Chorioamnionitis, neuroinflammation, and injury: timing is key in the preterm ovine fetus

Antenatal infections (i.e., chorioamnionitis) are an important risk factor for preterm birth and a major contributor to neonatal morbidity and mortality [1, 2]. Intra-amniotic exposure to microorganisms and subsequent induction of inflammatory mediators in the amniotic cavity can initiate a fetal systemic immune response that is characterized by increased plasma interleukin (IL)-6 and IL-8 concentrations [3], and (persistent) changes in essential immunological organs including the fetal spleen and thymus [4, 5]. At the crosstalk between fetal peripheral blood and the brain (i.e., blood-brain barrier), this systemic inflammatory response can initiate a detrimental neuroinflammatory response which is primarily mediated by microglia and peripheral immune effector cells [6, 5]. This cerebral inflammatory response is a risk factor for preterm brain injury and concomitant adverse neurodevelopmental outcomes including cognitive, behavioral, and attentional impairments and motor dysfunctions (i.e., cerebral palsy) [7, 5].

In a pre-clinical chorioamnionitis model, we showed that short-term (2 days) intra-amniotic exposure to lipopolysaccharide (LPS) resulted in systemic inflammation, overt microgliosis, and changes in myelin basic protein (MBP) immunoreactivity (IR) in the fetal ovine brain [8]. However, this systemic and cerebral phenotype was substantially different following longer exposure time (7 days) indicating that time-dependent peripheral and cerebral changes occur following intra-amniotic inflammation. Moreover, we and others have shown that inflammation can modulate a second inflammatory stimulus through either preconditioning or sensitization of the fetal brain [9, 8, 10]. Taken together, this emphasizes that inflammation as pathogenic mediator for brain damage is not a single trigger within a short time frame but more a dynamic process over an extended period of time. Therefore, detailed studies elucidating the time-dependent effects of antenatal infection/inflammation in relation to neurological injury and development are crucial to gain insight in the pathophysiological changes in the fetal brain following antenatal stress. Importantly, such temporal insight in the induction of brain injury following antenatal stress is also essential to define the therapeutic window of opportunity for neurotherapeutics.

One of the most promising treatment options for preterm neonates at high risk for brain injury is erythropoietin (EPO), an important cytokine for brain development [11, 12]. Multiple experimental and clinical studies have demonstrated efficacy of EPO administration to prevent injury to the preterm brain, including severe periventricular leukomalacia, without adverse effects [13‐21]. In contrast, other clinical trials do not report improvement in neurodevelopment following EPO treatment [22, 23]. Effects of EPO are mediated by its receptor, which is abundantly present on (pre) oligodendrocytes, astrocytes, microglia, and neurons. EPO binding triggers phosphorylation of two monomers, which in turn phosphorylates and activates the signaling kinase Jak-2 facilitating effects including its anti-inflammatory, anti-oxidative, and anti-apoptotic properties [24, 25]. In addition, EPO enhances neuro- and oligodendrogenesis, oligodendrocyte maturation, and myelin production which are indispensable events in injury repair and normal neurodevelopment [26]. We hypothesize that changes in basal levels of EPO receptor activation in response to inflammation or perinatal stress might explain at least part of the differences in clinical outcomes following EPO treatment.

Considering the clinical need for understanding the time-dependent cerebral changes following intra-amniotic inflammation, we performed a detailed analysis of the temporal dynamics of intra-amniotic LPS-induced systemic and cerebral inflammation and subsequent fetal brain injury. In addition, to optimize EPO treatment in the clinical setting, we analyzed the temporal expression of the phosphorylated EPO receptor (pEPOR) in the course of intra-amniotic inflammation.

The main findings of this study are that intra-amniotic exposure to LPS results in (1) an acute onset and biphasic fetal systemic inflammatory response; (2) persisting microgliosis with a limited pro-inflammatory phase; (3) cell death and (pre) oligodendrocytes loss only following long exposure to inflammation (8 days +) but (4) a striking regional sensitivity to changes in neuronal architecture following short (hours) and long (days) LPS exposure, and (5) biphasic regulation of pEPOR expression. We demonstrate that fetuses, intra-amniotically exposed to LPS, develop biphasic opposing peaks in the systemic inflammatory response, illustrated by changes in IL-6 and IL-8 concentrations. This data agree with previous observations from this model on increased circulatory monocytes and neutrophils and subsequent lymphocytopenia [4] and agrees with previous clinical data showing increased levels of these cytokines, associated with adverse neurological outcomes [41, 9]. Furthermore, the systemic response to LPS is consistently associated with the production of a vast milieu of inflammatory cytokines including IL-6 [42]. These signals across the blood-brain barrier trigger a neuroinflammatory response including activation of microglia and astrocytes [6, 5]. Moreover, IL-8 is known as potent chemotactic and neutrophil-activating factor, and increased serum IL-8 levels are reported in newborns with MRI-defined cerebral abnormalities and abnormal neurodevelopmental outcomes [43, 44, 9, 45, 46]. In line, we observed an increased number of MPO+ cells within the cerebral white matter at 15 days after LPS exposure.

Our study also provides meaningful information about the cerebral inflammatory response initiated by intra-amniotic LPS exposure. We observed a rapid increase in the inflammatory mediator COX-2 and microglial marker IBA-1 in the preterm brain. COX-2 activity results in prostaglandin-E2 production which has wide ranging inflammatory actions on the brain [47], but is typically associated with a pro-inflammatory state in microglia [48] and also reported in astrocytes [49]. Weaver-Mikaere et al. demonstrated in a fetal ovine-derived mixed glial culture that COX-2 activation is the most important mechanism leading to inflammation-mediated white matter injury [50]. Blocking the COX-2 pathway has recently been shown to prevent hypomyelination and behavioral impairment in mice with neonatal white matter injury [49]. In this model, it has also been reported that there is also no change in GFAP-IR across the groups, but additional analysis revealed that despite this, astrocytes together with microglia were still an important source of oligodendrocyte-injurious COX-2 [49]. An additional injurious mechanism of increased COX-2 is that its primary product prostaglandinE2 stimulates glutamate release from astrocytes [51], which is suggested to be essential in the pathophysiological mechanism underlying neonatal encephalopathy [52]. Additionally, the dynamic temporal microglial response in our inflammatory model, indicated by increased IBA-1IR, is consistent with distinct phases of cerebral inflammation in response to other injurious factors in perinatal brain injury, including a hypoxic-ischemic or excitotoxic insult [9]. Microglial activation is proposed to be essential in cerebral injury and dysmaturation associated with exposure to maternal fetal infection/inflammation [53]. Of interest for the application of neurotherapeutics is that microglia activation based on the simple proxy of IBA-1IR was maintained for at least 8 days following LPS exposure, but increased COX2 as surrogate for a pro-inflammatory state was only elevated for 1 day. Further studies to determine a systemic surrogate of microglial activation state might re-invigorate the utility of previously discarded immunomodulatory neurotherapeutics if we understood when immunosuppression would be beneficial.

Considerable debate has raged in the field of preterm neuropathology regarding the role for cell death in encephalopathy of prematurity [54‐57]. Experimental data has shown that microglia activation results in pro-inflammatory cytokine release which in turn can lead to apoptosis. However, experimental data has also demonstrated that moderate activation of microglia leads only to the maturation arrest of oligodendrocytes and not cell death [58]. Our study supports a coherent integration of clinical and experimental data as it shows that with a clinically relevant exposure paradigm, the duration of exposure is essential; oligodendrocyte death only occurs after 8 days of LPS exposure. Although outside the scope of this study, research into additional variables including pathogen type and maternal/fetal health and genetics also undoubtedly play a role understanding cohort-specific observations on neuropathology. We also wish to highlight that the significant loss of (pre) oligodendrocytes was not accompanied by myelin loss in our study. It has been previously postulated that the type of robust reduction in myelin protein levels that we would be able to measure with our analysis technique only occurs later in the course of brain injury due to normal kinetics of myelin production [59]. Furthermore, in an inflammation-induced oligodendrocyte injury model, early analysis of myelin proteins [58] and genes [60] has indeed failed to observe reductions despite later robust hypomyelination and behavioral deficits. Detailed analysis of myelin structure could be considered to investigate early markers of myelin injury in future studies.

Together with disturbed white matter development, alterations in gray matter development are implicated in long-term neurological sequelae following intra-amniotic infections and preterm birth [39, 37, 38]. During normal gray matter development, a decrease in MAP-2IR occurs which is indicative for refining of dendritic branching and spines [40]. Regulation of spine development is part of the basic homeostatic functions of microglia, which also include supporting immature cortical neuronal survival and stimulating myelination [61]. As such, increase in MAP-2 IR that we found in the acute phase following LPS exposure suggests a delay in cortical development, a phenomenon that has also been suggested from analysis of preterm infants using MRI [62]. Moreover, in humans suffering from schizophrenia, increased dendritic arborization was found in the hippocampus compared to healthy humans when stained for MAP-2 [63]. Concerning the specific mechanistic link between increased inflammation and increased MAP-2, it is noteworthy that COX-2 overexpressing neuronal-like cells showed significantly increased neurite outgrowth and branching, which was partially reverted by COX inhibitor [64].

Despite multiple studies identifying EPO as very promising neuroprotective candidate in newborn infants [65, 66, 14, 67, 13, 17, 68‐70], other studies did not find therapeutic effects on neurodevelopmental outcomes [71, 72, 23]. These inconsistencies might be explained by heterogeneity between study cohorts including, as outlined in this study for the first time, differences in EPO signaling in response to inflammation. Interestingly, after an initial increase in the receptor phosphorylation levels, a distinct and long-lasting decrease of the pEPOR occurred. Activation of the EPOR phosphorylates activates the kinase Jak-2 resulting in initiation of a complex anti-apoptotic signaling cascade [73, 25]. Accordingly, we are unsurprised that a decrease in the level of pEPOR is accompanied by an increase in the number of apoptotic cells (caspase-3+) within the same region in the fetal brain. This suggests that the reduced activation of the EPO receptor in our study results in increased apoptosis in the preterm brain and in experimental paradigms this relationship between EPOR and cell death has been demonstrated [25]. In regards to the total expression of EPOR, others have also shown that this remains stable or increases [74], and this elevation occurs within hours in response to pro-inflammatory cytokines like TNFa [75], whereas the level of its parent ligand, EPO, is downregulated under pro-inflammatory conditions [76]. Thus, understanding the inflammation-induced dysregulation of EPO signaling provides a therapeutic window where treatment with EPO may be most efficacious [76]. Altogether, the considerable changes in EPOR expression within the fetal brain in the course of inflammation stresses the need for biomarkers to determine the onset of intra-amniotic infections.

One important limitation of a large animal study is the relatively low number of animals per group. Given the relatively small animal numbers per group, we reported actual p values and tended to interpret p values between 0.05 and 0.1 as biologically relevant. This assumption decreases the chance of a false-negative finding but increases the chance that one of these differences is a false-positive result.

Altogether, the cerebral changes as found in our translational model of inflammation-related brain injury of the preterm infant could be reasonably expected to lead to deficits in learning, memory and social skills, and/or motor disabilities that can persist into adulthood [34]. However, these outcomes need to be confirmed in longitudinal studies, incorporating behavioral analysis and MRI but nonetheless inform the development of interventions to protect and/or regenerate the preterm brain in order to allow normal development. By gaining more insight into the temporal processes underlying inflammation-induced preterm brain injury, we provide an increased understanding of the pathophysiological changes in the fetal brain with the goal of helping to design new interventions or implement interventions more effectively to prevent perinatal brain damage.