Analysis of perfusate and bronchoalveolar lavage fluid
Assessed lungs were divided into two groups based on decision after EVLP: transplanted (n = 7) or non-transplanted (n = 9). Decision on transplant suitability was based on strict criteria summarized in Supplementary Table S2. Samples were analysed retrospectively for protein expressions.
For an initial general assessment of tissue injury, we measured LDH levels. LDH was found in substantially lower levels in perfusate from lungs meeting criteria for transplant (T) after EVLP compared to those that did not meet the criteria for transplant (NT) lungs. Suitability for transplant was evident after 2 h of perfusion: mean (T) = 0.150 U (SD 0.030, n = 14) and mean (NT) = 0.223 U (SD 0.082, n = 18) [t(30) = 2.77, P = 0.03] (Fig. 2). The LDH perfusate levels remained significantly different when assessed over all sample time points: mean (T) = 0.149 U (SD 0.033, n = 24) and mean (NT) = 0.259 U (SD 0.088, n = 40) [t(62) = 5.45, P < 0.001]. No difference in LDH levels was seen in BALF samples collected pre- and post-perfusion.
Figure 2: Protein expressions during EVLP in perfusate from transplanted and non-transplanted donor lungs. Protein levels expressed as medians with interquartile ranges. **P<0.05. EVLP: ex vivo lung perfusion; LDH: lactate dehydrogenase; IL: interleukin; TNF-α: tumour necrosis factor alpha; HMGB-1: high-mobility group box-1; Time: hours of ex vivo lung perfusion after start of pulmonary artery perfusion.
To assess the integrity of the pulmonary vascular compartment, we measured Syndecan-1 levels as a marker of endothelial glycocaylx disruption. Syndecan-1 levels were significantly lower in perfusate samples from transplanted donor lungs: mean (T) = 111 (pg/ml) (SD 98, n = 24) compared to mean (NT) = 281 (pg/ml) (SD 219, n = 40) [t(62) = 4.14, P < 0.001] (Fig. 3).
Figure 3: Protein expressions in perfusate and BALF from transplanted and non-transplanted EVLP donor lungs. Interleaved scatter plots with molecular marker levels expressed in (pg/ml) and lines representing means. *P<0.1. **P<0.05. ***P<0.001. BALF: bronchoalveolar lavage fluid; EVLP: ex vivo lung perfusion; IL: interleukin; TNF-α: tumour necrosis factor alpha; LDH: lactate dehydrogenase; HMGB-1: high-mobility group box-1.
We noted a similarly consistent pattern towards lower release of proinflammatory cytokines into the circulating perfusate from transplanted compared to non-transplanted lungs. The average level of IL-8 was nearly eight times higher in perfusate samples from non-transplanted lungs: mean (T) = 165 (pg/ml) (SD 155, n = 24) compared to mean (NT) = 1310 (pg/ml) (SD 1510, n = 35) [t(57) = 2.27, P = 0.06]. Levels of the anti-inflammatory cytokine IL-10 increased noticeably over the first 2 h of perfusion in the transplanted group, with high initial fold changes in both perfusate and BALF (Fig. 4). Perfusate levels then diminished and were less than half those in non-transplanted lungs when measured over the full assessment: mean (T) = 9 (pg/ml) (SD 11, n = 24) compared to mean (NT) = 23 (pg/ml) (SD 32, n = 40) [t(57) = 2.02, P = 0.07]. No difference in levels of IL-10 in BALF was seen between the groups.
Figure 4: Protein median fold changes in perfusate and BALF from EVLP donor lungs. Bar graphs showing median fold changes with interquartile range. BALF: bronchoalveolar lavage fluid; EVLP: ex vivo lung perfusion; IL: interleukin; TNF-α: tumour necrosis factor alpha; LDH: lactate dehydrogenase; HMGB-1: high-mobility group box-1.
A panel of three damage-associated molecular patterns, IL-33, HMGB-1 and S100A9, was used to assess the extent of cellular injury in the donor lung. Lower levels of IL-33 and HMGB-1 were found in perfusate from transplanted donor lungs. The average IL-33 levels in perfusate were mean (T) = 3 (pg/ml) (SD 5, n = 24) and mean (NT) = 9 (pg/ml) (SD 10, n = 40) [t(62) = 2.93, P = 0.02]. HMGB-1 was highly expressed in both perfusate and BALF. The HMGB-1 levels in perfusate were mean (T) = 8318 (pg/ml) (SD 4474, n = 24) and mean (NT) = 10 545 (pg/ml) (SD 4,313, n = 40) [t(62) = 2.10, P = 0.07]. No difference in lavage fluid levels of IL-33 or HMGB-1 was seen between the groups.
We noted a consistent pattern towards increasing perfusate protein levels and a more pronounced separation between transplanted and non-transplanted lungs over time of perfusion (Fig. 2). This pattern was most noticeable for the investigated inflammatory cytokines. The only protein marker not showing this pattern was S100A9, which appeared lower in the non-transplanted group.
A multivariate analysis model was fitted to assess the predictive value of combining two or more protein markers. The optimal model was established by leave-one-out cross-validation and is shown in Fig. 5 as a scatterplot of IL-1β and IL-8 perfusate levels from the 16 assessments. The shaded prediction region was derived from the regression analysis with covariates IL-1β, IL-8 and (IL-8)2 after 2 h of perfusion. The model suggests that the closer to the centre of the area, the higher was the probability of a donor lung being found suitable for transplant after EVLP assessment. This early IL-1β–IL-8 signature remained equally as robust when applied to the full set of samples (Fig. 5B). An ROC curve showing the potential benefit of the model in predicting EVLP outcome compared to random is shown in Fig. 5C, with a sensitivity of 0.81 and specificity of 0.92 at 2 h of perfusion.
Figure 5: (A) Early IL-1β-IL-8 signature in EVLP perfusate after 2 h of perfusion. (B) IL-1β-IL-8 signature in EVLP perfusate after full assessment. (C) ROC curve. (A) and (B) Scatter plots of IL-1β and IL-8 levels in perfusate samples from transplanted and non-transplanted EVLP donor lungs. The shaded area is the prediction region from the optimal logistic regression model with covariates IL-1β, IL-8 and (IL-8)2. (C) Receiver operating characteristic (ROC) curve of the model as a predictor of EVLP outcome. Calculated sensitivity (y axis) is plotted against the 1-specificity formula (x axis) of the logistic regression function 2 h into perfusion and after full assessment.
EVLP: ex vivo lung perfusion; IL: interleukin.
The modelled proteins were lastly assessed for correlations to post-transplant outcome measures. Donor IL-1β levels in the perfusate at time of transplant decision showed a significant negative correlation to recipient PaO2:FiO2 24 h post-transplant (r = −093, P = 0.03) (Fig. 6). No other correlations remained significant after corrections for multiple testing in this cohort.
Figure 6: Association between donor IL-1β level in EVLP perfusate at the time of transplant decision and recipient P/F ratio 24 h post lung transplantation. Pearson correlation plot. **P<0.05. EVLP: ex vivo lung perfusion; IL: interleukin; PaO2:FiO2: partial pressure of oxygen on 100% inspired oxygen.