Monocyte-Astrocyte Networks Regulate Matrix Metalloproteinase Gene Expression and Secretion in Central Nervous System Tuberculosis In Vitro and In Vivo¹

Tuberculosis of the CNS (CNS-TB)³ is the most serious form of infection with Mycobacterium tuberculosis. Of those who receive treatment for this disease, more than half die or are disabled and if left untreated CNS-TB is fatal (1, 2). CNS-TB is characterized by severe tissue destruction associated with excessive inflammation (3). However, the mechanisms by which M. tuberculosis causes destruction of cerebral tissues are not clearly defined.

Matrix metalloproteinases (MMPs) are a family of zinc-containing enzymes which together can catabolize all components of the CNS tissue matrix (4, 5). MMP activity is regulated at a transcriptional level and by secretion of inactive zymogens requiring cleavage activation. Control of MMP activity in situ is regulated by tissue inhibitors of metalloproteinases (TIMPs) (6). Unopposed MMP activity is associated with cerebral injury in CNS diseases, including multiple sclerosis where high MMP-9 concentrations have been related to relapsing, active forms (7, 8). MMP-9 (92kDa gelatinase) degrades tenascins, fibronectin, and type IV collagen, which are critical in blood-brain barrier (BBB) function and CNS matrix composition (9). A role for MMP-9 activity in inflammatory tissue destruction is supported by studies in knockout mice in which BBB disruption following ischemia is reduced (10).

MMP-9 concentrations are increased in cerebrospinal fluid (CSF) from patients with CNS-TB (3, 11) and we showed that MMP-9 concentrations are associated with neurological complications and death (12). The effect of active tuberculous on CSF TIMP-1 concentrations is variable and may be increased or remain unchanged (11, 12). In our previous study, TIMP-1 CSF concentrations were not related to clinical signs of disease severity (12). The NF-κB-signaling pathway is implicated in differential regulation of TIMP-1 and MMP-9 (13) and is activated during host responses to M. tuberculosis in other tissues (14). Importantly, the MMP-9 promoter contains binding sites for NF-κB, whereas there are none in the TIMP-1 promoter (13, 15, 16).

Astrocytes are the most numerous cell population within the CNS, outnumbering neuronal cells by a factor of 10 (17). In the healthy brain, MMPs secreted by astrocytes are involved in angiogenesis, tissue remodeling, and neurite extension (18, 19). Due to their ubiquitous presence and capacity to secrete both MMPs and TIMPs, astrocytes have the potential to play a central role in tissue destructive processes. For example, astrocytes are a source of MMP-9 in ischemic injury (20).

In CNS-TB, peripheral mononuclear cells migrate into brain parenchyma, where they are involved in inflammation (21, 22). We investigate the hypothesis that cytokine networks between peripherally derived monocytes and astrocytes are critical in controlling MMP gene expression, secretion and activity. These networks are distinct from the intrinsic microglial immune network and have not been studied in this context before.

For preparation of zymogram gels, AccuGel 29:1 (30% acrylamide, 29:1 acrylamide:bis-acrylamide), ProtoGel stacking and ProtoGel running buffers were purchased from National Diagnostics. Triton X-100 was purchased from BDH. Coomassie blue tablets were obtained from Pharmacia Biotech. For standards in zymography, recombinant MMP-1, -2, and -9 were purchased from Oncogene. Twelve percent casein minigels were purchased from Invitrogen Life Technologies. For Western blotting, sheep anti-human pro-MMP-9,-1 Ab and peroxidase-conjugated donkey anti-sheep IgG were purchased from The Binding Site; anti-human MMP-7 Ab was purchased from Merck. For the MMP-3 western blots rabbit anti-human Ab was purchased from Chemicon International and HRP-conjugated anti-rabbit Ab from Cell Signaling Technology. Rabbit anti-human p65, IκBa and IκBb Abs were purchased from Santa Cruz Biotechnology. The p65 alkylation-agent helenalin was purchased from BIOMOL. For immunohistochemistry, mouse monoclonal anti-human MMP-9 Ab (clone 15W2) and mouse monoclonal anti-human TIMP-1 Ab (clone 6f6a) were obtained from Novocastra. Polyclonal rabbit anti-human glial fibrillary acid protein (GFAP) Ab were purchased from DakoCytomation. All other reagents were purchased from Sigma-Aldrich.

M. tuberculosis H37-Rv was maintained in Middlebrook 7H9 medium supplemented with 10% albumin-dextrose-catalase enrichment medium, 0.2% glycerol, 0.02% Tween 80, and 2.5 μg/ml amphotericin. M. tuberculosis was used at mid-log growth phase at OD 0.60 (Biowave Cell Density Meter; WPA) in all experiments. M. tuberculosis endotoxin level was measured by the amoebocyte lysate assay (Associates of Cape Cod) and was <0.3 ng/ml LPS.

Human astrocytoma cell lines U373-MG and U87-MG (ECACC nos. 89081403 and 89081402 respectively) were maintained in Eagle’s MEM supplemented with 10% FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, 1% nonessential amino acids, and 100 μg/ml ampicillin according to the supplier’s instructions. All experiments were performed in serum-free medium before passage 15.

Primary human blood monocytes were prepared from single-donor buffy coat residues obtained from healthy donors. (National Blood Transfusion Service) by density gradient centrifugation (Ficoll Paque; GE) followed by adhesion purification for 1 h. Monocyte purity was >95% by FACS analysis (FACSCalibur; BD Biosciences). Monocytes were infected with M. tuberculosis at a multiplicity of infection (MOI) of 10 in RPMI 1640 with 2 mM glutamine. Conditioned medium was harvested at 24 h and M. tuberculosis was removed by filtration through a 0.2-μm Anopore membrane (23). Conditioned medium from infected monocytes was termed CoMTB and control medium from uninfected monocytes was CoMCon.

In direct infection experiments, U373-MG cells were incubated with M. tuberculosis at an MOI of 10 at 37°C for 72 h. Cell culture supernatants were harvested at this time and filtered to remove tubercle bacilli. For conditioned medium experiments, confluent U373-MG or U87-MG cells were stimulated with a 1/5 dilution of either CoMTB or CoMCon. Tissue culture medium was harvested at specified time points and centrifuged at 12,000 relative centrifugal force (rcf) to remove cellular debris and samples were frozen for later analysis.

Astrocytes were lysed with TRI Reagent (Sigma-Aldrich) and total RNA was extracted. One microgram of total RNA was reverse transcribed using 2 μg of random hexamers (Amersham Biosciences) and 200 U of Superscript II reverse transcriptase (Invitrogen Life Technologies), according to the supplier’s instructions. PCR were done on the ABI Prism 7700 (Applied Biosystems) according to previously described methods (24, 25), with each reaction containing 5 ng of reverse-transcribed RNA in 25 μl. Primer and probe sequences for the MMPs and TIMPs are described elsewhere (25). The cycle threshold (C_T) at which amplification entered the exponential phase was determined and this number was used as an indicator of the amount of target RNA in each sample; a lower C_T indicates a higher quantity of starting RNA. The C_T can be used to compare relative amounts of different transcripts at the RNA level, although this does not necessarily reflect protein levels. To accurately determine the quantitative change in RNA levels, standard curves were prepared by making 2-fold serial dilutions of one sample; these dilutions were subject to real-time PCR as above. Standard curves for C_T vs input RNA were prepared, and relative levels of starting RNA in each sample were determined. Changes of C_T from very low to very high expression are expressed on a 5-point scale which has been previously used. To account for differences in the amount of total RNA the results of each MMP member were normalized to 18S ribosomal RNA (primers and probes from Applied Biosystems) levels from the same sample.

MMP-9 and -2 activity was detected by gelatin zymography using standard methodology (26). In brief, standards and prepared cell supernatants were loaded with 5× loading buffer (0.25 M Tris (pH 6.8), 50% glycerol, and 5% SDS, bromphenol blue) and run on 11% acrylamide gels impregnated with 0.1% gelatin as substrate. After 3.5 h at 180 V (buffer 25 mM Tris, 190 mM glycine, and 0.1% SDS), the gel was renatured in 2.5% Triton X for 1 h with agitation. After two washes in collagenase buffer (55 mM Tris base, 200 mM sodium chloride, 5 mM calcium chloride, and 0.02% Brij (pH 7.6)), gels were incubated overnight in fresh collagenase buffer at 37°C. Gelatinolytic activity was detected using 0.02% Coomassie blue in 1:3:6 acetic acid: methanol: water.

MMP-1 and -7 activity was measured by casein zymography. Standards and prepared cell supernatants were mixed with 5× loading buffer (as for gelatin zymography) and were loaded onto casein minigels. Samples were separated by electrophoresis at 125 V for ∼2 h. After a 1 h wash in 2.5% Triton X, gels were rinsed in collagenase buffer (as for gelatin zymography), then equilibrated in fresh collagenase buffer for half an hour before a 40-h incubation in collagenase buffer at 37°C. Caseinolytic activity was revealed by staining for 1 h in 0.1% Coomassie blue followed by de-staining in 1:3:6 acetic acid: methanol: water for 1–2 h.

All experimental samples were run in parallel with 2 ng of rMMP to standardize between gels. Gel images were digitized with a Trans-illuminator (UVP) followed by proteolytic band quantification using LabWorks (version 4.5). The results of each sample were normalized to the standards.

Sandwich ELISAs were used to assay TIMP-1 and -2 secretion according to the manufacturer’s instructions (R&D Systems). The lower limit of detection was 31 pg/ml.

The NE-PER extraction kit (Pierce Biotechnology) was used to obtain nuclear and cytoplasmic extracts. Briefly, confluent cells were stimulated and incubated until the specified time point. Cells were scraped into ice-cold 1× PBS and spun at 100 rcf to produce a cell pellet. The cell pellet was resuspended in cold CER1 (with Halt protease inhibitors; Pierce Biotechnology). After an incubation of 10 min, CER2 was added to break down the cytoplasmic membrane. After centrifuging (16,000 rcf), the cytoplasmic extract was collected and frozen immediately. Nuclear Extract Reagent was added to the remaining nuclear pellet and after a 40-min incubation and centrifugation (16,000 rcf) the nuclear extract was harvested and frozen.

Western blotting was used to confirm MMP-9 secretion, to detect MMP-1, -3, and -7, to measure NF-κB nuclear translocation and to follow degradation of cytoplasmic IκB. After mixing 40 μl of prepared cell supernatants with 2× loading buffer (10% glycerol, 5% 2-ME, 2% SDS, 0.06 M Tris (pH 6.8), bromphenol blue), each sample was heat denatured and run on a 10% acrylamide gel at 200 V (running buffer 25 mM Tris base, 192 mM glycine, 0.1% SDS) for 3 h. After separation, proteins were transferred to a nitrocellulose membrane (GE) and blocked for 1 h with 5% milk protein/0.1% Tween 20. Then membranes were incubated with the primary Abs overnight at 4°C. The dilutions of the primary Abs were 1/1000 for MMP-1/-3/-7/-9, NF-κB (p65 subunit) and IκBα/IκBβ, respectively. After washing, the membrane was incubated with peroxidase-conjugate secondary Ab (1/1000 dilution, MMP-1/-3/-7/-9; 1/2000 dilution, p65, IκBα/IκBβ) for 1h. Protein bands were visualized on Hyperfilm ECL (GE) by chemiluminescence.

To investigate the activation of the multiple subunits of NF-κB a specific transcription factor assay (TransAM; Active Motif), which is five times more sensitive than EMSA, was performed. Nuclear extracts were added to a 96-well plate containing immobilized oligonucleotides encoding an NF-κB consensus site (5′-GGGACTTTCC-3′). Active NF-κB contained in the nuclear extract specifically bound to this oligonucleotide. The primary Abs used to detect p50, p52, p65, RelB, or RelC recognize an epitope accessible only when the active form of these factors is bound to its target DNA. An HRP-linked secondary Ab was added and the color change determined by spectrophotometry at 450 nm. Competition experiments demonstrated specificity of binding by adding 20 pM/well either wild-type or mutated NF-κB oligonucleotide before assaying with the p65 Ab.

To examine the spatial distribution of MMP-9 and TIMP-1 in infected and uninfected CNS tissue in vivo, immunohistochemistry for MMP-9, TIMP-1, and GFAP was performed from five patients with culture-proven M. tuberculosis infection and one noninfected control. Sections of 4-μm thickness were dewaxed and endogenous peroxidase activity was blocked with 0.6% hydrogen peroxide for 15 min. Sections were microwaved for 20 min in citrate buffer (0.01 M citrate (pH 6.0)) and blocked with 5% normal goat serum for 10 min. The primary Abs (MMP-9 at 1/1000, TIMP-1 at a 1/400 dilution and GFAP at a 1/500 dilution) were applied in 0.01 M PBS/azide/BSA for 1 h at room temperature. Ab was detected with the Menarini nonbiotinylated kit according to the manufacturer’s instructions. Peroxidase activity was developed with the 3,3′-diaminobenzidine system (Menarini). Slides were counterstained with Cole’s hematoxylin, dehydrated, and mounted. All experiments were performed with appropriate isotype-matched control Abs.

Data are presented as means ± SD of three samples and represent experiments performed in triplicate on at least two separate occasions, unless otherwise stated. Statistical analysis was performed using SPSS (version 13.0). Paired groups were compared with the Student t test. Multiple intervention experiments were compared with one-way ANOVA followed by Tukey’s multiple comparison. A p value of <0.05 was taken as statistically significant.

The expression of all known human MMPs in U373-MG cells in response to CoMTB and CoMCon was analyzed by real-time PCR. In astrocytes stimulated with CoMTB, expression of MMP-1, MMP-7, MMP-8, MMP-9, MMP-10, MMP-14, and MMP-19 was significantly up-regulated by 24 h (Fig. 1). In addition, expression of MMP-2, MMP-3, and MMP-12 were increased by 48 h (Table I), whereas MMP-16 and MMP-28 mRNA were increased at 72 h. MMP-17 expression was significantly decreased by CoMTB at 24 h and this inhibitory effect persisted over the entire 72 h.

The data in Fig. 1 and Table I show changes in MMP mRNA normalized to 18S rRNA and indicate relative not absolute levels of gene expression in astrocytes. C_T data was used to compare levels of different genes and to assess the abundance of each gene, where a low C_T value demonstrates high level gene expression (Fig. 2). Of the MMP mRNAs increased by CoMTB as opposed to CoMCon, MMP-1, -3, -7, and -9 are the most abundantly expressed, with CoMTB increasing MMP-9 levels from a low expression level to a high level compared with the effect of CoMCon. CoMTB also increases MMP-1 and MMP-3 expression from moderate to high levels at 24 h. Elevated MMP-1 mRNA accumulation persists at 48 and 72 h. MMP-2 and MMP-14 were expressed at constitutively very high levels. High MMP-7 expression was induced by both CoMCon and CoMTB at 24 h although higher expression was detectable in CoMTB-stimulated cells at 48 h. Other astrocyte MMPs up-regulated by CoMTB were only expressed at a low to moderate levels (Table I).

Considering the changes in RNA expression together with the more quantitative C_T data on the magnitude of increased gene expression, the initial findings are that MMP-1, MMP-3, MMP-7, and MMP-9 are the principal astrocyte MMPs specifically up-regulated to high levels in response to CoMTB.

On the basis of mRNA expression data, the secretion of MMP-1, MMP-2, MMP-3, MMP-7, and MMP-9 were investigated further. Kinetics studies showed that by 24 h, CoMTB induced astrocyte MMP-9 secretion. MMP-9 concentrations increased up to 72 h, after which they stabilized (Fig. 3,A). Incubation of the zymogram in 10 mM EDTA abolished MMP-9 bands (data not shown) which together with western analysis (Fig. 3,A) confirmed that enzymatic activity was due to MMP-9. MMP-9 secretion was undetectable in cell supernatants taken from CoMCon stimulated cells at all time-points. Low levels of MMP-9 are present in CoMTB which accounts for the MMP-9 secretion observed at time = 0 h. Levels of MMP-9 secretion in response to CoMTB stimulation was similar in both U373-MG and U87-MG astrocytes (data not shown). In contrast to stimulation with CoMTB, stimulation of astrocytes by M. tuberculosis at MOI 0.1–10 did not induce the secretion of MMP-9 (Fig. 3 B).

MMP-2 (72 kDa), also detected on gelatin zymography, was secreted constitutively by astrocytes. No significant difference in MMP-2 secretion was observed between CoMCon and CoMTB-stimulated cells (Fig. 3,C). This was consistent with gene expression data. Neither MMP-1 (52 kDa) nor MMP-7 (28 kDa) secretion from astrocytes was detectable, on Western analysis (data not shown) or casein zymography (Fig. 3 C). MMP-3 (52 kDa) secretion was not detectable using ELISA or Western blotting, these data contrast with mRNA expression findings for these MMPs.

As the net proteolytic activity is determined by the balance between MMPs and TIMPs, gene expression of the four human TIMPs was analyzed (Fig. 4 A). No significant change in gene expression levels were observed for TIMP-1 and TIMP-4 in CoMTB and CoMCon-stimulated astrocytes at any time point. TIMP-3 expression was significantly up-regulated by CoMTB at 72 h. Conversely, TIMP-2 expression was significantly down-regulated by CoMTB at 72 h. The C_T values show that all four TIMPs are expressed at constitutively high or very high levels in astrocytes (data not shown). TIMP-1 and TIMP-2 are potent inhibitors of MMP-9 activity and since these TIMPs are very highly expressed in astrocytes stimulated with CoMTB, analysis of their secretion was investigated.

Kinetics studies showed TIMP-1 secretion from both CoMTB and CoMCon-stimulated cells (Fig. 4 B). Secretion of TIMP-1 was ∼2-fold higher in CoMCon and CoMTB-stimulated cells than in control cells at 120 h. In contrast, there was a 3-fold increase in TIMP-2 secretion in CoMTB-stimulated cells when compared with CoMCon-stimulated or unstimulated cells at 120 h; a divergence that was apparent by 24 h.

MMP-9 but not TIMP-1 or -2 contains a binding site for NF-κB in its promoter which may allow differential regulation of gene expression. To investigate whether increased astrocyte MMP-9 secretion in response to CoMTB is mediated by NF-κB, the DNA-binding activity of the NF-κB subunits, p65, p52, p50, RelB, and RelC was examined. In human astrocytes (Fig. 5,A), CoMTB induced a 6-fold increase in the levels of active p65 in astrocyte nuclear extract in comparison to CoMCon-stimulated and unstimulated cells (p < 0.01). In addition, a 2-fold increase in p50 activity was observed in CoMTB-stimulated cells (p < 0.05). DNA-binding activity of p52, RelB, and RelC subunits was unaffected by stimulation with CoMTB. Western analysis showed that the NF-κB p65 subunit was translocated to the nucleus within 10 min of stimulation with CoMTB (data not shown). Nuclear NF-κB p65 was persistently higher in CoMTB-stimulated cells up to 24 h (Fig. 5,B). Correspondingly, kinetic analysis of the nuclear activation of the p65 subunit showed that astrocyte p65 is rapidly activated after stimulation with CoMTB by 30 min. After 2 h, the level was 19-fold higher in CoMTB-stimulated cells than in CoMCon-stimulated cells (Fig. 5 C).

NF-κB activity is normally regulated by IκBα/IκBβ. IκBα was degraded in astrocytes within 10 min of stimulation by CoMTB (Fig. 5 D). IκBα remained absent from the cytoplasm up to 2 h. IκBβ degradation was a later event not occurring until 30 min after stimulation with CoMTB. IκBα levels had returned to baseline by 24 h while IκBβ remained degraded at 24 h (data not shown). CoMCon did not cause degradation of IκBα or IκBβ at any time point.

To investigate the functional importance of the p65 subunit of NF-κB in CoMTB-initiated up-regulation of astrocyte MMP-9 secretion, experiments were performed after pretreatment with helenalin which specifically blocks p65 activity via an irreversible alkylation (27). Preincubation of astrocytes for 2 h with 2 μM helenalin resulted in a 9-fold decrease in MMP-9 secretion (p < 0.05) in CoMTB-stimulated astrocytes to near control concentrations (Fig. 5,E). No effect on cell viability was found observed. In addition, 1 μM helenalin resulted in a 4-fold reduction in MMP-9 secretion from 367.4 ± 72.1 to 94.9 ± 33.4 (p < 0.05) (Fig. 5,E) and caused a >50% decrease in DNA-binding activity of p65 (Fig. 5 F).

Dexamethasone treatment during CNS-TB is associated with reduced mortality and this effect is not thought to be due to general immunosuppression (28). Dexamethasone (0.1 μM) inhibited MMP-9 secretion from CoMTB-stimulated astrocytes by 49 ± 25% (p < 0.01; Fig. 6,A). Incubation with 10 μM dexamethasone inhibited astrocyte MMP-9 secretion in response to CoMTB by 62 ± 6%. In contrast, even at maximal concentrations dexamethasone had no effect on TIMP-1 or TIMP-2 secretion (Fig. 6, B and C). These data suggests that dexamethasone may attenuate the tissue destructive potential of CoMTB-stimulated astrocytes.

Immunohistochemical analysis comparing the distribution of MMP-9 and TIMP-1 expression in brain biopsies from patients with CNS-TB was performed and compared with brain tissue from noninfected patients (Fig. 7). MMP-9 was expressed in all astrocytes in CNS-TB tissue at high levels, this contrasted to the situation in patients without CNS-TB where MMP-9 was expressed by astrocytes at very low levels. These findings are consistent with our in vitro data. Interestingly, TIMP-1 expression was down-regulated in tissue from CNS-TB patients in comparison to noninfected controls. This contrasts with our in vitro data where, TIMP-1 was not affected or increased by CoMTB.

Astrocytes were the cell type that most commonly stained positive for MMP-9. Tissue-resident macrophages and microglia are seen to express high levels of MMP-9, but these are present in the tissues at much lower numbers than the astrocytes (data not shown). GFAP staining confirmed the presence of astrocytes in both control and infected tissues.

There is emerging evidence for the involvement of MMP activity in pathological tissue destruction in CNS diseases including MS, cancer, and tuberculous meningitis (29, 30, 31). Although astrocytes may be an important source of cytokines, chemokines, and reactive nitrogen species in inflammatory CNS disease (32, 33), this is the first study to demonstrate that astrocytes are an important source of MMP-9 in CNS TB. MMP-9 secretion and expression from astrocytes is induced by M. tuberculosis through a monocyte-dependent network. In contrast, this network does not affect astrocyte TIMP-1 gene expression and secretion. We found increased astrocyte MMP-9 secretion in CNS tissue from patients compared with controls; levels of TIMP-1 did not change in patients. M. tuberculosis-dependent up-regulation of MMP-9 from astrocytes is antagonized by the corticosteroid dexamethasone which is used to limit tissue destruction in patients. NF-κB particularly the p65 subunit has a key role in controlling unopposed increased MMP-9 secretion.

Analysis of mRNA expression of all human MMPs revealed that CoMTB up-regulated MMP-1, -3, -7, and 9 most significantly. Gene expression of a range of other MMPs was up-regulated to a lesser extent. MMP-9 secretion was up-regulated in CoMTB but not CoMCon-stimulated U373-MG and U87-MG human astrocytic cells. MMP-9 secretion does not increase after 72 h, which is consistent with mRNA data which showed MMP-9 gene expression stabilizing after 48 h. MMP-9 is not usually detectable in the CSF of healthy individuals. However, it does appear in a range of inflammatory diseases of the CNS where it is thought to contribute to damage of the parenchymal tissues and the BBB (10, 34). In tuberculous meningitis patients, CSF MMP-9 concentrations are significantly associated with mortality and local tissue damage (12). Astrocytes did not secrete MMP-9 after direct infection with M. tuberculosis. This may be partly due to the fact that astrocyte TLR expression is limited, expressing primarily TLR2 and TLR3 in the adult CNS (35). Although TLR2 is able to recognize mycobacterial components when in association with the TLR1/TLR6 complex, there is no evidence for either TLR2 or TLR3 doing this in isolation (36). This indicates that immune networks are required to stimulate the astrocyte MMP-9 response to M. tuberculosis. The exact mediators of the astrocyte response to CoMTB are unclear; preliminary data indicate that the situation is complex and this is subject of ongoing research.

The MMP expression profile of CoMTB-stimulated astrocytes changes over time, with some MMPs including MMP-7, MMP-9, MMP-14, and MMP-19 being expressed at increased levels at 24 h but not at 72 h. In contrast, MMP-2, MMP-3, and MMP-12 are not expressed until 48 h whereas MMP-16 and MMP-28 are not up-regulated until 72 h. This could be due to these genes being late expressed or due to mediators secreted by astrocytes in response to CoMTB stimulating their secretion as part of a secondary autocrine response. MMP-1, -3, and -7 secretion was not detectable, in contrast to the increased levels of gene expression induced in response to CoMTB. Possibly, additional stimuli are required to initiate secretion of these MMPs. Alternatively, it may be that these MMPs are in a cell associated form although for MMP-1 this would be somewhat unusual. Astrocyte MMP-1 secretion has not been reported to our knowledge and given increasing evidence that MMP-1 is directly toxic to neuronal cells, CNS-resident cells might be blocked from secreting it (37, 38). These data underline the importance of confirming expression data with secretion analysis.

MMP-17 was the only MMP persistently down-regulated by CoMTB. MMP-17, also called membrane-type-4 MMP, shares <40% sequence homology with the other four membrane-type MMPs, and does not possess a cytoplasmic domain (39). The biological function of MMP-17 is currently unclear.

Immunohistochemical analysis showed that normal CNS astrocytes express very low levels of MMP-9. Such MMP-9 may be involved in physiological processes such as tissue remodeling. In contrast, in CNS-TB tissue, all astrocytes stain for MMP-9 at significantly higher levels than in control tissue. Astrocytes were the cell type that most commonly stained positive for MMP-9. Although individual macrophages and microglia may express higher levels of MMP-9 than astrocytes, they are present in tissues at lower numbers.

The functional cytokine network postulated in this study is directed by peripheral monocytes and distinct from the microglial immune network (40). However, other glial cells are likely to play a key role in development of dysregulated astrocyte MMP activity. In particular, the microglial cytokine response to M. tuberculosis is likely to be similar to that of the monocyte (41). Thus, microglial-derived cytokines may also play a role in the activation of MMP-9 secretion from astrocytes in vivo.

The net proteolytic activity of astrocyte secretions depends upon the balance between TIMPs and MMPs. TIMP-1, TIMP-2, and TIMP-4 expression was constitutive and not up-regulated by CoMTB. TIMP-3 expression was down-regulated by CoMTB at 24 h and up-regulated at 72 h. It is possible that a feedback loop induces delayed TIMP-3 expression. TIMP-1 and TIMP-2 were expressed at higher levels than TIMP-3 and TIMP-4. As TIMP-1/2 are major inhibitors of MMP-9 (42), these were examined further. TIMP-1 secretion was constitutive and there was no significant difference between CoMCon- and CoMTB-stimulated cells. This contrasted to our in vivo data from CNS-TB patient samples in which astrocyte TIMP-1 expression was decreased compared with control tissue. The reason for this discrepancy is not clear although interestingly, reduced TIMP-1 expression is also observed in neuronal cells in CNS-TB samples. It is possible that the Ab used for immunohistochemistry does not detect bound forms of TIMP-1 as well as the Abs used in the TIMP-1 ELISA which would provide a technical explanation for the discrepancy. Alternatively, the finding may be real with the complex situation in patients with established disease resulting in decreased TIMP-1 compared with that found in cellular studies of relatively short duration. Overall the present study demonstrates that TIMP-1 concentrations in infected astrocytes are not elevated compared with those in controls and will not result in reduced proteolytic activity. This finding is consistent with our data showing that CSF levels of TIMP-1 were not increased in CNS TB patients (12). However, TIMP-2 secretion although not gene expression, was increased by CoMTB. TIMP-2 has complex functions and it is impossible to predict the effect of this on net proteolytic activity in vivo. Besides inhibiting MMP-2, at high concentrations TIMP-2 forms a complex with MMP-14 which may activate MMP-2 (43).

The MMP-9 but not the TIMP-1 promoter contains binding sites for NF-κB (13). NF-κB exists in the cytoplasm as a family of five rel-related subunits; p65, p50, p52, c-rel, and RelB, which have varied stimulatory and inhibitory effects on promoter regions of different genes. p50/p65 is the commonest activating heterodimer of NF-κB and is involved in activation of pulmonary epithelial cells in response to CoMTB (14). Homodimers of p65 exist in vivo and are strong transcriptional activators (44, 45). CoMTB activates p65 and to a lesser extent p50, suggesting that p65 homodimers and p50/p65 heterodimers drive astrocyte MMP-9 secretion. p65 was translocated to the nucleus within 10 min of CoMTB stimulation but activity persisted out to 24 h. In response to CoMTB, cytoplasmic IκBα was degraded within 10 min, whereas IκBβ degradation was delayed occurring after 30 min. IκBβ remains degraded at 24 h which is similar to the IκB response seen in TNF-α stimulated HeLa cells (46). Whereas IκBα regulates transient NF-κB activation, a key function of IκBβ is to maintain persistent NF-κB activity (47). The functional involvement of NF-κB and the importance of p65 in MMP-9 up-regulation from CoMTB-activated astrocytes was confirmed using the sesquiterpene lactone helenalin. P65 activity and MMP-9 secretion was reduced by a factor of 8 × 2 μM helenalin with no significant cell death. Sesquiterpene lactones are the active component in traditional anti-inflammatory remedies using plants from the genus Arnica (27, 48).

Dexamethasone is used as an adjunct to antituberculosis chemotherapy in the treatment of CNS TB because it decreases mortality, although the mechanisms underlining this effect are not known (28, 49). No detectable dexamethasone-induced differences in inflammatory markers such as CSF leukocytosis or TNF-α concentration has been demonstrated but MMP concentrations were not examined (28). We show that dexamethasone antagonizes up-regulation of astrocyte MMP-9 secretion by CoMTB. Dexamethasone did not affect the secretion of TIMP-1 or TIMP-2. These data suggest that dexamethasone may reduce the proteolytic potential of astrocytes and help prevent development of a matrix-degrading phenotype in the CNS.

In conclusion, the present study demonstrates that astrocyte gene expression and secretion of MMP-9 is significantly up-regulated by M. tuberculosis through a monocyte-dependent network. In contrast, TIMP-1 is not significantly up-regulated, supporting the idea that a matrix-degrading phenotype develops in CNS-TB. Importantly, the cellular findings were confirmed in patients with CNS-TB. Both dexamethasone and inhibition of NF-κB prevent MMP up-regulation from astrocytes in response to M. tuberculosis-dependent networks. This first study to demonstrate that astrocytes are a major source of MMP-9 in CNS-TB indicates that these glial cells may have a key role in CNS tissue destruction.

The authors have no financial conflict of interest.