Identication of Lipid Biomarkers for Skeletal Tuberculosis Using an Untargeted Metabolomics Approach

Background The identication of novel biomarkers in the human body is urgently needed to improve diagnosis and the clinical management of skeletal tuberculosis (STB). This study aimed to identify potential lipid markers to differentiate Mycobacterium tuberculosis infection from other infections, and establish a metabolite biomarker panel suitable for STB diagnosis from abscess samples. Methods Abscess specimens were collected from STB patients and patients diagnosed with other skeletal infections. Then we comparatively explored the lipid metabolomes of abscess specimens from STB and non-STB patients using untargeted lipid metabolomics approach.


Introduction
Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains the major public health concern worldwide, with an estimated 10.0 million incident cases in 2019 [1]. Although it most commonly affects the lungs, it can virtually affect any other anatomic site, termed as extrapulmonary tuberculosis (EPTB) [2,3]. Nearly 10% of global TB cases are EPTB cases, posing challenges for TB control efforts [1]. Skeletal tuberculosis (STB) is one of the most common sites of EPTB, both in terms of relative frequency and the substantial potential for permanent disability [2,4]. Timely diagnosis and treatment are critical in achieving successful outcomes and preventing disability associated with STB [5]. Unfortunately, similar to other forms of EPTB, STB is harder to diagnose due to its extremely paucibacillary nature [4]. Therefore, the identi cation of novel biomarkers in the human body is urgently needed to improve diagnosis and the clinical management of this EPTB form.
The wide application of omics technologies has extended our knowledge of new diagnostic and treatment strategies for infectious diseases, including TB [6,7]. Metabolomics is one of the newcomers to the omics revolution that identify and quantify of the complete metabolome of a speci c biological system in the nonbiased condition [8]. To date, several metabolomics studies have been conducted with the aim of explore speci c metabolite biomarkers in assisting TB diagnosis [9][10][11]. Some of the compounds identi ed as potential biomarkers were veri ed in subsequent reports, indicating their promising approach for clinical application [8]. However, the majority of previous studies were performed in blood samples to determine biomarkers for pulmonary TB; little is known about metabolites originating directly from tubercle bacilli, as well as metabolites altered in the host due to the infection in the specimens from EPTB patients.
Lipids play an important role in the pathology of tuberculosis, which is often associated with malnutrition and wasting syndrome [12]. Previous studies con rmed that the blood lipid level in host individuals is associated with bactericidal effect of rst-line antibiotics and phagosome maturation, suggesting the important role of lipids in bacteria-host interaction [13]. Given that, we comparatively explored the lipid metabolomes of abscess specimens from STB and non-STB patients using untargeted lipid metabolomics approach, to identify potential lipid markers to differentiate Mycobacterium tuberculosis infection from other infections. We aimed to establish a metabolite biomarker panel suitable for STB diagnosis from abscess samples.

Materials And Methods
Participants Participants were recruited from Beijing Chest Hospital, a tuberculosis-specialized hospital, between January 2018 and December 2019. Abscess specimens were collected from 10 STB patients with con rmed ethology by mycobacteria culture and/or GeneXpert MTB/RIF. For control group, abscess specimens were collected from 4 patients diagnosed with other skeletal infections as listed in Table 1.
Five milliliters of abscess samples were obtained from each patient, and then transported to the BioBank of Beijing Chest Hospital for storage. All samples were centrifuged at 12,000 g for 15 min, ltered with a disposable 0.22 µm cellulose acetate and stored at -80°C for further analysis. All procedures are performed in studies involving human participants were approved by the Ethics Committee of Beijing Chest Hospital. Signed informed consent was obtained from each participant prior to enrolment.

Lipid Extraction
One milliliter of each sample was pipetted in a 2 ml centrifuge tube and lyophilized in a freeze-dryer. Then samples were rstly bath sonicated for 15 min with 400 μl ice-cold 75% methanol to lyse cells. After mixing samples with 1 ml methyl-tert-butyl ether (MTBE, Merck, Darmstadt, Germany) for 1 h at room temperature, 0.25 ml of deionized water was added into mixture, and incubated at room temperature for 10 min. Then the samples were centrifuged for 15 min at 4 ℃. The upper organic phase was collected and freeze-dried.
Additionally, to ensure data quality for metabolism pro les, quality control (QC) samples were prepared by pooling aliquots of all samples that were representative of the all samples under analysis, and used for data normalization. Dried extracts were then dissolved in 50% acetonitrile. Each sample was ltered with a disposable 0.22 µm cellulose acetate and transferred into 2 ml HPLC vials and stored at -80°C prior to analysis.

UHPLC-MS/MS analysis
Metabolomics LC-MS analysis was conducted on a Thermo (Waltham MA, USA) UltiMate 3000 UHPLC system coupled to a Thermo Q Exactive Orbitrap mass spectrometer. A Hypersil GOLD C18 (100×2.1mm, 1.9 μm) (Thermo Scienti c) was used for lipid separation. The mobile phase A was prepared by dissolving 0.77g of ammonium acetate to 400ml of HPLC-grade water, followed by adding 600ml of HPLC-grade acetonitrile; the mobile phase B was the ow rate was set as 0.3 mL/min. The gradient was 30% B for 0.5 min and was linearly increased to 100% in 10.5 min, and then maintained to 100% in 2 min, and then reduced to 30% in 0.1 min, with 4.5 min re-equilibration period employed. Both electrospray ionization (ESI) positive-mode and negative mode were applied for MS data acquisition. The positive mode of spray voltage was 3.0 kV and the negative mode 2.5 kV. The ESI source conditions were set as follows: Heater Temp 300 ℃, Sheath Gas Flow rate, 45arb, Aux Gas Flow Rate, 15 arb, Sweep Gas Flow Rate, 1arb, Capillary Temp, 350 ℃, S-Lens RF Level, 50%. The full MS scans were acquired at a resolution of 70,000 at m/z 200, and 17,500 at m/z 200 for MS/MS scan. The maximum injection time was set to for 50 ms for MS and 50 ms for MS/MS. MS data was acquired using a data-dependent Top10 method dynamically choosing the most abundant precursor ions from the survey scan (200-1500 m/z) for HCD fragmentation. Stepped normalized collision energy was set as 15, 25, 35 and the isolation window was set to 1.6 Th. Blank samples (75% ACN in water) and QC samples were tested after every six samples for quality control purpose.

Data preprocessing and ltering
Lipids were identi ed and quanti ed using Lipid Search 4.1.30 (Thermo). Mass tolerance of 5 ppm and 10 ppm were applied for precursor and product ions. Retention time shift of 0.25 min was performed in "alignment". M-score and chromatographic areas were used to reduce false positives. The lipids with less than 30% RSD of MS peak area in QC samples were kept for further data analysis.

One-way analysis
The discriminating metabolites were obtained using a statistically signi cant threshold of variable in uence on projection (VIP) values obtained from MetaboAnalyst 5.0 (http://www.metaboanalyst.ca) and two-tailed Student's t test (p value) on the normalized raw data at univariate analysis level. The p value was calculated by one-way analysis of variance (ANOVA) for multiple groups' analysis. Metabolites with VIP values greater than 1.0 and p value less than 0.05 were considered to be statistically signi cant metabolites. Fold change was calculated as the logarithm of the average mass response (area) ratio between two arbitrary classes. On the other side, the identi ed differential metabolites were used to perform cluster analyses with R package.

Lipid identi cation
In this study, untargeted lipid metabolomics were conducted to investigate the metabolic dysregulation in paraspinal abscesses of STB. A total of 957 features in ESI + mode and 584 in ESI-mode were extracted from the lipidomic data. These 1541 lipids were classi ed into 39 lipid subclasses. Pooled QC samples were used to monitor the stability of the LC-MS system (Fig.1). When conducting differential lipid analysis between STB and non-STB, we used univariate analysis methods include Fold Change Analysis (FC Analysis), t test, and Volcano Plot that integrates the rst two Analysis methods. We used |FC|>1.5 and p value < 0.05 as the screening criteria to declare signi cant differences in the comparison group ( Fig.2A).

Differential metabolites between STB and non-STB
Among 1541 lipid metabolites, 55 lipid metabolites were signi cantly changed in STB patients compared to the control group. The VIP, FC, p-values and area under a ROC curve (AUC) values of differential metabolites were listed in Table S1. Signi cant up-regulation was clearly visualized in phosphatidylethanolamine (PE), phosphatidylinositol (PI), lysophatidylinositols (LPI), lysophosphatidylcholines (LPC) and lysophosphatidylethanolamine (LPE). Concomitantly, triacylglycerols (TG) and phosphatidylcholines (PC) were signi cantly down-regulated in patients affected by STB. In order to evaluate the rationality of different lipids, and to display the relationship between samples and the differences in lipid expression patterns in different samples more comprehensively and intuitively, we used qualitatively signi cant differences in lipid expression to perform hierarchical analysis on the two sets of samples. Fig.2B shows the results of signi cant difference in lipid hierarchical clustering analysis.

Potential diagnostic biomarkers of STB
As demonstrated in volcano plot ( Fig. 2A) and heatmap (Fig. 2B), the biomarkers had a clearly differential distribution between two groups. The AUC value was used to evaluate the diagnostic ability of biomarkers for STB. We presented diagnostic performances of 10 representative lipid metabolites consisting of 4 phospholipids, 4 lysophospholipids and 2 triacylglycerols (as the order of least q-values, Fig. 3A-D and Fig. S1). It is worth noting that the two down-regulated monoacyl chain phosphatidylcholine, triacylglycerols and one up-regulated LPE showed excellent diagnostic potential, the AUC values of which were higher than 0.9 and the highest value was 1.0. Due to the small sample size of this study, to further optimize the diagnostic performance of the biomarkers, we combined these 5 ether lipids as a panel. In multivariate ROC analysis, this panel yielded a better AUC value of 0.991 for diagnosis of STB patients (Fig. 3E).

Pathway analysis of STB
Based on the list of signi cantly regulated lipids, MetaboAnalyst (http://www.metaboanalyst.ca) was applied to investigate which pathway might be markedly perturbed. The result of the pathway analysis was graphically presented in Fig.4. From the enrichment analysis results, the Glycerophospholipid metabolism pathway had a statistically signi cant raw p-value (raw p < 0.05, as shown in the Y-axis).
Pathway impact results indicated that the Glycerophospholipid metabolism pathway presented higher impact than the other pathways, as indicated in the X-axis value. Combining the above two analysis results, we postulated that the Glycerophospholipid metabolism pathway to be a markedly perturbed pathway that correlated with the lipid rearrangement process induced by MTB infection in STB.

Discussion
In this study, using a lipid metabolomics approach, we comparatively explored the altered metabolomes that would differentiate MTB infection from other infections. Our data demonstrated that the lipid metabolic pro les of STB and non-STB patients were signi cantly different from that of abscess specimens, providing potential novel biomarkers for STB diagnosis. The varying metabolites were mainly related to the pathways of lysophospholipid, glycerophospholipid and phosphatidylcholine metabolism.
In particular, the levels of two LPEs were signi cantly higher in the STB cases than those in non-STB control. LPE is a constituent of cell membranes in the human host, which is derived from the hydrolysis of PE [14]. We speculate that the elevated abscess levels of LPEs originates from the destruction of cell membranes within necrotic lesions. The accumulation of LPEs may re ect more severe bone and tissue destruction by tubercle bacilli in STB patients than other patients. Our ndings may also be in line with previous observations of higher plasma levels of LPEs in rats with severe induced-liver injuries [15,16]. Therefore, it is also possible that the high LPE levels may be an indicator of disease severity in STB patients. In addition, previous experimental studies have shown that LPEs can serve as immune modulators that can stimulate the activation of multiple immune cells, including macrophage and natural killer T cell [17,18]. Similar results were found in a recent study by Lau and colleagues that the higher plasma levels of LPE in sepsis patients signi cantly correlated with proin ammatory cytokines [19]. Thus an interesting question yet to be answered is whether LPE may be involved in the immunity against tubercle bacilli. Further studies are urgently needed to verify this hypothesis.
Phosphocholines are reservoirs and transporters of fatty acids, phosphate, glycerol, and choline, and are also essential nutrients that maintaining health in adults [20]. In a recent metabolism study, a decrease in PC was noted in plasma samples of pulmonary TB patients [21]. In consistent with previous observations, we also observed lower abscess levels of PC in STB cases than non-STB cases. A hostpathogen metabolic ux model revealed that tubercle bacilli are able to consume PC on mycobacterial growth, thereby resulting in its decrease during infection [22]. Therefore this altered metabolism in PC may be the results of consuming nutrients by intracellular bacteria. In addition to energy source, PC participate in the innate immunity to ght intracellular bacteria [23]. Exogenous application of PC inhibits pro-in ammatory signaling in macrophages, thereby facilitating the survival of mycobacteria [23]. In view of these ndings, it is exciting to speculate whether their decrease in abscess specimens is actively regulated by host immune cells, thereby resulting in induction of MTB killing. The change in expression pattern of genes involving in PC metabolism in MTB-infected lesions is of importance to elucidate the molecular mechanism of PC against intracellular pathogen.
In addition, we also found that triacylglycerols, another major source of carbon and energy for MTB, were substantially decreased in abscess specimens from STB patients. This decrease may re ect the energy wasting in patients with STB, as noted in PC. A recent study on MTB by Daniel et al demonstrated that the bacteria could use host triacylglycerol to acquire a dormancy-like phenotype in macrophage [24]. According to this model, the decreased level of triacylglycerol served as an effective indicator for formulating dormant, non-replicating tubercle bacilli in lesions. In this aspect, the decreased triacylglycerol may be repurposed for energy storage in tubercle bacilli, thus improving their survival under hypoxia stress (Fig. 5). Further studies are also needed to identify these fatty acids that function in the lipid metabolism to MTB survival within host niches.
We also acknowledged several limitations to the present study. First, Due to the low recovery rate of pathogen from abscess specimens, only four patients with con rmed etiology were included in the control group. The small sample size may weaken the overall signi cance of our study. Second, although our method offered great sensitivity and speci city to differentiate STB from non-STB cases, the biomarkers require further validation using a separate larger sample cohort. Finally, we observed the varying regulation of lipid pro les between two different disease groups; however, its underlying molecular mechanism remains unclear. Despite these limitations, our results rstly provide important foundation for facilitating the diagnosis of STB patients using a panel of differential lipids.

Conclusions
To conclude, our data rstly characterize the lipid signatures of abscess specimens from STB patients. The LPEs are signi cantly upregulated in the STB cases than those in non-STB control, whereas phosphocholines and triacylglycerol are markedly downregulated in the STB cases. The panel of ve lipid biomarkers exhibits great capacity for differential diagnosis of STB and non-TB cases. Further studies are required for validating the performance of this novel diagnostic panel in a separate larger sample cohort.

Consent for publication
Informed consent for publication was obtained from all participants.

Potential con icts of interest
The authors declare no conflict of interest regarding the publication of this paper.   Metabolites volcano plot and heatmap of signi cantly different metabolites A. Volcano plot of differential lipids classi cation of the STB group and the non-STB group. The abscissa is a FC; the ordinate is the p value of one-way analysis of variance (ANOVA). Lipids with p value <0.05 obtained by ttest and FC 1.5 veri ed by Fold Change Analysis were identi ed as signi cantly differential metabolites.
Colored plots indicate upward trend and downward trend of substances, and gray plots indicate that they are not statistically signi cant. B. Heatmap of signi cantly different metabolites in STB and non-STB samples. Cells in each row represent individual samples. Red and bule color indicate increased and decreased levels, respectively.  Statistics of KEGG substance enrichment The ordinate represents the enriched pathway, and the abscissa represents the rich factor. The results show signi cant enrichment in glycerophospholipid metabolism (red color).