INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

CIRCULATING CYTOKINES INCREASE in many pathophysiological conditions, including infection, inflammation, tissue injury, stress, and heart failure (8, 14, 37, 56). TNF- and other proinflammatory cytokines elicit a broad spectrum of biological responses via their peripheral and central nervous system (CNS) effects. Fever, anorexia, sleep, stimulation of the hypothalamic-pituitary-adrenal (HPA) axis, and stimulation of sympathetic drive are among the CNS-mediated responses (7, 51). Because the blood-borne cytokines are too large to readily cross the blood-brain barrier (BBB), how circulating cytokines trigger the central neural circuits that mediate these functions is a widely studied issue (7, 17, 51). Three possible routes by which circulating cytokines might signal the brain have been proposed: 1) activation of sensory neurons at specific sites in hindbrain and forebrain that lack a BBB, called circumventricular organs (7); 2) activation of visceral sensory afferent nerves, particularly the abdominal vagus (20, 21, 26); and 3) induction of soluble mediators within the cells of the BBB that readily penetrate to initiate neural mechanisms.

Recent literature has provided evidence in support of the latter hypothesis and in particular for the role of PGE₂ as a likely mediator of the central effects of blood-borne cytokines (14, 42, 51, 52). This literature (30, 42) has focused primarily on cytokine activation of the HPA axis with increases in circulating adrenocorticotropic releasing hormone and corticosterone as the outcome indicators. The primary effector neurons for this glucocorticoid component of the cytokine response are the neurosecretory corticotropin-releasing factor (CRF) containing neurons in the paraventricular nucleus (PVN). Functional neuroanatomic studies (17, 18) have identified PGE₂-sensitive catecholaminergic neurons in the medulla oblongata as an obligatory component of the cytokine-induced glucocorticoid response.

A second important feature of the cytokine response is augmented sympathetic drive, manifested by increased levels of circulating epinephrine and norepinephrine and by increases in sympathetic drive to several vascular beds, particularly spleen, adrenal gland, and tail artery (2, 34, 44, 47, 55). However, the existing literature does not suggest a mechanism that links cytokine activation of the HPA axis and glucocorticoid release with increases in sympathetic drive. In fact, the studies exploring the mechanisms of cytokine-induced increases in sympathetic drive have focused on PGE₂ effects at the forebrain rather than the hindbrain level. Notably, those studies were performed in the general context of the immune response.

The present study arises from an entirely different perspective, namely, a consideration of the potential effects of cytokines on sympathetic regulation of cardiovascular and renal function. In preliminary studies (66), we observed that intracarotid administration of TNF-, targeting the forebrain region, consistently evoked increases in renal sympathetic nerve activity (RSNA), arterial pressure (AP), and heart rate (HR) in anesthetized rats. In the present study, also in anesthetized rats, we tested the hypothesis that these excitatory responses to blood-borne TNF- are mediated by prostaglandins acting centrally on parvocellular PVN neurons with descending influences on presympathetic neurons in rostral ventrolateral medulla (RVLM) and on RSNA. The issue is important in the context of chronic heart failure, in which circulating cytokines and sympathetic drive are high.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

These studies were performed in accordance with the "Guiding Principles for Research Involving Animals and Human Beings" of the American Physiological Society (1). The experimental procedures were approved by the University of Iowa Institutional Animal Care and Use Committee.

All experiments were performed on adult male Sprague-Dawley rats (300-350 g; Harlan Sprague Dawley). The animals were housed in the University Animal Care Facility and exposed to a normal 12:12-h light-dark cycle.

General preparation. Rats were anesthetized with urethane (1.5 g/kg ip), and supplemental doses of urethane (0.1-0.3 g/kg ip or iv) were given when increases in AP, HR, or respiratory rate were observed during surgery or recording. The level of anesthesia was periodically reassessed during the surgical procedures and experimental recording by examining pupillary size and nociceptive reflex responses and by continuously monitoring AP and HR. The left femoral artery was cannulated with PE-50 tubing filled with heparinized saline (20 U/ml) for the recording of AP, which was monitored with a Hewlett-Packard 7754A chart recorder (HP Medical Products Group, Andover, MA). The left femoral vein was cannulated with PE-20 tubing for the administration of drugs. A rostrally directed PE-20 cannula was inserted into the left common carotid artery, with the tip placed in the bifurcation for intracarotid artery (ICA) injection. Animals were intubated and most breathed spontaneously. Some animals underwent bilateral vagotomy in the cervical region or decerebration. These animals were mechanically ventilated (Harvard Small Animal Ventilator) with room air and paralyzed with pancuronium bromide (1.5 mg · kg¹ · h¹). Ventilation rate was set between 65 and 70 breaths/min. Core temperature was maintained at 37 ± 0.2°C with a rectal thermometer and a temperature controller (model K-100, Baxter Healthcare, Valencia, CA). The head was fixed in a stereotaxic frame (David Kopf Instrument, Tujunga, CA).

Intracerebroventricular injection. In some animals, a 29-gauge stainless steel cannula was implanted into the left lateral cerebral ventricle [stereotaxic coordinates: 0.91.0 mm posterior to bregma; 1.4-1.6 mm lateral to midline; 3.2-3.3 mm ventral to dura (23)]. The intracerebroventricular position of the cannula was confirmed by the staining of all four ventricles after injection of 5 µl Pontamine sky blue at the end of the experiments.

PVN microinjection. In some experiments, a 29-gauge guide cannula was inserted 0.5 mm above the PVN region. A thin 35-gauge (128-µm OD; 51.2-µm ID) stainless steel injection cannula was attached to PE-10 tubing, which was then connected to a 0.5-µl Hamilton microsyringe. The tip of the injection cannula was inserted into the guide cannula and then adjusted to a length extending 0.5 mm beyond the tip of the guide cannula. PGE₂ (50 ng in 100 nl) was microinjected over 30 s into the left PVN. Control experiments were performed in which the same dose of PGE₂ was injected lateral or caudal to the PVN. At the end of the experiments, 100 nl of 2% Pontamine sky blue were injected into the same location for histological verification.

Mid-collicular decerebration. A small left parietal craniotomy was made 1-2 mm rostral to the interaural line to avoid the sigittal/transverse venous sinus. A thin homemade blunt spatula (0.3 × 0.8 × 40 mm) was inserted horizontally from left to right side and then swept back across the depth of the brain stem, repeating if necessary. The basilar artery was preserved intact. The decerebration was verified by visual inspection of the brain after fixation. Only animals with complete decerebration were included in the study. A 1-h stabilization period was allowed before resuming the experiment.

Electrophysiological recording. A small craniotomy was made above the region of interest, and a glass micropipette was placed in the left PVN or left RVLM to record extracellular single-unit activity, using previously described techniques (65). Stereotaxic coordinates for PVN were 1.6-2.1 mm posterior to bregma, 0.3-0.5 mm from midline, and 7.0-8.0 mm ventral to dura, and for RVLM were 11.8-12.8 mm posterior to bregma, 1.8-2.1 mm from midline, and 8.0-10.0 mm ventral to dura (49). Recordings of RSNA were obtained from the left renal nerve using methods previously described (59).

Protocols. The recording session began at least an hour after completion of the surgical preparation. PVN and RVLM neurons were tested initially for responses to blood pressure changes induced by an intravenous bolus of phenylephrine and nitroprusside, as previously described (65). RVLM neurons were also tested for correlation with the AP pulse.

Data acquisition and analysis. The AP signal, the rectified and integrated voltage from the renal nerve recording, the transistor-transistor logic (TTL) pulses indicating multifiber action potentials in the raw RSNA exceeding a selected voltage, and the TTL pulses indicating PVN and RVLM single unit activity were fed into an on-line data-acquisition system consisting of a Cambridge Electronics Design (CED, Cambridge, UK) 1401 Plus computer interface coupled with a Gateway Pentium personal computer. HR was derived from the AP tracing. Mean arterial pressure (MAP), HR, RSNA, and single-unit discharge were averaged over 1-min intervals. A 3-min baseline was used as control. Changes in RSNA were calculated as a percent change from the baseline activity. In the representative tracings, both integrated and windowed RSNA are shown, but only the integrated activity was used in the analysis of grouped data. Statistical significance among multiple comparisons was determined by one-way or two-way repeated-measures ANOVA followed by post hoc Fisher's least squares difference test. Student's t-test was employed for analysis of paired or unpaired data. Values are expressed as means ± SE. P < 0.05 was considered to indicate statistical significance.

Anatomy/histology. At the conclusion of each experiment, the rat was killed with an overdose of urethane. The brain was removed and fixed in a 10% formalin solution for at least 3 days and then sectioned (40 µm) on a cryostatic microtome (OM2563, Triangle Biomedical Sciences, Durham, NC). The sections were thaw-mounted on microscope slides and then stained with 1% aqueous neutral red. The recording and microinjection sites marked with Pontamine sky blue were identified with a light microscope, and the locations of other recording sites were extrapolated with respect to this reference point. Recording and injection sites were plotted on representative schematic tracings of the PVN and RVLM, based on the rat atlas of Paxinos and Watson (49).

Drugs. Phenylephrine hydrochloride, PGE₂, and sodium nitroprusside were purchased from Sigma (St. Louis, MO). Recombinant rat TNF- was obtained from Research Diagnostic (Flanders, NJ). Ketorolac tromethamine was obtained from Abbott (Chicago, IL). All drugs were dissolved in aCSF for ICA or intracerebroventricular injection and for PVN microinjection or in saline for intravenous injection. ICA injections were given in a volume of 10-20 µl flushed by 25 µl aCSF (pH 7.5). The same volume of aCSF was administered ICA as a control.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Effect of ICA injection of TNF- on PVN or RVLM neuronal activity, RSNA, HR, and AP. Ipsilateral ICA injection of TNF- (0.5 µg/kg) dramatically activated hypothalamic PVN neurons (Figs. 1A and 2A, n = 9) or brain stem RVLM neurons (Figs. 1B and 2B, n = 10) and simultaneously increased RSNA, HR, and AP. The onset latency for this response was ~10-15 min with duration at least 90 min recorded. The peak responses occurred ~30 min after TNF- administration. PVN neuronal firing rate increased from a baseline of 2.2 ± 0.4 to a peak of 5.8 ± 0.8 spikes/s (164%, P < 0.001). Two PVN neurons showed no responses to TNF- and were not included in grouped data. Ten of 13 RVLM neurons tested were excited by TNF- from a baseline of 12.2 ± 2.9 to a peak of 21.0 ± 3.5 spikes/s (72%, P < 0.001); two RVLM neurons showed no changes in firing rate, and one was inhibited after TNF- injection. All RVLM neurons tested show pulse-related firing triggered from the peak of AP and were baroreceptor sensitive. The combined data from these studies showed that RSNA increased by 42.2 ± 5.7% from baseline (n = 19, P < 0.001), MAP increased in response to TNF- from a baseline of 95.1 ± 2.7 to 106.5 ± 2.5 mmHg (12.0%, n = 19, P < 0.001), whereas HR increased from a baseline of 342.2 ± 8.6 to 374.4 ± 9.7 beats/min (9.4%, n = 19, P < 0.001). The same volume of aCSF administered ICA had no effects on PVN (n = 2) or RVLM (n = 3) neuronal activity, RSNA, AP, or HR (n = 5; data not shown).

View larger version (34K): [in this window] [in a new window]   Fig. 1.   Representative tracings showing the responses of a paraventricular nucleus (PVN) neuron (A) and a rostral ventrolateral medulla (RVLM) neuron (B) to TNF- injected (arrows) into ipsilateral carotid artery (ICA). Simultaneously recorded renal sympathetic nerve activity (RSNA; discharge rate and integrated voltage), heart rate [HR; bpm (beats/min)], and arterial pressure (AP; mmHg) all increased.

View larger version (37K): [in this window] [in a new window]   Fig. 2.   Grouped data showing the effect of ICA TNF- (arrows) on PVN (A) and RVLM (B) neuronal discharge rate, RSNA, HR, and mean arterial pressure (MAP). Data points are means ± SE averaged over 1-min intervals.

Effect of vagotomy on intravenous TNF--induced sympathetic responses. Because vagus nerves may convey the blood-borne TNF- signal to the brain for sympathetic activation, we tested this possibility. Cervical vagotomy (n = 6) did not prevent the increases in RVLM neuronal activity (10.5 ± 1.1 to 18.3 ± 1.0 vs. 12.2 ± 2.0 to 18.7 ± 2.8 spikes/s, vagotomy vs. intact), RSNA (34.7 ± 3.6 vs. 35.8 ± 7.4%), HR (352.2 ± 9.6 to 377.7 ± 9.5 vs. 334.2 ± 10.7 to 365.8 ± 7.8 beats/min), and MAP (92.1 ± 1.1 to 107.8 ± 2.5 vs. 93.3 ± 1.3 to 107.2 ± 2.3 mmHg) induced by intravenous TNF- in the intact animals (n = 4). As shown in Fig. 3, similar responses for both groups in amplitude, latency, and duration were observed. The baseline HR was higher in the vagotomized rats, as expected. This group of experiments indicates that the vagus nerves do not mediate the sympathetic response to circulating TNF-.

View larger version (32K): [in this window] [in a new window]   Fig. 3.   A: representative tracing illustrating the effect of intravenous TNF- on HR, AP, RSNA (shown both as integrated nerve activity and as windowed spike activity), and the discharge rate of an RVLM neuron in a vagotomized rat. Vagotomy did not prevent the sympathoexcitatory response induced by intravenous TNF- (arrow). B: grouped data showing that the responses to intravenous TNF- were not different (P > 0.05) in rats with intact or sectioned vagus nerves. Values are means ± SE, averaged over 1-min intervals.

Effect of mid-collicular decerebration on intravenous TNF--induced sympathetic responses. In these experiments, we examined the contribution of the forebrain to the sympathetic responses to intravenous TNF- by removing its influence with a mid-collicular decerebration. After decerebration, the RSNA, AP, and HR increased. The baseline values for the decerebrate rats were higher than those of the intact rats (decerebrate vs. intact: RSNA 15.1 ± 2.1 vs. 11.2 ± 1.7 mV; MAP 111.9 ± 4.6 vs. 95.3 ± 3.3 mmHg; HR 377.0 ± 10.3 vs. 348.6 ± 11.2 beats/min). As shown in Fig. 4, intravenous TNF- did not induce the expected increase in HR (10.1 ± 1.0 vs. 1.0 ± 1.2%; intact vs. decerebrate; P < 0.001), MAP (24.7 ± 4.5 vs. 2.2 ± 3.1%, P < 0.001), or RSNA (37.0 ± 9.8 vs. 10.9 ± 3.4%, P < 0.001) in these anesthetized decerebrate rats (n = 9). Figure 4A illustrates the complete absence of HR and AP responses and the attenuation of the RSNA response to intravenous TNF- in an anesthetized decerebrate rat. Although the upward shift in baseline values after decerebration urges a cautious interpretation, it may be tentatively inferred that higher centers are necessary for the intravenous TNF--induced sympathoexcitation. These results also suggest that the forebrain has a tonic inhibitory effect on sympathetic output in the urethane-anesthetized rat.

View larger version (34K): [in this window] [in a new window]   Fig. 4.   A: representative tracing illustrating the effect of intravenous TNF- on HR, AP, and RSNA (shown both as integrated nerve activity and as windowed spike activity) in a decerebrate rat. HR and AP responses were completely abolished, and the RSNA response was substantially reduced by decerebration (dec). B: grouped data showing that the RSNA response to intravenous TNF- was substantially reduced in the decerebrate rats compared with intact group (P < 0.001), but the baseline levels of all 3 variables were higher in the decerebration group. Values are means ± SE, averaged over 1-min intervals.

Effect of intracerebroventricular PGE₂ on PVN neuronal discharge and RSNA. Prostaglandin is a possible mediator for TNF--produced sympathetic responses. We tested the central effect of exogenous PGE₂ on PVN neuronal activity and sympathetic output. Central PGE₂ administration (50 ng icv n = 8) elicited significant sympathoexcitatory responses: PVN single-unit discharge increased from 2.1 ± 0.3 to 5.2 ± 0.4 spikes/s; RSNA increased from 31.8 ± 8.5%; MAP increased from 90.6 ± 2.9 to 104.5 ± 2.5 mmHg; HR increased from 329.1 ± 10.6 to 396.9 ± 13.0 beats/min (P < 0.001 vs. baseline or aCSF control) as shown in Fig. 5. These responses are similar in amplitude to the TNF--induced responses but with relative shorter onset latency (2-3 min) and duration (30-40 min). Higher doses of PGE₂ had a more prolonged effect on RSNA, AP, and HR (data not shown). Intracerebroventricular injections of aCSF in the same volume (5 µl, n = 6) induced no sympathetic response, as illustrated in Fig. 5B. Intravenous injections of the same dose of PGE₂ (50 ng iv, n = 2) also had no effects on these parameters.

View larger version (30K): [in this window] [in a new window]   Fig. 5.   A: representative tracing illustrating the effect of intracerebroventricular (ICV) PGE2 (arrow) on HR, AP, RSNA (shown both as integrated nerve activity and as windowed spike activity), and the discharge rate of a PVN neuron. Intracerebroventricular PGE2 induced a sympathoexcitatory response similar to intravenous TNF-. B: grouped data showing the time course of effects of intracerebroventricular PGE2 on PVN neuronal activity, RSNA, HR, and MAP, compared with vehicle [artificial cerebrospinal fluid (aCSF)] treatment (P < 0.001). Values are means ± SE, averaged over 1-min intervals.

Effect of microinjection of PGE₂ directly into PVN on RVLM neuronal discharge and RSNA. Because intracerebroventricular PGE₂ elicited PVN neuronal and sympathetic responses, we tested the possibility that PGE₂ might act directly on PVN neurons to increased sympathetic drive. Microinjection of PGE₂ (50 ng, n = 8) directly into PVN activated descending sympathoexcitatory outputs: RSNA increased 28.9 ± 3.3% from baseline (P < 0.001); MAP increased from a baseline of 89.6 ± 2.7 to 97.6 ± 3.0 mmHg (P < 0.001); HR increased from a baseline of 328.9 ± 9.1 to 367.6 ± 9.4 beats/min (P < 0.001); and RVLM single-unit activity increased from a baseline of 10.9 ± 1.7 to 19.5 ± 3.2 spikes/s (P < 0.001). As shown in Fig. 6, the responses are similar to those of intracerebroventricular PGE₂, with rapid onset (2-3 min) and similar amplitude. However, the duration of responses was longer compared with intracerebroventricular injection, especially for HR response (more than 90 min). Control microinjections of PGE₂ (50 ng in 100 nl) at sites (see Fig. 8) ~1 mm lateral to PVN, in the lateral hypothalamic nucleus, and ~1 mm caudal to PVN, in and dorsal to the arcuate nucleus, had minimal or no effect on HR, MAP, and RSNA. The same volume of aCSF (100 nl, n = 7) microinjected into PVN also induced no significant sympathetic responses (Fig. 6B). These results suggest that PGE₂, whether locally produced or diffusing across BBB, can directly activate PVN neurons to elicit a sympathoexcitatory response.

View larger version (35K): [in this window] [in a new window]   Fig. 6.   A: representative tracing illustrating the effect of microinjection of PGE2 directly into PVN (arrow) on HR, AP, RSNA (shown both as integrated nerve activity and as windowed spike activity), and the discharge rate of an RVLM neuron. PGE2 excited PVN neurons and induced a sympathoexcitatory response similar to intravenous or ICA TNF-. B: grouped data show the time course of effects of PGE2 microinjected into PVN on RVLM neuronal activity, RSNA, HR, and MAP, compared with vehicle (aCSF) treatment (P < 0.001). Values are means ± SE, averaged over 1-min intervals.

Effect of blocking central prostaglandin synthesis with a COX inhibitor on TNF--induced sympathetic responses. The above experiments suggest that prostaglandins within the brain can mimic the cardiovascular and autonomic effects of blood-borne TNF-. In this group of experiments, we tested the hypothesis that prostaglandin synthesis within the brain mediates the cardiovascular and autonomic responses to blood-borne TNF-. Rats were treated with the COX inhibitor ketorolac, administered intracerebroventricularly before ICA injection of blood-borne TNF- compared with control treatment (aCSF, icv, n = 9), ketorolac treatment (150 µg/kg icv, n = 6) blocked the responses to ICA TNF- (0.5 µg/kg): PVN neuronal activity (216.7 ± 46.0 vs. 4.3 ± 4.2%, a CSF vs. ketorolac, P < 0.001), RSNA (34.7 ± 4.5 vs. 4.5 ± 4.8%, P < 0.001), AP (11.8 ± 1.9 vs. 4.9 ± 2.7%, P < 0.01), and HR (10.5 ± 2.6 vs. 4.0 ± 3.0%, P < 0.001) (Fig. 7). Because ketorolac does not readily cross the BBB (50, 54), this observation suggests that the sympathoexcitatory responses to circulating TNF- are dependent on central prostaglandin synthesis. The intracerebroventricular ketorolac had no effect on baseline PVN neuronal activity, RSNA, AP, and HR. Thus, endogenous prostaglandin synthesis in the CNS does not appear to have a tonic action on basal sympathetic activity in the urethane-anesthetized rat.

View larger version (38K): [in this window] [in a new window]   Fig. 7.   A: representative tracing illustrating the effect of central pretreatment (30 min) with the COX inhibitor ketorolac (ket; 150 µg/kg icv) on responses of HR, AP, and RSNA (shown both as integrated nerve activity and as windowed spike activity), and the discharge rate of a PVN neuron to ICA TNF-. B: grouped data showing that pretreatment with ketoralac almost completely blocked the sympathoexcitatory responses to ICA TNF- (P < 0.001). Values are means ± SE, averaged over 1-min intervals.

Recording and microinjection sites in PVN and RVLM regions. Figure 8 shows the locations of the single-unit recording sites in PVN (A) and RVLM (B) areas and the microinjection sites in PVN and surrounding tissue (8A). The PVN neurons tested in this study were distributed mainly in the medial regions close to the third ventricle. Within this region, there was no obvious difference in anatomic distribution of neurons responsive to blood-borne TNF- or centrally administered PGE₂. The specific functions of the responsive PVN neurons were not determined in the present study. However, local microinjection of PGE₂ into this same region of PVN activated downstream RVLM neurons and RSNA and increased AP and HR, as described above. Neurons recorded in RVLM were all barosensitive, but their projection sites were not determined. Within RVLM, there was no obvious difference in anatomic distribution between neurons activated by ICA or intravenous TNF- or by microinjection of PGE₂ into PVN.

View larger version (38K): [in this window] [in a new window]   Fig. 8.   Schematic reconstruction of recording and microinjection sites within PVN (A) and RVLM (B) regions. Locations shown are TNF--responsive neurons () and nonresponsive neurons () within PVN and RVLM, PGE2-responsive PVN neurons (), RVLM neurons () responding to PVN microinjection of PGE2, and PGE2 microinjection sites within PVN () and outside PVN (). The distance from bregma is indicated. Sections are modified from Paxinos and Watson (49). AH, anterior hypothalamic nucleus; ARC, arcuate nucleus; DMD, dorsomedial hypothalamic nucleus; f, fornix; IO, inferior olive; LHA, lateral hypothalamic area; LPGi, lateral paragigantocellular nucleus; ME, median eminence; NA, nucleus ambiguus; NTS, nucleus of the solitary tract; py, pyramidal tract; VMH, ventromedial hypothalamic nucleus; 3V, third ventricle.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

This study provides new insights into the mechanisms by which circulating TNF- activates the central excitatory pathways regulating cardiovascular and renal function. Findings novel to this study include: 1) ipsilateral intracarotid administration of TNF- stimulates both PVN and RVLM neurons and simultaneously recorded RSNA and increases AP and HR; 2) intracerebroventricular pretreatment with a COX inhibitor produces nearly complete blockade of the PVN and RSNA response to ICA TNF-; 3) PVN neurons and RSNA are activated by intracerebroventricular administration of PGE₂; and 4) RVLM neurons and RSNA are activated by microinjection of PGE₂ into PVN. These data confirm with direct electrophysiological recordings the hypothesis that the excitatory influences of acutely administered blood-borne TNF- on AP, HR, and RSNA are mediated by the central actions of prostaglandins. Moreover, they indicate a cardiovascular regulatory role for prostaglandin production within the CNS perhaps by a direct action of prostaglandins on hypothalamic neurons, in contrast to the indirect action via activation of catecholaminergic brain stem neurons that has been proposed for HPA axis activation (17). The RVLM neurons recorded in this study were pulse synchronous and baroreceptor modulated and thus likely a component of a descending pathway relaying signals from PVN to sympathetic preganglionic neurons in the intermediolateral cell column (IML), although other sources of excitatory input to the RVLM neurons and direct projections from PVN to IML (36) may also have contributed to the observed sympathoexcitatory responses.

The extant literature regarding the central neural influences of cytokines focuses on their role as the mediators of immune or stress responses. In both of those conditions, the defining feature of the CNS response is activation of parvocellular PVN neurons that contain CRF. The majority of these neurons are neurosecretory and project to the median eminence, where CRF is released into the pituitary portal system to stimulate the release of ACTH. The final product of this cytokine activation of the HPA axis is corticosterone in the rat or cortisol in humans.

The central pathways mediating HPA activation and ACTH release have been well studied. In a series of functional anatomic studies that are illustrative of current thinking with regard to mechanisms for activation of the HPA response, Ericsson and colleagues (17, 18) demonstrated that the activation of the neurosecretory CRF neurons by systemically administered IL-1 is dependent on the initial activation by PGE₂ of catecholaminergic neurons in the medulla oblongata that ascend to innervate parvocellular PVN. Consistent with that model is the finding that chemical destruction of catecholaminergic terminals in PVN reduces the HPA axis response to systemic IL-1 (9, 48, 60) and the effect of stimulation of the A₁ and A₂ catecholamine cell groups on tuberoinfundibular neurons in PVN (11). Interruption of inputs from vagal afferent fibers (17) and from forebrain (18) and hindbrain (17) circumventricular organs (17), two other potential routes by which cytokines might activate central neurons, did not significantly alter the response. In contrast, systemic administration of the COX inhibitor indomethacin attenuated cytokine activation of the critical medullary and hypothalamic neurons (17). Strong evidence now supports the concept that transduction of the cytokine signal across the BBB is mediated by PGE₂ produced by endothelial cells in the cerebral microvasculature, which is rich in cytokine receptors.

A second important component of the HPA response to cytokines is an increase in circulating catecholamines, epinephrine from adrenal medulla and norepinephrine from active sympathetic nerve terminals. However, the available data regarding cytokine stimulation of the sympathetic nervous system do not conform to the generally accepted model for cytokine activation of the glucocorticoid response. In the present study, for example, PGE₂ administered into the lateral ventricle elicited increases in PVN neuronal activity and RSNA, as well as increases in AP and HR, and PGE₂ microinjected directly into PVN elicited increases in RVLM neuronal activity and RSNA, as well as the cardiovascular response. These results mimicked the response to intracarotid injection of TNF-, which was largely blocked by intracerebroventricular administration of a COX inhibitor. In comparable studies (44), injection of PGE₂ into the lateral ventricle or directly into the preoptic area elicited fever and thermogenesis in brown adipose tissue, operating over an alternative sympathoexcitatory pathway via the rostral raphe pallidus. In work examining cytokine activation of the splenic nerve, important in feedback regulation of the immune response in lymphoid tissues, systemic endotoxin (34) and third ventricular PGE₂ (2) and third ventricular IL-1 (24) have all been shown to increase splenic nerve activity (2), and the endotoxin and IL-1 effects were blocked by intracerebroventricular pretreatment with a COX inhibitor. In aggregate, these studies strongly suggest that blood-borne cytokines initiate sympathetic responses at the forebrain level via the soluble mediator PGE₂. There is also the suggestion from this and one previous study (35) that cytokine-induced PGE₂ production inside the BBB plays an important role in driving sympathetic systems.

Systemically administered cytokines or endotoxin, which induces cytokine release, reportedly increase splenic (2, 34, 47, 55), adrenal (47), and lumbar (55) sympathetic activity, have variable effects on renal sympathetic drive (34, 47) and activate sympathetic mechanisms mediating the febrile response (44). If mediated centrally by PGE₂, these responses may be largely dependent on the site-specific distribution of PGE₂ receptors. When introduced into the lateral ventricle, PGE₂ induces prominent c-fos activation of several forebrain structures that might mediate sympathetic drive (31), including the medial preoptic area and the parvocellular PVN, as well as in several hindbrain regions involved in cardiovascular reflex control, including the nucleus of the solitary tract and the ventrolateral medulla. Because these observations were made 30 min or longer after PGE₂ injection, the relative time course for activation of the different structures is not known. PGE₂ receptors (EP1-EP4) committed to different function are present in many of these same brain regions (62). For example, the EP1 receptor appears to mediate the splenic nerve responses (2), and the EP3 receptor on neurons of the preoptic area is said to mediate the febrile response to cytokines (44). Intracerebroventricular PGE₂ also elicits sympathetically mediated (22) increases in AP and HR, but the specific receptor involved has not been identified. It therefore seems reasonable to speculate that PGE₂ activation of more than one receptor type in more than one forebrain structure might contribute to cytokine-induced alterations in sympathetic drive. Differential activation of sympathetic pathways from the hypothalamus has been described (40). Thus, activation of EP3 receptors in preoptic area might induce fever and an increase in HR, AP, and sympathetic nerve activity to brown adipose tissue via a central projection to raphe pallidus (41) and then IML, whereas activation of EP4 receptors in PVN might increase in HR, AP, and RSNA by another central pathway to RVLM and then IML. With regard to the potential role of the PVN in the cardiovascular response, both push-pull (58) and microdialysis (28) studies have shown that IL-1 stimulates the release of PGE₂ from the hypothalamus, and the present study demonstrates a prominent cardiovascular response to microinjection of PGE₂ into PVN, mimicking the response to ICA TNF-. In the present study, we did not employ selective prostaglandin receptor agonists and antagonists, but such studies may ultimately help clarify the involvement of different hypothalamic sites and different PGE₂ receptors in coordinating the sympathetic response to circulating cytokines.

In considering the relationship between hypothalamic PGE₂ and sympathetic drive, it is important to recognize that PGE₂ is only one of several putative neurotransmitters in this region of the brain that are affected by circulating cytokines. Systemic injection of IL-1 increases hypothalamic norepinephrine concentration, paralleling the increases in the plasma corticosterone levels (13, 15). Norepinephrine is usually considered in light of its inhibitory effects via -2 adrenergic receptors, but it may excite PVN neurons via activation of -1 adrenergic receptors (25, 29, 45). A high norepinephrine content in PVN is found in states of enhanced sympathetic excitation, including hypertension (61) and heart failure (6), the latter associated with chronic high circulating cytokine levels and augmented sympathetic drive. Both peripheral endotoxin (53) and central PGE₂ (31) dramatically and selectively upregulate CRF-R1 receptors in PVN, where they are not normally expressed. A CRF antagonist can be shown to block the increase in splenic nerve activity elicited by intracerebroventricular PGE₂ (27). On the other hand, a CRF antagonist had no effect on the fever or the HR and AP changes induced by intracerebroventricular PGE₂ (43). Because norepinephrine, CRF, and CRF-R1 receptors are all products of the HPA response described in the model above, an interaction at hypothalamic level between PGE₂ and these substances might be the link that coordinates the glucocorticoid and sympathetic responses. Of course, other neurotransmitters/mediators also present within the PVN (4, 32, 33, 63) may also be involved in the sympathetic response to cytokine stress.

Most previous studies evaluating the alternative mechanisms by which the circulating cytokines signal the CNS have not dealt specifically with the cardiovascular variables or the renal sympathetic response. In the context of HPA axis activation, some studies have suggested that vagal sensory mechanisms play an important role (21). Others have suggested that the abdominal vagi mediate central responses to cytokines or endotoxin when administrated intraperitoneally but not when intravenously injected (20, 26). However, one recent study (55) dealing specifically with cytokine effects on sympathetic discharge to various vascular beds reported that increases in sympathetic drive elicited by systemically injected IL-1 are not dependent on the vagus nerves; in that study, the sympathetic responses were actually enhanced by vagal denervation. Our data are consistent with those findings, demonstrating that the increases in RVLM neuronal activity, AP, HR, and renal sympathetic drive induced by acute intravenous injections of TNF- were not different in rats with vagi intact or sectioned and thus appear to be independent of any central influence that might be mediated by vagal afferent signals.

We also tested the effect of a mid-collicular decerebration on the cardiovascular and renal sympathetic responses to intravenous TNF-. This procedure, of course, interrupts both ascending and descending connections between forebrain and hindbrain. We found that the sympathetic response to systemic TNF- was substantially reduced in the decerebrate rats, suggesting that the forebrain is required for the full response to systemic TNF-. However, the interpretation of these data is rendered somewhat tenuous because of differences in the baseline variables in the decerebrate rats; an augmentation of sympathetic drive in response to TNF- could have been masked in these rats.

Several limitations of this study deserve attention. First, our interpretation assumes that the RVLM neurons recorded are components of a descending sympathoexcitatory pathway from PVN to RVLM to IML. A previous anatomic study showed that neurons in RVLM and PVN activated by LPS have direct projections to the spinal cord (64). All the RVLM neurons we studied were pulse phasic and baroreceptor sensitive, but we did not perform antidromic stimulation to confirm their projection pathway. Likewise, we did not determine the projection site of the recorded PVN neurons. Moreover, while RVLM neurons were responsive to ICA TNF- and to PVN PGE₂, we did not inject a PGE₂ antagonist into PVN to demonstrate that the ICA TNF- activated these neurons via a PGE₂-sensitive site in PVN. Thus, although the available data suggest that a PGE₂-sensitive site in PVN is activated by blood-borne TNF- to drive HR, AP, and RSNA, at least in part via a synapse in RVLM, additional studies will be required to prove the point.

Another concern relates to the possibility that the effects of the PGE₂ injections, both into lateral ventricle and into PVN, might have resulted from effects outside the intended targets. As mentioned, intracerebroventricular infusion of a large dose of PGE₂ (31) activated neurons in both forebrain and hindbrain sites at the earliest interval checked (30 min). Our dose was smaller, and the responses were relatively brisk in onset, but we cannot exclude PGE₂ effects mediated by remote hindbrain sites. In the case of the PVN microinjection, the dose and volume were intended to stimulate a sufficient number of neurons to elicit a downstream sympathetic response; they were not intended to map specific responsive regions within PVN. The dose we used is well within the range (1-1,000 ng, most commonly 25-100 ng) used by others for microinjection in the hypothalamic (38, 39) and brain stem (17) regions, but it is possible that some of the drug may have spilled into the CSF. However, with regard to the specific possibility of downsteam activation of the RVLM by leakage into the ventricular system, microinjection of the same or higher dose (50-150 ng) PGE₂ directly into RVLM has little or no effect on HR, AP, or RSNA in our preparation (unpublished data). With regard to possible influences on relevant brain tissue in the immediate vicinity of the PVN, it is notable that dorsomedial, ventromedial, and ventrolateral hypothalamic nuclei did not respond to a larger intracerebroventricular dose of PGE₂ (31), and we observed minimal responses at sites outside PVN.

Finally, a caveat to consider is that the present study and the previous studies investigating the central pathways and neurotransmitters mediating the cytokine response have relied on responses to acute injections of cytokines, endotoxin, or PGE₂. In chronic disease states like advanced heart failure, in which persistent high circulating levels of all three of these may be present (3, 12, 16, 46), conditions in the brain are likely altered. Cytokines and endotoxin facilitate the entry of peripherally produced PGE₂ into the brain (10). In addition, increased circulating levels of endotoxin (56) and PGE₂ (57) may induce brain production of cytokines, and brain cytokines may elicit local production of other factors, including PGE₂ (23). Other mechanisms of cytokine signaling from the periphery including transport across BBB (5) and vagal afferent activation (36) may also come into play under these conditions. We recently demonstrated that TNF- is present in hypothalamic neurons in a rat model of ischemia-induced heart failure (19).

In summary, we have presented electrophysiological evidence that the excitatory cardiovascular and renal sympathetic responses to circulating TNF- are mediated by the actions of prostaglandins, likely PGE₂, synthesized within the CNS. The exact prostaglandin receptor subtypes and the sites within the brain that are responsible for these effects remain to be determined, but our data strongly suggest an important action at the level of the PVN of the hypothalamus. These observations may have particular relevance to the altered neurohumoral state in heart failure, in which circulating cytokines and sympathetic drive are both chronically elevated.