Loss of blood-brain barrier (BBB) integrity is believed to be an early and significant event in lesion pathogenesis in the inflammatory demyelinating disease multiple sclerosis (MS), and understanding mechanisms involved may lead to novel therapeutic avenues for this disorder. Well-differentiated endothelium forms the basis of the BBB, while astrocytes control the balance between barrier stability and permeability via production of factors that restrict or promote vessel plasticity. In this study, we report that the proinflammatory cytokine IL-1β, which is prominently expressed in active MS lesions, causes a shift in the expression of these factors to favor plasticity and permeability. The transcription factor, hypoxia inducible factor-1 (HIF-1), plays a significant role in this switch. Using a microarray-based approach, we found that in human astrocytes, IL-1β induced the expression of genes favoring vessel plasticity, including HIF-1α and its target, vascular endothelial growth factor-A (VEGF-A). Demonstrating relevance to MS, we showed that HIF-1α and VEGF-A were expressed by reactive astrocytes in active MS lesions, while the VEGF receptor VEGFR2/flk-1 localized to endothelium and IL-1 to microglia/macrophages. Suggesting functional significance, we found that expression of IL-1β in the brain induced astrocytic expression of HIF-1α, VEGF-A, and BBB permeability. In addition, we confirmed VEGF-A to be a potent inducer of BBB permeability and angiogenesis, and demonstrated the importance of IL-1β-induced HIF-1α in its regulation. These results suggest that IL-1β contributes to BBB permeability in MS via reactivation of the HIF–VEGF axis. This pathway may represent a potential therapeutic target to restrict lesion formation.
The pathological characteristics of multiple sclerosis (MS)3 are demyelination, progressive axonal loss, inflammation, and a reactive astrogliosis (1). The lesions of the disease are predominantly perivascular, and magnetic resonance imaging studies have shown that local breakdown of the blood-brain barrier (BBB) is a prominent event in lesion pathogenesis (2, 3). Loss of BBB integrity is the first detectable sign of an MS relapse and correlates closely with clinical exacerbation (4). Studies have also shown that breakdown of the BBB in the spinal cord is an early and significant event in the pathogenesis of an animal model of MS, experimental autoimmune encephalomyelitis (EAE) (5), and that the extent of BBB permeability in EAE correlates with the severity of its clinical course (6). Taken together, these findings suggest that the CNS vasculature is the site of important changes early in the pathogenesis of an MS lesion.
The BBB is an endothelial barrier, consisting of a well-differentiated network of microvessels that use complex tight junctions to restrict paracellular diffusion (7). It is known that the ability of endothelial cells to form tight junctions is not intrinsic, and that their barrier phenotype is induced by the CNS environment (8). Astrocytes have been strongly implicated in the induction and maintenance of the barrier by both grafting studies (9) and coculture experiments (10, 11), and these cells are thought to act as important regulators of the balance between endothelial stability and permeability (12). Genetic and tissue culture approaches have also recently suggested a role for pericytes in angiogenesis and BBB formation (13, 14), and reciprocal effects of endothelial cells on the differentiation of both astrocytes and pericytes have also been reported (15, 16).
A reactive astrogliosis is a prominent pathological feature of the lesions of MS (1), and transgenic studies have implicated this response in the regulation of multiple aspects of CNS inflammation and regeneration, including breakdown and repair of the BBB (17). A critical factor in the induction of a reactive astroglial response is believed to be the proinflammatory cytokine IL-1, which is expressed by macrophages and microglia in acute and chronic-active MS lesions (18, 19). The concentration of IL-1 in the CSF of MS patients has been shown to correlate with disease activity (20), and specific IL-1 genotypes, or the balance between IL-1 and its receptor antagonist IL-1Ra, are also known to be associated with clinical severity and/or progression of the disease (21). Expression of IL-1β in the CNS induces reversible breakdown of the BBB (22), and this effect has been linked to IL-1β-induced activation of astrocytes and the resulting astrogliosis (18).
In the current study, we used a microarray-based approach to explore potential mechanisms linking astrocyte reactivity and BBB permeability, and this approach revealed that IL-1β up-regulated genes associated with vessel plasticity during development. Significantly, Il-1β induced hypoxia-inducible factor-1α (HIF-1α), the regulatory subunit of the transcription factor HIF-1, known to play a central role in developmental angiogenesis (23, 24). Interestingly, although cytokine induction of HIF-1α has been reported previously in other systems (25, 26, 27), its developmental up-regulation is more commonly observed as a hypoxia response, with this effect being achieved via a posttranslational mechanism (28). Up-regulation of HIF-1α leads to HIF-1 activation and binding to hypoxia response elements in the promoters of specific target genes (29, 30), perhaps the most important of which is vascular endothelial growth factor-A (VEGF-A), a potent angiogenic factor that also has been linked to BBB permeability (31, 32, 33).
To test the functional significance of our microarray findings, we conducted extensive functional analyses both in vitro and in vivo, and to determine relevance to MS, we performed immunohistochemistry on human tissue samples. Taken together, the results of these studies suggest that IL-1β-mediated activation of the HIF–VEGF axis in primary human astrocytes plays a significant role in the BBB permeability that occurs as part of MS lesion pathogenesis. We suggest that this pathway may be worthy of attention as a potential therapeutic avenue to restrict lesion formation and/or progression in the disease.
Materials and Methods
Primary human fetal astrocyte cultures were established from four different brains and were propagated as described previously (34). Tissue collection and use was approved by the relevant Institutional Clinical Review Committee. Primary human brain microvessel endothelial cells (HBMVEC) were purchased from Cell Systems, and were grown on attachment factor-coated surfaces in complete serum containing medium from the same company.
Human cDNA arrays were obtained from the Albert Einstein College of Medicine cDNA Microarray Facility (〈http://microarray1k.aecom.yu.edu〉), and each slide contained an unbiased, random collection of 27K cDNA probe elements (Incyte Genomics). Astrocytes derived from four separate fetal CNS tissue samples were studied in independent experiments. Total RNA (control or cytokine-treated) was extracted using the RNeasy Mini kit (Qiagen) and analyzed using an Agilent 2100 bioanalyzer. Samples without evidence of degradation were amplified once using the MessageAmp antisense RNA (aRNA) kit (Ambion). Following reverse transcription of aRNA, cytokine-treated samples were labeled with Cy5-fluorescent nucleotides (Amersham Biosciences), while control samples were labeled with Cy3-fluorescent nucleotides. Cy5- and Cy3-labeled cDNA samples were cohybridized to microarray slides, which were then washed and scanned using a GenePix 4000b scanner (Axon Instruments). Data were filtered, and the signal ratio of Cy5 to Cy3 was calculated for each spot to derive a relative expression value, which was then used to compare gene expression in the different experimental groups.
Quantitative real-time PCR (Q-PCR)
Cultures of primary human astrocytes were serum starved for 72 h, then treated as described, and RNA was harvested using an Absolutely RNA RT-PCR Miniprep Kit (Stratagene). cDNA was generated, and Q-PCR performed and validated, as described previously (37, 38). Each transcript in each sample was assayed in triplicate, and mean detection threshold (CT) values were used to calculate Fp (fold-change ratios between experimental and control samples) for each gene. Amplicon size and reaction specificity were confirmed by agarose gel electrophoresis. Primers used were: HIF-1α (sense) GAG CAG ACT GTA AAC AAG, (antisense) CAC AAA GAG CAA AAG GAA TG, product 245 bp; VEGF-A (sense) CAG ATT ATG CGG ATC AAA CC, (antisense) GAA CGC TCC AGG ACT TAT AC, product 183 bp; VEGF-C (sense) GCT AAG ATT GTA CTG TTT TCC, (antisense) CTT GGC TGT TTG GTC ATT G, product 225 bp; NRP-2 (sense) ACT CAA CCA GTA TGC CCA GC, (antisense) CAA AGT CAC AAC ACC ACC AC, product 309 bp; and Gravin/SseCKS (sense) CTT TAC ATC CTT TTT CCT AC, (antisense) GCA CAA ACA CAA TCC AGT ATC, product 216 bp.
Small interfering RNA (siRNA) experiments
HIF-1α siRNA was purchased from Dharmacon. Primary human astrocytes were plated to 70% confluence in 24-well plates and were transfected in fresh medium with 5 nM siRNA using TransIT-TKO transfection reagent (Mirus). Controls included nontargeting siRNA to rule out nonsequence-specific effects and transfection reagent alone (sham control). Cells were serum starved for 6 h, then treated as described. Efficiency of transfection and extent of gene silencing were assessed by Western blotting for HIF-1α and for HIF-1β (constitutively expressed control), and culture supernatants were collected in parallel for analysis by ELISA.
Adenoviral gene transfer: cell cultures
Astrocytes in 100-mm culture dishes were infected with adenovirus encoding a proteolysis-resistant (S32,36E) superrepressor form of IκBα (Ad-IκBα) or adenovirus containing the CMV promoter, but not IκBα (Ad-Ctrl), at a concentration of 104 viral particles per cell (39). Infection was allowed to proceed for 24 h, and cultures were then washed and treated as described.
SDS-PAGE and Western blotting were performed as reported previously (40).
Primary HBMVEC cultures were starved in medium containing 0.5% FCS for 4 h, trypsinized and replated into the same medium in 24-well plates on growth factor-free Matrigel (BD Discovery Labware), and treated as described. At the times specified, live cultures were examined and photographed using a Leica TCS-SP confocal microscope (Leica Microsystems). Image analysis was performed using ImageJ software (version 1.30; National Institutes of Health).
Immunofluorescence: cell cultures
Cultures grown on glass confocal plates (Mat-Tek) and treated as described were fixed and stained using previously published methods (41). Primary Abs were used at a dilution of 1/100 as stated in the text, while secondary Abs conjugated to AlexaFluor 488 and/or AlexaFluor594 (Molecular Probes) were also used at a 1/100 dilution. Cultures were examined and photographed using an Axiovert 200 inverted fluorescence microscope imaging system with Axiovision software (Zeiss).
Animal experiments were conducted using adult male Sprague Dawley rats (200–300 g), and were approved by the relevant Institutional Review Committee. Animals were anesthetized and placed into a stereotaxic frame (David Kopf Instruments). In some experiments, VEGF165 (100 ng, in 3 μl of PBS/BSA vehicle), or the same volume of vehicle control, was slowly microinjected into the cerebral cortex at (x = 3.2, y = 4.8, z = 2.0) in a controlled fashion using standard methods. In other experiments, 107 PFU of the replication-deficient adenoviral IL-1β expression vector AdhIL-1β (a gift from Dr. J. Gauldie, McMaster University, Ontario, Canada) or empty vector control AdDL70 was microinjected at (x = 3, y = 4.5, z = 1) in the same total volume of PBS (42, 43). Animals were allowed to recover, then 24 h to 14 days later were anesthetized with an overdose of pentobarbital (60 mg/kg) and perfused with heparinized saline followed by 10% buffered formalin. Brains were postfixed for 24 h, then cryoprotected in 20% sucrose for 48 h before being stored in isopentane at −20°C. Coronal sections (40 μm) were cut through the entire brain using a Leica CM3050 cryostat and were mounted on glass slides and stored at −20°C.
Immunohistochemistry: frozen sections
For staining using diaminobenzidine, sections were washed twice in PBS for 10 min each, incubated in methanol 0.3% hydrogen peroxide 30 min, rinsed in ddH2O for 2 min, washed once in PBS, then blocked in PBS 1% goat serum 0.2% Tween 20 1 h at room temperature (RT). After washing in PBS, sections were incubated overnight at 4°C with an Ab specific for fibrinogen (1/400), then washed and incubated in appropriate biotinylated secondary Ab (Vector Laboratories) at 1/250 for 30 min in blocking buffer at RT. Following additional washing in PBS, sections were incubated in streptavidin-HRP (NEN) 1/100 in a commercial blocking solution (NEN) for 30 min at RT, then washed three times in TBS 0.05% Tween 20, developed in diaminobenzidine, and mounted. All sections were examined and photographed using a Spot2 CCD camera (Diagnostic Instruments) attached to an Axiophot upright microscope (Zeiss), and image analysis was performed as described above.
For immunofluorescence staining, sections were rinsed twice in PBS, then incubated in PBS 0.1% glycine 10 min, blocked in PBS 0.3% Triton X-100 10% goat serum for 1 h, and incubated with primary Abs in blocking buffer overnight at 4°C. Abs used were HIF-1α (rabbit), VEGF-A, and VEGF-C, all at a dilution of 1/100, either alone or in combination with GFAP (also 1/100). After washing three times in PBS 0.3% Triton X-100, sections were then incubated in relevant secondary Abs conjugated to AlexaFluor 488 and/or AlexaFluor 594 (1/100; Molecular Probes) in blocking buffer for 1 h at RT, washed again three times, and counterstained with 4′,6′-diamidino-2-phenylindole (DAPI). All samples were examined and photographed using a Zeiss LSM 510 META laser scanning confocal system attached to an Axiovert 200 inverted fluorescence microscope. In some experiments, Z-series stacks were collected from the confocal microscope using 1 μm on the z-axis.
Immunohistochemistry: paraffin sections
Early postmortem tissues (4–18 h) from 10 MS patients staged as described (44), 5 age- and sex-matched normal controls, and 4 patients with other neurological diseases, were studied.
For all Ags except IL-1, the following protocol was used: sections were dewaxed and rehydrated, then incubated in PBS 3% hydrogen peroxide for 30 min, blocked with a commercially available blocking buffer (NEN), and incubated with relevant primary Abs in blocking buffer for 2 h at RT, then overnight at 4°C. After washing, sections were incubated in relevant goat biotinylated secondary Abs (1/200; Vector Laboratories) for 30 min, then developed using standard ABC protocols (Vector Laboratories) (GFAP) or tyramide signal amplification protocols (NEN) (all other Ags), followed by diaminobenzidine (brown) or NBT (blue) (both from Sigma-Aldrich). In the case of RCA-1, sections were incubated in biotinylated lectin, then developed using ABC protocols (Vector Laboratories) (GFAP) or tyramide signal amplification protocols (NEN) (all other Ags) followed by diaminobenzidine. For double labeling, sections were stained for HIF-1α or VEGF-A (rabbit), followed by GFAP (mouse). Control sections were processed using species- and isotype-matched irrelevant primary Abs, IgG fractions, and no primary Ab controls. All sections were then dehydrated, mounted, and examined as described above.
For IL-1, staining was performed using an alternate protocol, as described previously (45).
Statistical analysis of quantitated results was performed with Prism software (GraphPad), using appropriate tests. For multiple comparisons, the test used was one-way ANOVA followed by Bonferroni posttest. Student’s t test was used to compare two groups of matched samples. In both cases, p < 0.05 was considered significant.
IL-1β regulates astrocytic expression of factors that control endothelial permeability
Our initial evidence for a link between astrocyte reactivity and BBB permeability came from cDNA microarray analysis of human astrocytes activated by cytokines known to be present in the MS lesion (Fig. 1⇓). Astrocytes derived from four separate fetal CNS tissue samples were treated with either IL-1β (10 ng/ml) alone, or IL-1β plus IFN-γ (10 ng/ml), for 6 or 24 h, and the gene expression patterns of these cultures were compared with untreated matched controls using microarrays of 27,000 cDNA elements representing 18,000 unique human cDNAs. As shown in Fig. 1⇓a, IL-1β induced the expression of genes associated with endothelial permeability and new vessel formation, including the developmental transcription factor HIF-1α, VEGF-A and VEGF-C, and the VEGF coreceptor NRP-2. Conversely, IL-1β also down-regulated gravin/SSeCKS, a potent maturation and stability factor for the BBB (46). The induction of HIF-1α and VEGF-A was particularly interesting and potentially significant, because these two genes are functionally linked during development, in which new vessel formation is driven by activation of HIF-1 and resulting expression of its downstream target VEGF-A (23, 24, 29).
To validate our microarray data at the RNA level, we used Q-PCR. As shown in Fig. 1⇑, b–g, these experiments confirmed the results of our microarray screen, showing that IL-1β up-regulated factors associated with vascular permeability, and down-regulated antipermeability factors. They also validated an additional aspect of our microarray data, which was that IFN-γ, another proinflammatory cytokine expressed in active MS lesions, potentiated the effects of IL-1β on the expression of HIF-1α and VEGF-A and VEGF-C. Interestingly, when used alone, IFN-γ had minimal effects on the expression of these genes; the presence of IL-1β was required for the activity of IFN-γ to become apparent (Fig. 1⇑, b–d).
Protein-level data show that IL-1β induces both HIF-1α and its target gene, VEGF-A
We next used ELISA and Western blotting to confirm these results at the protein level (Fig. 2⇓). First, we conducted sandwich ELISA for the secreted factors VEGF-A and VEGF-C. VEGF-A is known to be a potent inducer of vascular plasticity, acting via binding to VEGFR-2/flk-1 (47). The primary role of VEGF-C is in lymphangiogenesis, acting via VEGFR-3/flt-4 (48), although it has also recently been linked to trophic effects on neural progenitors (49). These experiments confirmed that IL-1β strongly induced both VEGF-A and VEGF-C in human astrocytes, with levels peaking at 24 h following treatment, and also confirmed that induction of both factors was potentiated by cotreatment with IFN-γ (Fig. 2⇓, a and b).
In the developing embryo, VEGF-A is a target of the transcription factor, HIF-1, which exists in its active form as a heterodimer of HIF-1α and the constitutively expressed subunit, HIF-1β (29). Western blotting experiments confirmed that IL-1β strongly induced HIF-1α in human astrocytes, and that this effect was significantly potentiated by IFN-γ (Fig. 2⇑c). Induction of HIF-1α was first detected at 6 h following cytokine treatment and peaked at 24 h before dropping to undetectable levels by 72 h (Fig. 2⇑d). In contrast, and as expected, expression of HIF-1β was constitutive and not sensitive to IL-1β (Fig. 2⇑c). Western blotting also confirmed that IL-1β induced NRP-2 in primary human astrocytes (Fig. 2⇑e).
In addition to up-regulating factors associated with BBB permeability, our data also showed that IL-1β induced a slow and progressive down-regulation of the BBB maturation/stability factor gravin/SSeCKS in astrocytes, with a reduction in the levels of the protein first becoming apparent at 24 h, and progressing through at least 72 h following treatment (Fig. 2⇑f). The time course of these changes was significantly slower than those observed in the IL-1β-mediated up-regulation of HIF-1α (compare with Fig. 2⇑d).
Expression of HIF-1α, VEGF-A, and VEGFR2/flk-1 in MS lesions
In the developing embryo, the HIF-VEGF pathway is known to act as a critical mechanism in the induction of vascular plasticity that is associated with angiogenesis (23, 24, 50, 51). Changes in vascular permeability are the first detectable signs of new lesion formation in MS (4), and we hypothesized that IL-1β-induced reactivation of the HIF-VEGF axis in astrocytes might act as a significant contributing factor in the pathogenesis of these events. As an initial step in testing this hypothesis, we performed immunohistochemistry for HIF-1α, VEGF-A, VEGF-C, VEGFR2, VEGFR1, VEGFR3, NRP-2, and IL-1, on autopsy tissue from 10 confirmed cases of MS, and five age-and sex-matched normal controls. The results of these studies are shown in Fig. 3⇓. Fig. 3⇓, a–d, shows a series of matching images of an active MS lesion centered on a vessel, stained with Luxol Fast Blue to illustrate demyelination (Fig. 3⇓a), and immunostained for HIF-1α (Fig. 3⇓b), VEGF-A (Fig. 3⇓c), and IL-1 (Fig. 3⇓d). We found that all three genes were expressed in active MS lesions but not in adjacent normal-appearing white matter (Fig. 3⇓, b–d), and our experiments further demonstrated that HIF-1α and VEGF-A, in particular, localized to similar areas within MS plaques. Using cell type-specific markers, we were also able to identify the populations involved. As our in vitro experiments had predicted, in active MS lesions both HIF-1α and VEGF-A were expressed by GFAP+ hypertrophic astrocytes (Fig. 3⇓, e–i). Interestingly, HIF-1α also was expressed by a small number of macrophages/microglia (data not shown). Neither HIF-1α nor VEGF-A were expressed in normal control tissue. As expected, the receptor for VEGF-A, VEGFR2/flk-1, was expressed by endothelial cells lining CNS vessels (Fig. 3⇓j), and was found in the brains of both MS patients and normal controls. As has been published previously, IL-1 localized to cells with the typical morphology characteristic of RCA-1+ macrophages and microglia (Fig. 3⇓, k and l) (19). Immunoreactivity for IL-1 was not observed in normal control tissue. Isotype and primary Ab omission controls were also performed and were found to be negative. Staining was not observed for VEGF-C, VEGFR1, VEGFR3, or NRP-2 in either MS or control samples (data not shown).
IL-1β induces HIF-1α, VEGF-A, and VEGF-C, and BBB permeability, in an animal model
We next tested the relevance of these findings in the context of the adult CNS, using an animal model (Fig. 4⇓). First, we investigated whether expression of IL-1β in the CNS was associated with BBB permeability and whether this was accompanied by HIF-1α and VEGF-A expression. To do this, we used stereotactic microinjection to introduce a replication-deficient adenoviral vector expressing IL-1β or empty vector control into the left cerebral hemisphere of adult rats, and then monitored the resulting changes over a period of 14 days. We found that expression of IL-1β in the cortex was associated with progressive BBB breakdown, as assessed by immunohistochemistry for fibrinogen. BBB permeability was observable at 8 days and was maximal by 14 days postinjection (Fig. 4⇓a), and appeared to extend from the microinjection site along the corpus callosum and into the striatum of the ipsilateral hemisphere. It did not cross into the contralateral hemisphere, and we also observed little or no BBB permeability in the CNS of animals injected with empty vector control (Fig. 4⇓a). These changes were quantitated by computer-assisted morphometry and compared using statistical methods, and the results were significant (Fig. 4⇓b; p < 0.01). Using confocal imaging of immunofluorescently stained sections, we demonstrated that this BBB permeability was accompanied by expression of HIF-1α and VEGF-A in affected areas (Fig. 4⇓, c and d; area shown in each is adjacent to the injection site, marked by asterisk in Fig. 4⇓a). Interestingly, VEGF-C also was expressed in these same areas (Fig. 4⇓e), a result did not match our findings in human MS tissue samples. In addition, we found that areas of BBB permeability and HIF-1α, VEGF-A, and VEGF-C expression were noticeably hypercellular (Fig. 4⇓, c–e, blue channel); histological examination suggested that this was attributable to inflammation. To identify the populations expressing the three genes, we performed double-staining experiments using cell type-specific markers. The results of these studies indicated that all three genes were strongly expressed by reactive GFAP+ astrocytes, as predicted by our in vitro experiments (Fig. 4⇓, c–e, inset; note colocalization of green and red channels). This was particularly apparent for HIF-1α (Fig. 4⇓c) and VEGF-C (Fig. 4⇓e). In the case of VEGF-A, in addition to colocalization with GFAP, we also observed linear profiles at the border of the lesion that were VEGF-A+ GFAP− (Fig. 4⇓d, arrows). Immunoreactivity for the three genes did not colocalize significantly with other markers tested. In contrast with the changes in animals injected with IL-1β expression vector, we observed minimal immunoreactivity for HIF-1α, VEGF-A, or VEGF-C, and no hypercellularity, in the CNS of animals injected with empty vector control (Fig. 4⇓, c–e).
To determine whether part or all of the effect of IL-1β on the BBB might be attributable to VEGF-A induction, we introduced either recombinant VEGF-A or vehicle control into the CNS of adult rats using stereotactic microinjection and monitored the resulting changes (Fig. 4⇑, f and g). Using this approach, we confirmed that microinjection of VEGF-A was associated with significant BBB breakdown, which was readily observed by 24 h after injection of VEGF-A and which localized to areas around the injection site (Fig. 4⇑f). In contrast, minimal effects were observed in the CNS following injection of vehicle control. Quantitation of these changes by computer-assisted morphometry, and comparison using statistics, showed that they were significant (Fig. 4⇑g; p < 0.01). Taken together, the results of these experiments show that IL-1β induces expression of HIF-1α, VEGF-A, and VEGF-C, and BBB permeability in the adult CNS. They also are compatible with the hypothesis that at least some of the effects of IL-1β on the BBB may be mediated via VEGF-A.
Cultures of human brain microvessel endothelium exhibit robust responses to VEGF-A
Because VEGF-A had striking effects in the rodent CNS, we next tested whether it was also able to elicit responses in endothelial cells from the human brain (Fig. 5⇓). Using an immunofluorescence approach, we first determined whether primary cultures of HBMVEC express VEGFR2/flk-1 (Fig. 5⇓, a and b). We confirmed that these cells were positive for the cell type-specific markers CD31/PECAM-1 and Factor VIII-related Ag (Fig. 5⇓a) and found that they also expressed VEGFR2/flk-1, and that staining for this receptor localized to the cell surface in these cultures (Fig. 5⇓b). We next determined whether receptor expression was associated with functional responses to VEGF-A using an angiogenesis assay (Fig. 5⇓, c and d). Cells were plated onto a gel substrate, and we compared the behavior of VEGF-treated cultures with controls. We found that cells treated with VEGF-A formed tubular structures resembling capillaries over a period of ∼24 h, whereas untreated control cultures remained confluent and apparently quiescent (Fig. 5⇓c). These changes were quantitated and were found to be significant (p < 0.001; Fig. 5⇓d).
IL-1β uses NF-κB to induce HIF-1α, which triggers the formation of active HIF-1 complexes and potentiates VEGF-A expression in primary human astrocytes
Having established the potential significance of VEGF-A expression in the adult CNS, we next examined the role of IL-1β-induced HIF-1α in regulating VEGF-A expression (Fig. 6⇓). First, we determined whether induction of HIF-1α was associated with translocation to the nucleus in primary human astrocyte cultures (Fig. 6⇓a). HIF-1α is known to be expressed in in the developing brain (52), and the results of gene targeting studies suggest that its expression in the CNS triggers the formation of HIF-1α/HIF-1β heterodimers, which migrate to the nucleus and induce the expression of target genes including VEGF-A (50). To determine the subcellular localization of HIF-1α induced in astrocytes by cytokine treatment, we used immunofluorescence, and this approach demonstrated that HIF-1α expressed in response to IL-1β or IL-1β + IFN-γ localized to the nucleus in astrocyte cultures, suggesting that its up-regulation was sufficient to trigger the formation of active HIF-1 complexes (Fig. 6⇓a).
We next investigated the mechanism underlying IL-1β-mediated HIF-1α expression, using a molecular approach. IL-1β activates two main signal transduction mechanisms, the NF-κB and AP-1 pathways. Activation of NF-κB is brought about via phosphorylation and degradation of its inhibitor IκBα, whereas AP-1 is activated following phosphorylation of the MAPK p38 and JNK/SAPK (35). In separate experiments, we used an adenovirus expressing a proteolysis-resistant (S32,36E) IκBα superrepressor to inhibit the NF-κB pathway (39) and siRNA specific for p38 MAP kinase to inhibit AP-1 activity. Interestingly, we found activation of NF-κB to be the dominant mechanism underlying HIF-1α induction. Robust induction of HIF-1α was observed in response to IL-1β in human astrocytes infected with an empty control adenovirus, whereas induction was abolished in cultures infected with virus expressing the IκBα superrepressor (Fig. 6⇑b, upper). This effect was specific: the IκBα construct had no effect on the expression of HIF-1β (Fig. 6⇑b, lower). In contrast, down-regulation of p38 using siRNA had no effect on HIF-1α induction (data not shown). Taken together, these results suggest that IL-1β induces HIF-1α via an NF-κB-dependent mechanism in primary human astrocytes.
Finally, using siRNA we directly addressed the role of HIF-1α in the induction of VEGF-A (Fig. 6⇑, c and d). Human astrocyte cultures were transfected with siRNA specific for HIF-1α, or nontargeting control, or transfection reagent alone (sham), then treated with IL-1β + IFN-γ for 24 h. Induction of VEGF-A and expression of HIF-1α and HIF-1β were then monitored in parallel, using sandwich ELISA and Western blotting. HIF-1α was strongly induced in response to cytokine treatment, but this induction was abrogated in cultures transfected with HIF-1α siRNA (Fig. 6⇑c, upper). This effect was specific: we observed no effect on HIF-1β expression (Fig. 6⇑c, lower). Importantly, we found that the siRNA-mediated abrogation of HIF-1α expression strongly down-regulated, but did not completely abolish, the induction of VEGF-A, which was reduced by ∼50% in these cultures (Fig. 6⇑d). Taken together, the results of these experiments show that HIF-1α and VEGF-A induced in response to IL-1β are functionally linked. HIF-1α clearly exerts a strongly positive regulatory effect on VEGF-A, feeding forward to potentiate its expression, although interestingly, it is not required for basal VEGF-A induction (summarized in Fig. 7⇓).
In the mammalian CNS, the BBB has been shown to consist of at least two major components, the barrier itself, which is endothelial, and a second, astrocytic, component, which has been implicated in the induction and maintainence of the barrier phenotype by both grafting studies (9) and coculture experiments (10, 11). The existence of an additional, pericyte component has also recently been demonstrated by genetic and tissue culture approaches, although fewer details are currently available regarding the mechanisms used by these cells to regulate the properties of the BBB (13, 14). More is known about the role of astrocytes in these events, and based on available data, it is believed that astrocyte processes form endfeet that collectively surround CNS microvessels and regulate the endothelial barrier. In the current study, we tested the hypothesis that inflammatory cytokines produced during the formation of an MS lesion alter the pattern of astrocyte gene expression to favor a state of endothelial plasticity and vascular permeability. Using a microarray-based approach, we found that, in human astrocytes, the proinflammatory cytokine IL-1β, which is prominently expressed in the active lesions of MS, induced a pattern of gene expression compatible with this hypothesis. Two of the genes induced by IL-1β in our study, HIF-1α and VEGF-A, were of particular interest in the context of BBB breakdown. Both are involved in developmental angiogenesis, they are known to be functionally linked, and animal studies have shown the importance of the HIF–VEGF axis in vascular development (23, 24, 29, 50).
The role of HIF-1 in developmental angiogenesis has been highlighted by both gene targeting and cell culture studies. Mouse embryos that lack HIF-1α die at mid-gestation, with multiple vascular abnormalities and mesenchymal cell death (37), while targeted deletion of HIF-1α in the brain is associated with hydrocephalus and increased apoptosis of neural cells, which coincides with vascular regression within the telencephalon (50). Cell culture experiments have shown that HIF-1α is made continuously in defined cell types, and that although it is usually rapidly hydroxylated and degraded (53, 54), its accumulation (and resulting activation of HIF-1) can be triggered via two distinct mechanisms. First, the hydroxylases that target HIF-1α for degradation are inhibited under conditions of reduced oxygen tension, allowing it to accumulate (55) (Fig. 7⇑a). Second, its induction also is regulated by a number of different transcription factors (56), some of which are known to be cytokine-sensitive (25, 26, 27). Relevant to this latter mechanism, in the current study we found that HIF-1α expression in response to IL-1β was NF-κB dependent in primary human astrocytes (Fig. 7⇑b). A link between NF-κB and HIF-1α expression has been suggested by a previous study, although data from the manuscript in question do not match ours perfectly: the authors found that the effect of NF-κB was mediated via activation of cyclooxygenase-2 and acted on HIF-1α posttranscriptionally (26), whereas we detected a clear effect of IL-1β on HIF-1α at the RNA level. Interestingly, we also found that expression of HIF-1α and its downstream effectors VEGF-A and VEGF-C was potentiated by cotreatment with IFNγ (see Figs. 1⇑b and 2⇑c), although IFNγ alone was not sufficient for HIF-1α induction. IFNγ signaling is known to activate classical JAK-STAT signaling and the formation of STAT1 homodimers, but has also been linked to activation of additional signaling cascades, including the PI3 kinase pathway (57), and effects on mRNA stabilization (58). We are currently investigating the contribution of one or more of these pathways to HIF-1α expression and activation of HIF-VEGF signaling.
VEGF-A is believed to be the single most important driver of vascular formation in the body, and it is known to be required to initiate the formation of immature vessels by vasculogenesis or angiogenic sprouting, both during development and in the adult (31, 59). As such, its expression requires exquisite spatial and temporal regulation. The phenotype associated with disruption of the VEGF-A gene, or its receptor VEGFR2/flk-1, is even more extreme than that observed in HIF-1α knockout animals, such that loss of a single allele of VEGF-A leads to embryonic lethality due to severe vascular abnormalities (47). The consequences of inappropriate VEGF-A expression are almost as severe as those observed following its inappropriate deletion. For example, overexpression of VEGF-A in the retina leads to robust new angiogenesis, but of leaky vessels growing in abnormal patterns that ultimately result in severe damage to the tissue (60). The mechanisms that regulate VEGF-A are complex, and it is known to be sensitive to a number of different transcription factors. Of these, HIF-1α is the best characterized and is believed to play an important role in the developing CNS (29), although NF-κB, AP-1, and Smad and STAT proteins have also all been shown to exert regulatory effects upon VEGF-A (reviewed in Ref. 61). Our data are compatible with the concept that more than one transcription factor is important in the induction of VEGF-A by IL-1β in astrocytes. Althoughour experiments using siRNA show that HIF-1α is a potent regulator of VEGF-A, they also demonstrate that some VEGF-A induction persists when HIF-1α is abrogated, suggesting the involvement of an additional pathway, such as IL-1β-activated NF-κB or AP-1 signaling. For example, one potential scenario might see IL-1β-activated NF-κB inducing both HIF-1α and VEGF-A, with HIF-1α then feeding forward to potentiate and fine-tune expression of VEGF-A. Another might involve induction of HIF-1α by NF-κB (one arm of the IL-1β signaling cascade) and VEGF-A by AP-1 (the other arm), with the subsequent HIF-1α-mediated potentiation of VEGF-A representing integration of the two arms of the signaling network. Either scenario might be appealing due on the grounds of simplicity and/or elegance, although at the present time their existence is a matter for conjecture.
Given the potent effects of VEGF-A, studies have examined its role as an end mediator in initiating or contributing to pathological situations in adult tissues, including the CNS. Although initial work suggested that VEGF-A was not a significant factor in inducing BBB permeability in the adult (62), more recent experiments have clearly shown that its administration to the CNS in adult animals disrupts BBB integrity (63, 64) and regulates inflammation and angiogenesis (65). The results of our experiments are compatible with these data: we found that a single microinjection of recombinant VEGF-A was sufficient to initiate BBB permeability in the adult rodent CNS (see Fig. 4⇑, f and g). A recent study also has described VEGF-A expression in EAE and MS (32), and our data strongly support and extend this previous work. Taken together, the results of our experiments suggest that IL-1β-induced HIF-1α represents an important regulator of VEGF-A expression and BBB permeability in the context of the active MS lesion, as is the case in the developing CNS, and it is this conclusion, that puts the effects of VEGF-A into the context of mechanism, which is perhaps our most novel and compelling finding.
Although the data contained in this manuscript clearly implicate the HIF–VEGF-A axis in BBB permeability in the lesions of MS, they also leave unanswered a number of interesting questions, which are currently the subject of further study in our laboratory. For example, the mechanism underlying the changes in BBB permeability observed in our experiments remains unresolved. At its most basic level, the BBB acts as a physical and metabolic barrier and also inhibits the transport of bioactive molecules (reviewed in Ref. 66). The physical barrier relies on complex tight junctions between adjacent endothelial cells, which force molecular traffic to take a transcellular route across the BBB. Among the molecules identified as making important contributions to tight junction structure are members of the claudin family, which appear to be important contributors to the properties of the physical barrier, as well as occludin, which has been shown to function in tight junction regulation, and the junctional adhesion molecules A–C, together with adaptor and regulatory proteins, including ZO-1–ZO3, MAGI-1–MAGI-3, cingulin, and CASK (reviewed in Ref. 12). We have shown previously that at least some of these proteins, including members of the claudin family and occludin, are sensitive to the effects of inflammatory cytokines in some cell types, including astrocytes (67), and ongoing work in our laboratory is currently investigating cytokine regulation of tight junction components in more detail in both human brain microvessel endothelial cells and astrocytes. Our initial results suggest that multiple members of the claudin family, as well as other junctional components, are cytokine sensitive, and we hope to have a more detailed picture of cytokine regulation of BBB tight junctions in the coming months.
Another question left unanswered by our current data concerns the presence and role of VEGF-C in the inflamed adult CNS. Although we were unable to detect VEGF-C in our experiments using MS tissue samples, we observed robust VEGF-C staining in tissue from adult rats microinjected with IL-1β expression vector (see Fig. 4⇑e). The reason for this dichotomy remains unclear and is the subject of current investigation in our laboratory. VEGF-C has previously been associated with a role in lymphangiogenesis, acting via VEGFR-3/flt-4 (see Ref. 31), but a recent study has shown that in the CNS it acts as a potent trophic factor for neural progenitor cells, including oligodendrocyte precursors (49). The potential role of VEGF-C in the injured and/or inflamed adult CNS is therefore a subject of potential interest in the context of neuroprotection, and we are currently investigating its expression and function in the CNS in more detail.
Potential therapeutic strategies for MS can be divided into two general categories: 1) treatments designed to reduce or prevent the occurrence of new lesions within the CNS and, hence, new clinical episodes; and 2) therapies aimed at repairing the damage that already exists. On the basis of our findings, we suggest that targeting of the HIF-VEGF pathway may represent a potentially novel therapeutic avenue in the first of these two categories. Interestingly, it has recently been suggested that combining different treatments aimed at reducing new lesion formation in MS may lead to additive therapeutic benefits and hence a significant improvement in the clinical outcome (68). Because the HIF-VEGF axis represents a pathway that is not currently targeted by existing MS therapies, it may be potentially useful in this regard, because treatments aimed at reducing its activation might usefully be combined with other, preexisting therapeutic options.
We thank Dr. Bradford Poulos, Director of the Human Fetal Tissue Repository at the Albert Einstein College of Medicine, for tissue collection; and Dr. Robert Hennigan and Paul Carman of the Mount Sinai School of Medicine Microscopy Shared Resource Facility for help with imaging studies. We also thank Dr. Aldo Massimi and Shufen Chen of the Albert Einstein College of Medicine Microarray Facility for help with cDNA microarray experiments and Dr. Jack Gauldie (McMaster University, Ontario, Canada) for providing the replication-deficient adenoviral IL-1β expression vector AdhIL-1β used in this study.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was supported by U.S. Public Health Service Grants NS46620 (to G.R.J.), NS40137 (to C.F.B.), NS11920 (to C.F.B. and C.S.R.), MH55477 (to S.C.L.), National Multiple Sclerosis Society Fellowship FG-1739 (to Y.Z.), the Jayne and Harvey Beker Foundation (to G.R.J.), and Einstein CFAR P30 AI051519 (to S.C.L.). The Mount Sinai School of Medicine/Microscopy Shared Resource Facility is supported, in part, with funding from National Institutes of Health/National Cancer Institute Shared Resources Grant R24 CA095823.
↵2 Address correspondence and reprint requests to Dr. Gareth John, Corinne Goldsmith Dickinson Center for Multiple Sclerosis, Department of Neurology, Mount Sinai School of Medicine, 1 Gustave L. Levy Place, New York, NY 10029. E-mail address:
↵3 Abbreviations used in this paper: MS, multiple sclerosis; BBB, blood-brain barrier; EAE, experimental autoimmune encephalomyelitis; HBMVEC, human brain microvessel endothelial cells; HIF-1, hypoxia-inducible factor-1; NRP, neuropilin; VEGF, vascular endothelial growth factor; siRNA, small interfering RNA;RT, room temperature; Q-PCR, quantitative real-time PCR; DAPI, 4′,6′-diamidino-2-phenylindole.
- Received August 3, 2005.
- Accepted August 1, 2006.
- Copyright © 2006 by The American Association of Immunologists