Suppression of the Febrile Response in Late Gestation: Evidence, Mechanisms and Outcomes

Abstract

Fever is a beneficial host defence response. However, fever caused by the immune stimulant, lipopolysaccharide (LPS), are attenuated in many species during pregnancy, particularly near term. A number of parallel mechanisms may be responsible, and these vary in magnitude according to the time of gestation, type of inflammatory stimulus and species of animal. Some studies report a reduction in the plasma levels of circulating pro-inflammatory cytokines such as tumour necrosis factor-α, interleukin-1β and interleukin-6 along with increased levels of anti-inflammatory cytokines such as interleukin-1 receptor antagonist. Associated with the attenuated febrile response to LPS is a reduction in the activation of the prostaglandin synthesising enzyme, cyclo-oxygenase 2, resulting in reduced levels of the obligatory prostaglandin mediators of the febrile response in the brain. There is also a reduction in the sensitivity of the brain to the pyrogenic action of prostaglandins, which does not appear to be due to a change in the levels of hypothalamic EP3 prostaglandin receptors. The suppression of fever at term may be important for the health of the neonate because fever in pregnant mothers may be harmful to the late-term foetus and neonate.

Fever: an overview

The response of an organism to an invading pathogen consists of a highly orchestrated series of events that deal with the current invasion and prepare the body to combat future incursions. These adaptations are carried out by two arms of the immune system: the acquired immune response and the innate immune response. In this review, we will only deal with innate immune response, which differs from the acquired immune response in that it does not require previous exposure to the antigen, along with the subsequent antibody response. The innate immune response allows the organism to recognise and respond to specific molecular patterns of bacteria, viruses, molds, etc., that could threaten the animals’ defences. Immune cells possess receptors that are known as Toll-like receptors (TLRs); each TLR subtype responds to a different pathogen. For example, TLR3 responds to viral double stranded RNA, whereas TLR4 recognises lipopolysaccharide (LPS) molecules in the cell walls of Gram-negative bacteria (1).

This innate response can be studied in a controlled manner; LPS is available commercially, enabling it to be administered to activate responses mimicking those of a Gram-negative bacterial infection (2). Similarly, polyinosinic:polycytidylic acid (poly I:C) is a synthetic compound that targets the TLR3 and thereby activates viral-like responses. Because of the availability of such compounds, we now have a relatively good understanding of the body’s innate immune response to infection, in particular that to LPS, which is the most commonly utilised molecule. Although the response of the body to activation of TLRs is multi-faceted, consisting of both behavioural and physiological responses, the most commonly monitored innate immune response is the febrile response and, for the purposes of this review, we will consider the febrile response as representative of the innate immune response.

The cascade of events that occurs subsequent to introduction of LPS to the body has been described in detail and comprehensively reviewed (3, 4). In brief, when animals are given LPS, cells of the immune system such as macrophages, monocytes and Kupffer cells, which express TLR4, recognise and bind LPS. In the presence of appropriate ancillary molecules (e.g. CD14, LPS binding protein), a cascade of intracellular signalling molecules is generated that ultimately activate transcription factors such as nuclear factor κB (NFκB) and extracellular related kinases. An early component of the response is the activation of the complement cascade which results in generation of the anaphylatoxin, C5a. This complement activates cyclo-oxygenase (COX)-1 and possibly COX-2 that initiate rapid synthesis (and release) of prostaglandins (PGs). These prostaglandins either make their way through the circulation to quickly activate the brain or possibly act on prostaglandin receptors on afferent nerves, in particular the vagus. On a slightly slower time scale (approximately 30 min or more), cells of the innate immune system synthesise and release a variety of pro- and anti-inflammatory cytokines. These include pro-inflammatory cytokines such as interleukin (IL)-1β, IL-6 and tumour necrosis factor (TNF)-α, as well as anti-inflammatory cytokines such as IL-1 receptor antagonist (IL-1ra) and IL-10. The circulating cytokines bind to cytokine-specific receptors on brain endothelial cells, circumventricular and perivascular glial cells, and possibly on afferent cranial and spinal nerves. Activation of the cytokine receptors stimulates signalling pathways utilising transcription factors such as NFκB in the case of IL-1β, or the Janus activated kinase-signal transduction and activator of transcription-3 (JAK-STAT-3) pathway in the case of IL-6. Within the perivascular endothelial cells, inducible COX-2 is activated, resulting in the synthesis of prostaglandins that can be either pro-inflammatory in nature (e.g. PGE₂) or anti-inflammatory (e.g. PGJ₂). PGE₂ binding to EP3 receptors on GABAergic neurones in the anterior hypothalamic and pre-optic area (AH/POA) and organum vasculosum of the lamina terminalis (OVLT) is critical for the febrile response. PGE₂ inhibits these GABAergic inhibitory neurones and the resulting disinhibition allows projection neurones in the dorsomedial hypothalamus to activate downstream pathways, ultimately leading to brown adipose tissue thermogenesis and vasoconstriction (5). In addition to prostaglandins, there is additional synthesis within the brain of cytokines, chemokines and other transmitter molecules that also appear to contribute to various aspects of the fever response (6).

Although this review is most closely focused upon the febrile response, it is important to remember that exposure to LPS and activation of TLR4 causes a number of other responses important in the host defence response. These responses include activation of the hypothalamic-pituitary-adrenal (HPA) axis, sensitisation of sensory afferents leading to hyperalgesia, anorexia, and a type of sickness behaviour that is manifested as a lack of interest in social interactions (7, 8). Although such effects all appear to involve the same peripheral cytokines as the febrile response, there is some evidence that the central pathways involved are different. For example, brain IL-1β is required for the behavioural effects of peripheral cytokines, but not for their thermogenic (9) or anorexic (10) effects. On the other hand, central IL-6 can cause hypothalamic pituitary activation and fever but not sickness behaviour (11). The specific pathways and targets for these cytokines are not yet well described.

Attenuated fever in late pregnancy and early lactation

Fever, is now known to be important in helping the body fight infection. There is now good evidence that a lack of a febrile response is associated with increased morbidity and mortality (12) as the elevated body temperature provides a less hospitable environment for pathogens to thrive and reproduce. Given the importance of fever in host defence, it is of considerable interest that the febrile response may be naturally suppressed. This was first observed some 30 years ago when it was reported that sheep at the time of parturition do not mount a normal febrile response to LPS (13). The finding of an attenuated febrile response to LPS in sheep during late pregnancy and early lactation has since been replicated in one study (14) but not in another (15). However, it has also been reported in guinea pigs (16) and rats (17–19) (Fig. 1A, B) but, curiously, not in rabbits (20). The febrile response in rats is also reduced when activated by IL-1β (21), as well as by a localised peripheral inflammation induced by intramuscular turpentine (22) (Fig. 1C). There may be two components to this period of ‘endogenous antipyresis’; there is a very short, almost absolute suppression of the febrile response appearing 1–3 days before parturition and returning within hours after birth. In addition, there is a general reduction in the febrile response that occurs throughout the entire gestational and lactation period and that may be associated with a generalised reduction of central nervous system responses to both immune and psychological stressors (23, 24).

Fig. 1.

Fever is reduced near term in rats. (A) Magnitude of the febrile response to lipopolysaccharide (LPS) (Escherichia coli, 25 μg/kg, i.p) in pregnant and lactating rats in the hours before and after parturition. Reproduced with permission (18). (B) Body temperature responses in nonpregnant or day 18 pregnant female rats in response to saline or LPS (E. coli 50 μg/kg i.p.) given at time = 0. Reproduced with permission (19). (C) Body temperature response of nonpregnant (NP) or gestational day 18 (GD18) pregnant rats in response to i.m. injection of turpentine (TURP) or saline (SAL) at time = 0. Insert shows the fever indices for the data from the graph. Reproduced with permission (22). *P < 0.05; **P < 0.01.

Peripheral signalling molecules in pregnancy

To determine whether a reduction in pyrogenic cytokines or an increase in anti-inflammatory cytokines underlie the attenuated febrile responses, we measured circulating levels of the pyrogenic cytokines, IL-1β, IL-6, TNF-α and interferon-γ 2 h after LPS was given on gestational day (G)15, G22 and lactation day 5 (25). We were unable to observe any convincing reductions in any of these cytokines that could account for the reduced fevers. Likewise, when we measured levels of the anti-inflammatory cytokines IL-1ra and IL-10 in response to LPS, levels were similar at all time points of gestation. Interestingly, other studies have observed an almost complete suppression of LPS induced IL-1β and IL-6 at parturition (26). In a different model of a brief localised inflammatory response due to intra-muscular turpentine, Aguilar-Valles et al. (22) reported a significant reduction at G18, not only of fever, but also of circulating IL-6. Increased levels of IL-1ra in late pregnancy have also been reported both in rats (17, 19) and humans (27), which could account for both the reduced fevers and COX-2 levels that we observed. Most likely, different laboratory studies report different levels of cytokines because some contrast the virgin female with the pregnant, whereas others compare mid- and late pregnancy. In comparing the latter states, the emphasis is on the specific alterations that lead to the suppression of fever at term. In the case of comparisons between the virgin and pregnant, these changes could explain the generalised suppression of many stress responses during pregnancy. In speculating why different types of inflammation (systemic versus local) are associated with different cytokine profiles during gestation, Luheshi suggests that the magnitude of a systemic inflammatory challenge may overwhelm the system, obscuring differences that may be evident in the more localised inflammatory model (22). As one potential influence upon cytokine generation in the plasma is corticosterone, it is possible that, under different conditions, pregnant rats may respond with HPA activation in a different manner to release different quantities of corticosterone. However, all the evidence suggests that corticosterone secretion after IL-1β is reduced at term (28) and our experiments did not implicate it either (25).

Pregnancy and lactation are associated with major changes in circulating levels of reproductive steroids. Throughout pregnancy, there is a general elevation of progesterone that then falls precipitously just before birth. By contrast, oestrogen, which increases steadily during gestation, peaks at parturition. In an effort to determine whether these hormones could contribute to the reduced febrile response and possibly to altered cytokine levels, we treated ovariectomised female rats with oestrogen and progesterone. Fever responses to LPS are reduced under these circumstances and this reduction was associated with reduced COX-2 expression in the hypothalamus (29). Plasma IL-1β levels were reduced but IL-6, which is the major circulating cytokine, was unchanged. Thus, some aspects of the febrile response in this endocrine state (chosen to mimic the pre-ovulation hormonal surge) resemble that of the rat at parturition. Although these experiments in which endocrine hormone levels are manipulated indicate a potential role of these hormones in regulating febrile responses, further studies may be required in which a hormone regimen more closely mimicking that of near term is imposed.

Given the uncertainty of the contribution of altered levels of peripheral cytokines to the reduced febrile response, we then considered the possibility that their actions on the brain would be attenuated. We investigated this by measuring the levels COX-2 in the AH/POA throughout pregnancy and parturition. We found that unstimulated levels were generally reduced during late pregnancy and lactation. However, of greater relevance to the suppression of the febrile response near parturition was the fact that LPS induction of COX-2 was suppressed at the time of parturition relative to earlier in pregnancy and at lactation day 5 (30). Similar findings were reported for COX-2 and other PGE synthesising enzymes such as microsomal PGE-1 synthase, which is downstream from COX-2 (22, 31). In keeping with the reduced induction of the PGE biosynthetic enzymes, there is a general reduction of the PGE levels in the OVLT and ventricular fluid after LPS near parturition (31, 32). It remains to be determined whether there are alterations in other prostaglandin synthesising enzymes that might also lead to generation of antipyretic prostaglandins. One potential antipyretic prostaglandin is 15-deoxy-delta 12,14 PGJ₂ (33), but the enzyme responsible for its synthesis is also lower at term (25).

The major circulating cytokine, IL-6, is known to cause phosphorylation and translocation of STAT-3 in the AH/POA and OVLT; therefore, we carried out another series of experiments to evaluate this signalling sequence throughout gestation and lactation. First of all, in a replication of our previous experiments, we observed similar circulating IL-6 levels throughout mid- and late gestation as well as in lactation after LPS. When we counted the numbers of cells exhibiting nuclear STAT-3, or the levels of phosphorylated STAT3 in the AH/POA and OVLT, there were no differences at the various stages of reproduction (34). However, with a turpentine model of peripheral, localised inflammation, there was attenuation of an IL-6 activated transcription factor (SOCS3) in the brain (22).

The conclusion from these studies is that the febrile process may be altered differently according to either the dose of the immune stimulant or the location of the inflammation (i.e. localised versus systemic).

Central nervous system changes in pregnancy

Although the data suggested reduced levels of prostaglandins in the brain, we (35) and others (36, 37) have also investigated whether activation of the EP receptors within the brain was equally effective in causing fever throughout late pregnancy and parturition. Body temperature responses to intraventricular PGE₂ were reduced at parturition. This was associated with reduced central drive to tissues important in febrile thermogenesis (Fig. 2). In particular, we showed that the temperature of interscapular brown adipose tissue, which provides the major source of thermogenesis in rat during fever, and the activity of sympathetic nerves leading to this tissue were both reduced at parturition. Interestingly, other PGE₂ elicited effects, such as increased heart rate and blood pressure, were also reduced at parturition. We considered that this might be due to a reduction in the level of EP₃ receptors in the hypothalamus, but protein levels of the receptor within the hypothalamus were unchanged (30). This of course does not rule out a change in levels of EP receptors elsewhere, or of a different subtype, but the action of PGE₂ at the critical EP₃ receptors (38) appears to be unchanged.

Fig. 2.

Prostaglandin fever is reduced near term in rats. (A) Representative rectal and interscapular brown adipose temperatures in response to i.c.v. injection (arrows) of PGE_2, (50 ng) on days 16–17 of pregnancy (P), near term at days 19–20 of pregnancy and at 1–2 days after parturition (L1–2).(B) Net rectal temperature changes at the various times in response to PGE₂ at time = 0. (C) Net interscapular temperature changes in response to PGE₂ at time = 0. Reproduced with permission (35).

It is possible that the reduction in PGE₂ fever is due to generation at parturition of a central antipyretic molecule that might either abrogate prostaglandin signalling or act downstream to prostaglandin- activated neurones to inhibit their thermogenic effects. There is considerable evidence in male rats that arginine vasopressin (AVP) is an endogenous antipyretic acting in the ventral septal area to suppress fever (39). Female rats, however, have much less AVP immunoreactivity than do males in relevant brain areas (40) and do not appear to depend to a great extent upon AVP to lower PGE₂ fevers (41). However, there is some evidence from pregnant guinea pigs that immunoreactivity for AVP is increased in late pregnancy (42), making it possible that there is increased synthesis and release of this peptide in parturient females. To test the possibility that parturient females recruit de novo AVP antipyresis, we measured AVP released into the ventral septal area of female rats in response to PGE₂; however, there was no increase, even in late pregnancy such as we saw in males (35). Thus, this did not support the hypothesis that pregnant female rats recruited AVP as an endogenous antipyretic at parturition. Nonetheless, ovarian hormones are known to alter the synthesis of many receptors in the brain and another possibility is that there might be de novo synthesis of AVP-V1-subtype receptors in the brain at parturition. However, we found that neither V1 receptor mRNA, nor protein were altered at parturition (43). Although this finding does not support a role for increased activation of AVP receptors at parturition, receptor levels in a sample of hypothalamic tissue do not give a precise indication of receptor activity or distribution. Thus, receptors may be more efficiently translocated to the plasma membrane at parturition or the coupling of the receptor to intracellular signalling pathways may be altered, as seen in some other experimental conditions (44). This may result in increased AVP ‘tone’. To obtain a better idea of possible endogenous antipyretic activity in the brain during fever at parturition, we applied a V₁-receptor antagonist into the cerebral ventricles during PGE₂ exposure at parturition. This would be expected to cause an elevation and prolongation of the febrile response if AVP action was somehow amplified due to increased numbers or efficiency of receptors. However, in our experiments, the AVP receptor antagonist did not restore fever to its normal magnitude, suggesting that parturient rats resemble virgin rats in using antipyretic strategies not requiring AVP to initiate antipyresis (35). Similar results have been reported elsewhere (45).

We still do not know the mechanism that underlies the dramatic reduction in the effects of PGE₂ in the brain around the time of parturition. There are many other molecules that can alter fever responses when introduced into the brain and further work will be required to determine if any of these are involved. Similarly, there are many additional changes that occur in the brain in response to sex hormones that could also affect the PG stimulated events during late pregnancy. For example, there is increased opioid inhibition that affects at least some of the ascending noradrenergic fibres that innervate the hypothalamus (46); recall that, in guinea pigs, ascending noradrenergic fibres are critical for the febrile response (3). In addition, the elevated levels of progesterone found in the circulation can be converted within the brain to a neuroactive steroid such as allopregnanolone, and the activity of the enzyme that does so is regulated by oestrogen. Neuroactive steroids have been known for many years to interact with GABA_A receptors (47) and some responses to IL-1β have been shown to be attenuated due to elevated neurosteroid action at GABA receptors (48). In addition, the subtype of the GABA receptor that is sensitive to neurosteroids undergoes changes in response to peripheral sex hormones (49). Given the involvement of GABAergic neurones in the response to prostaglandins (5), it is possible that fluctuations in the numbers of GABA receptors and their efficacy at term may play a role in the reduced prostaglandin responses observed. Admittedly, this is speculative, but amenable to experimental investigation.

Very recently, some compelling evidence was provided indicating that elevated brain NO arising from higher levels of its enzyme, NOS, accounts for at least part of the reduction in LPS fever in rats around parturition (50). NO has many effects in the brain and it is not known at the present time if it would also act to reduce a prostaglandin-initiated fever. There is evidence in guinea pigs that it inhibits the release of norepinephrine within the preoptic area, which, in this species, is an important neurotransmitter within pathways mediating the febrile response (51). As norepinephrine is thought to act upstream to generation of prostaglandins and activation of EP₃ receptors, this could be one way by which NO might inhibit fevers. Another possibility is that NO would directly attenuate COX-2 expression, although its effects on COX-2 expression vary according to the cell type and time frame studied (52).

Why is fever suppressed?

Whatever the reason for the reduced febrile response, the fact that it is seen in a number of species and may involve alterations at several levels of the neuroimmune axis suggests that it is an important part of the maternal physiology at term. The question that then arises is why should this be? Is there survival value, either for the mother or the newborn to have a suppressed febrile response? There are few data that have addressed this question but several possibilities come to mind.

First, there is the possibility that the suppression of fever is not that critical and what is really important is a reduction in the sickness behaviour that accompanies the fever. Sickness behaviour is characterised by, among other things, a withdrawal from social interactions (7). The time immediately prior to, and around birth is important for the establishment of maternal behaviour and, in some animals for maternal-infant bonding (53). It is possible that anything that interferes with this (i.e. sickness behaviour) might be detrimental to offspring survival. Although this is only speculative at present, there is certainly evidence that LPS given to lactating rats slightly impairs pup retrieval and affects the motivational aspects of nest building (54). Similarly, maternal aggression towards an intruder is attenuated in lactating mice given LPS at high doses (55).

A second consideration is that, when the pregnant mother develops a fever, the foetus does as well (56). As foetal temperature is higher than maternal temperature, the foetal hyperthermia secondary to the maternal fever will be greater than in the mother. There is some evidence (although controversial) that high maternal fevers early in pregnancy are detrimental to birth outcomes in humans (57), possibly because of the well known teratogenic effects of high temperatures (58). Intrapartum maternal fever has also been considered a predictor of neonatal morbidity and transient adverse effects (59, 60). As delivery is often associated with hypoxia of the foetal brain, and hypoxic/ischemic damage is greater at higher temperatures (61), it would likely be important not to subject the foetus to febrile temperatures around the time of parturition. Such speculation on the ‘beneficial effects’ of reduced fevers at near term in women are of course relevant only if the phenomenon can be demonstrated in humans, but this has not yet been carried out.

Perspectives

Progress has been slow in determining the cause(s) of the reduced febrile response at parturition. In Fig. 3 we present a schematic summarising the relevant changes that have been observed in the various laboratories investigating this phenomenon. There is certainly a reduced neuronal response to prostaglandins, the obligate mediators within the brain of the febrile response. Somewhere downstream to the activation of the EP₃ receptor the prostaglandin activated pathways are inhibited, or at least activated to a lesser extent. It is possible that an elevated level of brain NO is responsible for this, but the mechanisms by which this is brought about are yet unknown and are excellent targets for further investigation. With respect to the peripheral cytokine generation during inflammation, there is a bewildering spectrum of responses, with some labs reporting reduced levels of pro- and higher levels of anti-inflammatory cytokines, and other labs reporting no change. Possibly, levels of circulating cytokines are not representative of biological activity of the cytokines. More likely the differences arise from the fact that some studies compare near term to mid-pregnancy, whereas others compare near term to virgins. Further studies addressing the interaction of oestrogen and progesterone and the peripheral mediators of the febrile response are required, as well as an investigation of some of the other earlier mediators such as the complement system. Many such studies will be necessary before the fascinating case of ‘endogenous antipyresis’ at near term will be fully understood.

Fig. 3.

Summary diagram illustrating known changes in the components of the febrile response that are altered in late pregnancy. Solid arrows indicate stimulatory pathways and dashed lines indicate inhibitory pathways. Thickness of lines is indicative of the magnitude of response, with thinner lines indicating smaller responses. Lines of similar thickness in the nonpregnant and pregnant state indicate no change.