Association between the renin–angiotensin system and chronic lung allograft dysfunction
- Gregory Berra1,2,10,
- Sofia Farkona2,10,
- Zahraa Mohammed-Ali2,
- Max Kotlyar3,
- Liran Levy1,2,
- Sergi Clotet-Freixas2,
- Phillip Ly1,2,
- Benjamin Renaud-Picard1,2,
- Guan Zehong1,2,
- Tina Daigneault1,2,
- Allen Duong1,2,
- Ihor Batruch4,
- Igor Jurisica3,5,
- Ana Konvalinka2,6,7,8,9,11⇑ and
- Tereza Martinu1,2,6,8,11
- 1Toronto Lung Transplant Program, University Health Network, Toronto, ON, Canada
- 2Toronto General Hospital Research Institute, University Health Network, Toronto, ON, Canada
- 3Krembil Research Institute, University Health Network, Toronto, ON, Canada
- 4Dept of Laboratory Medicine and Pathobiology, Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, University of Toronto, Toronto, ON, Canada
- 5Depts of Medical Biophysics and Computer Science, University of Toronto, Toronto, ON, Canada
- 6Multi-Organ Transplant Program, University Health Network, Toronto, ON, Canada
- 7Dept of Medicine, Division of Nephrology, University Health Network, Toronto, ON, Canada
- 8Institute of Medical Science, University of Toronto, Toronto, ON, Canada
- 9Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada
- 10G. Berra and S. Farkona contributed equally as first authors
- 11A. Konvalinka and T. Martinu contributed equally as lead authors and supervised the work
- Corresponding author: Ana Konvalinka (ana.konvalinka{at}uhn.ca)
Abstract
Chronic lung allograft dysfunction (CLAD) is the major cause of death after lung transplantation. Angiotensin II (AngII), the main effector of the renin–angiotensin system, elicits fibrosis in both kidney and lung. We identified six AngII-regulated proteins (Ras homolog family member B (RHOB), bone marrow stromal cell antigen 1 (BST1), lysophospholipase 1 (LYPA1), glutamine synthetase (GLNA), thrombospondin 1 (TSP1) and laminin subunit β2 (LAMB2)) that were increased in urine of patients with kidney allograft fibrosis. We hypothesised that the renin–angiotensin system is active in CLAD and that AngII-regulated proteins are increased in bronchoalveolar lavage fluid (BAL) of CLAD patients.
We performed immunostaining of AngII receptors (AGTR1 and AGTR2), TSP1 and GLNA in 10 CLAD lungs and five controls. Using mass spectrometry, we quantified peptides corresponding to AngII-regulated proteins in BAL of 40 lung transplant recipients (stable, acute lung allograft dysfunction (ALAD) and CLAD). Machine learning algorithms were developed to predict CLAD based on BAL peptide concentrations.
Immunostaining demonstrated significantly more AGTR1+ cells in CLAD versus control lungs (p=0.02). TSP1 and GLNA immunostaining positively correlated with the degree of lung fibrosis (R2=0.42 and 0.57, respectively). In BAL, we noted a trend towards higher concentrations of AngII-regulated peptides in patients with CLAD at the time of bronchoscopy, and significantly higher concentrations of BST1, GLNA and RHOB peptides in patients that developed CLAD at follow-up (p<0.05). The support vector machine classifier discriminated CLAD from stable and ALAD patients at the time of bronchoscopy (area under the curve (AUC) 0.86) and accurately predicted subsequent CLAD development (AUC 0.97).
Proteins involved in the renin–angiotensin system are increased in CLAD lungs and BAL. AngII-regulated peptides measured in BAL may accurately identify patients with CLAD and predict subsequent CLAD development.
Abstract
Components of the renin–angiotensin system are increased in chronic lung allograft dysfunction (CLAD) fibrosis. Angiotensin II-regulated proteins in bronchoalveolar lavage identify concurrent and predict future CLAD in lung transplant recipients. https://bit.ly/3eRYez7
Introduction
Lung transplantation is a life-saving therapy for patients with end-stage lung disease. Despite improvements in early post-operative mortality and long-term outcomes, survival after lung transplant remains poor compared with other solid organ transplants [1]. Chronic lung allograft dysfunction (CLAD) is the major long-term cause of death after lung transplant, affecting ∼50% of recipients at 5 years post-transplant. Two main clinical CLAD phenotypes, both characterised by fibrosis in different compartments of the lung, have been identified: bronchiolitis obliterans syndrome (BOS), marked by fibrosis within small airways, leading to obstructive pulmonary physiology, and restrictive allograft syndrome (RAS), a more aggressive disease seen in ∼20% of CLAD patients, characterised by diffuse airway and also parenchymal lung fibrosis [2–8]. CLAD mechanisms remain poorly understood and there are no effective treatments. Finding early markers of CLAD and identifying new therapeutic targets represent key goals in lung transplantation.
Acute rejections, infections and environmental exposures drive chronic inflammation and lung allograft injury. The chronic injury, in turn, leads to upregulation of pro-fibrotic mediators, which augment fibroblast proliferation, migration and extracellular matrix (ECM) deposition [9–11]. Studies of tissue fibrogenesis have shown that the “local” renin–angiotensin system mediates fibrosis in multiple organs, including kidney and lung [12–16]. Angiotensin II (AngII), the main effector of the renin–angiotensin system, promotes fibrosis in both kidney and lung [17]. Based on similarities between CLAD fibrosis and kidney allograft interstitial fibrosis and tubular atrophy (IFTA), an important cause of kidney allograft dysfunction and loss, we propose that mechanisms at play in IFTA may also be relevant in CLAD.
We recently identified six AngII-regulated proteins (Ras homolog family member B (RHOB), bone marrow stromal cell antigen 1 (BST1), lysophospholipase 1 (LYPA1), glutamine synthetase (GLNA), thrombospondin 1 (TSP1) and laminin subunit β2 (LAMB2)) in primary kidney cells, and demonstrated that these proteins are dysregulated in kidney grafts with IFTA and increased in urine of IFTA patients [18–20]. For the sake of simplicity, we will be using the protein name to refer to both the protein and respective genes (RHOB, BST1, LYPLA1, GLUL, THBS1 and LAMB2) throughout this article. There is evidence that these proteins are associated with, or implicated in, fibrosis in the lung. For instance, TSP1 and LAMB2 were significantly increased in lung tissue from patients with idiopathic pulmonary fibrosis compared with controls [21]. TSP1 catalyses the activation of transforming growth factor (TGF)-β, and is implicated in alveolar macrophage-dependent TGF-β activation in mouse and rat models of bleomycin-induced pulmonary fibrosis [22, 23]. While it is unknown how LAMB2 contributes to idiopathic pulmonary fibrosis pathobiology, it is associated with cell adhesion, differentiation and migration [24], which have all been identified as important factors in fibrogenesis [25, 26]. miR-29, a microRNA involved in LYPA1 regulation [27], is a central rheostat of fibrosis in different organs and model systems [28–31]. More specifically, restoration or overexpression of miR-29 blocked the initiation and progression of fibrosis in mouse lung, kidney and heart in vivo [31]. Lastly, BST1 was linked to lung inflammation, radiation fibrosis of the lung [32] and bleomycin-induced fibrosis in mice [33]. To further support our rationale, we built an interactome that shows that these six AngII-regulated proteins are expressed both in kidney and lung, interact closely with the AngII receptors, and are functionally associated with known pro-fibrotic proteins (supplementary material and supplementary figure S1).
Based on the aforementioned available data, we hypothesised that AngII is active in CLAD fibrosis and that AngII-regulated proteins are measurable in bronchoalveolar lavage fluid (BAL) of lung transplant patients as potential markers of CLAD. To test this hypothesis, we sought to determine the expression of AngII receptors (AGTR1 and AGTR2) and AngII-regulated proteins in lung tissue and BAL from CLAD and control patients. We also evaluated whether AngII-regulated proteins in BAL could predict subsequent CLAD development in patients at risk of CLAD.
Patients, materials and methods
Patient selection and sample collection
This study was approved by the University Health Network (Toronto, ON, Canada) Research Ethics Board (protocol 15-9531). Specimens were collected at Toronto General Hospital after written informed consent. The experimental workflow is presented in figure 1. For details, see supplementary material.
Cohort for lung tissue reverse-transcriptase quantitative real-time PCR, immunostaining and Western blotting
CLAD lung tissue was obtained from 33 patients undergoing retransplantation or autopsy between May 2015 and August 2019. CLAD diagnosis and phenotype were adjudicated according the 2019 International Society for Heart and Lung Transplantation consensus [2] by at least two independent reviewers [34]. Control lung tissue was obtained from 10 donor lungs at the end of cold ischaemic time prior to transplantation and from 10 patients undergoing lobar resection for suspected cancer. See figure 2a and supplementary material.
Cohort for BAL protein analysis
This cohort was selected from 224 consecutive lung transplantation patients who underwent a post-lung transplantation surveillance or indication bronchoscopy at our centre between January and April 2019 (bronchoscopy and BAL protocol details are provided in the supplementary material). 20 mL samples of BAL were sent for clinical assessment and any excess was processed for research. After BAL centrifugation at 3220×g for 20 min, the supernatant was aliquoted and stored at −80°C. All patients were prospectively screened for CLAD and acute lung allograft dysfunction (ALAD). ALAD was defined as an acute drop in forced expiratory volume in 1 s (FEV1) ≥10% compared with the highest FEV1 in the previous 4 months or the immediately preceding FEV1 available, in the absence of extrapulmonary causes. 17 control patients with stable lung function were randomly selected from the 175 patients without FEV1 drop and with sufficient BAL. The concentrations of each peptide and combinations of these peptides were compared between three final groups: 17 stable patients, 10 ALAD patients and 13 CLAD patients (n=40). Of these 40 patients, 27 patients did not have CLAD at the time of bronchoscopy (17 stable patients and 10 ALAD patients). The CLAD status of these 27 patients was reassessed at follow-up, using the last available FEV1 for each patient obtained 9–12 months after bronchoscopy. See figure 2b and supplementary material.
Analyses of lung tissue and BAL
Paraffin-embedded lung tissue was assessed using haematoxylin/eosin, elastic trichrome and immunostaining for AGTR1, AGTR2, TSP1, GLNA, PanCK, CD45 and α-smooth muscle actin (α-SMA). AGTR1, AGTR2, TSP1, GLNA, LYPA1 and BST1 transcripts were measured using reverse-transcriptase quantitative real-time PCR (RT-qPCR). TSP1 and GLNA protein expression levels were assessed by Western blotting. Targeted mass spectrometry assays were developed to quantify AngII-regulated proteins RHOB, BST1, LYPA1, GLNA, TSP1 and LAMB2 in BAL. Classifiers based on combined peptide concentrations were built using machine learning algorithms, and their diagnostic accuracy and predictive value were assessed in discriminating or predicting CLAD. See figure 1 and supplementary material.
Statistical analyses
Prism version 8.0 (GraphPad, San Diego, CA, USA) was used for the analysis of group characteristics: the t-test or Wilcoxon–Mann–Whitney test was used for continuous variables; the Chi-squared test and Fisher's exact test were used for categorical variables. p<0.05 was considered significant.
All other statistical and machine learning analyses were conducted in R version 3.5.1 [35]. Associations between peptide concentrations and patient clinical characteristics were tested: Spearman correlation for continuous variables, Mann–Whitney U-tests for binary patient characteristics and Kruskal–Wallis tests for ordinal patient characteristics. A heatmap of unadjusted p-values was plotted using the ggplot2 3.3.0 package [36]. Peptide concentrations were used to predict CLAD status, using four machine learning methods (mlr 2.15.0 package [37]): linear discriminant analysis, naïve Bayes, quadratic discriminant analysis and support vector machine (SVM). Predictions were done using leave-one-out cross-validation. Receiver operating characteristic (ROC) curves from predictions were plotted using the pROC 1.15.3 package [38]. p-values of predictions were calculated using the verification 1.42 package [39].
Results
AngII receptors and AngII-regulated proteins and transcripts in explanted CLAD lungs
We analysed AngII receptors (AGTR1 and AGTR2) and AngII-regulated proteins (TSP1 and GLNA) in 33 CLAD and 20 control lungs (figure 2a). CLAD patients had a significantly lower median (interquartile range (IQR)) age at sample collection (30 (24–48) years) compared with controls (62 (40–68) years), but the lung allograft age was comparable to the age of the control lungs. There was no significant difference in sex or smoking exposure history between CLAD and control patients. Of 33 CLAD specimens, 15 displayed a RAS or mixed phenotype and 18 had a BOS phenotype. Detailed patient characteristics are provided in supplementary table S1. The cohort used for immunostaining is a subset of the RT-qPCR cohort, with similar clinical characteristics as the larger RT-qPCR cohort (supplementary table S2).
As evident from trichrome staining, CLAD lungs had higher ECM deposition compared with controls (figure 3a). Quantification of immunostaining demonstrates significantly higher numbers of AGTR1+ cells in CLAD compared with controls (unpaired t-test, p=0.022). There was no difference in the mean number of AGTR1+ cells in BOS versus RAS (figure 3b and supplementary figure S2). Moreover, the mean number of AGTR2+ cells was not significantly different in CLAD versus controls or in BOS versus RAS (figure 3b). Quantifying the staining as the percentage of the total number of cells did not affect the results (supplementary figure S3). Finally, the AGTR1/AGTR2 ratio was significantly higher in CLAD versus controls but not in BOS versus RAS (figure 3b). Both AGTR1 and AGTR2 co-stained with CD45, but not with α-SMA or PanCK (supplementary figure S4).
Surprisingly, AGTR1 gene expression was 4 times higher in controls compared with CLAD lungs (Mann–Whitney test, p<0.05). In contrast, AGTR2 transcript levels tended to be higher in CLAD than in controls (Mann–Whitney test, p=0.07). There were no statistically significant differences in AGTR1 or AGTR2 transcript levels between BOS and RAS (supplementary figures S5–S7).
TSP1 and GLNA staining was assessed in the same cohort of CLAD and control lungs (figure 4a, and supplementary figures S8 and S9). Positive TSP1 staining was identified in areas of dense fibrosis. Cells in the alveolar regions stained positively for both TSP1 and GLNA. We noted a statistically higher immunoreactivity of GLNA in CLAD versus controls (Mann–Whitney test, p<0.0013), and consistent trends towards higher staining of TSP1 in CLAD versus controls and of both GLNA and TSP1 in RAS versus BOS lungs (figure 4b). TSP1 (p=0.03) and GLNA (p=0.20) protein expression levels (supplementary material) were higher in the RAS cases compared with controls by Western blotting (supplementary figure S10). Immunofluorescence for TSP1 and GLNA showed coexpression with CD68 (monocyte/macrophage marker) (supplementary figure S11).
We found a significant positive correlation between AngII-regulated proteins TSP1 and GLNA and the extent of parenchymal fibrosis in each CLAD specimen, estimated by blinded histological scoring (figure 4b, supplementary material and supplementary figure S12). When dividing the patients into two groups based on fibrosis extent, higher TSP1 and GLNA immunoreactivity was identified in patients with ≥30% parenchymal fibrosis (figure 4b).
TSP1 gene expression was significantly higher in CLAD compared with control lungs (Mann–Whitney test, p=0.02) and in RAS compared with BOS (Mann–Whitney test, p=0.006) (figure 4c). GLNA, LYPA1 and BST1 transcripts were similar in CLAD versus controls and in BOS versus RAS lungs (figure 4c, and supplementary figures S6, S7 and S13).
Quantification of AngII-regulated proteins in BAL of lung transplant recipients
Baseline patient characteristics at the time of bronchoscopy were comparable between stable, ALAD and CLAD patients (table 1).
Using optimised parallel reaction monitoring (PRM) assays (supplementary figures S14 and S15), we detected two native peptides of TSP1, BST1, GLNA and LYPA1 protein and one RHOB peptide in a pool of BAL from controls and CLAD patients. We constructed linearity curves (supplementary figure S16) for BST1, GLNA, LYPA1 and RHOB. LAMB2 protein was not detected in BAL. TSP1 was detected in the pooled BAL, but at levels which would not make the quantification confident.
We then quantified AngII-regulated peptides in individual BAL samples. First, we examined whether AngII-regulated peptide concentrations correlated with clinical characteristics at the time of bronchoscopy (figure 5a). AngII-regulated peptide concentrations correlated significantly and directly with the degree of neutrophilic inflammation and cytomegalovirus (CMV) viraemia. GLNA, RHOB and LYPA1 peptides also correlated with the recipient's sex. Importantly, there was no correlation between the peptide concentrations and the time elapsed between transplantation and BAL collection.
Second, we compared AngII-regulated peptides between patient groups, based on the presence of CLAD or ALAD at the time of bronchoscopy. Although a trend was observed for higher concentrations of RHOB, BST1 and GLNA peptides in BAL from CLAD patients compared with the other two groups, this difference was not statistically significant (figure 5b). We then examined whether the concentrations of individual peptides and combinations of the seven peptides using machine learning approaches could segregate patient groups. Individual peptide concentrations were marginally better compared with chance alone (figure 5c). However, the SVM classifier accurately discriminated CLAD from stable patients with area under the curve (AUC) 0.92 (figure 5c). The SVM classifier also discriminated between CLAD and no CLAD (stable+ALAD) with AUC 0.86 (figure 5d).
We wondered whether AngII-regulated protein concentrations in BAL could predict development of CLAD in the subsequent 12 months. We thus focused on the 27 patients in the BAL cohort who did not have a diagnosis of CLAD at the time of bronchoscopy, but who had a subsequent 9–12-month follow-up FEV1 (figure 2b). The baseline clinical and bronchoscopy-based characteristics were largely comparable between patients who developed CLAD at follow-up (n=6) and patients who did not (n=21) (table 2).
Concentrations of RHOB, BST1 and GLNA peptides (IQAYDYLECSAK, GFFADYEIPNLQK, ALLSPEQR and LVLCEVFK) at the time of bronchoscopy were significantly higher in patients who developed CLAD at follow-up compared with those who did not (figure 6a). We noted a consistent trend in the same direction for the remaining peptides. We then examined the ability of individual or combined peptide concentrations to predict CLAD development using machine learning. Individual peptide concentrations had a modest-to-good accuracy with AUC 0.58–0.80 (figure 6b). The SVM classifier that included combined peptide concentration had excellent accuracy with AUC 0.97 in predicting subsequent CLAD development (figure 6b).
Discussion
We examined the presence of AngII receptors and our previously reported AngII-regulated proteins in CLAD lungs, and measured the concentration of AngII-regulated proteins in BAL from lung transplantation recipients. We made three major observations: 1) AngII receptor AGTR1 is increased in CLAD versus control lungs and AngII-regulated proteins TSP1 and GLNA correlate with the degree of lung fibrosis; 2) classifiers based on combined AngII-regulated peptides measured in BAL accurately distinguish CLAD from stable and ALAD patients; and 3) AngII-regulated peptides in BAL predict CLAD development.
Applying immunofluorescence, we found significantly higher numbers of AGTR1+ cells, with a higher ratio of AGTR1+/AGTR2+ cells in CLAD. The renin–angiotensin system plays an important role in fibrosis of the lung and other organs, and this system may have an instrumental role in CLAD. Pre-clinical and clinical studies showed that the renin–angiotensin system and its main effector AngII are active in lung fibrosis [40–45]. These deleterious effects can be reversed by renin–angiotensin system inhibitors in mice [40]. AGTR1 is the mediator of pro-fibrotic and pro-inflammatory effects of AngII [44]. A consistently higher expression of AGTR1 has been shown in fibrotic lung conditions [15, 16, 41, 42]. In contrast, AGTR2 is thought to have more modulatory effects [15, 44]. While both AGTR1 and AGTR2 increase during wound healing [46], existing data on AGTR2 expression in different fibrotic conditions are conflicting [47]. Surprisingly and contrary to the protein levels, transcript levels of AGTR1 were lower and transcript levels of AGTR2 were higher in CLAD. While we have no definitive explanation for this observation, we formulate several hypotheses. First, CLAD samples displayed advanced fibrosis, where active transcription may have been reduced. Downregulation of AngII receptor transcripts in advanced fibrosis was shown previously [48]. Second, AngII receptor activation, via a negative feedback loop, may induce transcriptional downregulation, perhaps through the β-arrestin pathway [49]. Treatment of mesangial cells with AngII markedly suppressed AGTR1 mRNA expression [50]. Third, NF-κB inhibition by calcineurin inhibitors in lung transplant recipients downregulates AGTR1 transcription [51]. We next examined the cell types expressing AngII receptors and found positive co-staining for both AGTR1 and AGTR2 with CD45+ cells, suggesting the expression of AngII receptors by immune cells, as previously reported [52–54].
Among AngII-regulated proteins, GLNA displayed higher immunostaining, while TSP1 displayed a trend towards an increase in CLAD. Their immunostaining correlated with the extent of parenchymal fibrosis within CLAD lungs. Furthermore, RT-qPCR analysis showed significantly higher TSP1 expression in CLAD. TSP1 is a structural and regulatory ECM glycoprotein [55]. TSP1 mediates kidney fibrosis in vitro [56] and in vivo [57, 58], and has been linked to ischaemia reperfusion injury and fibrosis in kidney allografts in both animal models and patients [59–61]. Finally, TSP1 mRNA levels were significantly increased in urine cell pellets from kidney transplant patients with fibrosis (IFTA) compared with controls [62], while TSP1 urine protein excretion was significantly increased in IFTA [20]. GLNA is the ATP-dependent enzyme that catalyses the synthesis of glutamine from glutamate and ammonia. GLNA is required for proliferation of fetal skin fibroblasts, and is expressed in fibroblasts and immune cells of the tumour micro-environment. GLNA differential expression had been used to monitor regression of fibrosis in hepatic injury [63]. The other AngII-regulated proteins that we examined in lung tissue and monitored in BAL have also been involved in fibrosis. BST1 gene knockout protects against fibrosis in a mouse model of renal ischaemia reperfusion injury [64]. RHOB belongs to a family of GTPases upregulated by TGF-β and is involved in actin cytoskeleton remodelling in fibroblasts [65, 66]. LYPA1 protein is involved in nitric oxide metabolism. MIR29B1 gene mutation resulted in significantly increased fibrosis and LYPA1 protein expression and lower nitric oxide in the renal outer medulla in rats [27]. These AngII-regulated proteins thus participate in fibrosis, although it remains unclear if they are controlled by a common regulator, downstream of AGTR1.
For the measurement of the AngII-regulated proteins in BAL samples, we developed PRM assays that selectively monitor peptides unique to our proteins. PRM assays are highly selective, are cost-effective due to their multiplexing capabilities [67, 68] and are suggested for proteins for which no high-quality antibodies have been generated [69]. The limitation of PRM is poor sensitivity for very low abundance proteins in a complex matrix. Our PRM assays showed good performance for quantification of four out of six AngII-regulated proteins, i.e. BST1, GLNA, LYPA1 and RHOB, but we could not quantify TSP1 or LAMB2. This could be explained by the restricted sensitivity of PRM assays for peptide detection in BAL (e.g. TSP1) or peptide absence in BAL (e.g. LAMB2). Given the usual infrastructure and expertise in clinical laboratories, our targeted proteomics assays could be implemented for clinical monitoring. In fact, similar targeted mass spectrometry-based methodology is routinely used to quantify proteins, metabolites and drug levels in the clinic [70].
We quantified the respective AngII-regulated peptides in BAL of patients with stable lung function, ALAD and CLAD. Interestingly, all AngII-regulated peptides correlated strongly with the presence of neutrophils in the BAL fluid and the majority of them correlated with CMV viraemia. Our proteins thus hold promise for identification of patients with lung allograft inflammation. Individual AngII-regulated peptide concentrations could not discriminate between CLAD and non-CLAD patients, at the time of bronchoscopy. In contrast, a classifier based on the combined peptide concentrations exhibited significantly improved diagnostic accuracy. This is interesting and potentially biologically important, as it suggests coordinated secretion of these peptides is relevant and supports the value of combining markers.
Even more interestingly, patients who developed CLAD within 1-year follow-up demonstrated higher concentrations of BST1, GLNA and RHOB proteins compared with patients who remained stable. LYPA1 protein also showed a trend towards higher concentration in the CLAD group. While individual peptide concentrations predicted development of CLAD with relatively good accuracy, a classifier based on combined peptide concentrations enabled prediction of CLAD status with an almost perfect accuracy of 97%. Our data clearly show that the classifiers based on combined peptide concentrations yielded improved diagnostic accuracy and predictive value compared with individual proteins. Numerous studies have harnessed artificial intelligence and machine learning algorithms for improvement of diagnosis, proactive disease monitoring and personalised medicine [71–73]. Nevertheless, reassessment of these classifiers in an independent validation cohort is needed and is under way.
Our study has several strengths. This highly translational approach furthered insights into the expression of the renin–angiotensin system in CLAD and enabled quantification of AngII-regulated proteins in BAL, which may represent promising biomarkers. We took advantage of a unique patient cohort of lung transplant recipients with archived tissue and BAL samples. We used a multimodal approach as we tested our hypothesis in tissue and in BAL, and employed different (including cutting-edge) techniques to confirm our observations (immunostaining, RT-qPCR, Western blotting and mass spectrometry). As the direct activity of the renin–angiotensin system is difficult to assess due to the short half-lives of angiotensin peptides, we combined the measurement of AngII receptors with AngII-regulated proteins to gain insight into renin–angiotensin system activity. Finally, using machine learning algorithms, we demonstrate that the combination of AngII-regulated peptides can diagnose CLAD and predict later development of CLAD with high accuracy.
Notwithstanding its strengths, our study has some limitations. In this cross-sectional study, causality cannot be determined. Future analyses of longitudinal BAL samples will determine the stability and reproducibility of AngII-regulated peptides. The selection of the control group is challenging in studies involving lung tissue. For instance, we cannot formally exclude the influence of smoking, neoplasia (donor lungs) or ischaemia (lobectomy samples) on the expression of renin–angiotensin system proteins and AngII-regulated peptides. Finally, in the future, we will validate these observations in a larger cohort of lung transplant recipients and study mechanisms via in vitro and mouse orthotopic lung transplantation models.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-02975-2020.Supplement
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-02975-2020.Shareable
Acknowledgements
The authors would like to thank the Toronto Lung Transplant Program Biobank team (Toronto, ON, Canada) for helping with sample collection and retrieval as well as Paul Chartrand, manager of the Latner Thoracic Surgery Research Laboratories (Toronto, ON, Canada), for facilitating experiment logistics. The authors also appreciate the tremendous dedication and hard work of all students, research fellows and technicians who helped with lung tissue collection and processing, often at odd hours of the night: Matthew White, Sajad Moshkelgosha, Betty Joe, Gafoor Puthiyaveetil, Akihiro Takahagi, Mitsuaki Kawashima, Tatsuaki Watanabe, Christina Lam, Goodness Madu, Nadia Sachewsky, Daniel Vosoughi, Madalina Maxim, Kristen Bounstra, Yohei Taniguchi and Alexander Tigert. Special thanks to Stephen Juvet (University Health Network, Toronto, ON, Canada) for his input regarding the general design of the study, immunostaining and RT-qPCR experiments. Special thanks to Davor Brinc (University Health Network, Toronto, ON, Canada) for consultation on PRM troubleshooting.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Author contributions: G. Berra: clinical screening, cohort design, sample collection and organisation, immunostaining, qPCR, Western blotting, data analysis, writing of manuscript. S. Farkona: experimental design, mass spectrometry, Western blotting, data analysis, writing of manuscript. Z. Mohammed-Ali: mass spectrometry. M. Kotlyar: bioinformatic analyses, machine learning algorithms, help with manuscript writing. L. Levy: cohort design, collection and handling of samples, help with manuscript writing. S. Clotet-Freixas: Western blotting. P. Ly: immunostaining. B. Renaud-Picard: clinical screening, collection of samples, histological grading. G. Zehong: help with immunostaining protocols. T. Daigneault: collection of samples. A. Duong: collection and organisation of samples. I. Batruch: technical assistance with mass spectrometry. I. Jurisica: protein interactome generation, bioinformatic analyses supervision. A. Konvalinka and T. Martinu: conception of the research question, cohort design, experimental design, supervision of experiments, data analysis, manuscript writing and editing.
Conflict of interest: G. Berra has nothing to disclose.
Conflict of interest: S. Farkona has nothing to disclose.
Conflict of interest: Z. Mohammed-Ali has nothing to disclose.
Conflict of interest: M. Kotlyar has nothing to disclose.
Conflict of interest: L. Levy has nothing to disclose.
Conflict of interest: S. Clotet-Freixas has nothing to disclose.
Conflict of interest: P. Ly has nothing to disclose.
Conflict of interest: B. Renaud-Picard has nothing to disclose.
Conflict of interest: G. Zehong has nothing to disclose.
Conflict of interest: T. Daigneault has nothing to disclose.
Conflict of interest: A. Duong has nothing to disclose.
Conflict of interest: I. Batruch has nothing to disclose.
Conflict of interest: I. Jurisica reports grants and nonfinancial support (in-kind contribution to grants) from IBM, personal fees for lectures and other (reimbursement of expenses) from Novartis and Canadian Rheumatology Association, outside the submitted work.
Conflict of interest: A. Konvalinka has nothing to disclose.
Conflict of interest: T. Martinu reports grants from Canadian Donation and Transplantation Research Program and Di Pochi funds, during the conduct of the study; grants from Sanofi, nonfinancial support from APCBio, outside the submitted work.
Support statement: The authors disclose receipt of the following financial support for this research: Canadian National Transplant Research Program, Kidney Foundation of Canada Predictive Biomarker Grant, Canadian Institute for Health Research (CIHR), Kidney Research Scientist Core Education, National Training (KRESCENT) program, Di Pochi funds, Toronto General and Western Hospital Foundation, University Health Network Multi Organ Transplant Program, and Les Hôpitaux Universitaires de Genève. Computational analyses were supported in part by grants from the Ontario Research Fund (34876) and Canada Foundation for Innovation (29272, 225404 and 33536). Funding information for this article has been deposited with the Crossref Funder Registry.
- Received July 30, 2020.
- Accepted March 6, 2021.
- Copyright ©The authors 2021. For reproduction rights and permissions contact permissions{at}ersnet.org