Matrix metalloproteinase-9 mediates hypoxia-induced vascular leakage in the brain via tight junction rearrangement

Abstract

Blood-brain barrier (BBB) disruption, resulting from loss of tight junctions (TJ) and activation of matrix metalloproteinases (MMPs), is associated with edema formation in ischemic stroke. Cerebral edema develops in a phasic manner and consists of both vasogenic and cytotoxic components. Although it is contingent on several independent mechanisms, involving hypoxic and inflammatory responses, the single effect of prolonged hypoxia on BBB integrity in vivo was not addressed so far. Exposing mice to normobaric hypoxia (8% oxygen for 48 h) led to a significant increase in vascular permeability associated with diminished expression of the TJ protein occludin. Immunofluorescence studies revealed that hypoxia resulted in disrupted continuity of occludin and zonula occludens-1 (Zo-1) staining with significant gap formation. Hypoxia increased gelatinolytic activity specifically in vascular structures and gel zymography identified MMP-9 as enzymatic source. Treatment with an MMP inhibitor reduced vascular leakage and attenuated disorganization of TJ. Inhibition of vascular endothelial growth factor (VEGF) attenuated vascular leakage and MMP-9 activation induced by hypoxia. In conclusion, our data suggest that hypoxia-induced edema formation is mediated by MMP-9-dependent TJ rearrangement by a mechanism involving VEGF. Therefore, inhibition of MMP-9 might provide the basis for therapeutic strategies to treat brain edema.

Figure 1

Western blots showing reduction of occludin expression after hypoxic exposure. Total protein was extracted from the brains of control (C) mice and mice exposed to 8% oxygen for 48 h (H). (A) Western blots showing reduced occludin expression in hypoxia, but little effect on claudin-5 and Zo-1 expression. Representative samples of three animals in both groups are shown. (B) Quantification and normalization to actin showed significant reduction of occludin expression and little change of claudin-5 and Zo-1 in hypoxia. Values are mean±s.d. (^*P<0.05; n=7 to 14 animals).

Figure 2

Hypoxia correlates with localization changes of occludin and Zo-1. Mice were exposed for 48 h to 20% (control) or 8% oxygen (hypoxia). At least six coronal brain sections per animal were stained for claudin-5, occludin, or Zo-1 and co-stained with platelet endothelial cell adhesion molecule-1 (Pecam-1). Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). Representative pictures for both groups are shown. (A) Vessels in the cortex of control mice showed a continuous, linear labeling of occludin and Zo-1 along the whole vessel (arrows). In hypoxic brains, a discontinuous, less regular distribution of Zo-1 and occludin in the vessels was noted (arrowheads) (n=6 animals). Little change was seen in the distribution pattern of claudin-5. (B, D) Three-dimensional reconstruction after confocal microscopy demonstrates localization changes of occludin and Zo-1. The TJ proteins (green) are located at the cell margins (arrows) of Pecam-1-positive endothelial cells (red). Hypoxic vessels showed an irregular and diffuse staining and gap formation (arrowheads). (C, E) To quantify disruption and gap formation, gap length was measured (in % of complete staining). Hypoxia significantly increased gap length and disruption of occludin and Zo-1. Values are mean±s.d. (^*P<0.001; n=5 animals).

Figure 3

In situ zymography shows increased gelatinolytic activity in brain microvasculature after hypoxic exposure. Cryosections were incubated with gelatin-FITC substrate for 3 h and vascular structures were co-stained with platelet endothelial cell adhesion molecule-1 (Pecam-1; red). (A) Representative pictures showing no gelatinolytic activity (green) in normoxic control vessels (arrowheads) but clearly increased activity in hypoxia (arrows). Inlays show higher magnification of a labeled vascular structure. (B) Quantification revealed a two- to four-fold increase in the number of microvessels showing gelatinolytic activity in cortex, thalamus, and hypothalamus (^*P<0.0001, ^**P<0.000001; n=5 animals per group).

Figure 4

Hypoxia upregulates cerebral MMP-9. (A) Representative gelatin zymography demonstrating upregulation of the proform of MMP-9 (proMMP-9) activity after 48 h of hypoxia (H), compared with normoxic control (C). A second band corresponding to the cleaved-activated MMP-9 enzyme was detected in hypoxia only. Using the specific MMP inhibitor p-aminobenzoyl-gly-pro--leu--ala-hydroxamate (AHA), proMMP- and MMP-9 upregulation in hypoxia was completely blocked. ProMMP-2 activity was unchanged and no activated MMP-2 was detected in hypoxia. ST: standards as positive controls for MMP-2 and MMP-9. (B) Densitometric quantification showing significantly increased proMMP-9 in hypoxia, and no change for proMMP-2. The MMP inhibition strongly reduced hypoxia-induced activity of proMMP-9, even below the normoxic activity level, but had no effect on pro-MMP-2 (C) Elevated activation of MMP-9 in hypoxia was completely blocked by AHA. Bars show mean values (±s.d.) of relative gelatinolytic activity (^*P<0.05, ^**P<0.005; n=3 to 6 animals per group).

Figure 5

The MMP inhibition blunts vascular leakage in vivo. Mice received 60 mg/kg p-aminobenzoyl-gly-pro--leu--ala-hydroxamate (AHA) intraperitoneal (ip) (+AHA) or vehicle as control and were exposed to 8% oxygen for 48 h (H) or kept at room air (C). Sodium fluorescein injection was used to quantify vascular permeability. After cardiac perfusion, fluorescence was quantified per mg of brain tissue (in relative fluorescence units, r.f.u.). The MMP inhibition significantly reduced hypoxia-induced hyperpermeability, but had no effect under normoxic conditions. Values are means±s.d. (^*P<0.05, ^**P<0.001; n=4 to 5 animals per group).

Figure 6

The MMP inhibition reverts rearrangement and gap formation of occludin and Zo-1. After pretreatment with p-aminobenzoyl-gly-pro--leu--ala-hydroxamate (+AHA) or vehicle mice were exposed to hypoxia for 48 h at 8% oxygen (H) or kept at room air (C). Thereafter, cryosections from cortex were stained for occludin and Zo-1, and co-stained with platelet endothelial cell adhesion molecule-1 (Pecam-1). Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). (A, C) Three-dimensional reconstruction after confocal microscopy depicting continuous, linear, and sharp labeling of occludin and Zo-1 along the whole vessel (arrows), and discontinuous, irregular and diffuse staining and gap formation (arrowheads). (A, B) Pretreatment with AHA significantly reduced hypoxia-induced redistribution and relative gap formation by 30% for occludin, and (D) by 60% for Zo-1. Values are means±s.d. (^*P<0.05, ^**P<0.001; n=3 to 5 animals per group).

Figure 7

The VEGF inhibition reduces hypoxia-induced MMP-9 activity. (A) Mice receiving ip 10 mg/kg bevacizumab (Avastin) (+αVEGF) or vehicle were exposed to 8% oxygen for 48 h (H) or kept at room air (C). Sodium fluorescein was used to quantify vascular permeability. Fluorescence was quantified per mg of brain tissue (in relative fluorescence units, r.f.u.). The VEGF inhibition significantly attenuated hypoxia-induced vascular permeability, but had no effect under normoxic conditions. Values are means±s.d. (^*P<0.05, ^**P<0.001; n=6 animals per group). (B) Representative gelatin zymography demonstrating MMP-9 upregulation after of hypoxia (H) compared with normoxic control (C). Treatment with bevacizumab significantly attenuated MMP-9 upregulation in hypoxic brains. Bars show mean values (±s.d.) of relative gelatinolytic activity (^*P<0.05; n=3 to 6 animals per group).