Lipopolysaccharide-Induced Fever Depends on
Prostaglandin E2 Production Specifically in
Brain Endothelial Cells
Linda Engström, Johan Ruud, Anna Eskilsson, Anders Larsson, Ludmila Mackerlova,
Unn Kugelberg, Hong Qian, Ana Maria Vasilache, Peter Larsson, David Engblom,
Mikael Sigvardsson, Jan-Ingvar Jönsson and Anders Blomqvist
Linköping University Post Print
N.B.: When citing this work, cite the original article.
Original Publication:
Linda Engström, Johan Ruud, Anna Eskilsson, Anders Larsson, Ludmila Mackerlova, Unn
Kugelberg, Hong Qian, Ana Maria Vasilache, Peter Larsson, David Engblom, Mikael
Sigvardsson, Jan-Ingvar Jönsson and Anders Blomqvist, Lipopolysaccharide-Induced Fever
Depends on Prostaglandin E2 Production Specifically in Brain Endothelial Cells, 2012,
Endocrinology, (153), 10, 4849-4861.
http://dx.doi.org/10.1210/en.2012-1375
Copyright: Endocrine Society
http://endo.endojournals.org/
Postprint available at: Linköping University Electronic Press
http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-84885
Lipopolysaccharide-Induced Fever Depends on
Prostaglandin E2 Production Specifically in Brain
Endothelial Cells
Linda Engström,* Johan Ruud,* Anna Eskilsson, Anders Larsson, Ludmila Mackerlova, Unn Kugelberg, Hong Qian, Ana Maria Vasilache, Peter Larsson, David Engblom, Mikael Sigvardsson, Jan-Ingvar Jönsson, and Anders Blomqvist
Department of Clinical and Experimental Medicine (L.E., J.R., A.E., A.L., L.M., U.K., H.Q., A.M.V., D.E., M.S. J.-I.J, A.B.), Divisions of Cell Biology, Microbiology, and Molecular Medicine, and Department of Medical and Health Sciences (P.L.), Division of Radiation Physics, Faculty of Health Sciences, Linköping University, S-581 85 Linköping, Sweden
Immune-induced prostaglandin E2 (PGE2) synthesis is critical for fever and other centrally elicited disease symptoms. The production of PGE2 depends on cyclooxygenase-2 and microsomal pros-taglandin E synthase-1 (mPGES-1), but the identity of the cells involved has been a matter of controversy. We generated mice expressing mPGES-1 either in cells of hematopoietic or nonhe-matopoietic origin. Mice lacking mPGES-1 in henonhe-matopoietic cells displayed an intact febrile re-sponse to lipopolysaccharide, associated with elevated levels of PGE2 in the cerebrospinal fluid. In contrast, mice that expressed mPGES-1 only in hematopoietic cells, although displaying elevated PGE2 levels in plasma but not in the cerebrospinal fluid, showed no febrile response to lipopoly-saccharide, thus pointing to the critical role of brain-derived PGE2 for fever. Immunohistochemical stainings showed that induced cyclooxygenase-2 expression in the brain exclusively occurred in endothelial cells, and quantitative PCR analysis on brain cells isolated by flow cytometry demon-strated that mPGES-1 is induced in endothelial cells and not in vascular wall macrophages. Similar analysis on liver cells showed induced expression in macrophages and not in endothelial cells, pointing at the distinct role for brain endothelial cells in PGE2 synthesis. These results identify the brain endothelial cells as the PGE2-producing cells critical for immune-induced fever.
(Endocrinology 153: 4849 – 4861, 2012)
T
he febrile response, a hallmark of infection and in-flammation, has been shown to depend on induced prostaglandin (PG) E2 synthesis by the sequential action of cyclooxygenase-2 (Cox-2) and microsomal prostaglan-din E synthase-1 (mPGES-1) (1–5). It has also been dem-onstrated that the febrile response is dependent on PGE2-binding to its EP3 receptors in the median preoptic nucleus in the hypothalamus (6), demonstrating the critical role of intracerebral PGE2 for fever.Although the role of Cox-2, mPGES-1, and PGE2 for fever thus is well established, there have been considerable
arguments as to in which cells the critical PGE2 synthesis takes place. Several observations point to the brain vas-culature, in which Cox-2 and mPGES-1 are abundantly expressed upon immune challenge (7–9). The brain vas-culature also expresses receptors for the cytokine IL-1 (7, 8, 10), making it a dedicated structure for the transduction of peripheral immune signals to the brain through induced PGE2 synthesis (11). Studies in which the different cellular elements in the vascular wall have been characterized with specific markers (8, 12, 13) have identified the endothelial cells as the component of the blood vessel that expresses
ISSN Print 0013-7227 ISSN Online 1945-7170 Printed in U.S.A.
Copyright © 2012 by The Endocrine Society
doi: 10.1210/en.2012-1375 Received April 5, 2012. Accepted July 11, 2012. First Published Online August 7, 2012
* L.E. and J.R. contributed equally to this work.
Abbreviations: Cox-2, Cyclooxygenase-2; CSF, cerebrospinal fluid; GFP, green fluorescent protein; KO, knockout; LPS, lipopolysaccharide; mPGES-1, microsomal prostaglandin E synthase-1; PG, prostaglandin; PI, propidium iodide; qPCR, quantitative PCR; WT, wild type.
mPGES-1 and that hence is critical for the induced PGE2-synthesis and the PGE2-evoked and centrally elicited re-sponse. However, other studies, among them work that has focused on the expression of Cox-2, have advocated that cells of hematopoietic origin, in particular the perivas-cular macrophages, immune cells that are located in the vascular wall just beneath the endothelial cells, as well as immune cells located in peripheral tissues such as the liver and lung, may be equally important and under certain circumstances the sole source of centrally acting PGE2 (14 –17; but cf. Refs. 9, 18, and 19). However, for both views the evidence is circumstantial, and functional ex-periments that pinpoint the role of the different cellular compartments have been lacking.
Here we addressed the issue of in which cells the for the febrile reaction critical PGE2 synthesis takes place by ex-amining central and peripheral PGE2 production and the accompanying temperature response in mice chimeric for mPGES-1, implying that they expressed mPGES-1, and hence were capable of induced PGE2 synthesis, either in their cells of hematopoietic or nonhematopoietic origin. This enabled us to differentiate between the response me-diated by, e.g., the nonhematopoietic endothelial cells and that mediated by hematopoietically derived cells, includ-ing perivascular macrophages as well as peripheral im-mune cells. Our results show that nonhematopoietic cells are both necessary and sufficient for elevating central PGE2 and eliciting fever after a peripheral immune chal-lenge with lipopolysaccharide, whereas hematopoietically derived cells, although implied in peripheral PGE2 pro-duction, contributes little to central PGE2 levels and are unable by themselves to elicit a febrile response. Comple-mentary immunohistochemical stainings on brain tissue and quantitative PCR on dissociated brain vascular cells further identified the endothelial cells as the source of in-duced PGE2 synthesis. These findings demonstrate that the critical site for the immune-induced PGE2-mediated febrile response is the brain vascular endothelial cells.
Materials and Methods
Animals
Wild-type C57BL/6 mice (Scanbur, Sollentuna, Sweden), and
Ptges⫹/⫹and Ptges⫺/⫺mice (20), backcrossed onto the C57BL/6 background, were used. For transplantation experiments (donor bone marrow) Ptges⫺/⫺mice were crossed with a green fluores-cent protein (GFP)-expressing strain [C57BL/6-Tg(CAG-EGFP)C14-Y01-FM131Osb; kindly provided by Dr. Masaru Okabe, Osaka university, Osaka, Japan (21)]. The resulting heterozygous offspring were then crossed to generate
GFP⫹Ptges⫺/⫺and GFP⫹Ptges⫹/⫹mice. Ptgs2 [Cox-2 knock-out (KO)] mice (22) were from Taconic (Ry, Denmark). The
animals were housed one to five per cage on a 12-h light, 12-h dark cycle (lights on at 0800 h). All experimental procedures were approved by the Animal Care and Use Committee at Linköping University.
Irradiation and bone marrow transplantation
The mice (3– 6 months old) were irradiated to an absorbed dose of 9 Gy in two fractions. Approximately 24 h after irradi-ation, they were injected iv with 2 ⫻ 106
freshly prepared GFP⫹CD45⫹ bone marrow cells and allowed to survive for about 5 months. For details, see Supplemental Materials and Methods, published on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org.
Temperature recordings
For surgical procedures, see Supplemental Materials and Methods.
Intraperitoneal injection of lipopolysaccharide (LPS)
LPS from Escherichia coli (Sigma-Aldrich, St. Louis, MO; O111:B4; 120g/kg body weight) diluted in 100 l saline was injected ip at around 1000 h. Temperature data were obtained using ip transmitters (Data Science International, St. Paul, MN) implanted 1 wk before the injection. After a washout period of 1–2 wk, animals that had been given LPS were injected with saline, and vice versa, and body temperature recorded as de-scribed above.
Intravenous injection of lipopolysaccharide or IL-1
Mice were injected iv with LPS (30 or 1g/kg) or recombinant murine IL-1 (30 g/kg; Preprotech, Rocky Hill, NJ) through an indwelling jugular catheter implanted 3 d before injection. The catheter was exteriorized at the back of the neck and connected to a swivel system (CMA Microdialysis, Solna, Sweden) on the top of the cage, permitting injection without handling the mice.
Immunoassays for PGE2 and PGE2 metabolites in plasma and cerebrospinal fluid (CSF)
Following asphyxiation of the mice with CO2, blood was
drawn from the right atrium, transferred to EDTA-coated tubes (Sarstedt, Landskrona, Sweden) to which were added indometh-acin (10M), and centrifuged at 7000⫻ g for 7 min at 4 C. The
plasma was immediately frozen on dry ice and kept at⫺70 C. CSF was then collected from the cisterna magna using a Ham-ilton syringe, and immediately frozen. Samples that contained traces of blood were discarded. The whole procedure from when the animals were killed until CSF was withdrawn took less than 10 min. The concentration of PGE2 (CSF) and PGE2 metabolites (plasma) was determined using high sensitivity PGE2 enzyme immunoassay kit (Assay Designs, Ann Arbor, MI) and PGE me-tabolite enzyme immunoassay kit (Cayman, Ann Arbor, MI), respectively. For details, see Supplemental Materials and Methods.
Immunohistochemistry
Immunohistochemistry was performed according to stan-dard protocols in this laboratory (23). For details, see Supple-mental Materials and Methods.
Flow cytometry and quantitative PCR (qPCR)
Blood
Venous blood was collected into ice-cold PBS containing 2.5 mMEDTA. After lysis of erythrocytes, cells were resuspended in PBS and analyzed for GFP expression by flow cytometry.
Brain and liver
Single cells were prepared as described in Supplemental Ma-terials and Methods. The live cells [propidium iodide negative (PI⫺)] from each sample were first sorted using yield sorting mode and then resuspended in PBS with 10% fetal bovine serum and incubated with rat antimouse FC-Block (CD32/16), CD45 PE-Cy5.5, CD206-Alexa 647 (brain) or F4/80 PE (liver), and CD31 PE-Cy7 for 10 and 15–20 min, respectively, on ice. Dead cells were again excluded by PI staining. CD31⫹CD45⫺, CD45⫹CD31⫺, and CD45⫹CD206⫹ or CD45⫹F4/80⫹ cells were gated based on fluorescence minus one controls and isotype staining controls for each antigen expression (for antibody suppliers, see Supplemental Materials and Methods). The cell populations were then sorted directly into RNeasy lysis buffer (QIAGEN, Hilden, Germany) with-mercaptoethanol (Sigma; 143 mM) after purity analysis and stored at⫺80 C until use. For qPCR, total RNA was extracted, reversely transcribed, and the cDNA preamplified. For details, see Supplemental Materials and Methods.
Statistics
Temperature data were analyzed by a two-way ANOVA, fol-lowed by Bonferroni post hoc test. Data on PGE2 or PGE2 me-tabolites were analyzed by a one-way ANOVA followed by Newman-Keuls multiple comparisons test, or, for comparisons between treatments within the same group of chimera (plasma at 40 min), with a Student’s t test.
Results
Creation of chimeric mice by whole body ␥-irradiation and bone marrow transplantation
Immune-induced PGE2 is produced by mPGES-1, the PGE2 synthase encoded by the Ptges gene. We have pre-viously demonstrated that mPGES-1 is critical for fever (2– 4). Mice chimeric for mPGES-1 were created by sub-jecting littermates of wild-type (WT) and Ptges⫺/⫺(KO) mice to potentially lethal whole body ␥-irradiation (9 Gray) followed by transplantation of CD45⫹ enriched GFP⫹bone marrow cells from mice of the opposite geno-type (KO and WT, respectively). Hence, two different chi-meras were created: WT mice with Ptges⫺/⫺ hematopoi-etic cells and Ptges⫺/⫺mice with WT hematopoietic cells. In addition, WT mice transplanted with WT bone mar-row, and Ptges⫺/⫺mice transplanted with Ptges⫺/⫺bone marrow were generated and included as controls.
To allow as complete reconstitution of hematopoietic cells as possible, while at the same time assuring that mice did not become senescent, mice were allowed to survive for 5
months after irradiation and transplantation. Examination at this time point of LPS-induced IL-1 in plasma of chimeric mice (see Supplemental Materials and Method) showed a normal inflammatory response (Supplemental Fig. 1).
The degree of reconstitution was examined in periph-eral blood, brain, liver, and lung from tissue harvested after animals had been immune challenged for recordings of the temperature response. Overall, the data showed extensive replacement of native hematopoietic cells with transplanted cells.
Analysis of reconstitution in peripheral blood, as de-termined by flow cytometry of the proportion of GFP⫹ cells among the blood leukocytes, revealed that approxi-mately 90% of the white blood cells originated from trans-planted bone marrow cells (Supplemental Fig. 2).
Examination of tissue sections through the brain revealed GFP⫹cells (i.e. cells derived from transplanted cells) in the walls of blood vessels, with only few GFP⫹cells being seen in the brain parenchyma (Fig. 1, A–C). The morphology of the cells in the two compartments differed; whereas the cells in the walls of the blood vessel had an amoeboid appearance characteristic of perivascular macrophages (Fig. 1B), those found in the parenchyma displayed characteristics of micro-glial cells (Fig. 1C). Brain sections from six mice, encompass-ing all four groups of hybrid mice, which had been subjected to an LPS challenge to monitor their temperature response, were stained for both GFP and the mannose receptor (CD206) (Fig. 1D), which identifies perivascular macro-phages (24). Quantitative analysis on six randomly selected fields from each of the six mice showed that 87.5% (SEM8.7) of the CD206⫹ population also expressed GFP and that 91.5% (SEM5.4) of the GFP⫹ population also expressed CD206. Similar results were obtained in four transplanted mice that had not been given LPS, verifying that the recon-stitution had not taken place as a consequence of the immune challenge but was present before that. Thus, the data show that approximately nine of 10 GFP⫹cells in the examined brains were perivascular macrophages, and nine of 10 of the perivascular macrophages were derived from transplanted hematopoietic cells.
In additional hybrid mice, the proportion of GFP⫹cells among the CD206⫹population in the brain was determined by flow cytometry, providing corroborative evidence for the high degree of reconstitution of the perivascular cells from transplanted GFP expressing hematopoietic cells (Supple-mental Fig. 3). Accordingly, the irradiation and transplan-tation procedure generated a highly selective and efficacious replacement of native perivascular cells with transplanted cells. The specificity of the reconstitution was further sup-ported by labeling for GFP and the endothelial cell marker von Willebrand factor. In no case did GFP⫹cells also express the von Willebrand factor (Fig. 1E) and were in fact located
on the parenchymal side of the endothelial cells, consistent with their perivascular identity.
Reconstitution in the liver was examined using staining for GFP and the pan-macrophage marker F4/80 (25) (Fig.
1F). Quantitative analysis on confocal micrographs from 6 transplanted mice
showed that 67.7% (SEM 4.1) of the
liver F4/80⫹ population expressed
GFP, and that 84.6% (SEM1.9) of the total GFP⫹population was F4/80⫹. In the lung (Fig. 1G), staining for GFP and the lung macrophage marker CD68 (26, 27) showed that 78.3% (SEM2.3) of the CD68⫹cells expressed GFP and that 64.7% (SEM3.4) of the GFP⫹ pop-ulation expressed CD68. Thus, in both liver and lung, two thirds or more of the immune cells were replaced by trans-planted cells. Although it was slightly lower than in blood and brain, this de-gree of reconstitution should, when KO bone marrow cells were transplanted to WT mice, imply a very large reduction in any PGE2 synthesis of hematopoietic cells, and, most importantly, it should largely restore the PGE2 synthesis in he-matopoietic cells when WT cells were transplanted to mice of KO background.
mPGES-1 in nonhematopoietic cells is critical for the febrile response to ip LPS
The temperature response of the four different groups of transplanted mice 5 months after irradiation and transplantation is shown in Fig. 2A. Mice were given an ip injection of LPS
(120 g/kg body weight), and core
body temperature was recorded contin-uously by telemetry. As seen, WT mice transplanted with WT bone marrow (WT3 WT) displayed a characteristic fever with two clearly discernible phases, the first initial phase being partly obscured by the hyperthermia elicited by the handling during the in-jection (28). This temperature curve is similar to that observed in naïve WT mice using the same immune challenge paradigm (4). As also expected, similar to what is seen in mice with a global mPGES-1 gene deletion (2, 4), KO mice transplanted with KO bone marrow (KO3 KO) displayed a rapid temperature fall immedi-ately after the initial temperature peak elicited by the han-dling stress during the injection and then remained at a
FIG. 1. Reconstitution of hematopoietically derived cells in mouse brain, liver, and lung after
whole-body lethal irradiation and transplantation with CD45-sorted GFP-positive cells. A, Bright-field micrograph showing GFP-positive cells associated with blood vessels (arrows) in the brain. B, Fluorescent micrograph showing GFP-positive cells surrounding a blood vessel stained for the von Willebrand factor (vWF). C, Confocal micrographs showing (C1) cluster of GFP expressing microglial cells. (C2) is the same field showing the hematopoietic cell marker CD45, and (C3) the glial cell marker Iba1. Note that all GFP-positive cells, originating from transplanted cells, express CD45 and Iba1 (arrowheads), whereas not all CD45- and Iba1-positive cells in this field are transplanted cells, i.e. they are GFP negative (arrow with double
arrowhead). An Iba1-positive cells that is CD45 low (and GFP negative) is also shown (arrow).
D, Confocal micrographs showing colocalization of GFP with the mannose receptor (CD206), identifying the GFP-labeled cells as perivascular macrophages. Arrowheads point at dual-labeled cells, and arrow points at single-dual-labeled cell. E, Absence of colocalization between GFP and the vWF-stained endothelium. Note that the GFP cells are located on the
parenchymal side of endothelial cells. F, Dual labeling of GFP and the Kupffer cell marker F4/ 80 in the liver. Symbols are as in D. G, Dual labeling of GFP and the monocyte and
macrophage marker CD68 in the lung. Symbols are as in D. bv, Blood vessel. Scale bar, 100 m (A and B), and 20 m (C–G).
body temperature below or around that of control mice injected with saline. The temperature response of the two chimeras resembled that of mice carrying the same background genotype. Thus, WT mice transplanted with KO bone marrow (KO3 WT) showed a febrile response that was similar to that of the WT3 WT mice, and although the body temperature was slightly lower throughout most of the observation period, the differ-ence was not statistically significant. In contrast, KO mice transplanted with WT bone marrow (WT3 KO) were unable to mount a febrile response, similar to the KO3 KO mice. Like the latter they displayed a rapid temperature fall after the initial stress-induced hyper-thermia, and although their body temperature thereaf-ter for most of the observation period was slightly higher than that of the KO3 KO mice, this difference was not statistically significant.
The initial phase of fever in response to iv LPS is dependent on mPGES-1 in nonhematopoietic cells
Because the handling-induced hy-perthermia associated with the ip injec-tion may obscure rapid LPS elicited temperature responses (28, 29), we also examined the febrile response in chime-ric mice injected iv with LPS through an indwelling jugular catheter, a proce-dure that permitted injections during low stress condition for the mice. The dose (0.75g; ⬃30 g/kg body weight) was selected to yield a clearly discern-ible first phase of fever (29). As shown in Fig. 2B, WT mice, both when trans-planted with WT or KO bone marrow, displayed with a latency of 10 –15 min after the LPS injection a rapid and tran-sient temperature elevation, which peaked at 25– 40 min after the injec-tion. In contrast, KO mice, irrespective of whether they were transplanted with KO or WT bone marrow, showed no febrile response but instead a tempera-ture fall that started 10 –20 min after injection (Fig. 2B).
The febrile response to LPS in chimeric mice follows PGE2 levels in the cerebrospinal fluid
The aforementioned data strongly suggest that PGE2 production in cells of the hematopoietic lineage plays but a minor role for the febrile response and that the nonhema-topoietic cells instead are critical. To further address this issue, we examined in the different chimera the relation-ship between PGE2 levels in the brain and blood and the febrile response. Cerebrospinal fluid was withdrawn from the cisterna magna 40 min and 3 h after ip injection with LPS and analyzed for the PGE2 content by immunoassay. These time points were chosen because they correspond to the short first phase of fever and to the late, third, sustained phase of fever, respectively [(29) and present data]. More-over, these phases have been suggested to be elicited by distinct mechanisms (16). We found a significant elevation of the PGE2 level at 40 min in the cerebrospinal fluid of LPS-treated WT mice, irrespective of whether they had been transplanted with WT or KO bone marrow (Fig. 3A). LPS-treated KO3 KO mice displayed PGE2 levels in the cerebrospinal fluid similar to those seen in saline injected
FIG. 2. Temperature response to LPS in chimeric mice. A, Body temperature of chimeric mice
after ip injection of 120g/kg LPS or saline. The x-axis shows time in relation to injection (at 0 h). The body temperature for chimeric mice injected with saline is shown as an average for all mice taken together and is displayed for each separate group in the figurine. ***, P⬍ 0.001 between WT3 WT mice treated with LPS and NaCl, respectively; ###, P⬍ 0.001 between the same treatments in KO3 WT mice. B1, Body temperature of chimeric mice after iv injection of 30g/kg LPS. The x-axis shows time in relation to injection (at 0 h; vertical line). *, P⬍ 0.05, **, P 0.01, and ***, P 0.001, between KO3WT and WT3KO mice treated with LPS. Error bars,SEM. Note that although the temperature peak in WT3 WT is higher than that in KO3 WT mice, body temperature immediately before injection also differed so that temperature rise in both groups was about the same. Note also that the temperature fall in WT3 KO and KO3 KO mice was similar to that seen in mPGES-1 null mice (B2),
demonstrating that the radiation did not change the temperature response of the hybrid mice.
controls. LPS-treated WT3 KO mice showed somewhat higher levels than both KO3 KO mice and saline injected controls, but the difference was not statistically signifi-cant. At 3 h after the injection, LPS-treated WT mice trans-planted with either WT or KO bone marrow both dis-played still higher PGE2 levels in the cerebrospinal fluid (Fig. 3A). PGE2 levels in the KO3 KO mice remained low. Similar to the findings at 40 min, WT3 KO mice displayed in average somewhat higher PGE2 levels than the KO3 KO mice; however, again this difference was not statistically significant. Throughout, these data show that mice that display fever (WT3 WT and KO3 WT) display elevated levels of PGE2 in the CSF, whereas mice that do not display fever (WT3 KO and KO3 KO) show low lev-els of PGE2 in the CSF. Regression analysis (Supplemental Fig. 4) showed a strong and statistically significant rela-tionship (r2⫽ 0.88; P ⬍ 0.001) between body temperature
and PGE2 levels in CSF of LPS-injected mice.
The febrile response to LPS is only weakly related to plasma levels of PGE2 metabolites
AnalysisofplasmalevelsofPGE2metabolites[thisassaywas choseninsteadofdirectmeasurementofPGE2becausePGE2in
plasma is difficult to measure reliably (30–32)] showed induced expression 40 min after ip LPS injection in all hybrids except the KO3KO group (Fig. 3B). Notably, the rescue group, i.e. ani-mals in which WT bone marrow was transplanted to KO mice (WT3KO), showed a strong induction of PGE2 metabolites in plasma at 40 min, similar to WT3WT and KO3WT mice, indicating a significant role for hematopoietic cells for plasma PGE2 levels at this time point. However, as was shown in Fig. 2, WT3KO mice, in contrast to the WT3WT and KO3WT mice,didnotdisplayanyfebrileresponseat40min,andthedata thus did not show any clear relationship between plasma PGE2 levels and body temperature.
At 3 h after the LPS injection, plasma concentration of PGE2 metabolites (Fig. 3B) remained high in the WT3 WT group and low in the KO3 KO group, with KO3 WT and WT3 KO mice displaying levels in be-tween. Notably, although the concentration of PGE2
me-tabolites was not significantly different between
KO3 WT and WT3 KO mice but differed significantly in both these groups from WT3 WT mice, they displayed very different temperature responses, with KO3 WT mice showing a febrile response similar to WT3 WT mice and WT3 KO mice showing no fever (see Fig. 2). Again, these data thus show that the plasma PGE2 levels have at best a minor influence on body temperature. This conclusion was also confirmed by the regression analysis of the rela-tionship between body temperature and plasma levels of PGE2 metabolites (Supplemental Fig. 4).
Identification of PGE2 synthesizing cells
The data from chimeric mice clearly show that fever is dependent on intact PGE2 synthesis in nonhematopoietic cells and is best associated with elevated central but not peripheral PGE2 levels, suggesting a critical role for the cerebral endothelium. As a final step, we therefore exam-ined the expression of the inducible PGE2-synthesizing enzymes Cox-2 and mPGES-1 in the brain as well as in peripheral tissues. For the analysis of Cox-2, we used im-munohistochemistry. Although mPGES-1 has been un-equivocally demonstrated in rat brain tissue by both im-munohistochemistry and in situ hybridization by us and others (7, 8, 12), neither method produced any labeling in mouse brain tissue [consistent with the sparse evidence in the literature (33)]. Therefore, we instead used qPCR on dissociated brain cells separated by flow cytometry.
Prostaglandin-synthesizing enzymes are induced in brain endothelial cells but not in perivascular macrophages
Immunohistochemical staining for Cox-2 in the brain showed prominent induction in blood vessels 3 h after LPS injection (120g/kg, ip) (Fig. 4A). The labeling was seen
FIG. 3. PGE2response to LPS in chimeric mice. A, PGE2in CSF after ip injection of 120g/kg LPS. For observations at 40 min, n ⫽ 7 in the NaCl group (pooled from different chimera), n⫽ 4 in WT3WT, n ⫽ 5 in KO3 WT, n⫽ 4 in WT3KO, and n ⫽ 3 in KO3KO. For 180 min, n⫽ 6 in WT3WT and KO3WT, n ⫽ 5 in WT3KO and n ⫽ 4 in KO3 KO. *, P⬍ 0.05 vs. the NaCl-treated group; #, P ⬍ 0.05 vs. WT3 KO and KO3 KO mice. B, PGE2 metabolites in plasma at 40 and 180 min after ip injections of 120g/kg LPS or saline. At 40 min, n ⫽ 4 (NaCl) and n⫽ 8 (LPS) in WT3WT, n ⫽ 5 and 7 in KO3WT, n ⫽ 5 and 6 in WT3 KO, and n⫽ 4 and 5 in KO3KO. At 180 min, n ⫽ 8 in the NaCl group, n⫽ 6 in WT3WT and KO3WT, n ⫽ 5 in WT3KO, and n⫽ 3 in KO3KO. *, P ⬍ 0.05, **, P ⬍ 0.01, and ***, P ⬍ 0.001
vs. NaCl-injected mice; #, P⬍ 0.05 and ##, P ⬍ 0.01 vs. LPS-injected
in round nuclei outlining the vessel wall (Fig. 4B). Simul-taneous staining for the von Willebrand factor demon-strated perinuclear Cox-2 labeling surrounded by von Willebrand positive cytoplasm (Fig. 4F). Dual labeling for Cox-2 and GFP showed that the Cox-2-positive cells were distinct from the perivascular macrophages expressing GFP (Fig. 4, B–D). This was not a phenotype change of perivascular macrophages derived from transplanted he-matopoietic cells because cells labeled for the macrophage marker CD206 in nontransplanted wild-type mice neither expressed Cox-2 after LPS challenge (Fig. 4E).
Because the expression pattern of Cox-2 in the brain has been suggested to depend on the dose of LPS injected,
and possibly also the route of adminis-tration, we also injected a low dose of LPS iv (1g/kg). This procedure previ-ously has been reported to preferen-tially induce Cox-2 in perivascular cells and not in endothelial cells (15). How-ever, the same picture was seen as after injection of 120g/kg LPS ip: Cox-2 la-beling was observed in endothelial cells only, whereas perivascular macrophages were unlabeled (Fig. 4G). Similarly, iv in-jection of IL-1 (30 g/kg body weight), another immune stimulus, which, al-though being induced by LPS (34), also has been suggested to elicit Cox-2 pref-erentially in perivascular cells (15), only labeled endothelial cells (Fig. 4H). Hence, taken together these data demon-strate that Cox-2 induction occurs exclu-sively in endothelial cells.
qPCR analysis on cells obtained by flow cytometry (Fig. 5) showed a prom-inent up-regulation (9 times; P⬍ 0.05) of mPGES-1 mRNA in endothelial cells (CD31⫹CD45⫺) 3 h after ip LPS injec-tion (Fig. 5D). In contrast, mPGES-1 mRNA levels in perivascular cells
(CD206⫹CD45⫹) were low to
unde-tectable, both in LPS treated and NaCl treated mice (Fig. 5D). Notably, a sim-ilar qPCR analysis on cells obtained from the liver by flow cytometry (Fig. 6) showed strong up-regulation of mPGES-1 mRNA in macrophages (Kupffer cells) and undetectable values in endothelial cells, i.e. a pattern quite opposite to that seen in the brain, thus further supporting the distinct role of the brain endothelium in induced PGE2-synthesis. qPCR analysis of mPGES-1mRNAinthehypothalamusofchimericmicefurther supported the finding that hematopoietically derived brain cells do not take part in the PGE2 production (Supplemental Fig. 5). Taken together with the immunohistochemical data on Cox-2 expression, being localized to endothelial cells in the mouse brain, these findings hence identify the endothelial cells as the PGE2-producing cells in the brain during an immune challenge.
Cox-2 is induced in immune cells in the liver and in nonhematopoietic cells in the lung
Immunohistochemical staining of the liver 3 h after ip injection of LPS showed both in WT and transplanted mice induced Cox-2 expression in clusters of cells that
FIG. 4. Cox-2 expression in mouse brain does not occur in perivascular macrophages. A,
Low-power views of sections through the mouse brain: A1, constitutive expression in the hippocampus formation (hf) in NaCl-treated WT mice; A2, induced expression 3 h after LPS injection (120g/kg body weight) in wild-type mice; and A3, absence of expression (including that in hf) after LPS injection in Cox-2 KO mice. B, Dual labeling for GFP and Cox-2 in an irradiated and bone marrow-transplanted mouse. Both GFP-positive cells (arrows) and profiles with perinuclear Cox-2 expression are associated with blood vessels (bv), but GFP-positive cells also occur in blood vessels (right) that do not express Cox-2. C and D, Blood vessels at higher magnification. Cox-2 is not expressed in GFP-labeled cells (arrows). E, Mannose receptor (CD206)-positive cells (i.e. perivascular macrophages) are distinct from the cells that express Cox-2 (confocal micrographs; from wild-type mice). F, Perinuclear Cox-2 labeling in von Willebrand factor-stained endothelial cells. G, Cox-2 expression (red) after iv injection of a low dose (1g/kg body weight) of LPS or of IL-1 (30 g/kg body weight) (H). Perivascular cells (green) did not display any Cox-2 immunoreactivity. Note the unlabeled perinuclear region (arrows), which would be the preferential site of Cox-2 immunoreactivity. Scale bar, 500m (A), 100m (B), 50 m (C), 20 m (D–F), and 25 m (G and H).
were preferentially seen at the periphery of the lobule but occasionally also were more centrally located (Fig. 7A). Dual labeling revealed extensive overlap with F4/80
im-munoreactivity and in transplanted mice also with GFP, demonstrating that the Cox-2 induction occurred in im-mune cells, such as Kupffer cells (Fig. 7, B and C). In the
FIG. 5. Flow cytometry showing induced expression of mPGES-1 in endothelial cells (CD45⫺CD31⫹) but not in perivascular cells (CD45⫹CD206⫹) 3 h after ip injection of LPS. A, Representative fluorescence-activated cell sorting profiles show sorting strategy. Dead cells and debris were first excluded using forward scatter channel (FSC) area (A) vs. side scatter channel (SSC)-A and FSC height (H) vs. FSC width (W) and further excluded by PI staining (upper panels). The cell populations were then obtained by two-step sorting. First, the live cells (PI⫺) were enriched using a yield-sorting mode (upper panels). Then the live (PI⫺) CD45⫺CD31⫹and CD45⫹CD206⫹cells were gated based on fluorescence minus one (FMO) controls and isotype staining controls (lower panels). The CD45⫹CD206⫺cells, seen in the middle column of the lower panel in A represents microglial cells, which, although mostly being CD45low, in some areas display average to high immunoreactivity for CD45 (cf. Fig. 1C). B, Purity analysis of the sorted cell populations CD45⫺CD31⫹and CD45⫹CD206⫹cells. Note that although the sorting for CD45⫺CD31⫹cells (endothelial cells) in A did not yield a very large population (cf. also Fig. 6), the population selected for further analysis was clearly shown to be distinct. C, Quantitative PCR verification of the cell sorting. CD206 mRNA expression was normalized against that in the endothelial cells (EC) from mice injected with NaCl. Perivascular macrophages (PVC) display 1000-fold higher expression of CD206 than the endothelial cells (n⫽ 3 in each group). D, LPS-induced mPGES-1 mRNA expression. The⌬cycle threshold (CT) values, obtained by comparison with glyceraldehyde-3-phosphate
dehydrogenase, were normalized against the endothelial cells from mice treated with saline (n⫽ 3). In perivascular cells from one LPS-treated and two saline-treated mice, mPGES-1 mRNA levels were undetectable; the CT values were then set to 30, as a conservative sensitivity limit for preamplified samples. Error bars,SEM. *, P⬍ 0.05 compared with endothelial cells from saline-treated mice.
lung, ip LPS injection induced at the 3-h time point Cox-2 expression in endothelial-like cells in the walls of vessels (Fig. 8A), suggestive of pulmonary veins (Fig. 8D). Al-though extensive GFP labeling was seen in the lung of bone marrow-transplanted mice, no GFP positive cells ex-pressed Cox-2 (Fig. 8B, B1), and Cox-2 expression neither
colocalized with CD68 (Fig. 8C, C1), indicating that Cox-2 expression occurred in nonhematopoietic cells. These data thus confirmed the presence of LPS-induced synthesis of PGE2 in peripheral organs in both immune cells and nonimmune cells, being consistent with elevated plasma levels of PGE2 in both WT3 KO and KO3 WT
FIG. 6. Flow cytometry showing induced expression of mPGES-1 in liver macrophages (CD45⫹F4/80⫹) but not in liver endothelial cells (CD45⫺CD31⫹) 3 h after ip injection of LPS. A, Representative fluorescence-activated cell sorting profiles shows sorting strategy, which was the same as for brain samples (see Fig. 5). The pan-macrophage marker F4/80 was used instead of CD206 [used for the sorting of the brain tissue (see Fig. 5)] because it is a marker for Kupffer cells and because CD206 (that identifies brain macrophages) does not differentiate between liver macrophages and endothelial cells (50). B, Purity analysis of the sorted cell populations CD45⫺CD31⫹and CD45⫹F4/80⫹cells. Note that although the sorting for CD45⫺CD31⫹cells (endothelial cells) in A did not yield a very large population (cf. also Fig. 5), the population selected for further analysis was clearly shown to be distinct. C, LPS-induced mPGES-1 mRNA expression. The⌬CT value for liver Kupffer cells (KC) (CD45⫹F4/80⫹), obtained by comparison with GAPDH, was normalized against the⌬CT value for Kupffer cells from mice treated with saline (n ⫽ 3). In endothelial cells (EC) (CD45⫺CD31⫹), mPGES-1 mRNA levels were not detectable (nd), neither in LPS-treated nor NaCl-treated mice (n⫽ 3). Error bars,SEM. ***, P⬍ 0.001 compared with endothelial cells from saline-treated mice.
hybrids (cf. Fig. 3). They also showed that the transplanted cells had the capacity to produce prostaglandins at the ap-propriate sites, as evidenced by their induced Cox-2 expres-sion in the liver.
No induced expression of PGE2 synthesizing enzymes is seen at the time of the first phase of fever
Examination of brain, liver and lung at 40 min after the ip LPS injection showed no clear Cox-2 induction, and qPCR on brain endothelial cells and perivascular macrophages showed low levels of mPGES-1 mRNA, irrespective of whether mice had been injected with LPS or saline (not shown).
Discussion
The present study shows that the febrile response to LPS in mice is dependent on the presence of mPGES-1, the inducible terminal enzyme in the PGE2 synthesizing cas-cade (35), in cells of nonhematopoietic origin.
Further-more, we demonstrate that the febrile response is associated with increased PGE2 levels in the cerebrospinal fluid, whereas there was only a weak relation-ship between body temperature and PGE2 metabolites in plasma, indicating that the PGE2 production that gives rise to fever takes place in the brain. Finally, we show that endothelial cells are the site of Cox-2 induction in response to LPS and that these cells also show high induc-ible levels of mPGES-1. Because Cox-2-and mPGES-1-dependent PGE2 synthe-sis is necessary for LPS induced fever (1, 2, 4), the present data hence identify the brain endothelial cells as the cells critical for the febrile response.
Although this study is the first to identify the cells that are critical for the PGE2 production that gives rise to fe-ver, some previous observations that are consonant with the present finding should be noted. Thus, Scammell et al. (36) have demonstrated that local in-jection of a Cox inhibitor into the an-terior preoptic region attenuates the fe-ver elicited by systemic administration of LPS, and recently Vardeh et al. (37) showed that conditional deletion of Cox-2, the isoform critical for fever, in neural lineages did not affect the febrile response to LPS. Taken together, these two studies show that prostaglandin production in the brain is involved in the generation of fever but that this occurs in brain cells other than neurons and glia, being well in line with the present observations. Furthermore, quite recently Ridder et al. (38) showed that the MAPK kinase TGF-activated kinase 1 in brain endothelial cells is needed for IL-1-induced COX-2 production and that mice lacking the Tak1 gene in brain endothelial cells displayed a blunted fever response to IL-1. Although that study, unlike the present work, does not directly target the critical final step in the induced PGE2-synthesis, it is in strong support of an essential role of endothelial cells for immune-induced fever. Our findings that cells of hematopoietic origin play at best a minor role for the LPS induced febrile response, not only during the second and third phases of fever but also during the first initial phase, is at odds with the prevailing hypothesis, based largely on studies in rats, not mice, that has ascribed the initial fever to hematopoietic cells and in particular those in peripheral organs. That idea, which has the inherent weakness that PGE2 in the blood is rapidly
FIG. 7. 2 expression in the liver is associated with transplanted cells. A, Overview of
Cox-2 expression in liver 3 h after ip injection of LPS. cv, Central vein; p, portal vessels. Arrows point at clusters of Cox-2 immunoreactivity. B, F4/80-positive cells express Cox-2. Arrowheads point at dual-labeled cells, and arrow points at single-labeled cell. C, GFP-expressing cells are Cox-2 immunoreactive. Scale bar, 100m (A) and 20 m (B and C).
converted in vivo to its 13,14-dihydro-15-keto metabo-lite, with more than 90% of circulating PGE2 being cleared by a single passage through the lungs (30, 39), is based on the findings that iv injection of PGE2 in supra-physiological concentrations elicits fever (16, 40, 41), that mPGES-1 mRNA, which, however, is not the rate-limiting enzyme for the induced PGE2-synthesis (1), is induced more rapidly in peripheral organs than in the brain (42), and that LPS-induced fever can be attenuated by iv but not intracerebroventricular injection of PGE2-neutralizing antibodies (16). The latter experiment suffers from the obvious limitation that iv injected antibodies in fact may have better access to blood-brain barrier cells than those injected intracerebroventricularly. Importantly, it was never examined in the latter study (16) whether intrace-rebral injection of neutralizing antibody could attenuate the later phases of fever that are known to be associated with strong induction in the blood-brain barrier of PGE2-synthesizing enzymes or that the fever during these phases
were unaffected when the antibody was injected iv, making the interpretation of the obtained data hazardous.
Additional evidence for a role of pe-ripheral immune cells in fever comes from a study on mice that were hybrids for the Toll-like receptor 4 (43). In that study, mice that lacked functional Toll-like receptor 4 in hematopoietic cells showed absence of the first phase of fe-ver after iv injection of LPS (43), whereas the presence of functional re-ceptors in hematopoietic cells in mice of mutant background rescued the initial febrile response, suggesting a critical role for hematopoietic cells for the ini-tiation of fever. Although we here show in a similar experimental paradigm the opposite phenotypes, i.e. intact initial fever in WT mice with a KO genotype in hematopoietic cells, and absence of ini-tial fever in KO mice with WT hema-topoietic cells, the findings in the two studies are not necessarily mutually ex-clusive. The obvious reconciling inter-pretation is that intact Toll-like recep-tor 4 is necessary in hematopoietic cells for the release of proinflammatory cy-tokines (44), which in turn act on en-dothelial cells in the brain to induce PGE2 release and thereby eliciting fe-ver. The mediator acting on the endo-thelial cells may be IL-1 because IL-1 type 1 receptors are exclusively expressed on endothelial cells and not on perivascular cells (10) and selective genetic ablation of IL-1 type 1 receptors on brain endothelial cells yields mice refractory to the fever producing properties of IL-1 (45). It should in this context also be noted that brain macrophages are not likely to be able to respond to pe-ripherally injected LPS because brain uptake of circulating LPS is extremely low (46).
Previous morphological studies on the expression of PG synthesizing enzymes in the brain have largely been done in rats, in which both Cox-2 and mPGES-1 have been demonstrated with immunohistochemistry and in situ hy-bridization. Although the reports on which cells that ex-press Cox-2 are contradictory, as discussed above, mPGES-1 expression has consistently been ascribed only to endothelial cells (8, 12, 13, 47). In the single previous study in the mouse in which cell specific markers were used, Cox-2 expression was shown in endothelial cells (48), and so far mPGES-1 has been demonstrated in the mouse
FIG. 8. Cox-2 expression in the lung is distinct from the transplanted cells. A, LPS-induced
Cox-2 immunoreactivity in blood vessel-associated cells (A2). A1 shows staining pattern in NaCl-treated mice, and A3 shows background staining in LPS-treated Cox-2 KO mice. vl, Vessel lumen. B, Dual staining for Cox-2 and GFP. Detail at higher magnification in B1 shows that Cox-2-labeled profiles and GFP-positive (i.e. transplanted) cells are distinct. C, Dual staining for Cox-2 and the monocyte and lung macrophage marker CD68. Detail in C1 shows that although CD68-expressing cells are interspersed among the Cox-2-labeled profiles, the two populations are distinct. D, The Cox-2-expressing profiles are located to the walls of vessels that stain for the von Willebrand factor. Scale bar, 50m (A–D) and 10 m (B1 and C1).
brain only by in situ hybridization, without identifying the expressing cells (33). We were unable to obtain an mPGES-1 signal in the brain either by immunohistochemistry or in situ hybridization, despite previous successful stainings of rat brain tissue with both techniques (7, 11); however, our al-ternative approach of qPCR analysis of endothelial cells and perivascular cells isolated by flow cytometry, clearly dem-onstrated a strong LPS-induced induction in endothelial cells, with, in contrast, very low mRNA levels in perivascular cells, pinpointing the endothelial cells as the PGE2 producing cells in the brain vasculature.
Although the increased PGE2 levels in the brain at the peak of LPS fever, i.e. at around 3 h after the injection, are consistent with induced transcription and subsequent translation of PGE2-synthesizing enzymes, there was no evidence at the time of the initial phase of fever neither for induced Cox-2 expression in brain (or in liver or lung), nor for induced mPGES-1 transcription in brain vascular cells. This is in contrast to what has been reported in rats (16), suggesting that species differences may exist. Neverthe-less, also at this time point, PGE2 levels were increased in the cerebrospinal fluid in the chimeras that displayed fever (WT3WT and KO3WT) as compared with the chimeras that were afebrile (WT3KO and KO3KO). Considering that the temperature elevation starts within less than 15 min after iv injection of LPS (Fig. 2B), the data suggest that this early PGE2 release is not dependent on newly translated prostaglandin synthesizing enzymes but occurs from a pre-existing pool. Notably, the time course of the initial phase of fever follows the dynamic of the temperature response when PGE2 is injected intracerebroventricularly (49): a rapid rise followed by a somewhat slower decrease and is hence con-sistent with a momentary release of PGE2 and its subsequent removal from the brain. The present data are therefore con-sonant with the interpretation that this release, similar to that which results in the later phases of fever, occurs from central mPGES-1-expressing nonhematopoietic cells, i.e. brain vas-cular endothelial cells.
Acknowledgments
Author contributions are as follows: L.E., A.L., L.M., and U.K. carried out the temperature recordings; J.R., L.E., and A.E. did the immunohistochemistry; L.E. and J.R. performed the immu-noassays; P.L. established the method for whole-body irradia-tion; J.-I.J., L.E., L.M., U.K., and A.E. did the bone marrow transplantations; J.-I.J. did the flow cytometry sorting and the analysis of the peripheral blood; H.Q. did the flow cytometry sorting and the analysis of the brain cells; A.E. and A.M.V. car-ried out the qPCR; A.B., J.-I.J., and M.S. conceived the experi-ments; L.E., J.R., A.E., A.M.V., D.E., J.-I.J., M.S., and A.B. an-alyzed the data; and A.B. wrote the paper.
Address all correspondence and requests for reprints to: Dr. Anders Blomqvist, Division of Cell Biology, Department of Clin-ical and Experimental Medicine, Faculty of Health Sciences, Linköping University, S-581 85 Linköping, Sweden. E-mail: anders.blomqvist@liu.se.
This work was supported by Grant 7879 from the Swedish Re-search Council, Grant 100533 from the Swedish Cancer Founda-tion, the Swedish Brain FoundaFounda-tion, and Gustav V:s 80-års Fond. Present address for A.L.: Department of Integrative Medical Biology, Umeå University, Umeå, Sweden.
Disclosure Summary: The authors have nothing to disclose.
References
1. Li S, Wang Y, Matsumura K, Ballou LR, Morham SG, Blatteis CM 1999 The febrile response to lipopolysaccharide is blocked in cy-clooxygenase-2(⫺/⫺), but not in cyclooxygenase-1(⫺/⫺) mice. Brain Res 825:86 –94
2. Engblom D, Saha S, Engström L, Westman M, Audoly LP, Jakobsson PJ,
Blomqvist A 2003 Microsomal prostaglandin E synthase-1 is the central
switch during immune-induced pyresis. Nat Neurosci 6:1137–1138 3. Saha S, Engström L, Mackerlova L, Jakobsson PJ, Blomqvist A 2005
Impaired febrile responses to immune challenge in mice deficient in microsomal prostaglandin E synthase-1. Am J Physiol Regul Integr Comp Physiol 288:R1100 –R1107
4. Nilsberth C, Hamzic N, Norell M, Blomqvist A 2009 Peripheral lipopolysaccharide administration induces cytokine mRNA expres-sion in the viscera and brain of fever-refractory mice lacking micro-somal prostaglandin E synthase-1. J Neuroendocrinol 21:715–721 5. Steiner AA, Chakravarty S, Robbins JR, Dragic AS, Pan J,
Herken-ham M, Romanovsky AA 2005 Thermoregulatory responses of rats
to conventional preparations of lipopolysaccharide are caused by lipopolysaccharide per se—not by lipoprotein contaminants. Am J Physiol Regul Integr Comp Physiol 289:R348 –R352
6. Lazarus M, Yoshida K, Coppari R, Bass CE, Mochizuki T, Lowell BB,
Saper CB 2007 EP3 prostaglandin receptors in the median preoptic
nucleus are critical for fever responses. Nat Neurosci 10:1131–1133 7. Ek M, Engblom D, Saha S, Blomqvist A, Jakobsson PJ,
Ericsson-Dahlstrand A 2001 Inflammatory response: pathway across the
blood-brain barrier. Nature 410:430 – 431
8. Yamagata K, Matsumura K, Inoue W, Shiraki T, Suzuki K, Yasuda
S, Sugiura H, Cao C, Watanabe Y, Kobayashi S 2001 Coexpression
of microsomal-type prostaglandin E synthase with cyclooxygen-ase-2 in brain endothelial cells of rats during endotoxin-induced fever. J Neurosci 21:2669 –2677
9. Inoue W, Matsumura K, Yamagata K, Takemiya T, Shiraki T,
Ko-bayashi S 2002 Brain-specific endothelial induction of prostaglandin
E(2) synthesis enzymes and its temporal relation to fever. Neurosci Res 44:51– 61
10. Konsman JP, Vigues S, Mackerlova L, Bristow A, Blomqvist A 2004 Rat brain vascular distribution of interleukin-1 type-1 receptor immuno-reactivity: Relationship to patterns of inducible cyclooxygenase expression by peripheral inflammatory stimuli. J Comp Neurol 472:113–129 11. Engblom D, Ek M, Saha S, Ericsson-Dahlstrand A, Jakobsson PJ,
Blomqvist A 2002 Prostaglandins as inflammatory messengers
across the blood-brain barrier. J Mol Med 80:5–15
12. Engblom D, Ek M, Andersson IM, Saha S, Dahlström M, Jakobsson PJ,
Ericsson-Dahlstrand A, Blomqvist A 2002 Induction of microsomal
pros-taglandin E synthase in the rat brain endothelium and parenchyma in ad-juvant-induced arthritis. J Comp Neurol 452:205–214
13. Ozaki-Okayama Y, Matsumura K, Ibuki T, Ueda M, Yamazaki Y,
Tanaka Y, Kobayashi S 2004 Burn injury enhances brain
microsomal prostaglandin E synthase in cerebral vascular endothe-lial cells in rats. Crit Care Med 32:795– 800
14. Elmquist JK, Breder CD, Sherin JE, Scammell TE, Hickey WF,
De-witt D, Saper CB 1997 Intravenous lipopolysaccharide induces
cy-clooxygenase 2-like immunoreactivity in rat brain perivascular mi-croglia and meningeal macrophages. J Comp Neurol 381:119 –129 15. Schiltz JC, Sawchenko PE 2002 Distinct brain vascular cell types manifest inducible cyclooxygenase expression as a function of the strength and nature of immune insults. J Neurosci 22:5606 –5618 16. Steiner AA, Ivanov AI, Serrats J, Hosokawa H, Phayre AN, Robbins
JR, Roberts JL, Kobayashi S, Matsumura K, Sawchenko PE, Ro-manovsky AA 2006 Cellular and molecular bases of the initiation of
fever. PLoS Biol 4:e284
17. Serrats J, Schiltz JC, García-Bueno B, van Rooijen N, Reyes TM,
Sawchenko PE 2010 Dual roles for perivascular macrophages in
immune-to-brain signaling. Neuron 65:94 –106
18. Cao C, Matsumura K, Yamagata K, Watanabe Y 1996 Endothelial cells of the rat brain vasculature express cyclooxygenase-2 mRNA in response to systemic interleukin-1: a possible site of prosta-glandin synthesis responsible for fever. Brain Res 733:263–272 19. Matsumura K, Cao C, Ozaki M, Morii H, Nakadate K, Watanabe
Y 1998 Brain endothelial cells express cyclooxygenase-2 during
li-popolysaccharide-induced fever: light and electron microscopic im-munocytochemical studies. J Neurosci 18:6279 – 6289
20. Trebino CE, Stock JL, Gibbons CP, Naiman BM, Wachtmann TS,
Umland JP, Pandher K, Lapointe JM, Saha S, Roach ML, Carter D, Thomas NA, Durtschi BA, McNeish JD, Hambor JE, Jakobsson PJ, Carty TJ, Perez JR, Audoly LP 2003 Impaired inflammatory and
pain responses in mice lacking an inducible prostaglandin E syn-thase. Proc Natl Acad Sci USA 100:9044 –9049
21. Okabe M, Ikawa M, Kominami K, Nakanishi T, Nishimune Y 1997 ’Green mice’ as a source of ubiquitous green cells. FEBS Lett 407: 313–319
22. Morham SG, Langenbach R, Loftin CD, Tiano HF, Vouloumanos
N, Jennette JC, Mahler JF, Kluckman KD, Ledford A, Lee CA, Smithies O 1995 Prostaglandin synthase 2 gene disruption causes
severe renal pathology in the mouse. Cell 83:473– 482
23. Engström L, Rosén K, Angel A, Fyrberg A, Mackerlova L, Konsman
JP, Engblom D, Blomqvist A 2008 Systemic immune challenge
ac-tivates an intrinsically regulated local inflammatory circuit in the adrenal gland. Endocrinology 149:1436 –1450
24. Galea I, Palin K, Newman TA, Van Rooijen N, Perry VH, Boche D 2005 Mannose receptor expression specifically reveals perivascular macrophages in normal, injured, and diseased mouse brain. Glia 49:375–384
25. Takezawa R, Watanabe Y, Akaike T 1995 Direct evidence of mac-rophage differentiation from bone marrow cells in the liver: a pos-sible origin of Kupffer cells. J Biochem 118:1175–1183
26. Holness C, Simmons DL 1993 Molecular cloning of CD68, a human macrophage marker related to lysosomal glycoproteins. Blood 81: 1607–1613
27. RamprasadMP,FischerW,WitztumJL,SambranoGR,QuehenbergerO,
Steinberg D 1995 The 94- to 97-kDa mouse macrophage membrane
pro-tein that recognizes oxidized low density lipopropro-tein and phosphatidylser-ine-rich liposomes is identical to macrosialin, the mouse homologue of human CD68. Proc Natl Acad Sci USA 92:9580–9584
28. Romanovsky AA, Almeida MC, Aronoff DM, Ivanov AI, Konsman JP,
Steiner AA, Turek VF 2005 Fever and hypothermia in systemic
inflam-mation: recent discoveries and revisions. Front Biosci 10:2193–2216 29. Rudaya AY, Steiner AA, Robbins JR, Dragic AS, Romanovsky AA
2005 Thermoregulatory responses to lipopolysaccharide in the mouse: dependence on the dose and ambient temperature. Am J Physiol Regul Integr Comp Physiol 289:R1244 –R1252
30. Hamberg M, Samuelsson B 1971 On the metabolism of prostaglan-dins E 1 and E 2 in man. J Biol Chem 246:6713– 6721
31. Samuelsson B, Granström E, Green K, Hamberg M, Hammarström
S 1975 Prostaglandins. Annu Rev Biochem 44:669 – 695
32. Granström E, Kindahl H 1982 Species differences in circulating prostaglandin metabolites relevance for the assay of prostaglandin release. Biochim Biophys Acta 713:555–569
33. Gosselin D, Rivest S 2008 MyD88 signaling in brain endothelial cells is essential for the neuronal activity and glucocorticoid release dur-ing systemic inflammation. Mol Psychiatry 13:480 – 497
34. Givalois L, Dornand J, Mekaouche M, Solier MD, Bristow AF, Ixart
G, Siaud P, Assenmacher I, Barbanel G 1994 Temporal cascade of
plasma level surges in ACTH, corticosterone, and cytokines in en-dotoxin-challenged rats. Am J Physiol 267:R164 –R170
35. Jakobsson PJ, Thorén S, Morgenstern R, Samuelsson B 1999 Iden-tification of human prostaglandin E synthase: a microsomal, gluta-thione-dependent, inducible enzyme, constituting a potential novel drug target. Proc Natl Acad Sci USA 96:7220 –7225
36. Scammell TE, Griffin JD, Elmquist JK, Saper CB 1998 Microinjec-tion of a cyclooxygenase inhibitor into the anteroventral preoptic region attenuates LPS fever. Am J Physiol 274:R783–R789 37. Vardeh D, Wang D, Costigan M, Lazarus M, Saper CB, Woolf CJ,
Fitzgerald GA, Samad TA 2009 COX2 in CNS neural cells mediates
mechanical inflammatory pain hypersensitivity in mice. J Clin Invest 119:287–294
38. Ridder DA, Lang MF, Salinin S, Röderer JP, Struss M, Maser-Gluth
C, Schwaninger M 2011 TAK1 in brain endothelial cells mediates
fever and lethargy. J Exp Med 208:2615–2623
39. Hamberg M, Samuelsson B 1971 Metabolism of prostaglandin E2 in guinea pig liver. II. Pathways in the formation of the major me-tabolites. J Biol Chem 246:1073–1077
40. Romanovsky AA, Ivanov AI, Karman EK 1999 Blood-borne, albu-min-bound prostaglandin E2 may be involved in fever. Am J Physiol 276:R1840 –R1844
41. Ootsuka Y, Blessing WW, Steiner AA, Romanovsky AA 2008 Fever response to intravenous prostaglandin E2 is mediated by the brain but does not require afferent vagal signaling. Am J Physiol Regul Integr Comp Physiol 294:R1294 –R1303
42. Ivanov AI, Pero RS, Scheck AC, Romanovsky AA 2002 Prostaglandin E(2)-synthesizing enzymes in fever: differential transcriptional regulation. Am J Physiol Regul Integr Comp Physiol 283:R1104–R1117
43. SteinerAA,ChakravartyS,RudayaAY,HerkenhamM,RomanovskyAA 2006 Bacterial lipopolysaccharide fever is initiated via Toll-like receptor 4 on hematopoietic cells. Blood 107:4000–4002
44. Chakravarty S, Herkenham M 2005 Toll-like receptor 4 on nonhe-matopoietic cells sustains CNS inflammation during endotoxemia, independent of systemic cytokines. J Neurosci 25:1788 –1796 45. Ching S, Zhang H, Belevych N, He L, Lai W, Pu XA, Jaeger LB, Chen
Q, Quan N 2007 Endothelial-specific knockdown of interleukin-1
(IL-1) type 1 receptor differentially alters CNS responses to IL-1 de-pending on its route of administration. J Neurosci 27:10476–10486 46. Banks WA, Robinson SM 2010 Minimal penetration of
lipopoly-saccharide across the murine blood-brain barrier. Brain Behav Im-mun 24:102–109
47. Oka Y, Ibuki T, Matsumura K, Namba M, Yamazaki Y, Poole S,
Tanaka Y, Kobayashi S 2007 Interleukin-6 is a candidate molecule
that transmits inflammatory information to the CNS. Neuroscience 145:530 –538
48. Rummel C, Inoue W, Sachot C, Poole S, Hübschle T, Luheshi GN 2008 Selective contribution of interleuk6 and leptin to brain in-flammatory signals induced by systemic LPS injection in mice. J Comp Neurol 511:373–395
49. Nilsberth C, Elander L, Hamzic N, Norell M, Lönn J, Engström L,
Blomqvist A 2009 The role of interleukin-6 in
lipopolysaccharide-induced fever by mechanisms independent of prostaglandin E2. En-docrinology 150:1850 –1860
50. Asumendi A, Alvarez A, Martinez I, Smedsrød B, Vidal-Vanaclocha
F 1996 Hepatic sinusoidal endothelium heterogeneity with respect
to mannose receptor activity is interleukin-1 dependent. Hepatology 23:1521–1529
