TPA increases cerebrovascular permeability through PDGF-CC

The demonstration that tPA injection into the cerebrospinal fluid (CSF) rapidly increases cerebrovascular permeability¹⁷ led us to propose that tPA regulates cerebrovascular permeability through a substrate-dependent receptor-mediated cell signaling event within the NVU. The NVU is composed of endothelial cells, astrocytes, microglia, neurons and smooth muscle cells or pericytes. To determine if tPA acts specifically on the CNS side of the NVU, we compared the effect of tPA injection into the CSF with tPA administered intravenously on Evans blue (EB) dye extravasation in nonischemic mice. This analysis revealed that tPA acts on the abluminal side of the NVU since increased cerebrovascular permeability was observed only when enzymatically active tPA was injected into the CSF, and not when administered intravenously (Fig. 1a, Supplementary Fig. 1 online). Since tPA activates latent PDGF-CC^20,22, we examined if active PDGF-CC also induced cerebrovascular permeability. Like tPA, the intraventricular injection of active PDGF-CC significantly (p < 0.01) increased cerebrovascular permeability to an extent that was similar to that observed with tPA (Fig. 1b). The combined treatment with both tPA and PDGF-CC was not significantly different from either treatment alone indicating no additive or synergistic increase in cerebrovascular permeability with the two agents. This suggests that tPA and PDGF-CC may act within a common pathway.

To test whether PDGF-CC was acting downstream of tPA, neutralizing antibodies against PDGF-CC (Supplementary Fig. 2 online) were co-injected with tPA. These data demonstrated that PDGF-CC-specific antibodies, but not pre-immune IgG, significantly (p < 0.05) inhibited the tPA-induced increase in cerebrovascular permeability (Fig. 1c). This indicates that PDGF-CC acts downstream of tPA and suggests that tPA activates latent PDGF-CC within the CNS.

PDGF-CC is a known agonist of the PDGFR-α²¹. Therefore, we tested whether two other PDGFR-α agonists, PDGF-AA and PDGF-BB²³, also increased cerebrovascular permeability when injected into CSF. Both PDGF-AA and PDGF-BB increased EB extravasation significantly (p < 0.01) compared to the PBS-treated group, demonstrating a PDGFR-α dependent process (Fig. 1d). However, in contrast to PDGF-CC-specific antibodies, neutralizing PDGF-AB-specific antibodies had no effect on tPA-mediated enhancement of BBB permeability (Supplementary Fig. 3 online).

Several studies have shown that tPA-induced cerebrovascular permeability is inhibited by antagonists of members of the low-density lipoprotein receptor family (LDL-Rs)^17,19,24. Therefore, we examined whether the LDL-R antagonist RAP also inhibited PDGF-CC-induced cerebrovascular permeability. In contrast to its inhibition of tPA-induced EB extravasation¹⁷, RAP had no effect on the ability of active PDGF-CC to increase cerebrovascular permeability (Fig. 1e). This implies that active PDGF-CC does not require LDL-Rs to mediate its effect on the NVU, suggesting that like tPA, the LDL-Rs act upstream of active PDGF-CC. Since both tPA and PDGF-CC bind to the low-density lipoprotein receptor related protein (LRP) (Supplementary Fig. 4 online), we speculated that LRP may facilitate activation of PDGF-CC by tPA. To test this we investigated tPA-mediated cleavage of PDGF-CC in the presence of mouse embryonic fibroblasts (MEF) that either express LRP or are genetically deficient in LRP²⁵. These data demonstrated that both cell types express similar amounts of latent PDGF-CC, but that only cells expressing LRP stimulate efficient activation of PDGF-CC by tPA (Fig. 1f). This suggests that in vivo cell-associated LRP can markedly enhance PDGF-CC activation by tPA.

Analysis of vascular morphology by electron microscopy 1 h after intraventricular injection of either tPA or active PDGF-CC indicates that both proteins induced specific morphological changes in cerebral vessels compared to control injection (Fig. 2). In control tissue, vessels were intact and tightly associated with surrounding tissue (Fig. 2a,d). In contrast, arterioles from mice receiving intraventricular injections of tPA or PDGF-CC were surrounded by open areas of apparent fluid accumulation (Fig. 2b–c,e–f). This phenotype was observed in multiple arterioles in tPA and PDGF-CC-treated animals but not in brain capillaries (Fig. 2b–c). The similarity in morphological changes observed in tPA and PDGF-CC-treated animals along with the similar degree of increased cerebrovascular permeability induced by both agents suggests that PDGF-CC and tPA provoke comparable vascular changes that occur primarily in arterioles.

PDGF-CC, tPA and PDGFR-α in the NVU

Immunohistochemical staining demonstrated that PDGF-CC localization throughout the brain closely resembled the reported expression of tPA²⁶, with the most abundant staining observed in cortex, striatum, and hippocampus (not shown). Higher resolution microscopy indicated that PDGF-CC was often closely associated with arterioles but not with capillaries, and was typically seen as a stained patch on one side of larger vessels (Fig. 3a). Immunohistochemical analysis of tPA demonstrated abundant vessel-associated tPA throughout the brain. In capillaries, the staining was largely confined to endothelial cells while in arterioles, tPA was also associated with perivascular cells (Fig. 3b). Thus, both PDGF-CC, and tPA were localized in the NVU of arterioles. The expression of PDGF-CC and its receptor PDGFR-α in brain was also investigated using mice doubly heterozygous for a nuclear-targeted histone H2B/green fluorescent protein (GFP) fusion protein inserted into the Pdgfrα locus²⁷ (Supplementary Fig. 5 online), and for a lacZ-reporter inserted into the Pdgfc locus²⁸. Brains from the Pdgfrα^+/GFP/Pdgfc^+/lacZ mice demonstrated both PDGFR-α and PDGF-CC expression in cells associated with arterioles (Fig. 3c, arrows). PDGF-CC expression was also seen in cells not directly associated with vessels (arrowheads).

To further characterize PDGFR-α-positive vessels, GFP-positive vessel fragments were isolated from brain. The most intensely GFP-positive vessels resembled arterioles and were covered with firmly attached GFP-expressing cells. Whole mount immunofluorescence staining of these vessel fragments for smooth muscle actin (SMA) revealed that the vessels were arterioles since they were strongly positive for SMA (Fig. 3d). However, the GFP-positive nuclei did not appear to directly co-localize with SMA and instead localized to the outside of the SMA-positive cells. In contrast, staining of the GFP-positive vessels for the astrocyte marker, glial fibrillar acidic protein (GFAP), showed that GFP-expressing nuclei appeared to be embedded in the GFAPpositive cells (Fig. 3e). To confirm these observations, intact brain sections from GFP-expressing mice were stained by immunofluorescence for GFAP, SMA, or endothelial cells (PECAM) and imaged together with GFP (Fig. 3f–h). These results showed that PDGFR-α largely co-localized with perivascular GFAP positive cells (Fig. 3f, yellow), and not in cells expressing SMA (Fig. 3g) or PECAM (Fig. 3h), demonstrating that perivascular astrocytes are the primary neurovascular cells expressing PDGFR-α. These data suggest that tPA, PDGF-CC and PDGFR-α are localized to arterioles in the CNS where they interact on the surface of perivascular astrocytes.

Since LRP is also known to be specifically associated with arteriolar perivascular astrocytes²⁹, we tested whether type I fibrous astrocytes in culture could enhance the activation of PDGF-CC by tPA, and whether active PDGF-CC could stimulate PDGFR-α phosphorylation. These studies demonstrated that astrocytes in culture expressed latent PDGF-CC that was cleaved by added tPA (Fig. 4a). Active PDGF-CC also stimulated PDGFR-α phosphorylation in astrocytes (Fig. 4b), as did the addition of tPA, suggesting a possible autocrine effect of PDGF-CC on astrocytes (Fig. 4c).

PDGFR-α regulates cerebrovascular permeability after stroke

To examine whether PDGFR-α was activated in a tPA-dependent manner during cerebral ischemia we developed a mouse photothrombotic model of stroke that is similar to a previously described model^30,31. This model permits thrombolytic treatment and can be analyzed either with or without reperfusion (Supplementary Fig. 6 online). Six hours after unilateral MCAO in wild-type and tPA-deficient mice, isolated brains were divided into hemispheres, ipsilateral and contralateral to the MCAO, and each hemisphere was analyzed separately for PDGFR-α activation (Fig. 5a,b). Following stroke, PDGFR-α phosporylation was induced ~2-fold in the ipsilateral hemisphere of wild-type mice compared to the nonischemic hemisphere, while in tPA-deficient animals no induction of PDGFR-α activation was seen. These data suggest that PDGFR-α activation within the brain following ischemic stroke is tPA-dependent through PDGF-CC activation.

To test whether MCAO-induced activation of the PDGFR-α regulates the cerebrovascular response to stroke, we treated mice with the PDGFR-α inhibitor³² Imatinib mesylate 1 h after MCAO. Animals treated with Imatinib exhibited a 33% reduction in EB extravasation after MCAO compared to control mice (Fig. 5c–e), suggesting that PDGFR-α signaling regulates cerebrovascular permeability after stroke. To determine if this effect was due to activation of PDGFR-α by PDGF-CC, PDGF-CC-specific antibodies were injected intraventricularly into mice immediately after stroke induction, and EB extravasation was evaluated 24 h after MCAO (Fig. 5f). These data demonstrate that like Imatinib treatment, EB extravasation was significantly (p < 0.05) reduced in animals treated with PDGF-CC-specific antibodies but not in mice receiving preimmune IgG. This supports the hypothesis that endogenous PDGF-CC mediates the regulation of cerebrovascular permeability after stroke via activation of PDGFR-α.

Since reducing the extent of BBB dysfunction during stroke may impact infarct expansion and improve stroke outcome, we compared infarct volumes in mice treated with Imatinib to controls 72 h after MCAO. Imatinib significantly (p < 0.05) reduced lesion volume by 34% (Fig. 5g), suggesting that interfering with the PDGF-CC/PDGFR-α system may improve stroke outcome.

A major risk associated with the use of thrombolytic tPA in ischemic stroke is the occurrence of hemorrhagic complications^33–36. Since bleeding represents an extreme example of loss of BBB integrity we examined whether inhibition of PDGFR-α with Imatinib administered 1 h after the onset of ischemia could reduce the extent of hemorrhagic complications associated with tPA-mediated thrombolysis initiated 5 h after the induction of MCAO. These data indicated that Imatinib treatment significantly (p < 0.05) decreased hemorrhage, reducing the amount of hemoglobin associated with the ischemic hemisphere by 50% compared to untreated mice (Fig. 6a–c). This suggests that the known association of thrombolytic tPA with hemorrhagic complications in some patients may be due in part to the activation of PDGF-CC by therapeutic tPA.

Proteins and antibodies

The recombinant core domain of human PDGF-CC was produced in baculo-virus infected insect Sf9 cells and purified as described²¹. Blocking antibodies to active PDGF-CC (Supplementary Fig. 2 online) were affinity purified rabbit IgG isolated from rabbit no. 615 as described²¹. Preimmune IgG from the same animal was isolated on a protein A-Sepharose column (Amersham Biosciences). Mouse tPA and affinity purified antibodies to mouse tPA were from Molecular Innovations.

Cell culture, and immunoblotting

All cells used were maintained in Dulbecco’s Modified Eagle Medium (DMEM; 41965, Gibco) supplemented with 10% fetal calf serum (FCS), 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. The cells were cultured at 37°C in a humidified 5% CO₂ atmosphere. The mouse astrocyte cell line C8-D1A was obtained from ATCC and grown in the presence of 1 mM pyruvate. Sub-confluent cultures of wild-type and Lrp^−/− MEF cell lines²⁵ were kept in serum-free DMEM over night with or without the addition of 1 µg/ml tPA. Conditioned serum-free medium was collected, and proteins precipitated using trichloroacetic acid (TCA) as described²¹. PDGF-C species were detected by immunoblotting using PDGF-CC-specific antibodies (no.615).

Intraventricular Injections

To perform intraventricular injection of tPA, active PDGF-CC, and PDGF-CC-specific antibodies, wild type C57BL/6J mice were anesthetized with chloral hydrate (450 mg/kg, IP), placed on a stereotactic frame, and injected at bregma – 2, mediolateral 0, and dorsoventral 2¹⁷. Injections contained 3 ìl of either active tPA (3 µM), active PDGF-CC core protein (3 µM), PDGF-CC-specific antibodies (0.4 mg/ml), preimmune IgG (0.4 mg/ml) or PBS. Cerebrovascular permeability was then determined 1 h later from EB extravasation as described in Supplementary Methods 1 (online), except that brains were not divided into hemispheres. All animal experiments were approved by the Institutional Animal Care and Use Committee of Unit for Laboratory Animal Medicine at University of Michigan.

Electron Microscopic Analysis

One hour after intraventricular injection of either PBS, tPA or active PDGF-CC, C57BL/6J mice were perfused with PBS followed by 4% formaldehyde with 1% glutaraldehyde and fixed with the same solution for 24 h. The brain tissues were then rinsed with 5% sucrose phosphate buffer 3 times over 30 min, then post-fixed in 1% aqueous OsO₄ for 1 h and then stained with 1% uranyl acetate for 1 h. After dehydration, the brain tissues were embedded in EPON (EMS), and ultra thin sections were cut and stained with uranyl acetate and lead citrate. Some sections were stained with toluene blue for analysis by light microscopy. For electron microscopic analysis, sections were examined with a Phillips CM12 transmission electron microscope.

Immunohistochemical localization

For immunohistochemistry, mice were perfused with 4% PFA in PBS, and the brains were post-fixed overnight in the same fixative, and then processed for paraffin-embedding and sectioned to 6 µm using standard protocols. Immunohistochemical staining of PDGF-CC and tPA expression was performed using affinity-purified rabbit IgG against human core PDGF-CC (10 µg IgG/ml) and affinity-purified rabbit antibody to mouse tPA (15 µg IgG/ml) respectively. Controls using normal rabbit IgG gave only background staining (data not shown). An Elite ABC Vectastain kit (Vector Laboratories) was used. Blocking of non-specific binding was performed using TNB blocking buffer (TSA indirect NEN Bioscience). For the tPA staining, antigen retrieval with 0.1 M citrate pH 3.0 at 95 °C for 20 min followed by treatment with 0.25% trypsin at 37 °C for 20 min was performed before the immunostaining as above.

Immunofluorescence localization

Details of the immunofluorescence localizations are presented in Supplementary Methods 2 online.

Analysis of receptor activation

PDGFR-α activation was determined essentially as described⁴⁴. Briefly, for cell culture studies sub-confluent astrocyte cultures were stimulated with 100 ng/ml active PDGF-CC for 90 min on ice, or treated with tPA overnight, and then lysed as described²¹. Cell lysates were directly subjected to immunoblotting, and receptor tyrosine phosphorylation was determined using a phosphotyrosine-specific antibody (PY99, Santa Cruz). As a loading control immunoblotting of the endoplasmic reticulum integral membrane protein calnexin was also performed (sc-6465, Santa Cruz). For analysis of receptor activation after MCAO, both wild-type and tPA knockout mice (KO), male 10 weeks old, were subjected to MCAO. Six hours later, animals were perfused with PBS for 4 min and the brains were removed and separated into two hemispheres. Ipsilateral and contralateral hemispheres were then processed for PDGFR-α activation by immunoprecipitation of the PDGFR-α using a specific PDGFR-α antibody followed by immunoblotting for receptor tyrosine phosphorylation using the same phosphotyrosine-specific antibody as in the cell culture studies above. The amount of precipitated PDGFR-α was monitored by reprobing the blots with a PDGFR-α-specific antibodies (R&D Systems).

Mouse model of ischemic stroke

Details of the mouse model of ischemic stroke are presented in Supplementary Methods 1 and in Supplementary Figure 6 online.

Hemoglobin content assay

Twenty four hours after MCAO, animals were anesthetized with chloral hydrate (450 mg/kg, I.P., Morton Grove Pharmaceuticals, Inc.) and euthanized by exsanguination. Animals were then perfused with PBS for 4 min and the brains were removed and separated into hemispheres ipsilateral and contralateral to the MCAO. Each hemisphere was homogenized (Model T8, IKA Works, Inc) in 475 µl PBS at 25,000 rcf for 30 s on ice in a 1.5 ml microfuge tube. Twenty five µl 10% Triton X100 was added to the tube to make the final concentration of 0.5%. After thorough mixing, the tube was kept at room temperature for 5 min and then centrifuged at 25,000 rcf at 4 °C for 30 min (Eppendorf centrifuge, model 5417R). Fifty µl of the supernatant was read at A410nm (SpectraMax M5, Molecular Devices) and the hemoglobin content was quantified relative to a purified hemoglobin (Sigma-Aldrich) standard curve as described⁴⁵.

Statistical Analysis

Data were analyzed using either a Wilcoxon Mann rank sum test or in cases where more than one group is compared, by ANOVA with repeated test (GraphPad, InStat). p values smaller than 0.05 were considered significant.