We used the experimental model of BPD based on preterm rabbits delivered at 28 days of gestation (i.e. equivalent to the saccular stage of human lung development [2]) and exposed to hyperoxia shortly after birth [5]. In details, rabbits were exposed to two degrees of hyperoxia with different inspired oxygen fraction (FiO2), that is: moderate (FiO2 = 0.7, n = 6) or severe hyperoxia (FiO2 = 0.95, n = 9). Hyperoxia was provided right after birth for 1 h at 60% relative humidity. Two additional groups of animals treated in normoxia (FiO2 = 0.21) were also studied: one consisted of preterm rabbits (n = 8) and another was a control group of term (42 days of gestation) healthy rabbits (n = 5). The rest of animal care and treatment was provided as previously described [5].

Lung ultrasound was performed in newborn rabbits at on the 7th and 14th day of postnatal life, for severe and moderate hyperoxia groups, respectively; pups exposed to severe hyperoxia always die after the first week of life, thus they could not be scanned at the 14th days. Preterm rabbits exposed to normoxia and control healthy rabbits also underwent lung ultrasound on the 14th day of life. Lung ultrasound was performed as previously described [5]: in detail, each hemithorax was scanned as a whole, using a micro-linear high-frequency (20 MHz) probe (Fujifilm Visualsonics inc, Toronto-ON, Canada), and rabbit lung ultrasound score (rLUS) was calculated, by averaging results from the two hemithoraces. rLUS was based on the classical lung ultrasound semiology assigning a 0-to-3 score to the observed ultrasound patterns (Fig. 1) [6]. The evaluation of rLUS is known to have a high inter-rater agreement [5].

Relationship between radial alveolar count (RAC) and quantitative lung ultrasound findings. Green, orange, and violet dots represent individual RAC data for each rabbit with interstitial (rLUS = 1), interstitial-alveolar (rLUS = 2) or consolidated (rLUS = 3) ultrasound pattern, respectively. Illustrative pictures of lung ultrasound pattern are shown above the dot graph. Box plots depict (from top to bottom) maximum, 75th, 50th, 25th, and minimum RAC values. The curves on the right-side represent the density (distributions) of data points. No rabbit had rLUS = 0. Data were analyzed with Kruskal–Wallis test (overall p = 0.036, n = 28). The horizontal line represents the significant post hoc comparisons given by Dunnett test. RAC and rLUS are dimensionless variables. RAC radial alveolar count, rLUS rabbit lung ultrasound score

After the ultrasound scan, rabbits were euthanatized as previously described [5]. In detail, lungs were removed and fixed with 10% buffered formalin under constant pressure for 24 h, then transferred to 70% ethanol, embedded in paraffin, and stained with eosin and hematoxylin following standard histology procedures. A section for each lung was used to calculate the radial alveolar count (RAC) in 40 fields/section [7]. Similarly to the rLUS calculation, the values of the right and left lung were averaged.

Research protocol was approved by the Italian Ministry of Health review board (n. 875/2021-PR) and met all relevant regulations about animal research. ARRIVE guidelines were followed [8]. A convenience sample size was decided based on previous studies about lung mechanics and pathology on the same experimental model [9]. RAC and rLUS were analyzed with Spearman (ρ) and Kendall (τ-b) correlation coefficients. RAC values were expressed as median [25th–75th percentile] and compared between rabbits with different ultrasound pattern using Mann-Withney test followed by Dunnett post hoc test. JASP 0.17.1 (JASP Team 2023) was used and p-values <0.05 were considered significant.