Preterm birth impairs postnatal lung development in the neonatal rabbit model

The animalium of KU Leuven provided pregnant rabbits (New Zealand White and Dendermonde hybrid). All experiments were approved by the Ethics Committee for Animal Experimentation (P060/2016 and P080/2017) and were performed according to local guidelines on animal welfare. Does were housed in individual cages with free access to food and water in a controlled 12 h/12 h light-dark cycle. For Cesarean section does were sedated with intramuscular ketamin 35 mg/kg (Nimatek®; Eurovet Animal Health BV, Bladel, The Netherlands) and xylazin 6 mg/kg (XYL-M®; VMD, Arendonk, Belgium). Subsequently they were euthanized with a mixture of 200 mg embutramide, 50 mg mebezonium, and 5 mg tetracain hydrochloride (IV bolus of 1 ml T61®; Intervet Belgium, Mechelen, Belgium). Immediately after death, the abdomen was opened and the uterus exposed to extract all pups through hysterotomy. After birth, pups were dried, stimulated and placed in an incubator (Dräger Incubator 7310; Dräger, Lübeck, Germany) at 32 °C, 50% relative humidity and 21% of oxygen. Pups were fed twice daily via an orogastric tube with a milk replacer (Day One®, Protein 30%, Fat 50%; FoxValley, Illinois, US) with additional probiotics, electrolytes and vitamins (Bio-Lapis®; Probiotics International Ltd., Somerset, UK) and immunoglobulins during the first 2 days (Col-o-Cat®, SanoBest, Hertogenbosch, Netherlands). The volume of feeds is increased from a total of 80 ml/kg body weight/day on day 0 of life to 200 ml/kg/day on days 3–7. On day 2, vitamin K1 is administered intramuscularly (0.25 mg/kg, Konakion pediatrique®; Roche, Basel, Switzerland), and from day 2 onwards, pups are given a daily intramuscular injection of benzylpenicillin (20,000 IU/kg Penicilline®; Kela, Sint-Niklaas, Belgium) and amikacin (20 mg/kg day, Amukin®; Bristol-Myers-Squibb). MicroCT and dark field imaging was performed under isoflurane anesthesia (2% isoflurane in 2 l/min of pure oxygen). Lung functions were terminal experiments performed in deeply anesthetized postnatal rabbit pups (ketamine 35 mg/kg and xylazin 6 mg/kg IM). Anesthetized animals were euthanized by exsanguination for dissection and lung harvest.

The study groups and timeline are illustrated in Fig. 1. Pregnant dams were randomized to either a Cesarean section on day 31 (term, alveolar stage of lung development) or day 28 (preterm, saccular stage). Of each group, randomly selected pups were harvested before their first breath: they constitute the fetal groups 28F (preterm, day 28) and 31F (term, day 31). In fetal pups left lungs were harvested for histology, and right lungs snap frozen for molecular analysis. The other pups were stimulated to breath by tactile stimulation. Pups surviving the first adaptation hour after birth were hand reared until day 7 (preterm pups) and day 4 (term pups; same corrected age): they constitute the postnatal groups P (preterm) and T (term). A subset of the postnatal pups underwent in vivo microCT and dark field imaging on day 0, day 3, day 5 and day 7 in P and on day 0, day 2 and day 4 in T (same corrected ages). On day 7/day 4, all postnatal pups underwent lung function testing before snap freezing of the right lung, and left lung harvest for histology.

Left lungs were harvested for histology, and therefore pressure fixed on 25cmH₂O of paraformaldehyde for 24 h. Volume of fixed left lungs was measured by water immersion (Scherle’s principle). Lungs were embedded in paraffine, and for every lung a central sagittal 5 μm slide was made for hematoxylin-eosin staining. Stained sections were scanned with a slide scanner (Axio Scan® Slide Scanner, Zen Zeiss, Oberkochen, Germany). Alveolar morphometry measurements are semi-automatically performed with a self-designed Fiji-plugin on 20 randomly selected fields per lung. Based on the counts mean linear intercept (Lm, estimating alveolar size) and mean transsectional wall length (Lmw, estimating septal wall thickness), and total alveolar surface (S, by taking the reference left lung volume in to account) were calculated [15, 16].

A second central 5 μm slide was stained with Miller’s elastic staining. For at least 15 arteries of 30-100 μm in diameter, the internal and external diameter of the medial-layer was determined on the shortest axis. Average relative thickness of the medial layer (MT%) was calculated per animal as described before [16]. On a third slide an immunohistochemical staining of CD31 was performed in order to assess vascular density. Tissue preparation for immunohistochemistry included deparaffinization, heat retrieval, endogenous blockage of peroxidase (0.5% H₂O₂ in methanol) and protein blockage. Endothelial cells were stained with a mouse anti-human CD31 antibody at 10.25 μg/ml (M0823, DakoCytomation). A secondary goat anti-mouse antibody (Z0420, DakoCytomation) was used in combination with an alkaline phosphatase anti-alkaline phosphatase antibody (STAR67, Bio-Rad, Hercules, CA, USA) and nitroblue tetrazolium as a chromogen. We manually counted vascular structures of 20-50 μm on 5 randomly selected high power fields (hpf) of 500x500μm.

Additionally, one slide per lung was stained with a Sirius Red collagen staining. The amount of collagen was estimated by automatic measurement of the stained area on 20 randomly selected lung fields (500μmx500μm) relative to the whole field.

mRNA was extracted from snap frozen right lung tissue using the Tri-Pure® Isolation reagent (Roche Diagnostics, Germany). RNA concentration was measured using the Nanodrop 1000® spectrophotometer (Thermo Scientific) and RNA integrity was assessed by checking the integrity of the 18S and 28S band by agarose gel electrophoresis. cDNA was synthesized with Taq Man® Reverse Transcription Reagents (ThermoFisher Scientific). The Platinum SYBR® Green qPCR Supermix-UDG with ROX (ThermoFisher Scientific) was used to detect expression of surfactant protein B and C, elastin, collagen 1A2 (COL1A2) and vascular endothelial growth factor (VEGFA). YWHAZ, HPRT and actin B were used as stable housekeeping genes to normalize mRNA levels. An overview of used primers, obtained from Integrated DNA Technologies (Haasrode, Belgium), can be found in Additional file 1: Table S1. Samples were run in triplicate on a StepOne Plus® instrument (ThermoFisher Scientific). Relative quantitation was determined using the comparative Ct method. Statistics are calculated in based on ΔΔCt-values, while fold change (FC) is used for visualization.

Snap frozen right lung tissue was homogenized according to the manufacturer’s instructions, in a protein lysis buffer containing Tris-HCL and EDTA. Surfactant protein B was quantified using a commercial ELISA kit (MBS2601137, MyBioSource, San Diego, California, USA).

For a subset of animals, a piece of skin was snap frozen for sex determination through a PCR of the specific region of Y-chromosome (SRY). DNA was isolated using the GenUP gDNA Kit (Biotechrabbit, Henningsdorf, Germany). The SRY fragment was amplified using PCR, and the obtained PCR products were analyzed by gel electrophoresis. In all samples rabbit GAPDH was used as an internal amplification control. Primers are listed in additional file 1: Table S1.

FlexiVent 5.2 (SCIREQ, Montreal, Canada) is used to perform pulmonary function tests on deeply anesthetized postnatal pups. The trachea of the pup is surgically exposed and an 18G metal cannula is inserted in the trachea and connected to the FlexiVent ventilator (120 breaths/min ventilation; tidal volume of 8 ml/kg). A pressure volume perturbation is performed to measure inspiratory capacity and static compliance. Snapshot is a single-wave forced oscillation perturbation used to assess dynamic compliance, elastance and resistance from a single compartment model. Finally, Prime-8 is a broadband forced oscillation technique and is used to assess tissue damping (frequency independent parenchymal energy dissipation) and tissue elastance (energy conservation of lung tissue). Three measures for each perturbation are obtained per pup and the mean value is used in the analysis. Measures for compliance, elastance and inspiratory capacity were corrected for weight. After lung function, deeply anesthetized animals are euthanized by exsanguination.

Rabbit pups were fixated on a Styrofoam bed in the supine position in a low-dose small animal microCT scanner (SkyScan 1278, Bruker micro-CT, Kontich, Belgium). The following acquisition parameters were used: 1 mm Al filter, 50 kVp X-ray source voltage, 918 μA source current, 55 ms exposure time, 9 projection images per 0.9° rotation step over a total angle of 180°. Respiratory gating was performed retrospectively by matching projection images to the signal of a camera registering breathing movements. A 3D datasets with a 50 μm isotropic reconstructed voxel size was created. We used software provided by the manufacturer (TSort, NRecon, DataViewer and CTan) to gate, reconstruct, visualize and analyze microCT data [17]. End-expiratory data is used for analysis.

Lung volume was manually delineated on all slices of the reconstructed microCT dataset, avoiding the inclusion of the heart and mediastinal structures, by a researcher (MA) blinded for group allocation. Total (end-expiratory) lung volume and mean lung density were calculated. Additionally, by applying a fixed density threshold (85 on a grayscale from 0 to 255), aerated lung tissue was segmented from the non-aerated lung tissue. Aerated and non-aerated lung volume was calculated as a percentage of the total lung volume. To calibrate mean lung density for Hounsfield units (HU), a phantom (air-filled tube of 1.5 mL inside a water-filled tube of 50 mL) was scanned [18]. Only the outcomes of the pups that survived until day 7 were used for the analysis (n = 7–9).

Dark field imaging is a form of phase-contrast x-ray imaging, and was performed on a Talbot-Lau interferometer with the pup held in a straight position on a Plexiglas mold. The interferometer operated at 40 kVp with a 2 μm G2 grating period and a system visibility of 22%. The radiation dose per round was 0.51 mGy [19, 20]. From the acquisitions, dark field images were reconstructed through a MATLAB-algorithm. The intensity of the dark field signal in 4 quadrants of every lung image was scored on a scale from 0 (no signal indicating little aeration) to 8 (strongest signal indicating complete aeration) by 3 independent observers who were blinded for the group allocation. The average score of the 4 quadrants was considered representative for the animal, and the average of the 3 observers is reported.

GraphPad Prism® 7.0 software (GraphPad, La Jolla, California, USA) was used for statistical analysis. Normality was assumed after visual inspection of the data. Statistical outliers were removed from the analysis if a Grubbs test was positive. All values are expressed as mean ± standard deviation. Unpaired t-testing with Welch correction for unequal variance was used to compare outcome measures in 28F to 31F and P to T. A mixed-effects model was used to compare the longitudinal outcomes between the different time points. Bonferroni-Sidak correction for multiple comparisons was used to compare groups on individual time points. Kaplan–Meier curves with post hoc log-rank testing were used to quantify survival of rabbit pups from fetal day 28 until postnatal day 7 (preterm) or 4 (term), assuming that intrauterine deaths in the term mothers occurred before day 28 (n = 3). A p-value of < 0.05 was considered statistically significant.

Fifteen preterm and 10 term fetal animals were harvested before the first breath, from respectively 5 and 3 mothers. Birth weight was significantly lower in 28F than in 31F (p = 0.0109; Additional file 2: Table S2). Lungs of preterm rabbits had a significantly higher Lm (p = 0.0012) and an increased Lmw (p < 0.0001) compared with term fetuses (Fig. 2a, b). The representative figures depict the more rounded and less complex airspaces, with a higher cellularity in 28F (Fig. 2c). Left lung volume was not significantly lower in 28F (p = 0.2927), but due to the simplified lung structure total alveolar surface (S) is significantly decreased (p = 0.0484; Fig. 2d, e). However, the difference in alveolar surface is proportional to the difference in birth weight (p = 0.5162 if alveolar surface is normalized to weight). Additionally, arterial medial thickness was higher in 28F compared to 31F (p = 0.0004; Fig. 2f); representative pictures in Fig. 2c. Vascular density however was not significantly different between 28F and 31F (p = 0.1500; Fig. 2g); respresentative pictures in Fig. 2c. Finally, preterm fetuses also demonstrate lower mRNA-expression of surfactant protein B and C than term fetuses (p = 0.0055 and p = 0.0282 respectively; Fig. 3a, b). Also at a protein level, there was a decreased amount of surfactant protein B present in the fetal lung tissue of 28F-pups, in comparison to 31F-pups (p = 0.0065; Fig. 3c).

Four mothers were randomized to term and 8 to preterm delivery. In the first hour after birth, 100% (47/47) of term pups and 55.8% (43/77) of preterm pups survived the adaptation to postnatal life. Fifteen term and 25 preterm randomly selected pups were included in the experiment. Survival was significantly lower in pups born preterm (60%; 15/25) then in animals born at term (100%; 15/15; p = 0.0105; Fig. 4a). Over time, weight evolved significantly different between preterm and term animals (p = 0.0005). On day 7/day 4 (day of harvest), preterm pups were significantly smaller than term counterparts (p = 0.0190; Fig. 4b; Additional file 2: Table S2).

In vivo microCT-derived end-expiratory lung volume grew significantly over time in both the preterm and term animals (p < 0.0001). Lung volume was not significantly different in both groups (p = 0.1297; Fig. 5a). Additionally, the proportion of aerated lung volume changed over time in both P and T (p < 0.0001). Lungs of pups in P had a significantly lower proportion of aerated lung volume (p = 0.0001), with significant posthoc differences on day 3/day 0 and day 5/day 2, and a strong trend towards less aerated volume on 7/day 4 (p = 0.0022, p = 0.0282 and p = 0.0504 respectively). Conversely, non-aerated lung volume decreased over time and was significantly larger in P (Fig. 5b). Consequently, also the mean lung voxel density decreased with time (p < 0.0001), with a generally significantly higher density in P compared to T (p = 0.0008). Posthoc testing only revealed a significant difference on day 3/day 0 (p = 0.0085; Fig. 5c). At visual analysis, lungs of preterm pups, especially in the early time points but also later, were characterized by atelectatic lung regions, diffuse ground glass areas and sometimes air bronchograms. Representative images of the transversal microCT-sections can be found in Fig. 5e.

In line with the microCT-results also the dark field intensity scores, reflecting lung aeration, increased over time in the preterm group (p = 0.0016), but not in the term animals (p = 0.6723) that obtained high scores already immediately after birth. Dark field intensity was overall higher in T in comparison to P (p = 0.013), however at posthoc comparisons the difference was only significant on day 3/day 0 (p < 0.0001; Fig. 5d). Representative dark field images can be found in Fig. 5e.

Seven days after birth, Lm and Lmw were not significantly different in preterm pups (P), versus term pups at the same corrected age (T; Fig. 6a-c). Because ex vivo left lung volumes were significantly smaller in P than T (p = 0.0042; Fig. 6d), total alveolar surface area (S) was decreased in P versus T (p = 0.0210; Fig. 6e). The difference in alveolar surface area was proportional to the difference in body weights of the groups (p = 0.3231 if alveolar surface is normalized to weight). Arterial medial wall thickness was not significantly different between P and T (p = 0.5386; Fig. 6c and f). Additionally; vascular density was comparable in P and T (p = 0.8740; Fig. 6c and g). Tissue mechanics were altered in preterm animals, with increased tissue damping and tissue elastance in P (p = 0.0104 and p = 0.0240 respectively; Fig. 7a, b). Airway resistance tended to be higher in P (p = 0.1008; Additional file 3: Table S3). At single wave oscillation, dynamic compliance corrected for weight was increased in P (p = 0.0079; Fig. 7c). Consequently, dynamic elastance corrected for weight was decreased in P (p = 0.0159; Additional file 3: Table S3). Resistance was also significantly higher in P (p = 0.0140; Fig. 7d). In pressure-volume perturbations, static compliance and elastance, and inspiratory capacity (all corrected for weight), were not significantly different (Additional file 3: Table S3).

We observed no significant changes in mRNA-expression of extracellular matrix components elastin and collagen 1A2 in P versus T (p = 0.1382 and p = 0.1593; Fig. 8a, b). However there was a marked variation in elastin mRNA expression in P. We also did not observe any significant change in collagen positive area on lung tissue slides (p = 0.7574; Fig. 8c). VEGFA mRNA-expression was significantly decreased in P in comparison to T (p = 0.0173; Fig. 8d). Finally, surfactant protein mRNA-expression was also increased in P versus T: significantly for surfactant protein B (p = 0.0089; Fig. 8e), but only a trend for surfactant protein C (p = 0.0733; Fig. 8f).

There were no significant differences between male and female preterm pups for the main outcomes of this experiment, however it was not powered for this comparison (Additional file 3: Figure S1).