Large White pigs (3–4 months-old, 24 in total) included matched pairs of male donors and female recipients originating from unrelated parental origins. They were split into three experimental groups, four pig couples per group ( Figure 1 ): untreated donors and recipients (UT), CS-treated recipients only (CR, recipients that received 20 mg/kg methylprednisolone IV (Mylan, Canonsburg, USA) 60 min before cross-circulation) and CS-treated donors and recipients (CDR, donors that received 20 mg/kg methylprednisolone IV 12 h before lung procurement and recipients that received 20 mg/kg methylprednisolone 60 min before cross-circulation).

The overall cross-circulation procedure has been detailed in (10) based on a model previously established by the Columbia University team (12). The lung harvest from donor pigs (n = 12) was performed as described (10) in the Animal Surgery and Medical Imaging Platform (CIMA-MIMA2-BREED-INRAE, Jouy en Josas, France). Recipient pigs (n = 12) were anesthetized as described (10) and Septotryl^® (0.08 ml/kg) (Vetoquinol) was injected IM prior to catheterization. Heparin was administered as an initial 25,000 U bolus, and maintained by continuous infusion (100 U/Kg/h). The superior vena cava was cannulated with a 20 F double lumen cannula as described (10). Carboxyfluorescein succinimidyl ester (CFSE, 25 mg, Sigma-Aldrich, Saint-Louis, Mississippi, USA) was administered IV in the perfusing pig, diluted in 4 ml DMSO + 40 µl heparin, 30 minutes before the initiation of cross-circulation. Donor lungs were placed in dorsal position on an XVIVO^® chamber (XVIVO Perfusion) and the trachea was cannulated with a 7.5 mm diameter cuffed endotracheal tube (Mallinckrodt, Staines-upon-Thames, UK) and connected to a respirator. The vascular tubing was spliced to connect the recipient pig to the dedicated circuit, marking the start of cross-circulation. Sedation was kept for 10 hours by permanent administration of 2 mg/kg/h propofol (Proposure^®, Axience, Pantin, France) with 0.6% isofluorane and analgesia was achieved by IV administration of 0.2 mg/kg nalbuphine every 3 hours.

We collected blood samples from the pulmonary artery and vein (PA and PV) cannula hourly to perform the hemo-gas analyses using an Istat^® kit. The static compliance (Tidal Volume/(plate pressure – positive end expiratory pressure)), ΔPCO2 (arterial PCO2 – venous PCO2) and ΔPO2/FIO2 (venous PO2 – arterial PO2/FIO2) were measured once per hour. The transpulmonary pressure (PA - PV) and the pulmonary vascular resistance were calculated (PA pressure – left atrial pressure) x 80/flow rate.

We collected blood samples directly from the extracorporeal circuit. Blood counts and biochemical profiling from plasma were performed on a MS4.5 analyzer and a M-Scan II analyzer (Melet Schloesing Laboratoires, Cergy-Pontoise, France).

Lung biopsies for cell dissociation and RNA extraction were sampled at T0H, T6H and T10H of cross-circulation using a surgical stapler (Endo GIA™ universal stapling system, Medtronic, Minneapolis, USA). For cell dissociation, 2 g of tissue were immediately immerged in cold hypothermic preservation media (HypoThermosol^® FRS, Stemcell Technologies Inc, Vancouver, Canada) and for RNA, about 50 mg of tissue were stored in RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) at -20 °C. For histology, biopsies were fixed in cold phosphate-buffered 4% paraformaldehyde for 24 hours and were subsequently paraffin-embedded.

Lung tissue samples were obtained from eight patients undergoing surgical resection for lung carcinoma at the Foch Hospital in Suresnes (France) ( Additional File 1 ). The lung tissue parts located away from the tumor were dissected and used for cell isolation.

Lung tissue was finely chopped and incubated at 37 °C for 45 minutes under rotation in RPMI 1640 supplemented with 100 IU/ml penicillin,100 mg/ml streptomycin, and 10% inactivated fetal calf serum (FCS) (Invitrogen, Paisley, UK), containing 3 mg/ml collagenase D, 0.25 mg/ml Dnase I (Sigma-Aldrich) and 0.7 mg/ml dispase II (Gibco^®, ThermoFisher Scientific, St Aubin, France). The cell preparation was subsequently filtered on a nylon mesh (1 mm diameter) followed by successive passages through cell strainers (500 µm, 100 µm, 40 µm). Erythrocytes were eliminated with an erythrocyte lysis buffer (10 mM NaHCO3, 155 mM NH4Cl, and 10 mM EDTA). The cells were washed in PBS, counted and 10⁸ cells were processed to cell surface staining. For subsequent pig cell sorting, the remaining cells were frozen in 10% DMSO + 90% FCS using a Mister Frosty freezing container (Nalgene, Rochester, NY, USA) and finally kept in liquid N2.

The isolated human lung cells were suspended at a concentration of 2 x 10⁶ cells/ml in RPMI containing 10% fetal calf serum, 100 U/ml penicillin and 100 µg/ml streptomycin, with or without 20 µg/ml methylprednisolone, for 12 or 42 hours (Sigma-Aldrich Merck Group, MA, USA). After culture, the lung cells were harvested using accutase treatment (Sigma-Aldrich Merck Group) and spun at 600 g. The cells (30 to 40x10⁶) were then incubated with an anti-CD172A mAb (DH59B clone, 5 µg/ml) for 30 minutes at 4 °C and the positively labeled cells were magnetically sorted using anti-mouse IgG coated beads and LS columns (Myltenyi, Bergisch-Gladbach, Germany) following the manufacturer’s recommendations. The purity of the CD172A^pos cells was > 92% estimated by flow cytometry control.

We performed the cell surface staining in RPMI supplemented with 10 mM Hepes, 5% horse and swine serum respectively (Gibco, Life Technologies Europe, Bleiswijk, Netherlands). We used swine-specific primary antibodies (Abs) and conjugated secondary Abs that are listed in Additional File 2 . We used isotype controls (mouse IgG1, IgG2b, and IgG2a) at the same concentration as the tested mAbs, based on the fluorescence minus one method (13). In some cases, a directly PE-conjugated IgG1 mAb (anti-CD163-PE and anti-CD3-PE) was used as a third step, and an excess of mouse IgG1 (50 μg/ml) was added to saturate the first IgG1 step. A DAPI staining (Sigma-Aldrich) was used to exclude dead cells or to assess cell death among CD3^pos T cells. Results were acquired on BD LSR Fortessa™ Cell Analyzer (BD-Biosciences). The FACS data were analyzed with the FlowJo software (version 10.7.1; Tree Star, Ashland, OR, USA). The gating strategy is shown in Additional File 3 that mainly originates from our recent paper (10).

Isolated pig lung cells were thawed from frozen stocks and reacted with anti-CD14, CD16 and CD172A mAbs followed by conjugated goat anti-mouse isotype-specific Abs ( Additional File 2 ). The anti-IgG1 conjugate was saturated by an excess of mouse IgG1 (50 μg/ml) and cells were labeled with anti-CD163-PE (IgG1). The cell subsets were sorted (“purity” mode) by flow cytometry on the Imagerie-Gif Cytometry facility (I2BC, Gif sur Yvette, France) using the MoFlo ASTRIOS sorter (Beckman-Coulter, Paris, France) and the Summit 6.3 software. Alveolar macrophages (AMs) were sorted as CFSE^negSSC^hiCD163^hi/CD172A^hi from total live lung cells. The MoCs were sorted as CFSE^posSSC^loCD172A^hiCD14^posCD16^lo (named CD14^pos) and CFSE^posSSC^loCD172A^hiCD14^loCD16^hi cells (named CD16^pos).

RNA from flow cytometry-sorted pig CD14^pos MoCs, CD16^pos MoCs and AMs as well as from immunomagnetically-sorted human CD172A^pos lung macrophages were extracted using the Arcturus PicoPure™ RNA kit (ThermoFisher Scientific). Calibrator RNA samples from pig and human lung cells were prepared in parallel. In addition, whole lung tissue biopsies collected in RNAlater were placed in Trizol, homogenized with 1.4 mm ceramic beads in a Precellys 24 bead grinder homogenizer (Bertin Technologies, St Quentin en Yvelines, France), and purified using the NucleoSpin RNA kit that includes a DNAse digestion step (Macherey-Nagel, Düren, Germany). RNA was quantified by Qubit™ RNA high sensitivity kit (Invitrogen™, Fisher Scientific SAS, Illkirch, France). In the case of sorted cells, RNA (8–100 ng) was reverse transcribed using random primers (5 µM) and oligo dT primers (2.5 µM) and the Multiscribe reverse transcriptase (Applied Biosystem, ThermoFisher Scientific), using equal starting quantities of RNA from test and calibrator RNA. Quantitative real-time PCR was carried out with 300 nM primers in a final reaction volume of 25 µl of 1 X SYBR Green PCR Master Mix (Applied Biosystem, ThermoFisher Scientific). In the case of lung tissue, RNA (400 ng) was reverse-transcribed using random primers (5 µM) and oligo dT (2.5 µM) primers and the Takara Prime Script RT reagent kit (TaKaRa, Kyoto, Japan). Quantitative real-time PCR was carried out with 300 nM primers in a final reaction volume of 25 µl of Q-PCR TB Green Premix ExTaq (TaKaRa). PCR cycling conditions were 95 °C for 30 seconds, linked to 40 cycles of 95 °C for 5 seconds and 60 °C for 30 seconds. The primers were designed using the primer express software (v2.0) or were bought from Sigma-Aldrich and are reported in Additional File 4 . Real-time qPCR data were collected by the Bio-Rad CFX Maestro system (Bio-Rad Laboratories Inc, Marne-la-Coquette, France) and expression of the different genes was calculated relatively to RPS24 expression in pig and to RPS18 in human. All PCR data were normalized to the internal calibrator (arbitrary units) and the values were calculated by the 2^−ΔΔCT method.

Formalin-fixed paraffin-embedded lung tissues at T0H from the UT and CDR groups were cut every 50 µm to produce three 5 µm tissue slices per sample. After antigen unmasking (100 °C, 20 minutes), the tissue sections were incubated with 30% H2O2, followed by Avidin/biotine (OriGene, Rockville, USA), saturated with 10% goat serum (Vector Laboratories, Newark, USA), reacted with 2.5 µg/ml anti-pig CD3 (clone PPT3, Clinisciences) followed by 1:400 dilution of biotinylated Goat anti-mouse IgG (Vector Laboratories) in Bond™ Primary Antibody Diluent (Leica, Wetzlar, Germany), revealed by the VECTASTAIN ELITE ABC HRPO and DAB substrate kits (Vector Laboratories) and counterstained with Hematoxylin (Diapath, Martinengo, Italy). The slides were imaged with a slide scanner (Pannoramic SCAN II, v3.0.2, 3DHistech, Medipixel Ltd, Budapest, Hungary). Ten randomly selected high-power fields per slide (0.1 mm² area) were observed from three slides per sample and the mean number of CD3^pos T cells per pig was calculated.

The human lung cell supernatants were thawed, spun at 10000 g and assayed using the Human CXCL10/IP-10 DuoSet ELISA (R&D Systems, MI, USA).

Freshly isolated pig lung cells from an untreated pig (2 x 10⁷ cells) were incubated with anti-MHC class II (MSA clone, 2 µg/ml) for 30 minutes, washed, and reacted for 15 minutes with 20 µl goat anti-mouse IgG-conjugated immunobeads (Myltenyi, Bergisch-Gladbach, Germany). After filtration on a 40 µm cell strainer, MHC class II^pos cells were enriched with a MS Myltenyi separation column (ref 130-042-201), following the manufacturer’s recommendations. Cells were counted using a counting chamber and checked for viability using trypan blue and showed over 90% viability. A cell sample (2 x 10⁴) was loaded onto the 10x Chromium to produce a sequencing library, which was processed according to methods provided by 10x Genomics 3’ end (v3 Chemistry, 3’ end). Cell cDNA was sequenced using the Truseq Illumina Stranded protocol and the Illumina NextSeq 500 sequencing machine (> 3x10⁸ reads/sample). The reads were aligned with Cell Ranger v3.0.1 on the pig genome assembly version 11.1 using the Ensembl release 99 of gene annotations. The sequencing results were pre-processed and normalized using Seurat v4.3.0. Cells expressing less than 500 genes were removed. A total of 11029 cells were used to generate an UMAP with clustering (dimensions of reduction used = 11, k.param nearest neighbors = 20, resolution = 0.4). The top differentially expressed genes in each cluster were extracted (log 2-fold change > 0.6) using the FindAllMarker function of Seurat ( Additional File 5 ), after removal of the ribosomal genes. The most highly expressed genes were then used to determine the cell type of each cluster, based on the expression of recognized hallmark genes for human lung (14–16).

The surgery was conducted at the Medical Imaging in Animal platform (accreditation B78-322-2) and the animals were hosted at the Animal Genetics and Integrative Biology unit at INRAE-Jouy (accreditation C78-719). The animal experiments were conducted in accordance with the EU guidelines and the French regulations (DIRECTIVE 2010/63/EU, 2010; Code rural, 2018; Décret n°2013-118, 2013). The experiments were approved by the COMETHEA ethic committee under the APAFIS number authorization 25174–2020011414322379 and were authorized by the French “ministère de l’enseignement supérieur et de la recherche”. The human part of the study was declared as “dossier de Conservation et de préparation à des fins scientifiques D’Eléments du Corps Humain”: (CODECOH) DC N° DC-2020-3981 (1). Experiments on human tissues were approved by the regional investigational review board (Comité de Protection des Personnes Île de France VIII, Boulogne-Billancourt, France). In line with the French legislation on clinical research and as approved by the investigational review board, all the patients gave their informed consent for the use of resected lung tissue for research.

For the physiological and biological parameters, data were Log10-transformed and analyzed with R. Statistical comparisons were performed across time points and between control and treated groups using repeated-measures ANOVA, when the assumptions for parametric testing were met. A Shapiro test was used to evaluate the normality of the data distribution in each group and timing. When normality was confirmed, a t-test was used following equal variance evaluation. If normality was not met, the non-parametric Wilcoxon signed-rank test was applied. Paired tests were performed for comparison between timing and unpaired tests were performed for comparison between groups. The p-values were systematically adjusted for multiple comparisons using the Bonferroni correction. The statistics of the genomic data are reported in the dedicated paragraph. Mean ± standard deviations were calculated.

Using a cross-circulation pig model, we compared the early cellular responses in lung grafts between three groups ( Figure 1 ): untreated donors and recipients (UT), CS-treated recipients only (CR) and CS-treated donors and recipients (CDR). The CR group represents the widely adopted standard of care. The respiratory functions and the blood biological parameters were similar and globally stable in the three groups during the 10-hour experiments, indicating that the CS treatments did not impact the lung physiology ( Additional Files 6 , 7 ). The representation of immune cell subsets was analyzed in the three groups, considering cells from the recipient (CFSE^pos) and from the donor (CFSE^neg) ( Additional File 3 , Figure 1 ). The pig lung immune subsets encompass CD3^pos, CD4^pos and CD8^pos T cells, NK cells, B cells, CD172A^hi monocytic cells (MoCs) including the CD14^pos and CD16^pos monocyte subsets (classical and non-classical/intermediate, respectively) (11), and CFSE^neg CD163^hi lung macrophages that mainly correspond to alveolar macrophages (AMs) (11, 17).

Notably, the global recruitment of the recipient’s cells (CFSE^pos) did not differ between the three groups (21.6 to 30.7% CFSE^pos cells among live cells at T10H, Additional Files 8 , 9 ). As reported in our previous study, the administration of CS to the recipient only had a minor influence on the recruited cell subsets’ composition (11). Interestingly, preconditioning the donor strongly reduced the amount of CD3^pos T cells among the recipient cells recruited in the graft (3.9 ± 1.4 in CDR vs 16.3 ± 2.7 in UT at T10H, adj-p < 0.01 with no significant difference between CR and UT), affecting both the CD4^pos and CD8^pos T cell compartment ( Figure 2 , Additional File 9 ). The B cell and CD16^pos MoC representation was not significantly changed by the treatments. Similarly in the CR and CDR compared to in the UT group, the NK cell representation decreased at T10H and the CD14^pos MoC representation increased at T6H. The PMN representation appeared higher in the CDR than in the CR group ( Figure 2 , Additional File 9 ).

Regarding the donor cells (CFSE^neg), we observed that preconditioning the donor modified the donor cell composition in the graft at T0H, resulting in a decreased representation of CD3^pos T cells and NK subsets and a higher representation of CD14^pos MoC and PMN cells ( Figure 3 , Additional File 9 ). The decrease in CD3^pos T cells by CS in the CDR group was confirmed by immunohistochemistry (p < 0.05, Figure 4 ). The cross-circulation alone modified the donor cell subsets’ representation over time in the three groups, with a tendency toward reduction of CD3^pos T cells and increase in PMN and MoC representation ( Figure 3 , Additional File 9 ). However, the CD3^pos cell percent among donor cells reached the lowest values in the CDR group at T10H (5.2 ± 1.2 in CDR, p < 0.01 vs UT and 9.2 ± 4.8 in CR, adj-p < 0.05 vs UT).

Altogether, preconditioning the donor with CS strongly reduced the presence of CD3^pos T cells originating from both the donor and the recipient, during the 10-hour perfusion in the lung.

In murine LT models, recipient classical MoCs and the donor AMs have been shown to play important roles in driving the primary graft dysfunction (2). We thus examined the effect of donor preconditioning on the activation and anti-inflammatory profile of recipient MoCs and donor AMs, in our pig model. For the activation profile analysis, the whole MoC and the AM populations were considered, as we were not able to distinguish the CD14^pos from the CD16^pos MoC subset, due to lack of convenient marker combinations. We found that preconditioning the donor strongly reduced the expression of MHC class II and CD80/86 on donor MoCs already at T0H, while this reduction needed 10-hour reperfusion to be reached in the CR group ( Additional File 10 ). However, preconditioning the donor did not reduce the activation markers’ expression on AMs. For the anti-inflammatory profile analysis, we measured the expression of inflammatory genes in the recipient CD14^pos and CD16^pos MoC subsets and in the donor AMs, and we calculated the IL10:TNFA, IL10:CXCL8, IL10:CCL2 and IL10:IL6 expression ratios ( Figure 5 , Additional File 11 ). Interestingly, in the AMs, the IL10:TNFA and IL10:CXCL8 ratios were more clearly enhanced in the CDR than in the CR group (adj-p < 0.05 for CDR vs UT in most cases, and there was no significance between CR vs UT). The anti-inflammatory balance was similarly higher in the recruited MoC subsets of the CDR and CR groups than in the UT group.

Therefore, preconditioning the donor in addition to treating the recipient with CS promotes the anti-inflammatory balance in AMs.

We explored putative mechanisms that may explain the CD3^pos T cell-decrease induced by preconditioning. We checked that cell death was not increased by the CS treatment in the CD3^pos T cells of the three groups ( Additional File 12 ). The CXCL9/CXCL10-CXCR3 axis is a target of the CS-mediated depletion of CD4^pos T cells in glomerulonephritis (18). As the CXCR3-interacting chemokines (CXCL9, CXCL10 and CXCL11) are produced by human AMs (19, 20), we analyzed their transcript expression in the single cell RNA-seq (scRNA-seq) data of MHC class II^pos pig lung cells ( Figure 6 ). The UMAP revealed 20 clusters that correspond to epithelial cells, fibroblasts, vascular cells, lymphocytes, dendritic cells, monocytes, and macrophages ( Figure 6A , Additional File 5 , see an interactive viewer https://tinyurl.com/yr3375sd). CXCL9, CXCL10 and CXCL11 were dominantly expressed by macrophage clusters that also express CD163, and almost not by lymphoid nor epithelial cells. CCL5, another T-cell attracting chemokine was expressed by all cell types ( Figure 6B ). These chemokine genes were also found expressed in inflammatory endothelial cells (cluster 10). We therefore sorted the CFSE^negCD163^hi cells (i.e. AMs) from our control and CS-treated groups. Compared to AMs from untreated pigs, AMs from CS pigs expressed less CXCL10 and CXCL11 mRNA at T0H, p < 0.05 ( Figure 7A ). Furthermore, upon cross-circulation, the CXCL10 and CXCL11 gene expression was significantly reduced in the AMs from the CDR group at T6H (adj-p < 0.05 vs UT) and T10H (adj-p< 0.01 vs UT), whereas it was less the case in the CR group ( Figure 7B ).

Therefore, preconditioning the pig lung donor with CS leads to a sustained decrease in T cell-attracting chemokine gene expression by AMs.

In order to assess whether CS are also able to modify the human lung immunological profile, total human lung cells from surgical resections were cultured with and without 20 µg/ml methylprednisolone - the plasma dose obtained in humans on bolus injection. First, CS significantly reduced CXCL10 protein produced by human lung cells ( Figure 8A , p < 0.005). As in pigs, scRNA-seq analysis showed that CXCL9, CXCL10 and CXCL11 mRNA were also mainly synthetized by lung macrophages and inflammatory endothelial cells while CCL5 transcripts were expressed by most cell types ( Additional File 13 ).

After 12-hour culture of human lung cells and cell detachment with accutase, we immunomagnetically-sorted the human macrophages based on CD172A expression. We did not use the CD163 marker as it is accutase-sensitive (21). The CS-treated human lung macrophages expressed significantly less CCL5, CXCL9, 10 and 11 transcripts compared to controls, in agreement with the in vivo pig results ( Figure 8B ). Furthermore, CS promoted an anti-inflammatory profile, with higher IL10:TNF, IL10:CXCL18 and IL10:IL6 transcript ratios in CS-treated than in control lung macrophages ( Figure 8C ). As in pigs, the IL10:CCL2 ratio was not significantly modified. These in vitro results suggest that the reduction of the T-cell attractive chemokine gene expression and the higher anti-inflammatory profile induced by preconditioning the donor in the pig model is likely to be translatable to humans.

Altogether our findings in a pig cross-circulation model and in vitro with human lung cells support that preconditioning the donor with CS improves the lung immunological environment.