Fig. 4. Daily abdominal temperature (mean ± s.d.) of unrestrained rabbits for the first seven days after implantation of osmotic pumps (day 1) continuously infusing saline (open symbols, n — 5) or a Gram-negative pyrogen, lipopoly-saccharide (0.2 ng^1 kg"1 min~', closed symbols, w = 5) sub-cutaneously. Arrow indicates time of implantation. Daily temperatures are averages of 5-min values for each rabbit. Dotted line represents mean daily body temperature of rabbits implanted with loggers only.

Substantial evidence now exists that a conclusion that was surmised over 100 years ago (Welch, 1888, vide supra) is accurate, and that, in most fevers, exogenous fever-producing agents act indirectly on thermoregulatory neurons, via cytokines. Injection of cytokines via intracarotid routes evokes fever of shorter latency and greater magnitude than injection via i.v. routes (King and Wood, 1958), cytokine mRNA and bioactive protein are synthesised in the brain in response to peripheral pyrogenic challenge (see for review, Turnbull and Rivier, 1999; Zeisberger, 1999; Rivest et al., 2000), and receptors to these cytokines have been localised within the CNS, the expression of which is increased after systemic LPS or cytokine administration (see for review, Rothwell et al., 1996). Microinjection of proinflammatory cytokines into specific hypothalamic areas evokes fever, at doses far lower than those required peripherally (see Kluger, 1991). Cytokine administration in vivo and in vitro alters the activity of hypothalamic thermosensitive neurons, in a way which is consistent with the development of fever (see for review, Boulant, 1996), and inhibiting brain cytokine action by central administration of specific cytokine antagonists or neutralising antisera attenuates fever (see Kluger, 1991). Knockout mice with specific cytokine deficiencies are unable to mount a fever response to central (and peripheral) exogenous pyrogen or cytokine challenge (see for review, Leon

2002). There is no evidence that cytokine activity outside the brain has any thermal consequences (Bryce-Smith et al., 1959; Frens, 1971).

Whatever the central mediators ultimately responsible for evoking fever, they will modulate the thermoregulatory system which operates as a negative-feedback system, with body temperature regulated around a 'setpoint', under the control of integrating neuronal networks in the CNS (see for review, Boulant, 1980). Data collected over numerous years of investigation of the relationship between body temperature and thermoregulatory effectors refute earlier descriptions of the thermoregulatory system as a monolithic negative-feedback system in which inputs form various sensors converged on a single central controller responsible for the full suite of effectors. Rather, it is now accepted that there is an assembly of neural networks dispersed widely in the CNS, but with key neurons concentrated in the preoptic region, which act as co-ordinating centres that control each of the separate autonomic and behavioural mechanisms (see Boulant, 1996). Early electrophysiological recordings from rostral hypothalamic neurons revealed the presence of a surprisingly large population of thermosensitive neurons, characterised by the way their firing rates changed during changes in local, hypothalamic temperatures (Nakayama et al., 1963; see for review, Boulant, 1980, 1996). Several attempts have been made to study how the firing rates of such neurons change during fever, and they point towards a system in which pyrogenic molecules decrease firing rates of warm-sensitive neurons, and increase firing rates of cold-sensitive neurons (as a consequence of synaptic inhibition), thereby suppressing heat-loss responses (panting and sweating), and enhancing heat production (shivering or non-shivering thermogenesis) and heat-retention responses (see Mackowiak and Boulant, 1996). These changes in firing rate, shown experimentally by local application of endogenous pyrogens (Nakashima et al., 1989,

1991), can be reversibly blocked by application of COX inhibitors (Hori et al., 1988; Xin and Blatteis,

When pyrogenic molecules affect thermoregulatory neurons, the outcome for the thermoregulatory system will be an apparent elevation of setpoint. The traditional view envisages a global elevation, so that the temperature of all elements of the body core will increase. We have developed the proposal by Satinoff (1978) of a parallel hierarchical system, parallel in that each effector could be assigned its own controller, and hierarchical in that some controllers have a greater capacity to influence thermoregulation than others, to include subsystem controllers responsible for the autoregulation of elements such as scrotal and brain temperature (see Mitchell and Laburn, 1997). Thus, if autoregulation fails or is overwhelmed, then a higher-ranking system can be invoked to regulate the temperature of the subsystem by regulating the whole system containing it (see Fig. 5). We recently advanced this concept one step further to suggest that different pyrogens might affect different controllers within a parallel hierarchical



Controlled variable

Fig. 5. A nested parallel hierarchial feedback model of thermoregulation, incorporating controllers for different effectors and subsystem controllers for components such as scrotal thermoregulation and selective brain cooling. If the autoregulation fails or is overwhelmed, then a higher-ranking system can be invoked, to regulate the temperature of the subsystem by regulating the whole system that contains it. Redrawn by Peter Kamerman from Mitchell and Laburn (1997), by permission.

Controlled variable

Fig. 5. A nested parallel hierarchial feedback model of thermoregulation, incorporating controllers for different effectors and subsystem controllers for components such as scrotal thermoregulation and selective brain cooling. If the autoregulation fails or is overwhelmed, then a higher-ranking system can be invoked, to regulate the temperature of the subsystem by regulating the whole system that contains it. Redrawn by Peter Kamerman from Mitchell and Laburn (1997), by permission.

thermoregulatory system in different ways, in a manner specific to that particular pyrogen (Mitchell and Laburn, 1997). So, during fever, some subsystems may be excluded from the elevation in setpoint.

Whether circulating pyrogenic molecules actually reach thermoregulatory neurons themselves, during systemic fevers, or whether they trigger brain-message pathways from sites outside the brain, remains unknown. In an elegant and extensive series of studies, using in situ hybridisation, Rivest and colleagues (Vallieres and Rivest, 1997; Vallieres et al., 1997; Lacroix and Rivest, 1998; Lacroix et al., 1998; Laflamme et al., 1999; Nadeau and Rivest, 1999; Lebel et al., 2000; Nadeau and Rivest, 2000; Zhang and Rivest, 2000) have mapped the potential brain circuitry solicited in response to systemic LPS or cytokine administration. The brain areas targeted and cell types activated are remarkably similar following systemic treatment with LPS, IL-ip or TNF, the most likely targets being areas devoid of blood-brain barrier, namely the sensory circumven-tricular organs (CVOs, an idea originally put forward by Hellon and Townsend (1983)), and the micro-vasculature itself. The actions of IL-6 are quite different and dependent on the origin (central or systemic) of IL-6 during challenge. Additionally, using methods of neuroanatomical tracing and Fos staining (a marker for neuronal activation), the patterns of activation in the CNS in response to peripheral LPS or cytokine administration have been mapped systematically (Ericsson et al., 1994; Wan et al., 1994; Sagar et al., 1995; Elmquist and Saper, 1996; Elmquist et al., 1996; Herkenham et al., 1998). Peripheral administration of exogenous pyrogenic stimuli leads to activation of central autonomic systems capable of producing profound behavioural and physiological changes characteristic of the acute-phase response. Moreover, the ventromedial preoptic area (VMPO), a cell group adjacent to the organum vasculosum lamina terminalis (OVLT), and parvo-cellular component of the PVN emerge as potential key sites ('hot spots') for the initiation of fever during endotoxaemia (Elmquist and Saper, 1996; Scammell et al., 1996, 1998). Although these contributions have indeed helped in identifying pathways of influence on neuronal systems that underlie the cerebral component of fever, and responses to systemic pyrogen challenge, the pathways by which thermoregulatory neurons are influenced during discrete localised infection and/or inflammation, in which the exogenous pyrogenic stimulus is contained at the site of inflammation, and in which the circulating cytokine profile may differ from that following systemic challenge (see below), remain to be identified.

Experimental models of fever: assembling the pieces

Parenteral injection of LPS is the most popular experimental model of the fever induced by bacterial septicaemia, in which the circulating exogenous pyrogen evokes fever primarily as a result of the direct stimulation of brain vascular endothelium via a membrane-bound receptor, and via the cytokines it induces (Dinarello et al, 1999). In contrast, injections of turpentine and of LPS into a pre-formed subcutaneous airpouch are used as experimental models of sterile localised inflammation and localised infection, respectively. In both these models, the exogenous pyrogen remains at the site of its administration, and so cannot have direct access to the brain. The airpouch model of inflammation allows monitoring of the clearance of the exogenous pyrogen, and of the release of local mediators at the site of inflammation, as well as their appearance in the plasma and elsewhere (e.g. brain and cerebrospinal fluid, CSF). The role of cytokines in these three experimental fever models has been much studied, primarily in rats and mice because of the ready availability of cytokine assays for those species.

In the case of the experimental model of systemic infection, that is in response to i.v. or i.p. LPS administration, TNF appears first in the circulation, followed by trace amounts of IL-1 and larger amounts of IL-6 (Givalois et al, 1994). The topographical, temporal and cellular induction of brain IL-1 (3, IL-6 and TNF mRNA are similar but not identical and at early time points, hybridisation signal is most prevalent in the choroid plexus, the meninges and in the CVOs implicated in fever (see for review and list of references, Turnbull and Rivier, 1999; Rivest et al, 2000), and COX-2 mRNA is detected in cerebral endothelial microvasculature, perivascular microglia and meningeal macrophages (Lacroix and

Rivest, 1998). Additionally, bioactive TNF and IL-6 appears in the hypothalamus in response to LPS administered i.p, although it appears that these cytokines do not derive from the blood and their activities do not correlate exactly with the fever (Klir et al, 1993; Roth et al, 1993). A similar time sequence of circulating cytokines has been observed in patients who are 'critically ill' (see Dinarello, 1997). In response to turpentine (that is, in the absence of a microbial pyrogen), there is a marked increase in plasma concentrations of IL-6 (and in some laboratories, IL-1), but not of TNF (Cooper et al, 1994; Luheshi et al, 1997; Turnbull et al, 1997). However, inhibition of TNFa or IL-1 action markedly attenuates the fever induced by turpentine, and reduces also the concentrations of IL-6 in the blood, implying that both these cytokines are involved in IL-6 secretion, but their involvement probably is confined to the local inflammatory site (Cooper et al, 1994; Turnbull et al, 1994; Luheshi et al, 1997). The effect of sterile local inflammation (turpentine) on brain cytokine expression has been less well studied, and no increase in brain IL-1 P, IL-6 or TNF mRNA has been detected (Turnbull et al, 1997). Finally, in response to intrapouch injection of LPS, there is a significant elevation in concentrations of TNF, IL-1 and IL-6 (in that sequence) at the site of inflammation (that is within the pouch), but only of IL-6 in the circulation and CSF (Miller et al, 1997b; Cartmell et al, 2000, 2001). Tumour necrosis factor disappears in the pouch at a time when pouch IL-1 and IL-6 concentrations are at their peak (Miller et al, 1997b). The source of circulating IL-6 appears to be the site of inflammation, and circulating IL-6 is essential for evoking the fever (Cartmell et al, 2000). Recent findings, although largely circumstantial, raise the possibility that the chemokine CINC-1, which is similar structurally to the pyrogenic human chemokine IL-8 (Zagorski and DeLarco, 1993), might be produced in the pouch in response to exogenous pyrogen and may be released into the circulation (Cartmell, T, Ball, C, Bristow, A.B, Mitchell, D. and Poole, S, unpublished data). It has been proposed that IL-8 and IL-6 both have important roles in transforming acute into chronic inflammation (Marin et al, 2001).

Other pyrogenic moieties of Gram-negative and Gram-positive bacteria, such as superantigens, peptidoglycans and MDP, like LPS, induce a circulating cytokine cascade after systemic administration. When MDP is administered to rats and guinea pigs, the blood cytokine profile is similar to that following i.p. LPS (see Roth et al., 1997a; Zeisberger, 1999). Administration of synthetic viral compounds activates a different pattern of cytokine production, which includes IFNs as final mediators in the cytokine cascade (Dinarello et al, 1984).

There remains uncertainty, even amongst fever researchers, concerning the relative contributions of the different cytokines, in particular of TNF, 1L-1 and IL-6, to the genesis of fever, especially given that the three cytokines affect the synthesis and secretion of each other and are capable of enhancing the others' effect in a synergistic manner. One source of confusion is the counter-intuitive observation that, although IL-6 is the most likely candidate for the final pyrogenic cytokine in the cascade, systemic injection of IL-6 in experimental animals usually does not cause fever (see Cartmell et al, 2000). Coordinated, regulated, release of the three cytokines certainly occurs sequentially in the periphery, with TNF appearing before IL-ip, and IL-ip appearing before IL-6. Since IL-ip can induce IL-6 release, but IL-6 cannot induce IL-1P release (in fact it suppresses expression of both IL-1 and TNF (Schindler et al, 1990)), IL-6 indeed appears to be the critical pyrogenic cytokine.

It also is the case that IL-1, not IL-6, is the most potent pyrogenic cytokine (gram for gram) when injected either systemically or into the brain (see Dinarello, 1996; Rothwell et al, 1996). Similarly, elimination of peripheral macrophages (the main source of circulating IL-1, but not IL-6) prevents the increase in plasma IL-1 and attenuates the attendant fever evoked by high-dose endotoxin in rats (DeRijk et al, 1991, 1993). Although IL-1 may circulate in the blood in severe, systemic infections (Cannon et al, 1990) in clinical situations, IL-1 remains restricted to the site of infection or inflammation (Engel et al, 1994; Luheshi et al, 1997; Miller et al, 1997a,b; Cartmell et al, 2001) and appears to evoke fever by local induction of other mediators, including IL-6 (Luheshi et al, 1996; Miller et al, 1997a,b; Cartmell et al, 2000), which enters the circulation from the site of infection and/or inflammation (Cartmell et al, 2000). Nevertheless, the role of IL-1 in the generation of fever is pivotal: IL-1 ß has a critical role in the activation of NFkB and PGs within endothelial cells of the blood-brain barrier in turpentine-induced inflammation (Laflamme et al, 1999). Also, fever in response to LPS is inhibited by administration of anti-IL-1 sera or IL-lra (Long et al, 1990; Smith and Kluger, 1992; Klir et al, 1994; Luheshi et al, 1996; Cartmell et al, 1999), and IL-1 ß is essential for fever evoked during poxvirus infection, despite the absence of detectable IL-1 ß in the plasma (Alcami and Smith, 1996). Recently, the cytokine leptin has been reported to evoke fever when administered peripherally or into the brain and is thought to do so via upregulation of IL-1 expression in the hypothalamus (Luheshi et al, 1999). Whether IL-1 is an essential component of the cascade in LPS-induced fever is questionable, however: IL-1 ß knockout mice respond with only a slightly reduced (Kozak et al, 1995) or even enhanced (Alheim et al, 1997) fever in response to administration of LPS, and although IL-1RI is seen to be essential for all IL-1-mediated signalling events (Labow et al, 1997), IL-1RI knockout mice develop fever in response to administration of LPS (Leon et al, 1996). Burysek and colleagues (1997) have shown that brown adipocytes can produce IL-1 and IL-6 in response to LPS, and Cannon and colleagues (1998) have shown that brown adipocytes are more sensitive to LPS (enhanced uncoupling protein (UCP)-l gene expression) in IL-1 ß knockouts than in wild types. If the ability to express, synthesise and release cytokines, such as IL-6, from brown adipose tissue is increased in knockout mice, as a response to LPS, this enhanced sensitivity might explain the ability of IL-1 P-deficient mice to respond normally to LPS (Cannon et al, 1998).

So, IL-1 ß appears to be an important, though perhaps not essential, cytokine mediator, acting where exogenous pyrogens encounter myeloid cells. The role of IL-6 appears to be that of the major circulating endogenous pyrogen. It is the only proinflammatory cytokine that can be detected (as bio-active and immunoreactive) in significant quantities in the circulation during fever (Nijsten et al, 1987; LeMay et al, 1990a,c; Luheshi et al, 1997; Miller et al, 1997b; Cartmell et al, 2000) and the rise in circulating and CSF IL-6 concentrations parallels the development of fever (LeMay et al, 1990b,c; Roth et al, 1993) and is dependent on IL-1 (LeMay et al.

1990b; Klir et al, 1994; Luheshi et al, 1996; Miller et al, 1997a). Moreover, circulating IL-6 mediates the fever response to localised LPS administration (Cartmell et al, 2000) and 'normal' levels of IL-6 in IL-lß and IL-1RI knockout mice may be responsible for the residual fever observed in response to LPS in these animals (Zheng et al, 1995; Leon et al, 1996; Alheim et al, 1997; Kozak et al, 1998b). Circulating IL-6 can enter the brain via an active transport mechanism (see Banks et al, 1995) and injection of IL-6 directly into the brain induces fever (Le May et al, 1990c; Klir et al, 1993; Lenczowski et al, 1999). Finally, IL-6 knockout mice exhibit fever when IL-6 is injected in the brain but fail to develop fever in response to a systemic injection of IL-6, IL-1, TNF and a low, but not high dose, of LPS (Chai et al, 1996; Kozak et al, 1997). Consequently, the likely mechanism for the absence of fever following sterile local inflammation in IL-1 ß knockout mice is the reduction in (IL-1-driven) IL-6 production (Zheng et al, 1995; as in IL-1 ß knockout mice, IL-6 knockout mice fail to develop fever to local injection of turpentine (Kozak et al, 1997)).

Several findings support the hypothesis that, in those fevers that involve both IL-1 and IL-6, the two cytokines act not just sequentially, but in concert. IL-6 alone does not cause fever when injected systemically, but does so in the presence of a non-pyrogenic dose of IL-1 ß (Cartmell et al, 2000). IL-lß, although not a prerequisite for the induction of IL-6 following injection of LPS, is essential for IL-6 release in sterile local inflammation (Zheng et al, 1995; Leon et al, 1996), and both IL-6 and IL-lß are essential for the synthesis of PGE2 during sterile local inflammation (Kozak et al, 1998b). Interestingly, ICE knockout mice fail to generate mature IL-lß in response to LPS (Li et al, 1995) but exhibit normal production of IL-lß and display normal plasma IL-6 responses when injected with turpentine (Fantuzzi et al, 1997).

When IL-1 or IL-6 affects body-core temperature, it always is to increase the temperature. Tumour necrosis factor, on the other hand, has been reported to act as both a pyrogen and a cryogen (an agent that lowers body temperature in the absence of fever): murine TNF (which binds to both p55 and p75 receptors in rodents), but not human TNF (which binds only to p55 receptors in rodents) evokes fever in rats (Stefferl et al, 1996). Treatment with TNF antiserum or soluble TNF (sTNF) receptors attenuates LPS and turpentine-induced fevers in a variety of species (Kawasaki et al, 1989; Cooper et al, 1994; Roth et al, 1998) and rolipram (a type-IV phosphodiesterase inhibitor that inhibits the production of TNF) significantly inhibits the first, but not the second, phase of fever evoked in response to both Gram-negative and Gram-positive bacteria in rabbits (Mabika and Laburn, 1999). In TNF p55/p75 knockout mice, however, TNF acts as an endogenous cryogen, attenuating LPS-induced fever (Leon et al, 1997). According to the same team of researchers, treatment with neutralising antibodies to TNF or administration of sTNF receptors, enhances the fevers induced by LPS, whereas treatment with non-neutralising antibodies attenuates the fever (perhaps as a result of higher levels of circulating TNF) and, finally, injection of a low dose of TNF (which on its own has no effect on body temperature) attenuates LPS-induced fever (Long et al, 1990; Klir et al, 1995; Kozak et al, 1995). Many other investigations, however, conducted in several species, confirm that endogenous TNF has pyrogenic activity. Intrigu-ingly, cytokine production in response to i.p. LPS is intact in TNF knockout mice (Marino et al, 1997). In guinea pigs, TNF may not have a role in the genesis of fever, but only in its maintenance (Roth et al, 1998).

The controversial role of TNF, and the residual uncertainties about the roles of IL-1 and IL-6, are disconcerting to those who are seeking a united and coherent cytokine substrate for fever. Seeking unity and coherence may be unrealistic, however. Differences in experimental conditions, the dose and type of pyrogen, the route of pyrogen administration and the choice of species (it is well documented that rodents are far less sensitive to pyrogens than are rabbits and humans) may result in different profiles of mediators that subsequently activate different centripetal pathways, and, consequently yield seemingly discrepant results. One should bear in mind, too, that LPS is not a single chemical substance. Its chemistry depends on the parent bacterial species. Different LPS batches and hence serotypes differ in biological properties and potencies (see Henderson et al, 1998). Also, the inability to detect the presence of a cytokine in the circulation of febrile animals using ELISAs (immunoreactive) or bioassays (bioactive) does not necessarily mean absence of that cytokine. For example, it has been reported that concentrations of IL-1 as low as 1 pg/ml may exert biological effects in vivo (Dinarello, 1996), and such concentrations are below the sensitivity of most ELISAs. Although bioassays offer the advantage of increased sensitivity, they are not species specific and can be affected by inhibitory molecules, such as soluble receptors and, in the case of IL-1, IL-lra (see Kluger, 1991; Dinarello, 1996). Moreover, data obtained via inhibitory pharmacological agents can be interpreted accurately only if the injected substance completely antagonises the targeted cytokine (e.g. IL-1) at all sites within the body. Finally, it is important to be cognisant of the fact that animals that have never seen, for example, IL-1 (i.e. IL-1 knockout mice), may have developed compensatory mechanisms due to the redundancy in the cytokine network (see Leon, 2002).

The conventional fever hypothesis (and some potential flaws)

The current working hypothesis for the sequence of molecular events in fever is as follows: exposure of the body to an exogenous pyrogen induces immune responses which include the release of soluble mediators, in particular the endogenous pyrogenic cytokines TNF, IL-1 and IL-6, from systemic mononuclear phagocytes at the site of infection/inflammation. These cytokines are released into the circulation where they communicate with the brain either directly or indirectly at the CVOs, to induce the synthesis and release of PGE2 into the brain and that PGE2 acts on neurons in the preoptic-anterior hypothalamic thermoregulatory area with an outcome of elevation of temperature 'setpoint'.

The working hypothesis is flawed because we already know that it does not hold true for all fevers. We previously have drawn attention to two 'bypass mechanisms' (see Mitchell and Laburn, 1997): the first relates to Gram-positive organisms, the pyrogenic action of which can bypass the cytokine system (Riveau et al„ 1980; Goelst and Laburn, 1991b), and the second relates to the cytokine macrophage inflammatory protein-lp (MIP-1 (3), that can act in the brain during fever independent of PG synthesis (Davatelis et al., 1989; Minano et al., 1996). Recently, the existence has been proposed of a pre-formed pyrogenic factor, continuously present in macrophages and unrelated to IL-1, IL-6 or TNF, which can be released immediately after LPS stimulation, and which acts indirectly and independently of PG synthesis (Zampronio et al., 1994a). Other examples of fever that appear to be evoked independent of PGs are IL-8-induced fever (Zampronio et al., 1994b) and substance-P-mediated fever (Szelenyi et al., 1997).

A further challenge to the current working hypothesis relates to the postulate that peripheral cytokines signal the brain to evoke fever by the transport of the cytokines themselves, or their downstream mediators, in the circulation. That postulate is not self-evidently true. The cytokines themselves are large, hydrophilic and unlikely to penetrate the blood-brain barrier easily (Rapoport, 1976), to access the relevant thermoregulatory structures. There is a discrepancy in the time between first detection of cytokines in the blood and onset of the febrile response (see Blatteis and Sehic, 1997). Provoked by localised inflammation, animals develop rapid fevers in the absence of concomitant elevations in detectable circulating cytokine concentrations. Given the problems of access, there is a view that circulating exogenous pyrogens themselves do not pass the blood-brain barrier in sufficient amounts to exert direct actions on neurons or glial cells. Rather, the cytokines act at targets outside the blood-brain barrier to induce signals that subsequently are relayed to neuronal circuits in the brain (see Kluger, 1991; Zeisberger, 1999; Rivest et al„ 2000). There is evidence, however, that circulating cytokines indeed are transported actively into the brain, by cytokine-specific carriers (see Banks et al., 1995). Although the proportion of circulating cytokines that this system transports is less than 0.3% of the cytokines present in the blood, in chronic infection and/or inflammation, in which circulating levels of endogenous cytokines are elevated for prolonged periods (in some instances, weeks), this small proportion may be sufficient to account for the sustained fever. Also, transport of circulating cytokines can occur via the paracellular route (passive diffusion) when blood-brain barrier integrity is compromised, such as occurs in CNS diseases or in response to large doses of LPS (Tunkel et al., 1991; De Vries et al., 1996), but the initial effects of cytokines or LPS, administered peripherally, can be observed more quickly, and at lower doses, for damage to the blood-brain barrier to be the usual route of access. Alternative humoral pathways include cytokine message transfer via 'leaky' areas, lacking a tight blood-brain barrier, that is the so-called sensory CVOs, and in particular the OVLT on the midline of the POA, and the subfornical organ (SFO) (see Stitt, 1990; Blatteis and Sehic, 1997; Zeisberger, 1999). At these sites, circulating cytokines might enter the perivascular space and interact with receptors located at terminals of glial cells. Circumventricular organs express CD 14, TLR2 and TLR4 (Laflamme and Rivest, 2001), implying that they can bind and respond directly to bacterial fragments. However, cytokines do not have to gain access to the brain tissue to signal thermoregulatory neurons. Cytokines (and even exogenous pyrogens) can bind to receptors expressed on endothelial cells of the cerebral microvasculature, and on meningeal macrophages, and can signal the thermoregulatory neurons through the synthesis and release of second messengers, such as NO and PGs (Dinarello et al., 1999; see for review, Rivest, 1999; Rivest et al., 2000; Engblom et al., 2002), which can penetrate into brain tissue. This activation of cerebral vascular cells seems to precede the activation of deep neural structures, indicating a role for these cells as an intermediate step between the circulating cytokines and the neural elements (Herkenham et al., 1998).

Communication between peripheral sites of infection and the brain may not require circulating cytokines or other mediators. Rather, the vagal nerve may convey communication between pyrogen-sensitive cells and the brain (Dantzer, 1994; Watkins et al., 1995). The evidence for such a pathway has burgeoned over the last five years (see Zeisberger, 1999; Autonomic Neuroscience 85: 1-55, 2000). The proposed peripheral mechanism is complement-induced cytokine or PGE2 synthesis from Kupffer cells, with IL-1 or PGE2 then exciting afferent fibres of the hepatic branch of the vagus nerve via their specific receptors. The vagal traffic might then be transported to the central projection areas of the vagus nerve within the nucleus tractus solitarius, and passed on to noradrenergic A1/A2 cell groups, which are located in this brainstem area and which project to the POA via the ventral noradrenergic bundle. Blatteis and colleagues (1997; Li et al., 1999b) have supported the idea of vagal signalling. They doubt that circulating endogenous cytokines could account for the fever, at least in response to i.v. LPS in guinea pigs, because fever onset precedes the appearance of sufficiently high concentrations of circulating cytokines, and have proposed that complement fragments, induced almost immediately (within seconds) within Kupffer liver cells (the sessile macrophages), in response to LPS, stimulate PGE2 synthesis in macrophages. This PGE2 might be transported in the circulation to the brain, but could also activate local hepatic vagal afferents (see Li et al., 1999b). Although vagal afferents may play a role in fever, especially if pyrogens are present at low doses in the abdomen rather than in the blood or elsewhere in the periphery (Bluthe et al., 1996; Romanovsky et al., 1997), the importance of this route, and the question of whether or not it has a special role in abdominal pyrogen assault, remains controversial (see Autonomic Neuroscience 85: 1-55, 2000). Vagal signalling certainly cannot be advanced as a general mechanism by which pro-inflammatory cytokines trigger the cascade of thermoregulatory events taking place in response to systemic pyrogenic challenge (Rivest et al., 2000). In a view similar to that advanced for vagal signalling, modest evidence has been provided recently for the participation of cutaneous afferents in the transport of immune information from the skin to the brain, in the genesis of fever (Ross et al., 2000).

The conventional fever hypothesis applies only to systemic pyrogen challenge and fever genesis. A different series of events occurs when an exogenous pyrogen gains access to the brain (e.g. in bacterial meningitis). Experimentally, the latency of fever induction and the duration of fever is longer, and the fever is less pronounced, if an exogenous pyrogen is administered i.e.v. rather than i.v. (see Coceani and Akarsu, 1998). This counter-intuitive observation, and indeed the molecular mechanism by which fever develops when exogenous pyrogens access the brain, remain unexplained.

Endogenous antipyretics

'Heat is the immortal substance of life endowed with intelligence... However, heat must also be refrigerated by respiration and kept within bounds if the source or principle of life is to persist; for if refrigeration is not provided, the heat will consume itself.'

( Hippocrates) Cited in Mackowiak and Boulanl, 1996

Whatever the stimulus, a febrile episode continues until the 'setpoint' has been restored to the normal level. While much is now known about the mediators and the mechanisms initiating fever, as we have described above, how fevers are maintained and limited, and what mechanisms underlie the magnitude and duration of fever has been largely neglected by fever researchers. Several endogenous antipyretic systems, physiologically active during fever, are proposed to play a role not only in defining the empirical upper limit of fever, but also in determining the normal course of fever and fine tuning the febrile response (see Mackowiak et al., 1997). It may require ongoing interplay between the endogenous pyrogenic factors and antipyretic factors to determine thermoregulatory setpoints, thermo-effector responses and the course of the fever (Tatro, 2000). Whether these various systems act in parallel, serially or each plays an independent role, and the relative extent of each system's influence, in different types of fever and at different phases of the febrile response, remain yet to be determined.

Vasopressin, melanocortin and the glucocorticoids, released as a result of activation of the HPA axis, are the major known endogenous antipyretics (substances that lower body temperature only when fever is present). The neuropeptide arginine vasopressin (AVP) was the first peptide proposed a candidate antipyretic for the following reasons: circulating concentrations of AVP are increased in pregnant animals at a time that they are unresponsive to fever-inducing agents (Cooper et al., 1979), AVP present in the fibres and terminals of the ventral septal area is released into the ventral septal area during fever and fever is prolonged when its action is inhibited (see Pittman and Wilkinson, 1992). Also, microinjection of AVP into the ventral septal area attenuates fever induced by various agents in several species, by its action at type I vasopressin receptors (see for review, Kasting, 1989). Arginine vasopressin has been presumed to influence events in the pyrogen-signalling cascade that lie downstream of the earliest steps activated by LPS and cytokines (see Tatro, 2000). The pro-opiomelanocortin-derived hormones, ACTH, oc-melanocyte-stimulating hormone (a-MSH, which shares the first 1-13 amino acid sequence of ACTH) and y-MSH inhibit fevers produced by a variety of pyrogenic stimuli and in a number of different species (see Catania and Lipton, 1993), independently of adrenal glucocorticoids (Lipton et al., 1981; Huang et al., 1998). Their site of action is proposed to be a central one, since the effective central dose required to evoke antipyresis is lower than that required systemically. Importantly, repeated central administration of ot-MSH does not induce tolerance to its antipyretic effect (Deeter et al., 1989) and central administration of antiserum to a-MSH augments the fever response (Shih et al., 1986). The biochemical pathway through which a-MSH exerts its effect is not known, although it does not inhibit PG synthesis and does not act as a receptor antagonist of IL-1 (see Lipton, 1990). Overall, the properties of a-MSH make it a stronger candidate, for a role as an endogenous peptide antipyretic, than AVP is (Mitchell and Goelst, 1994).

Adrenal glucocorticoids, released in response to cytokine stimulation of the CNS, are potently effective antipyretics and anti-inflammatories. Although there is evidence that the antipyretic actions of these molecules are mediated directly in the CNS at the level of the anterior hypothalamus (Chowers et al., 1968; Morrow et al., 1996), probably due to suppression of PG synthesis, glucocorticoids also act at peripheral targets to suppress local PG, cytokine (IL-1 and IL-6) and kinin production and action and to inhibit local tissue damage and inflammation. Corticoid suppression of pro-inflammatory cytokine production is thought to occur via a direct genomic effect, stimulating the transcription of IicBa, interfering with the potential NFkB binding to DNA-response elements, or indirectly by induction of NFKB-binding protein (NFkI) (see Turnbull and Rivier, 1999; Imasato et al., 2002). Glucocorticoid effects on fever suppression may also be due to inhibition of CRF synthesis, since fever induced by central administration of IL-8 or PGF2a (both of which rely on CRF action, but not PG synthesis, for the induction of fever) is almost entirely abolished by administration of glucocorticoids, whereas fever induced by central administration of PGE2 (which is independent of CRF release) remains unaffected (see Rothwell, 1990, 1991). Lipocortin or annexin, a glucocorticoid-inducible protein, acts also as an endogenous antipyretic agent and inhibits the pyrogenic actions of IL-ip, IL-6, IL-8, PGF and IFN, which cause fever via CRF release, but has no effect on the pyrogenic actions of IL-la and TNF, which are PG dependent (Carey et al, 1990; Strijbos et al., 1992; see Rothwell and Hopkins, 1995). Moreover, anti-lipocortin prevents the antipyretic actions of exogenous lipocortin and reverses the antipyretic effects of glucocorticoids (Carey et al, 1990; Strijbos et al, 1993). Worth mentioning is that glucocorticoids upregulate the production (by lymphocytes but not monocytes) of the potent anti-inflammatory and antipyretic cytokine (see below), IL-10 (Chrousos, 2000).

The anti-inflammatory cytokines, which oppose or downregulate inflammatory processes (see Dinarello, 1997), potentially could behave as endogenous antipyretics. Based on current evidence, IL-lra, which, like IL-1, is produced endogenously in response to inflammatory stimuli, and can act both peripherally and in the brain to attenuate fever without having any effect on 'afebrile' body temperature, is considered not to behave as an endogenous 'physiological' antipyretic. Interleukin-10, however, produced by Th2 lymphocytes and monocytes, may well function as an endogenous antipyretic: IL-10 potently inhibits the production of TNF-oc, IL-1 p, IL-6 and IL-8, and upregulates the expression of IL-lra (see Dinarello, 1997), elevated plasma IL-10 concentrations are reported in patients with sepsis and after the injection of LPS into experimental animals (Durez et al, 1993; Marchant et al, 1994b), administration of IL-10 protects mice from lethal endotoxaemia by reducing TNF release (Howard et al, 1993; Marchant et al, 1994a), neutralisation of endogenous IL-10 in endo-toxaemic mice results in an increased production of several pro-inflammatory cytokines and enhanced mortality (Standiford et al, 1995), IL-10 knockout mice have an increased likelihood of inflammatory bowel disease (Rennick et al, 1997), have higher mortality rates after experimentally induced sepsis (Berg et al, 1995), and develop an exacerbated and prolonged fever in response to i.p. LPS, but not to localised turpentine injection (Leon et al, 1999). Finally, we have shown that neutralising antisera to IL-10, administered at the site of inflammation, exacerbates the magnitude of Gram-negative and Gram-positive fever and profoundly prolong the duration of LPS-evoked fever (from 8 h to 72 h, Cartmell et al, 2003). Although some evidence is compatible with IL-10 acting within the CNS (Frei et al, 1994), in vivo the antipyretic action of IL-10 is more likely to derive from IL-10 acting at the site of inflammation, whether that is in the periphery or in the brain, probably via inhibition of local IL-1 production (Ledeboer et al, 2002). Central administration of exogenous IL-10 attenuates centrally administered LPS-evoked fever and IL-1 production, but IL-10 administered systemically has no effect on the fever or peripheral IL-1 production. Likewise, systemic administration of exogenous IL-10 attenuates systemically administered LPS-evoked fever and IL-1 production, but IL-10 administered centrally has no effect (Ledeboer et al, 2002). There are two schools of thought as to the role of TNF in fever, which have been discussed in detail earlier (see Experimental models of fever: assembling the pieces): those who favour an antipyretic role of TNF during LPS fever suggest it is mediated by endogenous IL-10.

In addition to the three classes of antipyretics mentioned above, a class of lipid compounds known as epoxyeicosanoids, derived from activation of the arachidonic acid cascade, not by COX, but by cytochrome P-450 mono-oxygenase enzymes, also may be candidate endogenous antipyretics (reviewed in Kozak et al, 2000). Cytochrome P-450 is detected in the rat medial POA (Hagihara et al, 1990) and although metabolites of cytochrome P-450 do not contribute to maintenance of resting body temperature (Nakashima et al, 1996), inhibitors of cytochrome P-450 augment the defervescence of fever (Nakashima et al, 1996; see Kozak et al, 2000). The antipyretic effect of epoxyeicosanoids is linked to negative regulation of the synthesis of IL-6 (Kozak et al, 1998a), demonstrating that prostanoids and epoxyeicosanoids play contrasting roles in the regulation of IL-6 production (see Kozak et al, 2000). Thus, endogenous pyrogens, in addition to activating the cyclooxygenase branch of a fatty acid cascade during fever genesis, also activate the epoxygenase branch, allowing synthesis of antipyretic metabolites of the cytochrome P-450 system. Based on their own data, as well as those from others, Kozak and colleagues (2000) have speculated that modulation of the P-450 pathway has much broader implications than simply the production of antipyresis, namely the modulation of inflammation.

Concluding remarks

Investigation into the molecular basis of fever has flourished with the increasingly ready availability of sophisticated molecular tools and the recognition of the importance of integrative research between multiple scientific disciplines. This discipline continues to undergo explosive growth particularly since the discovery of the structure of many pro-inflammatory and anti-inflammatory cytokines, and the revelation that certain other hormones (e.g. leptin) are members of the cytokine family. Whether the beneficial effects of fever prevail over harmful effects to the infected host still is unknown: this has obvious clinical significance particularly with regards to treatment of the febrile patient, for example, whether it would be preferable to modulate the potentially harmful and discomforting effects of infection without necessarily interfering with the fever. Renewed integrative investigations into the ongoing interplay between the endogenous pyretic and antipyretic mediators that determine thermoregulatory setpoints, thermoeffector responses and the course of the fever will lead to important advances both in the scientific and in the clinical contexts.



interleukin-18 receptor accessory

Was this article helpful?

0 0

Post a comment