Differential Brain and Cerebrospinal Fluid Proteomic Responses to Acute Prenatal Endotoxin Exposure

Chorioamnionitis (CA) is a risk factor for preterm birth and is associated with neurodevelopmental delay and cognitive disorders. Prenatal inflammation-induced brain injury may resolve during the immediate postnatal period when rapid brain remodeling occurs. Cerebrospinal fluid (CSF) collected at birth may be a critical source of predictive biomarkers. Using pigs as a model of preterm infants exposed to CA, we hypothesized that prenatal lipopolysaccharide (LPS) exposure induces proteome changes in the CSF and brain at birth and postnatally. Fetal piglets (103 days gestation of full-term at 117 days) were administered intra-amniotic (IA) lipopolysaccharide (LPS) 3 days before preterm delivery by caesarian section. CSF and brain tissue were collected on postnatal Days 1 and 5 (P1 and P5). CSF and hippocampal proteins were profiled by LC–MS-based quantitative proteomics. Neuroinflammatory responses in the cerebral cortex, periventricular white matter and hippocampus were evaluated by immunohistochemistry, and gene expression was evaluated by qPCR. Pigs exposed to LPS in utero showed changes in CSF protein levels at birth but not at P5. Complement protein C3, hemopexin, vasoactive intestinal peptide, carboxypeptidase N subunit 2, ITIH1, and plasminogen expression were upregulated in the CSF, while proteins associated with axon growth and synaptic functions (FGFR1, BASP1, HSPD1, UBER2N, and RCN2), adhesion (talin1), and neuronal survival (Atox1) were downregulated. Microglia, but not astrocytes, were activated by LPS at P5 in the hippocampus but not in other brain regions. At this time, marginal increases in complement protein C3, LBP, HIF1a, Basp1, Minpp1, and FGFR1 transcription indicated hippocampal proinflammatory responses. In conclusion, few days exposure to endotoxin prenatally induce proteome changes in the CSF and brain at birth, but most changes resolve a few days later. The developing hippocampus has high neuronal plasticity in response to perinatal inflammation. Changes in CSF protein expression at birth may predict later structural brain damage in preterm infants exposed to variable types and durations of CA-related inflammation in utero.


Introduction
Chorioamnionitis (CA) and the associated intrauterine inflammation are risk factors for spontaneous preterm delivery [1], and CA may associate with 40-70% of these births [2]. The links among CA, preterm birth, and cerebral lesions are not clear [1] and may depend on time, type, and duration of perinatal inflammation. CA may induce lesions in white matter, such as altered vascular microstructure and periventricular hemorrhage, as well as lesions in grey matter, such as increased neuronal apoptosis, abnormal neuronal circuit formation, and cognitive impairments [3][4][5][6]. Recent data also suggested that neuropsychiatric disorders, such as autism spectrum disorder and schizophrenia, might have a neurodevelopmental origin and be linked to feto-maternal inflammation [7][8][9][10]. CA-related brain lesions may be mediated in spatiotemporal ways by several mechanisms, including transport of cytotoxic molecules across the blood-brain barrier (BBB), transient activation of microglia accompanied by release of proinflammatory cytokines and reactive oxygen species, arrest of preoligodendrocyte maturation followed by hypomyelination, reduced synaptic density, impaired neurogenesis, and cell death [5,[11][12][13][14]. In preclinical models, particularly the hippocampus, a brain region involved in memory and cognition was affected by CA [13,15,16]. In preterm infants, CA may reduce the volume of the hippocampus [17].
The type and severity of inflammation-induced abnormalities in the fetal brain are critically dependent on the type, timing, and duration of in utero inflammation, since distinct neurodevelopmental programs are affected at different gestational ages [13,18,19]. Thus, the diagnosis and characterization of CA-related brain pathophysiology are difficult. Evidence from numerous experimental animal studies suggests that feto-maternal inflammation induces an imbalance in endogenous neurotrophic factors and immune molecules that impact the course of brain development [1,15,20,21]. However, the underlying mechanisms and optimal biomarkers of postnatal CA-induced brain injury remain unknown.
Cerebrospinal fluid (CSF) components have emerged as important mediators of extracellular signaling and are centrally involved in the maintenance of brain development and homeostasis [22][23][24][25]. In addition to ependymal cells that line the cerebral ventricles and regulate the concentration of active peptides in the CSF, neural progenitor cells extend processes into the ventricles and directly contact the CSF (reviewed in [26]). CSF, which is cycled throughout the brain perivascular space, provides a fluid for the glymphatic system and is responsible for the clearance of brain parenchyma [27]. Therefore, CSF not only carries plasma proteins but also is indicative of the endogenous brain microenvironment and is thereby an important source of brain pathophysiological markers [28]. The protein composition of CSF changes throughout an individual's lifetime, and transient changes in the levels of growth factors (e.g., IGFs, FGF2, NGF, and TGF-a) occur during the fetal and perinatal periods [22,26,[29][30][31]. Importantly, the total CSF protein levels are higher in preterm vs. term neonates, while the white blood cell counts do not differ [32].
We previously documented rapid changes in CSF protein composition in response to neonatal infection in preterm pigs [33]. Furthermore, we described how the brain differs both structurally and functionally between preterm and term pigs [34][35][36][37] and how the brain responds to different diets [38][39][40] and infections [41] in the days and weeks after preterm birth. Prenatally, intra-amniotic (IA) endotoxin (LPS) given to fetal pigs induces local inflammation at birth, as evidenced by elevated IA leukocyte and cytokine levels [42], and this may have consequences for systemic immunity development [43]. Responses in multiple organs were indicated by tissue markers of inflammation and neutrophil/ macrophage infiltration in lungs, gut, and kidneys [42,44,45]. Across these earlier studies, fewer effects were observed five days after birth, relative to the day of birth. This was consistent with changes in plasma proteome profiles, where 45 and 2 proteins were affected by fetal LPS at birth and day 5, respectively [44]. Screening of CSF biomarkers is frequently used to predict neuropathology in human adults [28] but less is known from infants, including those born preterm and exposed to CA prenatally. Using brain and CSF samples from the preterm pig studies referred to above, our aim was to demonstrate how a few days of fetal endotoxin exposure affects the CSF and brain proteomes at birth and day 5, potentially reflecting the short-term brain responses to acute CA in preterm infants.

Animal Experimentation and Tissue Collection
The animal experimental protocol was approved by the Danish Animal Experiments Inspectorate (license number, 2014-15-0201-00,418), which is in accordance with the guidelines from Directive 2010/63/EU of the European Parliament. The detailed design of the in vivo study is shown in Fig. 1A using BioRender (San Francisco Bay Area, USA). Briefly, 54 fetuses from three pregnant sows (Large White × Danish Landrace × Duroc) were randomly allocated to receive an intra-amniotic injection of either 1 mg LPS/fetus (Escherichia coli, serotype 055:B5; Sigma-Aldrich, Copenhagen, Denmark) or control (saline or no injection; CON) at 103 days of gestation (E103). After 3 days, at E106, which corresponds to 90% gestation (full-term birth is 117 ± 2 days), the fetuses were delivered by cesarean section and randomly allocated either to be sacrificed shortly after delivery at Day 1 (P1; CON n = 14, LPS n = 16) or reared for 5 days (P5; n = 12 for each group) as previously described [44]. There were 5 fetuses from LPS group found dead at delivery, which were excluded from all the follow-up analysis. The sex distribution across litters and groups is indicated in Supplementary Table S1. Among the survived fetuses, CSF samples were successfully collected from 5 and 7 pigs at P1, and 9 and 10 pigs at P5 for CON and LPS groups, respectively. The birth weight was similar between LPS and control groups. The changes in systemic inflammatory marker levels and blood cell counts were previously published [42]. Immediately after euthanasia, CSF samples were collected by suboccipital puncture and frozen at -80 °C for proteomics analysis. A small aliquot of CSF sample was used for counting leucocytes. The dissected brain was separated into hemispheres: The right hemisphere was immersed in 4% paraformaldehyde (PFA), and the isolated hippocampus from the left hemisphere was snap frozen and stored at -80 °C for further analysis.
The levels of IL-8 in liver and spleen samples were measured with porcine DuoSet ELISA kits (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's protocol.

Proteomics Analysis
The CSF samples were centrifuged at 1000 × g for 5 min to remove cell debris. Hippocampal tissue was homogenized in chilled homogenization buffer (5% sodium deoxycholate, 50 mM triethyl ammonium bicarbonate, pH 8.5). The protein concentration was determined by UV spectrometry (NanoDrop 2000; Thermo Scientific, MA, USA). A PVDF membrane-based proteomic sample processing method was used for on-membrane tryptic digestion using a multiscreen filtration plate basically following the procedure described by [46]. Briefly, the total CSF and hippocampal proteins (15 μg) were diluted with saturated urea followed by a reduction with TCEP (10 mM) and alkylation (50 mM chloroacetamide). The samples were loaded on primed membranes and washed twice with 300 μl of 50 mM TEAB buffer. Digestion buffer (100 μl, 5% TFE (v/v), 5% ACN (v/v), trypsin 1:60 in 50 mM TEAB) was added to each sample and incubated overnight at 37 °C in an air incubator with high humidity. Tryptic peptides were transferred to a collection tube, and the remaining peptides were extracted with 150 μl TEAB buffer supplemented with 40% acetonitrile (ACN, v/v) and 0.1% formic acid (FA; v/v). The pooled extracts were dried Fig. 1 A The outline of the experimental setup. Chorioamnionitis (CA) was induced by intra-amniotic infusion of LPS in pregnant sows at 103 days of gestational age (GA). Preterm piglets were born by cesarean section at 106 days GA, and cerebrospinal fluid (CSF) and tissue samples were collected either on Day 1 or Day 5 for proteomics, ELISA, qPCR and immunohistochemistry (IHC) analyses. B Leucocyte counts in the CSF of vehicle (control) and LPS-exposed pigs measured at Day 1 (control, n = 5; LPS, n = 7). C Expression of the proinflammatory cytokine IL-8 in the liver and spleen at P1 (control, n = 9; LPS, n = 11) and P5 (control, n = 14; LPS, n = 16). The values are presented as mean ± SEM, **p < 0.01 Molecular Neurobiology (2022) 59:2204-2218 2206 by vacuum centrifugation and resuspended in loading solvent (2% ACN, 0.1% trifluoroacetic acid, 0.1% FA in Milli-Q water) before being loaded into the LC-MS system.
The individual CSF and hippocampal samples were randomized and sequenced on a hybrid trapped ion mobility spectrometry (TIMS) quadrupole time of flight (QToF) mass spectrometer, i.e., timsTOF in tims-off mode (Bruker Daltonics, Bremen, Germany) coupled to a modified nanoelectrospray ion source (CaptiveSpray, Bruker Daltonics) with an applied voltage of 1800 V. Liquid chromatography was performed using a Dionex RSCL Proflow UHPLC setup (Dionex, Thermo Scientific, Waltham, USA). Each sample was loaded onto a 2-cm reverse-phase C18-material trapping column and separated on a 75-cm analytical column (both from Acclaim PepMap100, Thermo Scientific). The liquid phase consisted of 96% solvent A (0.1% FA) and 4% solvent B (0.1% FA in ACN) at a flow rate of 300 nl/min. The peptides were eluted from the column by increasing concentrations of solvent B (to 8% and subsequently to 30%) with a 35-min ramp gradient and introduced into the mass spectrometer by a CaptiveSpray emitter for electrospray ionization (Bruker; Germany). The mass spectrometer was operated in positive mode with data-dependent acquisition (DDA), alternating between survey spectra and isolation/ fragmentation spectra. All the samples were analyzed in duplicate in a random order. The average between two technical replicates were used to reduce the influence of noise in downstream data analysis.

Immunohistochemistry (IHC) and Image Analysis
Based on the pig brain atlas [47], a predefined brain region including the hippocampus was embedded in paraffin and serially sliced into 7-µm coronal sections. IHC was assessed with antibodies to glial fibrillary acidic protein (GFAP; 1:800, Dako, Denmark) and ionized calcium-binding adapter molecule 1 (Iba1; 1:1000, ab5076, Abcam, Denmark) using 3-4 sections per animal. The number of animals from each litter (L) included in IHC analysis was as follows: L1/L2/L3 for P1 CON, 2/2/4; P1 LPS, 3/5/3; and for P5 CON, 2/4/2; P5 LPS, 3/3/3. Tissue deparaffanization and labeling were performed as described previously [48]. Four images per section from the cortical region, periventricular white matter (PVWM), and hippocampus were captured using CAST-GRID software (Olympus Denmark A/S) and quantified with image analysis software PlabApp (Protein Laboratory, Denmark). For each staining, recorded images were adjusted to the same threshold level, and the positively immunostained area was quantified and expressed as an average percentage. An observer blinded to the study groups performed all the histological assessments. To minimize staining differences, all the sections were stained at the same time using the same preparations.

Gene Expression Analysis by Real-Time Quantitative PCR
Hippocampal samples were homogenized in QIAzol lysis reagent (Qiagen, Copenhagen, Denmark), and total RNA was isolated using an RNeasy Lipid Tissue Mini Kit (Qiagen, Copenhagen, Denmark) and converted to cDNA with a High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher, Waltham, MA, USA). Quantitative real-time PCR analysis of the expression of selected genes (see Supplementary Table S2 for primer list) was performed in duplicate using the LightCycler 480 SYBR Green I Master kit on a LightCycler 480 (both Roche, Basel, Switzerland). The gene expression level was normalized to hypoxanthine-guanine phosphoribosyltransferase 1 (HPRT1), and the relative expression was analyzed using the 2-△△CT method.

Data Analysis and Statistics
Proteomic data were searched with MaxQuant (v 1.6.2.3), using default settings, against the UniProt Sus scrofa reference proteome (UPID000008227) using carbamidomethyl cystine (fixed mod) and oxidized methionine (variable mod). Initial processing was performed in Pereus (v.1.6.5.0 for CSF and v. 1.6.2.3 for hippocampus). All the statistics were performed in R (version 4.0.2, Vienna, Austria), and graphs were generated with R studio (RStudio, Boston, MA, USA) and GraphPad Prism version 8.0.2 software (Graph-Pad Software, San Diego, CA). For proteomic analysis, the data were sorted into the effect protein group and exposure protein group using the criteria that an effect protein was detected in at least 80% of all the samples of each treatment group and an exposure protein was detected in less than 20% of all samples in one treatment group and above 80% of all samples in another group. Significant differences between the treatment groups were identified by a linear mixed effect model using treatment and sex as the fixed factors and litter as a random factor, and the analyses were conducted with the lme4 and multcomp packages. To control the type I error, p value tests were further adjusted into q values with Benjamini-Hochberg (BH)-adjusted false discovery rate (FDR, α = 0.1). A threshold of adjusted p ≤ 0.1 was set to identify effect differentially expressed proteins (DEPs) between groups. Significant exposure DEPs between groups were identified by Fisher's exact test, and a threshold of p ≤ 0.05 was used to identify exposure DEPs between groups. qPCR data were analyzed using a linear mixed-effects model followed by Tukey's post hoc multiple comparison test, with treatment and sex as fixed factors and litter as a random factor. A threshold of p < 0.05 was used to identify significantly regulated genes between groups. A Venn diagram was generated in R by the VennDiagram package.

Prenatal LPS Exposure Marginally Alters the CSF Protein Profile at Birth
The study design of the experiment is illustrated in Fig. 1A. Compared to CON, intra-amniotic (IA) exposure to LPS did not alter the number of infiltrating leucocytes in the CSF on P1 (Fig. 1B), suggesting that the BBB was intact, which is consistent with our previous observations [42]. Moreover, the levels of IL-8 in the liver and spleen tissues were not different between the CON and LPS groups on both P1 and P5 (Fig. 1C). Of 1398 CSF proteins identified and annotated after statistical filtration, 953 and 981 proteins were detected in at least one piglet on P1 and P5, respectively. Annotated proteins were further sorted into the effect and exposure protein groups. Compared with the data obtained in our previous study [44], the CSF proteome had partial overlap with the corresponding plasma proteome (44.3% overlap of total annotated proteins), indicating that (i) at least some of the identified CSF proteins were produced locally in the brain and (ii) CSF protein composition underwent time-dependent dynamic changes.
IA LPS exposure affected nine CSF DEPs on P1 and only one DEP on P5 (Supplementary Table S3). Compared with the CON group, the LPS group had five proteins that were significantly upregulated (q ≤ 0.05), including C3, complement factor B (CFB), hemopexin (HPX), plasminogen, and inter-alpha-trypsin inhibitor heavy chain 1 (ITIH1) (q ≤ 0.1 for the latter), at P1 ( Fig. 2A, B); these proteins have functions related to the immune response and mimic the similar trend of these protein levels in the plasma [44]. Moreover, the CSF levels of fibroblast growth factor receptor 1 (FGFR1) and brain acid soluble protein 1 (BASP1) were significantly downregulated at P1 but recovered to CON levels at P5 (Fig. 2C).
Moreover, an additional eight proteins were identified as "exposure DEPs" on P1, and these proteins were detected in the P1 group at more than 80% and less than 20%, and only one of them (CPN2) overlapped with plasma DEPs (Fig. 3A; Supplementary Table S4). Among these eight CSF DEPs, six proteins (talin1, atox1, aminoacylase, HSPD1, UBEN2N, and RCN2) were transiently undetected at P1 but returned to CON levels at P5 (Fig. 3B-G). In contrast, two DEPs, VIP and CPN2, showed the reversed trend, e.g., they were absent in the CON group at P1, while at P5, their levels were similar to the LPS levels ( Fig. 3H-I). These results are consistent with our clinical observations and plasma proteome profile results from previous studies [42,44]. Our results suggest that prenatal LPS exposure marginally alters the CSF protein profile at birth, but most of the effects resolved after a few days postnatally (neonatal adaptation phase). Overall, our results revealed plasma-shared and CSF-specific transient changes in protein levels in perinatal CSF in response to intrauterine LPS exposure.

Antenatal Inflammation Induced Hippocampal Microgliosis and Altered the Protein Profile at P5
Compared with the CON group, a significantly higher level of Iba1-ir was observed in the hippocampus of LPS-treated pigs at P5 (p < 0.05) but not at P1. Moreover, a decrease in Iba1-ir at P1 vs. P5 was observed in the PVWM in both the CON (p < 0.05) and LPS groups (p < 0.01) over time.
A similar trend was observed in the hippocampus of the CON group (p < 0.01), whereas for the LPS group, the time-dependent decrease in Iba1-ir was not significant (Fig. 4A,C). No correlation between sex and Iba1-ir was observed in any of analyzed brain regions both at P1 (male/ female, 8/11) and P5 (male/female, 8/9). No difference between treatment groups was observed for GFAP labeling in all the analyzed regions and across time points, although the developmental increase in GFAP immunoreactivity was recorded within PVWM alone, suggesting ongoing gliogenesis in this region (Fig. 4B).
Given that microglial activation was observed only at P5 and was limited to hippocampal formation, we next sought to investigate the molecular changes in the hippocampus at this time point. MS-based proteomics of hippocampal tissue identified a total of 4038 quantifiable proteins. While 682 proteins showed a p value < 0.05, only three proteins remained significant after FDR adjustment (Supplementary  Table S5). Among these proteins, the expression of both elastin and multiple inositol polyphosphate phosphatase 1 (MINPP1) was downregulated in the LPS group, and the expression of prickle homolog 2 (PRICKLE2) was upregulated in the LPS group ( Fig. 5A; Supplementary Table S5). In addition, the expression of three hippocampal proteins, melanoma inhibitor protein 3 (MIA3), claudin 11 (CLDN11) and plasmolipin (PLLP), was upregulated 3-5 fold in the hippocampus at P5, but the adjusted p value did not pass the threshold criteria (Fig. 5B).

IA LPS Transiently Changed the Expression of Immunomodulatory and Neuroplasticity Genes in the Hippocampus
To further validate the expression of selected hippocampal targets by an alternative approach, reverse transcription qPCR was performed. Compared with those in the CON group, the expression levels of elastin, CLND11, MIA3, PLLP, and PRICKLE2 were unaltered in the LPS group over time (data not shown), whereas the hippocampal expression levels of MINPP1 and FGFR1 were significantly upregulated at P5 (Fig. 6A), probably due to a compensatory response. In addition, the expression of BASP1 tended to decrease at P1 (p = 0.07) but increased at P5 (p < 0.05) in the LPS group. The expression of several other selected genes with immunomodulatory functions was significantly upregulated in the LPS group, including LBP (Lipopolysaccharide-binding protein), S100A9, Fig. 2 Prenatal intra-amniotic LPS exposure marginally alters the CSF protein profile at birth. Effect DEPs selected by the criterion that one protein was detected in at least 80% of all the samples from each group. A Venn diagram illustrating the overlapping DEPs between plasma and CSF. B The expression of the CSF DEPs C3, CFB, hemopexin, plasminogen, and ITIH1, which was upregulated by fetal LPS exposure, on P1. C The expression of the CSF DEPs FGFR1 and BASP1, which was downregulated by fetal LPS exposure, on P1. In P1: control, n = 5; LPS, n = 7. In P5: control, n = 9; LPS, n = 10. The data are presented as the mean ± SEM. *, q ≤ 0.05; #, q ≤ 0.01. C3, complement C3; CFB, complement factor B; ITIH1, inter-alphatrypsin inhibitor heavy chain H1; FGFR1, fibroblast growth factor receptor 1; BASP1, brain acid soluble protein 1 and HIF1A (p < 0.05), or showed a trend of being upregulated, i.e., C3 (p = 0.06; Fig. 6B), confirming the proinflammatory hippocampal response observed at the histological level.

Discussion
Neonates exposed to intrauterine infection/inflammation have an increased risk of neurological, cognitive and neuropsychiatric disorders later in life [7][8][9][10]. However, the underlying mechanisms and biomarkers of brain disturbances caused by the many different types and intensities of fetal inflammation have not been elucidated. In the current study, we characterized the proteomic responses of the CSF and brain after a brief period (3 days) of LPS exposure just before preterm birth, modelling infants born prematurely under conditions of maternal CA. We found transient and only moderate changes in CSF protein levels at birth, followed by limited, but significant, hippocampal microgliosis and protein expression changes five days later. Temporary exposure of the amniotic sac to LPS in pig fetuses induced CA-like symptoms with a moderate degree of systemic inflammatory response, probably related to LPS reaching the fetal skin, lung, and gastrointestinal tract via amniotic fluid [42]. Such peripherally administered LPS likely has limited ability to enter the brain, even after BBB disruption [49]. In contrast to systemic LPS stimulation, which results in a robust and broad cytokine response in the brain [50], intra-amniotic LPS exposure has previously been shown to elicit a limited proinflammatory cytokine response in fetal brain tissues [51]. Thus, we speculate that the observed changes in the central nervous system (CNS) were induced by multiple focal immune responses that indirectly affected the fetal brain via upregulation of the plasma, liver, and kidney inflammatory proteins that we described previously [44]. The absence of dramatic changes in the CSF proteome or leukocytes at birth, with limited brain tissue responses five days after birth, confirmed that the LPS-induced CNS inflammatory responses were relatively mild in this model. Among large-brain animals, most fetal neuroinflammation experiments, coupled with preterm birth, have been conducted in lambs. When delivered preterm to reflect immature brain structure and function, such lambs show poor viability unless they are provided very intensive respiratory support [13,52]. In contrast, 90% gestation preterm pigs can be relatively easy resuscitated and are viable for days or even weeks with supportive intensive care. This unique model enables the investigation of how prenatal and postnatal factors interact to determine neurological outcomes in preterm or nearterm neonates [20,33,[53][54][55]. The immune system and anatomical development of the brain in pigs may also be more similar to those of human infants than in many other models [53,54]. Furthermore, the large litter size of genetically related piglets (18-26 pigs per litter), birth-related clinical complications similar to those of human preterm infants and good survival after initial rearing in incubators make preterm pigs a particularly suitable model for studies of disordered neurodevelopment [34,41,48,56]. In this study, the upregulation of the expression of complement system elements, such as C3 and CFB, in the CSF at birth (P1) was correlated with the upregulation of the expression of these proteins in plasma, suggesting the systemic origin of these CSF proteins. However, at a later time point, i.e., P5, we observed the upregulated expression of C3 and other inflammation-associated genes, such as LBP, S100A9, and HIF1a, in the hippocampus, suggesting that a primary neuroinflammatory response occurred, particularly in the hippocampus, where the activation of microglia was also confirmed at the histological level. Importantly, in the brain, the function of complement cascade products, including complement C3, extends beyond inflammation. Such proteins are partly responsible for shaping the synaptic network in early brain development [57,58], and their expression is upregulated in different pathological settings related to neuroinflammation, demyelination, and neurodegeneration (reviewed in [59,60]. Altered C3 levels were also noted in the CSF of patients with bacterial meningitis [61]. In addition, during brain development, C3 is functionally important for migrating neurons [62] and may inhibit neurogenesis [63]. Interestingly, since the complement system is mechanistically involved in synapse elimination, its activation in the perinatal period was suggested to be linked to psychiatric conditions, such as schizophrenia and autism spectrum disorder [59]. Thus, the upregulation of C3 levels in the CSF followed by microglial activation might be a link between neuroinflammation and altered synaptic organization [60]. The observed increased levels of VIP, a neurotrophic factor and neurotransmitter centrally involved in the control of GABAergic transmission, and upregulated expression of the postsynaptic protein PRICKLE2 in the hippocampus on P5, both of which were previously linked to the pathophysiology of epilepsy and autism-like behavior [64,65], further strengthen this association.
The higher level of hemopexin and plasminogen in the CSF of LPS pigs might originate from plasma due to a compromised blood-brain barrier following LPS exposure [66]. However, recently both plasminogen and its activator were shown to be expressed and released within the CNS where they control dopamine release [67] and BDNF maturation by a plasmin-dependent mechanism [68]. Thus, elevated CSF plasminogen might further link inflammation with altered synaptic plasticity. Another acute phase plasma protein found to be upregulated in CSF, hemopexin, might also at least in part be produced by neuronal cells in response to inflammation. Hemopexin was previously shown to be produced by hippocampal neurons, glial cells, and ependymal cells lining the ventricular system [69]. Consistent with our results, hemopexin was shown to be upregulated in CSF following inflammation in rats [70], suggesting an ongoing neuroinflammatory response after IA LPS treatment.
Several downregulated CSF proteins, such as UBE2N [71], RCN2 [72], FGFR1, BASP1, and HSPD1 [73], were previously shown to be associated with synaptic homeostasis and maturation, and their decreased expression might further highlight the possible involvement of brief CA-induced inflammatory conditions in aberrant synaptic remodeling [74]. In particular, the HSPD1 protein, a constitutive mitochondrial heat shock protein expressed in astrocytes, neurons, and other neuronal cells [75], was recently identified as a long-lived protein associated with synapses [73], and its downregulated expression may thus reflect the long-lasting synaptic maldevelopment linking CA with childhood and adolescent cognitive dysfunctions [7,9,74]. Furthermore, the transient decrease in the abundance of CSF proteins that, among other functions, are involved in neuronal survival, such as ATOX1 [76], HSPD1, FGFR1, and cellular adhesion and angiogenesis, such as talin1 [77], as well as the upregulated expression of the septic biomarker CPN2 [78], which is involved in brain lipid metabolism [79], also indicate disrupted cerebral homeostasis following CA.
In connection with the findings listed above, the observed downregulation of FGFR1 and BASP1 protein expression in the CSF at P1 and the upregulated expression of these genes in the hippocampus at P5 are of particular interest. Both FGFR1 and BASP1 are abundantly expressed in the developing brain and regulate neurite outgrowth and synapse differentiation, hence being centrally implicated in neuronal plasticity [80,81]. In particular, BASP1 was previously shown to be involved in stimuli-induced synaptic changes [80]. However, FGFR1, in addition to its canonical function as a cell membrane receptor, can also be translocated to the nucleus, where it controls neuronal growth and differentiation and is suggested to be involved in the pathogenesis of schizophrenia, particularly via its nuclear signaling (reviewed in [82]). In addition, FGFR1 can directly interact with CD200 [83], an immunomodulatory glycoprotein expressed primarily on neurons, and thus modulate neuron-microglia communication [84]. We suggest that hippocampal upregulation of the expression of these genes at P5 might be a delayed compensatory reaction of tissues to the IA LPS-induced early decrease in the expression of these proteins in CSF at P1. The observed altered expression of these targets at the protein and gene levels adds further weight to the argument that CA-induced inflammation may affect synaptogenesis/synaptic plasticity a few days after exposure and potentially predispose the appearance of late-onset neuronal dysfunctions. Microglia are a highly plastic cerebral resident myeloid cell population representing 10-15% of total brain cells and are key contributors to the neuroinflammatory response after CNS insults and peripheral pathological conditions [74]. Despite methodological differences across models, exposure to LPS or viral infection was shown to increase microglial activation and/or cell density in rodent [85][86][87], ovine, and porcine models of fetal inflammation, although these models require different intervals and severity of IA exposure [13,16,20,88]. Of note, in our model, reactive microgliosis, reflected by increased Iba1+-ir, was observed on P5, particularly in the hippocampus, which is the brain region centrally involved in memory and cognitive processes and previously shown to be highly sensitive to fetal inflammatory challenges [18,89]. The CA-induced hippocampal neuroinflammatory response in preterm pigs may explain why preterm infants exposed to CA have smaller hippocampal volumes [17], which in turn correlates with poor working memory at later time points [90].
In our model, short IA LPS stimulation showed an undistinguishable astroglial response, as indicated by GFAP-ir, in different brain regions and at both early-(P1) and later (P5) postnatal time points. In contrast, in a rat model of feto-maternal inflammation, GFAP expression was transiently increased in the hippocampus and other brain regions (PVWM, prefrontal cortex) of infected pups at P7 [85,91]. We suggest that in pigs, a longer postnatal follow-up analysis is required to detect the potential effect of in utero LPS stimulation on astrocyte activation. The inflammatory threshold (time, duration, and type) required to elicit fetal brain damage in preterm pigs is unknown and remains to be investigated in this model.
In addition to the impact of CA inflammation, we observed significant postnatal developmental changes in GFAP-ir specifically in PVWM, a white matter bundle especially vulnerable to inflammatory agents in premature infants [92]. Similar to preterm infants, preterm pigs were characterized by an immature microstructure of white matter [55]. The observed developmental increase in GFAP+ir in this region is supported by the strong role of astrocytes in ongoing myelination processes [93], which in pigs peak during the perinatal period in a biphasic manner, i.e., 2 weeks before and 3 weeks after birth [94]. Our results (Fig. 4B) are also consistent with previous observations that white matter-derived astrocytes expressed higher levels of GFAP than astrocytes from the gray matter [95]. Similar to astroglia, we observed significant developmental changes in Iba1+-ir, albeit in an opposite direction; its level in both the hippocampus and PVWM was significantly decreased from P1 to P5. Consistent with this observation, a decrease in hippocampal microglia number was previously shown in rat pups within the first four postnatal days [96] and in prenatal pigs [20]. The early postnatal decrease in Iba1+-ir might be indicative of the maturation process of microglial cells, whose phenotype is known to evolve from amoeboid (activated and/or proliferating) to mature ramified shapes in the perinatal period [15].
Due to the limited access to biological samples in infants, animal models may provide understanding of underlying organ-specific, pathophysiological mechanisms of prenatal insults. However, there are also limitations with our CA preterm pig model as compared to human CA. In human, CA is normally induced by bacterial colonization/infection that usually starts in the mother's urinary tract, which is usually associated with many maternal symptoms, such as fever, high plasma leukocytes, and tachycardia [97]. In our pig model, the LPS was injected directly into the amniotic fluid which induced mild CA and extensive intra-amniotic inflammation for the fetuses, however, very limited maternal symptoms [42]. Moreover, even though LPS is a canonical trigger in animal models of CA [13,[98][99][100], CA in humans can be induced by multiple common organisms, including Ureaplasma (47%), Mycoplasma (30%), Gardnerella vaginalis (25%), bacteriodes (30%), gram negative rods including Escherichia coli (8%), and group B Streptococcus (15%) [97]. Thus, our model might only mimic the condition induced by gram-negative bacterial infections.
Overall, our data indicate that a brief period of in utero LPS exposure induces a moderate, yet significant, change in CSF protein levels at birth, coupled with later postnatal histopathological region-dependent alterations in the developing pig brain. The changes in hippocampal gene and protein expression indicate transient neuroinflammation and an adapted synaptic plasticity response. The detected CSF proteins may serve as biomarkers of later structural brain alterations in infants exposed to a relatively short period of inflammation before preterm birth. Such biomarkers may help to identify preterm infants eligible for therapies to improve recovery from these CA-related neurological sequelae. The preterm pig is a suitable model to further define how different times, durations, and types of prenatal inflammation may affect the developing immature brain and how early biomarkers in biofluids, such as plasma, CSF, or urine, could be important for early diagnosis, preventive treatments, and early therapies.