The transport mechanism mediating brain uptake of tumor necrosis factor (TNF)-α has been studied. When 125I-labeled rat TNF-α was used in internal carotid artery perfusions in rats, the cytokine showed transcytosis through the blood-brain barrier in intact form (permeability-surface area product 0.34 ± 0.13 μl · min−1 · g−1). Uptake was inhibited by low nanomolar concentrations of unlabeled rat TNF-α. Human TNF-α, which does not interact with the p80 TNF receptor in rodents, showed no brain uptake. mRNA expression of both p60 and p80 receptors could be demonstrated in native brain microvessel preparations. These transcripts increased to 149% (p60) and 127% (p80) of control 4 h after a systemic immune stimulation (2 mg/kg bacterial endotoxin ip). Lipopolysaccharide treatment did not alter the rate of brain uptake of TNF-α measured between 4 and 24 h later. In conclusion, a receptor-mediated mechanism is responsible for the transcytosis of TNF-α. Saturable transport, requiring the p80 receptor, occurs at concentrations encountered under pathophysiological conditions and therefore constitutes a relevant mechanism of communication between the immune system and the brain.
- brain perfusion
- Northern blot
systemically released proinflammatory cytokines, like tumor necrosis factor-α (TNF-α), act as signals in the complex network of immune-neuro-endocrine interactions. They cause unequivocal effects in the central nervous system (CNS), such as fever and stimulation of the hypothalamo-pituitary-adrenal axis (5, 9, 25). However, the mechanisms of transmittal from the circulation to the brain, in particular the role of the blood-brain barrier (BBB), are still not fully understood. Studies utilizing diverse experimental approaches suggest at least four distinct pathways of communication (9,17). Cytokines may act on cognate receptors on brain cells after physically entering the organ, either by 1) transport across the BBB or 2) by gaining access at circumventricular organs, i.e., at sites where the BBB is absent. Alternatively, 3) cytokines may trigger the release of second messengers (prostaglandins, nitric oxide) by brain microvascular endothelial cells and perivascular microglia. Finally, 4) cytokines may stimulate afferent nerve fibers in the periphery to send signals to the brain parenchyma.
Regarding the first mechanism, the intact mammalian BBB does not allow nonspecific uptake of proteins by diffusional transport. However, as initially demonstrated for insulin (16) and transferrin (18), brain capillary endothelial cells express receptor-mediated transport systems for peptides and proteins. These enable transcytosis of their respective ligands from blood to brain extracellular space. An analogous receptor-mediated transport of proinflammatory cytokines through the BBB could contribute to their centrally mediated effects. In this regard, TNF-α is of particular interest, as it is the first cytokine released into the circulation after stimulation of the immune system with endotoxin (6), and it triggers the inflammatory cascade in concert with interleukin (IL)-1 (14). In contrast to IL-1, TNF-α concentrations in plasma rise into the high picomolar range under pathophysiological conditions in humans and experimental animals. Brain uptake of TNF-α from the systemic circulation has been described (3, 11, 21,30), and a saturable, specific transport mechanism at the BBB has been postulated (21). Available evidence suggests that the two known TNF receptors, designated TNFR1 (receptor type-1, also known as p55/p60 receptor or CD120a) and TNFR2 (receptor type-2, also known as p75/p80 receptor or CD120b) (4, 42), are involved in the transport. These receptors are expressed on many cell types, and endothelial cells are among the primary targets of TNF-α (28). The detection of TNF receptor transcripts within the rat brain has been reported by use of in situ hybridization (29). In the basal state, a low-expression level of TNFR1 over blood vessels was described, whereas the TNFR2 signal was negative. After systemic injection of either lipopolysaccharide (LPS) or recombinant TNF-α, the message for the TNFR1 was upregulated, and TNFR2 became detectable. Recent experiments in receptor knockout mice indicated that both TNFR1 and TNFR2 are required for brain uptake of TNF-α from the circulation (32), with no uptake seen in double-knockout animals.
A local immune response or tissue damage in the CNS enhances BBB transport of TNF-α from the systemic circulation, as shown in experimental autoimmune encephalomyelitis (30) and spinal cord injury (33). It is not known whether a systemic immune stimulation would also modulate brain uptake of TNF-α. In the case of insulin, systemic endotoxin treatment apparently increased the transport rate of the peptide hormone at the BBB (47).
In the present studies, we addressed three questions: 1) whether brain uptake of TNF-α is reflected in binding of the cytokine by freshly isolated brain microvessels, 2) whether native brain microvessels express both TNF receptor types, and 3) how systemic immune stimulation with endotoxin would affect receptor expression and brain uptake of TNF-α.
The transport of TNF-α through the BBB in rats was measured with or without prior intraperitoneal administration of LPS. These transport studies were performed using the internal carotid artery perfusion technique. This method is particularly suitable for brain uptake studies of labile substances, because it eliminates systemic metabolism. The latter point is relevant with respect to TNF-α: in rats, a distribution half-life of only 5 min and a mean residence time of 24 min have been reported (45). In the present experiments, uptake into brain parenchyma was differentiated from vascular sequestration by the capillary depletion method. Binding and cellular uptake of TNF-α were also studied in receptor-binding assays with use of freshly isolated brain cortical microvessels and cell membrane preparations from these microvessels. Northern blot analysis of brain microvessel-derived RNA was applied to determine the TNF receptor gene expression in untreated rats and after systemic LPS administration.
MATERIALS AND METHODS
Male Sprague-Dawley rats (Harlan-Winkelmann, Borchem, Germany) weighing 200–300 g were used in the animal experiments. D. Gemsa (Institute for Immunology, Philipps University, Marburg, Germany) kindly provided L929 mouse fibroblasts. Recombinant rat TNF-α was purchased from PeproTech (Rocky Hill, NJ) and from the National Institute for Biological Standards and Control (Potters Bar, UK). Recombinant human TNF-α was also obtained from PeproTech. LPS from Escherichia coli 0111:B4 [lethal dose of 50% (LD50) = 12.3 mg/kg in mice] was obtained from Difco Laboratories (Detroit, MI). The rat TNF-α ELISA was obtained from Endogen (Woburn, MA), and the EZ4Y cytotoxicity assay was purchased from Biomedica (Vienna, Austria). Na125I was purchased from Amersham Pharmacia (Braunschweig, Germany) and iodogen from Pierce (Rockford, IL). If not specifically mentioned, all other chemicals were of analytical grade and were purchased from Sigma (Deisenhofen, Germany).
TNF-α was radioiodinated with the iodogen method (19) to specific activities between 600 and 3,000 μCi/nmol (referring to trimeric TNF-α). Briefly, 2 μg of TNF-α in 30 μl of 0.1 M Na phosphate (pH 7.4) were allowed to react for 3 min with 0.5 mCi Na125I in an iodogen-coated microfuge tube. The reaction mixture was transferred to a second tube with 100 μl of 1% K127I. The labeled protein was purified from free125I by Sephadex G25 gel chromatography. An aliquot of the peak fractions was tested for TCA precipitability. Only a tracer of >98% precipitability was used within 48 h of labeling. Rat serum albumin (RSA) was labeled with [3H]N-hydroxysuccinimidyl propionate, as described (7).
Bioactivity of labeled and unlabeled rat TNF-α was determined using the L929 mouse fibroblast assay. Briefly, L929 fibroblasts were seeded in 96-well plates at 25,000 cells/well and grown in RPMI 1640 with 10% fetal calf serum. After 24 h, the medium was replaced with 50 μl of fresh medium containing Actinomycin D (2 μg/ml), and 1 h later the TNF-α dilutions in 50 μl of medium were added. After 18 h, incubation at 37°C cell viability was quantified by the EZ4Y assay according to the manufacturer's instructions. To obtain the exact TNF-α protein concentration, 125I-labeled TNF-α was quantified with the ELISA specific for rat TNF-α.
Internal carotid artery perfusion technique.
Unilateral vascular brain perfusions were performed in anesthetized rats (100 mg/kg ketamine and 4 mg/kg xylazine im) via retrograde cannulation of the external carotid artery after cauterization of the occipital artery, superior thyroid artery, and pterygopalatine artery. The common carotid artery was ligated just before initiation of the perfusion. Krebs-Henseleit buffer containing 1% bovine serum albumin (BSA) equilibrated with 95% O2-5% CO2 was perfused at a flow rate of 1.25 ml/min by use of a peristaltic pump (Spetec, Munich, Germany). The perfusate contained 1 μCi/ml of125I-labeled rat TNF-α (0.33–1.67 nM, referring to trimeric TNF-α) or human TNF-α and 10 μCi/ml 3H-RSA. Inhibition experiments were performed by adding unlabeled rat TNF-α (0.83–16.7 nM) to 125I-labeled rat TNF-α tracer in the perfusate. Perfusion times of 1, 5, or 10 min were chosen. To keep intravascular volume constant during the 5-min and 10-min perfusions, blood was withdrawn at the same rate via a catheter in the femoral artery. Perfusions were terminated by decapitation. The brain was quickly removed and cleaned from meninges, and the ipsilateral hemisphere (forebrain without hypothalamus and olfactory bulb) was weighed. The tissue was gently homogenized on ice in physiological buffer, and “capillary depletion” analysis was performed, as described (43) with the following modifications: dextran (molecular mass 60–90 kDa) was added to a final concentration of 20%, and the centrifugation was performed at 3,200g for 15 min at 4°C in a table top centrifuge with swinging bucket rotor. Efficiency of separation into the vascular pellet and the postvascular supernatant was tested in a pilot series by measurement of the capillary-enriched marker enzymes alkaline phosphatase and γ-glutamyl transpeptidase. The results were comparable with those of the original protocol, with 93.5 ± 2% (alkaline phosphatase) and 93.2 ± 1.6% (γ-glutamyl transpeptidase) of enzyme activities accumulated in the pellet fraction (means ± SD, n = 4, based on enzyme activities per mg protein). The pellet and aliquots of homogenate and postvascular supernatant were digested with tissue solubilizer (Soluene 350, Canberra Packard, Dreieich, Germany) and measured in a Wallac 1210 liquid scintillation counter with a dual-isotope program (Wallac, Ohu, Finland). The results are expressed as the apparent volume of distribution, VD, which is calculated as The apparent rate of brain uptake of TNF-α can be expressed in terms of the permeability-surface area product (PS) at the BBB where T is perfusion time (1 min, 5 min, 10 min). When a unidirectional transport (influx) is assumed, the rate of uptake may then be determined by linear regression analysis of data obtained after the various perfusion periods (8). The corresponding values for the intravascular marker VD(RSA) were subtracted from the homogenate, supernatant, and pellet VD of TNF-α.
TCA precipitation of brain postvascular supernatant.
Two milliliters of ice-cold 20% TCA in water were added to 0.5 ml of postvascular supernatant, briefly vortexed, incubated on ice for 10 min, and centrifuged at 4,300 g for 5 min at 4°C. The TCA-precipitable pellet and the TCA supernatant were measured in a γ-counter. The TCA-precipitable fraction of radioactivity was expressed as where cpm is counts per minute.
After 5-min internal carotid artery perfusion with125I-TNF-α, the perfused hemisphere was homogenized for 10 s in 3 ml of ice-cold 0.9% NaCl with an Ultra Turrax (Kinematica, Lucerne, Switzerland). A 300-μl aliquot was loaded on a 3-mm discontinuous 5%/15% SDS-PAGE gel. The gel was dried and exposed to autoradiographic film (Kodak X-OMAT) for 2 wk.
Isolation of rat brain capillaries.
Animals were killed under halothane anesthesia, and the brains were removed. Meninges, choroid plexus, and white matter were carefully removed from the cortical shell. The tissue was mechanically homogenized in 20% dextran and centrifuged at 3,200 g for 15 min at 4°C. Microvessels were purified by sieving through an 80-μm nylon mesh and pouring onto a glass bead column held by a 40-μm nylon mesh as described (35). The microvessels were harvested from the glass beads and the 40-μm nylon mesh and were spun down by gentle centrifugation (300 g, 5 min, 4°C). The capillary pellet was resuspended in Ringer-HEPES buffer (RHB) with 0.1% BSA and visually inspected by light microscopy. An aliquot was removed for protein measurement by the bicinchoninic acid (BCA) assay (Pierce, Rockford, IL).
Binding of 125I-TNF-α to rat brain capillaries.
Freshly isolated capillaries (50 μg of protein per vial) were used for binding studies and incubated in a final volume of 0.45 ml RHB-0.1% BSA with 125I-TNF-α at a concentration of 0.167–0.27 nM (100,000 counts · min−1 · vial−1) at 4°C or 37°C for up to 120 min. Unlabeled TNF-α (33 nM) was added in competition studies. At the end of the incubation times, the samples were centrifuged and the supernatant was removed. In some assay series, the pellet was subjected to a brief acid wash to remove cell surface-bound tracer (6 min at 4°C in 0.12 M NaCl, 0.028 M Na acetate, and 0.02 M Na barbital, pH 3) (35). Radioactivity of the pellet was measured in a γ-counter. Protein was subsequently determined by solubilization (500 μl of 1 M NaOH, 60°C for 30 min) and BCA assay (Pierce).
Binding of 125I-TNF-α to isolated rat brain capillary membranes.
The membrane fraction was prepared from freshly isolated rat brain capillaries according to Lidinsky et al. (26). After cell lysis in 0.01 M Tris · HCl (pH 7.4), the basement membrane was separated from cell membranes by sonication and spun out (25,000 g, 30 min). The supernatant with cell membranes was used immediately or stored in 0.05 M Tris · HCl (pH 7.4) at −70°C.
Binding studies with membranes (97- or 78-μg membrane protein per vial) in a final volume of 0.45 ml RHB-0.1% BSA with125I-TNF-α (0.0167 or 0.33 nM) were performed at 4°C overnight. For competition experiments, 33 nM unlabeled TNF-α or 300 nM human leptin was added. At the end of the incubation, 4 ml of buffer were added to each tube. Membrane-bound radioactivity was separated from free tracer by rapid filtration through preconditioned Whatman GF/C filters. Radioactivity retained on filters was determined by γ-counting.
Total RNA and poly(A)-RNA preparation.
Total RNA from cortical microvessels of 12 rats was extracted with the single-step phenol-chloroform extraction method (13). The isolated microvessels were solubilized in 1 ml of lysis buffer containing 4 M guanidine thiocyanate, 25 mM Na citrate, 10% (wt/vol)N-lauroylsarcosine, and 0.7% β-mercaptoethanol. The lysates were vortexed and incubated on ice for 10 min. A centrifugation step was performed at >10,000 g for 15 min at 4°C. The upper aqueous phase contained the isolated total RNA. After addition of 1/10 volume of 2 M Na acetate (pH 4.0), the RNA was extracted with an equal volume of phenol-chloroform-isoamylalcohol (25:24:1). Isopropanol was used for precipitation of the RNA pellet, and 70% ethanol was used for washing the RNA. The total RNA (yield ∼20 μg, measured by UV detection at 260 nm) was used for RT-PCR. Poly(A)-RNA for Northern blots was sampled from brain capillaries, choroid plexus, pineal glands, and anterior and posterior pituitaries from two groups of rats, either untreated control rats or rats after injection of 2 mg/kg ip LPS, 4 h before decapitation (n = 12/group). The tissues of each group were pooled and homogenized with a Teflon-glass homogenizer in 5 ml of 0.1 M Tris · HCl with 1% SDS, 0.5 M LiCl, 10 mM EDTA, and 5 mM dithiothreitol (pH 8.0). poly(A)-RNA was isolated using the oligo(dT)25 Dynabeads single-step method according to the manufacturer's instructions (Dynal Biotech, Hamburg, Germany).
Primers for amplification of p60 (TNFR1) were selected from the rat sequence (EMBL accession code m63122) for amplification of bp 813 to 1167 of the cDNA sequence by using the HUSAR software. Primers for cDNA of rat p80 (TNFR2) were taken from the partial sequence of 256 bp (EMBL accession code u55849). Four additional primer pairs for overlapping regions spanning the coding region were selected using the mouse p80 cDNA sequence (EMBL accession code m60469).
One microgram of total RNA was reverse transcribed with Superscript II-reverse transcriptase (GIBCO, Karlsruhe, Germany) and pd(N)6 random primer hexamers (Amersham Pharmacia Biotech, Freiburg, Germany) for 90 min at 42°C, and 2 μl of reverse transcription product (cDNA) were used for PCR without further purification. Hot start PCR with addition of 2.5 U AmpliTaq polymerase (Applied Biosystems, Weiterstadt, Germany) at 94°C was performed in a total volume of 50 μl for 35 amplification cycles with annealing temperatures of 55°C and annealing times of 40 s. PCR products (10 μl) were resolved by electrophoresis on 1.5% agarose gels.
Northern blot hybridization.
Enriched poly(A) RNA (2.5–5 μg) was separated by electrophoresis on denaturating agarose gels (1.5% agarose, 1× MOPS, 6 M formaldehyde 37%), transferred to nylon membranes (Nytran NY 12N, Schleicher & Schuell, Dassel, Germany), and cross-linked by UV irradiation. To obtain p60 and p80 cDNA probes, RT-PCR products were T/A cloned into the pGEM-T vector (Promega, Mannheim, Germany) by following the manufacturer's protocol. A 0.5-kb cDNA fragment for GLUT1 was cut from the Bluescript KS plasmid containing the bovine GLUT1 cDNA clone pGT51 (10). A 1.2-kb cDNA probe pRB15 for mouse C1q was excised from pCR 1000 by EcoRI digest (39). As a standard, a commercial human GAPDH cDNA fragment was used (Clontech, Heidelberg, Germany). All cDNA probes were labeled to high specific activity (>109 cpm/μg) using [α-32P]dCTP (Hartmann Analytic, Braunschweig, Germany) and Prime-it RmT random primer labeling kit (Stratagene, Austin, TX).
Membranes were prehybridized in prehybridization buffer (50% formamide, 0.9 M NaCl, 0.06 M Na phosphate, 6 mM Na EDTA, and 0.5% SDS, pH 7.4) with 0.1 mg/ml denaturated salmon sperm DNA (GIBCO) for 3–4 h at 42°C, followed by hybridization with32P-labeled cDNA probes. Hybridization was performed at 42°C for 24 h. The membranes were then washed in 6× sodium chloride-sodium phosphate-EDTA (SSPE)-0.3% SDS for 15 min at 42°C, in 2× SSPE-0.3% SDS for 15 min at 42°C, and in 1× SSPE-0.3% SDS for 15 min at 42°C to a final stringency of 0.5× SSPE-0.3% SDS for 30 min at 50°C. Autoradiographic signals were analyzed by a phosphoimager (Fujix BAS 1000) in combination with densitometric software (MacBAS, Fuji Photo Film). All data were corrected for mRNA loading by normalization with the signal obtained with the GAPDH probe. For photographic reproduction, autoradiograms were developed after exposure to X-ray films (X-OMAT, Kodak).
Statistical analysis was performed with the InStat program (GraphPad, San Diego, CA). Comparison of the means of two groups was performed by Student's t-test; ANOVA was used for multiple group comparisons. The significance level was set at P < 0.05. Linear regression analysis was performed with GraphPad Prism.
The brain uptake of radiolabeled rat TNF-α was measured in rats by use of the internal carotid artery perfusion method. We confirmed that the labeling procedure did not compromise bioactivity of TNF-α by measuring the EC50 of unlabeled and125I-labeled TNF-α in a cytotoxicity assay with L929 cells. EC50 for tracer was not significantly different from unlabeled TNF-α, with values of 9.5 ± 1 and 11.4 ± 3.7 pg/ml, respectively (unpaired t-test, n = 4,P > 0.1). Brain perfusions over 1, 5, and 10 min with rat 125I-TNF-α provided evidence of time-dependent organ uptake. Consistent with a relatively slow transport, the VDof TNF-α after a 1-min perfusion was not yet significantly different from the vascular volume, with VD values of 10.77 ± 1.38 and 12.27 ± 0.75 μl/g for 125I-TNF-α and3H-RSA, respectively. However, after 5- and 10-min perfusions, the VD of rat 125I-TNF-α was significantly higher than that of the vascular marker3H-RSA. This is shown in Fig.1, A and B. Separation of brain tissue into the vascular component and postvascular supernatant by capillary depletion revealed that, after correction for intravascular content, the uptake could largely be attributed to transcytosis through the BBB. The vascular pellet contains only a minor fraction of total brain radioactivity. For example, the pellet VD of 0.99 ± 0.29 μl/g after 5 min corresponds to only 6% of the total tracer in brain. The radioactivity in postvascular supernatant represented intact tracer. This can be concluded from the TCA-precipitable fraction in these samples of 91.3 ± 1.0 (mean ± SE, n = 43; pooled data from all 5- and 10-min perfusions), which approached the precipitability of the freshly labeled tracer. The integrity of the tracer in brain was further demonstrated by SDS-PAGE and autoradiography. A single band at the expected molecular mass of the TNF-α monomer (17 kDa) was detected (Fig.2).
The permeability-surface area product (PS) at the BBB for TNF-α was estimated from 1-, 5-, and 10-min perfusions by linear regression analysis, as shown in Fig. 1 B. Uptake into whole brain and into postvascular supernatant occurred at rates of 0.81 ± 0.13 and 0.34 ± 0.13 μl · min−1 · g−1, respectively.
Figure 1 C provides evidence that a TNF-α-specific mechanism is involved in that transport. Co-perfusions of rat125I-TNF-α with unlabeled rat TNF-α at concentrations between 0.83 and 16.7 nM led to a decrease in brain VDcompared with tracer perfusion alone (concentration = 1.25 nM), down to VD values as seen for RSA. Further evidence of a specific and TNF receptor-related mechanism can be derived from the perfusion experiment with human125I-TNF-α, depicted in Fig. 1 A. Human TNF-α has no affinity to the rodent p80 receptor (24). There was no difference between the VD of human125I-TNF-α and 3H-RSA after a 5-min brain perfusion.
We studied next whether the in vivo brain uptake of TNF-α is reflected in binding and uptake by isolated rat brain microvessels. The result is shown in Fig. 3 A. There was a time-dependent increase in the amount of rat125I-TNF-α bound by a brain microvessel preparation. Part of the tracer binding was displaced either by competition with unlabeled rat TNF-α or by a mild acid wash. As apparent from Fig.3 B, similar results were obtained with a membrane preparation derived from the isolated brain microvessels. A 100-fold molar excess of rat TNF-α displaced ∼30% of the binding, whereas no competition was observed by a 900-fold molar excess of the unrelated polypeptide leptin (negative control). The relatively high level of nonspecific binding of TNF-α with both the intact microvessels (Fig.3 A) and the membranes (Fig. 3 B) did not allow the performance of nonlinear binding analysis or Scatchard transformation. Modifications of the binding buffer or of the separation method (filtration or centrifugation) did not reduce nonspecific binding (data not shown).
Analysis of gene expression at the mRNA level was conducted to further characterize the binding sites. As shown in Fig.4, RT-PCR with RNA from rat microvessels resulted in the generation of amplicons of the predicted lengths for both TNF receptor types. Furthermore, sequencing of the overlapping amplified segments of the rat type 2 receptor (p80) covered 1,300 nucleotides in the coding region (deposited in GenBank under accession no. AF420214). The sequence homology with the published murine type 2 receptor was 90% at the mRNA level and 85.5% for the deduced amino acid sequence. No additional PCR products of different lengths that could indicate p80 isoforms were observed. Northern blots with32P-labeled cDNA probes encompassing the transmembrane domain (in the case of the p60 transcript) or part of the cytosolic domain (in the case of the p80 transcript) showed distinct bands for both messages in a brain capillary polyA-RNA preparation (Fig.5). The signals for both receptors were clearly detectable in native capillaries, albeit weak. Consistent with the high enrichment of endothelial cells in the capillary preparation, a strong signal for the BBB endothelium-enriched message GLUT1 was detected. In contrast, the absence of a signal for C1q argues against the contribution of perivascular microglia as a mRNA source in our microvascular samples. For comparison, Fig. 5 also shows blots of other CNS-derived tissues that do not possess an endothelial barrier system. The same pattern of expression of TNFR1 and TNFR2 was seen in all tissues. No evidence of additional transcripts could be detected in the capillary preparation for either receptor. Systemic treatment with LPS as immune stimulant resulted in moderate upregulation of the message for both TNF receptors in brain capillaries to 149% of control level (p60) and 127% of control level (p80), as given in Table1. The corresponding upregulation of p60 mRNA in the other tissues ranged from 136% (choroid plexus) to 225% (pineal gland), and from 190% (choroid plexus) to 348% (posterior pituitary) for p80 mRNA.
The question of whether an increase in mRNA for the two TNF receptors at the BBB results in an increased brain uptake of TNF-α was addressed by comparing the BBB transport in vehicle (0.9% NaCl)-injected animals and that in rats which had received intraperitoneal injections of LPS between 4 and 24 h earlier. As shown in Fig. 6, the preceding systemic immune challenge with LPS did not cause significant changes in brain uptake of rat 125I-TNF-α as measured by 5-min brain perfusion. This was true for brain homogenate (Fig. 6) as well as for the vascular pellet and postvascular supernatant (not shown). The VD of the vascular marker 3H-RSA remained unchanged from untreated controls, also, confirming our earlier observations that LPS administration does not open the BBB to proteins nonspecifically (7).
The results of the study are compatible with the following conclusions. 1) A saturable transport system for TNF-α is present at the BBB, which recognizes rat TNF-α but not human TNF-α. The p80 receptor appears to be required for the BBB transport.2) The TNF receptors p60 and p80 are constitutively expressed by brain microvascular endothelial cells of untreated rats.3) The rat p80 receptor shows 90% sequence homology with the corresponding mouse cDNA. No isoforms were detected at the mRNA level. 4) Systemic treatment with LPS moderately upregulated the mRNA levels of both receptors in brain microvasculature and in brain tissues without endothelial barrier function. 5) The LPS treatment did not change the rate of uptake of TNF-α at the BBB in vivo, suggesting that the receptor protein level is not increased on the endothelial cell surface.
The rate of BBB transport of TNF-α was measured using the internal carotid artery perfusion method. Although the PS product for TNF-α was low, it was clearly much higher than would be expected for a macromolecule of its size (molecular mass of the homotrimer = 51 kDa). For comparison, a PS value of ∼0.01 μl · min−1 · g−1for native albumin has been reported in intravenous studies (36). Albumin does not show measurable brain uptake in short-term perfusion experiments, as used here, and is a suitable intravascular marker. When the fraction of TNF-α tracer truly penetrating the BBB is considered (represented by the postvascular supernatant after capillary depletion), the PS product amounted to 0.34 ± 0.13 μl · min−1 · g−1. This value is similar to the PS product of 0.2 μl · min−1 · g−1for murine TNF-α in mice, as measured by the intravenous bolus technique (21, 32). It is also in the same range as the PS product of 0.32–0.39 μl · min−1 · g−1reported for the low molecular mass substance sucrose (342 kDa) (37, 41). The latter comparison with a substance often used as vascular marker highlights the importance of excluding metabolic artifacts. This requirement was met in the present perfusion studies, as shown by the integrity of the tracer in brain tissue, which was independently determined with TCA precipitation and SDS-PAGE.
The saturability of TNF-α transport is consistent with specific, receptor-mediated transcytosis as the underlying mechanism. Although exceptions exist, as postulated for epidermal growth factor (31), in which the cell surface receptor mediating the classical physiological effect of a peptide hormone is distinct from its transport protein at the BBB, such may not apply to TNF-α. Evidence from the present studies strongly argues for the involvement of at least the known p80 receptor in brain uptake of the cytokine. Uptake in the rat model was seen only with homologous rat TNF-α, whereas human TNF-α was not transported (Fig. 1 A). It is known that p80 receptors are species specific, and rodent p80 receptors do not recognize human TNF-α (24). The absence of brain uptake of human TNF-α thus indicates critical involvement of the p80 receptor in the BBB transport process. In this regard, the present result is in good agreement with recently reported data in knockout mice (32). These studies showed a decreased rate of transport of murine TNF-α at the mouse BBB, when either one of the TNF receptor types was absent, and no brain uptake in double knockout animals.
The range of concentrations over which the saturation of BBB transport occurs (between 1 and 17 nM, Fig. 1) is compatible with the p80 receptor. The dissociation constant (K d) of p80 for TNF-α has been reported as 0.42 nM (20). The affinity of the p60 receptor, on the other hand, seems to be higher, with a K d of 19 pM determined under physiological conditions on human cell lines (20).
The characterization of the receptors in binding and internalization studies with isolated brain microvessels and membrane preparations proved difficult. We were able to demonstrate specific binding sites for 125I-labeled rat TNF-α on these microvessels by the partial competition with a molar excess of TNF-α. However, a calculation of maximal binding (Bmax) andK d was not feasible. Presumably the endothelial expression of these binding sites at the protein level is relatively low compared with the degree of nonspecific binding of TNF-α to microvessels. The latter could be due to adsorption to cytoskeletal elements, which are partially exposed in microvessel preparations such as those used here (29).
However, we could readily identify both the p60 and the p80 receptor on Northern blots of native rat brain capillary mRNA. These transcripts were of endothelial origin, as demonstrated by the presence of a strong GLUT1 signal, which is known to be selectively expressed at the BBB (34). Absence of the message for C1q, which represents an exclusive marker of microglia and perivascular macrophages in brain tissue (38), served as a negative control in support of the pureness of the microvessel preparation. Microglia are known not only as a source of cytokines but also to constitutively express both types of TNFR (15). Therefore, the lack of a signal for C1q in the capillary mRNA indicates that these cells were not present in the microvessel preparation.
Our results for p60 and p80 mRNA expression in brain microvessels after systemic LPS administration extend the observations with in situ hybridizations performed by Nadeau and Rivest (29). These authors found low constitutive expression of the p60 transcript over brain parenchymal blood vessels, whereas they did not detect basal p80 expression by BBB endothelium. The discrepancy between the negative in situ hybridization data and the present findings with respect to the basal p80 signal on brain microvessels may be explained by the different methods. We achieved high sensitivity in our Northern hybridizations due to the use of poly(A)-RNA from brain microvessels of six pooled brains. After provocation of a systemic immune response with intraperitoneal injection of LPS, the mRNA levels of p60 and p80 in brain capillaries were moderately elevated in the present study, and these Northern blot data 4 h after LPS are consistent with peak stimulations by LPS or TNF-α found between 3 and 6 h in the in situ hybridization study over brain parenchymal blood vessels and over choroid plexus (29). The signals seen on our Northern blots did not reveal additional transcripts in brain capillary endothelial cells compared with non-BBB tissues. Similar transcript sizes for the p60 and p80 receptor have been reported in rat bronchial epithelium (2) and rat glia (15). In addition, all of the PCR products for TNFR1 and TNFR2 described here were of the expected size, as confirmed by sequencing. In this regard, it is interesting to note that a novel isoform of the human TNFR2, icp75TNFR, which is intracellularly expressed at a low abundance, was recently isolated (40) and found in nonbrain endothelial cells (HUVEC) by these authors.
Concerning the expression of TNFR1 and TNFR2 in pituitary and pineal gland, TNF-α binding was detected autoradiographically in the anterior pituitary (46), and the transcripts for the two receptors have been measured in the corticotroph-derived cell line AtT-20 (23). Functional evidence for pineal TNF receptors is derived from a decrease in pinealocyte serotonin secretion and stimulation of pineal microglia of explant cultures under TNF-α treatment (44). Receptor expression at the mRNA level has not been previously reported in the pineal gland.
No significant changes occurred in the rate of brain uptake of rat125I-TNF-α between 4 and 24 h after intraperitoneal administration of LPS. We have previously shown that LPS administration at low (50 μg/kg) and high doses (2 mg/kg) did not result in a measurable breakdown of the BBB in rats for vascular markers of small and high molecular weight (sucrose and serum albumin, respectively) (7). The present data with RSA confirmed that finding. Therefore, no increase in the rate of brain uptake of TNF-α due to a damaged BBB had to be expected in our study. As outlined, the present results (Fig. 1) clearly argue in favor of TNF receptor-mediated transcytosis of TNF-α through the BBB. This is supported by the expression of both TNF receptor types on native brain microvessel endothelial cells and by the saturability of transport at low nanomolar concentrations. TNF receptor-mediated transport could theoretically change in either direction during the course of an immune challenge by LPS with its associated release of cytokines. 1) The receptor expression may be upregulated. Our Northern blot data support this option at the level of gene transcription, and this is also in agreement with the published in situ hybridization data (29). In primary endothelial cell culture from mouse brain, TNF-α caused a peak increase in mRNA for TNFR1 and TNFR2 levels at 6 h, whereas the corresponding increase in protein expression was significant after 12–24 h (27). However, at the protein level, in vivo effects of LPS on TNF receptor regulation in brain endothelial cells have not been reported.2) Ligand-stimulated receptor shedding may actually decrease the number of surface-expressed TNF receptors. Systemic TNF-α levels are increased after LPS. Receptor shedding is a physiological response that is considered protective under septic conditions (1) and could protect the BBB from damage in vivo. Shedding of both types of TNF receptors by endothelial cells has been observed within the first 2 h after ligand stimulation (27).3) Even without shedding, the number of TNF receptors on the cell surface may be downregulated by ligand-induced internalization, which has also been demonstrated on endothelial cells (12). Finally, a combination of two or all of the mechanisms listed above could result in a condition in which no net increase or decrease of cell surface receptor expression would occur.
From a physiological point of view, a substantial increase of the transport rate of TNF-α into brain under inflammatory conditions may be undesirable, considering the potent cytotoxic effect of the cytokine. However, the amount of the cytokine transported would increase proportionately with plasma concentrations, as long as the concentrations stay within a range seen in physiological conditions. These concentrations are insufficient to saturate the BBB transport system for TNF-α (Fig. 1). Here, the rate of BBB transport after an LPS stimulus (Fig. 6) was measured at perfusate TNF-α tracer concentrations of ∼1 nM, which already correspond to high levels seen in conditions like sepsis (48).
In conclusion, our data strengthen the concept that receptor-mediated transport of TNF-α at the BBB occurs under native conditions and after systemic immune stimulation. This could form part of a feedback regulation between the immune system and the CNS, analogous to a feedback loop regulating food intake, which seems to be mediated by the BBB transport of leptin and may be disturbed in obesity (22). The transport of the cytokine across the BBB would complement the well described direct effect of circulating TNF-α on the endothelial cells (29), which releases soluble mediators such as prostaglandins and nitric oxide and activates the hypothalamo-pituitary-adrenal axis.
We acknowledge the expert technical assistance of G. Hohorst and S. Schaefer-Dewald. The plasmid containing the GLUT1 cDNA was kindly provided by Dr. R. Boado, and the C1q cDNA probe was a gift from Dr. W. Schwaeble. We are indebted to Drs. K. Bauer and J. Weidanz for stimulating discussions.
This research was supported in part by the Deutsche Forschungsgemeinschaft (SFB 297). B. Osburg is the recipient of a stipend from the Daimler-Benz Foundation, Germany.
Address for reprint requests and other correspondence: U. Bickel, Dept. of Pharmaceutical Sciences, Texas Tech Univ. Health Sciences Center, School of Pharmacy, 1300 Coulter Dr., Amarillo, Texas 79106 (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
July 31, 2002;10.1152/ajpendo.00436.2001
- Copyright © 2002 the American Physiological Society