Leukocyte migration into the central nervous system (CNS) is mediated by chemokines expressed on CNS endothelial cell surfaces. This study investigated the production of chemokines and expression of chemokine receptors by human brain endothelial cells (HBECs) in vitro and in situ. Four chemokines (CCL2, CCL5, CXCL8, and CXCL10) were demonstrated by immunohistochemistry in endothelial cells in brain samples from patients with multiple sclerosis. CXCL8 and CCL2 were constitutively released and increased by primary HBECs and the brain endothelial cell line hCEMC/D3 in response to tumor necrosis factor and/or interferon γ. CXCL10 and CCL5 were undetectable in resting endothelial cells but were secreted in response to these proinflammatory cytokines. Tumor necrosis factor strongly increased the production of CCL2, CCL5, and CXCL8; interferon γ upregulated CXCL10 exclusively. CCL3 was not secreted by HBECs and seemed to be confined to astrocytes in situ. The chemokine receptors CXCR1 and CXCR3 were expressed by HBECs both in vitro and in situ; CXCR3 was upregulated in response to cytokine stimulation in vitro. In contrast, CXCR3 expression was reduced in noninflammatory (silent) multiple sclerosis lesions. The particularly high levels of CXCL10 and CXCL8 expressed by brain endothelium may contribute to the predominant TH1-type inflammatory response observed in chronic inflammatory conditions such as multiple sclerosis.
Chemokines are low molecular weight chemotactic cytokines that regulate leukocyte trafficking into tissues and are thought to be key mediators of inflammatory responses. More than 40 different chemokines that bind to 18 different chemokine receptors have been identified to date (1). The position of the first and second conserved cysteine residues determines the basis for the classification of chemokines into 4 families (i.e. CC, CXC, CX3C, and XC or the α, β, δ, and γ subfamilies in the old nomenclature, respectively). Many chemokines can bind to different receptors within the same subgroup, and most receptors can bind to several different chemokines; this is thought to ensure the redundancy of the system (2). Indeed, most of the knockout mice for chemokines or chemokine receptors are viable with the exception of knockouts of CXCL12 and its receptor CXCR4 (3-5).
Multiple sclerosis (MS) is thought to be an autoimmune inflammatory disorder of unknown cause in which a strong TH1-type immune response associated with production of the inflammatory cytokines interferon γ (IFN-γ), tumor necrosis factor (TNF), and transforming growth factor β2 (TGF-β2) develops in the central nervous system (CNS). Activated T lymphocytes and macrophages cross the blood-brain barrier (BBB) and target myelin, creating demyelinated plaques that are associated with chronic inflammation (6). A range of chemokines and their receptors are highly expressed on infiltrating leukocytes and on CNS resident cells of MS lesions. These include CCL2, CCL3, CCL5, CCL19, CCL21, and CXCL8 and CXCL10 and their cognate receptors (7-11). Among these, the increased expression of CXCL10 and its receptor CXCR3 has been shown to be a prominent feature of reactive astrocytes (12) and infiltrating CD3+ T cells within MS lesions (13). CXCL10 is an IFN-γ-inducible chemokine, and its receptor CXCR3 is highly expressed on active IFN-γ-producing TH1 cells. Consequently, once a TH1-type response has developed in the CNS, the induction of IFN-γ-inducible chemokines may selectively attract additional TH1 cells to the site of inflammation, thus expanding the chronic inflammatory response. In addition, CXCL8 expression has been shown to be increased in MS, although its exact role in MS pathogenesis remains unknown (14).
Of the CC chemokines, CCL2, CCL3, CCL5 and their receptors CCR1, CCR2, and CCR5 show increased expression on microglia and astrocytes or on infiltrating lymphocytes and macrophages in MS lesions (15). An increase in CCL2, CCL3, and CCL5 also occurs in murine experimental autoimmune encephalomyelitis (EAE), an animal model of MS (16); this suggests that CCL2 signaling could be important for attracting macrophages to sites of neuroinflammation.This view is supported by the observation that CCL2 knockout mice showed reduced clinical severity of EAE (17); in 1 study, deficiency of its receptor CCR2 prevented mice from developing EAE (18). Other potentially important CC chemokines in MS pathogenesis are CCL19 and CCL21, which have been shown to be involved in T-lymphocyte migration into the CNS in EAE and MS (7).
Most studies of chemokine expression in MS or EAE have focused on chemokine production by CNS parenchymal cells or infiltrated leukocytes, assuming that chemokines expressed in the parenchyma will be transported to the luminal surface of the endothelium. The potential for chemokines to move to the lumen is very limited, however, because brain endothelia normally have continuous tight junctions that prevent paracellular diffusion of proteins. Moreover, transcytosis is also limited because brain endothelial cells have relatively few transport vesicles compared with other endothelia. Therefore, for these reasons, chemokine production by brain endothelial cells themselves may be particularly important for regulating leukocyte traffic into the CNS; that is, they may represent the primary source of chemokines that interact directly with circulating leukocytes. Because endothelia from different tissues secrete different subsets of chemokines, the specific combinations of chemokines produced may have a major role in determining the characteristics of inflammation in each tissue (19); this is particularly true for brain endothelium.
The aim of this study was to identify the pattern of chemokine expression by human brain endothelial cells both in the resting state and in response to inflammatory cytokines. For the in vitro analysis, we used primary human brain endothelial cells (HBECs) and an immortalized HBEC line, hCMEC/D3 (20). The chemokines investigated included a subset of chemokines potentially involved in MS pathogenesis (i.e. CXCL8, CXCL10, CCL2, CCL3, and CCL5) because there is evidence that they may be produced by vascular endothelium (19). The findings in vitro were correlated with observations of chemokine expression by the endothelial cells of cerebral capillaries in situ in normal CNS and in areas of neuroinflammation in MS postmortem tissue samples.
We also investigated the expression of chemokine receptors on brain endothelium in vitro and in situ. The functional significance of endothelial chemokine receptors is not fully understood (21). They may be involved in clearance of chemokines from the cell surface by endocytosis (22-24). It has also been proposed that they are involved in transcytosis (25, 26). Alternatively, some chemokines have angiogenic or angiostatic properties, so the presence of receptors could allow these cells to respond appropriately to signals for proliferation (27). Berger et al (28) have previously demonstrated the expression of CXCR1, CXCR3, CCR2, and CCR5 by cultured HBECs, but there is little comparative data on cells in situ (25). We therefore examined MS tissue using immunohistochemistry and immunogold labeling combined with transmission electron microscopy to identify specific chemokine receptors (CXCR1 and CXR3) to establish whether our findings in vitro reflect the expression of the receptors in situ. By identifying the sets of chemokines and their distribution, the work indicates which of the chemokines present in neuroinflammation are localized in the endothelium, where they can control leukocyte transmigration.
Materials and Methods
Human Brain Tissue and Cell Culture
Human brain tissue from the frontal or temporal cortex was obtained with ethical approval either postmortem from patients diagnosed clinically or neuropathologically with MS through the UK Multiple Sclerosis Tissue Bank (London), and the MRC London Brain Bank for Neurodegenerative Diseases (Institute of Psychiatry, London). As indicated in Table 1, there were 10 female and 3 male patients with a mean age of 55.6 ± 4.37 years and a range of 34 to 92 years. Control samples were from 3 epilepsy patients undergoing temporal lobe resection (King's College Hospital, London) according to local ethical committee guidelines (Protocol 99-002). Tissues obtained from the frontal and temporal lobes of MS patients contained demyelinated plaques and/or nonaffected (nondemyelinated) regions (Table 1). Brain tissues were used for immunohistochemical and electron microscopic analyses (frozen, formalin-fixed, and paraffin-embedded samples; Table 1). Fresh brain tissue was available from 5 cases (Table 1; Cases 4-7 and 9) for establishing primary brain endothelial cultures; this was from the nonaffected white matter in 2 cases, from a silent lesion in 1 case, and from the white matter of 2 cases in which lesion activity had not been determined.
Primary human brain endothelial cells were isolated according to previously described techniques (29) with the modification that cells were cultured in medium containing 0.5 μg/ml puromycin for 5 days to remove contaminating cells (30). For isolating primary HBECs, 1-cm3 tissue blocks were required to obtain sufficient capillaries to generate pure endothelial cultures. In MS brain tissue, the blocks were taken from areas that contained MS lesions (as determined by the naked eye) but which also contained periplaque areas and surrounding normal-appearing white matter (NAWM). Human brain endothelial cells reached confluence 2 to 3 weeks postisolation and were used between Passages 2 and 4. Purity was confirmed by immunostaining for von Willebrand factor (vWf). Human brain endothelial cells and the hCMEC/D3 cell line (20) were cultured in EBM-2 (Cambrex, Wokingham, UK) supplemented with vascular endothelial growth factor, insulin-like growth factor, basic fibroblast growth factor, hydrocortisone, ascorbate, gentamicin, and 2.5% fetal bovine serum but were rested in the same medium without growth factors for at least 48 hours before assay.
Typing of Lesions in MS Brain Tissue Samples
Lesions within the white matter in MS tissues were characterized according to (i) classical histologic staining with Luxol fast blue, (ii) immunostaining with myelin-associated proteins (myelin basic protein and proteolipid protein; neuropathologic samples), and/or (iii) immunostaining patterns for CD68 and major histocompatibility complex (MHC) class II antigens (31). Active (demyelinating) lesions are hypercellular and can be characterized by infiltrating MHC-II-positive macrophages throughout the lesions; chronic active lesions are demyelinated with hypocellular centers but are hypercellular at the periphery, with activated macrophages located primarily at the edge of the lesion, and chronic inactive or “silent” lesions are hypocellular and lack MHC-II-positive cells (31).
Sectioned materials were immersed in methanol/hydrogen peroxide solution, permeabilized with 0.1% Triton X-100 in PBS for 1 hour, and nonspecific binding sites were blocked using 10% normal rabbit serum for 1 hour. The primary antibodies (CD68, clone PGM1, 1/20 and MHC-II, clone CR3/43; DakoCytomation, Cambridgeshire, UK) were incubated on the slide overnight at room temperature and processed according to established immunoperoxidase protocols (Dako 3-step avidin-biotin-horseradish peroxidase complex protocol; DakoCytomation) using 1:100 biotinylated rabbit anti-mouse immunoglobulin (Ig)G as secondary antibody. In the sampled tissues, “active” lesions were identified by the abundance of CD68-positive/MHC-II-positive mononuclear phagocytes within a lesion core; “chronic active” lesions as CD68-positive/MHC-II-positive primarily at the edge of the lesion, and CD68-positive/MHC-II-negative within the center; “silent” lesions were CD68-negative/MHC-II-negative (31; Supplemental Fig. 1). A total of 13 to 16 active/chronic active lesions from 5 cases and silent lesions from 6 cases (Table 1) were investigated for chemokine and chemokine receptor expression in situ.
Immunosperoxidase Staining of Chemokines In Situ
The immunoperoxidase staining procedure for chemokines and chemokine receptors in situ (tissue samples) was performed as previously described (32) with minor modifications. Formalin-fixed and fresh-frozen tissue blocks were sectioned serially at 10-/20-μm thickness using a cryostat and were collected on Superfrost plus slides. Frozen sections were air-dried for 90 minutes at room temperature, immersed in methanol containing 2.5% of a 30% hydrogen peroxide solution, incubated in Hanks balanced saline solution containing 1% bovine serum albumin, 1% of a (1 mol/L) stock of MgCl2 and CaCl2, and 0.01% Tween 20. Sections were next incubated with 10% appropriate normal sera (rabbit or swine) made up in PBS for 90 minutes before incubation with primary antibody solution for 24 to 36 hours at 4°C (1/500, 1/100, or 1/100 dilutions of mouse monoclonal IgG1 antibodies for CXCL8, CCL2, and CCL3, respectively; 1/1000 or 1/500 dilution of rabbit polyclonal IgG antibodies for CXCL10 and CCL5, respectively; PeproTech, London, UK). Negative controls were included where the primary antibody step was omitted and sections were incubated with normal sera alone. After 3 washes for 5 minutes each in PBS, sections were incubated in relevant biotinylated secondary antibodies (rabbit anti-mouse IgG, 1:100; swine anti-rabbit IgG, 1:100; DakoCytomation) for 120 minutes, washed in 3 changes of PBS, and incubated with avidin-biotin-horseradish peroxidase complex (ABC-HRP Standard Kit, Vector Laboratories, Peterborough, UK) made up according to the manufacturer's instructions for a further 90 to 120 minutes. After a final 3 washes in PBS, the sections were reacted with 0.5 mg/ml 3,3′-diaminobenzidine made up in PBS to which 20 μl of hydrogen peroxide was added, and the reaction was timed for optimal staining (3-5 minutes). Nuclei were weakly counterstained with Harris hematoxylin solution if required, and sections were differentiated in 0.1% acid alcohol solution (10 seconds) before being rinsed under running tap water. Sections were then dehydrated in a graded series of alcohol (50%, 75%, 90%, 95%, and 100% ethanol), cleared in xylene, and coverslipped using DPX mountant (VWR, Ltd., Dorset, UK). Slides were viewed using a brightfieldX microscope. Formalin-fixed, paraffin-embedded tissue blocks were also sectioned at 3 μm using a base-sledge microtome, dewaxed in xylene, and transferred to 100% alcohol before immersion in methanol/hydrogen peroxide solution and immunohistochemical staining as described above.
Immunofluorescence Labeling of Chemokine Receptors In Situ and in Culture
For chemokine receptor staining, fresh tissues were also fixed using 2% paraformaldehyde and 3.75% acrolein in 0.1 mol/L phosphate buffer (PB) for 1 hour and left overnight in 2% paraformaldehyde only. The sections were cut at 50 μm using a Vibratome (Leica Microsystems, Milton Keynes, UK). For immunolabeling to detect vWf/glial fibrillary acidic protein (GFAP) and CXCR1/CXCR3, an antigen retrieval step was included by incubating for 20 minutes in 0.01 mol/L citrate buffer, pH 6, at 95°C. All sections were then permeabilized with 0.1% Triton X-100 in PBS for 1 hour. Nonspecific binding was blocked using 10% normal goat serum for 1 hour. Sections were incubated with mouse anti-CXCR1/anti-CXCR3 (10 μg/ml; R&D Systems, Oxon, UK) for 1 hour, and after 3 washes, followed by Cy3-conjugated goat anti-mouse IgG (1/100; Chemicon,Hampshire, UK). Sections were then incubated with a rabbit polyclonal anti-human vWf (1/800; Sigma, Dorset, UK) or anti-GFAP (1/100; Chemicon) antibodies in PBS overnight at room temperature and detected using goat anti-rabbit-Ig conjugated to fluorescein isothiocyanate (1/200×; Chemicon) incubated for 1 hour. Sections were mounted using Mowiol. Negative controls were included where sections were incubated with normal sera or isotype-matched control antibodies in place of primary antibodies.
For characterization of chemokine/ receptor expression in vitro, cultured endothelial cells were first grown to 50% confluence on collagen-coated Labtek multiwell chambered slides (Nunc, Scientific Laboratory Supplies, Nottingham, UK), and fixed using 4% paraformaldehyde in PBS for 10 minutes. The cells were permeabilized using 0.1 % Triton X-100 in PBS for 5 minutes. Nonspecific binding sites were then blocked using 10% normal goat serum in PBS for 1 hour, followed by 1-hour incubation with primary antibodies at room temperature. Chemokine receptor expression was determined using a panel of phycoerythrin- or fluorescein-conjugated mouse monoclonal antibodies specific for CCR1, CCR2, CCR5, CXCR1, CXCR2, and CXCR3 (R&D Systems) or their respective isotype-matched antibody controls as previously described (15). These were applied for 1 hour at room temperature at the recommended dilutions. After 3 washes, the slides were mounted with a glass coverslip with Dako Fluorescent Mounting Medium (DakoCytomation) and viewed with a fluorescence microscope (BX61; Olympus, Tokyo, Japan).
Immunogold Labeling and Electron Microscopy
For chemokine receptor staining, fresh tissue was fixed using 2% p-paraformaldehyde and 3.75% acrolein in 0.1 mol/L PB for 1 hour and left overnight in 2% paraformaldehyde. The sections were cut at 50 μm using a vibratome. The floating sections were treated with 1% sodium borohydride in PB for 30 minutes and then freeze-thawed to permeabilize the tissue. The sections were incubated in cryoprotectant solution and then rapidly immersed in chlorodifluoromethane, followed by liquid nitrogen, and then in 3 successive PB washes. Blocking solution was added for 30 minutes, and sections were incubated with primary mouse antibody (mouse IgG2A anti-human CXCR1, 5 μg/ml, or mouse IgG1 anti-human CXCR3, 1 μg/ml; R&D Systems) or their respective isotype-matched antibody controls for 24 hours at room temperature and a further 24 hours at 4°C in 0.1% bovine serum albumin in TBS. Nonspecific binding of the secondary antibody was blocked with the incubation/washing buffer for 10 minutes after a rinse with PBS. Sections were then incubated for 2 hours with goat anti-mouse IgG secondary antibody (dilution 1/50) conjugated to colloidal gold (British Biocell International, Cardiff, UK) in the incubation/washing buffer. After a 5-minute wash with the incubation/washing buffer and 3 rinses with PBS, bound (1 nm) gold particles were fixed using 2% glutaraldehyde in PBS for 10 minutes, followed by a wash in a 0.2-mol/L citrate buffer solution. The sections were reacted with a silver enhancement solution (British Biocell International). The silver enhancement reaction was stopped by 2 washes in the citrate buffer and 3 washes with PB. The immunogold stained tissue was postfixed for 1 hour in 2% osmium tetroxide in PB for electron microscopic viewing. They were then dehydrated in a series of ethanol solutions (50%, 70%, 80%, and 95% concentrations for 5 minutes and twice in 100% concentration for 10 minutes) and twice in propylene oxide for 10 minutes before being embedded in Epon 812 between sheets of Aclar plastic. Epon polymerization was performed by incubation at 60°C for 48 hours, and a sample of the tissue was mounted on the tips of Epon Block. Ultra-thin sections of 70 nm were cut with a diamond knife (Diatome; TAAB, Gillingham, UK) and collected on copper mesh grids. The sections on grids were counterstained with uranyl acetate for 20 minutes and lead citrate for 6 minutes and examined with a JEOL JEM1010 electron microscope attached to a Gatan Bioscan digital camera (Joel, Welwyn Garden City, UK). Micrographs at 8,000× magnifications were taken, scanned with Epson Perfection 4870 photo, and visualized with the Photoshop 5.5 program. For quantification of the density area (number of gold particles/μm2), the negative film of the photograph was scanned and used to count the number of gold particles within endothelial cells. The surface of the endothelial cells was determined from the digital images using a National Institutes of Health image program on a Macintosh G4 computer. The subcellular localization of gold particles was recorded as cytoplasmic, luminal, or abluminal.
Reverse-Transcriptase-Polymerase Chain Reaction
RNA was isolated from primary HBECs grown to confluence in 25-cm2 flasks coated with collagen. Total RNA was extracted using 1 ml of TRIzol LS (Invitrogen, Paisley, UK) according to the manufacturer's protocol and stored at -80°C. For the reverse-transcription (RT) procedure, the protocol given by the supplier Promega (ImProm-IITM Reverse Transcriptase) was followed. Briefly, 2 μg of RNA was mixed with 1 μg of random primers at 70°C for 5 minutes then annealed on ice. Reverse transcriptase was performed by addition of 40 μl RT mix at 25°C for 5 minutes, followed by 60 minutes at 40°C. To terminate the reaction, the mix was incubated for a further 15 minutes at 70°C. The negative control for each sample consisted of RNA samples annealed with random primers without the RT. The resulting cDNA was either stored at −20°C or taken directly into the polymerase chain reaction (PCR) amplification procedure.
The PCR amplification steps were performed by adding 4 μl of cDNA from the RT reaction to 20 μl of PCR mix (50 mmol/L KCl, 10 mmol/L Tris-HCl, pH 9.0, at 25°C, and 0.1% Triton X-100), 1.5 mmol/L MgCl2, 0.2 mmol/L of deoxyribonucleotide triphosphates, 1 μmol/L of each primer, and 1 U of Taq polymerase) and placed into an iCycler PCR machine (Bio-Rad Laboratories, Hertfordshire, UK). Polymerase chain reaction conditions were 5 minutes at 94°C, followed by 40 cycles (94°C, 45 seconds; Ta indicated below, 45 seconds, 72°C, 60 seconds) finishing at 72°C for 7 minutes. Primers, Ta, and size of PCR product were as follows: CCR1, F5′ACGAAAGCCTACGAGAGTG3′, R5′GGTGAACAGGAAGTCTTGG3′, Ta 50°C, 240 bp; CCR2, F5′GATTACGGTGCTCCCTGTC3′, R5′GCCACAGACATAAACAGAATC3′, Ta 50, 496 bp; CCR5, F5′GCTGAGACATCCGTTCCCCTACA3′, R5′GGTGACCGTCCTGGCTTTTA3′, Ta 58°C, 477 bp; CXCR1, F5′GTGATGCTGGTCATCTTATACAG3′, R5′TTGTTTGGATGGTAAGCCTGG3′, Ta 52°C, 230 bp; CXCR2, F5′CGAAGGACCGTCTACTCATC3′, R5′AGTGTGCCCTGAAGAAGAGC3′, Ta 53°C, 519 bp; CXCR3, F5′GGAGCTGCTCAGAGTAAATCAC3′, R5′GCACGAGTCACTCTCGTTTTC3′, Ta 53°C, 200 bp. Each sample was run in parallel with cyclophilin primers acting as a positive control (33). Polymerase chain reaction products were visualized using agarose gel electrophoresis.
Capture Enzyme-Linked Immunosorbent Assay
Primary HBECs (Passage 2) were cultured on collagen-coated 24-well plates until confluent and treated for 48 hours with cytokines at the following concentrations: TGF-β at 25 ng/ml, TNF at 25 ng/ml, and IFN-γ at 100 ng/ml (R&D Systems). Culture supernatants were then collected and frozen at −20°C until further analysis. Chemokines were measured by sandwich ELISA (R&D Systems) according to the manufacturer's protocols. The standard range for all 4 chemokines tested was between 0.03 and 8 ng/ml. Only absorbance values within the linear part of the standard curve corresponding to supernatants and/or diluted supernatants were used for each experiment. For experiments on polarized secretion of chemokines, transwell polyester membrane inserts (Corning Costar, Buckinghamshire, UK [0.4-μm pore, 12-mm diameter]) were first coated with rat collagen (Sigma-Aldrich; 0.005 % w/v, 1 hour at room temperature) and then with human fibronectin (Sigma-Aldrich; 5 μg/ml, 1 hour). Primary HBECs were grown to confluence (∼1 × 105 cells/cm2) with a culture media change every 2 to 3 days and incubated for 2 days postconfluence before treatment.
Chemokine receptor expression on hCMEC/D3 cells in the absence or presence of cytokines (25 ng/ml TNF and 100 ng/ml IFN-γ for 24 hours) was determined using a panel of fluorescently labeled antibodies (R&D Systems) as previously described (33). Briefly, hCMEC/D3 cells were grown to confluence, washed, and trypsinized using 0.25% trypsin/EDTA (Invitrogen). Cells were fixed using 1 ml 4% p-paraformaldehyde in PBS for 10 minutes at 4°C and then centrifuged at 300 × g for 5 minutes. Cells were then permeabilized using 0.1% Triton X-100 in PBS for 1 minute at room temperature, centrifuged at 300 × g for 5 minutes, resuspended in 1 ml of blocking solution (0.1 mg/ml of human IgG/10% normal goat serum in PBS), and incubated for 30 minutes at 4°C. Cells were counted and resuspended at 8 × 106 cells/ml. For the assay, 25 μl of the cell suspension (2 × 105 cells) was added to 10 μl of appropriate antibodies at the manufacturer's recommended concentrations. Appropriate isotype-matched controls were used. Cells were incubated with antibodies for 1 hour at 4°C then washed once using PBS and resuspended in 0.4 ml PBS for analysis. Flow cytometry data were acquired and analyzed using the FACScalibur flow cytometer and CellQuest software (Becton Dickinson, Oxfordshire, UK).
Significance was determined by a 1-way or 2-way analysis of variance, followed by a post hoc Tukey t-test; p < 0.05 was considered significant. For CXCL8 production by HBECs, the data were not normally distributed, and a Kruskal-Wallis and Mann-Whitney test was used.
Expression of Chemokines by Human Cerebral Endothelium in MS
Endothelial cells in nondemyelinated, NAWM, and cortical gray matter showed a vesicular staining pattern for CCL2 (monocyte chemoattractant protein 1; Fig. 1A) and CXCL8 (interleukin 8; Fig. 1H). Higher-power magnification (Fig. 1A' and A") illustrates this expression more clearly for CCL2. These chemokines, CCL3 (macrophage inflammatory protein 1α), CCL5 (RANTES), and CXCL10 (IP-10), were detected on the surface of endothelial cells and/or on surrounding cells, but there were differences in the patterns of expression between nonplaque and plaque tissue. Antibody to CCL2 heavily stained endothelial cells in large vessels and surrounding cells within the demyelinated plaque (Fig. 1C); expression in adjacent nonaffected cortical areas (white and gray matter) was patchy on the vasculature and was sometimes associated with perivascular cells (Fig. 1B, asterisk). The localization of CXCL8 closely matched that described for CCL2 within the plaque area (Fig. 1I) and in adjacent nonaffected regions (not shown). Similarly, CCL5 showed a finely peppered, patchy distribution at the surface of endothelial cells (Fig. 1F). In contrast, CCL3 staining in the plaques was primarily detected on fine processes radiating in a perpendicular direction away from the blood vessels (Fig. 1D). This pattern of immunoreactivity was seen to a much lesser extent on endothelial cells of blood vessels in nonaffected areas (Fig. 1E). The CCL3 staining pattern suggests that it is primarily associated with astrocyte processes and end-feet rather than the endothelium or other perivascular cells. CXCL10 was more diffusely expressed on vascular endothelium in plaques (Fig. 1G). These results implicate human brain endothelial cells as a source of chemokines in vivo both in resting (CCL2, CXCL8) and inflammatory (CCL2, CCL5, CXCL8, CXCL10) conditions.
Chemokine Secretion by Primary HBECs and hCMEC/D3 Cells
Chemokine levels constitutively released into the culture medium by cultured primary HBECs isolated from 5 MS and 3 epileptic tissue donors and by hCMEC/D3 cells were determined by ELISA (Fig. 2). Of the 5 chemokines tested, only CCL2 and CXCL8 were constitutively produced by all cells tested. This finding corroborates the immunohistochemical observation that only these 2 chemokines are present in the NAWM in situ. Basal CCL2 levels released by primary HBECs from MS brain tissue ranged from 1.4 to 22.7 ng/ml, which was similar to the range of amounts produced by HBECs from epilepsy resection tissue, that is, 4.4 to 14.3 ng/ml. CXCL8 constitutive levels were more variable; levels in primary HBECs from MS brain tissue ranged from 3.2 to 206 ng/ml, and from epileptic tissue, HBECs ranged from 45 to 142 ng/ml. Endothelium from both sources may be considered “resting” because they had been removed from potential microenvironmental proinflammatory stimuli for at least 1 week in vitro before they were assayed.
After cytokine stimulation, chemokine release into the culture medium increased to similar levels in cultures of primary HBECs from both MS brain tissue and temporal lobe resections (Fig. 2), but specific cytokines or cytokine combinations differentially affected chemokine secretion. The most potent activator of CXCL8 and CCL2 production by primary HBECs was TNF alone, but the effect of TNF on CXCL8 secretion was effectively blocked by coincubation with either IFN-γ or TGF-β. Both CXCL10 and CCL5 were maximally induced by IFN-γ in combination with TNF, whereas IFN-γ alone was sufficient to induce a considerable increase in CXCL10 production. CCL3 was not released into the culture medium either basally or in any of the cytokine treatment conditions tested (data not shown). This correlates with the results in situ in which CCL3 seemed to be associated with astrocytes rather than endothelial cells.
The hCMEC/D3 cells largely exhibited the same pattern of chemokine secretion as primary HBECs with 2 exceptions: 1) the basal levels of CCL2 and CXCL8 secretion were lower than those observed for primary HBECs and 2) TNF alone was sufficient to induce CXCL10 levels (Fig. 2).
Because chemokines are presented to circulating leukocytes, we investigated whether primary HBECs grown on filters were polarized in their secretion either constitutively or after stimulation with TNF and IFN-γ. Constitutive CCL2 and CXCL8 levels were similar in both the upper and lower chambers (Fig. 3). In contrast, after stimulation with TNF and IFN-γ, the concentrations of all chemokines were higher in the upper than in the lower chamber, although increased apical levels were statistically significant only for CCL2, CXCL8, and CXCL10 (Fig. 3). These results indicate that cytokine-induced chemokine secretion was preferentially directed to the apical side of the endothelium.
Chemokine Receptor Expression by Cultured Brain Endothelial Cells
Chemokines released into the circulation might act on brain endothelial cells in an autocrine manner; alternatively, the receptors may be involved in the clearance of free chemokines from plasma. Therefore, we investigated the expression by cultured brain endothelial cells of the chemokine receptors to which the previously investigated chemokines bind, namely, CXCR1-3 and CCR1, CCR2, and CCR5. Using semiquantitative RT-PCR, primary HBECs were found to express CCR1, CCR5, and CXCR1-3 mRNA but not CCR2 (Fig. 4A). At the protein level, CXCR1 and CXCR3 immunoreactivities were present at high levels, whereas those of CXCR2 and CCR5 were lower and CCR1 and CCR2 were not detected by immunocytochemistry (Fig. 4B). CXCR1 and CXCR3 seemed mainly to be localized within the cells, particularly around the nucleus. The patterns of chemokine receptor expression did not differ between cells originating from MS brains and those obtained from epileptic patients (data not shown).
The patterns of chemokine receptor expression of the hCMEC/D3 cell line were similar. CXCR1 and CXCR3 showed the highest levels of expression among the chemokine receptors tested. We used this cell line to quantify changes in chemokine receptor expression induced by cytokines using FACScan analysis. After incubation of hCMEC/D3 with TNF and IFN-γ for 24 hours, only CXCR3 expression was significantly increased (p < 0.05; n = 3), whereas CCR1 expression was induced (p < 0.05; n = 3). No changes in expression were observed for CXCR1, CXCR2, CCR2, or CCR5 (Fig. 5).
Chemokine Receptors in MS Brains
Because CXCR1 and CXCR3 seemed to be expressed at high levels by cultured HBECs and CXCR3 was upregulated by cytokine stimulation, we analyzed these chemokine receptors in MS brain sections by immunohistochemistry. CXCR1 and CXCR3 immunoreactivities in active (CD68+, MHC-II+) and silent (CD68+, MHC-II−) lesions were compared with those in NAWM (CD68-, MHC-II−). Double labeling with anti-vWf and either CXCR1 or CXCR3 revealed chemokine receptor staining associated with blood vessels in NAWM (Fig. 6A). Additionally, other cells within the brain parenchyma that were negative for vWf were also positively labeled for CXCR1 and CXCR3. Because human astrocytes have been previously shown to express CXCR1 and CXCR3 (29), double labeling with either CXCR1 or CXCR3 and GFAP, a specific marker for astrocytes, was performed to determine whether the positive chemokine receptor expression by blood vessels was due to associated astrocytic end-feet. Clear colocalization of CXCR1 (Fig. 6B) and CXCR3 (not shown) with GFAP was observed in the astrocytic processes surrounding blood vessels. In addition, cells enclosed within the astrocytic end-feet were also positively labeled for CXCR1 (Fig. 6B) and CXCR3 (not shown), although whether these cells were endothelial cells could not be ascertained. In MS lesions, CXCR1 (not shown) and CXCR3-positive (Fig. 6C) stainings were diffuse throughout the lesion, probably due to astrogliosis and/or leukocyte infiltration.
Subcellular Localization of CXCR1 and CXCR3 by Immunogold Labeling and Electron Microscopy
To determine whether endothelial cells expressed chemokine receptors in situ, the immunogold technique using monoclonal antibodies to CXCR1 and CXCR3 was performed on sections of MS brain tissue. Immunogold labeling with the monoclonal CXCR1 and CXCR3 antibodies revealed gold particles along the plasma membrane and cytoplasm of the endothelial cells, pericytes, and astrocytic end-feet (Fig. 7A, B, D). Leukocytes, whether infiltrated or interacting with the endothelial cells on the luminal side, were also labeled for CXCR1 (Fig. 7C) and, to a lesser extent, CXCR3 (not shown).
In the context of endothelial cells, labeling was easily identified at both the luminal and abluminal plasma membranes and within the cytoplasm of the capillary endothelial cells (Fig. 7D). The total numbers of gold particles on endothelial cells per surface area for CXCR1 and CXCR3 did not vary significantly among NAWM, active lesions and silent lesions, with the exception of CXCR3 in silent lesions, which was significantly lower than in NAWM or active lesions (Table 2). In NAWM, CXCR1 and CXCR3 antigenic sites were mainly localized at the cytoplasm; in particular, for CXCR1, approximately 87% of gold particles were located in the cytoplasm compared with 65% for CXCR3 (Table 2). The distribution of antigenic sites between the luminal and abluminal membranes was different for CXCR1 and CXCR3, with a higher percentage of gold particles on the luminal membrane compared with the abluminal membrane for CXCR1; the opposite was observed for CXCR3 (Table 2). The endothelial subcellular distribution of CXCR1 in MHC-II+ (active) and MHC-II− (silent) lesions did not differ significantly from that observed in NAWM. In contrast, a significant reduction in CXCR3 immunolabeling on the luminal membrane and a significant increase in CXCR3 immunolabeling within the cytoplasm were detected in MHC-II+ lesions (Table 2). Control experiments performed to assess labeling specificity showed a negligible number of gold particles randomly distributed when the primary antibody was omitted.
Many chemokines are found in the CNS in MS patients (11), but the chemokines that are expressed on the luminal surfaces of the endothelium likely control the patterns of leukocyte migration into the CNS parenchyma. Chemokines may be synthesized by the endothelium (19) or produced within tissues and transported across the endothelium in transport vesicles, including caveolae (22, 26). Transcytosis is likely to be important in tissues such as the lung in which the bulk transport systems are well developed (34). In contrast, the endothelial barrier is strong and transcytosis is limited in the CNS, and chemokine secretion by the endothelium itself is likely to be more important. Endothelia from different tissues vary in their chemokine secretion profiles and the rate of chemokine clearance from cell surfaces (35). Moreover, chemokine binding to the cell surface depends on the glycocalyx; brain microvascular endothelium has a particularly high negative charge due to its sulfate glycosaminoglycans that can interact with and retain positively charged chemokines (36). Thus, identifying chemokine production by brain endothelium is particularly important for understanding the distinctive patterns of leukocyte migration that occur in CNS inflammatory conditions.
We demonstrated that CCL2 and CXCL8 are produced and secreted by resting brain endothelium in vitro and by endothelium in normal-appearing brain tissue in situ. These chemokines, along with CXCL10 and CCL5, are induced after activation by inflammatory cytokines in endothelial cells in vitro and in areas of inflammation and demyelination in MS tissue. Chemokine production by brain endothelium is distinct from that in other endothelial subtypes, including primary microvascular endothelium from lung, dermis, and liver and saphenous vein endothelium (19). Our results are in agreement with previous reports demonstrating production of CCL2 and CXCL8 by HBECs isolated from temporal lobes of epileptic patients under resting conditions (37) and after stimulation with cytokines (38) or with supernatants derived from allogenic or myelin basic protein-reactive TH1 cells (39). We found that the chemokine profiles from primary brain endothelium were similar, regardless of whether the cells came from MS patients or temporal lobe resection; they were also broadly similar to the results with the cell line hCMEC/D3. This suggests that in MS, HBECs do not show increased chemokine production per se but rather respond appropriately to the inflammatory environment to which they are exposed. The levels of production of CXCL8 and CXCL10 by brain endothelium were, however, higher than in non-CNS tissue-derived endothelia (19). The finding with CXCL10 is notable because this chemokine acts on CXCR3, which is strongly expressed on activated TH1 cells, that is, the population that is believed to be critical for in MS pathogenesis.
In contrast to other chemokines, CXCL8 secretion by HBECs was more variable in resting and cytokine-stimulated cells. Both epilepsy tissue- and MS-derived HBECs seemed to secrete variable amounts of CXCL8 depending on the individual donor rather than on tissue type. This variability has also been observed in response to the human immunodeficiency virus 1 protein tat (40) and in endothelial cells from non-CNS tissues in response to cytokines (19), and it is possible that the number of cell divisions is a critical factor regulating the storage of CXCL8 in endothelial cells (41). In our primary cultures, the number of cell divisions at the time of the assay likely varied between donors because the yield of capillary fragments, and thus, the number of cell divisions to attain confluence is dependent on many factors, including postmortem time, quantity of tissue, and cause of death. Another possibility involves interindividual variation among different donors. Indeed, in a recent study, CXCL8 plasma levels of healthy blood donors varied greatly (42), and variability has also been demonstrated among different ethnic groups (43).
The downregulatory effect of IFN-γ and TGF-β on TNF-induced upregulation of CXCL8 secretion is notable. TGF-β is considered to be an anti-inflammatory cytokine, and it has been detected in active MS lesions (44). CXCL8 inhibition by TGF-β was only partial, however, suggesting that it may not exert its anti-inflammatory effects in the presence of high levels of proinflammatory cytokines. The inhibitory effect of IFN-γ on TNF-induced CXCL8 production has been reported in other cell types such as monocytes (45). Activation of nuclear factor-κB and activating protein 1 by TNF is required for CXCL8 transcription and is inhibited in the presence of IFN-γ in endothelial cells (46). The suppression of CXCL8 production by IFN-γ might thus be considered a protective effect against CXCL-8-mediated neutrophil infiltration. Another important finding in the present study is that brain endothelium does not produce CCL3 in vitro. This is consistent with the observations in situ, and taken together, the findings suggest that astrocytes rather than endothelial cells are the main source of CCL3. These results demonstrate the importance of correlating in vitro and in situ results.
How does chemokine expression by HBECs relate to leukocyte infiltration? Apical release of chemokines by brain endothelium in vitro may result in rapid dilution of a chemokine by the blood stream in vivo. Indeed, chemokines released in the circulation would be rapidly degraded by proteases or their actions neutralized by decoy chemokine receptors such as the Duffy antigen receptor for chemokines on erythrocytes (47). Alternatively, chemokines released by endothelial cells may bind to the endothelial glycocalyx either on the secretory cell itself or on other endothelial cells further along the capillary wall, thereby trapping immune cells within the inflamed area either at the lumen or within the perivascular space. This scenario might apply to CXCL10 because high levels of its receptor are detected on the abluminal side of brain endothelial cells in situ. Indeed, previous studies stress the importance of chemokines in directing leukocyte trafficking into the CNS. In chronic relapsing EAE in mice, disease severity correlated with CCL3 production during the initial acute phase but more closely with CCL2 levels during relapse (48). These observations and the fact that CCL2-null mice do not develop EAE indicate a potential role for CCL2 in the development of neuroinflammation (18). Whether CCL2 is essential in EAE has, however, been questioned by other studies that show that EAE can develop in a number of CCR2-deficient mice (49). Neuropathologic examination in this model showed a higher proportion of neutrophils and fewer macrophages than in normal animals, suggesting that macrophages can be partly replaced by neutrophils in EAE, although it still implies that CCL2/CCR2 interactions are important for monocyte migration into the CNS. The levels of CCL3 in the CSF in MS patients have a weak positive relationship to the level of cells present (50), but this may merely reflect the fact that inflammation activates astrocytes to produce CCL3 and does not necessarily imply that CCL3 is required to drive leukocyte transmigration. Indeed, other evidence suggests that CCL2 is also important in controlling monocyte migration in MS, and that migrating cells lose their CCR2 receptor as they transmigrate (10).
There is also considerable evidence that at least some endothelial-derived chemokines regulate leukocyte trafficking into the CNS, specifically of TH1 cells. For example, supernatants from TH1, but not TH2, cells induce production of CXCL10, CXCL8, and CCL2 by HBEC (39). Using an in vitro human BBB model, Prat et al (37) demonstrated that antibody neutralization of CCL2 considerably reduced migration of T lymphocytes isolated from MS patients across HBECs, and CCL2 has been shown to be crucial for the TH1 immune response in EAE (17). Using intravital microscopy, these studies were expanded to show that treatment with anti-CCL2 or anti-CCL5 antibodies prevent leukocyte adhesion, but not rolling, in EAE (51). Our present results indicate that brain endothelium itself is a major source of CCL2 and CXCL10, both of which have been implicated in the development of TH1-type inflammatory reactions in MS.
The expression of chemokine receptors on brain endothelium as determined by fluorescence microscopy and fluorescence activated cell sorter analysis shows some similarity to other endothelia with high expression of CXCR1 and CXCR3. Our findings are generally in agreement with those of Berger et al (28), who showed expression of CXCR1-3 by HBECs, although the expression of CXCR2 was low in the present study. In addition, we confirmed a low expression of CCR5 in agreement with other in vitro (28, 52) and in situ (53) studies. In contrast, although CCR2 expression has previously been reported in HBECs (28), and more recently also in mouse brain endothelial cells at the protein and mRNA level (54), we could not demonstrate CCR2 expression either at the protein or transcript level. Species differences in CCR2 expression cannot be ruled out at present. In the study by Berger et al (28), HBECs were positive by immunofluorescence using goat polyclonal antibodies to CCR2A but not with those to CCR2B, whereas we used a more specific CCR2 monoclonal antibody. In addition, Andjelkovic et al (52) showed binding sites to CCL2 in isolated human brain capillaries, although this may either be due to other cell types expressing CCR2 such as perivascular cells/pericytes or to other endothelial chemokine receptors such as the Duffy antigen receptor for chemokines.
The function of endothelial cell chemokine receptors has not been defined. Interestingly, CXCR1 and CXCR3 bind to CXCL8 and CXCL10, respectively, which might allow secreted chemokines to act in an autocrine fashion. Previous studies have suggested that signaling via CXCR1 is angiogenic, whereas signaling via CXCR3 is angiostatic (27). Another potential function for endothelial chemokine receptors is to clear the plasma of free chemokines so that leukocytes do not become activated unless they are triggered by chemokines held on the endothelial glycocalyx. Finally, it has been proposed that chemokine receptors could be involved in the transport of chemokines across the endothelium. Our observation that the subcellular localization of CXCR3 is altered in MS lesions compared with NAWM is consistent with this hypothesis. Indeed, there seemed to be an increase in the intracellular pool of CXCR3 and a decrease in the abluminal endothelial cell membranes in MS lesions. Because CXCL10 induces the internalization of its receptor (55), the observed increase in intracellular CXCR3 in MS lesions where CXCL10 is increased might explain the decrease of the receptor on the abluminal side. CCR2 has been suggested to act as a transporter of its chemokine ligand (CCL2) across the BBB (56), and it is possible that the same applies for CXCR3 and its ligand because they are greatly increased in the lesions, particularly active lesions. Indeed, we found that CXCR3 is upregulated by cytokine treatment in a human brain endothelial cell line. Our in vitro results should, however, be approached with caution because hCMEC/D3 cells do not retain all characteristics of the BBB phenotype (e.g. high transendothelial resistance ), and the pattern of chemokine expression was somewhat different between primary HBECs and the immortalized cell line. Whether CXCR3 transports CXCL10 across human brain endothelium remains to be determined.
In summary, we have shown that human brain endothelium cultured in vitro expresses a similar pattern of chemokines and chemokine receptors to that seen in situ. Brain endothelium responds to cytokine stimulation by secreting chemokines and does so with the same response pattern as other microvascular endothelium. The results also highlight higher secretion of CXCL8 and CXCL10 by HBEC than by other endothelia, which might contribute to the distinct TH1 cell pattern of chronic inflammation in the brain in diseases such as MS.
The authors thank Dr. R. Selway and staff at King's College Hospital, London, the UK Multiple Sclerosis Tissue Bank, and the MRC London Brain Bank for Neurodegenerative Diseases (Dr. Safa Al-Sarraj, Clinical Director) for provision of human tissues for use in investigations.