All animals were treated according to current guidelines for animal well-being, and experiments were approved by the Ethics Committee for Animal Experimentation of the Faculty of Medicine (P080/2019). Experiments are reported according to ARRIVE guidelines63. Time-mated rabbits (hybrid of Dendermonde and New Zealand White) were housed in individual cages at 21˚C, 42% humidity, with a 12-hour day/night cycle and free access to food and water. Conception day was considered day 0 of pregnancy (GA = 0). At GA 25 (full term 31.5 days), does underwent induction of FGR. Briefly, rabbits were administered induction anesthesia with IM ketamine (35 mg/kg Nimatek®, Eurovet Animal Health BV, Bladel, The Netherlands) and xylazine (5 mg/kg XYL-M® 2%, VMD, Arendonk, Belgium), and antibiotic prophylaxis (10 mg/kg Baytril® 2.5% SC, Bayer, Diegem, Belgium), tocolysis (10 mg/kg Depo-Provera® SC, Pfizer, Puurs, Belgium), and analgesia (0.03 mg/kg Vetergesic® SC, Ceva Animal Health, Brussels, Belgium) prior to surgery. Anesthesia was maintained with a continuous IV infusion of ketamine (8-16mg/kg/h) and xylazine (2.4-4.8mg/kg/h), while monitoring vital signs. Following laparotomy, 30–40% of the vessels going to each placenta were ligated in one random horn with Vicryl® 5 − 0 (Ethicon®, Diegem, Belgium), leaving the contralateral unligated horn as internal control. The abdomen was closed with Vicryl® 2 − 0 and Monocryl® 3 − 0 (Ethicon®, Diegem, Belgium) for fascia and skin, respectively. The surgical wound was infiltrated with levobupivacaine (2 mg/kg Chirocaine®, Abbvie, Wavre, Belgium) and sprayed with aluminium (Kela®, Hoogstraten, Belgium).
Rabbits were monitored daily until delivery by caesarian section at GA 30. Following delivery, does were euthanized using IV phenytoin/pentobarbital (140 mg/kg Euthasol®, Kela, Bladel, The Netherlands). Kittens were pet dried, numbered with a permanent marker, and kept in a warmed (34°C) and humidified (55% RH) incubator. Later that day they were stimulated to urinate, weighed, and fed a commercial milk substitute (Day One, protein 30%, fat 50%; Fox Valley, Lakemoor, IL) with added probiotics (Bio-Lapis; Probiotics International, Somerset, UK) and immunoglobulins (Col-o-Cat; SanoBest, Hertogenbosch, The Netherlands). The following morning litters were allocated for brain or lung assessment.
At the time of caesarean delivery rabbits were anesthetized with the same protocol as for FGR induction, put on a warming pad under a warming light, and uterine horns were exposed and continuously rinsed with warm saline (37˚C). The VisualSonics VEVO 2100 (Toronto, Ontario, Canada) high-resolution micro ultrasound platform and a VisualSonics MS-250 transducer were used for image acquisition (center frequency 21 MHz; bandwidth 13–24 MHz; geometric focus 15 mm; maximum image width 23 mm; maximum image depth 30 mm; footprint 28 x 5.75 mm). Flow velocity waveforms of the umbilical artery were obtained in both ovarian end fetuses by locating the umbilical vessels using color Doppler and placing the pulsed Doppler sample gate over the umbilical artery, keeping the angle to a minimum. The pulsatility index (PI) was calculated offline using the VisualSonics analysis software.
After delivery, placentas (until and including the decidua) were carefully separated from the implantation sites in the uterine wall, washed in PBS, trimmed from umbilical cord and membranes, and immerse-fixed in 4% paraformaldehyde (PFA) for 72h, blotted dry and weighed. The smallest lobe was sectioned in 2 coronal portions, embedded in paraffin blocks, and cut in 4µm slides. 2 slides per placenta were then double stained for cytokeratin and lectin to assess trophoblast and fetal capillaries. The detailed staining protocol is added in the Supplementary materials.
All slides were digitally scanned with the Zeiss AxioScan Z1 imaging platform (AxioScan® Slide Scanner, Carl Zeiss MicroImaging GmbH, Munich, Germany).
Placental zones (labyrinth, junction, decidua) were manually delineated using QuPath software64 and their area relative to the total placental area was calculated. The structures within the labyrinth zone were manually quantified as described before65. Briefly, using a 4x4 point counting grid on 16 random fields at 20x magnification, the number of maternal blood spaces, fetal capillaries, and trophoblast cells was quantified and expressed as a percentage of the total count. A blinded observer (I.V.) performed all histological evaluations.
Placental computed microtomography
A subset of rabbits was used for fetoplacental perfusion with a contrast agent under the same anesthesia protocol at GA30. Following laparotomy, both uterine horns were removed and immediately placed on ice. Fetuses and placentas were exposed, and a 24G cannula was inserted in the umbilical vein of each fetus, while an umbilical artery was cut open to serve as vent. Initially, warm heparinized saline was perfused, followed by a barium-based contrast solution consisting of 0.9% sodium chloride solution, barium sulfate (E-Z PAQUE® High Density Barium Sulfate 98.75%, Bracco imaging, Milan, Italy), and porcine skin gelatin (Gelatin 48723, Sigma-Aldrich, Darmstadt, Germany), previously filtered and warmed to 60˚C. Perfusion was stopped when the contrast solution was clearly seen in the capillary bed of the placenta. Thereafter, placentas were separated from the conceptuses and immersed in 4% PFA for 24h, washed, and stored in PBS. A total of 16 placentas (eight FGR and eight controls) from four different litters were included for analysis.
X-ray computed microtomography scans were performed using SkyScan 1172 micro-CT system (Bruker microCT, Kontich, Belgium) with X-ray energy of 80 keV. The image pixel size was set to 8.98 µm and a 0.5 mm Al filter was applied. Projection images were captured between the rotation steps of 0.2°. Each projection was merged from 6 or 8 images averaged from 5 frames to minimize the image noise and captured with different field of view to cover the entire sample, resulting in 3872 by 3872–4956 pixels. The total scan duration was up to 15h.
For image reconstruction, GPU powered NRecon software (v.220.127.116.11, Bruker micro-CT) was used. Image processing and analysis was done by CT analyzer software (v.18.104.22.168, Bruker micro-CT,). The reconstructed image stacks were median filtered with 2 pixels circular kernel, and histogram based manual global threshold was applied. The distribution of vessels thickness, and parameters of volume and surface of vascular tree were calculated. Additionally, color-coding visualization of the local thickness was applied. An observer blinded to group assignment (B.L.) performed all evaluations.
On postnatal day 1 (PND1), kittens underwent a validated NBA protocol18,66. Short-term motor assessment comprised scoring of gait, posture, locomotion, head and limb activity, and activity duration. Afterwards, the cranial nerves, pain response, and righting reflex were tested for sensory evaluation. All assessments were filmed and later scored by an observer blinded to the groups (I.V.). A full description of NBA protocols can be found in the Supplementary materials.
Immediately after NBA, animals were deeply sedated with IM ketamine (35 mg/kg) and xylazine (6 mg/kg), and transcardially perfused with 0.9% saline + heparin (100 u/mL; 3 minutes at 30mL/min) followed by 4% PFA (5 minutes at 30ml/min). Their brains were removed from the skull and further immerse fixed in 4% PFA for 48h. A subgroup of kittens (n = 23) was perfused with heparinized saline, followed by 4% PFA and 2% (10 mM) gadoteridol (ProHance® 1mL of 279,3mg/mL solution). Their brains were kept inside their craniums for a maximum of 7 days to undergo MRI, after which they were removed and processed for histological assessment.
Ex vivo MRI was performed on perfused fixed brains using the active staining technique. Briefly, a Bruker Biospec 9.4 Tesla small animal MR scanner (Bruker Biospin, Ettlingen, Germany; horizontal bore, 20 cm) equipped with actively shielded gradients (600 mT/m) was used. Data was acquired using a 72mm internal diameter quadrature volume coil for transmission decoupled with a rat brain quadrature shaped surface coil for signal reception (volume resonator, Rapid Biomedical, Rimpar, Germany). Data was acquired with a high-resolution 3D Flash sequence (TE/TR 5.5/50 ms; flip angle 70 degrees; slice thickness 0.35 mm with no interslice gap, data matrix 392×392×392; isotropic resolution 89µm; 4 averages, acquisition time 1h15min) and a SE-EPI sequence (8 segments; TE/TR: 25/150ms; slice thickness 0.4mm with no interslice gap, data matrix 192×160×160; isotropic spatial resolution of 208 µm; 64 directions and 3 b-values per direction of 800, 1000, 1500 s/mm2, acquisition time 11h43min).
MRI volumes were processed to quantify microstructural properties in the following brain regions: frontal cortex, corpus callosum, caudate nucleus, putamen, internal capsule, corona radiata, hippocampus, thalamus, and hypothalamus. Automatic atlas-based parcellation was obtained by diffeomorphic registration of the atlas template in Ferraris et al 67 to the acquired structural images, and then translated to the diffusion space. Fractional anisotropy (FA) and mean diffusivity (MD) were estimated using dipy python libraries68,69, and their mean values in regions of interest were computed. An observer blinded to group assignment (E.M.) performed all evaluations. A complete description of MRI processing can be found in Supplementary materials.
Of 23 acquisitions (13 controls 10 FGR), 4 (2 in each group) were excluded from analysis due to important artifacts. Statistical comparison between groups was performed in one random cerebral hemisphere. An interim power calculation, based on the FA data in the hippocampus, confirmed a power of 95% to detect these differences.
Following fixation, brains were paraffin embedded and serially sectioned at 4 µm. Three sets of four serial coronal sections every 100 µm were taken at each of the following two levels, as previously described18: level 1 started at the medial septal nucleus and level 2 at the hippocampal formation.
Six slides per brain (three slides per level) were stained with Cresyl Violet (CV; C5042-10G; Sigma- Aldrich, Overijse, Belgium), and two slides per brain (one slide per level, eight slides in total) were incubated with each of the following four primary antibodies: mouse monoclonal anti-human Ki67 (M724001-2; Agilent, Diegem, Belgium), mouse monoclonal anti-glial fibrillary acidic protein antibody (GFAP) (G6171, Sigma-Aldrich, St Louis, MO, USA), anti-NG2 chondroitin sulfate proteoglycan antibody (MAB5384, Millipore, Billerica, MA, USA), or a terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) method for fluorescent in situ end labeling of double- stranded DNA fragmentation (Apoptag S7110; Millipore). The secondary antibody was Alexa Fluor® 488 goat anti-mouse conjugate (Invitrogen) or Alexa Fluor® 647 goat anti-mouse conjugate. Sections were counterstained with Hoechst 33342 (Sigma-Aldrich, Bornem, Belgium). Eight brain areas were assessed, namely, the frontal cortex, corpus callosum, caudate nucleus, putamen, hippocampus (CA1, CA3, dentate gyrus), and thalamus (anteroventral nucleus).
Details on image acquisition and quantification can be found in Supplementary materials.
Pulmonary function testing
Pressure-volume and forced oscillation maneuvers were performed via a tracheostomy on PND1 using the FlexiVent system (SciReq; FlexiVent, Montreal, QC, Canada). Animals were sedated with ketamine (35 mg/kg) and xylazine (6 mg/kg) before tracheostomy. An 18-gauge metal cannula was inserted into the trachea and secured with an airtight suture. Kittens were ventilated with a tidal volume of 10 mL/kg and positive end-expiratory pressure of 3 cmH2O at a rate of 120 breaths/min. Before lung function tests, two deep inflation maneuvers were performed until reaching a pressure of 30 cmH2O to maximally inflate the lungs and standardize lung volume. Both pressure-volume (inspiratory capacity, static compliance, and static elastance) and forced oscillation tests (tissue damping, tissue elastance, central airway resistance, respiratory system resistance, dynamic compliance, and dynamic elastance) were performed as previously described70. The mean of three separate measurements for each maneuver, with a coefficient of determination > 95%, was calculated and used as a single data point for analysis.
Histological lung assessment.
After PFT, lungs were removed en bloc via thoracotomy, the trachea was cannulated with a 20-gauge catheter, and the left lung was pressure fixed for 24h at a constant hydrostatic pressure of 25 cmH2O in 4% PFA70. After fixation, left lung volume was calculated using water displacement before paraffin embedding. Alveolar morphology was measured on digitally scanned 5µm hematoxylin and eosin (H&E)-stained slides using a semi-automated, validated Fiji-plugin (ImageJ) (http://fiji.sc/Fiji) that randomly selected 20 fields per lung71, according to stereological principles. Calculations of mean linear intercept (Lm), alveolar air space (Lma), and interalveolar septal thickness (Lmw) were made as previously described29. Vascular morphology was evaluated using immunohistochemistry. A primary α-smooth muscle actin (α-SMA) antibody (mouse anti-human, M0851; DakoCytomation, Glostrup, Denmark) was used in combination with a horseradish peroxidase-conjugated secondary antibody (goat anti-mouse, 115-035-044; Jackson ImmunoResearch, Ely, UK). Aminoethyl carbazole was used as a chromogen. A minimum of 10 pulmonary arteries with an external diameter of 30–100 µm per lung slide were examined, measuring both internal and external diameter of the media to calculate the vascular medial thickness27. An observer blinded to group assignment (D.B.) performed all pulmonary histological evaluations.
Data were analyzed using Prism 9 for MacOS (GraphPad Prism, San Diego, CA, USA). Data distribution was checked for normality using a Shapiro-Wilk normality test and presented as mean with standard deviation or median with interquartile range, as appropriate. Data comparison was done by Fisher’s exact, unpaired t-test or Mann Whitney U-test, depending on data characteristics. A p value of < .05 was considered significant. Adjustment for multiple testing was applied for DTI data using the Benjamini, Krieger and Yekutieli method with 10% FDR.