The Paradoxical Effects of Chronic Intra-Amniotic Ureaplasma parvum Exposure on Ovine Fetal Brain Development

Neonatal brain injury acquired during pregnancy remains a major cause of adverse neurodevelopmental outcomes throughout life [1,2]. Chorioamnionitis, which is defined as a microbial invasion and infection of the amniotic cavity, is one of the most important risk factors for adverse neurodevelopmental outcomes of the newborn [3,4]. Ureaplasma spp. are the most common isolated microorganisms associated with chorioamnionitis [5]. Clinical recognition of amniotic fluid (AF) infections is challenging given its asymptomatic course despite sustained fetal exposure to intrauterine inflammation, particularly during the critical period of fetal brain development [6].

Intra-amniotic exposure to Ureaplasma spp. is associated with development of fetal and neonatal brain injury [7,8,9,10]. Clinical data show that there is an increased risk of intraventricular hemorrhage and impaired neurodevelopmental outcomes later in life after intra-amniotic Ureaplasma spp. exposure [7,9,10]. This association was confirmed by Normann et al. [11] who showed that intra-amniotic Ureaplasmaparvum (UP) exposure resulted in increased microglial activation, delayed myelination, and disturbed cortical development of the fetal murine brain. In contrast, clinical studies have reported that antenatal exposure to Ureaplasma spp. and brain injury do not correlate [12,13]. Diversity in microbial interplay and the timing, duration, and severity of the inflammatory response after the onset of chorioamnionitis are considered to determine the neurodevelopmental outcome, which most likely explains the considerable differences in antenatal UP exposure and brain injury incidences among studies [6,14]. In particular, the onset of cerebral inflammation during the brain's most vulnerable period from 23 to 32 weeks of gestation can have detrimental consequences for the fetal brain, particularly white matter damage. Multiple animal models have demonstrated that the brain becomes more (i.e., sensitization) or less (i.e., preconditioning) susceptible to a second injurious hit following preexposure to inflammation [15,16]. Besides cerebral inflammation, epigenetic mechanisms (such as DNA methylation and DNA hydroxymethylation) may mediate the processes leading to brain injury in response to environmental challenges in utero [17]. In line with this, DNA methylation levels in genes involved in growth and development are found to be increased in premature infants with chorioamnionitis compared to infants without chorioamnionitis [18].

Moreover, alterations of phospholipids, which are highly abundant in the brain and play important functions in cell membrane formation, as energy reservoirs and as precursors for second messengers (i.e., arachidonic acidAA) [19] have been implicated in multiple brain pathologies. In particular, changes in lipid metabolism, as seen in lysosomal storage diseases, can cause severe impaired brain function, with lipids accumulating within the brain [20].

Detailed investigations of the interactions between different infectious triggers and the timing and duration of inflammatory exposures in the context of a polymicrobial syndrome such as chorioamnionitis are essential to understanding the complex and diverse neurodevelopmental outcomes after birth. We therefore investigated the effects of chronic intra-amniotic UP exposure in the presence or absence of a second (acute) inflammatory stimulus on fetal neurodevelopment. We used a well-established translational ovine model of intrauterine inflammation in which fetuses were chronically exposed to intra-amniotic UP, followed by acute exposure to Escherichia coli-derived lipopolysaccharide (LPS). Cerebral outcome was studied by analyzing inflammation, structural injury, epigenetic markers, and the lipid profile composition of the fetal brain.

The animal procedures were performed with the approval of the Animal Ethics Committee of The University of Western Australia (Perth, WA, Australia).

Thirty-seven date-mated merino ewes were randomly assigned to study groups of 5-7 animals. Fetuses of either sex were used. Ewes received an ultrasound-guided intra-amniotic injection of culture media (2 mL) as the control or strain HPA5 of UP serovar 3 (2 × 10⁵ colony-changing units [CCU]) at 80 days of gestation (term ∼150 days). To minimize any inflammatory effects from culture media, both UP and control injections were created from stock cultures/sterile media diluted 1:100 in sterile saline. To assess the effect of an additional inflammatory hit following long-term preexposure with UP, both groups received a second intra-amniotic injection of 10 mg E. coli-derived LPS (O55:B5; Sigma-Aldrich, St. Louis, MO, USA) at 2 or 7 days before preterm delivery at 122 ± 2 days of gestation or an equivalent dose of saline (SAL; controls) (Fig. 1).

All fetuses were surgically delivered via Cesarean section at 122 ± 2 days of gestation (equivalent to 32-34 weeks of human gestation) and euthanized with intravenous pentobarbitone (100 mg/kg) immediately after birth. AF, blood, and cerebrospinal fluid were collected at delivery and cultures for UP were performed. Brains were immersion fixed in 4% paraformaldehyde.

Samples of AF (1 mL) collected by amniocentesis at LPS or control saline injection, as well as plasma, cerebrospinal fluid, and AF collected at Cesarean section delivery, were cultured for UP growth as previously described [21]. For each animal, 20 μL biological fluid was serially diluted 1:10 in Ureaplasma selective medium (Mycoplasma Experience plc., Reigate, UK) in triplicate for each sample and incubated at 37°C. Assays were performed in 96-well plates and bacterial growth was quantified via titration of the urease activity (conversion of urea to ammonium ions leading to a pH color change). Plates were observed until the bacterium-mediated color change ceased and the titration of the bacteria present was determined.

The proinflammatory cytokine interleukin (IL)-6 was measured in fetal plasma as a marker for systemic inflammation using a sheep-specific sandwich enzyme-linked immunosorbent assay (ELISA). Briefly, a mouse anti-ovine monoclonal antibody (MAB1004, working concentration 1:200; Millipore, Darmstadt, Germany) was the coating antibody. Plasma samples (100 μL) were loaded per well in duplicate and incubated for 2 h at room temperature. Incubation with the detection antibody (rabbit anti-ovine IL-6, AB1839, working concentration 1:500; Millipore) was performed for 60 min, followed by incubation for 30 min with 100 μL goat anti-rabbit HRP (working concentration 1:500; Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA). After washing, incubation with 3,3′,5,5′-tetramethylbenzidine substrate solution was performed for 15 min. The reaction was stopped by addition of 50 μL 2 N sulfuric acid. Optical density was measured using a microplate reader at 450 nm. Concentrations were expressed relative to a standard curve of ovine IL-6 recombinant protein (ImmunoChemistry Technologies, Bloomington, MN, USA).

After fixation, the right hemisphere was divided into 4 defined anatomical regions. The region of the posterior hippocampus/mid-thalamus was embedded in paraffin and serial coronal sections (4 µm) were cut with a Leica RM2235 microtome. At this level, we analyzed the hippocampus and cerebral white matter for inflammatory and structural changes since these regions are most affected following intrauterine infection at this developmental stage [22]. Four slides per staining per animal were used (i.e., every 10th consecutive slide) for immunohistochemical analysis. Hematoxylin and eosin (H&E) staining was performed for morphological and anatomical analysis. Adjacent sections were stained as previously described using a rabbit anti-ionized calcium binding adaptor molecule 1 (IBA-1) antibody (Wako Pure Chemical Industries, Osaka, Japan) for microglia, a rabbit anti-glial fibrillary acidic protein (GFAP) antibody (Z0334; DAKO) for astrocytes, a rat anti-myelin basic protein (MBP) antibody (MAB386; Merck Millipore) for myelin sheaths, a rabbit anti-oligodendrocyte transcription factor 2 (Olig2) antibody (AB9610; Millipore) for oligodendrocyte lineage cells, rabbit anti-myeloperoxidase (MPO) (A0398; DAKO) for neutrophils, mouse anti-cluster of differentiation (CD) 68 (M0718; DAKO) for active microglia/phagocytizing macrophages, and rabbit anti-CD3 (A0452; DAKO) for T lymphocytes, and mouse anti-5-methylcytosine (5-mc) (GWB-CB561B; Genway) and rabbit anti-5-hydroxymethylcytosine (5-hmc) (39769; Active Motif) were used as epigenetic markers.

Endogenous peroxidase activity was inactivated with 0.3% H₂O₂ treatment (or 1% H₂O₂ for 5-mc and 5-hmc). Antigen retrieval was performed by microwave boiling of tissue sections in citrate buffer (pH 6.0). Nonspecific binding was blocked by incubation with bovine, goat, or donkey serum in PBS. Sections were incubated overnight at 4°C with the diluted primary antibody (5-hmc 1:2,500; IBA-1, GFAP, and MBP 1:1,000; 5-mc 1:500; and Olig2 and MPO 1:200). The following day, sections were incubated with the specific secondary antibody and staining was enhanced with a Vectastain ABC Peroxidase Elite Lit (Vector Laboratories Inc., Burlingame, CA, USA) and (nickel) 3,3′-diaminobenzidine staining. If required, appropriate background staining was performed.

A more detailed molecular analysis of the cerebral tissue was done by matrix-assisted laser desorption ionization mass spectrometry imaging (MALDI-MSI) to map variations in lipid profiles of the white and grey matter. MALDI-MSI to image the lipid distribution can be invaluable in understanding complex lipid changes, and it has been used to study these molecular patterns in models of brain injury [23]. With MALDI-MSI we avoid all extraction and purification steps for lipid analysis while retaining their spatial distribution. For this technique, postfixation tissues of controls, 42UP, 2LPS, and UP&LPS groups were frozen in liquid nitrogen and subsequently samples were cryo-sectioned (10-µm thickness) in a cryostat (Leica CM3050S), deposited on indium tin oxide high-conductive slides (Delta Technologies, USA), and stored at -20°C. Subsequently, the matrix solution consisting of norharman (7 mg/mL) in 2:1 chloroform:methanol was sprayed on top of the tissue section by a vibrational sprayer (Suncollect; SunChrom, Germany) for positive ion mode and 9-aminoacridine (10 mg/mL) in 70% ethanol for negative ion mode MALDI-MSI analysis. Digital optical scans of all of the tissue sections were obtained prior to MALDI-MSI experiments using a 2,400-dots-per-inch desktop scanner. The resulting digital images were imported into MALDI Imaging HDI software v1.4 (Waters Corporation). A MALDI-quadrupole time-of-flight SYNAPT HDMS G2Si system (Waters Corporation) operating with a 200-Hz Nd:YAG laser was configured to acquire data in positive and negative V-reflectron modes. Data were acquired at a raster size of 100 × 100 µm. Instrument calibration was performed using a standard calibration solution of red phosphorus.

Principal components analysis (PCA) and discriminant analysis (DA) were used to investigate spectral similarities and differences between all samples. PCA was performed as a data compression and noise-filtering step before the application of DA, and only one fourth of the functions (n = 216) were used as input for DA. In short, PCA is an unsupervised statistical method that aims to pool a maximum amount of variance in a minimum number of independent variables. Data pretreatment, PCA, and DA were performed using our in-house built ChemomeTricks toolbox for MATLAB version 2014a (MathWorks, Natick, MA, USA). Peak assignments were performed according to the bibliography and LIPID MAPS software (http://www.lipidmaps.org/).

H&E-stained sections were analyzed by 2 independent investigators and a neuropathologist, blinded to the experimental set-up, to assess overall brain structure and inflammatory changes. The sections were examined for the presence of gliosis, hemorrhages, (cytotoxic) edema and structural damage, including cyst formation.

Immunohistochemical stainings were analyzed using a light microscope (Leica DM2000) equipped with Leica QWin Pro version 3.4.0 software (Leica Microsystems, Mannheim, Germany). Regions of interest of the white matter and hippocampus were defined as previously described [24]. These regions were chosen since they are most affected by intrauterine infections at this developmental stage. In the white matter 4-6 adjacent images were taken at ×100 magnification and from the hippocampus 1 image at ×20 magnification was taken. To assess regional vulnerability within the hippocampus, separate images were taken at ×200 magnification of the cornu amonis 1 and 2, 3, and 4 and the dentate gyrus. Area fractions of IBA-1, GFAP, MBP, 5-mc, and 5-hmc expression were determined using a standard threshold to detect positive staining with Leica QWin Pro V 3.5.1 software (Leica, Rijswijk, The Netherlands). Regions of interest were delineated in the image with large blood vessels and artefacts excluded from analysis. Since the level of DNA methylation and hydroxymethylation can differ per cell, the integrated density of 5-mc and 5-hmc was calculated by multiplying the area fraction by the mean grey value, and these values were normalized to the data of the control group as previously described by Lardenoije et al. [25]. In addition to area fraction analysis, IBA-1- and GFAP-positive cells were counted in 3 fields of view within the regions of interest at a magnification of ×400. The Olig2-positive cells were counted using Qwin software and expressed as cells/mm². MPO-positive cells were counted focusing on the cerebral vasculature, meninges, and choroid plexus. To quantify the density (cells/mm²) of MPO-, CD68-, and CD3-positive cells, representative images of the choroid plexus present in the lateral ventricles aligning the hippocampus were counted using ImageJ software The images were acquired and analyzed by an independent observer who was blinded to the experimental groups.

All values are shown as means with 95% CI or SD and p < 0.05 was considered statistically significant. Comparison between different experimental groups was performed using analysis of variance (ANOVA) or with a random intercept mixed model in case of repeated measurements per animal (e.g., different sections per brain). We applied log-transformation to obtain normal distributed data when data or variables were positively skewed before statistical testing. Statistical analysis was performed using IBM SPSS Statistics version 22.0 (SPSS, RRID: SCR_002865; IBM Corp., Armonk, NY, USA). Considering the relatively low number of animals per group, we depicted the exact p values in Figure 4-6.

Chronic UP infection in animals inoculated at 80 days of gestation was confirmed by culture of AF at the time of subsequent LPS or saline injections by amniocentesis (Fig. 2). No significant differences in UP levels were observed among the 3 groups (42 UP, 1.1 ± 0.8 × 10⁶ CCU/mL; UP&2LPS, 1.1 ± 0.8 ×10⁷ CCU/mL; UP&7LPS, 9.4 ± 0.7 × 10⁶ CCU/mL). Cultures of AF at the time of delivery were positive for UP in all experimentally infected animals, except for 1 of the animals in the UP&7LPS group. No endogenous UP growth was observed in the AF of animals that were not inoculated with UP (SAL, 2LPS, and 7LPS groups). No UP growth in cerebrospinal fluid or plasma was observed in any animal.

Overall, no sex differences in susceptibility were observed in any of the readouts. LPS exposure for 2 days decreased the body weight (SAL vs. 2LPS, p = 0.002) and increased the brain-to-body ratio (SAL vs. 2LPS, p = 0.038) compared to controls (Table 1). These significant changes were not observed in animals that were chronically exposed to UP prior to 2 days of LPS exposure (UP&2LPS). Moreover, no change in brain weight or brain-to-body ratio was observed in the animals of the 42UP, 7LPS, and UP&7LPS groups compared to the control animals.

Elevated plasma IL-6 concentrations were found in 50% (3 out of 6) of the animals exposed to LPS for 2 days and in 20% (1 out of 5) of the animals exposed to LPS for 42 days compared to controls (Fig. 3). Plasma IL-6 concentrations in the SAL, 42UP, 7LPS, and UP&7LPS animals were not detectable.

Qualitative analysis of the H&E-stained sections revealed increased cell densities in the gyral crest of the white matter which primarily consisted of glial cells. These gliotic foci were most prominent in 3 out of 6 (50%) of the animals exposed to LPS for 2 days. Furthermore, in 1 out of 6 (16%) of the animals exposed to LPS for 42 days and in 1 out of 6 (16%) of the animals exposed to LPS for 7 days, these gliotic foci were present. Control animals and animals of the UP&LPS combined groups had mild to no gliotic foci. No evidence of structural changes including intraventricular hemorrhages and cystic lesions was present in any of the experimental groups.

The neuroinflammatory changes, as indicated by the more pronounced presence of gliotic foci in the animals exposed to LPS for 2 days, were further evaluated by cell counts and area fraction analysis of the microglial marker IBA-1 and the astrocytic marker GFAP in the cerebral white matter and hippocampus. Chronic intra-amniotic exposure to UP decreased GFAP immunoreactivity (IR) (SAL vs. 42UP, p = 0.020) and the number of astrocytes (SAL vs. 42UP, p = 0.100) compared to controls (Fig. 4). IBA-1 IR- and IBA-1-positive cell numbers remained unaltered following chronic intra-amniotic UP exposure (Fig. 4). In contrast, acute exposure to LPS increased IBA-1 IR (SAL vs. 2LPS, p = 0.008) and the number of IBA-1-positive cells (SAL vs. 2LPS, p = 0.036) in the cerebral white matter (Fig. 4). In addition, morphological analysis revealed a higher density of amoeboid microglia present in the white matter after 2 days of LPS exposure (Fig. 4 insets). However, if the animals were chronically exposed to UP prior to LPS, no IBA-1 IR or IBA-1-positive cell increase was observed in the cerebral white matter at 2 or 7 days post-LPS challenge (Fig. 4d, f). Equally, LPS administration did return GFAP IR in chronically UP-infected animals to control levels (Fig. 4c). This preconditioning effect of UP was also found in the hippocampus, in which an increase in IBA-1 IR was found at 2 and 7 days following LPS exposure, but not in the groups with preexposure to UP (SAL vs. 2LPS, p = 0.002; SAL vs. 7LPS, p = 0.000) (data not shown). No changes in GFAP IR or GFAP-positive cell numbers following LPS exposure were found in the white matter (Fig. 4) or in the hippocampus (data not shown).

Since the choroid plexus is primarily dominated by macrophages, T lymphocytes and dendritic cells for continuous immune surveillance, and the resolution of cerebral inflammation [26], we analyzed the distribution of CD68+ macrophages, CD3+ lymphocytes, and MPO+ neutrophils within the choroid plexus. MPO-positive cells tended to be increased following 7-day LPS exposure compared to control animals and this increase was prevented by preexposure to UP for 42 days (SAL vs. 7LPS, p = 0.086; Table 2). In line, this increase in MPO-positive cells at 7 days post-LPS exposure was accompanied by a decrease in IBA-1 IR in the cerebral white matter compared to animals exposed to LPS for 2 days. No differences in CD68- or CD3-positive cells were found in the choroid plexus. No CD3- and MPO-positive cells were detected in the brain parenchyma.

White matter injury was studied by assessing the densities of mature (MBP) and overall (Olig2) oligodendrocytes in the cerebral white matter. An apparent increase in Olig2-positive cells was found for all groups relative to control levels; however, this only reached significance for chronically UP-infected animals (SAL vs. 42UP, p = 0.012) and animals exposed to LPS for 2 days (SAL vs. 2LPS, p = 0.037) (SAL vs. 7LPS, p = 0.211; SAL vs. 42UP&2LPS, p = 0.558; SAL vs. 42UP&7LPS, p = 0.467; Fig. 5c). In addition, MBP IR tended to be decreased at 42 days following UP exposure (SAL vs. 42UP, p = 0.097; Fig. 5). Short-term LPS exposure for 2 days resulted in a decrease in MBP IR within regions of overt microgliosis (SAL vs. 2LPS, p = 0.001) which was prevented by preexposure to UP. At 7 days of LPS exposure no changes in MBP IR were found (2LPS vs. 7LPS, p = 0.000).

We analyzed 5-mc and 5-hmc integrated density as epigenetic markers for DNA-methylation in the dentate gyrus of the hippocampus. We focused our analysis on the dentate gyrus since this is the region in the hippocampus were neurogenesis takes place and DNA methylation and demethylation are important contributors to this process [27,28]. Both short-term and chronic UP exposure resulted in an increase in the gene repression marker 5-mc integrated density compared to controls (SAL vs. 2LPS, p = 0.008; SAL vs. 7LPS, p = 0.002; SAL vs. 42UP, p = 0.008) (Fig. 6). The increase in 5-mc following LPS exposure tended to be prevented by preexposure to UP. An increase in transcription activation marker 5-hmc integrated density was only found in animals exposed to LPS for 2 days compared to controls and it was prevented by preexposure to UP (SAL vs. 2LPS, p = 0.000; 2LPS vs. 7LPS, p = 0.000; Fig. 6).

Using MALDI-MSI, we demonstrated unique regional differences in lipid composition in the preterm ovine brain between animals from the control, 42UP, 2LPS, and UP&LPS groups. Figure 7 shows the reconstructed image that represents the molecular differences in lipids between white (red area) and grey matter (green area). The lipid composition, characteristic of healthy white and grey matter of the preterm brain, was not altered following chronic UP exposure (Fig. 7a-c). In particular, in the white matter of control and 42UP animals, typical tentative assigned m/z values of different phosphocholine (PC) species such as m/z 782.5 PC 34:1+Na⁺, m/z 810.5 PC 36:1+Na⁺ or sphingomyelin m/z 725.5 SM 34:1+Na⁺ are found which are known to be representative for the white matter [29,30] (Fig. 7c). m/z 756.5 PC 32:0+Na⁺, which is a characteristic grey matter lipid [31], was detected in the grey matter of control and 42UP animals, demonstrating that control and chronic UP-exposed animals had a similar and constitutive lipid profile.

Mosaic PCA images demonstrated that short-term LPS exposure resulted in lipid accumulation in the white matter and diffusion of white matter-specific molecular patterns into the grey matter and vice versa (Fig. 7d-i; principal components 4 and 6). These changes were reduced when LPS exposure was preceded by 42 days of UP infection (Fig. 7e, h). In particular, the abundance of the white matter-specific component m/z 725.5 in the grey matter of LPS-exposed animals was not present in the grey matter of control or UP animals (Fig. 7g-i; negative spectrum of principal component 6). m/z 734.5 PC 32:0+H⁺ was accumulated in the grey matter of LPS animals (negative spectrum of principal component 4), and increased in the white matter (positive spectrum of principal component 6). In addition, principal component 6 showed that other peaks such as m/z 760 PC 34:1+H⁺ accumulate in the white matter at 2 days of LPS exposure, which did not correspond with the pattern seen in our controls or those of other studies [29] in which these peaks were evenly distributed.

Other lipid species including phosphatidylinositols (PI) and sulfatide (SF) can be identified using the negative ion mode in MALDI-MSI. Regions of interest corresponding to the white matter were selected based on the results with the positive ion mode. DF1 (Fig. 8a) revealed that the highest differences were observed between the control group (negative scores) and the LPS group (positive scores). The UP&LPS group had negative scores and therefore possessed a molecular composition more similar to that of the control group. The DF1 spectrum (Fig. 8b) showed the lipid composition of the white matter of animals exposed to LPS for 2 days. This spectrum showed that LPS exposure reduced the amount of tentative assigned SF species such as m/z806.5 SF 18:0-H^-, m/z 888.6 SF 24:1-H^-, and m/z 890.6 SF 24:0-H^-, which were described as lipids characteristic of white matter [28]. These species were mainly present in the white matter of the control, UP, and UP&LPS groups (spectrum not shown), whereas m/z 885.5 PI 38:4-H^- or m/z 718.6 PC 31:0-H^- was representative of the white matter in animals of the 2LPS group (Fig. 8b).

Chronic intra-amniotic UP exposure decreased GFAP IR and increased Olig2-positive cells and 5-mc IR in the brain. These changes have potential clinical implications postnatally.

The observed decrease in GFAP IR and the number of astrocytes (GFAP+ cells) at 42 days of UP exposure is important because these cells possess several essential functions in brain development, including regulation of extracellular glutamate homeostasis, providing structural and metabolic support to surrounding cells (e.g., oligodendrocytes) and modulate neuronal connections [32]. Changes in astrocyte function or density result in altered neurological outcomes. In particular, altered astrocyte protein expression (GFAP) and disrupted astrocyte maturation have been implicated in the pathogenesis of neurodevelopmental disorders such as autism and cerebral palsy [33,34]. Moreover, Sharma et al. [35] showed that LPS injection in the spinal cord of rodents decreased astrocytes, which was followed by hypomyelination. This suggests that white matter injury, a hallmark of preterm brain injury, can still occur in these fetuses considering the loss of GFAP IR at day 42 after UP exposure. Collectively, the astrocyte cell and protein loss in our study indicates that chronic UP exposure during the second trimester of gestation predisposes to brain pathologies that are often seen in newborns.

Second, the increase in oligodendrocyte lineage cells, as seen following 42 days of UP exposure, might indicate replenishment of oligodendrocytes upon an initial loss in the acute phase following UP exposure [36]. Importantly, UP was administered at 80 days of gestation, which is the premyelinating stage of brain development, with abundant vulnerable immature preoligodendrocytes sensitive to glutamate receptor-induced injury [37]. Interestingly, mature oligodendrocytes tended to be decreased following chronic UP exposure. Given these combined findings of increased Olig2+ cell numbers and reduced MBP+ IR, it is tempting to speculate that this indicates a maturation arrest of oligodendrocyte progenitor cells, a key feature of white matter injury in preterms [38,39,40]. This oligodendrocyte maturation arrest can be linked to the decreased astrocytes that we found in this study. Astrocytes are essential contributors to extracellular glutamate homeostasis, which is disturbed by a loss of astrocytes [32]. Since immature oligodendrocytes are particularly vulnerable to glutamate receptor-induced injury, this can lead to oligodendrocyte injury [37]. In addition, oligodendrocytes rely on astrocytes for their metabolic support via gap junctions [41]. Failure of metabolic support for oligodendrocytes following an astrocyte loss results in energy failure and eventually maturation arrest or death. Alternatively, it is tempting to speculate that oligodendrocyte maturation arrest may be connected to the apparent increase in the DNA methylation marker 5-mc at 42 days of UP exposure, which is a very important repressor of gene transcription [42]. This theory is supported by several reports stating that changes in epigenetic regulatory mechanisms contribute to disturbed maturation and differentiation of immature oligodendrocytes [43,44,45]. Moreover, inflammation induced epigenetic changes during early development can cause substantial lasting neurodevelopmental impairments later in life [46,47]. Altogether, these data offer a mechanistic insight into the association between intra-amniotic UP exposure and the increased incidence of adverse neurodevelopmental outcomes postnatally.

Interestingly, the cerebral phenotype following short-term LPS exposure was remarkably different compared to chronic UP exposure. In particular, we demonstrated that short-term LPS exposure induced a rapid and temporal increase in the number of microglia and decreased myelin IR, reflecting diffuse cerebral inflammation with hypomyelination. Microglia are important in inflammatory perinatal brain injury [43]. Aberrant or excessive microgliosis can be detrimental for the immature brain, resulting in white matter injury [14], which corresponds to the loss of myelin that we found in our study. The cerebral inflammatory response following LPS in this study seems to be temporal since no increase in microglial density was found following 7 days of LPS exposure. This dynamic response of activated microglia is consistent with distinct phases of cerebral inflammation [48] and can be explained by the presence of neutrophils in the choroid plexus, which are known to be important for the resolution of cerebral inflammation [26]. However, our immunohistochemical analysis does not rule out the possibility that phenotypic conversion of microglia might be induced following short-term LPS or chronic UP exposure, which might influence the cerebral immune response.

Although such short-term UP effects were not investigated in our model, in a study performed by Normann et al. [11] in rodents short-term UP exposure during early pregnancy resulted in an increased microglial density and a decreased MBP density in the fetal brain. This cerebral phenotype is consistent with our data following short-term LPS exposure, indicating that timing and not the inflammatory trigger is more important for the neurological outcome of the fetal brain.

Besides the DNA methylation marker 5-mc, the hydroxylated product and transcription activation marker 5-hmc was increased following 2 days of LPS exposure. 5-hmc is very important for proper neurodevelopment and it is altered in the umbilical cord of babies born after preeclampsia and gestational diabetes mellitus [49] and in the hippocampus of 7-week-old mice exposed to noninfectious stress [50]. In addition, 5-hmc alterations are associated with severe neurodevelopmental disorders such as Rett syndrome, which is caused by mutations in the MeCP2 gene that encodes for proteins that directly bind to methylated DNA domains [51]. Therefore, the alterations we found in epigenetic markers following acute LPS exposure might explain, at least in part, the association between chorioamnionitis and the development of psychopathology later in life [52]. Since epigenetic changes are reversible, these findings provide new therapeutic targets to prevent long-lasting neurodevelopmental morbidities following prenatal stress [17].

In addition, our lipid data provide supporting evidence that short-term LPS exposure results in lipid accumulation and “diseased” lipid patterns in the preterm brain. Such lipid accumulation in the brain is associated with severe neurological damage and altered brain functions [20]. In addition, we noted a relative decrease in myelin-specific sphingolipids in the white matter of the LPS-exposed animals, which confirms and extends our findings concerning the loss of MBP in these animals. The abundance of PI following 2 days of LPS exposure was primarily seen within the region of increased IBA-1 IR. In line with this, the phosphorylated form of PI, phosphoinositide, is known to be upregulated in microglia and contributes to the activation of microglia following ischemia [53].

Third, we found that chronic intra-amniotic UP exposure prevented an increase in IBA-1 IR and IBA-1+ cells, 5-hmc IR, lipid profile changes, and a decrease in MBP IR upon a second inflammatory hit with LPS. This phenotype, also referred to as “preconditioning”, has been previously described in animal models in which preexposure to inflammation induced by LPS renders the brain less susceptible to a second hypoxic-ischemic insult, thereby resulting in less brain injury [54,55]. This preconditioning effect of chronic UP exposure could be explained by the work of Cao et al. [56], which showed in pregnant sheep that microglia, once activated in vivo by intra-amniotic LPS exposure, display diminished inflammatory responses following reexposure to LPS. Moreover, they stated that the memory acquired by microglia upon the first exposure to inflammation might be mediated by epigenetic regulatory processes [56]. Although this hypothesis needs to be tested in future studies, it is noteworthy that changes in the global level of 5-hmc and 5-mc were observed in our study following acute LPS exposure that was prevented by chronic UP exposure. Clearly, long-term protection after inflammation-induced preconditioning needs to be confirmed in a longitudinal study, but it is considered to be permanent since structural and functional protection for up to 8 weeks was established following hypoxic preconditioning in a neonatal rodent model [57].

One important limitation of a large animal study is the relative 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.

In this double-hit study, in which sequential different infectious hits were tested, we showed that microbial interactions, the moment of onset, and the duration of these potential injurious triggers determine the neurological outcome. These findings seem to be an important explanation for the diversity of neurological outcomes associated with intra-amniotic UP exposure. Altogether, these data emphasize that an accurate history of infections during pregnancy is essential to guide neonatal management, which warrants the need for biomarkers to diagnose antenatal infections.

This work was supported by a National Institutes of Health (Bethesda, MD, USA) grant (HD 57869), a Royal Society (London, UK) grant (IE130066), and a Financial Markets Foundation for Children (Sydney, NSW, Australia) grant (EOI-2013-059). F.P.Y.B. received funding from the European Union's Horizon 2020 research and innovation program under the Marie Sklodowska-Curie Program, TargetCare (ITN-2014-ETN 642414). The authors would like to thank Nico Kloosterboer and Hellen Steinbusch for their excellent technical assistance.

The authors of this paper declare that there are no actual or potential conflicts of interest. The authors affirm that there are no financial, personalm or other relationships with other people or organizations that have inappropriately influenced or biased this research.