Transcriptional analysis identifies potential novel biomarkers associated with successful ex‐vivo perfusion of human donor lungs

Abstract Background Transplantation is an effective treatment for end‐stage lung disease, but the donor organ shortage is a major problem. Ex‐vivo lung perfusion (EVLP) of extended criteria organs enables functional assessment to facilitate clinical decision‐making around utilization, but the molecular processes occurring during EVLP, and how they differ between more or less viable lungs, remain to be determined. Methods We used RNA sequencing of lung tissue to delineate changes in gene expression occurring in 10 donor lungs undergoing EVLP and compare lungs that were deemed non‐transplantable (n = 4) to those deemed transplantable (n = 6) following perfusion. Results We found that lungs deemed unsuitable for transplantation had increased induction of innate immune pathways and lower expression of oxidative phosphorylation related genes. Furthermore, the expression of SCGB1A1, a gene encoding an anti‐inflammatory secretoglobin CC10, and other club cell genes was significantly decreased in non‐transplantable lungs, while CHIT‐1 was increased. Using a larger validation cohort (n = 17), we confirmed that the ratio of CHIT1 and SCGB1A1 protein levels in lung perfusate have potential utility to distinguish transplantable from non‐transplantable lungs (AUC .81). Conclusions Together, our data identify novel biomarkers that may assist with pre‐transplant lung assessment, as well as pathways that may be amenable to therapeutic intervention during EVLPAQ6.


INTRODUCTION
Lung transplantation is an effective treatment for selected patients with life-threatening end-stage lung disease. Unfortunately, a shortage of donor organs limits access to transplantation for many patients who might benefit from this treatment. 1 One strategy to address this challenge is to utilize lungs from extended criteria donors (ECD) with advanced donor age, a smoking history, reduced lung function or comorbidities. 2,3 ECD lungs potentially have an increased risk of poor outcomes if sub-optimal organs are inadvertently used. 4,5 Ex-vivo lung perfusion (EVLP) provides an opportunity to functionally assess, and potentially recondition, lungs deemed unsuitable for immediate transplantation, to aid organ utilization decisions and increase the donor lung pool. 6,7 The decision to accept organs for transplantation after EVLP is currently based on visual appearance and physiological parameters such as oxygenation, lung compliance, pulmonary vascular resistance, and peak airway pressure. However, reported discard rates of perfused lungs are highly variable, ranging from 10 to 60%, suggesting that some donor lungs are being inappropriately used and others unnecessarily declined for transplant. 8 The cellular and molecular events occurring during EVLP have not been well characterized, and how these processes differ between lungs that are deemed useable compared with those that are discarded is unknown. In addition, there is a pressing need to identify novel predictive biomarkers that can be utilized during EVLP. Cellular injury during lung retrieval and storage results in the release of danger signals that activate immune cells, so-called sterile inflammation, causing further collateral tissue damage. 9 We previously demonstrated the feasibility of measuring pro-inflammatory and tissue injury signals in EVLP perfusion fluid to identify unsuitable donor lungs. 10 Furthermore, we recently demonstrated that interleukin (IL)-1β and tumor necrosis factor (TNF) levels in perfusate after 30 min of EVLP are predictive of in-hospital mortality post-transplantation. 10 Although these candidate proteins in perfusate may have potential utility as biomarkers, their selection is based on published literature that identified candidate proteins associated with primary lung graft dysfunction 11,12 and does not provide a comprehensive means of assessing the molecular pathways impacted by EVLP nor of identifying novel biomarkers.
Here we utilized RNA-sequencing (RNASeq) to generate an unbi-

METHODS
Further methods are given in the supplement.

Study subjects and protocol
In this study, we utilized samples of whole donor lung tissue col-

Sample collection
For collection of tissue for RNASeq a standardized sampling protocol was followed to allow collection from either the right middle lobe or lingula both before and after the EVLP assessment using a GIA surgical

RNASeq analysis
BCL files were demultiplexed using CASAVA, fastqs aligned to hg38 using hisat2, and counts determined using Featurecounts. Differential expression was carried out using DESeq2.

Transcriptomic analysis of lungs pre and post EVLP reveals induction of immune pathway genes
Between April 2012 and July 2014, 53 pairs of human donor lungs deemed unsuitable for immediate transplantation underwent EVLP in the DEVELOP-UK clinical trial. 13 Pre-and post-EVLP samples were available from n = 10 donor lungs, of which n = 6 "Passed" EVLP assessment and were deemed transplantable, and n = 4 "Failed" assessment and were deemed non-transplantable based on trial protocol criteria 13 (Table 1).
We investigated gene expression changes during EVLP by performing RNAseq on pre-and post-EVLP samples ( Figure 1A); 293 genes were significantly up-regulated, and 78 genes significantly downregulated during EVLP ( Figure 1B, C). Several immune genes were evident in the top 20 up-regulated genes, including CD83 and interferon regulatory factor (IRF)1 ( Figure 1C). We also observed an increase in components of the NFkB transcription factor complex during EVLP ( Figure S1A), as well as NFKBIE, a negative regulator of the NFkB pathway ( Figure 1C). Heat shock proteins (HSPs) are protein chaperones that contribute to the canonical protective cellular response to a variety of physiological stressors 14 but also act as innate immune stimulators when released extracellularly. 15 The HSP70 family proteins HSPA1A and HSPH1 (also known as HSP105) were among the genes most significantly upregulated during EVLP ( Figure 1C).
We next assessed changes in the expression of molecular pathways rather than individual expression genes. By considering the net expression of all the genes in a pathway rather than individual genes we are able to better identify patterns in the data whilst also gaining statistical power by reducing the multiple testing burden. Gene set enrichment analysis (GSEA) using gene-sets curated in the "Hallmarks" database confirmed immune system activation during EVLP, identifying "TNFA signalling via NFkB" pathway as the most enriched in the dataset, as well as enrichment of other immune pathway genes including "Inflammatory response", and "Interferon gamma response" pathways ( Figure 1D, E). Notably, a number of metabolic pathways were also enriched including "Oxidative phosphorylation" and "Fatty acid metabolism" ( Figure 1D), both of which contribute to the generation of adenosine triphosphate (ATP), a critical energy source for cells.
Together, these data show that during EVLP, a number of immune pathways are induced, as well as pathways involved in energy generation.  We assessed chemokine and cytokine gene expression, and found a significant increase in TNF, CXCL1 and CXCL5 (neutrophil-recruiting chemokines), and CCL20 (a chemoattractant for CCR6-expressing CD4 T cells and innate lymphoid cells) during EVLP ( Figure 1F). At a protein level, CCL2, CXCL8, and ICAM1 in the EVLP perfusate positively correlated with transcript levels in pre-EVLP biopsies ( Figure S1B). In contrast, post-EVLP transcript levels showed no significant correlation with any perfusate cytokine ( Figure S1C). This suggests that proteins released during perfusion reflect pathways already activated prior to EVLP.

Lungs deemed non-transplantable have increased induction of immune pathway genes following EVLP compared with those deemed transplantable
One of the aims of EVLP is to assess organ function in order to determine suitability for transplantation. Of the n = 10 lungs studies, n = 6 lungs were deemed transplantable ("Pass"), and n = 4 were deemed non-transplantable ("Fail") (Table 1, Figure 2A Figure 2B). CHIT1 is a hydrolase produced by activated alveolar macrophages and has been investigated as a serum biomarker in several lung diseases, including chronic obstructive pulmonary disease. 16 Chitinase-like proteins (including CHI3L1) are expressed by alveolar macrophages and play a role in lung inflammation, remodeling and fibrosis. 17 Taken together, this analysis suggests marked activation of alveolar macrophages in non-transplantable lungs.
Two genes were significantly upregulated in transplantable lungs, MTRNR2L1, (encoding humanin-like 1, one of a family of proteins that protect cells from oxidative stress 18,19 ), and SCGB1A1, a secretoglobin expressed by non-ciliated bronchiolar epithelial cells and a marker of club (previously known as "Clara") cells 20 ( Figure 2B). SCGB1A1 (also known as club cell protein 10 (CC10)), is one of several protective proteins produced by club cells. 21   The interaction of perfusion and outcome was compared and the resulting differential expression data was used to rank genes for GSEA analysis against the Hallmarks database. All significantly enriched pathways (FDR q value < .05) have been plotted, size of point is inversely correlated to the FDR q value, red points indicated positively enriched pathways and blue negatively enriched. (C). Individual enrichment plots from analysis in A. Line plot indicates the running enrichment score and violin plot the distribution of genes from the pathways of interest within the ranked gene list. (D). Transcriptional changes were assessed for pre and post EVLP separately for those organs which passed EVLP verses those that failed. These results were filtered for only the genes in the indicated pathway above each plot and violin plots of log2 fold changes for each gene were plotted. Blue is organs which have passed EVLP, and orange genes which have failed. The P value is for comparison of the two groups using a Mann-Whitney test. (E). Top 10 genes by rank of leading edge genes for the indicated GSEA pathway when comparing the interaction of outcome and perfusion stage

Increase in NLRP3 inflammasome-associated genes in lungs deemed non-transplantable
Perfusate IL-1β has been investigated as a predictive biomarker of successful EVLP and post-transplant outcome in ECD lungs. 23 The NLRP3 inflammasome acts as a major hub to generate IL-1β in response to damage-associated molecular patterns (DAMPs). 24 Inflammasome associated genes were increased during perfusion, particularly in lungs which failed EVLP assessment compared to those that passed ( Figure 4A, B). When considering all 10 lungs analyzed, we found a trend towards an increase in IL1B transcripts following EVLP, with non-transplantable (Fail) lungs showing variable and non-significantly higher IL1B transcripts pre-EVLP ( Figure 4C). IL1B gene expression levels in pre-( Figure 4D) but not post-EVLP (data not shown) biopsies significantly correlated with IL1β protein concentration in perfusate at the end of EVLP. Overall, these data suggest that NLRP3 inflammasome activation occurs both prior to and during EVLP, and that this process is more marked in lungs that are deemed non-transplantable after EVLP.
One effect of IL1β is to stimulate the production of neutrophilrecruiting chemokines, including CXCL8. There was a non-significant increase in CXCL8 transcripts following EVLP, particularly in nontransplantable lungs ( Figure 4E) and variable increases in other neutrophil-attracting chemokines and adhesion molecules ( Figure S3).

Assessment of RNA and protein correlates for potential biomarkers
We investigated whether the genes differentially expressed in "Pass" and "Fail" lungs ( Figure 2A) might be potentially useful biomarkers of successful EVLP. We selected two candidates from our RNASeq data, SCGB1A1 and CHIT1, because they were significantly differentially expressed genes in "Pass" and "Fail" lungs respectively, expressed at a reasonable level, encoded secreted proteins, and quantification reagents/assays were commercially available. Perfusate samples were available on n = 17 lungs from the DEVELOP UK study, taken 150 min after the start of perfusion. 7/17 of these lungs were part of the previously described RNASeq study. 9/17 lungs passed EVLP assessment and n = 8 failed and were deemed non-transplantable. We observed an increase in CHIT1 and a decrease in SCGB1A1 in the perfusate of nontransplantable lungs ( Figure 6A). The ratio of these two measurements was a better predictor of EVLP outcome compared to either individually, with a reasonable positive and negative predictive value for lung utilization when applied to a receiver operating characteristic (ROC) analysis (Area under the curve = .8, Figure 6B).

F I G U R E 5
Effect of perfusion on the expression of heat shock protein transcripts. (A). Heatmap of all heat shock protein transcripts in perfusion. The expression profiles were hierarchically clustered and three groups identified using a k-means approach. (B). Transcriptional changes were assessed for pre and post EVLP separately for those organs which passed EVLP versus those that failed for each group of HSP genes. Blue is organs which have passed EVLP and orange genes which have failed. The P value is for comparison of the two groups using a Mann-Whitney test F I G U R E 6 Validation of potential protein level biomarkers for success of perfusion in n = 18. (A). Concentration of CHIT1 in perfusate samples taken after 150 min of perfusion with the Lund protocol. Data has been normalized for total lung capacity. (B). Concentration of SCGB1A1 as for A. (C). A ratio of the proteins from A and B was calculated to produce a test statistic which was used for subsequent ROC analysis

DISCUSSION
In order for EVLP to be effectively utilized in clinical practice as a platform for the functional assessment, reconditioning and potential treatment of donor lungs, an increased understanding of the cellular and molecular events occurring is required. In this study, we aimed to address this knowledge gap by performing an unbiased analysis of the lung transcriptome prior to and following EVLP. This revealed that several hundred genes were differentially expressed during EVLP, with an induction in a number of innate immune (including TNF and IL1B) and HSP genes. These data suggest that the tissue damage that occurs during donor lung procurement and hypothermic transport has a substantial capacity to stimulate sterile inflammation during EVLP. Furthermore, the increase in HSPs, (which form part of a canonical cellu-lar stress response), demonstrate the activation of cell-intrinsic compensatory mechanisms to promote organ viability. HSPs were originally categorized into several families according to their molecular weight, with HSPB family members among those classified as "small" HSPs. Our analysis showed three patterns of HSP gene expression in lungs, and unusable lungs failed to upregulate a subset of HSP genes (dominated by HSPB genes that encode small HSPs). Of note, transgenic over-expression of the small HSP, HSPB1 (HSP27) has been shown to protect murine hearts and livers from apoptosis in the context of ischemia re-perfusion injury 25,26 and HSPB1 also has antiinflammatory effects in airways epithelial cells. 27 These data suggest that detection of upregulation of this HSP gene subset might be a useful indicator of more suitable lungs that warrants assessment in larger datasets.
Our analysis of post EVLP biopsies identified 12 genes that were statistically significantly upregulated in unusable lungs, and two genes that were upregulated in transplantable lungs. These genes may have potential as novel transcriptional biomarkers but are also likely to reflect functionally significant pathways activated in Pass and Fail lungs. CHIT1 (encoding chitotriosidase) was the gene most significantly upregulated in unusable lungs. CHIT1 is the most highly expressed chitinase in humans 28 and is secreted by activated macrophages.
Murine models suggest that CHIT1 may enhance TGFβ receptor signaling following lung injury. 29 Furthermore, in humans, increased CHIT1 expression has been found in areas of the lung affected by granulomata and fibrosis in tuberculosis, sarcoidosis, idiopathic pulmonary fibrosis, scleroderma, and chronic obstructive lung diseases. Our data raise the possibility that CHIT1 pre-transplant expression in lungs may be a useful biomarker that identifies donor lungs likely to perform poorly during EVLP.
CHI3L1 was also significantly upregulated in unusable lungs.
Chitinase shown to be induced in alveolar macrophages during mycobacterial infection. 32 Other genes upregulated in non-transplantable lungs included SLC7A5 and GBP5. SLC7A5 is an amino acid transporter that mediates lysine influx and contributes to IL1β production via mTOR complex 1 (mTORC1)-induced glycolytic reprograming of activated human monocytes 33 whilst GBP5 promotes selective NLRP3 inflammasome assembly in macrophages in response to soluble stimuli. 34 Taken together, our data suggest that macrophage activation is more marked in unusable lungs compared with those deemed transplantable.
Whether this is due to higher levels of tissue damage and associated DAMPs, or due to a higher macrophage-intrinsic capacity to respond to DAMPs remains to be determined.
Two genes were significantly upregulated in transplantable lungs, MTRNR2L1, (encoding humanin-like 1), and SCGB1A1, a secretoglobin expressed by club cells. 20 The humanin gene family encodes for small secretory peptides 35 that have been shown to protect cells from oxidative stress, starvation and hypoxia 19 and MTRNR2L1 transcription is increased by oxidative stress and toll-like receptor 4 agonists, 36 therefore, its upregulation may enhance cellular integrity in lungs in the face of cell stressors associated with the organ procurement process.
Club cells are the major secretory cells of the human small airways that have anti-protease activity via the secretion of α1-antitrypsin (an important neutrophil elastase inhibitor) and SLPI (secretory leucoprotease inhibitor, a potent serine protease inhibitor). 20 SCGB1A1 has been shown to have direct anti-inflammatory actions inhibiting NFkB activation and CXCL8 secretion in airway epithelial cells. 21,37 Our data raises the possibility that club cells function to protect lungs from some of the deleterious processes occurring during the organ donation and procurement process and to inhibit collateral damage caused by neu-trophils recruited following reperfusion. It will be possible to test this hypothesis in future studies by investigating club cell number and activation in pre-implantation biopsies in a larger cohort of transplanted lungs.
The major limitation of our study is the relatively small sample size (n = 10) but this is not dissimilar to other transcriptomic studies in lung transplantation. 38,39 However, our use of paired pre-and post-EVLP samples, as well as gene pathway analysis, substantially increased our statistical power to demonstrate meaningful increases in functionally relevant molecular pathways. Our study identified a number of novel RNA biomarkers that may indicate successful EVLP and we went on to measure the protein concentration of two of these biomarkers, CHIT1 and SCGB1A1, in perfusate in a larger validation cohort. This analysis indicated that these two measurements can be combined to predict successful EVLP with reasonable sensitivity and specificity, and merit prospective assessment in future clinical trials. Our results also suggest a number of potential targets for therapeutic intervention to improve lung suitability during EVLP that will require validation in larger studies. In particular, inhibition of the NLRP3 inflammasome and enhancement of club cell function may be of utility. The former is readily translatable, given the development of these agents for the treatment of sterile inflammation in other contexts, such as crystal-induced arthropathies and autoinflammatory conditions. 40 In summary, our data suggest that lungs deemed suitable for transplantation following EVLP have reduced induction of multiple immune pathways and are better able to generate ATP. Our study identifies specific biomarkers that may hold utility in identifying suitable lungs and pathways amenable to therapeutic intervention during EVLP that will inform the design of future clinical trials in this area.

CONFLICTS OF INTEREST
None of the authors have any finical conflicts of interest to disclose. wrote the manuscript, and A.J.F. edited the manuscript. All authors reviewed and commented on the manuscript.

DATA AVAILABILITY STATEMENT
Data will be made available following all reasonable requests to the corresponding author.