Introduction

Rheumatoid arthritis (RA) is a systemic inflammatory disorder involving the synovial joints. It is treated by conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), or biologic or targeted synthetic DMARDs (b/tsDMARDs).

Study participants with RA are at increased risk of becoming infected with nontuberculous mycobacterial pulmonary disease (NTM-PD)1, which is diagnosed by microbiologic and clinical findings2. Although NTM-PD is caused by several species of nontuberculous mycobacteria, NTM-PD has mainly been caused by Mycobacterium avium complex (M. avium and M. intracellulare) in Japan3. The risk of NTM-PD is thought to be elevated for RA study participants treated with b/tsDMARDs4,5,6. NTM-PD is a chronic lung disease7,8,9,10,11 and causes bronchiectasis. RA is also associated with bronchiectasis12,13,14, which leads to difficulties in diagnosis of NTM-PD in RA patients. Serum anti-glycopeptidolipid core antibodies have been used as biomarkers for NTM-PD associated with RA, although the sensitivity of these biomarkers has not been proven adequate15,16, Thus, biomarkers for NTM-PD are needed to identify its presence in study participants with RA.

Lipidomic and metabolomic analyses have been used to study low-molecular-weight compounds and to clarify disease pathogenesis and identify novel biomarkers associated with various diseases17. They have recently been used to investigate abnormalities in the metabolism of lipids, which might be involved in the pathogenesis of various diseases, including NTM-PD18,19,20. However, the findings of these investigations have not been validated for NTM-PD associated with RA. Here, we report our investigations on the lipidomic profiles in urine specimens of RA study participants with NTM-PD, in order to identify novel biomarkers for NTM-PD associated with RA.

Study participants and methods

Study participants

This study enrolled 90 study participants with RA from Tokyo National Hospital, Himeji Medical Center, Yokohama Medical Center, Nagasaki Medical Center, Asahikawa Medical Center, Sagamihara National Hospital, and Fukushima Medical University. All study participants fulfilled the criteria for RA21,22. The diagnosis of RA of NTM-PD was in accordance with the criteria of the American Thoracic Society2. Of 45 RA study participants, NTM-PD was caused in the participants by M. avium (36 individuals), M. intracellulare9, M. gordonae2, and M. ulcerans subsp. Shinshuense1 with overlap. RA patients with or without NTM-PD were included and the RA patients under 20-year-old were excluded from this study.

There were 2 groups of RA study participants with NTM-PD and 2 groups of control RA study participants with NTM-PD. Group 1 consisted of NTM-PD( + ) RA cases 1–1 to 1–20 and NTM-PD( − )RA control cases 1–1 to 1–20, and group 2 consisted of NTM--PD( + )RA cases 2–1 to 2–25 and NTM-PD( − )RA control cases 2–1 to 2–25. The controls were matched for age (differences between the ages of each case and control were less than < 15 years); sex; and use of corticosteroids, csDMARDs, or b/tsDMARDs. Group 1 NTM-PD( + )RA cases were enrolled from Tokyo National Hospital, and Group 2 NTM-PD( + )RA cases were enrolled from the other institutions. All the control study participants were recruited from Tokyo National Hospital.

This study was submitted to and approved by the National Hospital Organization Central Institutional Review Board (R4-0,819,001) and the ethical research committee of Fukushima Medical University School of Medicine. Written informed consent was provided by each study participant. This study was conducted in accordance with the principles of the 2024 revision of the Declaration of Helsinki.

Lipidomic profile analysis of urine specimens

Urine specimens were obtained from the study participants and centrifuged at 1200 rpm for 10 min, and the supernatants were transferred to vials and stored at − 70 ℃ before being analyzed. The lipidomic profiles of urine specimens were analyzed by liquid chromatography (LC) triple quadrupole mass spectrometry (MS) at the Lipodome Lab (Lipidome Lab Co., Ltd., Akita, Japan) based on the methods described previously in accordance with the Lipidome Lab Multi-Phospholipid Scan package23,24. Briefly, urine samples were treated with methanol and lipids were extracted, as previously described25. LC–MS/MS analyses were performed using Xevo TQ-XS with ACQUITY UPLC H-Class (Waters Corporation, Milford, MA, USA). Lipids were separated on Waters X-Bridge C18 (Waters Corporation, 3.5 μm, 150 mm × 1.0 mm). Data were analyzed with MetaboAlign (Mitsui Knowledge Industries Co., Ltd., Tokyo, Japan).

Statistical analysis

The web server MetaboAnalyst 6.0 (https://www.metaboanalyst.ca/home.xhtml) was used to analyze the results of the urine phospholipid assays26. The zero value was replaced with one fifth of the minimum positive value. Normalization with auto-scaling was performed as mean-centered and divided by the standard deviation of the mean of each phospholipid concentration. The Ward method and Euclidean distance were used to perform hierarchical clustering, and heatmaps with dendrograms were constructed.

Univariate analysis and the paired Student t-test were used to compare normalized phospholipid levels. Patient groups 1 and 2 were subjected to meta-analysis, and multiple comparisons were corrected by calculating the false discovery rate (FDR) Q value. Multiple logistic regression analyses were employed and deviation from 0 was evaluated for coefficients by the Wald test.

Receiver operator characteristic (ROC) curve analysis was performed to obtain areas under the ROC curves (AUROCs), which were used to select candidate phospholipids. Multivariate analysis and partial least squares-discriminant analysis (PLS-DA) were performed to evaluate compound phospholipid biomarkers for NTM-PD associated with RA. Two phospholipids with the highest AUROC values were analyzed.

Results

Clinical manifestations of the study participants with RA

Table 1 summarizes the demographic features of the NTM-PD( + )RA and NTM-PD( − )RA study participants. The body mass indices (BMIs) of participants with NTM-PD( + )R A were lower than the BMIs of participants with NTM-PD( − )RA, although the participants with NTM-PD( − ) RA were matched for age; sex; and treatments administered (corticosteroid, csDMARDs, or b/tsDMARDs). The NTM-PD( + ) RA participants had higher anti-glycopeptidolipid antibody concentrations than the NTM-PD( − )RA participants.

Table 1 Characteristics of study participants with rheumatoid arthritis.
Full size table

Urine phospholipid profiles of participants with RA with or without NTM-PD

Lipidomic analysis detected a total of 377 phospholipids (Supplementary Table S1). The concentrations of 340 phospholipids in the urine from participants with NTM-PD( + )RA were compared with the specimens from participants with NTM-PD( − )RA after normalization using auto-scaling (Supplementary Figure S2, Supplementary Table S2). Univariate analysis found significant differences between the two groups of participants. Table 2 shows the meta-analysis of the comparisons, which found significant differences after multiple comparisons adjusted by FDR. The concentrations of lysophosphatidic acid (LPA) 22:5 (FDR Q = 0), LPA 20:5 (FDR Q = 1.13 × 10−5), and LPA 22:2 (FDR Q = 0.0008) were lower in participants with NTM-PD( +) RA and the concentrations of phosphatidic acid (PA) 36:1 were higher (FDR Q = 0.0017) than those levels in the controls. Although urine phospholipids with Q values less than 0.005 were shown in Table 2, the differences in phosphatidylethanolamine (PE) 38:3, PE 38:4, PE 40:5, lysophosphatidylserine (LPS) 22:1, and phosphatidylinositol (PI) 38:4 were discordant between group 1 and 2. The results of multiple logistic regression analysis of BMI and urine phospholipids were shown in Supplementary Table S3, indicating the lack of associations with NTM-PD in some phospholipids (PE 38:3, PE 38:4, PE 40:5, PI 38:4) after adjustment. Thus, some phospholipids (LPA 22:5, LPA 20:5, LPA 22:2, PA 36:1, and LPA 20:3) were still robustly associated with NTM-PD.

Table 2 Metabolites in study participants with nontuberculous mycobacterial pulmonary disease associated with rheumatoid arthritis.
Full size table

Hierarchical clustering with visualization on a heat map were performed for these phospholipids (Supplemental Figure S1). Hierarchical clustering of the phospholipids did not indicate any apparent differences between the overall patterns of phospholipids between NTM-PD( + )RA and NTM-PD( − )RA participants, although in some individual the levels of phospholipids did vary. Since differences in individual phospholipids were found, differences in phospholipid groups might be detected. Phospholipid groups were also compared to evaluate the total concentrations of phospholipid groups (Supplemental Table S4 and S5). However, the differences between NTM-PD( + )RA and NTM-PD( − )RA participants were not detected. Thus, the urinary concentrations of some phospholipids were significantly skewed in participants with NTM-PD( + )RA .

Potential biomarkers for study participants with RA and NTM-PD

ROC curves of the phospholipids were constructed and the AUROCs were calculated (Supplementary Table S2, Fig. 1A–E, LPA22:5 AUROC 0.977, 95% confidence interval [CI] 0.942–0.996, PA 36:1 AUROC 0.811, 95% CI 0.694–0.900, LPA 20:5 AUROC 0.791, 95% CI 0.686–0.876, PIP3 36:4 AUROC 0.751, 95% CI 0.644–0.843, PA 38:4 AUROC 0.751, 95% CI 0.637–0.843). A PLS-DA model was created using the two phospholipids with the highest AUROC values (LPA22:5 and PA 36:1). The AUROC for the 2 phospholipids was 0.988 (95% CI 0.958–1.000, Fig. 1F). The mean accuracy, which was determined by 100 cross validations, was 0.959; and a permutation test was performed (permutation P < 0.001). Thus, the PLS-DA model of the two phospholipids produced compound biomarkers with good performance.

Fig. 1
figure 1

Results of ROC curve analyses. (A–E) ROC curve analysis of phospholipids with highest AUC values (LPA 22:5, AUC 0.977 95% confidence interval [CI] [0.942–0.996], PA 36:1, 0.811 [0.694–0.900], LPA 20:5, 0.791 [0.686–0.876], PIP3 36:4, 0.751 [0.644–0.843], PA 38:4 0.751 [0.637–0.843]) in RA participants with or without NTM-PD. (F) ROC curves of the PLS-DA model with two phospholipids with higher AUC values (LPA 22:5 and PA 36:1) comparing RA participants with or without NTM-PD. The AUC value of the ROC curve is 0.988 and 95% CI of the AUC is 0.958–1.000. RA: rheumatoid arthritis; NTM-PD: nontuberculous mycobacterial pulmonary disease; NTM-PD( + )RA: RA participants with NTM-PD; NTM-PD(-)RA: RA study participants without NTM-PD; AUC: areas under the curve; ROC: receiver operating characteristic; PLS-DA: partial least squares-discriminant analysis; LPA: lysophosphatidic acid; PA: phosphatidic acid; PIP3: phosphatidylinositol triphosphate.

Full size image

Discussion

In this study, meta-analysis identified differences between the urinary levels of some phospholipids in study participants with RA plus or minus NTM-PD. PLS-DA was used to produce a compound urinary biomarker consisting of the two phospholipids LPA22:5 and PA36:1 for NTM-PD associated with RA. Although serum LPA22:5 levels were reported to be increased in diabetic kidney disease27, the roles of the phospholipids, LPA 22:5, LPA 20:5, LPA 22:2, PA 36:1, and LPA 20:3, in the pathogenesis of NTM-PD have not been reported. Since four of these five phospholipid levels were significantly decreased in the participants with NTM-PD( + )RA, these candidate phospholipids for biomarkers of NTM would not be derived from NTM microbial cell components and changes in phospholipid metabolism were considered to be associated with the pathogenesis of NTM-PD. A serum lipidomic analysis has revealed that concentrations of the hydroxy fatty acid esters in study participants with mycobacterium infections vary compared with the concentrations in healthy control participants20. A urine metabolomic analysis of study participants with NTM-PD revealed changes in the concentrations of amino acids in their urine specimens19. A metabolomic investigation of sputum specimens from study participants with cystic fibrosis and NTM-PD has been reported18. However, these studies were not conducted in RA populations. Phospholipids were not analyzed in these previous studies. The candidate phospholipids for biomarkers of NTM-PD in RA found in the present study could not be detected in these previous metabolomic studies. Our study findings that demonstrated differences between the profiles of these metabolites in the participants with NTM-PD( + )RA and the participants with NTM-PD( − )RA suggest that the phospholipid metabolism in individuals with NTM-PD( + )RA is different from that metabolism in individuals with NTM-PD( − )RA. However, the results of serum metabolomic analyses in study participants with RA and healthy controls have been compared, and there were obvious differences between the metabolomic profiles of the two study groups28. In our study, the profiles of some phospholipids varied between NTM-PD( + )RA and NTM-PD( − )RA participants, but the number of the metabolites with significant differences between NTM-PD( + )RA and NTM-PD( − )RA participants seemed to be lower than the number of metabolites with significant differences between RA participants and healthy control participants.

Our study sample size was relatively small. Although in Japan, NTM-PD associated with RA is mainly caused by Mycobacterium avium complex, NTM-PD associated with RA in Europe is predominantly caused by Mycobacteroides abscessus. Independent large-scale studies on the lipidomic profiles of NTM-PD in different populations are necessary for the validation of our results.

To our best knowledge, this is the first report on the use of a lipidomic analysis to identify NTM-PD in the urine specimens of our study participants with RA. The results of univariate analyses showed significantly different urine levels of some phospholipids in RA participants with NTM-PD. The PLS-DA analysis identified a candidate biomarker consisting of two phospholipids that might serve as a biomarker for NTM-PD in study participants with RA. The biomarker was more sensitive than the anti-glycopeptidolipid antibodies used as current biomarkers for NTM-PD in RA study participants. Lipidomic profiling should be useful for generating biomarkers with increased sensitivity for NTM-PD in study participants with RA.