Highly glycosylated CD147 promotes hemorrhagic transformation after rt-PA treatment in diabetes: a novel therapeutic target?

This study was carried out following the Guide for the National Science Council of the People’s Republic of China. Eight-week-old male Wistar rats with initial body weight of 200–220 g were randomized into two groups: diabetes and non-diabetes. The non-diabetes group was injected with saline, while the diabetes group received a standard intraperitoneal injection of streptozotocin (60 mg/kg; Sigma, St. Louis, MO, USA) to induce diabetes. Seven days after streptozotocin administration, rats with a blood glucose concentration of > 15 mmol/L were retained as successfully fitting the model as previously described [19].

Both groups were subjected to middle cerebral artery occlusion (MCAO) to generate an ischemic stroke model as described previously [20]. In brief, the rat was anesthetized with ketamine/xylazine (65/6 mg/kg) and put onto a heating pad and feedback control system (FHC, Bowdoin, ME, USA) to keep the body temperature at 37 °C. A siliconized filament was inserted into the internal carotid artery and advanced until the origin of the middle cerebral artery to produce a sudden drop of cerebral blood flow (CBF) to below 25% of baseline measured using a laser Doppler monitor (Periflux PF3, Perimed, Stockholm, Sweden). After 60 min, the filament was removed.

Immediately after reperfusion, rt-PA (10 mg/kg) or saline was administered as an intravenous bolus injection of 1 mg/kg followed by a 9 mg/kg infusion for 30 min with a syringe infusion pump (World Precision Instruments). Thus producing four groups: diabetes + saline, diabetes + rt-PA, non-diabetes + saline, and non-diabetes + rt-PA.

To examine the effectiveness of Exendin-4, rats were randomized into four groups: control + saline, control + rt-PA, Ex-4 + saline, Ex-4 + tr-PA (n = 6 each group). All animals were subjected to MCAO, and the administration of rt-PA was performed as described above. Exendin-4 (Sigma-Aldrich, USA; 10 mg/kg) was administrated by the caudal vein before rt-PA or saline injection [21].

The infarct volume was measured 72 h after MCAO using 2,3,5-triphenyltetrazolium chloride (TTC, Sigma). Only those rats in which CBF below 25% of baseline was achieved and with a return to > 80% after reperfusion were included. Infarct volume was calculated as described previously [22, 23]. To measure BBB permeability, Evans blue dissolved in saline (m/v, 2%) was injected into the caudal vein 72 h after MCAO. Three hours following injection, animals were sacrificed by an overdose of anesthetic and immediately were subjected to transcardiac perfusion with saline. Then brain hemispheres were removed and homogenized in 3 mL of N,N-dimethylformamide (Sigma-Aldrich), followed by incubation for 18 h at 55 °C and subsequent centrifugation. The supernatants were subjected to spectrophotometry at 620 nm.

The hemoglobin content in brain tissues was quantified by spectrophotometric assay. The hemispheric brain tissue was homogenized with PBS and centrifuged for 30 min (13,000 g). The hemoglobin-containing supernatants were collected. A 20-μL aliquot of supernatant was mixed with 80 μL of Drabkin’s reagent (Sigma Diagnostics), followed by incubation at room temperature for 15 min followed by optical density (OD) measurement at 540 nm. The hemorrhage volume was expressed in equivalent units by comparing OD values with a reference curve generated using homologous blood as previously described [23, 24].

The immunofluorescent staining was performed according to previously reported methods [27]. In brief, the rats were anesthetized, first perfused with saline, and then fixed with buffered paraformaldehyde (4%). The brains were removed and immersed in 4% paraformaldehyde overnight for post-fixation and then in 30% sucrose solution overnight. Fifteen-micrometer coronal sections were obtained on a cryostat. The slices were blocked with PBS containing 5% BSA, 10% goat serum, and 0.3% Triton-X 100. Next, the slices were incubated overnight at 4 °C with the primary antibodies: anti-GFAP (1:1000; Millipore, Billerica, MA, USA), anti-NeuN (1:500; Abcam), anti-CD31(1:200; Santa Cruz), anti-CD147(1:500; R&D Systems) or anti-caveolin-1(1:200; BD Biosciences, San Jose, CA, USA), and then Alexa Fluor 488- or 595-labeled secondary antibodies (Molecular Probes Inc., Eugene, OR, USA) for 2 h at room temperature. The tissue sections were washed twice in PBS and then immersed in diphenyl phenylindole (DAPI) solution (1:1000 dilution; Molecular Probes) for 10 min. After being rinsed in distilled water, the sections were fixed with a coverslip using the anti-fade mounting medium.

For immunoprecipitation, HEK293 cells were grown in 60-mm dishes and transiently transfected with 1 μg of plasmids containing HA-caveolin-1 or Flag-CD147 for 48 h in the presence or absence of advanced glycation end products (AGEs, 200 mg/L). Cells were then collected, lysed in immunoprecipitation buffer (1 × PBS, 0.5% Triton X-100, Complete Mini protease inhibitor (Roche Applied Science)) and rotated for 1 h at 4 °C. After centrifugation at 17,500 g for 15 min at 4 °C, the supernatant was collected for immunoprecipitation. Primary monoclonal anti-HA antibody (2 μg, Roche Applied Science) or mouse IgG (2 μg, Santa Cruz) was bound to Dynabeads Protein G (Invitrogen) at room temperature for 30 min by rotation. The bead-antibody complexes were washed in PBS containing 0.01% Tween-20 and then incubated with the above supernatants (500 μL) overnight at 4 °C with rotation. The bead-antibody-antigen complexes were then washed with citrate phosphate buffer (pH 5.0) and boiled in sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer for 5 min to elute the immune complexes. The samples were then separated in gradient gels (5–15%) and immunoblotted with anti-CD147 antibody.

For other immunoblotting experiments, the brain tissues from the striatum supplied by the middle cerebral artery or cells lysed with radioimmunoprecipitation assay buffer (RIPA) containing protease inhibitors (Sigma) were separated by SDS-PAGE and then transferred onto nitrocellulose membranes. The membranes were blocked by 5% (w/v) skim milk (Becton, Dickinson and Company, Le Pont de Claix, France) for 1 h and then incubated overnight at 4 °C with following primary antibodies: anti-ZO-1 (1:1000; Invitrogen), anti-Occludin (1:1000; Invitrogen), anti-collagen IV (1:2000; Abcam), anti-CD147 (1:1000; R&D Systems), anti-Caroline-1 (1:500; BD Biosciences), and anti-β-actin (1:5000; Sigma). Secondary antibodies conjugated with horseradish peroxidase (HRP) were used, and immunoreactivity was visualized with chemiluminescence (SuperSignal Ultra, Pierce, Rockford, IL, USA). Protein bands were analyzed and quantified using Scion Image system.

MMP-2/9 activity was detected according to the manufacturer’s instructions (Invitrogen). In brief, equal amounts of protein samples (30 μg) were mixed with SDS sample buffer and loaded onto 10% Tris-glycine gels containing 0.1% gelatin. Human MMP-2/9 standard was also loaded parallelly in each gel. After electrophoresis, gels were washed with renaturing buffer for 1 h and incubated in developing buffer at 37 °C for 48 h. Gels were stained with 0.5% Coomassie blue R-250 for 1 h, and the stain was then removed appropriately. MMP-2/9 band intensity was quantified and normalized to human MMP-2/9 standard.

Blood was collected 24 h after MCAO by retro-orbital bleeding into polypropylene tubes containing 3.8% trisodium citrate. Platelet-poor plasma was prepared by centrifugation. The amounts of CD147, MMP2, and MMP9 protein in the serum samples were measured using commercial kits (R&D systems, MN, USA) according to the manufacturer’s instructions.

We induced cerebral ischemia in rats by MCAO and treated it with rt-PA. rt-PA thrombolysis was effective in reducing brain infarct volume in both non-diabetic and diabetic rats, but significantly exacerbated intracerebral hemorrhagic transformation (HT) (Fig. 1a–c). Such an undesired side effect became more severe in diabetic rats because the level of HT was already increased by diabetes (Fig. 1c). These results, obtained in rats, are in accordance with our clinical observation in human patients.

Next, we examined the BBB permeability, which is central to HT in ischemic stroke after thrombolytic therapy, using Evans blue dye extravasation assay. The results show that the diabetic rats had higher BBB permeability than non-diabetic controls, while rt-PA therapy increased BBB permeability (Fig. 1e). We further measured the major tight junction proteins, zonula occludens-1 (ZO-1) and occludin, and the major vascular basal membrane protein collagen IV in the peri-infarct region (Fig. 1d, f, g). The results show that in diabetic rats, rt-PA therapy reduced the expression of these proteins in both the non-diabetes and diabetes groups, with the exception of ZO-1. Although these factors were lower in diabetic rats than in the non-diabetic animals. Collectively, we conclude that the exacerbation of HT in rt-PA-treated diabetic rats is due to the increased BBB permeability as a result of accelerated degradation of tight junction proteins and vascular basal membrane proteins.

It has been reported that rt-PA induces the production of serum MMP-9 and MMP-2, which damage the neurovascular unit and promote BBB disruption [28‐30]. We examined the serum MMP-2 and MMP-9 activity at 72 h after MCAO and found that the development of diabetes increased the serum levels of MMP-2 and MMP-9, which were further elevated after rt-PA treatment (Additional file 1: Figure S1A–B). Similarly, CD147, as a key regulator of MMP expression, was upregulated in the serum by diabetes and rt-PA-treatment (Additional file 1: Figure S1C). Although the circulating MMPs contribute to early BBB disruption and HT, the involvement of brain-derived proteases cannot be denied. Therefore, we further examined the MMP-2/9 activity in ischemic brain tissues and CD147 protein in the peri-ischemic region. The results indicated that both the activity of tissue MMPs and CD147 protein were significantly increased by diabetes and rt-PA treatment (Fig. 2a, b, d, e). CD147 occurs in two forms, the 33 kDa LG-CD147 and 40–65 kDa HG-CD147 with different MMP-inducing activity [17]. Both forms were upregulated in diabetic rats, but only HG-CD147 was further upregulated after rt-PA treatment (Fig. 2b). As rt-PA treatment did increase LG-CD147 level in non-diabetic rats, the irresponsiveness of LG-CD147 in diabetic rats seems a result of its complete induction by diabetes. Together, this data implies that diabetes evokes production of brain-derived MMPs by promoting the glycosylation of CD147.

It has been demonstrated that caveolin-1 not only upregulates CD147 glycosylation but also influences the induction of CD147-dependent MMPs in malignant tumor cells [31, 32]. However, whether caveolin-1 is related to diabetes-induced CD147 glycosylation in the central nervous system has not been reported. We examined the level of caveolin-1 expression in the brain and found that this too was also upregulated in association with the expression of HG-CD147 (Fig. 2a, c).

To determine which cells were bearing the increased amount of CD147 and caveolin-1 in the diabetic ischemic brain, we performed immunofluorescent double staining at the peri-ischemic region with cell type-specific markers: GFAP as an astrocyte marker, CD31 as an endothelial cell marker, and NeuN as a neuron marker. As shown in Fig. 2f, in the brain of rt-PA-treated diabetic rats, CD147 co-localized with GFAP-positive astrocytes and CD31-positive endothelial cells, but not with NeuN positive neurons. Consistently, we can determine that caveolin-1 is located in astrocytes and endothelial cells rather than neurons.

In order to further understand the effects of CD147 glycosylation on the production of MMPs in diabetes, in vitro experiments were conducted. As CD147 is predominantly expressed on endothelial cells and astrocytes, BMECs and primary astrocytes were isolated and cultured. BMECs were exposed to high glucose or AGEs to mimic the diabetic stimulation. The cell layers and conditioned media were then subjected to analysis for CD147 and activity of MMPs, respectively. As shown in Additional file 2: Figure S2, high glucose and AGEs had a similar effect on the glycosylation of CD147, and we subsequently chose AGEs for further investigation.

AGEs-rt-PA treatment significantly increased the expression of CD147 proteins in cultured BMECs, of which most were HG-CD147 (40–65 kDa; Fig. 3a). In addition, AGEs-rt-PA treatment increased the secretion of MMP-2 and MMP-9 by 2.5 and fivefold, respectively (Fig. 3c, d). Furthermore, the amounts of two major tight junction proteins, ZO-1, and occludin in BMECs decreased significantly after AGEs-rt-PA treatment (Fig. 3e, f). These results are consistent with those observed in rat models, suggesting that the BMEC-derived CD147 and MMPs (MMP-9 in particular) may participate in the BBB disruption as a consequence of ischemic stroke and rt-PA-induced hemorrhagic complications.

To confirm the role of CD147 glycosylation in MMP induction, we first suppressed the N-glycosylation in BMECs using the general inhibitor tunicamycin (TM). TM treatment was found to successfully deplete N-glycan from CD147, since most CD147 occurred as a 27 kDa band on immunoblot, representing the non-glycosylated form of CD147 (Fig. 3a). As expected, TM treatment significantly suppressed the induction of the activities of MMPs and prevented the degradation of tight junction proteins in BMECs (Fig. 3c–f).

As tunicamycin treatment lacks target protein specificity, we further utilized the specific dominant negative form of CD147, the N152Q mutant. The overexpression of N152Q in BMECs potently reduced the portion of HG-CD147 (Fig. 4a, b) and was found to be as effective as TM treatment in suppressing MMPs and protecting tight junction proteins (Fig. 4d–g).

Taken together, these results demonstrated clearly that HG-CD147 is a key molecule in the diabetes-related BBB disruption and thrombolytic therapy-induced hemorrhagic transformation.

Based on immunofluorescence staining results, CD147 and caveolin-1 are mainly located on endothelial cells and astrocytes. To know whether CD147 glycosylation could also influence the production of MMPs in astrocytes, we isolated primary astrocytes and treated them with rt-PA in the presence or absence of AGEs. AGEs-rt-PA treatment significantly increased the glycosylation of CD147 and increased the activity of MMPs (Fig. 5a, c, d). Unlike in BMECs, more MMP-2 was induced than MMP-9 in astrocytes (Fig. 5c, d—a three- to fourfold increase vs 1.8-fold increase).

Tunicamycin treatment and CD147 N152Q transfection were also applied to astrocytes. The results showed that either pharmacological blockage or genetic interference of the glycosylation of CD147 significantly inhibited the induction of MMPs (Fig. 5c, d, h, i). This data was consistent with those in BMECs.

Caveolin-1 is a scaffold protein capable of collecting signaling molecules, including CD147, into the caveolae and regulating their activities. It has been reported that caveolin-1 may participate in the process of CD147 glycosylation [31, 33]. However, whether caveolin-1 affects CD147 glycosylation in DM is yet to be verified. In this study, we found that upregulated caveolin-1 and CD147 were co-localized on both astrocytes and endothelium in the diabetic model (Fig. 6a), indicating a potential interaction between these two proteins. In addition, the specific suppression of N-glycosylation of CD147 did not impact the expression of caveolin-1 (Figs. 4c and 5g), implying that caveolin-1 functions as an upstream of CD147. To corroborate this, we knocked down caveolin-1 expression (by 80%, data not shown) using siRNA in both astrocytes and BMECs treated with AGEs and rt-PA. We found that a significant proportion of highly glycosylated CD147 was changed into the lowly glycosylated form (Fig. 6b). Furthermore, the activity of MMPs decreased significantly (Fig. 6c, d). These results indicate that caveolin-1 is required for the glycosylation of CD147 in astrocytes and endothelium under rt-PA treatment and diabetic circumstance.

We then examined whether caveolin-1 directly interacts with CD147. To this end, we transfected HEK 293 cells (due to their high transfection efficiency) with Flag-CD147 and HA-caveolin-1 constructs in the presence or absence of AGEs. The expression of these fusion proteins was confirmed by immunoblotting with whole cell lysates (Fig. 6e, input panel). The cell lysates were then immunoprecipitated with anti-HA antibody or matched IgG followed by immunoblotting with CD147 antibody. The results indicated that CD147 was pulled down by anti-HA antibody but not matched IgG (Fig. 6e, ip panel). More CD147 protein was pulled down in the presence of AGEs, which appears to be a result of the increased amount of HG-CD147 after AGEs stimulation (Fig. 6e, ip panel). We conclude that caveolin-1 binds directly to CD147 and regulates its glycosylation.

Our data indicates that diabetes promotes secretion of MMPs through CD147 glycosylation, causing BBB damage. Upon rt-PA treatment, the damage is further exacerbated and leads to significant HT. Considering the possible clinical application of our findings, we examined whether diabetes medication could protect BBB and ameliorate thrombolysis-induced HT without increasing the risk of hypoglycemia. Exendin-4 (Ex-4) is an endogenous glucagon-like peptide-1 receptor (GLP-1R) agonist that promotes insulin secretion. Our previous research shows that Ex-4 decreases warfarin-related HT in diabetic rats by inhibiting MMP-9 activity [21]. In this study, we examined the effect of Ex-4 on rt-PA-related HT in diabetic rats and found that it was also effective in reducing stroke volume (Fig. 7a) and alleviating HT (Fig. 7b). The permeability of BBB elevated by rt-PA treatment was reversed by Ex-4 (Fig. 7c), which may be attributed to the lowered activity of MMPs (Fig. 7e, f) and protection of tight junctions (data not shown). Finally, we examined the CD147 concentration in the brain tissue and found Ex-4 reduced the HG/GL-CD147 ratio in both groups with or without rt-PA (Fig. 7d), supporting the notion that Ex-4 inhibits MMP-2/9 possibly through regulating the HG/GL-CD147 ratio.