Fever is an ancient host-defense response and a common symptom of infection and systemic inflammation. Since Milton and Wendlandt [1] discovered the pyrogenic activity of prostaglandins (PGs) of the E series, and Vane [2] found that nonsteroidal anti-inflammatory drugs block fever by inhibiting PG synthesis, it has been accepted that fever is mediated by PGs, specifically PGE₂ [3–6]. PGE₂ synthesis occurs in three steps: (1) membrane phospholipids are converted to arachidonic acid by phospholipase A₂ (PLA₂); (2) arachidonic acid is converted to PGH₂ by cyclooxygenase (COX); and (3) PGH₂ is isomerized to PGE₂ by a terminal PGE synthase (PGES) [6,7]. It has been shown in rats [8–14] and mice [15,16] that COX-2 and microsomal PGES-1 (mPGES-1) are transcriptionally up-regulated in endothelial and perivascular cells of brain microvessels between 1.5 and 12 h after administration of pyrogenic doses of bacterial lipopolysaccharide (LPS). Furthermore, Scammell et al. [17] have shown that microinjection of the COX inhibitor ketorolac into the preoptic region attenuates the febrile response over 1.5–6 h after intravenous (i.v.) injection of LPS in rats. These results indicate that febrigenic PGE₂ is produced centrally.

It should be considered, however, that the initiation of fever precedes by approximately 1 h the earliest time point at which PGE₂-synthesizing enzymes have been shown to be up-regulated in the brain. In a thermoneutral environment, i.v. LPS typically causes in rats and mice a polyphasic fever, and the first phase of this response starts at approximately 0.5 h post-LPS [18,19]. Because the first phase is sensitive to ambient temperature and can be readily masked by the stress hyperthermia associated with animal handling and LPS injection [19,20], this phase often escapes detection and remains the least studied component of the febrile response. The first phase of LPS fever was not investigated in any of the abovementioned studies of the source of febrigenic PGE₂. We [21–23] and others [24–26] have hypothesized that, unlike the second and subsequent febrile phases, the first phase of fever is triggered by peripherally produced PGE₂. Over the last two decades, several studies have attempted to test this hypothesis, but the results obtained have been inconclusive, contradictory, or incomplete (for details, see Results and Discussion). In particular, the location (inside or outside the brain) and phenotypes of the cells involved in the initiation of fever are unknown, as are the steps of the PGE₂-synthesizing cascade that are initially activated to trigger the fever response. By closing these gaps, the present study identifies the cellular and molecular bases of the initiation of fever.

The question as to whether peripherally (i.v. or intra-arterially) administered PGE₂ causes fever remains controversial. Although there are reports of peripherally injected PGE₁ and PGE₂ being pyrogenic in several species of laboratory animals [24,27], there are at least as many documented failures to induce fever by peripheral administration of PGE [24,28,29]. The latter, negative results can be explained, at least partially, as due to self-aggregation of PGE in aqueous solutions and the subsequent loss of biological activity. Indeed, PGE₂ was found to be highly pyrogenic in rabbits when infused in an albumin-bound (monomeric), but not in a free (aggregated) form [21]. Albumin is the principal carrier of PGE₂ in the circulation, and up to 99% of circulating PGE₂ is albumin-bound [30].

In the present study, a 2:1 (molar ratio) PGE₂–albumin complex was prepared by adding PGE₂ (all reagents are from Sigma-Aldrich, St. Louis, Missouri, United States, unless specified otherwise) and bovine serum albumin (BSA) to pyrogen-free saline, and then sonicating this mixture for 3 min and incubating it at 37 °C for 1 h. In a thermoneutral environment, the rats were infused i.v. with BSA-bound PGE₂ (280 or 560 μg/kg, 100 μl/kg/min, 10 min). Based on the assumptions that PGE₂ is evenly distributed in the extracellular compartment (20% of the body mass) and that its half-life is 1 min [31], it can be estimated that the protocol used elevates the plasma concentration of PGE₂ by 350 pg/ml (low dose) or 700 pg/ml (high dose) at 12 min after the beginning of infusion. These concentrations are within the physiological range [24,32]. Whereas BSA had no thermoregulatory effect, the PGE₂–BSA complex caused a dose-dependent rise in deep body (colonic) temperature (T_c; Figure 1A). This fever response was brought about, at least in part, by tail skin vasoconstriction, as evident from a decrease in the heat loss index (the quotient of two temperature gradients: skin-ambient/colonic-ambient [33]). Hence, when administered in its most relevant form (albumin complex) and at physiologically relevant doses, peripheral PGE₂ is pyrogenic in rats.

Figure 1 Circulating PGE2 Initiates LPS Fever in Rats: Circumstantial Evidence

How circulating, albumin-bound PGE₂ causes fever remains speculative. Activation of vagal afferents by PGE₂ has been proposed [34], but the fact that vagotomy does not affect the first febrile phase (for discussion, see [35,36]) makes this mechanism unlikely. An alternative scenario seems more plausible. Binding to albumin prevents the rapid enzymatic inactivation of PGE₂ [37,38], thus allowing it to reach a distant site. A good candidate for such a site is the preoptic hypothalamus, which is highly sensitive to the pyrogenic effect of PGE₂ [39]. Once dissociated from albumin at the target site, PGE₂ may be carried into the brain tissue by transporters expressed at the blood–brain barrier (BBB) [6,26,40]. It should be noted, however, that this scenario is speculative and needs to be tested experimentally.

Having shown that peripheral PGE₂ is pyrogenic in rats, we asked whether blood levels of PGE₂ are elevated at the onset of the first phase of LPS fever. Fever was induced by administering 0111:B4 Escherichia coli LPS (10 μg/kg) nonstressfully via the extension of a preimplanted venous (jugular) catheter to rats kept in a thermoneutral environment (see Materials and Methods for details). The first febrile phase started at approximately 30-min post-LPS and was brought about, at least partially, by tail skin vasoconstriction (Figure 1B). At 40 min (the time corresponding both to the maximal rate of rise in body temperature and to the maximal thermoeffector activity that underlies this rise), samples of venous and arterial blood were collected from LPS-treated (febrile) and saline-treated (afebrile) rats, and the concentration of PGE₂ in the venous and arterial blood was measured by enzyme immunoassay. The venous blood gathers PGE₂ synthesized in the tissues, and the arterial blood delivers it to the brain, the presumptive site of the febrigenic action of circulating PGE₂ [39]. Consistent with the marked catabolism of PGE₂ in the lungs [41], the level of PGE₂ was lower in the arterial than in the venous blood plasma in both afebrile and febrile rats (Figure 1C). However, both the venous and arterial concentrations of PGE₂ were substantially (~2.5 times) higher in the febrile rats as compared to the afebrile controls. These data show that the level of circulating PGE₂, most importantly in the arterial blood, is increased at the onset of the first febrile phase.

Several studies aimed at determining the source of febrigenic PGE₂ have compared the antipyretic effects of nonsteroidal anti-inflammatory drugs administered peripherally (i.v. or intraperitoneally) and centrally (intracerebroventricularly [i.c.v.]). The drugs used included indomethacin [25], nimesulide [32], and keterolac (present study; unpublished data). All these studies faced multiple methodological problems, including acute thermoregulatory effects of the drug administered i.c.v. (present study), the ability of drugs to cross the BBB, and consequently, their tendency to be distributed evenly between the peripheral compartment and the brain [32]. We proposed [6] that selective neutralization of circulating PGE₂ using an antibody is a better approach to test the hypothesis that peripherally produced PGE₂ initiates fever. Being large proteins (160 kDa), antibodies cannot cross the BBB; this eliminates uncertainty common in experiments involving nonsteroidal anti-inflammatory drugs. The antibody used in the present study was raised against a PGE₂–BSA complex in rabbits. It displayed a high affinity to PGE₂ (association constant of 6.3 × 10¹⁰ M⁻¹, as determined by Scatchard plot) and a low cross-reactivity with other prostanoids (<15% for PGF_1α and PGB₂, and <9% for PGA₂, PGF_2α, and PGB₁). The rats were pretreated i.v. with the anti-PGE₂ antibody (neat antiserum; 100 μl/kg/min, 120 min) or with normal rabbit serum, and LPS was injected 18 h later, i.e., at the time when the injected antibody is expected to achieve a steady-state level in the circulation [42]. The results of this experiment are shown in Figure 2A–2C. The antibody (but not normal serum) suppressed the first phase of LPS fever: both the rise in T_c and the associated decrease in the heat loss index were delayed and significantly attenuated (Figure 2A). Immediately after the temperature response was recorded, a sample of venous blood and the whole brain (cleared of blood) were collected for immunoenzymatic determination of the anti-PGE₂ antibody. The antibody was found at a high concentration in the blood plasma, but was below the detection limit in the brain tissue (Figure 2C). To rule out the possibility that a minute, undetectable amount of antibody in the brain might have accounted for the suppression of fever, we administered a low dose (2.7 μl/min, 15 min) of the anti-PGE₂ antibody or normal serum i.c.v., and injected the rats with LPS 18 h later. The rats injected with the anti-PGE₂ antibody i.c.v. had a detectable level of the antibody in the brain (Figure 2C), but their febrile response to LPS was unaffected (Figure 2B). A large fraction of the antibody given i.c.v. leaked into the blood, presumably reflecting the asymmetric nature of the BBB (its major role is to limit transport in the blood-to-brain direction, but not in the opposite direction) or possibly because the BBB was breached in this experimental group by the implanted i.c.v. cannula. Importantly, however, the plasma antibody concentration in the rats treated with the i.c.v. antibody was approximately 60 times lower than that in the rats treated with the i.v. antibody (Figure 2C). It is concluded that minute amounts of the anti-PGE₂ antibody in the brain (even when detectable) are not sufficient to suppress the initiation of fever, and that the cause of the delayed and attenuated first febrile phase observed in the rats pretreated with i.v. antibody was neutralization of PGE₂ outside the BBB.

Figure 2 Circulating PGE2 Initiates LPS Fever in Rats: Direct Evidence

Having demonstrated that circulating PGE₂ is indeed responsible, at least partially, for triggering LPS-induced fever, we investigated which step of the PGE₂ biosynthetic pathway is activated at the onset of the febrile response. Previously, we reported that the onset of the first phase of LPS fever is associated with large increases of COX-2 and mPGES-1 mRNAs in the lung and liver and with a moderate increase of COX-2 (but not mPGES-1) mRNA in the hypothalamus [22]. However, it remained to be determined whether the observed transcriptional changes translate into changes in the corresponding protein contents at such an early time point (40 min) after LPS administration. We had also shown [22] that neither cytosolic PLA₂-α (cPLA₂-α) nor either of the two secretory PLA₂ studied (II and V) is transcriptionally up-regulated at the onset of fever. This finding, however, does not exclude the possibility that cPLA₂ is activated posttranscriptionally by phosphorylation, the principal mechanism of activation for this enzyme [43]. In the present study, we determined the contents of phosphorylated cPLA₂ (p-cPLA₂), COX-2, and mPGES-1 proteins by Western blot in the lung, liver, and hypothalamus at 40 min after injection of LPS or saline, a time that corresponds to the onset of the first febrile phase in LPS-treated rats (Figure 1B). COX-2–positive cells were also studied in all three tissues by immunohistochemistry using two different protocols of sample preparation (see Material and Methods). None of the enzymes studied was increased at the protein level in the hypothalamus of the LPS-treated rats as compared to the saline-treated controls (Figure 3). Neither did the immunohistochemical analysis reveal any increase in the number of COX-2–positive hypothalamic cells at the onset of fever, although the same antibody readily detected a surge in the number of COX-2–positive endotheliocytes in the hypothalamic microvasculature at later stages of LPS fever, in both the present study (positive controls; unpublished data) and previous studies [11,12]. In the lung, LPS increased the contents of p-cPLA₂ and COX-2 (Figure 3), and augmented the number of cells containing COX-2 (Figure 4), but did not alter the protein level of constitutively expressed mPGES-1 (Figure 3). In the liver, the immunohistochemical analysis (which is more sensitive) revealed a surge in the number of COX-2–positive cells at the onset of fever (Figure 4), whereas the Western blot analysis (less sensitive) found a tendency for an increase in the overall content of COX-2 and no changes in the content of either p-cPLA₂ or mPGES-1 (Figure 3). We also found that inflammatory signaling (assessed by a decrease in the content of the nuclear factor-κB inhibitor, IκB-α [44]) was activated in the lung and liver, but not in the hypothalamus, at the onset of LPS fever (Figure 3).

Figure 3 Mechanism of Activation of PGE2 Synthesis at the Onset of LPS Fever in Rats

Figure 4 Identification of the Pulmonary and Hepatic Cells Producing PGE2 at the Onset of LPS Fever in Rats

These results show that the onset of the first febrile phase is associated with activation of inflammatory signaling and increased PGE₂ synthesis in the periphery. The early activation of PGE₂ synthesis involves phosphorylation of cPLA₂ (lung) and transcriptional up-regulation of COX-2 (lung and liver). Transcriptional up-regulation is the main (although not the only [45,46]) mechanism of activation for this enzyme [6,7]. Hence, the increased circulating level of PGE₂ at the onset of the first febrile phase may be explained by the following enzymatic events in the lung and liver: production of arachidonic acid by activated (phosphorylated) cPLA₂ → conversion of arachidonic acid to PGH₂ by up-regulated COX-2 → isomerization of PGH₂ into PGE₂ by constitutively expressed mPGES-1. Whereas the physiological importance of cPLA₂ and mPGES-1 in the first febrile phase remains to be confirmed in studies with pharmacological or genetic blockade of these enzymes, the indispensable role of COX-2 (and the uninvolvement of COX-1) in the first phase of LPS fever have been demonstrated in our recent study in knockout mice [47].

Preferential location of the synthesis of febrigenic PGE₂ in the liver and lungs (but not in the brain) deserves special discussion. The fact that the i.v. antibody attenuated the first febrile phase but did not abolish it completely (Figure 2A) may be due to incomplete neutralization of circulating PGE₂. However, it may also reflect a contribution of centrally produced PGE₂ (e.g., by a small number of hypothalamic cells that express COX-2 constitutively) to the development of the first phase of LPS fever. Although we cannot rule out such a contribution, it is noteworthy that multiple methods used in our present and previous [22] studies (Table 1) found a profound activation of PGE₂ synthesis in the periphery, but hardly any signs (none at the protein level) of activation of hypothalamic PGE₂ synthesis.

Table 1 Inflammatory Signaling and PGE2 Synthesis Are Selectively Activated at the Onset of the First Phase of LPS Fever in the Periphery (Lung and Liver) but Not in the Brain (Hypothalamus)

To identify the pulmonary and hepatic producers of PGE₂, we first determined how the cells that become COX-2 positive at the onset of LPS fever relate to the histological elements revealed by eosin staining; this analysis was performed in freshly frozen samples. In the lung, COX-2–positive cells were found to cluster around alveoli, often forming what looked like cell chains (unpublished data). In the liver, the parenchyma did not stain for COX-2, and the vast majority of COX-2–positive cells were located in the stromal compartment, often in close proximity to sinusoids. Some COX-2–positive cells were also found around the central vein (a small vein that gathers the blood from sinusoids) and in the visceral peritoneum covering the liver (unpublished data). We then double-stained lung and liver for COX-2 and either the macrophage marker ED2 [48] or the endothelial marker RECA1 [49]; this analysis was performed in paraformaldehyde-fixed samples (Figure 4). In the lung, 89 ± 6% (mean ± standard error [SE] of five samples) of COX-2–positive cells were macrophages (ED2 positive), and 11% were unidentified (ED2 and RECA1 negative). In the liver, 83 ± 2% of the COX-2–positive cells were macrophages (ED2 positive), 9 ± 1% were endotheliocytes (RECA1 positive), and 8% remained unidentified. The key role of macrophages in the initiation of fever agrees with our recent finding that the first febrile phase depends entirely on the recognition of LPS (via the Toll-like receptor-4) by bone marrow-derived cells [50].

In summary, the present study shows that the first phase of LPS fever is initiated (at least partially) by PGE₂ that originated in peripheral tissues. Activation of PGE₂ synthesis at the onset of the first phase of LPS fever involves phosphorylation of cPLA₂, transcriptional up-regulation of COX-2, and possibly other mechanisms. The vast majority of the PGE₂-producing cells are macrophages. These findings challenge the predominant view that fever is initiated exclusively by inflammatory mediators produced at the level of the BBB. These findings, however, do not contradict the principal role of the centrally produced PGE₂ in the second and subsequent febrile phases.

For collection of tissue samples for the immunoassay and Western blot protocols, each rat was perfused through the left ventricle (right atrium cut) with 100 ml of 10 mM phosphate-buffered saline (PBS; pH 7.4). The right medial lobe of the liver, the medial lobe of the right lung, and either the hypothalamus (for Western blot) or the entire brain were collected rapidly and snap frozen in liquid nitrogen. The samples were stored at −80 °C.

Tissue specimens for the immunohistochemistry protocols were collected using two methods. For immunohistochemical analysis of freshly frozen (non-fixed) tissues, a rat was perfused with PBS (50 ml). The liver, right lung, and brain were excised, frozen in dry ice powder, and stored at −80 °C. The specimens were cryosectioned (section thickness, 14 μm) immediately before the analysis. For immunohistochemical analysis of fixed tissues, a rat was perfused with 50 ml of heparinized (10 U/ml) PBS followed by 200 ml of a 4% paraformaldehyde solution in PBS. The liver, right lung, and brain were excised, postfixed in the paraformaldehyde solution for 2 h, transferred to PBS containing 20% sucrose, and kept in this solution at 4 °C for 24 h. The specimens were then frozen, cryosectioned (section thickness, 30 μm), and immediately subjected to the immunohistochemical analysis.

Technical assistance of D. L. Oliveira is appreciated.