Post-hypoxia Invasion of the fetal brain by multidrug resistant Staphylococcus

Herein we describe an association between activation of inflammatory pathways following transient hypoxia and the appearance of the multidrug resistant bacteria Staphylococcus simulans in the fetal brain. Reduction of maternal arterial oxygen tension by 50% over 30 min resulted in a subseiuent significant over-expression of genes associated with immune responses 24 h later in the fetal brain. The activated genes were consistent with stimulation by bacterial lipopolysaccharide; an influx of macrophages and appearance of live bacteria were found in these fetal brains. S. simulans was the predominant bacterial species in fetal brain after hypoxia, but was found in placenta of all animals. Strains of S. simulans from the placenta and fetal brain were equally highly resistant to multiple antibiotics including methicillin and had identical genome sequences. These results suggest that bacteria from the placenta invade the fetal brain after maternal hypoxia.

17 ± 1 to 10 ± 1 mm Hg, respectively. This is a severity of hypoxia similar to that observed by acute exposure to approximately 3600 m elevation 6 .
In the hypothalamus, 357 and 280 genes were down-and up-regulated, respectively, 24 h after a 30 min period of hypoxia, compared to the normoxic control (Fig. 1). In the hippocampus, 240 and 270 genes were down-and up-regulated, respectively, after hypoxia (Fig. 2). The biological processes associated with genes upregulated by hypoxia included lipopolysaccharide response and binding, cytokine-mediated signaling pathway, positive regulation of cell migration in hypothalamus, and innate immune response, cytokine stimulus response, and lipid response in hippocampus (Tables S2, S3). Molecular functions included interleukin-6 binding activity, chemokine receptor binding, glucocorticoid receptor binding in hypothalamus, and lipopolysaccharide binding, cytokine receptor activity, and glucose transmembrane transporter in the hippocampus (Tables S4, S5). Significant associations were revealed with MAPK signaling pathways, cytokine-cytokine receptor interaction, hematopoietic cell lineage, and toll-like receptor signaling pathway in hypothalamus, and cytokine-cytokine receptor interaction, hematopoietic cell lineage, and chemokine signaling pathway in hippocampus (Tables S6, S7).
Given the above associations, the relationship between hypoxia and inflammation was explored using qPCR, including changes in expression of other genes that were not on the array, but would be predicted to also be increased if inflammatory pathways were activated. Hypoxia increased gene expression in inflammation-related pathways in hypothalamus including that of TLR2, CD14, NFKBIA, IL1B, CXCL10, PTGS2, and CASP8 (Fig. S1). Similarly, in hippocampus, expression of TLR2, TLR4, NFKB, MYD88, CXCL10, and CASP8 increased (Fig. S2). Apoptosis-related pathways (e.g., NIK, AKT, APAF1, BAX, BAD, CASP3) were not upregulated.

Evidence of increased microglia/macrophage abundance in fetal brain
While it was not surprising to find signs of inflammation after hypoxia, it was surprising that the pathways included toll-like receptors, TLR2 and TLR4. This suggested the presence of microglia or macrophages. To test the possibility that infiltration of the hypothalamus and hippocampus with microglia or macrophages might be causing this inflammation, sections of hypothalamus and hippocampus were immunostained for Iba-1, a marker of microglia and macrophages. Hypoxia increased the number of iba-1 positive cells in both hypothalamus and hippocampus (Fig. 2). This result parallels that previously found in cerebral cortex 4 and kidney cortex 7 . Immunostaining for the adrenergic receptor P2Y 12 , a more specific marker for microglia 8 , revealed increases

Evidence for bacteria in fetal brain
To test whether the fetal brain inflammation was accompanied by a bacterial invasion, gram staining was performed as well as qPCR of 16S rRNA using bacterial universal primers that do not amplify eukaryotic genes. The hypothalamus, hippocampus, and cerebral cortex of hypoxia-exposed fetuses were all found to have significantly more bacterial 16S rRNA 24 hours after transient hypoxia compared to the normoxic control (Fig. 3). As one potential route for entry into the fetal brain, hypoxia increased vascular permeability as measured by the effusion of albumin from the vascular space to the interstitial space (Fig. S3).
Analysis of the relative abundance of bacterial taxa as represented in 16S rRNA and on the number of 16S rRNA gene cop ies in each sample showed that Staphylococcus simulans was the most common species in the cerebral cortex (Fig. 3). The abundance of bacteria in the fetal cerebral cortex of normoxic fetuses was much lower, but not absent. Histological analysis of fetal cerebral cortex, hippocampus, and hypothalamus revealed the presence of gram-positive bacteria in both control (normoxia) and hypoxia-exposed animals, although the abundance was very low in the animals from the normoxia group. Bacteria with morphology consistent with Staphylococcus were observed in extravascular macrophages, especially in the sub-pial regions (Fig. 4). Bacteria present in the brain parenchyma appeared to be clumped and many were clearly intracellular within cells that were similar in size and shape to activated microglia (Fig. 4).

Presence of bacteria in placenta
Staphylococcus simulans was also most abundant in placentas of both normoxic and hypoxic ewes. To test for similarity of cortical and placental populations of bacteria, Staphylococcus simulans strains were cultured from a hypoxia-exposed placenta and fetal cerebral cortex and tested for antibiotic resistance. Strains from both locations were coagulase-negative and found to be equally resistant to ampicillin and kanamycin (both at 500 μg/ ml), vancomycin (250 μg/ml), ciprofloxacin (100 μg/ml), chloramphenicol, erythromycin, gentamicin, oxacillin, rifampicin, spectinomycin, tetracycline, trimethoprim (all at 50 μg/ml), and methicillin (30 μg/ml). The genomes of S. simulans cultures from the placenta and fetal cerebral cortex of a hypoxia-treated ewe were identical at the nucleotide level (Fig. 5), suggesting that a common source. Alpha and beta diversity of the bacteria in brain and placenta were calculated for hypoxic and normoxic treatments (Fig. 6). The normoxic placental samples possessed a significantly higher number of bacterial taxa than did the hypoxic placental samples. No bacteria were observed in the cerebral cortex of the normoxic samples. In contrast, bacteria were observed in the hypoxic cerebral brain tissues and the number of taxa was considerably lower than in the hypoxic placenta.

Discussion
The fetus is traditionally considered sterile 9 . In many cases of preterm birth, there is invasion of bacteria into the fetus, thought to be the result of an ascending vaginal infection 10 . A diverse bacterial population has been reported in the human placenta that more closely matches the diversity found in the human mouth 11 . In newborn infants, bacterial DNA is found in meconium (first stool) 12 , especially among preterm infants. The present study in sheep indicates that bacteria can become mobilized to the fetal brain as a consequence of hypoxic episodes. These results represent a sea change in our understanding of the presence of a microbiome in the fetus; the data strongly suggests that hypoxic stress led to invasion of the fetal brain by S. simulans from the placenta. Preterm reterm neonates are especially vulnerable to polymicrobial bloodstream infections, 70% of which arise from coagulase negative Staphylococcus 13 . Prior to birth, the neonatal adaptive immune response lacks robustness towards bacterial invasion. Based on our findings in preterm fetal sheep exposed to hypoxia, it is probable that the high abundance of S. simulans in both the cerebral cortex and placenta is directly correlated with a transiently suppressed or insufficient host immunological response after hypoxia. There are no reports of S. simulans crossing the blood-brain barrier but such reports do exist for S. aureus 14,15 . Regardless, the late gestation fetal blood brain barrier is likely to be far more permeable than that from a young lamb 16 . Our data suggest that maternal and/or fetal hypoxic stress mobilizes bacteria from the placenta or other tissue, allowing entry into into the fetal brain. Both maternal cells (maternal microchimerism) and virus transit the placenta 17,18 ; hypoxia may increase the transit rate of bacteria and/ or increase delivery of bacteria to the brain as a consequence of the increased brain blood flow, as well as the increased permeability of the brain 19 . The high level of antibiotic resistance from the cultured S. simulans strains may be the result of selection following five days of ampicillin treatment of the ewe prior to the study. Future experiments are needed to confirm the source of these bacteria and the identity of the exposures that led to antibiotic resistance in these bacteria. In addition, it is not known whether maternal stresses other than hypoxia can lead to Staphylococcus invasion into the fetus, or if these events would lead to premature birth in sheep or humans. Commensurate populations of bacteria on maternal skin or in maternal mouth, vagina, mammary glands, or gastrointestinal tract may comprise a background of exposure of the pregnancy to bacteria. Maternal stress -which causes an increase in maternal cortisol secretion -may transiently reduce the ability of the maternal immune system to clear the small numbers of bacterial cells that would otherwise be cleared rather than transit the placenta to the fetus. If this were true, the specific genus, species, and strain of bacteria found in the fetus would be informed by the maternal microbiome. In the present experiments, we found Staphylococcus simulans, a bacterium commonly found in mouth 20 , on skin 21,22 , and in mammary gland 23 of various species. One might expect that other species or populations of animals might respond to hypoxia similarly, but with other species of Staphylococcus or other taxa altogether, gain depending on the commensurate bacteria resident in the pregnant mother.
It is perhaps logical to assume that inflammation in the fetal brain is damaging, or that inflammation might have negative consequences with regard to programming of adult disease. Cortical white matter damage, for example, is common in infants suffering birth asphyxia 24 , and is assumed to result from direct cellular oxygen deprivation. In our study the relatively brief period of hypoxia was not associated with activation of apoptotic pathways, suggesting there is not rapid cellular damage due to the bacteria. Microglia and macrophages are also important participants in tissue development and maturation. For example, microglia are a critical part Figure 5. Circular map alignment of the genomes from S. simulans strains MR3 (from brain) and MR4 (from placenta). Alignment identity calculations were performed using the nucleotide BLAST algorithm implemented through BRIG. The inner black ring represents GC content for the sequences. Percent identity between the sequences is color coded at three percentage levels as shown in the legend (red for MR3, blue for MR4). Antibiotic resistance genes and their locations, as predicted by RAST, surround the map in green and are consistent between the genomes. Within the graphic is the average nucleotide identity between the genomes. Sequence identity is displayed as both a percent and a bit-score distribution. Calculations were performed in 1 kb windows, both one and two way, with a 95% identity threshold. of maintaining the balance between cell division and cell death in the developing brain and loss of microglial function might be causative with regard to known conditions involving cellular overgrowth 25 . Thus, whether the biology of fetal brain inflammation has great relevance to the ultimate outcome of the offspring is not known. Our results suggest, however, not only that bacteria may access the brain after periods of hypoxia in late pregnancy, but that the bacteria might be influenced by maternal antibiotic administration. The study has implications for treatment of mothers and newborns. For example, does maternal antibiotic treatment increase the risk of antibiotic resistant bacteria in the placenta and the fetus? Is anti-inflammatory treatment the best treatment for newborns at risk for hypoxic ischemic CNS damage?

Methods
Sixteen chronically catheterized singleton or twin ovine fetuses were studied between the gestational age of 122 ± 5 days (full term = 145-147 days). All experiments were approved by the University of Florida Animal Care and Use Committee. All laboratory procedures were performed in accordance with relevant guidelines and regulations.
After 24 h of fasting of the pregnant ewes, fetal surgery was performed on 116 ± 3 days of gestation as previously described 4 . After recovery from surgery, ewes and their fetuses were studied while conscious and freestanding in their pens with access to food. Experimental design was as previously described 4 . In this report, we present the results from the normoxic control (NC) and hypoxic control (HC) groups (n = 4/group). Blood gases and fetal cardiovascular and neuroendocrine responses to hypoxia were reported previously 5 .  (Shannon -right panel). Beta diversity compares the relationship of the community diversity between samples and is depicted as a principal component analysis plot (PCoA) using the weighted UniFrac distance measure. Bacterial diversity in cerebral cortex and placenta of fetal sheep exposed to normoxia or hypoxia.
SCIEntIFIC REPORTS | 7: 6458 | DOI:10.1038/s41598-017-06789-6 Fetuses were sacrificed 24 h post hypoxic stress, and various fetal tissues were snap frozen and stored at −80 °C. At sacrifice, fetal brain was dissected, processed and immunostained for Iba-1 as described previously 4,6,9 and for P2Y 12 using antiserum HPA014518 (1:200) 8 . Microarray analysis was performed and analyzed as previously described 26 . Real-time qPCR. RNA was extracted as previously described 4 . The same mRNA samples used for the microarray was also used for qPCR validation. The primers for measurement of ovine gene expression were designed based on the known Ovis aries and Bos taurus genomes (Table S1) for Taqman or Sybr green chemistry (Thermo-Fisher Scientific, Waltham, MA). Bacterial 16S rRNA was measured using bacterial universal primers 27 and 16S rRNA copy number determined using a standard curve of S. simulans DNA.
DNA extraction, 16S rRNA gene amplification and sequencing and diversity analysis. Placental and cerebral cortex DNA was extracted using the DNEasy Blood & Tissue Kit (Qiagen) following the manufacturer's guidelines. DNA was extracted and quantified spectrophotometrically, amplification of 16S rRNA genes, and barcoded sequencing of the resulting amplicons was done as described previously 12,28 . Diversity analyses were calculated and plots designed using phyloseq in R 29 .
Bacterial culturing and genome sequencing from brain and placental tissues. Tissues were mechanically homogenized in 300 μl of Brain Heart Infusion Broth (BHI, Sigma-Aldrich) for 30 seconds from which 10-fold serial dilutions were prepared and spread on BHI solid medium. Following incubation at 37 °C for 24 h, selected colonies were purified and their identity determined by amplifying the16S rRNA gene using 8 F and 1492 R primers followed by Sanger sequencing. Whole genomes from selected strains of Staphylococcus simulans were sequenced and using the PacBio RS II platform as described previously 30 . Genome annotation was performed using the NCBI Prokaryotic Annotation Pipeline 31 after submission to GenBank. Whole genome sequence comparison and circular map generation was done using BRIG 32 . One and two-way average nucleotide identity was calculated using the method of Goris et al. 33 . The genome sequences and raw sequence data are archived in Genbank accession numbers CP017428 and SRR4340965, respectively for S. simulans strain MR3; and CP017430 and SRR4340966, respectively for S. simulans strain MR4.
Bacterial diversity. Filtered reads were grouped into Operational Taxonomic Units (OTUs), which were then classified at the phylum level with "R" statistical software. Relative differences in clustering amongst placenta and cerebral cortex bacterial communities were calculated using the Bray-Curtis Index and were then graphed by Principal Component Analysis (PCoA) in order to illustrate beta diversity in the sample set. In this case, shorter distances between plotted points indicated an increase in similarity in terms of microbial composition, where UniFrac distances were calculated based on shared branches within the phylogenetic tree.
Alpha diversity (Shannon-Wiener Index) was calculated to estimate the species richness and biodiversity of bacterial communities in the fetal placenta and fetal cerebral cortex, while also taking into account treatment with either hypoxia or normoxia. Overall, our results mirror the microbial colonization patterns observed from 16S rRNA sequencing, transcriptomics, and qPCR analyses, lending evidence that there is higher diversity in placental samples and considerably lower diversity in the cerebral cortex, while also indicating higher abundance of bacterial species in normoxic samples.
Antibiotic Resistance Screening of S. simulans isolate 622B. Sensitivity of S. simulans cultures from brain and placenta to ampicillin, methicillin, oxacillin, kanamycin, vancomycin, tetracycline, rifampicin, gentamicin, chloramphenicol, ciprofloxacin, spectinomycin, trimethoprim, and erythromycin (Sigma Aldrich) was tested in BHI broth with antibiotics at various concentrations in 96-well plates. Viability of cells post-antibiotic treatment was tested by addition of 20 μL PrestoBlue.
Calculations and Statistical Analysis. Data are presented as mean values ± standard error of the mean (SEM). The ovine Agilent 15.5k array results were analyzed with R Studio's Limma package for R software v.2.15.1, employing moderated t-test using empirical Bayes method for small sample size per group as previously described 34 . The qPCR data were analyzed Student's t-test 35 . Gene networks were inferred using the GeneMania plugin of Cytoscape 36 . Gene functional annotation was analyzed using WEB-based GEne SeT AnaLysis Toolkit (WebGestalt) 37,38 . In all tests, the criterion for statistical significance was p < 0.05. Data availability. The datasets generated during and/or analyzed during the current study are available in the NCBI Gene Expression Omnibus repository (GSE82016 and GSE97916).