Quantitative Proteomics Identies Cannabinoid Receptor 2 as Candidate Marker of Idiopathic Pulmonary Fibrosis and Dysregulation of Endogenous Cannabinoids Pathway

Background: Idiopathic pulmonary brosis (IPF) a chronic, progressive, and lung brosis disease of unknown etiology with less effective treatment. It is important to discover new biomarker and therapeutic target for the diagnosis and cure of IPF. Method: Differential metabolic proles may be useful for the diagnosis of IPF and provide additional insight into the molecular mechanisms underlying IPF. Plasma samples from IPF, COPD and normal controls were investigated using liquid chromatography-quadrupole time-of-ight mass spectrometry (UHPLC/Q-TOF-MS) and these datasets were analyzed using multiple pattern performance methods. Multivariate statistical methods, pathway enrichment analysis and univariate receiver operating characteristic (ROC) curve analysis were performed. Results: OPLS-DA results showed that it exhibited signicant separation between any two groups. ROC curve analyses revealed that 8 metabolites with high AUC above 0.7 between three groups in the plasma samples. Pathway analysis revealed that 3 metabolites are involved in endogenous cannabinoids/cannabinoid receptor 2 (CB2) signaling. Moreover, we found that the specic elevation of CB2 could be a signature of PF(Pulmonary Fibrosis) lung tissues in rat model. Conclusions: LC-MS-based plasma metabolomics provides a important tool to identify the potential biomarkers for IPF. Taken together, we performed quantitative chemoproteomic proling and identied endogenous cannabinoids/CB2 as the key target of ICA in bleomycin-induced pulmonary brosis. study


Background
Idiopathic pulmonary brosis (IPF) is a progressive brotic lung disease characterized by excessive proliferation of broblasts and extracellular matrix accumulation that leads to the distortion of the lung parenchyma and the loss of lung function [1,2]. With a median survival of 3-5 years following diagnosis, IPF Patients usually present unspeci c symptoms, such as cough and dyspnea on exertion. The diagnosis of IPF is based on typical features of high-resolution computed tomography (HRCT) or lung biopsy excluding other idiopathic interstitial lung diseases (ILD) in a multidisciplinary setting [3,4]. The lung biopsy is invasive and has several risk procedures in those patients with an uncertain diagnosis and those thought to have IPF [5,6]. Moreover, IPF is a chronic and fetal interstitial lung disease with poor prognosis and limited treatment options [3,7,8]. Pirfenidone and Nintedanib are two anti-brotic therapies available for the treatment of IPF and have been approved since 2014 [9,10]. Hence, exploring the molecular mechanisms involved in the development of pulmonary brosis may lead to novel therapeutic strategy.

Page 4/20
Metabolomics is the study of inputs and outputs of biological pathways and often considered important representative of the function state of a cell or organ. Currently, an untargeted metabolomics analysis can detect many metabolites at the same time. To identify novel pathway or diagnosis biomarkers indicative of metabolic alterations, metabolomics is a cumulatively used experimental method. In this study, we try to identify the potential plasma metabolic biomarker used for diagnosing IPF.

Human subjects
All human plasma samples of patients with non-acute exacerbation IPF (n = 10), stable chronic obstructive pulmonary disease (COPD) (n = 10), and normal control (NC) subjects (n = 10) were obtained from Huashan Hospital, Fudan University and stored by -80℃ refrigerator until analysis. IPF was diagnosed according to the international guideline 2011 and 2018 [3,4]. COPD was diagnosed according to the the Global Initiative for Chronic Obstructive Lung disease [11]. The normal subjects had no appearance or history of respiratory diseases and malignant tumor. Data from the clinical history and physical examination were analyzed, as well as the results of the lung function, blood gas determination (Table 1). The study protocol was approved by the Ethics Committee of Huashan Hospital, Fudan University (Shanghai, China). Raw Data analysis, metabolite identi cation and pathway analysis The raw data les were converted into the mzXML format, and then using the XCMS package to perform peak alignment, retention time correction and extraction peak area [12]. Prior to multivariate statistical analysis, metabolite data were subjected to pareto-scaling. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were performed for pattern recognition analysis. In the OPLS-DA model, the variable importance in projection (VIP) score were calculated. The student's t-test was selected for pairwise comparison. The signi cantly changed features were screened with VIP 1, and P value 0.05 was considered to be statistically signi cant. Signi cantly changed features and pathway analysis were compared with the built-in database by searching KEGG (http://www.genome.jp/kegg/) [13].

Reagents and antibodies
Acetonitrile and methanol were acquired from Merck (Frankfurt, Germany

Animals and treatment
All animal experimental protocols were approved by the Animal Care and Use Committee of Fudan University (Shanghai, China). Male SD rat with mean weight 200-220g were obtained from Shanghai SLAC Laboratory Animal Centre (Shanghai, China), and then adaptive feeding for three days. 20 rats were randomly allocated to control group and BLM group(n = 10 per group). Rat pulmonary brosis model was induced by BLM. BLM was administered once by intra -tracheal instillation at 3 U/kg in 0.9% saline. At the end of the study, the rats were sacri ced by pentobarbitone overdose, and then lung tissues were collected. The construct protocol of Chronic obstructive pulmonary disease (COPD) rat modle was described in the our previous article [14].

In vivo micro-CT analysis
In vivo micro-CT examinations of rat lungs were carried out at the end of the study. CT images were acquired in high resolution mode on the trimodal Quantum GX micro-CT scanner (PerkinElmer, Waltham,

Characteristics of subjects and plasma metabolic pro ling by UPLC-Q-TOF-MS
Characteristics of the study population (patients with IPF, COPD and normal control) were presented in Table 1. To search the metabolic differences among IPF, COPD and normal control, we applied nontargeted LC-MS focusing on the metabolic pro les of the plasma samples and the related speci c metabolism pathways underlying IPF development. In order to identify the real biological differences, it is necessary to perform effective qualitative control (QC). The overlapping spectra was conducted for the UPLC-Q-TOF-MS total ion chromtograms of QC samples (Fig. 1A). Scatter plots and regression lines demonstrated correlation of ion current intensity of QC samples (Fig. 1B). All in all, these data showed that the retention time and ion current intensity of chromatographic peaks overlapped, indicating that the differences found in this study were biological differences rather than technical differences caused by measurement errors. Moreover, PCA analysis of the metabolic pro les measured by UPLC-Q-TOF-MS indicated differences between IPF patients, COPD patients and healthy control group (Fig. 1C). Hotelling's T2 analysis was employed to determine whether an outlier sample exists. It showed that most samples were within the 99% con dence interval (Fig. 1D). Taken together, results show that the graphical integration of outputs from multiple quantitative tools facilitates the evaluation of instrumental effects and judges the necessity for further data analysis.
Metabolomics analysis of plasma samples from normal control, IPF and COPD subjects.
PCA and OPLS-DA were performed to determine whether it was possible to distinguish NC, IPF and COPD subjects on the basis of the metabolic data. Obviously, PCA analysis showed that there was no signi cant separation among these three groups ( Figure. 2A). However, OPLS-DA results showed that it exhibited signi cant separation between any two groups and then the replacement inspection established a 200-fold OPLS-DA model by randomly changing the order of classi cation variable Y to obtain the R2 and Q2 values of a random model (Fig. 2B: left panel). All of these blue Q2 points, which were shown in left to right, were lower than the original Q2 points in positive ion mode (Fig. 2B: right panel). The same results were obtained in negative ion mode (Fig. 2C). All in all, these results suggested that the model was reliable without tting.
Selection of potential metabolite markers for IPF Box plots and classical univariate ROC curve analyses were used to further characterize the predictive value of these individual metabolites independently. The box plots showed the relative changes in the eight potential metabolites D-Proline, Deoxycholic acid, EDTA, Erucamide, Glycochenodeoxycholate, Myristic acid, Taurochenodeoxycholate and 11(Z),14(Z)-Eicosadienoic Acid between groups (Fig. 3A). Furthermore, ROC curves also showed a apparent discrimination of these eight metabolites between Groups with an area under the curve (AUC) > 0.7, as shown in Fig. 3B and Fig. 3C. The results of ROC curve analyses suggested that 8 metabolites with high AUC above 0.7 between groups in the plasma samples.
Screening and identi cation of differential metabolites The metabolite features were selected as VIP > 1.0 and P value < 0.1. As shown in Volcano plot Fig. 4A, the differential metabolities signi cantly up or down-regulated among NC, IPF and COPD (FC > 1.5 and P value < 0.05). The heatmap and hierarchical cluster analysis showed the distribution of signi cant differential metabolites between groups in the IPF, COPD and NC ( Figure. 4B). Box plots showed representative metabolite changes among NC, IPF and COPD groups (Fig. 4C). Compared with the IPF and COPD group, the NC group had higher levels of 3-Indoleacetonitrile (* p < 0.05; **p < 0.01). Glycochenodeoxycholate level was higher in IPF samples. However, Vanillin was signi cantly lower in NC samples. In order to understand the relationship of metabolite differences with IPF, correlation analysis was used to analyze the differential metabolites (Fig. 4D). These data showed that the individual discriminating metabolite may be useful to differentiated the IPF from NC and COPD patients, and the accuracy can be further up-regulated if more pre-selected discriminating metabolites would be used.

Pathway enrichment analysis and metabolite signatures re ecting aberrant metabolism in IPF
The KEGG pathway enrich analysis of the differential metabolites between the IPF, COPD, and NC groups were identi ed. The top 20 most enriched pathway terms are shown in the KEGG enrichment bubble diagrams (Figure. 5A). The main differential pathways in IPF compared to other groups indicated that the differential metabolites were mainly involved in the following pathways: Central carbon metabolism in cancer, Protein digestion and absorption, Aminoacyl-tRNA biosynthesis, Retrograde endocannabinoid signaling, and Linoleic acid metabolism, et al. The key differential metabolites between IPF and NC were identi ed to be involved in the pathways for Retrograde endocannabinoid signaling (Fig. 5B). Moreover, Box plots showed an apparent discrimination of the three metabolites associated with retrograde endocannabinoid signaling among the three groups (Fig. 5C).
Previous studies have shown that human lung-resident macrophages expressed a complete endocannabinoid system, the cannabinoid receptors was also expressed in lung cancer-associated macrophages and cannabinoid receptor activation selectively inhibited the release of angiogenic and chronic in ammation [15]. Recent studies shown that CB1 expression by ATII cells and alveolar macrophages was indicated by their colocalization with surfactant protein C (SP-C) and CD68 respectively[16]. However, little is known about cell-type-speci c localization and the function of CB2 in BLM-induced lung brosis. To further examine the relationship between the endocannabinoid signaling and pulmonary brosis, we detected the CB2 receptor protein expression in the lung tissues of rat models (BLM-induced pulmonary brosis, cigarette smoke-induced COPD, and normal control) corresponding to the three groups of clinical subjects. Immunohistochemistry results show that the protein expression of CB2 receptor in the lung tissues of BLM-induced rat pulmonary brosis model was signi cantly upregulated, whereas there was no signi cant change in the lung tissues of cigarette smoke-induced rat COPD model and normal control rats. Positive immunostaining for CB2 was observed in alveolar epithelial cells and myo broblasts in lung brogenesis (Fig. 5D). Collectively, these results demonstrated that endocannabinoid signaling may associated with pulmonary brosis.

Validation of dysregulated endocannabinoid signaling in rate model
The above results revealed that the endocannabinoid signaling may be associated with pulmonary brosis. To validate the ndings from the proteomics study, endocannabinoid/cannabinoid receptor system was selected for veri cation in the lung tissue samples by immunohistochemistry. Next, Highresolution micro-CT scan was carried out to assess lung morphologic changes. The reconstruction of micro-CT image analysis of BLM-induced pulmonary brosis rats showed remarkable lung structure destruction as compared to normal control rats, as illustrated in Fig. 6A. BLM instillation resulted in histological damage of lungs tissues in BLM-induced pulmonary brosis rats, characterized by consolidation of the parenchyma and the excessive deposition of collagen. α-SMA immunostaining of lung tissue was signi cantly increased in PF group, relative to Normal control (NC) group. In contrast, BLM-induced α-SMA expression was attenuated by the treatment of ICA (Fig. 6B).
In order to determine the role of CB2 receptor activation mediating the effect of BLM-induced lung brosis. As shown in Fig. 6C, we examined the CB1 and CB2 protein expression by using immunohistochemical staining. The expression of CB2 receptor protein was signi cantly up-regulated in the PF group after BLM instillation. Moreover, the expression of CB2 protein was predominantly in alveolar epithelial cells and myo broblasts. Taken together, the BLM-induced increase in CB2 levels was dramatically changed, and re ected the key role of CB2 blockade in treatment of IPF.

Discussion
The pathogenesis of IPF involved a series of pathological process, such as early in ammation, oxidative stress, myo broblasts proliferation, excessive wound healing and brosis. De cient understanding of the mechanisms underlying IPF has hampered the development of e cient tools for diagnosis and interventions [17]. For the discovery of new biomarkers, the application of metabolomics is emerging and signi cantly contributes to a deeper understanding of the metabolic pathways in IPF [18]. Moreover, previous studies have showed that dysregulated metabolism may be used to discover new biomarkers and targets [19]. In this work, a simple, repeatable and sensitive method was validated for the simultaneous determination of metabolites for diagnosis of IPF disease. Herein, a non-targeted UPLC-Q-TOF-MS plasma metabolism method was applied for exploring the metabolic characteristics to screen e cient predictors of IPF. The presented results demonstrated the metabolic pro ling of IPF patients differing from those of controls and COPD, which showed satisfying data quality. We compared plasma metabolic pro les of 30 subjects to identify its metabolic signatures. Previous studies on the metabolites of the three groups have been really rare.
Previous study reported that the advantage of plasma metabolic pro ling for characterizing metabolite signature of IPF and changes in metabolic pathways that might be helpful for understanding the metabolic mechanism of pulmonary brosis [20]. In the current study, PCA did not show a clearly different distribution in the positive/negative ion model. Then, we applied OPLS-DA statistical approach to select differential metabolites. 47 signi cantly changed metabolites related to IPF were found to be differentialy altered among the three groups in positive/negative ion model. Furthermore, recent many studies have been proved that establishing a diagnostic model to predict the IPF patients could potentially improve the diagnosis [21]. ROC curves of classi cation models base on the 8 metabolites with high AUC above 0.7 were plotted to distinguish IPF and normal control. Therefore, a combination of more than one discriminatory metabolite will be necessary to increase the diagnostic performance of IPF. To research their metabolism mechanisms, the enriched pathways of the metabolites were also analyzed according to the KEGG database, which re ected the key pathway related with differential metabolites. The most enriched pathway terms were biosynthesis of unsaturated fatty acids, caffeine metabolism, Arginine biosynthesis, linoleic acid metabolism, central carbon metabolism in cancer, protein digestion and absorption, retrograte endocannabinoid signaling and ABC transporters.
Endocannabinoid, interacts with CB1 and CB2 receptors of endocannabinoid system, help coordinate and regulate everything which we feel, think and do. The endocannabinoid/cannabinoid receptor system may be a rational therapeutic target in IPF and promote in ammatory by CB1 activation in many chronic in ammatory diseases [20]. In addition to promoting in ammatory, activation of CB1 also promote brosis progression in many organs, such as liver, kidney, heart and skin [22][23][24]. Simultaneously, it is reported that the brain-penetrant CB1 antagonist attenuates liver brosis in mice model [25]. However, the potential role of CB2 has not yet been investigated in patients with IPF or pulmonary brosis animal model. As shown in Fig. 5D, CB2 is predominantly expressed in ATII and myo broblast cells of lung. Rice W et al. reported that CB1 is always expressed in ATII cells, bronchial epithelial cells, and alveolar macrophages [26]. However, we found a higher expression of CB1 in alveolar macrophages than other cell populations following BLM treatment.
Although the FDA have approved pirfenidone and nintedanib for the cure of IPF patients, both compounds still have modest e cacy, but not lead to good overall survival.Therefore, the ideal application implies that simultaneously targeting alveolar in ammation and brotic process. Polypharmacology may offer a model for the way to drug discovery. As was demonstrated in Fig. 6, it was observed that the CB1 and CB2 were speci cally activated in rat PF model. Furthermore, Resat Cinar etal. reported that elevated activity of the endocannabinoid/CB1 system parallels disease progression in the patients with IPF and in mice with BLM-induced lung brosis, thus making it a feasible candidate for the treatment of IPF. Next, cannabinoid receptors antagonist or CB2 deletion will be employed to explore the detail mechanism of anti brosis and nd the downstream target of endocannabinoid/cannabinoid receptor system in the lung with PF. In summary, the results of this study indicate that the endocannabinoid/cannabinoid receptor system and related signal transduction molecules are important targets for the diagnosis and treatment of PF, and that the treatment of ICA protect against pulmonary brosis induced by BLM in rats for the rst time. Further research should be encouraged to clarify the mechanism of endocannabinoid/cannabinoid receptor system and evaluate the effectiveness of ICA in patients with IPF.

Conclusions
The present study demonstrates that metabolic pro ling of plasma by untargeted LC-MS analysis allows to discriminate among NC, IPF and COPD subjects. Therefore, the differential metabolites and enriched signaling pathways may involve in metabolisms of endocannabinoid/cannabinoid receptor system in pulmonary brosis. The outcome may be supporting CB2 as potential treatment target for PF.