intercellular space and loosely arranged vascular wall layers. The VECs were also swollen and loosely attached to the internal elastic membrane with the nucleus protruding to the lumen. In some cases, VECs were either found in the lumen or disappeared. Inflammatory Oncotarget s183 www.oncotarget.com cells were also observed attached to the VECs ( Figure  12). The negative controls showed uniformly arranged muscle fibers with normal cell morphology, uniform size, intercellular space and vascular endothelial structure in the vascular wall. The VECs were flat, normal-sized, regularly arranged and closely attached to the inner lumen of internal elastic membrane with the nucleus protruding into the lumen. VECs did not show any swelling and were not associated with any inflammatory cells. In the I/R + si-AK139328 group, muscle fibers were slightly swollen and unevenly sized but regularly arranged. No dissolution and fracture was found in the fibers. Some muscle cells showed slight swelling, intercellular space and fractures. VECs were regularly arranged although some were swollen. They were closely attached to the internal elastic membrane, with few VECs found in the lumen and few attached to inflammatory cells. In the I/R + Wortmannin group, the fiber swelling was more severe than the I/R group. The fibers were irregularly arranged, twisted, and fractured. Muscle tissues showed severe swelling, narrow   Oncotarget s185 www.oncotarget.com intercellular space and loosely arranged vascular wall layers. VECs were swollen and loosely attached to the internal elastic membrane with nucleus protruding into the lumen and associated with many inflammatory cells. This demonstrated that PI3K/Akt signaling pathway was necessary for protecting the integrity and organization of the muscle fibers and VECs during I/R ( Figure 12). Electron micrographs showed large lipid vacuoles in the I/R group, but fewer lipid vacuoles in the I/R + si-AK139328 group ( Figure 13).

Role of lncRNA AK139328 in the pathology in rat hindlimb I/R model
Next, we analyzed the various pathological indicators in the rat hindlimb I/R model. Pro-inflammatory cytokines, TNF-α and VCAM-1 levels were higher in the I/R groups, but reduced upon lncRNA AK139328 silencing (P < 0.05; Table 3). This demonstrated that I/R induced inflammation was PI3K/Akt dependent. Serum MDA levels and CK activity were higher in gastrocnemius muscle from I/R rats, but were lowered by silencing lncRNA AK139328 (P < 0.05; Table 4). Moreover, there were high numbers of serum CECs in I/R rats, but reduced by silencing lncRNA AK139328 (P < 0.05; Table 5). TUNEL assay showed increased apoptosis in gastrocnemius muscle upon I/R, but decreased upon lncRNA AK139328 silencing (P < 0.05; Figure 14). However, inhibition of PI3K/AKT pathway by wortmannin reversed the effects of lncRNA AK139328 silencing. This suggested that activation of the PI3K/AKT signaling pathway was required to repair from I/R injury.

DISCUSSION
In this study, we demonstrated that lncRNA AK139328 negatively regulated the PI3K/Akt signaling pathway in I/R injury. LncRNAs are a class of noncoding RNAs, more than 200 nucleotides in length that regulate gene expression at transcriptional and posttranscriptional levels as well as chromatin modification and genomic imprinting [19]. LncRNA CRNDE promotes the proliferation, migration and invasiveness of hepatocellular carcinoma (HCC) cells [20]. LncRNA EWSAT1 promotes human nasopharyngeal carcinoma cell growth [21]. Thus, lncRNAs play pivotal roles in regulating a wide range of biological processes and are critical in various diseases [22].
In a previous study, LY294002, a PI3K inhibitor, repressed the expression of thrombin-activating fibrinolysis inhibitor (TAFI) in human hepatoma HepG2 Cells by decreasing the stability of CPB2 transcripts [23]. Akt is activated as a protective mechanism against Notes: * , compared with the control group, P < 0.05; # , compared with the model group, P < 0.05; the experiment was repeated for four times and differences between groups were analyzed by t-test. Notes: * , compared with the sham group, P < 0.05; # , compared with the I/R group, P < 0.05; 64 rats were included in the experiment (8 in each group) and differences between groups were analyzed by t-test; Wom, wortmannin; I/R, ischemiareperfusion.
I/R injury causes tissue damage, complement activation, cytochrome release, cytokine activation, inflammation, edema, neutrophil platelet adhesion, capillary plugging, and thrombosis [34]. Spinal cord ischemia triggered local edema, vasospasm, free radicals, calcium influx into the axons, and loss of potassium from the intercellular space [35]. In our study, silencing lncRNA AK139328 inhibited edema, dissolution, muscle fiber damage (fracture and swelling) and association of immune cells to VECs in response to I/R injury. We also observed decreased MDA levels and CK activity in the plasma.  (5) sham + si-AK139328 (6) I/R + si-AK139328 (7) I/R + si-AK139328 + Wom and (8) I/R + Wom groups (n = 8 per group). β-actin was used as internal control. Note: * denotes P < 0.05 compared with sham group; # denotes P < 0.05 compared with I/R group; the experiment was analyzed by t-test; Wom = wortmannin; si-AK139328 NC = negative control siRNA; si-AK139328 = siRNA against lncRNA AK139328. Moreover, silencing lncRNA AK139328 decreased the presence of CECs in the serum and apoptosis in the gastrocnemius muscle cells by inhibiting the PI3K/Akt signaling pathway. The number of serum CECs indicates the severity of acute respiratory distress syndrome [36]. In cancer patients, CECs are shed from vessel walls and enter circulation as a result of endothelial dysfunction and correlates with tumor progression [37]. Treatment with Notes: * , compared with the sham group, P < 0.05; # , compared with the I/R group, P < 0.05; 64 rats were included in the experiment (8 in each group) and differences between groups were analyzed by t-test; Wom, wortmannin; I/R, ischemiareperfusion; MDA, Malondialdehyde; CK, creatine kinase. Oncotarget s190 www.oncotarget.com antioxidant N-acetyl cysteine reduced I/R injury in the femoral artery endothelium and gastrocnemius muscle by preventing tissue damage and gastrocnemius muscle apoptosis [38]. We demonstrated that silencing lncRNA AK139328 protected VECs, decreased the serum CECs and apoptotic gastrocnemius muscle cells via PI3K/Akt signaling pathway.
In conclusion, our findings suggest that silencing lncRNA AK139328 enhances survival and limits pathological damage to VECs due to I/R injury by activating the PI3K/Akt signaling pathway. Hence, lncRNA AK139328 is a potential target in I/R therapy. may be a potential biomarker for treatment of VECs injury caused by I/R. But more underlying molecular mechanism studies are in need to verify the result and explore the further potentials.

Cloning and analysis of rat lncRNA AK139328
We searched the rat genome database in GenBank by BLAST program [39] (taxid: 10116) (http://blast.ncbi. nlm.nih.gov/) using the full-length nucleotide sequence of mouse lncRNA AK139328 (full length of mouse AK139328 sequence is available at http://www.ncbi.nlm.nih.gov/ nuccore/AK139328). We determined homology between the nucleotide sequences of mice and rat lncRNA AK139328. Notes: * , compared with the sham group, P < 0.05; # , compared with the I/R group, P < 0.05; 64 rats were included in the experiment (8 in each group) and differences between groups were analyzed by t-test; Wom, wortmannin; I/R, ischemiareperfusion; CECs, circulating endothelial cells. To clone full length lncRNA AK139328, we isolated total RNA from rat VECs by the method of Trizol. Then, we used the reverse transcription kit to generate cDNA and performed 3′RACE by nested-PCR (3′RACE Inner Primer: GTGGCCCCTTGGTCTTGCAAAC, 3′RACE Outer Primer: GTGTCTGCAACCCGATAGGT). First, PCR was performed with the lateral primers. The PCR products were diluted 100 times and then the second round of PCR was performed. After the second round of PCR, the PCR products were electrophoresed in a 1% agarose gel, purified, cloned and sequenced.
For the 5′ RACE, RNA was incubated with calf intestinal phosphatase (CIP) at 37°C for 1 h. After terminating the CIP reaction, the RNA was extracted with phenol-chloroform. Then, the RNA was incubated with tobacco acid pyrophosphatase (TAP) at 37°C for 1 h followed by incubation with the 5′ RACE adapter at 37°C for 1 h. Then, we performed reverse transcription using the 3′-RACE PCR product as a template and subjected the cDNA product to 5′RACE by nested PCR as described for 3′RACE (5′RACE Inner Primer: AGTTTGCAAGACCAAGGGGCCAC, 5′RACE Outer Primer: ACTATTATGTCAATATATAA).

In vitro transcription & translation
We performed in vitro transcription and translation of the full-length rat lncRNA AK139328 cloned into pcDNA3.1 plasmid with the T7 promoter using the TnT quick transcription translation kit (Promega, Madison, WI, USA). Then, 1 μl of the reaction product was boiled at 100°C for 2 min with 15 μL SDS (sodium dodecyl sulfate) loading buffer and subjected to 4% ~ 12% NuPAGE SDSpolyacrylamide gel electrophoresis (Life Technologies Corporation, Carlsbad, CA, USA). Streptomycin affinity antibody was incubated with proteins labeled with biotinylated Transcend tRNA. Subsequently, the gel was placed in G250-10% acetic acid and -45% methanol, and after 4 h of greenhouse effect, it was washed twice and then photographed.RNA-FISH analysis in rat VECs We generated a specific FISH fluorescent DyLight 488-labelled RNA probe (Promega, Madison, WI, USA) for the rat lncRNA AK139328. The rat VECs were grown on treated slides for 12 ~ 16 h, washed twice with by 1 × PBST (PBS + 0.1% Tween-20) followed by incubation at room temperature with 4% polyoxymethylene for 15 minutes. The cells were subsequently treated with 4 μg/ ml proteinase K for 5 min, 200 μl glycine in 1 × PBST for 5 minutes, and 200 μl of acetylation reagent for 10 minutes. Then, the slides were incubated with 200 μl prehybridization solution at 65°C for 1 h followed by 300 ng/ml probe in 200 μl hybridization solution overnight at 65°C in a culture plate covered with paraffin wax. The hybridization solution was removed, and the slides were incubated serially with solutions 1, 2 and 3, respectively at 65°C for 10 mins. Then, they were incubated with 300 μl 0.5% bovine serum albumin (BSA) at room temperature for 1 h. Then, the slides were incubated with antibody diluted in 0.5% BSA at room temperature for 4 ~ 5 h, followed by DAPI (4′,6-diamidino-2-phenylindole) staining at room temperature for 5 min. The slides were sealed with antifade reagent and kept at 4°C before being photographed under an inverted fluorescent microscope (LWD300-38LFT, Shanghai Cewei Electrooptical Oncotarget s192 www.oncotarget.com Technology Co., Ltd., Shanghai, China) at 488 nm (DyLight 488) and at < 460 nm.

Experimental animals and grouping
This study was approved by the Animal Care Committee of College of Animal Sciences, Zhejiang University. Adult male Wistar rats (weight: 200 ~ 250 g) were obtained from the Institute of Animal Husbandry and Veterinary, Shanxi Academy of Agricultural Sciences (Taiyuan, China). Sixty-four rats were randomly assigned into 8 groups: WT (no treatment), I/R (no treatment after I/R), sham (injected with normal saline), I/R + si-AK139328 negative control (NC) (injected with si-AK139328 NC after I/R), sham + si-AK139328 (injected with siAK139328 and normal saline), I/R + si-AK139328 (injected with siAK139328 after I/R), I/R + si-AK139328 + Wom (injected with siAK139328 and wortmannin after I/R) and I/R + Wom (injected with wortmannin after I/R).

Rat I/R model
Male rats were anesthetized by abdominal injection of 40 mg/100 g urethane followed by midline laparotomy and systemic heparinization. Abdominal aorta was blocked 0.5 cm above the bifurcation of the common iliac artery with atraumatic artery clip such that the arterial pulse disappeared and lumen became flat below the clip to induce ischemia for 2 h. The clamp was removed to allow reperfusion for another 2 h. In the sham group, 1 ml normal saline was injected into the sublingual vein instead of blocking the abdominal aorta. For lncRNA AK139328 knockdown, 100 μl of 2.5 mg/kg AK139328 siRNAs were injected into the caudal vein of rats with hindlimb I/R at 9:00 a.m. on the same day as I/R and 24 h later. The sequences of the 3 siRNAs used were: sense 1, 5′ -CAGCUAUCACAUGCCAGCAUC AUAU-3′, antisense 1, 5′-AUAUGAUGCUGGCAUGUGAU AGCUG -3′; sense 2, 5′-GCAUCUAAAGCUGGUGGCAAU ACUA -3′, antisense 2, 5′-UAGUAUUGCCACCAGCUUUAG AUGC -3′; sense 3, 5′-CCUUGGACUUCUGACUGAAUGA ACU -3′, antisense 3, 5′-AGUUCAUUCAGUCAGAAGU CCAAGG -3′ (Invitrogen Inc., Carlsbad, CA, USA) [17]. For si-AK139328+Wom and I/R+Wom groups, 1 ml wortmannin (1:400 dilution; Beijing ZhongshanJinqiao Biotechnology Co., Ltd., Beijing, China) was injected into the sublingual vein after I/R.  Oncotarget s193 www.oncotarget.com small pieces. The homogenate was sieved through 100and 200-µm mesh screens. Microvascular tissues on the screen were centrifuged in DMEM and digested with 5 ml of 0.2% collagenase I for 20 min. The homogenates were centrifuged (12000 rpm/min, at 4°C) again and the cells from the pellet were cultured in 6 ml of complete DMEM medium supplemented with endothelial cell growth factor (ECGF) and 30% fetal bovine serum (FBS) for 48 h. The cells were trypsinized when they reached 80% confluence.

In vitro vascular endothelial cell I/R model
For oxygen-glucose deprivation experiments to simulate I/R, endothelial cells were washed in PBS twice and incubated in a closed anoxic tank with sugar-free Krebs solution at 37°C with 95% N 2 + 5% CO 2 for 4 h. Subsequently, the cells were cultured in serum-free DMEM medium for 12 h. These cells were included in the model group. Cells in the control group were cultured in a normal incubator for 16 h. Altogether, cells were divided into 6 groups: control, model, model + si AK139328 NC (transfected with si-AK139328 NC), model + siAK139328 (transfected with siAK139328), model + si-AK139328 + LY294002 [transfected with si-AK139328 and treated with 50 μM LY294002 (specific inhibitor of PI3K/Akt signaling pathway, which can permeate cells, specifically inhibit PI3K and PI3K/Akt signaling pathway) (Cell Signaling Technologies, Beverly, MA, USA)], and model + LY294002 (treated with 50 μM LY294002) groups. All transfections and treatments were performed 12 h after oxygen-glucse deprivation. LY294002 is a specific inhibitor of PI3K/Akt signaling pathway which was obtained from Cell Signaling Technologies, Beverly, MA, USA.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
Total RNA was extracted from gastrocnemius muscle tissues with Trizol (Life Technologies, USA) according to manufacturer's instructions. All primer sequences ( Table 6)

Western blotting
Gastrocnemius muscle tissues were quickly dissected from mice, frozen in liquid nitrogen and stored at -80°C. Protein lysates were prepared by incubating 100 µg muscle tissue with 500 μl lysis solution for 5 min. The homogenate was centrifuged at 12000 rpm for 10 min at 4°C and the protein concentration was determined with bicinchoninic acid (BCA) method.
The protein lysates (20 μg) were boiled at 100°C for 10 min in 5 × loading buffer and separated on a 10% SDS PAGE and subsequently transferred to polyvinylidene fluoride (PVDF). The blots were blocked with 5% skimmed milk for 1 h and incubated with primary rabbit anti-rat PI3K, p-PI3K, Akt and p-Akt (1:1000; Cell Signaling Technology, USA) antibodies at 4°C overnight. Then, after washing the membranes thrice with PBST for 10 minutes, they were incubated with HRP-conjugated goat anti-rabbit secondary antibody (1:4000) at 37°C for 1 h. Then, the blots were developed with ECL (Amersham, Arlington Heights, IL, USA) and then developed. The protein bands were quantified by the Labwork 4.6 software in UVP image processing system using GAPDH as internal reference.

Analysis of reactive oxygen species
After reperfusion, the central coronal sections from rat gastrocnemius muscles were paraffinembedded and cut into 5µm slides. Then, the slices were incubated with 2 µM dihydroethidium (DHE) in the dark for 15 minutes. The slides were observed under the fluorescent microscope. Rat VECs from all groups were incubated with 2 µM dihydroethidium (DHE) in the dark for 30 minutes. Then, the fluorescent intensity was measured in a microplate reader at 488 nm excitation wavelength.

Hemotoxylin and eosin (H&E) staining
Rat skeletal muscle tissue samples from the left lower limb (1.0 cm 3 ) and left external iliac artery (0.5 cm) were fixed in 10% neutral formalin, paraffin embedded and sliced. Tissue sections were stained with the H&E kit (Beijing Solarbio Science & Technology Co., Ltd, Beijing, China) according to manufacturer's instructions and imaged under a light microscope (IX53, Olympus Optical Co., Ltd, Tokyo, Japan).

Transmission electron microscopy (TEM)
Lower extremity artery of rats was harvested in sterile PBS. After removing endovascular blood and fat from the outer membrane of the aorta, blood vessels were opened longitudinally and placed in a 60 mm with DMEM containing 75 μg/ml endothelial cell growth supplement (ECGS). Then, 1 mm 2 pieces were cut with a surgical knife, placed in a culture plate with the inner membranes facing down and incubated at 37°C and 5% CO 2 for 20 min. Then, they were incubated in 1.2 mL DMEM medium with 75 μg/ ml ECGS for 3 days. The vessel pieces were then removed carefully so that the few cells growing around them were not disturbed and fresh medium was added every 2 ~ 3 days. Endothelial cells were confirmed by immunofluorescence for von Willebrand factor (vWF) antigen. The endothelial cells were then fixed in 2.5% glutaraldehyde for 4 h, followed by 1% crane acid for 2 h and 70% ethanol at 4 o C overnight. Then, the cells were embedded with Epon812 (ethoxyline resin), sliced with a LKB ultramicrotome and stained with uranium acetate and citric acid. Images were captured using the transmission electron microscope (IX53, Olympus Optical Co., Ltd, Tokyo, Japan).

ELISA
Blood (4 ml) from the abdominal aorta was centrifuged at 1000 g for 5 min and the supernatant (plasma) was preserved at -80°C. Plasma levels of TNF-α and VCAM-1 were measured according to manufacturer's instructions for the corresponding ELISA kits (Shanghai Westang Bio-tech Co., Ltd, Shanghai, China). Oncotarget s195 www.oncotarget.com

Immunohistochemical staining
Blood vessels were fixed in 4% paraformaldehyde and dehydrated with serial concentrations of ethanol. After paraffin embedding, the vessels were sectioned. Then, the sections were deparaffinized, hydrated and incubated with 1% hydrogen peroxide (H 2 O 2 ) solution for 10 min at room temperature to eliminate endogenous peroxidase activity. Then, the samples were blocked with goat serum and incubated with anti-caspase-3 antibodies (Beijing ZhongShan GoldenBridge Biological Technology Co., Ltd., Beijing, China) at 4°C overnight. The samples were further incubated with anti-mouse digoxigenin followed by biotinylated anti-mouse IgG at 37°C for 30 min. Then, streptavidin was added and the slides were developed with diaminobenzidine (DAB). Sections were then counterstained with hematoxylin and images captured with a light microscope.

Measurement of caspase-3 activity
Caspase-3 activity was measured with the assay kit (C1115, Beyotime Biotechnology Co., Shanghai, China) according to manufacturer's instructions. Tissues were homogenized on ice, centrifuged at 15,000 g for 10 mins and the supernatants were used for analysis. The reaction included incubation of tissue supernatants with Ac-DEVD-AMC (tetrapeptide fluorescent substrate of caspase-3) at 37°C for 1 h. The samples were read in the fluorescence spectrophotometer (POLARstar, Ortenberg, Germany) at 380 nm excitation and 405 nm emission. Quantification was performed using the standard curve. The experiment was repeated thrice.

Estimation of serum malondialdehyde (MDA) and creatine kinase (CK) activity
Serum MDA levels and creatine kinase activities were measured by the Lipid Peroxidation MDA Assay Kit (Beyotime Biotechnology Co., Shanghai, China) and Creatine Kinase Activity Assay Kit (Sigma, St Louis MO, USA), respectively, according to manufacturer's instructions.

Measurement of circulating endothelial cells (CECs)
Venous blood (1.6 ml) was mixed with 0.4 ml anticoagulant (3.8% sodium citrate) and centrifuged for 20 min at 2350 rpm. Then, 0.2 ml of 0.1% adenosine 5′-diphosphate (ADP) was added to the supernatant and centrifuged similarly. Further, 0.1 ml of re-centrifuged supernatant was mixed with a small amount of PBS for 10 min and added onto a cover glass slide. The average numbers of CECs were determined by counting the numbers of CECs in 0.9 μl under an ultramicroscope. Experiments were repeated thrice.

In situ detection of cell apoptosis in gastrocnemius muscle
Lower part of rat gastrocnemius muscle was fixed in 4% paraformaldehyde for 4 ~ 8 h, paraffin embedded and cut into 50 μm slices. Cellular apoptosis was determined by terminal deoxynucleotidytransferase mediated d-UTP nick end labeling (TUNEL) assay kit (Wuhan Boster Biological Technology Ltd., Wuhan, China). The slices were photographed with Nikon E600 and the apoptosis index (AI; ratio of apoptotic to total nuclei) was analyzed by the LabWorks 4.0 software.

Statistical analysis
All data were analyzed by SPSS 22.0 software (IBM Corp., Armonk, NY, USA) and presented as mean ± standard deviation. Differences between groups were analyzed by t-test. P < 0.05 was considered as statistically significant.