Background
The blood brain barrier (BBB) is a unique astrocyte-capillary-endothelial complex which maintains CNS homeostatic fluid balance, and serves as a first line of defense protecting the brain and parenchyma against pathogens, as well as blood-borne leukocytes and hormones, neurotransmitters and pro-inflammatory cytokines and chemokines [
1,
2]. The loss of BBB structural integrity and function plays a central role in the pathogenesis of neuroinflammatory diseases like multiple sclerosis, Alzheimer's disease, meningitis, brain tumors, intracerebral hemorrhage and stroke [
3‐
10]. Many reports in the literature indicate that loss of BBB in neuroinflammation represents a result of complex often continuous interactions between the BBB and immune cells, adhesive determinants and inflammatory cytokines, all of which may be relevant targets for therapy [
11‐
18]. While several studies have modeled interactions between astrocytes and brain endothelial cells, fewer studies have considered how this gliovascular unit might be dysregulated by the combined influences of metabolic stress and cytokine exposure.
Astrocytes are the most abundant glial cells in the CNS, playing crucial roles in cerebral ion homeostasis, neuro-transmitter regulation, structural and metabolic support of neuronal and endothelial cells and BBB maintenance [
19‐
21]. Furthermore, astrocytes provide an important link between neuronal and vascular units in the glucose-lactate shuttle and in modulating Ca
2+ responses [
22‐
29]. Importantly, astrocytes have been shown to play divergent roles in various pathologic conditions [
29‐
32]. For example, following ischemic strokes, astrocytes protect neurons [
33‐
35] by secreting several neurotrophic factors like glial cell-line derived neurotrophic factor [
36], neurotrophin-3 [
37,
38], transforming growth factor-β1 [
39], and vascular endothelial growth factor [
40]. Astrocytes can also secrete pro-inflammatory cytokines such as TNF-α, IL-1β, and IL-6 which would be anticipated to aggravate inflammatory injury to ischemic tissues [
41]. The roles played by astrocytes and astrocyte-derived factors in maintaining or injuring the post-ischemic BBB are complex, cell-specific and time-dependent. Several reports have indicate that astrocytes co-cultured with endothelial cells or astrocyte-conditioned media improve endothelial barrier integrity, however the potential effects of astrocytes on the cerebral endothelial cells during CNS stress contributing to the pathological loss of BBB are not yet as well understood [
20]. The mechanisms through which factors secreted by stressed astrocytes (e.g. in response to glucose, serum, or oxygen deprivation) dysregulate endothelial barrier during pathologies e.g. cerebral ischemia remains an area under intensive investigation [
42].
Cytokines exert diverse and cell-specific effects on BBB integrity [
43‐
46]. TNF-α and IFN-γ are among the best studied cytokines which cause differing permeability responses in different cell systems [
47]. For example, IFN-γ was shown to increase permeability in human colonic epithelial cells (T84), microvascular endothelial cells, human umbilical vein endothelial cells and cholangiocytes, but decreased permeability in human lung epithelial cells (Calu-3). TNF-α increases permeability of bovine pulmonary artery endothelial (BPAEC) monolayers, human colonic adenocarcinoma (Caco-2), HT29/B6 and cholangiocytes, but decreased solute permeability of uterine epithelial cells (UECs) [
47]. Further, TNF-α can either increase or decrease solute exchange depending on the type of insult in porcine renal epithelial cells (LLC-PK1) [
48,
49]. These effects are mediated by diverse mechanisms involving actin reorganization, monolayer motility, NF-kβ activation, apoptosis and reorganization of junctional proteins [
49‐
54].
Apart from direct actions of cytokines, factors secreted by astrocytes may also disturb BBB [
32,
42]. For example, matrix metalloproteinases ('MMP') -9 (MMP-9) and -13 (MMP-13), derived in part from astrocytes may contribute to post-ischemic BBB dysregulation [
55‐
57] and MMP-9 inhibition partially protects against ischemic stroke, decreasing infarct size and BBB breakdown. Conversely, Tang et al. have reported that MMP-9
-/- mice exhibit a more pronounced BBB damage and edema than controls (in a collagenase model of hemorrhage) [
58]. Many other mediators may be involved in mediating the deleterious effect of stressed astrocytes on BBB during pathological conditions.
In the present study we investigated the direct or indirect influence of cytokines (TNF-α, IL-1β and IFN-γ) on brain endothelium and astrocytes (individually or in synergy) on barrier during metabolic stresses using a 3-D in vitro BBB model with human, mouse brain endothelial cells, ECV-304 and astrocytes. The results of our current study indicate that under conditions of pathological stress, astrocytes indirectly modify endothelial barrier responses to cytokines, leading to strikingly different barrier conditions observed in the absence of astrocytes. The differential roles of astrocytes and cytokines in modulating brain endothelial barrier properties are also discussed.
Materials and methods
Reagents
Mouse rTNF-α, was purchased from Endogen (Woburn, MA) Thermo scientific (Rockford, IL), Mouse rIL-1β was purchased from Chemicon (Temecula, CA) or Endogen. Mouse rIFN-γ was purchased from Endogen. Human rTNF-α and rIFN-γ were purchased from Thermo-scientific. Human rIL-1β was purchased from Endogen. All other chemicals were purchased from Sigma (St. Louis, MO) unless specified.
Cell culture
Murine brain endothelial cells (bEnd.3) provided by Dr. Eugene Butcher (Stanford Univ.). Human fetal astrocytes (HFA) were provided by Dr. Danica Stanimirovic (Univ. of Ottawa). Both cell types were both cultured in DMEM supplemented with 10% fetal calf serum (Hyclone) and 1% Penicillin-Streptomycin-Amphotericin (PSA) ('complete medium' referred as 10% DMEM). Media were changed every 2
nd day. Human brain endothelial cell line (HBMEC-3) was kindly provided by Dr. Anat Erdreich-Epstein, (Children's Hospital of Los Angeles, California) and were cultured in RPMI with 10% FCS with 2 mM sodium pyruvate and 1% PSA. An additional human brain endothelial cell line (HCMEC-D3) was provided by Dr. P.O. Couraud, (Institut Cochin, Paris, France) [
59,
60]. HCMEC-D3 cells were cultured in rat tail collagen coated plates (100 ug/ml) in medium consisting of EBM2 supplemented with 5% FCS, 1.4 uM hydrocortisone, 10 mM HEPES, 1 ng/ml bFGF and 1% PSA. As an additional control, ECV-304, (ATCC, Manassas, VA) a bladder carcinoma with several endothelial-like properties was also used in this study [
61]; (these cells were cultured as described for HBMEC-3.)
In vitro barrier function studies
Brain endothelium (and ECV-304) was cultured on the apical surface of 8.0 μm PETP transwell inserts (Falcon) placed in a 24-well culture plates ('outer chamber'). The outer chamber contained 1 ml of medium with 0.5 ml media in the insert. To generate contact-independent co-cultures, the apical/inner surface of the insert was seeded with either human or mouse brain endothelial cells or ECV-304 cells; astrocytes were cultured in the basal/outer chamber.
To create a 'close-contact' co-culture system closely resembling the in vivo gliovascular unit, after human or mouse endothelial (HCMEC-D3 or bEnd-3) cells were cultured on the apical surface and astrocytes were cultured on the basal side of the insert. These cultures were established by allowing 100 μl of astrocyte cell suspension (approximately 20,000 cells) to adhere to the basal surface for 1 hr before seeding the apical surface of the insert with endothelial cells. Later, inserts with attached endothelial cells and astrocytes were transferred into the outer chamber.
Trans-endothelial electrical resistance (TEER)
Trans-endothelial electrical resistance was measured using an epithelial volt-ohmmeter (EVOM) (World precision instruments, Sarasota, FL). Cultures systems on inserts were exposed to treatments, and at time points, were transferred to the TEER chamber (using matching media conditions) and electrical resistance recorded (ohms/cm2, no = ohms/0.332).
Brain endothelial barrier permeability
Mouse brain endothelial cells (bEnd3) were grown in transwell inserts (apical side) and at confluence were treated with cytokines in both apical and basal sides. TEER was recorded at 24 h time intervals. At 3 d, 50 μl of FICT-dextran (120 kD) at a final concentration of 1 mg/ml (in culture medium) was added to the apical side of the brain endothelium. At various time points from 30 min to 6 h, 100 μl of medium from the basal chamber was used to measure the extravasated FITC-dextran to the basal side across the endothelium. Equal volume of media was supplemented to replace the volume of used medium. The experiment was terminated after 6 h. All the readings were measured at constant 'gain' settings. The values obtained were plotted on graph pad and checked for significance.
Cytokine treatments
Murine brain endothelial cells and human astrocytes were treated with matching mouse or human TNF-α (20 ng/ml), IL-1β (20 ng/ml) and IFN-γ (1000 U/ml) respectively. Depending on the study, cytokines (at specified concentrations) were added either to the apical or basal surface surrounding the insert (in contact-dependent or contact-independent systems).
MTT assay
Brain endothelial cells were grown in 96-well plates. At confluence, human and mouse brain endothelium was incubated with matching TNF-α (20 ng/ml), IL-1β (20 ng/ml), IFN-γ (1000 U/ml) for 4 d. At the end of incubation time period, cell energy metabolism was measured by washing cells 3X, and extracting in 300 ul of acetic acid/isopropanol. Absorbance of the acid/isopropanol-extracted products was then measured at 450 nm.
Statistics
Graphpad-3 InStat™ software was used to perform statistical analyses. One way-ANOVA or repeated measures ANOVA each with Dunnett's' post-hoc test or Bonferroni post-test were used to determine statistical significance. Sigmaplot™ was used to generate plots. *p < 0.05 was considered to be statistically significant, **p < 0.01 very significant, and ***p < 0.001 highly significant.
Discussion
The neurovascular unit is a highly organized functional complex composed of neurons, their associated glia and microvessels which match cerebral blood flow with metabolism [
19,
62‐
64]. This unit is further divided into gliovascular units in which astrocytes support the function of neurons and communicate with the associated microvasculature. Astrocytes play a central role in integrating this functional unit. These neuro- and gliovascular units sense changes in local metabolism and synchronize functions between the involved cell types during normal physiological regulation [
62,
65]. However, during pathological conditions the cumulative influences of several internal and external factors may significantly alter this balance, to compromise the normal BBB. Dysregulation of the BBB appears to be a critical step in the pathogenesis of many CNS disturbances. Severely compromised BBB function is observed in many clinical conditions including brain trauma, ischemic stroke, meningitis, glioma, Alzheimer's disease and multiple sclerosis [
3‐
10]. Such disruptions in the BBB play a pivotal role in aggravating many forms of cerebrovascular pathology by intensifying inflammatory responses within the CNS environment [
66].
IFN-γ has been reported to decrease endothelial barrier [
52,
67‐
69], however it is worth noting that most of these studies have been performed in non-CNS endothelial cells. Brain endothelial cells differ from other endothelial cells in many respects including highly organized tight junctions which restrict paracellular transport and depend on biochemical support and interaction with astrocytes and neurons [
70,
71]. We attempted to identify specific responses involving interactions between astrocytes, individual cytokines, individually and in combination, to isolate possible mediators of barrier dysregulation in cell- and cytokine-mediated pathological conditions. Interestingly, our present study found unique brain endothelial responses to astrocytes and cytokines (compared to other endothelial types). Treatment with cytokines (i.e. TNF-α, IFN-γ, IL-1β) did not reduce barrier, compared to controls and paradoxically, TNF-α (on mouse brain endothelium) and IFN-γ somewhat enhanced barrier in mono-culture conditions. The effect of these cytokines on brain endothelial barrier (also on ECV-304) persisted for 7 days. These results differ from some, (but not all) previous reports, and may reflect complex, cell- and species-specific interactions.
For instance, Wong et al., observed decreased electrical resistance in human 1° endothelial cultures after treatment with 500 U/ml of IFN-γ [
69]. We also observed a similar decrease in barrier when astrocytes (but not endothelial cells or ECV-304 alone) were treated with IFN-γ (in co-culture). Importantly, the observed barrier tightening effect of IFN-γ was eliminated and reversed when astrocytes were treated with IFN-γ in co-culture. This clearly shows that factors released by astrocytes exposed to IFN-γ (but perhaps not IFN-γ directly on endothelial cells) may trigger endothelial signaling and barrier breakdown. This finding indicates that negative barrier effects of IFN-γ on endothelial cells may be indirect, and reflect the production of factors produced by the astrocytes in our study. Stressed astrocytes may secrete several classes of factors, acting on brain endothelial cells (and other barrier forming cells, e.g. ECV-304) to compromise barrier. Activated astrocytes are known to release several factors like MMPs, that are involved in barrier breakdown [
72‐
74]. Clear differences in the effect of cytokines on barrier are seen in different sets of conditions in the present study. For example, while some reports suggest that IL-1β dysregulates barrier [
75], we found that barrier was maintained in brain endothelial monolayers treated with IL-1β (not different from controls). Moreover, when both astrocytes and brain endothelial cells were treated with cytokines in co-culture, trans-cellular resistance of co-cultures treated with TNF-α or IFN-γ were lower than controls indicating that astrocyte stimulation is required for barrier dysregulation rather than cytokines alone. Similar results were also found for ECV-304 cells. IFN-γ mediated barrier dysregulation involves a specific action on astrocytes rather than a direct effect on the brain endothelium (Figure
2,
4b,
5b and
6b). These results indicate that the specific actions of TNF-α in brain endothelial barrier dysregulation involves a synergy between endothelium, astrocytes and astrocyte-secreted factors and suggests that IFN-γ indirectly dysregulates barrier/permeability through activation of astrocytes.
To determine if TEER changes might parallel changes in cell energy metabolism, mitochondrial respiration was measured in both human and mouse brain endothelium upon exposure to cytokines for 4 days using MTT. In normal medium, brain endothelial cells were metabolically active and TNF-α and IFN-γ each significantly depressed metabolism of both mouse and human brain endothelial cells at days 3 and day 4 (significant change in metabolism vs. controls). These results indicate that the increase in barrier seen in human and mouse cells does not reflect metabolic depression. Moreover, it is possible that this decreased endothelial cell metabolism might be an adaptive response against cytokines which protects the barrier by conserving energy and preventing cell border contraction. This effect seems more prominent in IFN-γ treated brain endothelium. Therefore, IFN-γ may either modifies extracellular matrix (ECM) composition or alters endothelial junctions to prevent barrier dysregulation, a phenomenon which deserves further study [
76]. Importantly, while some prior reports indicate that IFN-γ injures cells during cerebral ischemia, recent reports also indicate that IFN-γ protects neurons from CD8 T cell mediated injury [
77‐
79]. Moreover, microglia treated with IFN-γ and transplanted
in vivo protect neurons by secreting neurotrophic factors [
80]. In the same context, the observed beneficial barrier tightening effect of IFN-γ may indicate another set of positive effects of IFN-γ in BBB modulation.
The barrier of brain endothelial cells (and ECV-304) was elevated by IFN-γ in all studies, except when astrocytes were treated with IFN-γ in co-culture with endothelial cells (and ECV-304). These results indicate that cytokines (e.g. IFN-γ) may initiate different barrier responses depending on the types of cells contacted, acute vs. chronic timing, and the cytokine involved. Several studies have tried to determine mechanisms through which astrocytes modulate endothelial barrier using contact-dependent and independent co-culture models. Taking into consideration that intimate contact with astrocytes might alter endothelial barrier; we studied the effect of cytokines in a contact-dependent transwell system. Interestingly, our results were similar in both models, and might reflect species-specific differences. Porcine endothelial and rat glial cells have been shown to be a useful system for contact-dependent BBB studies [
26]. Porcine and rat cells might thus be able initiate modulating signals despite species differences, which human and mouse co-cultures may not duplicate. Therefore, to match the species specificity both human brain endothelial monocultures (HCMEC-D3, HBMEC-3) as well as ECV-304, and human brain endothelial: human astrocyte co-cultures (HCMEC-D3/HFA, HBMEC-3/HFA) and ECV-304/HFA) were prepared and evaluated for cytokine responses. Interestingly, similar responses were observed using mouse brain endothelial (bEnd-3) mono-cultures and mouse brain endothelial: human fetal astrocyte (bEnd-3/HFA) co-cultures. While TNF-α and IFN-γ induced barrier in mono-cultures, cytokine treated co-cultures showed a rapid reduction in barrier. However, a pronounced decrease in mouse, human brain endothelium and ECV-304 barrier was observed when starved human astrocytes were used in species-matched co-cultures; barrier was severely decreased by 5 d compared to controls. This indicates that stressed astrocytes strongly promote barrier breakdown which may be further aggravated by elevated cytokine levels, rather than through direct effects of these cytokines on the endothelium.
Another important aspect of this study is the apparent resistance of endothelial cells to various stressful conditions. For example, although brain endothelium are quite resistant to external forces/factors, results with stressed astrocytes show that astrocytes can disturb endothelial barrier. During CNS disorders like ischemic stroke, stressed/activated astrocytes may increase production of cytokines/proteases and intensify other factors leading to BBB failure during CNS pathologies. Pro-inflammatory cytokines like TNF-α, IL-1α, IL-1β, IL-6, GM-GSF and chemokines like MCP-1 have been implicated in several forms of BBB breakdown [
11‐
15,
17,
18,
66]. Further, cytokine mediated chemokine modulation (e.g. IL-1β driven MCP-1) has also been implicated in BBB breakdown [
81,
82]. These results indicate that cytokines indirectly affect other barrier modulators. A consistent observation of this study is that astrocytes mediate cytokine mediated BBB breakdown. The pathophysiology of many CNS disorders such as cerebral ischemia, MS, glioma and brain trauma are closely associated with increased production of cytokines in the brain. The production of these resulting cytokines can strongly activate astrocytes to release factors that dysregulate BBB. Despite a paradoxical tightening of barrier in response to IFN-γ and TNF-α, the loss of barrier due to the effect of cytokines on astrocytes indicates that these coordinated cytokine-astrocyte interactions closely regulate pathological breakdown of the BBB and are model-specific.
Ganta Vijay Chaitanya, PhD, Department of Molecular and Cellular Physiology, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130
Walter Cromer, PhD, Department of Cell Biology and Anatomy, Louisiana State University Health Sciences Center- Shreveport, Louisiana-71130
Shannon Wells, MPH, Department of Molecular and Cellular Physiology, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130
Merilyn Jennings, BS, Department of Molecular and Cellular Physiology, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130
Anat Erdreich-Epstein, MD, Division of Hematology-Oncology, Departments of Pediatrics and Pathology, The Saban Research Institute at Children's Hospital Los Angeles and the Keck School of Medicine, University of Southern California, 4650 Sunset Boulevard, Mailstop#57, Los Angeles, California 90027.
P.O. Couraud, PhD, Department of Cell Biology, Université Paris Descartes, CNRS (UMR 8104), Paris, France, Inserm, U567,
Ignacio A. Romero, PhD, Department of Biological Sciences, The Open University, Milton Keynes, UK, Department of Medicine
Babette Weksler, PhD, Weill Medical College, New York, NY USA.
J. Michael Mathis, PhD, Department of Cell Biology and Anatomy, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130
Alireza Minagar, MD' Department of Neurology, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130,
J. Steven Alexander, PhD, Department of Molecular and Cellular Physiology, Louisiana State University Health Sciences Center-Shreveport, Louisiana-71130
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
GVC conceived and performed the majority of the experiments, analyzed data and wrote the study. WEC assisted in performing experiments and revising the manuscript. SW and MJ assisted with experiments and revisions. AEE provided HBMEC- 3 cells and helped in revisions of the study. POC, IAR, BW provided HCMEC-D3 cell lines and helped in revisions of the study. MJM assisted in interpretation and revision of the study. AM assisted in the interpretation and revision of the study. JSA helped conceive, analyze and interpret data and assisted in writing the manuscript. All authors read and approved the final manuscript.