All experiments were performed in accordance with the ARRIVE guidelines 2.0 (Supplementary File 1; Percie du Sert et al., 2020). The use of animals was approved by Monash Medical Centre Animal Ethics Committee (MMCA/2020/15) and was conducted in accordance with the Australian Code of Practice for the care and use of Animals for Scientific Purposes established by the National Health and Medical Research Council of Australia.

Pregnant Border-Leicester ewes carrying singletons or twins were utilized in this study. At 110 days of gestation (dGA; term is ~148 dGA), after withdrawal of food for at least 18 h, ewes were anaesthetized with intravenous sodium thiopentone (20 mg·kg⁻¹) and intubated. General anesthesia was maintained by inhalation of 1.5–3.5% isoflurane in oxygen. Under sterile conditions, a midline laparotomy was performed on pregnant ewes to expose the fetal head, neck and left forelimb. In the case of a twin pregnancy, only one fetus was operated on. Instrumentation of pregnant ewes and their fetuses has been previously described in Vidinopoulos et al. (2023). Briefly, a tracheostomy was performed to secure a modified reinforced endotracheal tube in the lower trachea and was connected to a saline filled large-bore ventilation tube. A separate saline-filled catheter was inserted into the upper trachea and connected to the ventilation tube externally to create an exteriorised tracheal loop, to allow normal flow of lung liquid. The left brachial artery and left jugular vein were catheterised for serial arterial blood sampling and antibiotic administration, respectively. The fetus was returned to the uterus and the catheters and tracheal loop were exteriorised via the ewe’s right flank. Postoperative analgesia was maintained for 3 days via a transdermal fentanyl patch on the left hind leg of the ewe (75 μg·h⁻¹; Jansen Cilag, North Ryde, NSW, Australia). Antibiotics were administered i.v. to the ewe (ampicillin, 800 mg and engemycin, 500 mg) and the fetus (ampicillin, 200 mg) for 3 consecutive days after surgery. Three to 5 days of post-operative recovery were allowed prior to commencing the experiment.

At 113–115 dGA, fetuses were randomly assigned to four groups: unventilated controls without (UVC; n = 7) or with intratracheal LPS (UVC + LPS; n = 7); or ventilated controls with (VENT + LPS; n = 7) or without intratracheal LPS (VENT; n = 8). Ventilation tubing was disconnected, and lung liquid passively drained. To induce localized acute inflammation, LPS (Escherichia coli, 055:B5, Millipore Sigma, MO, USA; 1 mg in 2 mL saline) was infused into the tracheal tube 1 h before the onset of in utero ventilation (Polglase et al., 2009). For UVC and VENT fetuses, 2 mL of saline was administered. After 1 h, the endotracheal tube was connected to a neonatal ventilator (Babylog 8000+, Dräger, Lübeck, Germany) and ventilated (Vidinopoulos et al., 2023). Both VENT and VENT + LPS fetuses were ventilated in utero with non-humidified air with a peak inspiratory pressure (PIP) 25–40 cmH₂O targeting a tidal volume (V_T) between 3 and 5 mL·kg⁻¹, positive end expiratory pressure (PEEP) 4 cmH₂O, flow 10 L·min⁻¹, 60 inflations·min⁻¹ and FiO₂ 21% for 24 h. The in utero ventilation technique in fetal sheep is an established model to investigate the mechanisms of ventilation-induced injury independent of haemodynamic instability involved with the cardiorespiratory transition and removes potential confounders that occur when ventilating preterm neonates for extended periods of time ex utero, including: maintaining oxygen, cardiovascular support, nutrition, temperature and corticosteroid exposure (Blanco et al., 1987; Allison et al., 2008). Further, it enables the mechanical ventilation of preterm fetal sheep at a much younger gestation than what would be viable ex utero, allowing us to investigate the sheep brain that is comparable to a preterm infant (Back et al., 2006).

Fetal arterial blood was sampled for blood gas measurements (ABL80 FLEX, Radiometer Medical ApS, Denmark) before LPS/vehicle infusion (Pre-LPS), before ventilation (Pre-Vent), +15, +30, +45, +60 min post-vent and +3, +6, +9, +12 h, +24 h post-vent. Pre-vent, +3, +6 and, +12 h post-vent plasma samples were collected for IL-6 analysis. At 24 h, the ewe and fetus were humanely euthanised with an intravenous overdose of sodium pentobarbitone (100 mg·kg⁻¹ i.v.; Valabarb Euthanasia Solution; Jurox, NSW, Australia) via the maternal jugular vein catheter. UVC fetuses were instrumented and received intratracheal vehicle/LPS but were not ventilated and were euthanised at the same age as the VENT groups. At post-mortem, the fetal brain was removed, weighed and hemisected. The left hemisphere was dissected coronally to obtain a ~1 cm block at the level of the ansate sulcus. The subcortical white matter (SCWM) was collected via microdissection of the white matter within the 1st and 2nd gyri, samples were pooled and immediately snap frozen in liquid nitrogen. Similarly, cortical gray matter (GM) of the 1st and 2nd gyri were pooled and immediately snap frozen in liquid nitrogen. The periventricular white matter (PVWM) was located as the white matter surrounding the ventricle above the subventricular zone, tissue was microdissected, collected and snap frozen. All collected tissue was stored at −80°C for RT-qPCR analysis. The right hemisphere was immersion fixed in 10% neutral-buffered formalin for immunohistochemical analyses.

Arterial blood was collected via the fetal brachial artery catheter before ventilation (Pre-Vent), and at 3, 6, and 12 h for assessment of plasma IL-6 using a sandwich enzyme-linked immunosorbent assay (ELISA) assay as described previously (Galinsky et al., 2020). Briefly, plates were read on a SpectraMax i3 microplate reader (Molecular Devices, CA, USA) at 450 nm to determine optical density. Standards (recombinant ovine IL-6; Kingfisher Biotech, MN, USA) were included and a standard curve was generated for every ELISA plate used (R² > 0.99). Due to a freezer malfunction, some plasma samples could not be assessed. The final group numbers for plasma IL-6 analysis were as follows, UVC, n = 6; UVC + LPS, n = 7; VENT, n = 6; and VENT + LPS, n = 5.

RNA extraction, cDNA preparation and analysis were conducted as described previously (Tran et al., 2023). Briefly, RNA was extracted from frozen brain tissue (20–30 mg) using an RNA extraction kit (RNeasy Mini Kit, Qiagen, Germany) following the manufacturer’s instructions. RNA yield was determined by spectrophotometry (Nanodrop, Analytical Technologies, Biolab). cDNA was transcribed from RNA, then pre-amplified at 50 ng·μL⁻¹ (SuperScript^® III First-Strand Synthesis System for RT-PCR kit; Invitrogen). Gene expression was analyzed using a Fluidigm Dynamic array Biomark HD system (Fluidigm, USA). Gene expression of 8 genes of interest (Table 1) was determined by relative expression calculated by change in cycle threshold (ΔCt) between each gene of interest and the geometric average of two endogenous housekeeping genes, YWHAZ and RPS18. These genes were selected based on markers identified form previous clinical and preclinical studies investigating inflammatory markers in the context of intrauterine inflammation and mechanical ventilation (Bohrer et al., 2010; Polglase et al., 2012; Barton et al., 2016; Stojanovska et al., 2022). Levels of mRNA expression relative to geometrical average of house-keeping genes were determined using the 2^−ΔΔCT method (Livak and Schmittgen, 2001) and expressed relative to the UVC group.

Immersion-fixed fetal brains were paraffin embedded and then microtome sectioned at 8 μm corresponding to section 720 according to the Michigan State University Sheep Atlas (Johnson et al., 2013; Figure 1). Two sections per animal per antibody were utilized. Immunohistochemical staining was conducted using primary antibodies: rabbit anti-neuronal nuclei (NeuN; 1:200; Abcam, UK; CAT#: 177487) for mature neurons; rabbit anti-ionized calcium binding adaptor molecule 1 (Iba-1; 1:250; Abcam, UK; CAT#: ab178846) for microglia; rabbit anti-glial fibrillary acidic protein (GFAP; 1:200; Abcam, UK; CAT#: ab68428) for astrocytes; rabbit anti-Olig-2 (1:200; Abcam, UK; CAT#:ab109189) for oligodendrocytes. Immunohistochemical staining protocol was conducted as described previously (Tran et al., 2023). Briefly, tissue sections were subjected to antigen retrieval using citrate buffer (10 mM Tri-sodium citrate in dH₂O, pH 6.0; Sigma Aldrich), PBS washes, endogenous peroxidase blocking, and overnight incubation in primary antibodies at 4°C in primary diluent of 3% normal goat serum in PBS. Sections were incubated in secondary goat biotinylated anti-rabbit IgG for 2 h (1:200; Vector Laboratories, UK; CAT#: BA-100) then incubated with avidin-biotin complex (ABC Elite kit; 1:1:200 in PBS; Vectastain^®, Vector Laboratories, UK) and visualized with 3,3′-diamniobenzidine solution (DAB; MP Biomedicals, Australia).

Prior to analyses, all slides were coded, and assessors (NTT, AS) were blinded to the treatment group. Slides were scanned at 40× magnification using Aperio Scanscope AT Turbo (Leica Biosystems, Germany). Regions of interest included the cortical gray matter (GM) and subcortical white matter (SCWM) of the first and second parasagittal gyri (i.e., GM 1st and 2nd gyri, SCWM 1st and 2nd gyri), and the periventricular white matter (PVWM) (Figure 1). For each region, 2 fields of view (FOV; 410 μm × 320 μm) were analyzed with FOV placement kept consistent across all antibodies. All analyses were averaged across a total of four FOV (two sections per subject with two FOV per region).

Total cell density for all antibodies was conducted and expressed as cells/field. For analysis of total cell density, cells were identified by their immunopositive densely stained round somas with diameters >6 μm. Specifically for Iba-1 analysis, total cell density was manually counted as cell bodies with immunopositive staining irrespective of the morphology of the processes, and ameboid microglia were identified by characteristic enlarged and round, densely stained soma with resorbed processes (Kettenmann et al., 2011). To assess area coverage of GFAP immunostaining, expressed as % area coverage, each FOV was exported from Aperio and the image was processed using ImageJ software (version 2.0.0-rc-69/1.52p, National Institutes of Health) and an optimized set threshold was used to calculate area coverage for all images using ImageJ. The optimized set threshold was conducted by calculating an average threshold range of 20 randomly selected FOV that would allow optimal detection of positive staining. The optimized threshold was then set for all slide assessments and the ‘area fraction tool’ in ImageJ was used to automatically quantify the percentage of positive immunostaining in each FOV. Microglial aggregations were also identified as dense clusters of positive staining and the aggregated area expressed as a percentage of the total brain regions of interest (SCWM, GM, PVWM).

All statistical analyses were conducted using GraphPad Prism (version 10.1.0; GraphPad Software, CA, United States). Data were assessed for normality using the Shapiro–Wilk Test. Data for animal characteristics and immunohistochemical were assessed using a parametric two-way ANOVA and assessed for the main effects of LPS (p_LPS), ventilation (p_VENT) and interactions between LPS and ventilation (p_{LPS X VENT}). Post-hoc analysis of significant interactions and group effects was used to determine differences between groups using a Tukey’s multiple comparisons test. For mRNA expression analysis, data was non-parametric and was log10-transformed to conduct two-way ANOVA analyses as above. For plasma IL-6 and blood chemistry measurements, a parametric three-way ANOVA was used to assess the main effects of LPS (p_LPS), ventilation (p_VENT) and time (p_TIME). Post-hoc analysis was performed on significant interactions using Tukey’s multiple comparison test or a Bonferroni’s multiple comparisons test where a significant main effect was observed. Data are presented as mean ± SD. A p < 0.05 was considered statistically significant for all analyses.

Fetal characteristics at post-mortem (112–119 dGA) of the four groups are presented in Supplementary Table 1. LPS fetuses were older and had higher body weights than non-LPS fetuses (p_LPS = 0.024). There were no differences in male to female ratios. Due to the small numbers for each group, sex differences could not be assessed.

Between 6 and 24 h of ventilation, arterial pH and PaO₂ decreased and PaCO₂ and lactate increased in VENT + LPS fetuses compared to both UVC and VENT fetuses (Figures 2A–E and Supplementary Table 2). Arterial pH was significantly lower in VENT + LPS fetuses at 12 h compared to UVC + LPS fetuses (p = 0.011; Figure 2A).

Plasma IL-6 levels in VENT + LPS fetuses were significantly higher compared to both UVC and VENT fetuses 12 h after starting ventilation (p = 0.014 and p = 0.017, respectively; Figure 3).

The effect of LPS and in utero ventilation on gene expression in the cortical GM, PVWM and SCWM of the fetal sheep brain are summarized in Figure 4.

In LPS-exposed fetuses, numbers of Iba-1-positive cells were increased in the SCWM (1st and 2nd gyri) and GM (1st gyri) (p_LPS = 0.010; p_LPS = 0.006; p_LPS = 0.003, respectively; Figures 5A,C). The increase in numbers of microglia in LPS-exposed fetuses was greatest in the VENT + LPS fetuses compared to VENT alone (SCWM 1st gyri: p = 0.021; SCWM 2nd gyri: p = 0.016 and GM 1st gyri: p = 0.037). In LPS-exposed fetuses, numbers of ameboid microglia were increased within the GM (1st gyri) (p_LPS = 0.044; Figures 5B,C).

Microglial aggregation area within the white and gray matter regions did not differ between groups (all p > 0.05; Supplementary Table 3).

In ventilated fetuses, numbers of astrocytes were significantly increased in the PVWM (p_VENT = 0.03; Figure 6). No other changes to astrocyte cell density or area coverage were found.

Numbers of Olig2-positive oligodendrocytes and NeuN positive neurons did not differ between groups for any of the white and gray matter regions examined (all p > 0.05; Figure 7).

The ventilatory strategy used in this study was one that aimed to simulate a gentle cardiorespiratory and cerebrovascular transition by using lower tidal volumes. While we did demonstrate that this ventilation strategy, in the absence of LPS, induces minimal neuroinflammation at 24 h, the inflammatory consequences following mechanical ventilation after intratracheal LPS exposure amplifies systemic and central nervous system pathways of neuroinflammation. We have shown previously that irrespective of the ventilatory strategy, continuous positive airway pressure (CPAP) or mechanical ventilation (PPV), prior inflammatory exposure resulted in similar degrees of inflammation within the lungs and the circulation (Polglase et al., 2009). Collectively, these data suggest that 24 h of mechanical ventilation per se may not be a major influence on neuroinflammation; but instead, what is critical is the degree of systemic and central nervous system inflammation at the time of birth. Overall, these data suggest that targeting systemic and central nervous system inflammatory pathways are likely to be key for improving interventions for preterm neuroprotection (Kelly et al., 2023).