The generation of a maternal immune or inflammatory response during pregnancy leads to a range of changes in the brains of the offspring which have strong parallels with those found in human behavioural disorders such as autism and schizophrenia. The administration of either bacterial lipopolysaccharides or the viral mimetic poly(I:C), (compounds frequently used in models of prenatal infection) can induce electrophysiological alterations in synaptic transmission[20–22] and elements of cognitive function[21, 23, 24] as well as changes in the levels of some major neurotransmitters. Here we have used poly(I:C) to activate an immune response via TLR-3 in the late stages of gestation, a time when susceptibility of the rat foetal brain is believed to be comparable with that of the human brain in the second trimester of pregnancy, since brain development is at a comparable stage.
The measurement of cytokine and chemokine levels here confirms the ability of poly(I:C) to activate the innate immune system since the expression of MCP-1 is increased 5 h after its injection. The lack of any change in IL-1β or TNF-α levels confirms the absence of any on-going, background immune system activity, as in vivo administration of poly(I:C) is known to induce a rapid and transient increase in plasma TNF-α, peaking around 2–3 h post-injection, with levels reportedly returned to control values by 4–6 h[13–15]. In addition, Gilmore and co-workers have previously shown that administration of Poly(I:C) to pregnant rats at E16 significantly elevated TNF-α in the maternal plasma at 2 h post-injection, with no increase detectable by 8 h. Poly(I:C) administered to mice at 12 mg/kg has previously been reported to induce plasma IL-1β 3 h post-treatment, whereas, in agreement with our finding, other groups have failed to detect increases in this cytokine after poly(I:C) administration in vivo[14, 15]. Examination of mRNA for cytokines after treatment with poly(I:C) also suggests a transient profile, with peak up-regulation of genes occurring 3-6 h after poly(I:C), with only a few genes remaining elevated 24 h post-injection.
On the other hand, it should be emphasised that these comments apply to maternal blood since brains from the embryos were not examined in this study. Thus, we cannot exclude the possibility that changes occurred in the foetal brain that are not mirrored in the maternal circulation, although we are not aware of any previous evidence for this. However, we can exclude the initiation of an inflammatory response that persists from the administration of poly(I:C) through to the P21 stage when the offspring were examined. This is based partly on published data showing that the elevation of cytokines does not persist much beyond 24hours after injection. In addition, we have noted that there is no change of inflammatory markers such as cyclo-oxygenase-2 or NFkB at P21 as would be the case if there existed a continuing activation of the immune system (Pisar, Forrest, Omari, Darlington and Stone, unpublished observations).
It is clear from the present study that a number of significant changes are induced by the inflammatory challenge. The synaptic vesicle proteins synaptophysin, synaptotagmin and VAMP-1 (synaptobrevin), are of particular interest in brain development, and VAMP-1 has been noted as one of 18 proteins whose gene increases continually from postnatal day P0 to P45. Clear changes in several synaptic vesicle release proteins have been demonstrated in brain tissue from schizophrenic patients[28–30] where the most marked changes are in the levels of synaptophysin and VAMP-1. Treatment of pregnant dams in the present study did not produce any significant change in the expression of synaptophysin or the synaptic vesicle calcium sensor protein synaptotagmin. This result is consistent with a study by Afadlal et al. in which prenatal stress induced by maternal restraint or corticosterone injections produced a decrease in synaptophysin levels only in animals at P7 or P14, but not at later ages such as the P21, post-weaning time examined here. On the other hand the levels of VAMP-1 were affected to a significant degree, emphasising the potential parallels between the effects of poly(I:C) and the changes reported in schizophrenia.
An increasingly recognized view of the molecular changes underlying schizophrenia postulates a hypofunction of NMDA receptors. It is already established that neuronal receptors for glutamate, especially those sensitive to NMDA, affect neuronal migration, synapse formation, and neurite growth, spine formation[35, 36] and neuronal plasticity[6, 37, 38]. The pharmacological blockade of NMDARs in neonatal rats causes a loss and disruption of synapses with profound abnormalities of brain structure and behaviour in adulthood[39–41], many of which resemble the behavioural abnormalities which occur in schizophrenia. Thus, an induced change in NMDA receptor function during development could modify (for example by accelerating or slowing, amplifying or suppressing) the normal processes of programmed neuronal elimination which shapes the early nervous system.
In light of this fundamental role of NMDAR in neuron and synapse formation and function, the present finding of substantially reduced expression of the major subunit of these receptors could imply that a change in NMDAR number or function could underlie some of the neurochemical and behavioural abnormalities which have been described after exposure of animals to LPS or poly(I:C)[21, 23, 24]. NMDARs are composed of GluN1 and GluN2 subunits which bind glutamate and exert a regulatory control over GluN1. The GluN1 subunit is encoded by one gene to produce several splice variants. The antibody used in this work was selected as it interacts with all of the known splice variants. The change in GluN1 expression, therefore, reflects a real overall change of NMDAR number or function. Changes similar to those observed here have been reported previously in the expression of GluN1 receptors, although the experimental design was substantially different, with LPS injections or viral infection being induced directly in the pups themselves during postnatal development, and not during gestation[40, 43]. Nevertheless, taken together, these various studies offer highly suggestive mutual support for the concept that activation of the immune system can interfere with NMDAR function during brain development.
The small GTPases RhoA and RhoB have repeatedly been linked with synaptic plasticity[16, 17, 44] and various aspects of neuronal development[45, 46], often mediating the effects of NMDAR. However, there were no clearly significant changes in their expression. Similarly, despite the role of the tyrosine kinase family of Eph receptors and their ephrin ligands in synaptogenesis and spine formation[48–50] the absence of any clear change in the level of EphA4 implies that it does not contribute to changes in neuronal development and behaviour produced by poly(I:C).
There was a significant increase in the expression of Sox-2 (sex-determining region Y-related HMG-box2), a transcription factor involved in stem cell maintenance in the brain as well as a major factor in regulating cell loss or survival during development. Sox-2 is often considered a marker of cells that are actively dividing but which are in the early phases of proliferation and are as yet undifferentiated. Sox-2 occurs primarily in brain regions that contain actively proliferating neural stem cells, possibly maintaining a ‘latent’ and undifferentiated state[53, 54]. Suppression or deletion of sox-2 can lead to increased apoptosis and is incompatible with embryonic survival, although overexpression does not promote proliferation. The pattern of changes seen in this group of neurogenesis markers, therefore, suggests a possible increase in the number of cells (as yet undifferentiated) at an early stage of proliferation. However, since no change was detected in the expression of PCNA, a nuclear protein which is increasingly used as a marker for the early phases of cell degeneration, the sox-2 expressing cells either do not express this protein at P21, or they have past the stage of producing PCNA. Similarly there was no change in levels of doublecortin, a molecule associated primarily with later stages of neurogenesis and whose expression declines during brain development. This observation suggests that poly(I:C) administration does not produce significant overall changes in neuronal or glial generation at P21.
A major question which arises following the induction of inflammation by poly(I:C) or bacterial polysaccharides (LPS) is whether any observed effects are a direct action of the injected agent, or are secondary, indirect consequences mediated via an intermediate compound. It is improbable that the immunostimulator molecules are themselves responsible, since they are unlikely to cross into embryos. In one study, radiolabelled LPS was administered to pregnant dams but the labelled compound was detected only in maternal tissues and not in the embryos.
The compounds that have been most often considered in the role of mediating immunostimulant effects in the brain are the cytokines. It is well recognised that both poly(I:C) and LPS can induce the expression and release of several pro-inflammatory cytokines although they can also modify the expression of neurotrophic proteins such as brain-derived growth factor (BDNF) and nerve growth factor (NGF), all of which have profound effects themselves on the CNS.
The administration of poly(I:C) does increase plasma levels of IL-1β, IL-6, TNF-α and interferon-β in adult rats when measured at the appropriate time points[12, 13] and both IL-1β and TNF-α have been detected within the brain after systemic administration of LPS or poly(I:C)[58–60]. This expression of cytokines has also been shown after maternal administration of LPS, when expression of IL-1β, IL-6 and TNF-α, as well as MCP-1 was demonstrated within the brains of the embryos. These results seem to depend acutely on the precise experimental protocol since others have found that the expression of TNF-α was reduced by LPS in the foetal brain, whereas poly(I:C) had no effect. Other workers have reported finding either no change of foetal brain TNF-α or an increase.
Some of the most compelling work is that which shows the ability of cytokine antagonists to prevent CNS features of inflammation such as microglial activation, neuronal death or cognitive dysfunction. Some of these effects can be prevented by an IL-1β receptor antagonist[58, 64], although the direct administration of IL-1β to pregnant dams has been reported not to result in changes in the offspring.
Certainly, both neurons and glial cells possess interleukin receptors and there have been many studies showing that a range of cytokines, including IL-1β, have marked effects in the CNS even after peripheral delivery. Work by Riazi et al.,[59, 60] has shown that poly(I:C) can increase the CNS levels of IL-1β, the cytokine which was probably responsible for the observed increase in CNS excitability and susceptibility to seizures induced by pilocarpine or pentylenetetrazol. Interestingly these effects were accompanied by an increased expression of GluN2A and GluN2C subunits which did not occur in the present study, presumably because of the difference between direct administration into the postnatal brain rather than administration into the mother followed by examination of the offspring.
Although most attention on cytokine mediation of poly(I:C) effects has been concentrated on IL-1β, there are several alternative candidates, such as IL-6. When labelled IL-6 was injected into pregnant rats, the protein was found in the amniotic fluid and in the foetus itself showing its ability to pass across the placental barrier. In addition, maternal IL-6 administration has been reported to reproduce many of the effects of poly(I:C) on offspring, a claim strongly supported by the finding that antagonism of IL-6 prevented the effects of poly(I:C)[9, 10].
The problem of identifying a likely mediator of poly(I:C) effects is complicated further by the fact that neurons and glia can express several cytokines including IL-1β and TNF-α, so that the passage of cytokines into the CNS may not even be required for them to mediate genetic changes within the brain. It may only be necessary that neurons or glia are activated by a small molecule factor which can pass readily between blood and brain. However, Oskvig et al. have noted that cytokine protein levels are increased within the foetal brain whereas the corresponding RNA molecules are not, implying that the foetal proteins are derived from the maternal circulation and not from foetal synthesis. There is evidence that poly(I:C) can increase the permeability of barriers and it may therefore contribute indirectly to an increased movement of cytokines into the foetus.
Overall, the profile of molecular changes described here suggests that, after exposure in utero to the viral mimetic poly(I:C), rat pups exhibit abnormal levels of several proteins known to be important in cerebral development. Even though several of the proteins studied show no significant change, it is important to realise that the absence of a change may be as scientifically important as the presence of a change. Firstly, it is valuable evidence that the schedule of immune activation used in the study does not produce a non-selective toxicity, with a generalised loss of cells and proteins. Secondly, when development of the brain can be understood in terms of the full gamut of molecular changes of expression it is likely to be just as important to know which aspects of neuronal development are unaffected by inflammation as to know which are. In addition, it is highly likely that for many aspects of brain development it is the relative change between molecules that may be the determining factor in abnormality, especially if different molecular changes occur at different rates during the course of development. Here again therefore it will be essential to know which proteins change in response to insult and which do not.
Finally, we elected to examine the overall protein expression in cerebral hemispheres. The apparent absence of a change may, therefore, mask more localised, significant changes, which need to be investigated in more detail.