Chronic obstructive pulmonary disease (COPD) is a chronic inflammatory disease of the airway. With high mortality and morbidity, it has caused a huge global burden and would be the third leading cause of death worldwide (Quaderi and Hurst, 2018). COPD is a heterogeneous disease with complicated pathogenic mechanism which has not been fully elucidated. In prior researches, whole genome sequencing, transcriptomics and proteomics have been widely utilized for biomarker screening and molecular mechanism research (Kan et al., 2017). A large number of genes and protein molecules have been determined to be associated with COPD (Stockley et al., 2019). Metabolites have also been implicated in various pathological processes in recent years (Johnson et al., 2016). Metabolites associated with diseases not only work as the promising biomarker for disease screening, but also provide an insight into the pathogenesis of disease (Arakaki et al., 2008). Therefore, researches identifying promising metabolite biomarkers to screen patients with COPD and to elucidate the pathogenic mechanisms associated with metabolites are warranted.

Metabolomic, the profiling of metabolites, is an emerging approach to identifying metabolite biomarkers and discovering the perturbed metabolic pathways that may help to elucidate the mechanisms governing COPD development (Johnson et al., 2016). Untargeted metabolomics can measure the widest array of metabolites in a clinical sample and has been implemented in recent years to uncover changes in metabolic pathways and identify novel biomarkers in many different diseases (Chen et al., 2019; Puchalska et al., 2019; Bunning et al., 2020; Liu et al., 2020; Thome et al., 2020). Metabolic disorders have also been identified in COPD patients. Previous studies have focused on the metabolism profiling of plasma (Paige et al., 2011), exhaled breath condensate (EBC) (Kilk et al., 2018), bronchoalveolar lavage fluid (BALF) (Yu et al., 2019), and induced sputum (Telenga et al., 2014) of patients or plasma and serum from the mouse model of emphysema or bronchitis caused by cigarette smoke (CS) (Cruickshank-Quinn et al., 2014; Ren et al., 2016). It is worth noting that samples in those studies could not reflect the metabolism of lung directly and the differential metabolites have not been validated in another independent set. Lung tissue from COPD patients is a more direct reflection of metabolism of the lung, however, has been rarely study. Though a recent study with a limited sample size reported the metabolomic of para-cancer tissue from lung cancer patients with COPD (Li et al., 2020), more studies with normal tissue away from lung nodules with a larger sample size are necessary.

Considering that lung tissue is a direct reflection of the metabolism of lung cells and has been rarely explored and the prevalence of COPD in men is more than twice that of women (Fang et al., 2018), we collected two independent sets of lung tissues from male COPD patients and controls to perform untargeted metabolomics analysis. We attempted to screen and further validate the differential metabolites associated with male COPD and disordered metabolic pathways. Furthermore, we also performed plasma metabolomic analysis to determine the overlapping metabolites between plasma and lung tissue; these metabolites would be more precise biomarkers to effectively discern COPD patients from controls.

In our study, we performed untargeted metabolomics in lung tissue of male COPD patients. We found that several classes of metabolites, including glycerophospholipids (GPs), amino acids and fatty acyls, were abnormally expressed. Pathway analysis suggested that many abnormal metabolic pathways were associated with male COPD. Among these pathways, glycerophospholipid metabolism and valine, leucine and isoleucine biosynthesis pathways were disordered in both the discovery and replication set. Besides, we also tested another independent plasma set to screen the overlapping biomarkers between the lung tissue and plasma. We subsequently screened out two overlapping metabolites containing phytosphingosine and l-tryptophan, which showed a high diagnostic performance for discriminating early male COPD patients from controls.

Metabolomic profiling is an emerging tool for detecting differential metabolites, disrupted pathways related to diseases and potential biomarkers. It has also been evaluated in COPD patients. However, most of those studies focused on the metabolome of plasma, exhaled breath condensate (EBC), bronchoalveolar lavage fluid (BALF) or induced sputum in COPD patients and did not validate their findings with a replication set (Kilk et al., 2018); (Paige et al., 2011); (Ran et al., 2019); (Telenga et al., 2014); (Yu et al., 2019). Lung cells can reflect the metabolic profiling of patients with COPD more directly, while the normal lung tissue from COPD patients has rarely been calculated to date. Sex can influence the metabolism (Darst et al., 2019). A nationwide prevalence study revealed that the prevalence of COPD in men is more than twice that of women (Fang et al., 2018). Recent studies also reported that the difference in metabolic profiles between male and female COPD patients (Gillenwater et al., 2021); (Naz et al., 2017). Therefore, lung tissue from male COPD patients in two independent set were collected to perform metabolomic analysis: a discovery set to detect and a replication set to validate. What’s more, we also analyzed metabolomic profiling of plasma to screen the overlapping metabolites biomarkers.

Glycerophospholipids (GPs), the most abundant phospholipids and one of the primary lipid types in cell membranes, consist of such lipids as phosphatidylethanolamine (PE), phosphatidylcholine (PC), phosphatidylglycerol (PG), phosphatidylserine (PS), phosphatidylinositol (PI), phosphatidic acid (PA), phosphatidylglycerophosphate (PGP) and CDP-diacylglycerol (CDP-DG) (Kimura et al., 2016). Predictably, these lipids play critical roles in lipid metabolism and health (van der Veen et al., 2017). The metabolism of GPs showed in Figure 6A. In recent studies, the perturbation of GPs has been observed in other diseases, such as osteoporosis (Mei et al., 2020; Bellissimo et al., 2021), type 2 diabetes (Lai et al., 2020; Schillemans et al., 2021), pleural effusion (Luo et al., 2020) or environmental microplastics (Zhao et al., 2021). In our study, three decreased PEs, nine increased PCs and eight elevated PEs were identified. How these abnormal GPs are involved in the pathological mechanism of COPD is currently unclear. A study reported that loss of PE induces mitochondrial dysfunction and oxidative stress (Heden et al., 2019). PE-deficient mice produced more H₂O₂, and mitochondrial PE deficiency is harmful to their metabolic and contractile functions, causing ventilatory failure and lethality. Another study showed that the increased PC levels may be caused by cigarette smoking and may participate in oxidative stress (Wang-Sattler et al., 2008; Li et al., 2016). In addition, it also reported that the destruction of the CDP-ethanolamine pathway for PE synthesis is related to cellular senescence (Wu et al., 2019) and the human T cell response could be triggered by PGs mediated by CD1b (Shahine et al., 2017). COPD is a chronic inflammatory disease associated with oxidative stress and cellular senescence (Birch et al., 2018). These evidences suggest that GPs may be related to COPD by participating in oxidative stress and inflammation, but this requires further verification.

In addition, both the discovery and validation sets showed perturbations in the class of amino acids by lung metabolomics analysis in our study. Meanwhile, disorders of the valine, leucine and isoleucine biosynthesis pathways have also been identified and validated (Figure 6B). Amino acids are essential nutrients and pivotal determinants of cell proliferation and stress responses (Shao et al., 2018). Not surprisingly, perturbations in amino acid metabolism have been identified in many diseases (Abooshahab et al., 2020; Gallagher et al., 2020; Xuan et al., 2020). Similar perturbations are observed in COPD. Plasma or serum amino acid levels are related to energy metabolism and inflammation in COPD, and their disturbances suggest that COPD is a wasting disease with protein degradation (Pouw et al., 1998; Engelen et al., 2007; Ubhi et al., 2012; Ruzsics et al., 2016). Through metabolome analysis of lung tissue, we first discovered the elevated branched chain amino acids (BCAAs) including leucine and isoleucine. BCAAs, as essential amino acids for mammals, are synthesized in bacteria, plants, and fungi but not in animals (Neinast et al., 2019; Bowerman et al., 2020). In lungs of COPD patients, abnormal bacterial diversity has been reported (Bowerman et al., 2020); (Pragman et al., 2012; Ren et al., 2018; Wang et al., 2019), which we speculated might partially interpreted the upregulated BCAAs in lungs. Notably, BCAAs also can be transported into the cells from extracellular medium by amino transporters and then be catabolized in a series of enzymatic reactions (Neinast et al., 2019); (Taylor, 2014). Anyhow, whether the disrupted bacterial diversity, the abnormal expression of amino transporters or key enzymatic associated with BCAAs catabolism contribute to the increased BCAAs needs further exploration.

Moreover, a large proportion of differential metabolites were fatty acyls. Most differential fatty acyls were acylcarnitines; among them, all acylcarnitines associated with l-carnitine were downregulated in the discovery set, and 5 of 6 acylcarnitines were decreased in the replication set. The carnitine shuttle, which refers to l-carnitine, acylcarnitines, CD36 and carnitine acyl transferase Ⅰ/Ⅱ, participates in the translocation of acyl-CoA from the cytoplasm to mitochondria and has an intimate relationship with β-oxidation. The imbalance of oxidation/antioxidation contributes to COPD occurrences (Rahman and Adcock, 2006). The decreased acylcarnitines may break the balance of β-oxidation, which enhances the development of COPD. As previous studies reported, the progress of the carnitine shuttle was disrupted, and a relatively low level of l-carnitine was identified in the lung, serum and BALF in mice exposed to CS (Agarwal et al., 2014; Conlon et al., 2016). Exogenous carnitine supplementation exhibited antioxidant ability (Vanella et al., 2017). In our study, we validated the abnormality of the carnitine shuttle in the lungs of COPD patients, which might play a role in the development of COPD by being involved in oxidative stress.

COPD is a heterogeneous disease for which few screening biomarkers using omics approaches are available². In our study, we finally screened two overlapping metabolites combining three independent sets as new promising biomarkers. Phytosphingosine alone displayed a strong ability to diagnose COPD with an AUC of 0.8931 (95% CI: 0.8224-0.9639), while l-tryptophan was moderately efficient in discriminating COPD with 0.69 AUC (95% CI: 0.5711-0.8064). The combined diagnostic efficiency of the two metabolites was more optimal (AUC = 0.911, 95% CI: 0.8460-0.9765). This satisfactory diagnostic potential may provide some guidance significance for clinically distinguishing COPD patients from controls in the future. Lung tissue is a direct, effective and accurate approach for evaluating the metabolic profiles of lung of COPD patient. While plasma, a comprehensive reflection of body metabolism, is a convenient, non-invasive but indirect way to explain the metabolic profiles of lung. In our study, phytosphingosine and l-tryptophan were two overlapping differential metabolites in plasma and lung tissue. Since most COPD cases in our study are relatively mild/moderate, we hold that phytosphingosine and l-tryptophan were two indicators for early screening of COPD patients.

Phytosphingosine (PHS) were increased in both plasma and lung tissue. PHS belongs to sphingolipids and exists in plants, yeast, and other mammalian tissues. In yeast and mammalian cells, it can be metabolized to odd-numbered fatty acids and incorporated into glycerophospholipids (Kondo et al., 2014). PHS has been reported to be associated with associated with apoptosis, migration, and inflammation. It enhanced apoptotic cell death in cancer cells through ROS-dependent and -independent AIF release or caspase 8 activation and bax translocation (Park et al., 2003; Park et al., 2005). Takahashi, M., et al. showed that phytosphingosine interacted with CD300b to promote neutrophil recruitment in the way of zymosan-induced, nitric oxide–dependent (Takahashi et al., 2019).Other two studies revealed that PHS and its derivatives ameliorated skin inflammation (Kim et al., 2006; Kim et al., 2014). Furthermore, PHS can activate the endoplasmic reticulum (ER) stress surveillance (ERSU) pathway (Breslow, 2013), which prevented inheritance of stressed ER (Babour et al., 2010; Piña and Niwa, 2015). Reasonably, we speculated PHS may be associated with COPD by affecting apoptosis, inflammation or ER stress. l-tryptophan (L-Trp) is an essential amino acid. Disruptions in l-tryptophan metabolism are reported in several neurological, metabolic, psychiatric, intestinal disorders and cancers, which is also considered as a pharmacological target in clinical practice (Platten et al., 2019; Modoux et al., 2021). The increased L-Trp in lung tissue and decreased L-Trp in plasma have been found in our study and the increased plasma L-Trp is consisting with previous studies (Wendt et al., 2016; Naz et al., 2019). L-Trp can be metabolized by indoleamine 2,3-dioxygenase (IDO), a rate-limiting enzyme in the kynurenine pathway (Cervenka et al., 2017). Interestingly, the reduced sputum IDO activity and expression while the increased serum IDO activity and expression have been reported (Maneechotesuwan et al., 2013; Naz et al., 2019), which may partially explain the divergent change of L-Trp between lung tissue and plasma. Nonetheless, more exploration and underlying mechanisms about the paradoxical changes of L-Trp and IDO in plasma and lung tissue are needed. All in all, PHS and L-Trp are not only the promising metabolite biomarkers to discern COPD from controls but also potential molecular that contributed to the pathogenesis of COPD.

There were several limitations in our study. Firstly, only male patients were enrolled in our study. Women with COPD are less usual than men with COPD but still common (with the prevalence of 8.1% (95% CI [6.8–9.3]) (Fang et al., 2018). The metabolic pattern of lung tissue in women with COPD needs further clarification to provide more awareness of the pathogenesis. Secondly, we did not divide the controls into non-smokers and smokers according to smoking status due to the scarcity of the number of samples. However, we ensured that there was no difference in smoking status between the COPD patients and the controls (Table 1). Further study with a larger of samples should be performed to compare the metabolic profiles of lung tissue in non-smokers, smokers and COPD patients. Finally, considering the instability of untargeted methods, our results also need more studies and targeted metabolomic analysis to further validate.

In summary, we characterized the metabolic profile of lung tissue from COPD patients. The metabolite profiles of COPD patients were different from those of controls. Additionally, glycerophospholipid metabolism and the valine, leucine and isoleucine biosynthesis pathways were significantly disrupted in COPD patients. Furthermore, we also investigated two metabolites, phytosphingosine and l-tryptophan, as overlapping metabolite biomarkers in three sets. Our results demonstrated that the abnormality of metabolic profiling and the perturbed pathways can help to elucidate the pathological mechanisms underlying COPD, and the screened metabolite biomarkers can effectively discriminate COPD samples from controls. These results may help to establish a foundation for future research attempting to characterize and treat COPD.