The experiments were performed in compliance with the EU guidelines and the French regulations (DIRECTIVE 2010/63/EU, 2010; Code rural, 2018; Décret n°2013-118, 2013). The procedures received authorization of the Ministry of Higher Education and Research and the experimental protocols were approved by the COMETHEA ethic committee under the number APAFIS# DAP2022120510144946 v3. The surgery was carried out at the Medical Imaging in Animal platform (accreditation B78-322-2, doi.org/10.15454/1.5572348210007727E12).

Lungs were harvested from five Large-White x Landrace x Pietrain pigs (48–55 kg). Animals were anesthetized with intramuscular ketamine (20 mg·kg^-1, Imalgene®) and xylazine (2 mg·kg^-1, Rompun®). Orotracheal intubation was performed after deepening anesthesia with 3%–5% isoflurane using mask, and intravenous injection of fentanyl (20 μg·kg^-1) and diazepam (0.1 mg·kg^-1). Anesthesia was maintained with 2%–3% inhaled isoflurane and continuous propofol infusion at 5 mg·kg^-1·h^-1 (Proposure®). Mechanical ventilation was initiated with a tidal volume (Vt) of 7 mL·kg^-1, a positive end-expiratory pressure (PEEP) of 5 cmH₂O, and a respiratory rate of 15 cycles·min^-1. Upon chest opening, a biopsy from the azygos lobe was harvested (designated as the “Time of death” time point). The animal was then euthanized by aortic clamping and exsanguination following intravenous administration of 25,000 IU of heparin. Euthanasia marked the onset of WI. After inflation, each lung was isolated. The right lung (control group) was immediately flushed with 2 L Perfadex® at 4 °C (XVIVO Perfusion, Gothenburg, Sweden) and stored in the refrigerator at 6 °C for 1 h. The left lung (WI group) remained inflated in situ for 2 h, before being similarly flushed and stored at 6 °C for 1 h. Each lung was then connected to an EVLP circuit following the Toronto protocol [10]. The circuit was primed with 1 L of Steen® solution supplemented with 0.5 g methylprednisolone, 0.75 g cefuroxime, and 3750 ui heparin. The lung size differences were taken into account to determine the flow of 16% (left) and 24% (right) of the theoretical cardiac output. The system reached 37 °C after approximately 30 min of perfusion. Ventilation settings were adjusted to a Vt of 3.5 mL/kg for the right lung and 2.5 mL/kg for the left lung, with a fraction of inspired oxygen FiO2 of 0.21, positive and expiratory pressure (PEEP) of 5 cmH₂O and a respiratory rate of 15 breaths per minute. The perfusate was deoxygenated using a gas mixture containing 8% CO₂, 86% N₂O, and 6% O₂. EVLP was maintained for a total duration of 6 h.

Pulmonary gas exchange capacity was evaluated using PaO₂ and PaCO₂ measured in the perfusate at an FiO₂ of 0.21, using i-STAT cartridges (Abbott, Chicago, USA). Respiratory mechanics and mechanical ventilation were assessed by measuring plateau pressure (Pplat), driving pressure (ΔP = Pplat–PEEP), pulmonary compliance (Vt/ΔP) and recruitment (compliance change during a PEEP test performed at 2 cmH₂O and 8 cmH₂O). The pulmonary strain was estimated as stress divided by specific elastance, with stress approximated by Pplat for isolated lungs, and specific elastance estimated in pigs at 6 cmH₂O [11]. Hemodynamics were evaluated by calculating pulmonary vascular resistance (PVR = [(mPAP – 0)/ECMO flow] × 80). Perfusates were collected at hourly intervals and frozen at −80 °C for lactate deshydrogenase (LDH), mitochondrial DNA and cytokine detection. Lung tissue samples (about 100 mg) were collected using a lung stapler from the azygos lobe just before exsanguination, from the superior lobe at the end of the cold ischemic phase and after 2-h perfusion, and from the inferior lobe after 4- and 6-h perfusion. These samples preserved in RNAlater (Merck, Darmstadt, Germany) or directly frozen in liquid nitrogen and kept at −70 °C. Additionally, a 100 mg lung tissue sample collected at 6 h was fixed in 4% buffered paraformaldehyde for histological analysis, then rinsed after 24 h and stored in 70% ethanol. Bronchoalveolar lavages were performed with 2 × 20 mL cold PBS, and samples were kept frozen at −70 °C.

Detailed histology protocols, the commercial kits employed (LDH, cytokine assays, carbonylated proteins, oxidized DNA), and mitochondrial DNA quantification by PCR (ND1 and ND4 genes) are described in Supplementary Material 1.

Total RNA was extracted from lung tissue using TRIzol reagent with bead homogenization (Precellys 24 bead grinder) and purified with the NucleoSpin RNA kit including DNase treatment. RNA quantity and integrity were assessed using the Qubit RNA assay kit and Agilent 2100 Bioanalyzer. Reverse transcription (1.2 µg RNA) was performed using the PrimeScript RT reagent kit. Gene expression was measured using custom TaqMan Array Cards (96 genes, 4 housekeeping genes, Supplementary Material 2) on a QuantStudio 7 Flex Real-Time PCR System. Data were analyzed using the 2^−ΔΔCT method normalized to RPS24, RPLP2, and GAPDH; genes with Ct > 32 were excluded. Principal component analysis was performed (mixOmics). Differentially expressed genes (DEGs) versus “Time of death” were identified using paired one-way tests with Benjamini–Hochberg correction (adjusted p < 0.1). Fold changes were calculated relative to “Time of death”. Comparisons between all conditions and the 2-h WI at time zero used paired non-parametric tests with multiple testing correction (adjusted p < 0.2). More details are provided in Supplementary Material 1.

The statistics of gene expression is described in the TAC array design and analysis. For the physiological and biological parameters, a Linear Mixed Model with fixed effects for group, time, and their interaction, and a random intercept for pig subject, was used to analyze repeated measures using R Studio2 v2024.09.0 + 375; the fixed effects were tested with a type III ANOVA. For the post hoc comparisons between lung groups, a Tukey test was performed (paired data), using estimated marginal means, and provided adjusted p-values. For histology and percent weight gain, the data were statistically compared using a paired t-test, after checking normal distribution (Shapiro-Wilk). Supplementary Material 3 reports the raw data, means, standard deviation as well as the statistical results (F-val, p-val).

Lungs were explanted from five pig donors (see Supplementary Material 1). The azygos lobe was sampled immediately post-mortem (“Time of death”). Each lung block was divided into right and left lungs to allow paired analysis. The right lung (control group) was kept at 6 °C for 1 hour and immediately subjected to 6-h EVLP. The left lung (WI group) remained for 2 h in the thoracic cavity, before undergoing the same cold preservation and EVLP protocol (Figure 1). At the time of surgery, the animals’ body temperatures were within the physiological range (38.5 °C–39 °C) and decreased to 35 °C–37 °C 2 hours after death. During EVLP, oxygenation was maintained in both groups, with PaO₂ values consistently exceeding 150 mmHg (Figure 2A, see Supplementary Material 3 for the raw data of Figures 2–5). Initial lung compliance was lower in the WI group than in the control group (32% lower, p-val = 0.02), and improved over EVLP time (time effect p-val = 0.008 for both groups, Figure 2B). The strain ratios were below 2.5 and not different between groups (Figure 2C). The initial vascular resistance was higher in the WI than in the control group (25.6% higher, p-val = 0.007), and then the difference became non-significant during EVLP (Figure 2D). At 6 h, weight gain was limited and comparable between groups (percent weight gain: 7.89 ± 11.76 in WI and 14.61 ± 12.23 in control group, Figure 2E). So overall, although 2-h WI initially impaired lung functions (reducing compliance and increasing vascular resistance), these effects were corrected by a 6-h EVLP.

Biological markers associated with ischemia–reperfusion injury were measured during EVLP (Figures 3–5). At baseline (“Time of death”), carbonylated protein levels in lung tissue reached mean values of 2.5 mmol/mg protein (Figure 3A). These levels decreased over time during EVLP (time effect p = 0.01), with no difference between groups at any time point. Similarly to carbonylated protein levels, oxidative DNA levels showed a temporal decrease in both groups (Figure 3B). Mitochondrial DNA concentrations in the perfusate increased over time (time effect p = 0.031), without intergroup differences (Figure 3C). LDH was progressively liberated in the perfusion liquid (time effect, p-val <0.001), without difference between groups (Figure 3D). The metabolic response reflected by lactate production increased over time (p-val in both groups <0.001) and showed higher values in the control group as compared to in the WI group particularly at 6 h (p-val = 0.002, Figure 3E). Finally, the levels of PECAM release also increased over time with higher values in the control group at 6 h (p-val in both groups = 0.0003, Figure 3F). Cytokine concentrations in the perfusate increased over time for TNFα, IL-6, and IL-10, with no differences between groups (Figure 4A). Similar findings were observed in bronchoalveolar lavage samples (Figure 4B). Histological assessment at 6 h showed low injury scores in both groups, with no significant difference (WI: 0.59 ± 0.19 vs. control: 0.42 ± 0.12; p = 0.145; Figures 5A–C).

We designed a 96 PCR-array using genes previously shown by us and others to be modulated by EVLP ([12–18], Supplementary Material 2). We also incorporated the PDK4 and FKBP5 genes that are upregulated by corticosteroids. Among the 96 genes, 84 genes were retained for analysis based on expression thresholds (Cq < 32; see Methods). Principal component analysis of 2^−ΔΔCT values showed overlap between samples obtained at “Time of death” and those from the control group at EVLP initiation (0 hour-EVLP) (Figures 6A,B). During EVLP, transcriptomic profiles from the control and WI groups followed similar trajectories and converged at 6 h. Interestingly, at 0-h EVLP, the WI group samples were located in the EVLP trajectory.

For each condition (group and time), a list of differentially expressed genes relative to the “Time of death” (DEGs, p-val <0.1) was established, revealing 69 genes that were differentially expressed at least once (Figure 7, Supplementary Material 4). Among these DEGs, genes related to inflammation (IL6, CXCL2, CXCL8, CCL2, IL17F, TGFA), IRI (CD14, TLR4, MyD88, STAT3) and oxidative stress (GADD45, MT1A, SOD2) were upregulated, whereas genes related to immune activation (such as CD1E, TCRD, XCL1, IFNG) and cytoskeleton signaling and cell adhesion/migration) (ACTG1, ACTB, Myo5C, ITGA4, ITGA6, CX3CR1) were downmodulated. PDK4 and FKBP5 were upregulated only during EVLP, reflecting effective response to corticosteroids. Figure 7 shows that among these 69 DEGS, 29 were significantly modulated by WI per se at 0-h EVLP whereas none were found in the control condition at 0-h EVLP. Thereafter during EVLP, there was no significant difference in the gene expression fold changes between the WI and control group at any time point. Furthermore at 2- and 4-h EVLP, the WI and control groups presented a majority of DEGs in common (40 out of 52 DEGs at 2 h, and 42 out of 58 DEGs at 4 h). At 6 h, the WI group showed a high dispersion of the FC values between pig lungs, leading to less DEGs in the WI group than in the control group (51 in the control group, 27 in the WI group, and 23 in common). Finally, when comparing gene expression during EVLP conditions to that of WI lungs at 0-h EVLP, we found that the absolute mean FC of most DEGS (42 out of 69) tended to be higher (p-val <0.2, Supplementary Material 4).