Immune-to-brain signaling and central prostaglandin E₂ synthesis in fasted rats with altered lipopolysaccharide-induced fever

Abstract

Acute starvation attenuates the fever response to pathogens in several mammalian species. The underlying mechanisms responsible for this effect are not fully understood but may involve a compromised immune and/or thermoregulatory function, both of which are prerequisites for fever generation. In the present study, we addressed whether the impaired innate immune response contributes to the reported attenuation of the fever response in fasted rats during LPS-induced inflammation. Animals fasted for 48 h exhibited a significant and progressive hypothermia prior to drug treatment. An intraperitoneal injection of LPS (100 μg/kg) resulted in a significantly attenuated fever in the fasted animals compared with the fed counterparts. This attenuation was accompanied by the diminution in the concentration of some [TNF and IL-1 receptor antagonist (RA)] but not all (IL-1β and IL-6) of the plasma cytokines normally elevated in association with the fever response. Nevertheless, fasting had no effect on the LPS-induced inflammatory responses at the level of the brain, as assessed by mRNA expressions of inhibitory factor(I)-κB, suppressor of cytokine signaling (SOCS3), IL-1β, cyclooxygenase (COX)-2, and microsomal PGE synthase (mPGES)-1 in the hypothalamus, as well as by PGE₂ elevations in the cerebrospinal fluid. In contrast, fasting significantly attenuated the fever response to central PGE₂ injection. These results show that fasting does not alter the febrigenic signaling from the periphery to the brain important for central PGE₂ synthesis but does affect thermoregulatory mechanisms downstream of and/or independent of central PGE₂ action.

fever is a common response to various types of infection and constitutes an important component of an adaptive strategy for fighting disease (14, 23). The development of fever is a finely tuned, complex event that involves both the peripheral immune system and the brain, through which a series of inflammatory and metabolic processes are regulated (44). The sum of the changes required to mount an effective fever response is an energy-demanding process that can be influenced by the host's energy status. This was clearly shown by numerous studies reporting that acute starvation, such that caused by fasting, negatively affects the fever response to pathogens in various species of experimental animals (22, 37, 50, 51, 53). In some of these studies, the underlying mechanisms responsible for the depression of fever were linked to a compromised metabolic function of the starved animals. For example, the impaired fever response to bacterial endotoxin reported in fasted guinea pigs was associated with blunted elevation of metabolic heat production (51). In a different study using fasted newborn rabbits, the depression of fever was reversed by allowing the animals to freely select warmer ambient temperatures in a thermal gradient (22), arguing that the selective attenuation of heat production, rather than the general downturn of thermoregulatory function, is the more likely mechanism responsible for the depression of fever during acute starvation. This selective ablation of metabolic thermogenesis coexisting with the preserved heat-seeking behavior resembles the thermoregulatory responses to starvation under noninflammatory conditions where food deprivation in laboratory animals rapidly induces depression of basal metabolic heat production and hypothermia, while increasing preference to higher ambient temperature, as a part of an adaptive strategy to conserve energy stores (47, 58).

As well as the depression of metabolism, however, acute starvation is known to trigger a number of other physiological changes, including the suppression of immune function, which in many cases adversely affects the host's defense mechanisms and subsequently decreases its ability to effectively combat infectious disease (32, 35, 54). Given that the initiation of fever is essentially dependent on the innate immune response to exogenous pathogens, accompanied by the production and release of a full complement of inflammatory immune mediators belonging to the cytokine family (25), the suppression of immunity is likely to contribute to the diminution of fever observed in starved animals (22, 37, 50, 51, 53). The link between circulating cytokines and fever is well established with a number of these mediators, most prominently IL-1β, IL-6, and TNF, implicated in triggering the central mechanisms regulating the fever response (6, 25, 33). The contribution of these cytokines to the activation of the central mechanisms involved in regulating the fever response has been regularly assessed by measuring the degree of activation of transcription factors such as NF-κB, which is activated by both TNF and IL-1β, and STAT3, which is activated by IL-6 (27, 28). Both of these pathways lead to the transcription and induction of cyclooxygenase (COX)-2 (8, 39, 45, 56), the rate-limiting enzyme for PGE₂ synthesis, and ultimately fever production (19, 31, 36, 48). Despite the important functional link between these immune mediators and fever, the role of cytokines in the attenuated fever response to exogenous pathogens during acute starvation has not previously been investigated in depth. For example, none of the aforementioned studies reporting the negative effects of acute starvation on the fever response (22, 37, 50, 51, 53) evaluated the changes in cytokine production. Nor have any studies investigated whether the central inflammatory mechanisms regulating the fever response (e.g., PGE₂ production) were affected in these situations as a consequence of a dampened immune response.

The aim of the present study, therefore, was to address, in detail, if and how the peripheral and central components of the inflammatory response to a single injection of LPS is affected in rats food deprived for 48 h. To achieve this, we measured the following variables in fed and fasted animals: 1) circulating levels of cytokines, 2) the activation of signaling pathways implicated in the cytokine action in the brain during fever (NF-κB and STAT3), 3) up-regulation of COX-2 and microsomal PGE synthase (mPGES)-1, two essential enzymes for brain PGE₂ production during fever (9, 10, 31, 57), and 4) levels of PGE₂ in the cerebrospinal fluid (CSF). In addition, the thermoregulatory response downstream of central PGE₂ action was tested by the direct injection of PGE₂ into the brain.

MATERIALS AND METHODS

Animals

Adult female Sprague-Dawley rats with initial body weight 200–225 g (Charles River, Saint Constant, Quebec, Canada) were used in all experiments. They were housed individually in a controlled environment at an ambient temperature of 21 ± 1.5°C on a 12:12-h light-dark cycle (light on from 0800 to 2000), with free access to water and standard laboratory chow (Rat chow #5012; Purina, St. Louis, MO) unless otherwise indicated. All experimental procedures were approved by the Animal Care Committee of McGill University pursuant to the Canadian Council of Animal Care guidelines.

Experimental Design

Experiments were performed over 4 days; 1 day for baseline measurement (day 0) followed by 3 days for feeding manipulation (days 1–3). Animals were weighed daily between 0930 and 1030. On day 0, rats were divided into two weight-matched feeding groups (fed and fasted). On day 1, food was removed from the fasted group immediately after the body weight measurement. Both groups of rats had free access to water throughout the experimental period. On day 3, each feeding group was subdivided into three treatment groups receiving a single injection between 0930 and 1030 (48 h after the start of fasting). For intraperitoneal injections, each subgroup (n = 7–8 per group) received either saline (1 ml/kg) or 100 μg/kg LPS (Escherichia coli O111:B4, lot 42k4120; Sigma, Oakville, Ontario, Canada). For intracerebroventricular injections, each subgroup (n = 7 per group) received vehicle (0.5% ethanol in saline), 10 ng/rat PGE₂ (Sigma), or 50 ng/rat PGE₂ injection at 5 μl volume.

Measurement of Body Temperature Using Remote Biotelemetry

Precalibrated temperature-sensitive radio transmitters (model:TA10TA-F40, Data Sciences, St. Paul, MN) were implanted via midline incision into the abdominal cavity of anesthetized rats (intramuscular; 50 mg/kg ketamine hydrochloride, 5 mg/kg xylazine hydrochloride, 0.5 mg/kg acepromazine maleate; total volume 1 ml/kg). The level of anesthesia was assessed by the withdrawal reflex to a toe pinch. 2% lidocaine (25–50 μl) was applied to the area of incision. Animals were allowed to recover for at least 7 days before experimentation. Transmitter output frequency (Hz) was monitored, at 10-min intervals for LPS-induced fever, and 2-min intervals for PGE₂-induced fever, by an antenna mounted in a receiver board situated beneath the cage of each animal. The output data from each transmitter were transformed into degrees centigrade using Dataquest A.R.T. software (Data Sciences).

Intracerebroventricular Cannula Implantation

Following the transmitter implantation described above, the animals designated for intracerebroventricular injection were implanted with 21-gauge guide cannulas (Plastics One, Roanoke, VA), as previously described (17). Briefly, rats were secured in a stereotaxic frame, the skull was exposed by midline incision, and 2% lidocaine (25–50 μl) was applied to the area of incision. The coordinates using bregma as a reference point were the following: 0.8 mm posterior, 1.5 mm lateral, and 3.0 mm ventral (from scull surface), resulting in the tip of guide cannula being situated 1 mm above the right lateral ventricle. The guide cannula was then fixed to the skull with dental cement (Lang Dental Mfg., Wheeling, IL), holding two supporting screws with a diameter of 1.2 mm secured on the skull, and was sealed with a “dummy” cannula designed to reach the tip of the guide. Rats were allowed to recover for at least 7 days postsurgery. Injections were performed in conscious free-moving animals, with a 26-gauge injection cannula that extended 2 mm beyond the tip of the guide cannula and connected to a 10-μl Hamilton syringe via polyethylene (PE)-20 tubing. Accuracy of the cannula placement was verified by an injection of black ink (5 μl). The dye was visible in the fourth ventricle in all animals examined.

Repeated Blood Collection via a Jugular Vein Catheter

Separate sets of rats were prepared for repeated blood collection. Animals received the anesthesia mixture described above, and the depth of anesthesia was assessed by the withdrawal reflex to a toe pinch. A single dose of ampicillin (subcutaneous; 150 mg/kg; Nanopharm, Ontario, Canada) was administered 10 min before the surgery, and 2% lidocaine (25–50 μl) applied to the area of incision. The external jugular vein was exposed through a small skin incision, and the linguofacial vein branching from the jugular vein was isolated and ligated 5 mm rostral from the bifurcation. A sterile Silastic catheter (0.97 mm OD; Dow Corning, Midland, MI) filled with heparinized (50 IU/ml) sterile saline was inserted 4 cm toward the jugular vein until the tip of the catheter reached the level of the right atrium; at which point it was secured. The other end of the catheter was tunneled under the skin, exteriorized at the nape and sealed with a stainless steel cap. Animals were allowed to recover from surgery for 2 days before experimentation. Blood was collected from conscious free-moving animals, by connecting the catheter to a 1-ml syringe containing 5 IU of heparin via PE-50 tubing with 22-gauged blunted needle tip. The sample volume (150 μl/collection) collected was replaced with sterile saline prewarmed to 37°C.

To compare the effect of fasting on cytokine production during LPS-induced inflammation, the fed and fasted rats were treated with a single intraperitoneal injection of either saline or LPS 100 μg/kg (n = 4 per group). Food was removed from the fasted groups 48 h before the injection, and blood was collected at −48, −24, 1, 2, 4, 8, and 24 h with respect to the injection time at 0 h. The samples were immediately placed on ice and centrifuged at 3,000 g for 5 min, and plasma was collected, aliquoted, and stored at −80°C until assays were performed.

Cerebrospinal Fluid and Brain Collection

To assess the levels of PGs in the CSF and to measure the expression levels of inflammatory genes in the hypothalamus, CSF and brains were collected from the fed and fasted rats treated with either saline (1 ml/kg ip) or LPS (100 μg/kg ip) (n = 5 or 6 per group). Animals were killed 2 h after the injection under deep anesthesia with a terminal dose of pentobarbital sodium (60 mg/kg ip). CSF was sampled from the cisterna magna with a 27-gauge needle connected to a microsyringe (250 μl) via PE-20 tubing. The sampled CSF was immediately frozen on dry ice and kept at −80°C. Blood was collected from the same animals via cardiac puncture using sterile heparinized syringes and plasma samples prepared for additional cytokine assay. Animals were then perfused with ice-cold saline, the brains were removed, and the hypothalami were dissected, frozen on dry ice and kept at −80°C until use.

Cytokine and Prostaglandin Measurement

Sandwich ELISA for TNF, IL-1β, IL-6, IL-1 receptor antagonist (RA), and leptin (NIBSC, Potters Bar, UK) were performed as previously described (40), except that plasma samples and biotinylated detection antibodies were diluted in a buffer containing 0.5 M NaCl, 10 mM phosphate, 0.1% Tween 20, pH 7.4. All plasma samples were diluted 1:10, except for IL-1β (1:5 dilution). Intra-assay and inter-assay variations were below 10%. The sensitivities of the assays were 3.9 pg/ml for TNF and IL-1RA, 7.8 pg/ml for IL-1β, IL-6, and leptin. All samples were assayed in duplicate.

PGE₂ and 15-deoxy-Δ^12,14-PGJ₂ (15-d-PGJ₂) levels in CSF were measured using an enzyme immuno assay (EIA) kit [PGE₂ EIA kit-Monoclonal, Cayman Chemical (Ann Arbor, MI); 15-deoxy-Δ^12,14-PGJ₂ EIA kit, Assay Designs (Ann Arbor, MI)], according to the manufacturer's protocol. The CSF samples were diluted 1:5 in an assay buffer provided as part of the kit. The sensitivity of the assay for PGE₂ and 15-d-PGJ₂ was 7.8 pg/ml and 36.8 pg/ml, respectively. All samples were assayed in duplicate.

RNA Extraction and RT-PCR

Total RNA was extracted from the hypothalamus in 1 ml of TRIzol (Invitrogen, Burlington, Ontario, Canada), according to the manufacturer's protocol. The first-strand cDNA was synthesized from 1 μg of total RNA using 200 units of Molony murine leukemia virus reverse transcriptase (Invitrogen), 5 μM of random hexamers (Applied Bioscience, Streetsville, Ontario, Canada), and 1 mM of dNTP mix (Sigma) in a total reaction volume of 20 μl. The cDNA product (0.9 μl) was added to 15 μl PCR reaction mix containing ReadyMix Taq PCR (Sigma) and 6 pmol of gene-specific primer sets for inhibitory factor (I)κBα, suppressor of cytokine signaling (SOCS)3, IL-1β, COX-2, mPGES-1 and β-actin (Alpha DNA, Montreal, Quebec, Canada) using a Gene Amp PCR system 9700 Thermocycler (Applied Biosystems, Foster City, CA). The following parameters were used: 1) denaturing; 95°C for 5 min, 2) amplification cycle; 95°C for 30 s, annealing temperature for 30 s and 72°C for 1 min, 3) final extension; 72°C for 10 min. Primers were designed to span a sequence derived from different exons [separated by intron(s) in the genomic DNA sequence] to minimize amplification from non-mRNA-derived templates. Inappropriate amplification from genomic DNA was negligible when amplification was performed with a template without reverse transcription. The gene accession numbers, primer sequences, annealing temperatures, and cycle numbers used are listed as following: IκBα (NM_01105720: forward, 5′-AACAACCTGCAGCAGACTCC-3′, reverse, 5′-GTGTGGCCGTTGTAGTTGG-3′; 60°C; 28 cycles), SOCS3 (NM_053565; forward, 5′-CCAGCGCCACTTCTTCAC-3′, reverse: 5′-GTGGAGCATCATACTGGTCC-3′; 60°C; 36 cycles), IL-1β (NM_031512; forward, 5′-CCCAAGCACCTTCTTTTCCTTCATCTT-3′, reverse, 5′-CAGGGTGGGTGTGCCGTCTTTC-3′; 60°C; 36 cycles), COX-2 (NM_017232; forward, 5′-TGATAGGAGAGACGATCAAGA-3′, reverse, 5′-ATGGTAGAGGGCTTTCAACT-3′; 57 °C; 32 cycles), mPGES-1 (NM_022415; forward, 5′-TTTCTGCTCTGCAGCACACT-3′, reverse, 5′-CATGGAGAAACAGGTGAACT-3′; 57°C, 36 cycles), β-actin (NM_031144; forward, 5′-GCCGTCTTCCCCTCCATCGTG-3′, reverse, 5′-TACGACCAGAGGCATACAGGGACAAC-3′; 60°C; 20 cycles). PCR products were separated by gel electrophoresis (1.5% agarose), and band densities were obtained using GeneTool image analysis software (Syngene, Frederick, MD). To normalize the expression level of genes between different samples, the levels were estimated as the ratio of geneX/β-actin. In a pilot experiment, the amount of PCR product (on a log scale) vs. the number of cycles was plotted, and the linear range of template amplification was determined for two samples from each treatment group. The cycle numbers were determined within the exponential phase of amplification for all treatment groups.

Data Analysis

Measurements of body temperature.

To evaluate the changes of core body temperature (T_core) over the course of fasting, the average T_core of the dark (between 2400–0600) and light (between 1200–1800) periods were calculated for each animal. This 6-h interval was chosen to exclude the temperature change caused by body weight measurement performed daily at 1000 in the light period (T_core returned to normal level by 1200). A corresponding interval was chosen for the dark period (starting from 2400 for 6 h).

Two measures have been used to assess the fever response: absolute T_core and the rise in T_core (ΔT_core: ΔT_core = T_core − preinjection T_core). However, these two measures are not interchangeable when animals have different preinjection T_core (e.g., day/night variation and different feeding conditions as in the present study). Satinoff and colleagues (11) demonstrated that absolute T_core, rather than ΔT_core, provides a better representation of the physiological characteristic of fever because the former is solely determined by the dose of pyrogen (PGE₂) and is largely unaffected by day/night variation in preinjection T_core. The validity of this definition was supported by several other studies investigating the fever response in animals with different preinjection T_core (18, 21, 34). Therefore, in the present study, we used absolute T_core rather than ΔT_core to assess the fever response.

To evaluate the overall change in T_core by pyrogenic stimuli in the fed and fasted animals, preinjection T_core and postinjection T_core were calculated. The preinjection T_core was defined as the average T_core during the period of −30 to 0 min prior to the injection. The postinjection T_core was defined as the average T_core between 0 and 60 min in the case of intracerebroventricular. PGE₂ injection, and in the case of intraperitoneal LPS injection, the average between 120 and 480 min to exclude stress-induced hyperthermia.

Statistics.

All data are presented as means ± SE and were analyzed using StatView software 4.57 (Abacus Concepts, Berkeley, CA). Two-way repeated-measures ANOVA (time × treatment) was used to analyze the data from body weight change, temperature course, and cytokine time course studies. Two-way ANOVA (feeding × inflammation) was used to analyze the data from mRNA expression and PG measurements at a single time point. Newman-Keuls multiple-comparison test was performed as a post hoc analysis when applicable. In all cases, P values less than 0.05 were deemed statistically significant.

RESULTS

Changes in Body Weight During Fasting and Inflammation

The body weight changes of the animals in all treatment groups are shown in Fig. 1A. The rats in the fasted groups were deprived of food immediately after the body weight measurement on day 1 between 0930 and 1030 until the end of the experiment, while the fed groups were fed ad libitum. All groups were injected intraperitoneally with either saline or LPS 100 μg/kg on day 3, at the time point corresponding to 48 h after the start of food deprivation in the fasted groups. The initial body weight on day 1 was similar across all groups (243.8 ± 7.4 g, 254.1 ± 6.2 g, 247.2 ± 5.9 g, and 253.0 ± 7.5 g for fed-saline, fed-LPS, fasted-saline, and fasted-LPS, respectively) and decreased significantly following food deprivation (P < 0.001; fed vs. fasted for days 2, 3, and 4). LPS injection resulted in a significant loss of body weight 24 h after treatment in the fed groups (P < 0.01; saline vs. LPS). In contrast, the LPS injection in the fasted groups caused no further loss of body weight compared with the saline-treated counterpart.

Fig. 1.

Changes in body weight and body temperature. A: body weight change. Animals were weighted between 0930 and 1030. Food was removed from the fasted groups after the body weight measurement on day 1. All groups of rats received an intraperitoneal injection of the respective drug on day 3 immediately after the body weight measurement. Values are expressed as means ± SE. Fed-saline and fasted-saline (n = 8), fed-LPS and fasted-LPS (n = 7). *Significantly different from saline-treated controls, P < 0.05; #Significantly different from fed counterpart, P < 0.05. B: daily cycles of core temperature (T_core). The time of injection is indicated by a thick arrow, and small thin-angled arrows indicate body weight measurement. Note that the handling of animals caused a transient hyperthermia in both the fed and fasted groups. The gray background represents the dark period (2000–0800). C: average T_core in the active and inactive phases. *Significantly different from the inactive phase on day 0, P < 0.05. #Significantly different from the active phase on day 0, P < 0.05.

Effects of Fasting on Basal Body Temperature

Figure 1, B and C illustrate the daily cycle of T_core of the fed and fasted rats that were treated with saline on day 3 as an example of T_core cycle under noninflammatory conditions. The T_core was higher in the active (dark) and lower in the inactive (light) phase displaying clear day/night variations in the fed condition. This daily rhythm was still sustained during fasting. However, as reported previously (49, 58), the fasted groups showed a progressive decline in the T_core (P < 0.01, time × feeding). To evaluate the change of T_core over the course of fasting, the average T_core in the active and inactive phases are shown in Fig. 1C. In the active phase, the average T_core of fasted rats tended to be lower (P = 0.07) on day 2 and became significantly lower (P < 0.001) on day 3 compared with its control value on day 0 (Fig. 1C). The decline of the average T_core in the inactive phase was more prominent than that in the active phase, reaching significance by day 2 (P < 0.001, day 0 vs. day 2) and was further decreased on day 3 (P < 0.05, day 2 vs. day 3).

Effect of Fasting on Lipopolysaccharide-Induced Fever

The fever response to a single intraperitoneal injection of LPS was examined on day 3. The average T_core of the fasted groups prior to the injection was significantly lower than those of the fed counterparts (Fig. 2). The saline injection caused a transient, stress-associated hyperthermia in both groups reaching a similar peak (Fig. 2A). Thereafter, the T_core of the fasted rats returned to a lower level than that of the fed group (170–600 min, P < 0.05, fed-saline vs. fasted-saline). LPS injection at the dose of 100 μg/kg caused a clear thermal response in the fed group (42), which was characterized by an initial mild hypothermia occurring immediately after the injection-associated hyperthermia, followed by a biphasic temperature increase (Fig. 2C). The fasted rats also showed a clear polyphasic fever response to LPS; however, the T_core was consistently lower compared with the fed counterparts (100–200 min and 250–390 min, P < 0.05, fed-LPS vs. fasted-LPS). In addition, the first fever peak was significantly attenuated (P < 0.05, 38.5 ± 0.2°C vs. 37.9 ± 0.2°C, fed-LPS vs. fasted-LPS) and delayed in the fasted rats (P < 0.001, 171.4 ± 5.1 min vs. 215.7 ± 8.4 min, fed-LPS vs. fasted-LPS). The second peak was also significantly attenuated in the fasted rats (P < 0.05, 39.1 ± 0.2°C vs. 38.4 ± 0.1°C, fed-LPS 100 vs. fasted-LPS), but this reached maximum at the same time in both groups (360.0 ± 11.1 min vs. 375.7 ± 18.1 min, fed-LPS vs. fasted-LPS).

Fig. 2.

T_core change after LPS challenge. T_core course of rats treated with an intraperitoneal injection of saline (A) or LPS 100 μg/kg (C) at time point 0 h. An additional group of fed rats received a fixed amount of LPS (23 μg/rat) equivalent to the amount given to the fasted group (100 μg/kg) as a control. Values are expressed as means ± SE. Fed and fasted saline (n = 8), fed and fasted LPS (n = 7), and fed (23 μg/rat) (n = 5). *Significantly different from fed counterpart, P < 0.05. Pre- and post-injection T_core of saline- (B) and LPS- (D) treated rats. Two-way (feeding × injection) ANOVA; interaction P > 0.05, feeding P < 0.05, injection P > 0.05 (B); interaction P > 0.05, feeding P < 0.001, injection P < 0.001 (D).

The fact that the fasted rats had a smaller fever response may be a factor of the smaller amount of LPS given, which was corrected for body weight (preinjection body weight, 260.6 ± 6.8 g vs. 226.6 ± 7.0 g, fed-LPS vs. fasted-LPS). To eliminate this possibility, an extra set of fed rats (preinjection body weight 264.5 ± 3.7 g) were injected with a fixed amount of LPS (23 μg/rat), which is equivalent to the absolute amount given to the fasted group at the dose of 100 μg/kg. The fever response induced by this amount of LPS was almost identical to the one seen in the fed rats treated with the fixed LPS dose, confirming that the attenuation of fever response in the fasted rats was not due to the smaller amount of LPS injected (Fig. 2C).

Figure 2, B and D, illustrate preinjection T_core and the average postinjection T_core in different treatment groups. It is clear from these graphs that, in the LPS-treated groups, both preinjection and postinjection T_core were similarly attenuated by fasting (two-way ANOVA, P < 0.001 for feeding and injection, P > 0.05 for feeding × injection interaction). Saline injection had no effect on the T_core. The same experiments were also conducted in male rats for comparison with a similar outcome, thus negating the effect of gender on the fever response (data not shown).

Circulating Cytokines

We next examined whether the attenuation of fever response to LPS was attributable to the alteration in immune-to-brain signaling involved in the development of fever (6, 25, 33). To this end, we characterized the plasma levels of proinflammatory and anti-inflammatory cytokines, mRNA levels of the genes involved in PGE₂ synthesis in the hypothalamus and PGE₂ levels in the CSF after an intraperitoneal LPS (100 μg/kg) injection. Figure 3 illustrates the time course of plasma TNF, IL-6, IL-1RA, and leptin, as well as the IL-1β levels at 2 h after injection in the four groups of rats. LPS treatment dramatically increased the plasma levels of TNF, IL-6, and IL-1RA in the fed rats. The same cytokines were also increased by LPS in the fasted group; however, the TNF and IL-1RA elevations were significantly attenuated. The LPS-induced TNF was significantly lower at 2 h (P < 0.05), and the IL-1RA levels were lower at 2, 4, and 8 h (P < 0.01) in the fasted than fed rats. On the other hand, the increase in IL-6 was almost identical between the fed and fasted rats. The fasted rats treated with saline, however, had significantly higher IL-6 levels at 2 h after the injection compared with the fed counterpart (P < 0.01). The present study did not determine the time course of IL-1β levels because of the limitation of the amount of sample available from the repetitive blood collection. Alternatively, its plasma levels were measured in rats killed at 2 h after LPS or saline treatment (from the same animals used for CSF PGs measurement presented below). Plasma IL-1β was found to be similar between the fed and fasted rats (1.4 × 10⁻¹ ± 0.2 × 10⁻¹ ng/ml vs. 2.0 × 10⁻¹ ± 0.6 × 10⁻¹ ng/ml, fed-LPS vs. fast-LPS), while in the in the same samples, IL-1RA levels were confirmed to be lower in the fasted than in fed rats (P < 0.01, 4.5 ± 0.1 ng/ml vs. 3.3 ± 0.1 ng/ml, fed-LPS vs. fast-LPS). The relative ratio between these two interlinked cytokines was calculated as a predictor of net IL-1 signaling (IL-1β/IL-1RA ratio) and was found to be significantly higher in the fasted than fed rats (P < 0.01, 3.1 × 10⁻² ± 0.4 × 10⁻² vs. 6.1 × 10⁻² ± 1.2 × 10⁻², fed-LPS vs. fast-LPS). Leptin was also measured in these studies. As reported previously (2), fasting dramatically reduced the plasma levels of this appetite-suppressing hormone. LPS treatment, on the other hand, did not have a significant effect on the leptin levels in either feeding condition. However, the time × treatment interaction was significant (P < 0.001) and post hoc tests revealed that the LPS-treated fed animals tended to have higher leptin levels compared with saline-treated controls at 8 h after injection, although the analysis did not reach statistical significance (P = 0.05). More specifically, the leptin levels in the saline-treated controls showed a profile of circadian change (7); its levels were higher in the earlier stage of the inactive phase (at 1, 2, and 24 h) and lower in the later stage (4 and 8 h). After LPS injection, the decline of leptin levels toward the end of the inactive phase was abolished and indeed reversed. Accordingly, the fed-LPS group had a tendency to have higher leptin levels than the fed-saline counterparts. In the fasted groups, however, the circadian change of leptin levels was no longer apparent, and LPS did not induce any further change. Interestingly, there was a relatively large interanimal variability of leptin levels within the same treatment groups, but an animal with higher leptin level at one time point consistently showed higher levels at other time points compared with those with lower levels. Therefore, to reduce the interanimal variability, the leptin levels of each animal were normalized to their baseline levels measured 48 h before the injection on day 1 (just prior to the start of food restriction). The leptin levels at the baseline were similar across all the treatment groups (1.6 ± 0.4 ng/ml, 1.3 ± 0.2 ng/ml, 1.2 ± 0.2 ng/ml, and 1.0 ± 0.2 ng/ml for fed-saline, fed-LPS, fasted-saline, and fasted-LPS, respectively). This analysis revealed that LPS treatment in the fed group significantly increased leptin levels at 4 and 8 h after the injection compared with saline treatment in the counterpart controls (P < 0.01).

Fig. 3.

Time course of circulating cytokine levels after LPS challenge. The fed and fasted rats were treated with an intraperitoneal injection of saline or LPS (100 μg/kg) at time point 0 h. Blood was sampled repeatedly starting at prefasting (48 h before injection) and from then at the time points indicated. Leptin levels are expressed as % change from prefasting baseline levels of each animal in an additional graph. Values are expressed as means ± SE. for (n = 4 per group). Plasma IL-1β levels were determined in a separate set of animals killed 2 h after injection (n = 5 or 6 per group) *Significantly different from fed counterparts, *P < 0.05, **P < 0.01. ##Significantly different from saline-treated counterparts, P < 0.01.

Inflammatory Gene Expression in the Brain

Figure 4 shows mRNA expression levels of IκB, SOCS3, IL-1β, COX-2, and mPGES-1 in the hypothalamus studied 2 h after LPS (100 μg/kg ip) injection. Despite the altered cytokine levels in the circulation, the feeding condition of the animals did not have any major effect on the mRNA expressions during LPS inflammation. Two-way ANOVA (feeding × LPS) revealed that LPS significantly (P < 0.001) increased mRNA levels of all the genes studied. The overall effect of feeding was significant only for the IκB mRNA levels, which were increased in the fasted compared with the fed animals. The levels of all the other genes tested were unaffected by the feeding condition with no significant interaction apparent.

Fig. 4.

Gene expression levels in the brain after LPS challenge. Semiquantitative RT-PCR analysis of suppressor of cytokine signaling 3 (SOCS3), inhibitory factor (I)-κB, IL-1β, cyclooxygenase-2 (COX-2), and microsomal prostaglandin E synthase-1 (mPGES-1) mRNA levels in the hypothalamus 2 h after injection. Values are expressed as means ± SE for 5 or 6 animals. Two-way ANOVA (feeding × LPS) revealed significant main effect of LPS (Ps < 0.001) for all genes studied. The effect of fasting was significant only for IκB. There was no significant interaction (feeding × LPS) on any gene tested. AU, arbitrary unit. *Significant effect of LPS. #Significant effect of feeding.

Prostaglandin Levels in the Cerebrospinal Fluid

Consistent with the expression levels of COX-2 and mPGES-1, LPS treatment significantly increased PGE₂ levels (P < 0.001) in the CSF, which was similar between the fed and fasted groups 2 h after injection (Fig. 5). It has recently been shown that 15-d-PGJ₂, another PG family member derived from the COX metabolite PGH₂, exerts antipyretic actions during LPS fever (38). Therefore, we also measured its CSF levels in the same groups of animals. Neither LPS nor feeding condition had any effect on the 15-d-PGJ₂ concentrations at the time point studied.

Fig. 5.

PGE₂ and 15-deoxy-Δ^12,14-PGJ₂ (15-d-PGJ₂) levels in the brain after LPS challenge. PGE₂ and 15-d-PGJ₂ levels in cerebrospinal (CSF) 2 h after injection were measured by enzyme immunoassay. Values are expressed as means ± SE. for 5 or 6 animals. Two-way ANOVA (feeding × LPS) revealed significant main effect of LPS (P < 0.001) on PGE₂ levels, while feeding had no effect. Interaction (feeding × LPS) was not significant. None of the parameter (feeding, LPS, interaction) was significant in 15-d-PGJ₂. *Significant effect of LPS.

Fever Response to Central Prostaglandin E₂ Injection

The fact that fasting did not affect the LPS-induced synthesis of PGE₂ in the brain, while it attenuated the fever response to LPS indicated that the attenuation is due to the altered thermal response either downstream or independent of PGE₂ action in the fasted rats. To test this hypothesis, both the fed and fasted rats were injected with two doses of PGE₂ (10 and 50 ng/rat) into the right lateral ventricle. Vehicle containing 0.5% ethanol in sterile saline was injected as a control. As expected, the average preinjection T_core were significantly lower in the fasted groups compared with the respective fed counterparts (37.0 ± 0.1°C, 37.0 ± 0.1°C, 37.0 ± 0.0°C, 36.3 ± 0.1°C, 36.1 ± 0.2°C, and 36.3 ± 0.2°C for fed-vehicle, fed- PGE₂10, fed- PGE₂50, fasted-vehicle, fast- PGE₂10, and fasted- PGE₂50). In the vehicle-injected groups, the T_core of the fasted group was consistently lower than that of the fed group (Fig. 6A; 0–60 min, P < 0.05). Both doses of PGE₂ caused a fever response in both the fed and fasted groups. The major difference between the fed and fasted rats was observed at the lower dose (10 ng/rat) of PGE₂ treatment (Fig. 6C). The T_core course was significantly attenuated in the fasted groups (0–22 min, P < 0.05). The peak T_core was lower in the fasted than the fed group (P < 0.01, 38.1 ± 0.1°C vs. 37.6 ± 0.2°C, fed-PGE₂10 vs. fasted-PGE₂10), while the time to reach this peak was similar between the groups (17.4 ± 1.4 min vs. 24.3 ± 4.4 min, fed-PGE₂10 vs. fasted-PGE₂10). The attenuation of fever was less prominent after the higher dose (50 ng/rat) of PGE₂ treatment (Fig. 6E). By 10 min after the injection, the T_core of fasted rats reached levels similar to those of the fed group. The peak T_core value and timings were similar between the groups (T_core, 38.6 ± 0.1°C vs. 38.4 ± 0.2°C; timing, 21.4 ± 2.0 min vs. 21.4 ± 1.1 min; fed-PGE₂50 vs. fasted-PGE₂50).

Fig. 6.

T_core change after central PGE₂ challenge. T_core course of rats treated with an intracerebroventricular injection of vehicle (A), PGE₂ 10 ng/rat (C), or PGE₂ 50 ng/rat (E) at time point 0 min. Values are expressed as means ± SE. for 7 animals. *Significantly different from fed counterpart, P < 0.05. Pre- and postinjection T_core of vehicle- (B), PGE₂ 10 ng/rat- (D) or PGE₂ 50 ng/rat- (F) treated rats. Two-way (feeding × injection) ANOVA; interaction P > 0.05, feeding P < 0.001, injection P < 0.05 (B); interaction P > 0.05, feeding P < 0.05, injection P < 0.001 (D); and interaction P < 0.05, feeding P < 0.05, injection P < 0.001 (F).

Figure 6, B, D and F illustrate the preinjection T_core and the average postinjection T_core in three different treatment groups. At 10 ng of PGE₂ injection, both the pre- and postinjection T_core were similarly attenuated by fasting (two-way ANOVA, P < 0.001 for feeding and injection, P > 0.05 for feeding × injection interaction). At the 50 ng dose, two-way ANOVA yielded a significant effect of both feeding (P < 0.05) and injection (P < 0.001), and a significant interaction (P < 0.05). Post hoc analysis revealed no significant difference between the fed and fasted groups on either preinjection T_core or postinjection T_core.

DISCUSSION

Previous studies have reported that acute starvation prior to inflammatory events attenuates the fever response in various species of experimental animals (22, 37, 50, 51, 53). The present study tested whether the attenuation of fever is due to diminished inflammatory responses to pathogen, resulting in reduced immune-to-brain pyrogenic signaling. Our results demonstrated that 48 h preinflammation fasting clearly lowered the T_core during the fever response to a systemic injection of LPS. Characterization of the LPS-induced inflammatory response revealed that preinflammation fasting diminished the elevations of circulating levels of TNF and IL-1RA, while the rise of IL-1β and IL-6 levels were unchanged. Fasting affected neither the mRNA upregulation of genes important for PGE₂ synthesis in the brain, nor the elevation of PGE₂ levels in the CSF. These results indicate that, despite the changes in some cytokine levels, the pyrogenic signaling to/within the brain is well preserved during acute starvation, and that a mechanism(s) independent of this cascade is (are) responsible for the attenuation of the LPS-induced fever. With this in mind, the final experiment in the present study examined the fever response to central PGE₂ injection and demonstrated that the preinflammation fasting also attenuated the T_core during the PGE₂-induced fever. With these results, we conclude that the alteration in the thermoregulatory mechanisms downstream and/or independent of PGE₂ action are responsible for the attenuation of LPS-induced fever during acute starvation.

It is widely known that malnutrition impairs various aspects of the immune function and ultimately affects the host's susceptibility to infection (5, 12, 32, 35, 54). In line with this, the present study found that preinflammation fasting blunted the induction of plasma levels of TNF and IL-1RA by systemic LPS injection. On the other hand, LPS-induced elevation of IL-1β and IL-6, two important pyrogenic cytokines besides TNF, was largely unaffected by the fasting. Interestingly, the IL-1β/IL-1RA ratio, an indicator of net IL-1β action, was significantly higher in the fasted than the fed animals, because of the diminished IL-1RA levels in the fasted group. These results imply that preinflammation fasting alters the balance between pro- and anti-inflammatory cytokines and thereby could enhance certain aspects of the inflammatory response and resistance to infection (55). Interestingly and unexpectedly, saline injection alone increased plasma IL-6 levels in fasted rats but not ad libitum fed animals. The reason for this difference is not clear, however, it could partly be a reflection of a heightened stress response (to injection/handling) in these animals resulting from the food deprivation, thus rendering them more susceptible to stress- induced IL-6 release (29) compared with ad libitum fed rats.

We initially predicted that the inflammatory response in the brain after LPS treatment would be altered as a consequence of the blunted peripheral cytokine production in the fasted rats. However, the LPS-induced levels of IκB and SOCS3 mRNAs, two commonly used markers for assessing the magnitude of the inflammatory response to afferent cytokine signals in the brain (27, 28), were unimpaired in fasted animals. Similarly, mRNA up-regulations of IL-1β, COX-2 and mPGES-1 in the brain as well as PGE₂ elevation in the CSF, critical components of fever development, were similar between the fed and fasted rats. Among the classical pyrogenic cytokines, namely TNF, IL-1β and IL-6, both TNF and IL-1β actions are linked to NF-κB activation (assessed by IκB mRNA upregulation in the present study) (27), while IL-6 is linked to STAT3 activation (assessed by SOCS3 mRNA up-regulation) (28). Therefore, the slight enhancement of the LPS-induced IκB levels in the fasted rats, despite the diminished TNF levels, can be explained by the enhanced net IL-1β action (IL-1β/IL-1RA ratio). This explanation also applies to the intact up-regulation of COX-2 and mPGES-1 (and consequently PGE₂ production) in the TNF-diminished fasted rats since NF-κB is known to be critical for the transcription of these genes (24, 39). An alternative, but not mutually exclusive, explanation is that circulating IL-6, the levels of which were unaffected by fasting, plays a major role in central PGE₂ production. Among the three major pyrogenic cytokines, IL-6 was demonstrated to consistently increase in the circulation after various types of pyrogenic challenges, including LPS, and the circulating levels of this cytokine were shown to correlate with the magnitude of fever more closely to those of TNF and IL-1β (1, 3, 26, 30, 43). In support of the significant role of this cytokine in the fever response, studies from our laboratory and those of others demonstrated that IL-6 deficiency or anti-IL-6 antiserum treatment strongly attenuates LPS-induced fever (3, 4, 45). In addition, we have recently reported that IL-6 directly induces the expression of COX-2 in the brain (45). The present study further supports the role of IL-6 in regulating central PGE₂ synthesis during fever.

Other than IL-6, we and others reported previously that neutralization of leptin, another circulating cytokine, with anti-leptin antiserum attenuated the fever response to LPS in ad libitum fed rats. The fever attenuation was accompanied by blunted up-regulations of brain IL-1β and COX-2, suggesting a role for leptin as a circulating pyrogen (13, 17, 46). However, in the present study, the reduction of leptin levels after fasting had no effect on the LPS-induced up-regulation of IL-1β or COX-2 mRNA in the brain, arguing against its proinflammatory roles. Moreover, in another study using obese Koletsky rats that lack leptin receptor, Steiner et al. (52) found that the deficiency of leptin signaling caused a prolonged hypothermia (indicative of a more severe systemic inflammation than fever) after LPS injection, indicating possible anti-inflammatory effects of leptin. In the same study, the hypothermia exhibited by the obese rats was accompanied by an exaggerated increase of plasma TNF, which is opposite to the diminished TNF rise in the fasted low-leptin rats in the current study. The basis of these discrepancies is currently unclear, however, both the fasted and obese animals have a number of hormonal and metabolic abnormalities due to the change in the nutritional status besides the direct effects of change in leptin signaling, and it is possible that these differences may have contributed to the different outcomes. Acute leptin neutralization with anti-leptin antiserum in ad libitum fed animals would help to delineate the role of leptin in mediating the CNS component of the inflammatory response since this model of transient leptin deficiency would circumvent starvation- or obesity-associated abnormalities.

The results from the present study did not show a causal link between the attenuation of fever and alteration in the pyrogenic inflammatory response during acute starvation. This is in contrast to the case of chronic protein-calorie malnutrition which is a major common cause of secondary immune deficiency (5). The attenuation of fever by chronic protein deficiency was ascribed to a significant attenuation in the cytokine production by monocytes and macrophages responding to exogenous pathogens in experimental animals (15, 16), as well as malnourished patients (20). Interestingly, the protein-deficient animals, as opposed to the fasted rats manifesting hypothermia in our study, were able to maintain normal body temperature under noninflammatory conditions, even when exposed to a cold environment down to 5°C (15), indicating that chronic protein deficiency has little effects on thermoregulatory function. The mechanisms of how nutritional deficiency affects the immune and thermoregulatory function are currently unclear and potential mechanisms are numerous (general shortage of energy, deficiency of specific macro- or micro-nutrients such as protein, zinc, iron, vitamin A, C, and E, and an adaptive hormonal response to starvation such as increased glucocorticoid levels). However, these different forms of malnutrition, both of which attenuate the fever response but in a different manner, will serve as useful models to investigate the precise mechanisms that specifically affect immune or thermoregulatory function.

Fasted rats showed significantly lower T_core at the time of injection compared with their fed counterparts (Figs. 2 and 6). These data support previous studies that acute starvation alone alters thermoregulatory mechanisms leading to hypothermia under non-inflammatory condition (49, 58). Interestingly, it was shown that the fasted rats remained hypothermic, even when they were allowed to select (and indeed moved to) warmer ambient temperatures in a thermal gradient (47). Moreover, the hypothermia was accompanied by a fall in the threshold T_core for cold-induced thermogenesis, and once the critical temperature was reached, the thermogenic abilities of these animals per se were unimpaired (47). Therefore, it was argued that the fasting-induced hypothermia is not simply due to a debilitation of thermogenesis but rather a regulated decline of body temperature. In the present study, both LPS and PGE₂ injections induced a clear rise in T_core in the fasted rats. In fact, the ΔT_core rise of the fasted rats was equivalent to, or even greater than (at 50 ng of PGE₂), that of their fed counterparts (Figs. 2D, 6, D and F). These results indicate that the febrigenic capacity of PGE₂-driven thermoeffectors remains largely intact in the fasted rats.

In analyzing the results from our fever studies we chose to use absolute T_core, rather than ΔT_core rise, because the former was deemed to represent a more accurate reflection of the upward shift of threshold temperature (the underlying thermoregulatory response of fever; see Data analysis). The elevated absolute T_core following pyrogen administration (LPS and 10 ng of PGE₂) was clearly and consistently lower in the fasted rats compared with the fed counterparts (Figs. 2 and 6). On the basis of these observations we described the fever response as “attenuated”. Nevertheless, the present data demonstrated that fasting did not per se attenuate the “febrigenic response” of animals, namely, LPS-induced central PGE₂ production as well as LPS- and PGE₂-induced thermogenesis (or heat gain) when judged by ΔT_core elevation (Figs. 2D, 6, D and F). This is obviously counterintuitive and somewhat difficult to explain. An additional, yet very important, point here is that the T_core is not regulated by a single unified system (previously known as the “set point” theory) but is controlled by multiple thermoeffector loops that defend different threshold temperatures and work relatively independently of each other (41). By taking this into account, our interpretation of the data is that the fever response (absolute T_core) is a combination of both PGE₂-dependent and -independent thermoreffector functions (each defending a unique threshold temperature), with the former remaining relatively intact, or even enhanced, during fasting while the latter process is largely responsible for the lowered T_core in the absence of, as well as, during fever. On the basis of our interpretation of these data, we conclude that fasting does not attenuate fever, in a physiological sense, but it does diminish the absolute T_core during fever via a themoregulatory function independent of the PGE₂-driven fever response.

Perspectives and Significance

The present study further highlights the complexity of the thermoregulatory system, especially when linked to the nutritional status of the organism. While it was clear that fasting altered the mechanisms regulating body temperature control, it did not affect the fever response. The fasted animals exhibited a remarkable resilience, given their diminished caloric status, to respond to exogenous pyrogens. This was to a large extent, reflected in all aspects of the “fever-pathways” tested both in the periphery and the brain. The main objective of this study was to examine the effect of fasting on immune-to-brain signaling, which remained more or less intact. We did, however, only study the contribution of the established “pyrogenic” cytokines, and it may be that other mediators, especially those linked with energy balance regulation, play an important part. Leptin could be a likely candidate, given that the decrease in the levels of this hormone is known to mediate a number of the neuroendocrine responses to starvation. In the 48-h-fasted rats used in the present study, we confirmed that the leptin levels in the circulation were dramatically reduced compared with ad libitum fed rats. The role of this cytokine-like hormone in fasting-induced hypothermia and alteration of the fever response to exogenous pathogens warrants further investigation.

GRANTS

This work was supported by the Canadian Institutes of Health Research and the National Sciences and Engineering Research Council of Canada. W. Inoue is a recipient of a Ph.D. scholarship from the Nakajima Foundation, Japan.