Tuberculosis is caused by Mycobacterium tuberculosis, one of the most widespread infectious diseases worldwide (Procop, 2016), which is always divided into two types: pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB). Because of its lower incidence rate, EPTB does not attract enough attention compared to PTB. Osteoarticular tuberculosis is one of EPTB, whose M.tb directly infects the bone and joint tissue or spreads to the bone and joint tissue from other parts, and it is hard to distinguish with rheumatic immune arthritis and bone tumor at its early stage based on current clinical laboratory and clinical imaging methods (Fan et al., 2018). Metabolomics is one of the omics proposed together with proteomics and transcriptomics in recent years (Dutta et al., 2020). It is defined as the study of the complete set of metabolites inside cells, tissues, organs, and biological fluids. It represents a major and rapidly evolving component of systems biology—a new integrative approach to deciphering the complexity of biological systems (Preez et al., 2017). The metabolite changes associated with the specific phenotype being investigated may be classified as characteristics of the perturbation, which, in the context of investigating a disease, could be used toward better disease characterization, diagnostics, treatment, and other clinical applications. The methods of metabolomics include nuclear magnetic resonance (NMR) and mass spectrometry (MS) (Laíns et al., 2019). NMR is appropriate for the detection of all hydrogen-containing compounds, since it determines the magnetic resonance of nuclei in a molecule, and it is considered an unbiased, robust, reproducible, non-destructive, and selective analytical platform, which requires almost no sample pretreatment. On the other side, however, NMR has a low sensitivity and is short of an analyte separation element (Shin et al., 2011). MS is defined as the process of forming gaseous ions, with or without fragmentation, which are then characterized by their m/z ratios and respective relative abundances. Direct MS infusion is a high-throughput method, requiring short time for each sample analysis, and has been applied successfully in metabolomics studies, but it is not preferred for the analyses of complex biological samples such as blood and urine due to matrix interference (Schoeman et al., 2012). For TB, there are also more metabolomics studies focusing on colony culture, sputum specimens, blood specimens, urine specimens, tissue specimens, etc. (Parida and Kaufmann, 2010)

Zhou et al. (2015) and Albors-Vaquer et al. (2020) used NMR to detect differences in the expression of metabolites in the serum of PTB, healthy adults, lung-related benign lesions, and lung cancers; the results showed that some amino acids have been changed among these groups, such as alanine, lysine, glutamate, glutamine, ketone bodies, lactate, and pyruvate. Mendes Rˆego et al. (Rêgo et al., 2021) used MS to detect serum metabolites among drug-sensitive tuberculosis and drug-resistant tuberculosis; there were also some amino acids that were changed such as isoleucine, proline, hercynite, betaine, and pantothenic acid. On another aspect, Vrieling et al. (2019), Luo et al. (2020) and Chen et al. (2021) used MS to detect serum metabolites among PTB and healthy control, tuberculous pleuritis and malignancy, and PTB and PTB with type 2 diabetes, respectively. The results of these studies also showed some changes in amino acids, phospholipids, sphingolipids, etc., which are similar to other studies. According to other studies, amino acid levels such as alanine, lysine, glutamate, and glutamine in patients with tuberculosis infection have decreased significantly, which may be related to the uptake of glutamate, glutamate, and alanine by Mycobacterium tuberculosis for their corresponding life activities (Harth and Horwitz, 2003; Agapova et al., 2019). The body’s immune process for tuberculosis is mainly the activation of T cells, which also changes the law of glucose metabolism. The disease process affects the metabolism of related immune cells and thus affects the secretion of cytokines such as IFN-γ, further leading to a decrease in immune effect (Lande et al., 2003).

According to the characteristics of Mycobacterium tuberculosis, in addition to the cell membrane and peptidoglycan of ordinary bacteria, there are also a large number of lipids and carbohydrates on its surface. These components have strong biological activity on eukaryotic cells, thus suggesting that it has a strong relationship with pathogenicity. For the mechanism of tuberculosis, lipid metabolism and lipid effector molecules play a vital role, such as regulating the production of cytokines, scavenging oxygen free radicals, and producing granulation inflammation and mitochondrial toxicity (Walpole et al., 2018). These related pathogenic components have also been reflected in other studies. Our study focuses on the blood specimens of osteoarticular tuberculosis, compared with PTB; this type of research is currently rarely reported. The results of our study showed that compared with healthy adults and rheumatoid arthritis patients, patients with osteoarticular tuberculosis also have significant changes in serum amino acid and lipid metabolism. Most amino acids such as glycine, L-arginine, and L-alanine were downregulated in osteoarticular tuberculosis patients’ serum; in contrast, most lipids such as phosphatidylcholine, sphingomyelin, and phosphatidylethanolamine were upregulated in osteoarticular tuberculosis patients’ serum.

Further, our study selected five potential serum metabolite biomarkers, namely, PC(o-16:1(9Z)/18:0), PC(20:4(8Z,11Z,14Z,17Z)/18:0), PC(18:0/22:5(4Z,7Z,10Z,13Z,16Z)), SM(d18:1/20:0), and SM(d18:1/18:1(11Z)), which belong to phosphatidylcholine (PC) and sphingomyelin (SM). On the aspect of KEGG enrichments, necroptosis, choline metabolism, sphingolipid signaling, retrograde endocannabinoid signaling, sphingolipid metabolism, and glycerophospholipid metabolism are maybe the main pathways. In most cases, these metabolites are related to lipid accumulation and obesity in the body (Wahl et al., 2012; Reinehr et al., 2015; Hellmuth et al., 2016). At the same time, some other pathological processes that cause enhanced lipid metabolism can also significantly increase these metabolites; for example, Wu et al. (2018) have reported that the sphingomyelinase/ceramide system, which has shown several times to be a crucial factor in the internalization, processing, and killing of diverse pathogens, also modulates the pro-inflammatory response and the state of mycobacteria in macrophages, which highlights the important role of lipid metabolism in the pathogenic mechanism of pathogens. For PTB, sphingosine-1 phosphate (S1P) and ceramide are central molecules and are decisive for sphingolipid signaling; otherwise, they are about the secretion of interferon (IFN)-γ during the course of infection and infiltration of pulmonary CD11b+ macrophages and expression of S-1P receptor-3 (S-1PR3) in the lungs during the course of infection (Braverman et al., 2016; Nadella et al., 2019). Takenami et al. (2018) have a research about the IgM and total IgG antibody response to cardiolipin (CL), phosphatidylcholine (PTC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), and sulfatide (SL-I) as biosignatures that can be used to diagnose PTB and its applicability for monitoring the efficacy of antituberculosis treatment; the antibody concentrations of PTB patients were significantly higher than those of healthy control, which also indicates that lipids play a vital role in the process of Mycobacterium tuberculosis infection. Also, fatty acids can stimulate the activation of dormant Mycobacterium tuberculosis in liquid medium (Shleeva et al., 2013). Eicosanoids, lipid mediators derived from arachidonic acid, have been associated with the modulation of the host response to Mycobacterium tuberculosis infection (Bafica et al., 2005; Tobin et al., 2012). Moreover, it also reported increased eicosanoid ratios in plasma in tuberculosis patients compared to healthy control (Mayer-Barber and Sher, 2015). However, related research about the lipid in osteoarticular tuberculosis or even other extrapulmonary tuberculosis is rare; the results of our research, whether from the differential metabolites screened out or the signal pathways obtained from bioinformatics, are closely related to the results of pulmonary tuberculosis-related research (Huang et al., 2020; Han et al., 2021). It provides more reference value for future research on diagnostic biomarkers from the perspective of metabolomics and provides some hints for the establishment of subsequent auxiliary diagnostic methods for these differential metabolites.

In the future study, we will expand the scope of clinical serum samples, from the single-center collection to the multicenter collection in different regions, in order to minimize the limitations of the experimental results. At the same time, serum samples of different tuberculosis diseases such as pulmonary tuberculosis can be added as follow-up controls to determine whether the relevant markers are widely related to tuberculosis, and then extended to not only the auxiliary diagnosis for osteoarticular tuberculosis but also the establishment of a new method of laboratory examination for tuberculosis for reference.