Potential novel biomarkers for chronic lung allograft dysfunction and azithromycin responsive allograft dysfunction

Chronic Lung Allograft Dysfunction (CLAD), manifesting as Bronchiolitis Obliterans Syndrome (BOS) or Restrictive Allograft Syndrome (RAS), is the main reason for adverse long-term outcome after Lung Transplantation (LTX). Until now, no specific biomarkers exist to differentiate between CLAD phenotypes. Therefore, we sought to find suitable cytokines to distinguish between BOS, RAS and Azithromycin Responsive Allograft Dysfunction (ARAD); and reveal potential similarities or differences to end-stage fibrotic diseases. We observed significantly increased Lipocalin-2 serum concentrations in RAS compared to BOS patients. In addition, in RAS patients immunohistochemistry revealed Lipocalin-2 expression in bronchial epithelium and alveolar walls. Patients with ARAD showed significantly lower Activin-A serum concentrations compared to Stable-LTX and BOS patients. Further, increased serum concentrations of Lipocalin-2 and Activin-A were predictors of worse freedom-from-CLAD in Stable-LTX patients. These biomarkers serve as promising serum biomarkers for CLAD prediction and seem suitable for implementation in clinical practice.


Decreased Activin-A in ARAD compared to Stable-LTX and BOS. Activin-A serum concentrations
were significantly decreased in ARAD patients compared to CLAD (p = 0.004), BOS (p = 0.004) and Stable-LTX (p = 0.002) patients. There was no difference among patients with RAS and ARAD (p = 0.305), RAS and BOS (p = 0.410), RAS and Stable-LTX (p = 0.166) or Stable-LTX and BOS patients (p = 0.601), respectively. Activin-A/Follistatin ratio was significantly decreased in ARAD compared to CLAD (p = 0.017) and BOS (p = 0.017) patients, respectively. Activin-A serum concentrations of all patient groups: LTX patients (RAS, BOS, ARAD, stable-LTX) as well as patients with end stage pulmonary fibrosis (CF and IPF) were depicted in Fig. 1D. Performing multiple testing between BOS, RAS, ARAD and stable LTX revealed significant differences in Activin-A serum concentrations (p = 0.008; Supplementary Fig. 1D).

Low serum concentrations of Activin-A in ARAD patients. Eighty-three percent of patients with
suspected BOS (41 out of 49) and 50% (6 out of 12) with suspected RAS were treated with azithromycin. While a treatment response to azithromycin was evident in 19 (46%) patients with suspected BOS, only one (16%) patient with suspected RAS was responsive to therapy. Patients with ARAD displayed reduced Activin-A serum concentrations (p = 0.001) and Activin-A/Follistatin ratio (p = 0.011) compared to patients not responding to azithromycin therapy. Activin-A serum concentrations (cut-off = 350 ng/ml) predicted responsiveness to azithromycin therapy (sensitivity = 0.772, specificity = 0.500, positive predictive value = 0.531 and negative predictive value = 0.750).
Increased serum cytokine expression in end-stage pulmonary fibrotic diseases. IPF patients showed significantly elevated serum concentrations of MMP-9 (p < 0.001) and TIMP-1 (p = 0.002) compared to healthy volunteers, respectively ( Fig. 2A,B). Similar to IPF, we detected higher serum concentrations of MMP-9

Discussion
While milestones in the perioperative management of LTX were achieved, CLAD, manifesting as BOS or RAS remains the main reason for adverse long-term outcome. The underlying pathomechanisms of the CLAD phenotypes are barely understood and an accurate diagnosis is often difficult to obtain. Biomarkers aiding in diagnosis and identifying pathophysiological pathways might spark new developments in CLAD management.
Our data identified Lipocalin-2 as a potentially useful biomarker to distinguish RAS from BOS and Stable-LTX. Lipocalin-2 is constitutively expressed and stored in neutrophilic granules making it a possible biomarker for neutrophil activity 17 . However, previous studies did not show differences in neutrophil counts in lungs of RAS compared to BOS patients 18 . Thus, whether elevated levels of circulating Lipocalin-2 in RAS patients  Table 2. Semi-quantitative analysis of pulmonary Lipocalin-2 expression. Immunohistochemistry for Lipocalin-2 was performed on lung specimens of patients with CLAD (RAS and BOS), end stage pulmonary fibrosis (IPF and CF) and healthy controls. Antibody reactivity in BO lesions of BOS patients, bronchial epithelium, alveolar walls and lung parenchyma was assessed. Y positive antibody staining, N negative antibody staining, BO bronchiolitis obliterans, BOS bronchiolitis obliterans syndrome, RAS restrictive allograft syndrome, IPF idiopathic pulmonary fibrosis CF cystic fibrosis. In the case of increased serum Lipocalin-2 in end-stage CF, neutrophils are a plausible source. The airways of adult CF patients are chronically besieged by neutrophilic inflammation and circulating Lipocalin-2 was reported to be elevated in acute exacerbations and Pseudomonas aeruginosa infections of CF patients 19 . It is strongly and selectively induced in bronchial and alveolar cells of inflamed lungs by Interleukin-1β 20 . We found Lipocalin-2 expression in epithelial cells of 90% of all CF patients. In line with our findings in IPF are correlations of Lipocalin-2 BAL levels with BAL neutrophilia and forced vital capacity, as well as no elevation of plasma Lipocalin-2. Further, Lipocalin-2 was expressed in airway epithelial cells that covered the honeycomb cysts and in bronchioles of IPF patients 21 .
Lipocalin-2 is known to build complexes with MMP-9, thereby stabilizing MMP-9 activity 22 . Airway remodeling is promoted by active secretion of MMP-9 by bronchial epithelial cells triggered by T-cells 23 In accordance to these findings, several studies linked elevated MMP-9 concentrations and MMP-9/TIMP-1 ratios in serum and BAL to CLAD, IPF and CF 24,25 . We detected higher MMP-9 serum concentrations (but not MMP-9/TIMP-1 ratio) in BOS patients compared to Stable-LTX patients and lower MMP-9/TIMP-1 ratios in CLAD compared to CF and IPF patients, suggesting a different importance of protease-antiprotease imbalance in pathogenesis. Divergent study results concerning cytokine profiles may pertain to different immunosuppressive therapy regimens. In previous studies, TIMP-1 was shown to be up-regulated in human pulmonary fibrosis 26 . Our study supported these results since TIMP-1 serum concentrations were higher in end-stage CF and IPF compared to healthy volunteers, but not compared to Stable-LTX or CLAD patients.
In a study on patients after LTX Activin-A serum concentrations decreased from postoperative week two to week twelve, while Follistatin serum concentrations remained unchanged until they increased 24 weeks post-LTX. Patients with primary graft dysfunction expressed lower serum Follistatin and higher Activin-A/follistatin ratios 27 . Our study was neither performed in the perioperative setting nor during a dramatic event such as primary graft dysfunction. We discovered Activin-A as a promising biomarker to distinguish ARAD patients from Stable-LTX and BOS patients, however not from RAS. Activin-A served as a predictive molecule to measure the responsiveness to azithromycin therapy. Low Activin-A concentrations in ARAD patients are in line with previous studies showing that Activin-A promotes inflammation at lower concentrations mediated by resting monocytes/macrophages, but inhibits inflammatory activity as concentrations increase 16 . Azithromycin was shown to exhibit inhibitory effects on neutrophils, such as diminishing neutrophil extracellular traps release 28 . Activin-A serum concentrations were not elevated in IPF and CF patients. Interestingly, Activin-A serum concentrations were reported to correlate inversely with lung function and nutritional status in CF patients 29 . In a CF mouse model with CF-like lung disease, Follistatin administration reduced Activin-A levels, mucus hypersecretion, and airway neutrophilia. ln CF patients, decreased serum concentrations of Follistatin and elevated Activin-A resulted in an increased Activin-A/Follistatin ratio 29 . While we detected decreased serum concentrations of Follistatin, we did not observe elevated Activin-A concentrations and Activin-A/Follistatin ratios in end-stage CF patients. These different results might stem from different CF disease severity: the aforementioned study sampled blood from stable CF patients during outpatient visits, while our study investigated patients with end-stage pulmonary   www.nature.com/scientificreports/ CF immediately before LTX. In our study, neither Activin-A serum concentrations nor Activin-A/Follistatin ratios were elevated in end-stage IPF patients. Our study has several limitations due to the retrospective analysis of prospectively collected blood samples, the single center experience and the limited sample size. Large-scale studies are warranted for further biomarker evaluation in CLAD, ARAD, CF and IPF patients. We are not suggesting that our absolute serum concentration values can be used to make any judgments about patient's diagnosis. While comparative results during one experiment could be repeated in separate ELISA tests. The commercially available ELISA tests were only sold for research use and not for diagnostic purposes. Since cut-offs were chosen via Youden Index out of our study results, sensitivity and specificity must be interpreted with caution. Early recognition of different CLAD phenotypes is essential for developing new therapies and improving outcomes. This study identified promising new biomarkers for better characterization and classification of CLAD phenotypes in addition to well-established diagnostic methods including spirometry and radiological imaging. Lipocalin-2 might serve as a potential diagnostic or even therapeutic target for RAS patients. Activin-A seems helpful to identify ARAD patients. Further prospective studies will be required confirming our biomarker results. The statistically significant difference in Lipocalin-2 serum concentrations between BOS and RAS deserves further investigations; as does Activin-A in ARAD patients by employing further methods including immunostaining.

Materials and methods
This single-centre explorative cohort study was performed at the institutional Division of Thoracic Surgery. All methods were carried out in accordance with relevant guidelines and regulations in the manuscript. Ethical approval was obtained from the institutional Ethics Committee of the medical university of vienna (EC-No:846/2010). All patients provided written informed consent prior to participation in the study. A graphical depiction of the study design is shown in Fig. 5.
Definition of CLAD phenotypes and ARAD. CLAD was defined as persistent pulmonary function decline (FEV 1 ≥ 20%) calculated from baseline (the best two FEV 1 measurements after LTX performed more than three weeks apart) after exclusion of non-CLAD causes by pulmonary function testing, bronchoscopy including transbronchial/endobronchial biopsies and BAL, blood work and radiological imaging. Patients suffering from suspected CLAD were treated with azithromycin 250-500 mg three times per week (irrespective of BAL neutrophilia) for 3 months. Patients with an FEV 1 improvement greater than 10% were classified as ARAD and not included in the analysis of CLAD. CLAD was further subdivided into RAS (concomitant drop in Total Lung Capacitiy (TLC) ≥ 10% compared to baseline levels and persistent opacities on chest imaging) and BOS (no restrictive pulmonary function pattern) 6 . Patient cohort. Demographic and clinical data were depicted in Table 3. For serum analysis, we enrolled 119 patients undergoing LTX. Among them, 41 patients were diagnosed with CLAD and dichotomized into BOS (n = 30) and RAS (n = 11) phenotypes. The included BOS patients displayed the following disease severity: BOS1 (n = 5), BOS2 (n = 13) and BOS3 (n = 12). Ten out of 11 RAS patients had a mixed phenotype.
Patients with ARAD (n = 22) were separately analysed and not included in the CLAD cohort. Sixty-three LTX patients with stable pulmonary function were included in this study (Stable-LTX). Blood samples of patients with stable-LTX were drawn at clinical routine check-up including surveillance bronchoscopy. Furthermore we enrolled patients with end-stage IPF (n = 31) and CF (n = 15) listed for LTX. Serum samples of 63 healthy volunteers served as a control group. None of the healthy volunteers suffered from acute or chronic diseases, such as asthma, chronic obstructive pulmonary disease, acute or chronic kidney diseases, autoimmune diseases or any cancer type. Furthermore, none of the healthy volunteers took medical drugs for at least three months before the blood draw.
Blood samples were aliquoted after centrifugation with 2851×g for 15 min at 4 °C and stored in 2 ml cryotubes at − 80 °C in until further analysis.
CLAD, Stable-LTX, ARAD and control groups were matched by gender and age. There were no statistically significant gender differences in any of the study groups (CLAD, comprising BOS and RAS, ARAD, stable-LTX and healthy volunteers). Mean age of the study groups ranged from 49.8 years in BOS to 51.7 years in IPF patients. There were no statistically significant age differences between ARAD, BOS, RAS, Stable-LTX and IPF patients. However, CF patients were significantly younger (mean age of 31.0 years) compared to patients with ARAD, BOS, RAS, stable LTX and IPF. Standard immunosuppressive treatments. Fifty patients (42.0%) received induction therapy with either alemtuzumab (n = 34, 68.0%), anti-thymocyte globulin (n = 15, 30.0%) or daclizumab (n = 1, 2.0%). Maintenance therapy was performed in all patients with tacrolimus, mycophenolat-mofetil (only in the nonalemtuzumab group) and corticosteroids. All patients undergoing LTX received postoperative anti-infectious prophylaxis therapy with piperazillin/tazobactam, a lifelong pneumocystis prophylaxis with trimethoprimsulfamethoxazole, prophylactic inhalation therapy with amphotericin B and gentamicin; and cytomegalovirus (CMV) prophylaxis including CMV hyperimmunoglobulines, together with valganciclovir.
Immunohistochemistry. Immunohistochemistry was performed with specimens of 50 patients allocated to the following subgroups: 20 patients who underwent re-transplantation for either BOS (n = 11) or RAS (n = 9), 20 patients who underwent primary LTX for either IPF (n = 10) or CF (n = 10) and 10 patients who served as healthy controls. Patients of the latter group underwent lobectomy for peripherally located lung cancer (pT1pN0M0). The tumour was located at least five centimetres from the investigated lung area and patients showed no evidence for advanced parenchymal lung diseases. Immunohistochemical staining was performed on formaldehyde-fixed, paraffin-embedded tissue specimens according to routine protocols 30 . Anti-Lipocalin-2/NGAL antibody ab115324 (Abcam, Cambridge, UK) and anti-goat IgG secondary antibodies (Vector Laboratories, Burlingame, CA, USA) were used for staining. Omission of the primary antibody served for negative control staining. Results were assessed by two independent investigators, including one specialist on pulmonary histopathology (O.F.). Antibody staining of alveolar lining cells, bronchial epithelium and the interstitium were evaluated on each slide. In case of interobserver discrepancies, results were discussed until consensus was achieved. BOS lesions were identified by Elastica van Gieson staining on adjacent slides. Statistical methods. Statistical data analysis was performed using SPSS software (Version 20; IBM SPSS Inc., Chicago, IL, USA). Graphical methods (histograms) were employed to test normality. Data were reported as median (range) for non-normal distributions. Mann-Whitney U test was used to compare two independent groups. Kruskal-Wallis rank test was used to perform multiple testing followed by Dunn's multiple comparison tests. P-values < 0.05 were considered statistically significant. For follow-up analysis, cytokine serum concentrations were dichotomized into low and high subgroups by using ROC-Curves for calculating Youden-Indices. We used the best cut-offs for maximizing sensitivity at the expense of specificity in order to create a test that is maximally sensitive. Accordingly, we performed Kaplan-Meier survival analysis and used log-rank analysis for outcome assessment. For the analysis of the prognostic value of serum cytokines, the date of serum withdrawal was taken as the starting point for follow-up analysis. We analysed Stable-LTX patients for freedom from CLAD and OS. In order to express the precision and repeatability of our ELISA experiments we calculated the intraassay and inter-assay CV. Therefore, we performed following calculation % CV = SD of plate means ÷ mean of plate means × 100. GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA) was used for data visualization. Boxplots were designed as follows: box: 1st to 3rd quartile, bar: median, whiskers: percentile 5-95, outliers: all shown as dots.