OUP user menu

Dexamethasone regulation of matrix metalloproteinase expression in CNS vascular endothelium

K. A. Harkness, P. Adamson, J. D. Sussman, G. A. B. Davies-Jones, J. Greenwood, M. N. Woodroofe
DOI: http://dx.doi.org/10.1093/brain/123.4.698 698-709 First published online: 1 April 2000


Matrix metalloproteinases (MMPs) have been implicated in the early breakdown of the blood–brain barrier in neuroinflammatory disease. Although expression of these enzymes by resident glial cells and recruited immune cells has been described, altered expression of MMPs by the CNS vascular endothelial cells may also contribute to barrier disruption. In the present study, the in vitro expression of MMP-2 and -9 as well as tissue inhibitor of metalloproteinase (TIMP)-2 by rat CNS microvascular endothelial cells has been determined and compared with that by endothelial cell lines derived from rat aorta and high endothelial venules. Primary cultures of rat brain microvascular endothelial cells as well as the rat brain (GP8/3.9) and rat retinal endothelial (JG2/1) cell lines constitutively expressed MMP-2, -9 and TIMP-2. In vitro activation of CNS endothelium with the pro-inflammatory cytokines, tumour necrosis factor-α and interleukin-1β, resulted in selective upregulation of MMP-9 activity, whereas no significant changes were seen in MMP-2 or TIMP-2 levels at 24 h. The addition of dexamethasone partially inhibited the cytokine-induced upregulation of MMP-9. Treatment of GP8/3.9 brain endothelial cells with active MMP-9 caused subtle but distinct alterations in the expression of the junctional protein, ZO-1. Quantitative differences found between CNS and non-CNS endothelial cells in the expression of both MMP-2 and -9, and in the expression of TIMP-2 demonstrate that CNS vascular endothelium is functionally distinct from non-CNS endothelium. These results suggest that cytokine-induced upregulation of MMP-9 expression by the CNS vascular endothelium may play a role in the pathogenesis of blood–brain and blood–retinal barrier breakdown in vivo.

  • blood–brain barrier
  • endothelium
  • matrix metalloproteinases
  • TIMP-2
  • ZO-1
  • AEC = aortic vascular endothelial cells
  • ELISA = enzyme-linked immunosorbent assay
  • HEV = high endothelial venule
  • IFN = interferon
  • IL = interleukin
  • MMP = matrix metalloproteinase
  • TIMP = tissue inhibitor of metalloproteinases
  • TNF = tumour necrosis factor


The microenvironment of the CNS is normally maintained by the presence of the blood–brain barrier. This is a complex cellular system comprising cerebral vascular endothelial cells linked by continuous tight junctions, resting on a basal lamina of collagen type IV, fibronectin, laminin and proteoglycans. The endothelial cells are in close apposition to pericytes and astrocytic foot processes, which are thought to contribute to the functional maintenance of barrier characteristics (Bradbury, 1985). Breakdown of the blood–brain barrier is a key feature of neuroinflammatory conditions and is associated with the influx of inflammatory cells, fluid and proteins, including complement and cytokines. Mediators of blood–brain barrier disruption include the classic vasoactive amines and other autocoids but, recently, matrix metalloproteinases (MMPs), a group of zinc-containing endopeptidases, have also been implicated in the pathogenesis of blood–brain barrier breakdown. Intracerebral injection of 72 kDa gelatinase (MMP-2) leads to barrier disruption which is partially inhibited by administration of tissue inhibitor of metalloproteinases (TIMP)-2 (Rosenberg et al., 1992). The 92 kDa gelatinase (MMP-9) also mediates barrier disruption following the intracerebral administration of tumour necrosis factor (TNF)-α and lipopolysaccharide (Rosenberg et al., 1995), although the mechanism by which proteases disrupt the barrier remains unclear. Proteases may have a direct effect on the external domain of junctional proteins, or lead to degradation of collagen type IV in the basal lamina, disrupting integrin binding between endothelial cells and the underlying basement membrane, which may in turn impact on junctional integrity.

Whatever the mechanism of action, evidence from both animal models and human studies supports a role for MMP-9 in blood–brain barrier disruption during stroke, multiple sclerosis, experimental allergic encephalomyelitis and other neuroinflammatory conditions (Anthony et al., 1997; Chandler et al., 1997; Cossins et al., 1997; Kieseier et al., 1998; Umehara et al., 1998; Yong et al., 1998). In a longitudinal study, higher serum levels of MMP-9 were demonstrated in patients with multiple sclerosis compared with controls, with significantly elevated levels during episodes of clinical relapse, which was also associated with gadolinium enhancement on brain MRI (Lee et al., 1999). However, what is not clear from these studies is the source of MMPs, although invading leucocytes, endothelial cells, astrocytes and microglia are all potential sources of MMP-9 (Colton et al., 1993; Hanemaaijer et al., 1993; Cottam et al., 1996; Gottschall and Deb, 1996; Unemori et al., 1996; Wells et al., 1996; Cross and Woodroofe, 1999). Considerable evidence now supports the view that MMPs are derived from perivascular cells, in particular recruited immune cells (Leppert et al., 1995; Cuzner and Opdenakker, 1999), although an alternative and likely source of MMP-9 is the cerebral endothelial cell. In a study of experimental allergic encephalomyelitis in Lewis rats, increased MMP-9 levels correlated with disease onset, and MMP-9 immunoreactivity was detected primarily in the meninges and around blood vessels, with a predominantly extracellular localization. However, co-localization of immunoreactivity for MMP-9 with T-cell or macrophage antigens was not observed, suggesting that the source of MMP-9 may not derive from recruited immune cells (Kieseier et al., 1998).

The suggestion that vascular endothelial cells could be a significant source of MMP-9 in early disruption of the blood–brain barrier is an attractive hypothesis. Alteration in the balance of MMPs and their natural inhibitors, TIMP, may lead to an overall increase in the permeability of the blood–brain barrier, giving rise to vasogenic oedema and unimpeded influx of inflammatory cells to the CNS. We have therefore investigated the expression of these proteolytic enzymes in primary cultures of rat cerebral endothelial cells and a variety of CNS and peripheral endothelial cell lines. The regulation of these specific proteases in response to stimulation with the pro-inflammatory cytokines TNF-α, interleukin (IL)-1β and interferon (IFN)-γ, which have been well documented as playing a role in CNS disease (Benveniste, 1992; Owens et al., 1994; Woodroofe, 1995), and their regulation by dexamethasone have been assessed.



Unless otherwise stated, all reagents were obtained from Sigma Chemical Company (Poole, Dorset, UK) and were of the highest grade available.

Primary brain endothelial cell cultures

Brain endothelial cells were isolated from two, 2- to 3-month-old female Lewis rats (Harlan, UK) following published techniques (Abbott et al., 1992). Cells were cultured in Hams F-10 medium containing 75 μg/ml endothelial cell growth supplement (First Link, UK), 40 μg/ml heparin, 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 5 μg/ml vitamin C and 20% plasma-derived serum (First Link, UK). Cultures were maintained at 37°C in 5% CO2/95% air in a humidified atmosphere and the medium was changed every 3 days for 7–10 days until confluent monolayers had formed. Cells routinely were phenotyped as endothelial cells by positive staining with von Willebrand factor, and any wells containing <90% endothelial cells by microscopic examination were rejected from further studies. A total of four separate preparations were carried out, each requiring two animals.

Rat aortic endothelium

Aortic endothelial cells (AEC) were isolated according to previously published methods (McGuire and Orkin, 1987). Rat aortas were removed by dissection, cut into 2–5 mm pieces, placed luminal side down onto collagen-coated 24-well plates and grown in RPMI medium (Life Technologies, UK) supplemented with 20% foetal calf serum, 75 μg/ml endothelial cell growth supplement (First Link, UK) 40 μg/ml heparin, 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. After 3 days, the explants were removed and outgrowing cells were expanded and passaged by trypsinization and re-plating onto collagen-coated plastic culture plates. At confluence, the cells had a `cobblestone' morphology, expressed von Willebrand factor and grew in medium containing d-valine (a capacity which is lacking in fibroblasts and smooth muscle cells). Cells were used after passage 3, which is the earliest stage at which sufficient cells were available for experimentation.

CNS endothelial cell lines

The rat brain (GP8/3.9) and rat retinal (JG2/1) endothelial cell lines were grown on bovine collagen-coated plastic 24-well plates (Falcon, UK) in the same medium as described above for primary cultures. These cell lines have been shown to retain their phenotypic and immunological characteristics in vitro (Greenwood et al., 1996; Adamson et al., 1999). All experiments were carried out on cell passages 8–12.

High endothelial venule cells

Rat high endothelial venule (HEV) endothelial cells (Ager, 1987) were generously supplied by Dr Ann Ager, (NIMR, London, UK). HEV cells were grown in RPMI medium (Life Technologies, UK) supplemented with 20% fetal calf serum, 75 μg/ml endothelial cell growth supplement (First Link, UK), 40 μg/ml heparin, 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin.

Cytokine activation of endothelial cells

Cell lines were seeded at an initial density of 1 × 105 cells/ml into 24-well cell culture plates (Falcon, UK) and grown to confluence prior to stimulation. Cells were washed twice in endothelial serum-free medium (Life Technologies, UK) and incubated in this medium for 1 h prior to stimulation. Cells were stimulated with recombinant rat cytokines, TNF-α, IL-1β and IFN-γ (Peprotech, UK), at concentrations of 0.1, 1.0 and 10 ng/ml in a total volume of 300 μl of serum-free medium for 24 h. Serum-free medium alone was used as a control. To assess the effect of dexamethasone on MMP expression, cells were grown in medium containing dexamethasone (1 μM) alone, or with dexamethasone (1 μM) in the presence of cytokine as described above, for 24 h. Supernatants were harvested from duplicate wells mixed and then centrifuged at 1000 r.p.m. for 10 min to remove cell debris prior to aliquoting and storage at –70°C. A minimum of three independent assays were performed for each group.

Zymographic assesment of matrix metalloproteinase activity

MMP activity was determined by SDS–PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis) zymography using pre-cast 10% polyacrylamide gels containing 10% gelatin (Novex, UK). Freshly defrosted culture supernatants were suspended in sample buffer (10% SDS, 0.1% bromophenol blue, 0.5 M Tris–HCl pH 6.8) at a ratio of 1 : 2. A total volume of 20 μl was resolved by SDS–PAGE. Gels were placed in 2.5% Triton X-100 for 1 h to remove SDS, and then incubated for 24 h at 37°C in buffer containing 5 mM CaCl2 and 50 mM Tris–HCl pH 7.4, prior to fixation and staining in 30% methanol, 10% acetic acid and 0.5% w/v Coomassie brilliant blue for 6 h. Gels subsequently were destained and proteolysis detected as a horizontal white band on a blue background. For confirmation that gelatinolytic bands resulted from metalloproteinase activity, identical experiments were performed in which gels were incubated with 5 mM EDTA, to inhibit MMP enzyme activity. Purified human recombinant MMP-2 and -9 (TCS Biologicals, UK) were used as standards. High molecular weight marker proteins were used to estimate molecular weight.

Images of gels were captured by scanning on a UMAX Astra 1200S scanner, using Corel Photopaint software. Images subsequently were analysed using Kodak Digital Science gel analysis software. Gelatinase activity values were obtained by measuring the comparative peak intensities. Statistical analysis of data was determined by paired Student's t-test.

Analysis of MMP9 and MMP2 mRNA

RNA was isolated from confluent 90 mm dishes of GP8/3.9, AEC and HEV cells using RNAeasy (Qiagen, UK) columns and quantitated by UV spectrometry at 260 nm. A 1 μg aliquot of RNA extracted from GP8/3.9 cells was primed with oligo(dT)15–18 and reverse transcribed with AMV reverse transcriptase (Gibco, UK). Single-stranded cDNA subsequently was amplified with oligonucleotide primer pairs complementary to MMP-9 [5′-AAATGTGGGTGTACTACAGG-3′ (1764–1783) and 5′-ACAGG-3′ (1764–1783) and 5′-TCAAGGTGTTGCCCACT-3′ (2073–2092)] and to MMP-2 [5′-CTATTCTGCCAGCCACT-3′ (1653–1670) and 5′-GACTTTGGTTCTCCAG-3′ (1943–1959)] sequences. PCR (polymerase chain reaction) cycling was repeated 35 times in buffer containing 10 mM Tris, pH 8.0, 50 mM KCl, 1.5 mM MgCl2, 200 μM dNTPs, 0.2 μM of each primer and 1 U of Taq DNA polymerase. cDNA was denatured at 94°C for 1 min, primers annealed at either 57 or 58°C for 30 s and DNA extended at 72°C for 1 min. Resultant PCR products were purified by agarose gel electrophoresis and recovered into solution using QIAquick columns (Qiagen, UK). PCR products were cycle sequenced and showed 100% identical sequence homology with MMP-2 and MMP-9. PCR products subsequently were labelled, with [α-32P]dCTP using random hexamers and Klenow fragment of DNA polymerase using standard methods, and used as probes in Northern analysis.

A 5 μg aliquot of RNA from GP8/3.9, AEC and HEV cells was denatured with formamide and resolved on a 1.2% agarose–formaldehyde gel in 20 mM MOPS pH 7, 1 mM EDTA and 8 mM sodium acetate. RNA was electrotransferred to Hybond-N+ at 3.55 mA/cm2 for 40 min and fixed by UV illumination at 700 J for 30 s. Membranes were pre-hybridized in 5 × SSPE (0.075 M NaCl, 0.05 M NaH2PO4, 0.004 M Na2EDTA), 10 × Denhardt's solution, 100 μg/ml salmon sperm DNA, 50% formamide and 2% SDS for 4 h at 42°C and sequentially hybridized overnight with PCR-derived probes in pre-hybridization solution for MMP-9 and -2 at 42°C. Unhybridized probes were removed after repeated washing with 2 × SSC (0.3 M NaCl, 0.03 M trisodium citrate) containing 0.1% SDS. Membranes were stripped after each probing using 0.1% SDS at 100°C. Membranes were exposed to X-ray film (Fuji, UK) at –70°C in the presence of intensifying screens.

MTT cellular proliferation assay

In order to establish whether alterations in MMP activities were due to altered protein production or differences in cellular proliferation, MTT assays were performed on the cell lines according to published techniques (Mosmann, 1983).

Detection of TIMP-2

Enzyme-linked immunosorbent assay (ELISA) studies were carried out to assess TIMP-2 expression in supernatants from endothelial cell lines, using the Biotrak ELISA kit (Amersham, UK) according to the manufacturer's instructions. No cross-reacting kits were available for detection of rat TIMP-1, MMP-2 or MMP-9.

Immunocytochemical analysis of MMP9 effects on brain endothelial ZO-1 expression

GP8/3.9 cells were grown to confluence on collagen-coated plastic eight-well chamber slides (Lab-Tek/Amersham Life Science, UK). Duplicate wells were stimulated with a total volume of 200 μl of serum-free medium containing 10 or 100 ng/ml human recombinant active MMP-9 (Oncogene, UK) for 30 min, 2, 4 or 24 h. Duplicate wells were also incubated with medium containing either 1 μM dexamethasone or 10 μM broad spectrum MMP inhibitor, KB8031 (Pharmingen, UK), for 24 h. Cells incubated in the presence of serum-free medium alone served as a control. Cells were first washed in PBS (phosphate-buffered saline) and then fixed with 4% paraformaldehyde for 20 min. After subsequent permeabilization with 0.25% Triton X-100 for 10 min and blocking for 30 min with PBS/10% foetal calf serum, cells were then incubated with a polyclonal rabbit anti-human ZO-1 (Zymed, Cambridge Bioscience, UK) (1 : 120 dilution) antibody overnight at 4°C. Cell monolayers were then re-blocked with PBS containing 10% foetal calf serum and 10% goat serum for 15 min prior to the addition of FITC (fluorescein isothiocyanate)-labelled goat anti-rabbit monoclonal antibody (1 : 50 dilution) (Jackson Immuno Research Laboratories Inc., Pa., USA) for 1 h in the dark. After washing three times with PBS between each step, coverslips were mounted with Vectashield (Vector Labs, UK). Negative controls included omission of the primary antibody or application of an irrelevant antibody. Slides were viewed and photographed on a Leica confocal fluorescence microscope.


Basal expression of MMPs by cultured CNS and non-CNS endothelial cells

Primary brain endothelial cells were found to express both MMP-2 and MMP-9 under resting conditions when cultured in serum-free medium for 24 h. Zymography of cell culture supernatants revealed bands which correspond to proMMP-2 (72 kDa), active MMP-2 (66 kDa), proMMP-9 (92 kDa) and active MMP-9 (84 kDa) (Fig 1). In addition, reverse transcription–PCR of mRNA extracted from GP8/3.9 cells showed expression of mRNA for both MMP-2 and MMP-9 under basal conditions (Fig. 2A), although the level of expression was low, as mRNA was not revealed by the less sensitive Northern blot analysis (Fig. 2B). The brain endothelial cell line GP8/3.9 (Fig. 7) and the retinal endothelial cell line JG2/1 also expressed MMP-2 and MMP-9 in quantities similar to primary cultures. The activity of MMP-2 was found to be much greater than the activity of MMP-9 in both of the CNS cell lines and in primary brain endothelial cells grown under identical conditions. It was confirmed that enzyme activity was the result of MMP activity by the inhibition of lysis zones following overnight co-incubation with EDTA (data not shown). AEC and HEV cells also expressed MMP-2 and MMP-9 under basal conditions (Figs 3 and 4). Quantitative differences in MMP expression were seen in endothelial cells from different tissue sites, with AEC expressing the greatest and CNS endothelial cells the lowest basal MMP-2 levels. The level of MMP-2 expression differed significantly between AEC and GP8/3.9 endothelial cells (P < 0.02) and between HEV and GP8/3.9 endothelial cells (P < 0.01) as assessed by the paired Student's t-test on data obtained from densitometric quantitation of zymograms (Fig. 5). The overall rank order of MMP-2 activity was AEC>HEV>GP8/3.9, although there was no significant difference in the levels of expression between AEC and HEV. It was confirmed that the differences were due to functional differences and not to differences in the total protein content of the supernatant, as bicichoninic acid protein analysis and SDS–PAGE did not demonstrate any significant differences in protein content between the supernatants derived from different endothelial cell types (data not shown). Unlike MMP-2 expression, no significant differences in basal MMP-9 expression were observed between the three cell lines (Fig. 5).

Fig. 1

Basal and cytokine-stimulated production of MMP-2 and MMP-9 by primary cultures of rat brain microvascular endothelium. (A) Gelatin zymogram of endothelial cell culture supernatants from brain showing proMMP-2 (72 kDa), active MMP-2 (66 kDa), proMMP-9 (92 kDa) and active MMP-9 (84 kDa). Lane 1, control; lane 2, cells treated for 24 h with TNF-α (10 ng/ml); lane 3, cells treated for 24 h with IL-1β (10 ng/ml); lane 4, cells treated for 24 h with IFN-γ (10 ng/ml). (B) Histogram representing the mean peak densitometric analysis of MMP-2 (black bars) and MMP-9 (shaded bars) activity from primary cultures of rat brain microvascular endothelium treated with cytokines as described above. Results are expressed as the mean percentage compared with basal levels ± SEM of n = 4 experiments. Significant difference from control (*P < 0.01; **P < 0.005).

Fig. 2

Analysis of MMP mRNAs from cell cultures. (A) Detection of MMP-2 and MMP-9 cDNA fragments by reverse transcription–PCR. A 1 μg aliquot of GP8/3.9 total RNA was primed with oligo(dT)15–18 and reverse transcribed with AMV reverse transcriptase. Single-stranded cDNA was amplified subsequently with oligonucleotide primer pairs specific to MMP-9 and MMP-2. (B) Effect of dexamethasone (Dex) on control and cytokine-stimulated MMP-9 mRNA expression in AEC, GP8/3.9 and HEV cells, assessed by Northern blot analysis.

Fig. 3

Basal and 24 h cytokine-stimulated production of MMP-2 and MMP-9 by AEC. (A) Gelatin zymogram of AEC supernatants stimulated with cytokines (10 ng/ml) for 24 h. Lane 1, control unstimulated; lane 2, with TNF-α; lane 3, with IL-1β; lane 4, with IFN-γ. (B) Histogram representing the mean peak densitometric analysis of MMP-2 (black bars) and MMP-9 (shaded bars) activity from AEC treated with cytokines as described above. Results are expressed as the mean percentage compared with basal levels ± SEM of n = 5 experiments. Significant difference from control (*P < 0.05).

Fig. 4

Basal and cytokine-stimulated production of MMP-2 and MMP-9 by HEV endothelial cells. (A) Gelatin zymogram of HEV supernatants stimulated with cytokines (10 ng/ml) for 24 h. Lane 1, control unstimulated; lane 2, with TNF-α; lane 3, with IL-1β; lane 4, with IFN-γ. (B) Histogram representing the mean peak densitometric analysis of MMP-2 (black bars) and MMP-9 (shaded bars) activity from HEV treated with cytokines as described above. Results are expressed as the mean ± SEM of n = 5 experiments.

Fig. 5

Comparative basal expression of MMP-2 and MMP-9 by CNS- and non-CNS-derived vascular endothelium. (A) Gelatin zymogram of cell supernatants. Lanes 1–3, GP8/3.9 cells; lanes 4–6, HEV cells; lanes 7–9, AEC. (B) Histogram representing the mean peak densitometric anlysis (± SEM) of MMP-2 (black bars) and MMP-9 (shaded bars) activity of n = 3 experiments. Significant difference from GP8/3.9 (*P < 0.01; **P < 0.02).

Comparison of MMP activity in cytokine-activated endothelium derived from different tissues

In primary CNS vascular endothelial cells, significant, dose-responsive upregulation of MMP-9 activity was observed following stimulation with TNF-α and IL-1β for 24 h (Fig. 6), as assessed by densitometric analysis of the zymograms (TNF-α, P < 0.005; IL-1β, P < 0.01) (Fig. 1). Similar significant increases from control were observed in GP8/3.9 (TNF-α, P < 0.005; IL-1β, P < 0.01) (Fig. 7) and JG2/1 cells (TNF-α, P < 0.005; IL-1β, P < 0.02) (Fig. 8). Using Northern analysis, these differences were also confirmed at the mRNA level for GP8/3.9 (Fig. 2B) and JG2/1 (data not shown) endothelial cells. Increases in MMP activities were not due to induction of cell proliferation, particularly in the faster growing cell lines, since MTT cell proliferation assays did not demonstrate any significant differences between control and cytokine-treated cells (data not shown). Unlike TNF-α and IL-1β, IFN-γ did not alter the level of MMP-9 activity as assessed by zymography (Figs 1, 7 and 8) and mRNA analysis (Fig. 2B). None of the cytokines tested had any significant effect on MMP-2 activity as indicated by consistent band sizes after zymographic analysis (Figs 1, 7 and 8). AEC did not show any alteration in MMP-9 activity following stimulation with TNF-α or IFN-γ, but did show a significant upregulation of MMP-9 activity following IL-1β treatment (P < 0.05) (Fig. 3). However, none of the cytokines tested had a significant effect on AEC MMP-2 activity (Fig. 3). HEV endothelial cells did not show any alteration in MMP-2 or MMP-9 activity following cytokine stimulation for 24 h (Fig. 4), illustrating a functional difference between the different endothelial cell populations studied.

Fig. 6

Dose-dependent upregulation of MMP-9 in GP8/3.9 rat brain microvascular endothelium following stimulation with TNF-α and IL-1β. Gelatin zymogram of supernatants derived from GP8/3.9 brain endothelial cells. Lane 1, control culture; lanes 2–4, cells treated for 24 h with TNF-α at concentrations of 0.1, 1.0 and 10 ng/ml; lane 5, control culture; lanes 6–8, cells treated for 24 h with IL-1β at concentrations of 0.1, 1.0 and 10 ng/ml.

Fig. 7

The effect of dexamethasone (Dex) on basal and cytokine-stimulated MMP-2 and MMP-9 expression by GP8/3.9 cells. (A) Gelatin zymogram of cell supernatants. Lane 1, serum-free medium; lane 2, dexamethasone (1 μM) for 24 h; lane 3, TNF-α (10 ng/ml) for 24 h; lane 4, TNF-α (10 ng/ml) and dexamethasone (1 μM) for 24 h; lane 5, IL-1β (10 ng/ml) for 24 h; lane 6, IL-1β (10 ng/ml) and dexamethasone (1 μM) for 24 h; lane 7, IFN-γ (10 ng/ml) for 24 h; lane 8, IFN-γ (10 ng/ml) and dexamethasone (1 μM) for 24 h. (B) Histogram representing the mean peak densitometric anlysis of MMP-2 (black bars) and MMP-9 (shaded bars) activity. Results are expressed as the mean ± SEM percentage compared with basal levels of n = 4 experiments. Significant differences from control (*P < 0.01; **P < 0.005; ***P < 0.001) and from cytokine treatment alone (P < 0.05; ††P < 0.005).

Fig. 8

Basal and cytokine-stimulated production of MMP-2 and MMP-9 by JG2/1 cells. Histogram representing the mean peak densitometric anlysis of zymograms for MMP-2 (black bars) and MMP-9 (shaded bars) activity. JG2/1 cells treated for 24 h with TNF-α (10 ng/ml), IL-1β (10 ng/ml) and IFN-γ (10 ng/ml). Results are expressed as the mean percentage compared with basal levels ± SEM of n = 5 experiments. Significant difference from control (*P < 0.02; **P < 0.005).

Effect of dexamethasone on MMP production in CNS endothelium

The effect of dexamethasone on MMP-2 and MMP-9 was studied in the CNS endothelial cell lines GP8/3.9 and JG2/1. Dexamethasone (1 μM) caused significant inhibition of constitutive MMP-2 activity (P < 0.001) compared with untreated cells and inhibition of MMP-2 production in the presence of TNF-α (P < 0.005), IL-1β (P < 0.05) and IFN-γ (P < 0.05). A significant downregulation of cytokine-induced MMP-9 activity was also observed in both brain GP8/3.9 cells (Fig. 7) and retinal JG2/1 cells (data not shown) following incubation with dexamethasone for 24 h. Zymographic analysis showed that dexamethasone inhibited MMP-9 expression in TNF-α- and IL-1β-stimulated GP8/3.9 endothelial cells by 62% (P < 0.05) and 80% (P < 0.05), respectively (Fig. 7). In addition, dexamethasone also significantly reduced both MMP-2 (P < 0.05) and MMP-9 (P < 0.005) in those cells activated with IFN-γ. These findings were also confirmed by Northern blot analysis on GP8/3.9 cells (Fig. 2B) and JG2/1 retinal endothelial cells (data not shown). These findings were not due to inhibition of cellular proliferation, as MTT assays did not show any significant difference in cellular proliferation between cytokine- and dexamethasone-treated CNS endothelial cells at 24 h (data not shown).

TIMP-2 expression by vascular endothelial cells

All the endothelial cells studied expressed TIMP-2 constitutively as assessed by ELISA. Levels within each group were not significantly altered following stimulation with TNF-α, IL-1β or IFN-γ (Table 1). TIMP-2 levels expressed by primary brain endothelium were not significantly different from either the GP8/3.9 brain endothelial cell line or the JG2/1 retinal endothelial cell line. However, TIMP-2 expression in both AEC (P < 0.05) and HEV (P < 0.005) was significantly greater than in primary brain endothelium, with HEV TIMP-2 levels being greater than those of AEC (P < 0.05). Samples were assayed in duplicate, and significant differences from control values were calculated by one-way ANOVA (analysis of variance) on each individual experiment followed by a multiple range test.

View this table:
Table 1

TIMP-2 levels in supernatants from CNS and non-CNS endothelial cell cultures

HEVAECPrimary brain endotheliumJG2/1GP8/3.9
Mean TIMP-2 levels (ng/ml) in supernatants from CNS (primary brain endothelium, GP8/3.9 brain endothelial cell line and JG2/1 retinal endothelial cell line) and non-CNS [high endothelial venule (HEV) and aortic (AEC)] vascular endothelial cells (n = 3). Results are expressed as mean ± SEM. Significant difference from AEC (*P < 0.05) and from primary brain endothelium (P < 0.05; P < 0.005). No significant difference was found between the CNS endothelial cell groups as assessed by ANOVA.
Control61.80 ± 3.37*40.39 ± 0.8521.61 ± 4.7129.65 ± 5.0125.31 ± 3.41
TNF-α74.62 ± 11.44*42.83 ± 10.2326.13 ± 1.9829.22 ± 10.9619.69 ± 2.39
IL-1β67.29 ± 7.07*32.69 ± 7.1220.82 ± 2.8927.36 ± 11.8823.47 ± 3.00
IFN-γ64.60 ± 2.73*39.17 ± 2.5923.39 ± 3.3933.17 ± 12.221.96 ± 1.07

ZO-1 expression following treatment of GP8/3.9 cells with MMP9

To evaluate whether MMP-9 alters the junctional complex, the level of expression and distribution of the junctional protein ZO-1 were assessed by immunocytochemical analysis on cells that had been cultured in the presence of active enzyme. The brain endothelial cell line (GP8/3.9) expressed the tight junction protein ZO-1 in the presence of serum-free medium (Fig. 9). The pattern of expression was junctional, forming a fine near-continuous line at the point of cell–cell contact. No gross alterations in ZO-1 staining were seen following incubation with active MMP-9 at 10 or 100 ng/ml up to 24 h. However, subtle but distinct and consistent changes in the level of expression were clearly observed. After treatment with 100 ng/ml active MMP-9 for 30 min (Fig. 9B), there was evidence of a small increase in the number of discontinuities in ZO-1 junctional expression. More striking than this was the consistent finding that after 1 h (Fig. 9C) there was an overall decrease in the intensity of expression which continued through to 6 h post-treatment (Fig. 9D). Following co-incubation with dexamethasone (1 μM) or the matrix metalloproteinase inhibitor (KB8301) for 24 h, no alterations in ZO-1 expression were observed (data not shown).

Fig. 9

Projected confocal immunofluorescent images of ZO-1 expression in control and MMP-9-treated GP8/3.9 brain endothelial cells. (A) Control monolayer, (B) 30 min of treatment with 100 ng/ml active MMP-9, (C) 1 h of treatment with 100 ng/ml active MMP-9, (D) 6 h of treatment with 100 ng/ml active MMP-9. All images are shown at the same magnification (bar = 60 μm).


The role of MMPs in CNS inflammation is well established, but the origin of their production is still the subject of considerable investigation. In the present study, we investigated the production of MMPs by CNS microvascular endothelium and have demonstrated that both MMP-2 and MMP-9 are expressed constitutively by these cells in vitro. Following stimulation of CNS vascular endothelial cells with the proinflammatory cytokines TNF-α and IL-1β, we also observed differential upregulation of MMP-9 expression. The production of MMPs by CNS vascular endothelial cells raises the possibility that these cells may be an important source of proteases, especially with regard to disruption of the blood–brain barrier. Although endothelial cells may not be the major source of production, because of their location they may play a critical role in the development of neuroinflammatory diseases, especially during the early stages prior to large-scale recruitment of leucocytes.

The production of MMP-9 in the CNS is clearly important in the pathogenesis of neuroinflammatory diseases. Histologically, MMP-9 immunoreactivity has been detected at low levels in blood vessels from normal human brain tissue, and in sections from actively demyelinating multiple sclerosis lesions a significantly greater number of positively stained vessels is observed (Cossins et al., 1997). Studies of experimental autoimmune encephalomyelitis, an animal model of multiple sclerosis, and experimental autoimmune neuritis, an animal model of Guillain–Barré syndrome, associated with breakdown of the blood–nerve barrier, have also shown that MMP-9 expression peaks at the time of maximum disease severity and that the endothelial cells appear to be a major early source of the enzyme in these models (Hughes et al., 1998; Kieseier et al., 1998). Our in vitro findings with CNS endothelium would therefore support a role for endothelial-derived MMP-9 in blood–brain barrier breakdown.

Steroids, such as dexamethasone, are used frequently in the clinical management of conditions associated with blood–brain barrier dysfunction. In this study, we investigated the effect of dexamethasone on CNS endothelial cell MMP production and found that it attenuated the cytokine-induced production of MMPs. The mechanism by which steroids affect blood–brain barrier permeability is poorly understood but their effect on MMP production may be one possible mechanism of action. These results are consistent with findings in patients with multiple sclerosis, where levels of MMP-9 in the CSF and number of gadolinium-enhancing lesions on MRI were reduced significantly following treatment with intravenous methyl prednisolone (Rosenberg et al., 1996). Inhibition of blood–brain barrier disruption in experimental allergic encephalomyelitis has also been demonstrated experimentally following administration of dexamethasone (Paul and Bolton, 1995).

The barrier function of the CNS endothelium is dependent on the integrity of tight junctions, which are a complex of intracellular proteins (Staddon and Rubin, 1996) including ZO-1 (Watson et al., 1991) and ZO-2 in association with the membrane-spanning proteins occludin (McCarthy et al., 1996) and claudin (Furuse et al., 1998). It is conceivable, therefore, that MMPs may have a direct effect either on the extracellular domain of the tight junctional proteins of the blood–brain barrier or, more likely, on the basement membrane, both of which may lead to barrier disruption. Loss or reduction of the expression of ZO-1 protein is associated with blood–CNS barrier breakdown in vitro (Gardner et al., 1997) as well as in vivo (Bolton et al., 1998). Treatment of GP8/3.9 endothelial cells with MMP-9 did not bring about large-scale disruption of ZO-1 expression. However, the expression of ZO-1 following treatment was clearly altered, with an increase in the frequency of discontinuities, which is known to correlate with increased permeability (Schulze et al., 1997), and a consistent decrease in the intensity of expression. Enzymatic digestion of endothelial cell basement membrane could lead to dysfunction of the integrin-dependent attachment to the extracellular matrix, to subsequent changes in intracellular signalling (Boudreau and Jones, 1999; Dedhar, 1999) and to alterations in junctional integrity.

It is now fully recognized that vascular endothelial cells from different tissues vary phenotypically. Thus, data derived from endothelial cells outside the CNS are not necessarily representative of what occurs within the CNS. Because of known differences in MMP expression between endothelial cell populations, we compared MMP production from brain and retinal endothelium with that of large vessel endothelium (aortic) and the specialized endothelium derived from HEV. We did not observe any significant differences between endothelial cells derived from the brain and retina, which confirms previous data illustrating their functional similarity (Greenwood, 1992). However, our data did show differences in both basal MMP expression and the responses to cytokine stimulation between CNS and non-CNS endothelial cell populations, with CNS endothelial cells being more susceptible to cytokine stimulation under the conditions studied. Our findings in rat aortic endothelial cells were consistent with previously published data in bovine (Unemori et al., 1989) and monkey aortic endothelium (Cottam et al., 1996), specifically in the selective upregulation of MMP-9 following IL-1β stimulation. This heterogeneity in endothelial cell responses was noted previously by Hanemaaijer and colleagues (Hanemaaijer et al., 1993). The differences do not reflect solely differences between macrovascular and microvascular endothelium, however, as Hanemaaijer and colleagues did not see MMP-9 upregulation in foreskin microvascular endothelium, following TNF-α stimulation, which differs from our findings in CNS microvascular endothelium.

All endothelial cells examined expressed TIMP-2 constitutively and did not show any significant alteration following cytokine stimulation. Our findings are in accordance with studies in the mouse CNS where high constitutive TIMP-2 expression was reported with no significant alteration in a variety of inflammatory conditions (Pagenstecher et al., 1998). The same group found that TIMP-1 levels were significantly upregulated in CNS inflammation. Unfortunately, we were unable to measure TIMP-1 levels in our system as no suitable antibodies or ELISA kits are available currently for the detection of rat TIMP-1. It is interesting to note, however, that the TIMP-2/MMP-2 profile in aortic endothelial cells appears to differ from that seen in HEV, i.e. the ratio of MMP-2 to TIMP-2 appears greater in aortic cells. This may reflect functional differences between macrovascular and microvascular endothelial cells. Furthermore, the level of TIMP-2 expression in HEV and AEC was significantly greater than for the CNS-derived endothelia, which also demonstrates functional differences between the different vascular endothelia.

Previously published studies demonstrate heterogeneity in the proteolytic profile of different CNS cell types both under basal conditions and in response to the proinflammatory cytokines. Microglia have been shown, like endothelial cells, to constitutively express MMP-2 and -9 in vitro (Colton et al., 1993; Gottschall and Deb, 1996; Cross and Woodroofe, 1999), whereas astrocytes only express MMP-2 under basal conditions (Colton et al., 1993; Wells et al., 1996). All three cytokines, TNF-α, IL-1β and IFN-γ, have been shown to upregulate both MMP-2 and MMP-9, in adult rat microglia grown under identical conditions to those used in this study (Cross and Woodroofe, 1999), thus highlighting functional cellular heterogeneity in response to cytokines within the CNS. These patterns of expression are in contrast to our findings in this present study and suggest that the secreted MMP profile found in primary culture supernatants has not been influenced significantly by potential contaminant cells.

From these data, we propose that the CNS vascular endothelium plays an active part in the breakdown of the blood–brain and blood–retinal barriers through the altered expression of MMPs. The current evidence from animal and human studies would suggest that increased endothelial MMP-9 expression is most significant in the early stages of inflammation.


This work was generously supported in Sheffield by the Ryder Briggs Charitable Neurological Trust and in London by the Wellcome Trust and Multiple Sclerosis Society.


View Abstract