Abstract Background Sarcoidosis is a systemic inflammatory disease, primarily affecting the lungs, with a prognosis that varies widely among patients. While some patients recover spontaneously after diagnosis, others experience disease progression. Currently, the metabolomic profile associated with pulmonary sarcoidosis and its different clinical outcomes remains poorly understood. Methods Serum samples from 29 pulmonary sarcoidosis patients and 10 healthy controls were analyzed using untargeted UPLC-MS/MS metabolomics. Univariate and multivariate analyses identified differentially expressed metabolites, followed by pathway enrichment to evaluate their biological relevance. Patients were further stratified into self-healing (n = 11) and progressive (n = 18) subgroups based on prognosis. Differential metabolites between subgroups were compared, potential biomarkers were selected, and their diagnostic performance assessed. Correlations with clinical parameters were also analyzed to explore associations with disease progression. Results Sarcoidosis patients showed distinct serum metabolic profiles compared to healthy controls, with 10 upregulated and 199 downregulated metabolites. Pathway analysis indicated enrichment in amino acid, lipid, and immune-related pathways. Between prognostic subgroups, 25 differential metabolites were identified. Uric acid, testosterone sulfate, allopregnanolone sulfate, and 24,25-dihydroxyvitamin D[3] emerged as key metabolites with prognostic value and moderate correlations with clinical parameters. Conclusions This study highlights distinct serum metabolic profiles associated with sarcoidosis prognosis, suggesting that specific metabolic alterations may aid in monitoring and predicting disease outcomes. These findings offer a foundation for future research into personalized treatment and management strategies. Supplementary Information The online version contains supplementary material available at 10.1186/s12890-025-03863-y. Keywords: Sarcoidosis, Metabolomics, Prognosis, UPLC-MS/MS, Serum Background Sarcoidosis is a multisystem granulomatous disease of unknown etiology, primarily affecting the lungs and lymphatic system, with an annual incidence of 1 to 15 cases per 100,000 people [[44]1]. Pulmonary sarcoidosis is diagnosed through a combination of chest imaging that reveals infiltrates or lymphadenopathy, bronchoscopy with biopsies confirming non-necrotizing granulomas, and excluding infections or other interstitial lung diseases [[45]2]. The pathogenesis of sarcoidosis remains unclear, and its clinical course is highly heterogeneous. While some patients experience spontaneous resolution or respond well to treatment [[46]3], others progress to chronic or fibrotic stages despite receiving systemic corticosteroid therapy, often resulting in irreversible lung damage and significantly reduced quality of life. A major clinical challenge in managing sarcoidosis lies in the absence of reliable, non-invasive prognostic biomarkers to accurately predict disease progression. Consequently, therapeutic decisions—particularly regarding the initiation and intensity of corticosteroid treatment—are often guided by clinical experience and empirical judgment, which may result in overtreatment or delayed intervention. Given the minimally invasive nature of serum collection, considerable efforts have been directed toward identifying serum-based biomarkers to support diagnosis, prognosis, and treatment stratification in sarcoidosis [[47]4–[48]6]. Considering the disease’s complexity and heterogeneity, omics-based and systems biology approaches offer powerful tools for uncovering the underlying biological mechanisms across different phenotypes [[49]7]. However, most previous studies have focused on differentiating various types of interstitial lung diseases [[50]8] or distinguishing sarcoidosis patients from healthy controls [[51]9], rather than addressing prognosis. This underscores the translational value of identifying novel biomarkers for personalized management of sarcoidosis. Metabolomics can delineate the metabolic changes in the process of immune response, discover the key pathways and potential biomarkers related to immune activation, inhibition or dysregulation, and provide an important basis for early diagnosis, therapeutic monitoring and targeted therapy of immune diseases. Previous NMR-based metabolomics studies have shown that the metabolic disturbances observed in sarcoidosis patients can be interpreted as a holistic fingerprint of their physiological or pathological state, potentially influenced by environmental exposures and sample-related factors [[52]10–[53]12]. Interestingly, the metabolomic profiles of sarcoidosis patients display overlapping characteristics with both uveitis [[54]13] and tuberculosis [[55]14], indicating that serum metabolomic analysis could serve as a non-invasive alternative for sarcoidosis diagnosis and management. Despite these advances, most earlier studies have been limited by narrow metabolite coverage and a lack of in-depth exploration of disease heterogeneity or prognostic stratification [[56]15, [57]16]. A recent NMR-based metabolomics study revealed distinct serum profiles between sarcoidosis resolvers and progressors, with elevated trimethylamine N-oxide and taurine, and reduced glycolate, alanine, and proline in progressors [[58]17]. These findings highlight the potential of metabolomics for monitoring disease trajectory. Nonetheless, comprehensive LC-MS-based profiling linked to clinical outcomes remains insufficiently explored. In this study, an untargeted metabolomics approach based on ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) was employed to analyze the serum metabolic profiles of pulmonary sarcoidosis patients. Additionally, metabolic differences among patients with varying prognoses were explored to gain further insights into disease progression. Methods Study design and participants The study was approved by the Institutional Review Board of Shanghai Pulmonary Hospital (K21-074Z). Participants were recruited at Shanghai Pulmonary Hospital between September 2021 and November 2024. Patients with pathologically confirmed sarcoidosis who had been followed for more than one year and had a clearly defined prognosis were included. To minimize potential confounding effects on metabolite analysis, individuals who had received corticosteroid treatment within the preceding six months were excluded. Peripheral serum samples were collected during the patients’ initial clinical visits. Written informed consent was obtained from all participants prior to enrollment. The diagnostic criteria for sarcoidosis followed the 2020 guideline of the American thoracic society [[59]18]. Pathological tissue biopsies were obtained from all patients, revealing non-necrotizing granulomas. Other granulomatous diseases, including tuberculosis, fungal infections, parasitic infections, tumors, and vasculitis, were excluded. Tuberculosis was specifically ruled out in patients with negative sputum smears using TB-PCR [[60]19], GeneXpert MTB/RIF Assay, and a multi-parameter scoring system for distinguishing sarcoidosis from sputum negative tuberculosis (variables in the scoring systems include the results of the tuberculin skin test, the presence of tachypnea, extrathoracic presentation, symmetrical lymphadenopathy, cavity or calcification and lesion location in radiography) [[61]20]. We evaluated patients’ clinical manifestations, serological markers (e.g., 1,3-β-d-Glucan Testing, galactomannan test, cryptococcal antigen latex agglutination test, QuantiFERON-TB Gold detection and autoantibody profiles), radiological features (including chest CT and PET/CT), as well as pathological features. All patients included in the study showed pathological evidence of non-necrotizing granulomas. Special stains (acid-fast, PAS, iron, and hexamine silver) were performed to further exclude tumors and granulomas caused by environmental factors or fungal infections. All patients were followed up for more than one year. The entry date was defined as the date when bilateral mediastinal lymph node enlargement or lung opacities suggestive of sarcoidosis were identified on chest CT scans. Follow-up evaluations, including telephone interviews or outpatient visits, were performed every three months. The mean follow-up time was 18.45 ± 4.99 months. The healthy control (HC) group consisted of 10 individuals. Based on clinical diagnosis, 29 patients were categorized into two subgroups: (1) the self-healing group (SG, n = 11), defined as those who, over multiple follow-up visits, exhibited significant reduction or complete resolution of pulmonary lesions on chest CT without corticosteroid treatment, accompanied by marked symptom alleviation or resolution; and (2) the progressive group (PG, n = 11), which included patients who, despite not receiving corticosteroid therapy, showed an increase in pulmonary lesions and/or worsening symptoms over multiple follow-ups, those who were clinically determined to require steroid treatment after comprehensive evaluation, or those who experienced significant disease progression or relapse despite undergoing corticosteroid therapy. Sample collection and preparation Before UPLC-MS/MS analysis, serum samples were thawed in an ice-water bath. Each sample was transferred to a 1.5 mL centrifuge tube and mixed with three volumes of pre-chilled methanol/acetonitrile (1:1, v/v) for protein precipitation. After thorough vortexing, the mixture was centrifuged at 4 °C. The resulting supernatant was divided into two aliquots, lyophilized, and stored at − 80 °C until further use. One aliquot was reconstituted in 50% methanol/water for reverse-phase chromatography, while the other was reconstituted in 80% methanol/water for hydrophilic interaction liquid chromatography (HILIC). Both were centrifuged prior to injection. Additionally, equal volumes from individual samples were pooled to prepare quality control (QC) samples. Liquid chromatography Chromatographic separation was performed using a Waters ACQUITY UPLC HSS T3 column (1.8 μm, 2.1 × 100 mm), maintained at 40 °C on a Waters Acquity UPLC I-Class system. The mobile phase comprised water containing 0.1% formic acid (A) and acetonitrile containing 0.1% formic acid (B), delivered at a flow rate of 0.4 mL/min. The gradient began with 0.5% B for 1 min, followed by a linear increase to 95% B over 20 min. Afterward, the column was flushed with 95% B for 5 min and re-equilibrated for 5 min before subsequent injections. For HILIC, separation was achieved using a Waters ACQUITY UPLC BEH HILIC column (1.7 μm, 2.1 × 100 mm), maintained at 45 °C. The mobile phases included A (10 mM ammonium acetate in 95% acetonitrile/5% water) and B (10 mM ammonium acetate in 50% water/50% acetonitrile), with a flow rate of 0.4 mL/min. The gradient program was as follows: 0–1.0 min, 2% B; 1.0–12.0 min, 2–45% B; 12.0–14.0 min, 45–90% B; 14.0–16.5 min, 90% B; and 16.5–20.0 min, 2% A. All samples were injected with a fixed volume of 3 µL. Mass spectrometry The Q-Exactive mass spectrometer (Thermo Fisher Scientific, MA, USA) equipped with an electrospray ionization source was operated in both positive and negative ion modes, with spray voltages of 3.8 kV and 3.2 kV, respectively. The capillary temperature was set at 320 °C, with sheath and auxiliary gas flow rates at 40 and 10 arbitrary units. The S-lens RF level was 50, and the auxiliary gas heater temperature was 450 °C. Data acquisition covered an m/z range of 70–1,050 at a resolution of 70,000. The eight most intense ions were selected for data-dependent MS2 analysis at a resolution of 17,500. System calibration and QC were rigorously maintained. Three blank samples and six QC samples were initially run to stabilize the system. QC samples were injected every 8–10 runs during the UPLC analysis to monitor data quality and assess variability. Data analysis The UPLC-MS/MS raw data were processed using Compound Discoverer 3.3 software, which utilized an untargeted metabolomics workflow for peak alignment, feature filtering, gap filling and metabolite annotation. The retention time (RT) tolerance was set to 0.2 min, and the mass tolerance was set to 5 ppm. The signal-to-noise (S/N) ratio threshold was 1.5, and a peak intensity threshold of 10,000 was applied. To ensure reliable features, a peak rating threshold of 5 was implemented, and only those peaks detected in at least five samples were retained. This approach aimed to minimize poorly reproducible features and ensure accurate and comprehensive matrix data generation. Metabolites were annotated using online and in-house databases, including the human metabolome database (HMDB), Massbank, mzCloud, ChemSpider, and mzVault, by matching detected MS/MS spectra or exact masses with reference data. In the positive ion mode, [M + H]^+ was the primary ion selected, along with [M + Na]^+, [M + K]^+, and [M + NH[4]]^+, while in the negative ion mode, [M − H]^− was used as the primary ion. Potential peaks originating from mobile phase contaminants such as acetonitrile, formic acid, or dehydration byproducts were also taken into account. Pathway enrichment analysis was performed via the Metaboanalyst 6.0 ([62]www.metaboanalyst.ca) [[63]21], and the related pathways of differential metabolites were analyzed by using the KEGG online database. Statistical analysis All statistical analyses were performed using GraphPad Prism 8 and R (version 4.4.2). Data normality was assessed using the Shapiro–Wilk test. For normally distributed data, values are expressed as mean ± standard deviation (SD), and comparisons between groups were performed using Student’s t-test or one-way ANOVA. For non-normally distributed data, results are presented as median (interquartile range), with group comparisons analyzed using the Mann-Whitney U test or Kruskal-Wallis test. Statistical significance was defined as p < 0.05 for all analyses. Categorical data were compared using the chi-square test. Multivariate statistical analyses were performed using MetaboAnalyst 6.0 or Metware Cloud ([64]https://cloud.metware.cn). Normalized data were subjected to principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA). To assess the risk of model overfitting in OPLS-DA, permutation tests were conducted 200 times. Differential metabolites were identified based on the OPLS-DA S-plot, Variable Importance in Projection (VIP) score (> 1.0), and independent-samples t-test (p < 0.05). Key metabolic alterations were further evaluated for their diagnostic potential using receiver operating characteristic (ROC) curve analysis. Spearman’s correlation coefficient was calculated for correlation analysis. Results Clinical characteristics of participants This study included 29 patients with sarcoidosis and 10 healthy controls, with no significant differences in age or BMI between the groups (Table [65]1). The patient cohort comprised 22 females and 7 males, with a mean age of 50.41 ± 11.16 years and a mean BMI of 24.06 ± 3.33. Based on prognosis, patients were further stratified into a progressive group (PG) and a self-healing group (SG). No statistically significant differences were observed between these two subgroups in terms of blood lipid levels, T-cell subsets, immune-related cytokines, or immunoglobulin levels (Table [66]2). However, KL-6 levels were significantly higher in the PG group compared to the SG group (Fig. [67]S1). Table 1. Demographic data of sarcoidosis patients and healthy control group Sarcoidosis (n=29) Control (n=10) p-value SG (n=11) PG (n=18) Age (years) 51.18 ± 12.06 49.94 ± 10.90 40.30 ± 6.62 0.27 Gender (male/female) 4/14 3/8 5/5 N/A BMI (kg/m^2)^* 25.39 (21.76, 27.31) 22.15 (20.98, 24.90) 24.53 (21.83, 26.81) 0.30 Race Asian Asian Asian / Smoking status  No 10 17 9 /  Former or current smoker 1 1 1 0.90 Follow-up time (months) 17.82 ± 3.09 18.83 ± 5.91 / 0.55 Organ involvement  Lung 11 18 /  Peripheral lymph node 6 6 / 0.26  Spleen 1 0 / 0.19  Skin 1 3 / 0.57  Eye 0 2 / 0.25 Organ involvement  1 5 8 / 0.96  2 4 9 / 0.47  ≥3 2 1 / 0.28 Scadding stage  0 0 0 / /  I 1 3 / 0.57  II 7 10 / 0.67  III 3 5 / 0.98  IV 0 0 / / Comorbidities  Hypertension 2 3 1 0.86  Diabetes mellitus 2 1 0 0.27  Coronary artery disease 0 1 0 0.55  Autoimmune diseases 0 0 0 / [68]Open in a new tab BMI Body Mass Index *Data are presented as median (25th, 75th percentile) Table 2. Clinical characteristics of sarcoidosis patients stratified into progressive and self-healing groups Characteristics SG (n = 11) PG (n = 18) p-value Stage (active/stable) 0/11 5/18 N/A SACE (IU/L) 35.90 ± 10.25 49.14 ± 29.82 0.1687 ESR (mm/h)^* 14.00 (4.00, 16.00) 16.5 (10.00, 22.50) 0.3218 24-h urinary calcium^* 4.97 (3.77, 6.01) (n = 9) 6.32 (4.75, 9.73) (n = 16) 0.2022 TG (mmol/L)^* 1.38 (0.76, 3.23) (n = 5) 1.26 (1.14, 2.31) (n = 9) 0.8981 TC (mmol/L) 1.870 ± 1.676 (n = 5) 1.640 ± 0.8586 (n = 9) 0.7360 HDL-C (mmol/L) 0.9800 ± 0.1564 (n = 5) 1.068 ± 0.2282 (n = 9) 0.4619 LDL-C (mmol/L) 2.826 ± 0.6677 (n = 5) 2.838 ± 0.7976 (n = 9) 0.9782 ApoA-I (g/L) 1.506 ± 0.3232 (n = 5) 1.424 ± 0.2667 (n = 9) 0.6194 ApoB (g/L) 0.7680 ± 0.1143 (n = 5) 0.8889 ± 0.2141 (n = 9) 0.2686 Lp(a) (mg/L) 147.1 ± 118.9 (n = 5) 81.86 ± 41.67 (n = 10) 0.1339 Hcy (IU/L) 9.291 ± 3.501 (n = 11) 11.06 ± 1.290 (n = 14) 0.0933 IL-1β (pg/mL) 1.323 ± 1.064 (n = 6) 1.904 ± 1.905 (n = 9) 0.5115 IL-2 (pg/mL) 1.487 ± 0.6791 (n = 6) 0.9144 ± 0.6692 (n = 9) 0.1307 IL-4 (pg/mL) 1.818 ± 1.262 (n = 6) 1.796 ± 0.8431 (n = 9) 0.9670 IL-5 (pg/mL) 0.9933 ± 0.8801 (n = 6) 0.8933 ± 0.8572 (n = 9) 0.8300 IL-6 (pg/mL)^* 4.26 (3.02, 6.69) (n = 6) 3.04 (2.50, 7.46) (n = 9) 0.3884 IL-8 (pg/mL)^* 5.18 (3.05, 21.65) (n = 6) 8.76 (8.41, 12.64) (n = 9) 0.0663 IL-10 (pg/mL)^* 2.51 (1.88, 3.87) (n = 6) 4.41 (2.64, 6.26) (n = 9) 0.0937 IL-12 p70 (pg/mL) 1.013 ± 0.9578 (n = 6) 1.354 ± 0.9092 (n = 9) 0.4979 IL-17 A (pg/mL)^* 1.87 (0.06, 5.64) (n = 6) 0.87 (0.12, 1.84) (n = 9) 0.5273 TNF-α (pg/mL) 1.963 ± 0.9074 (n = 6) 0.8756 ± 1.023 (n = 9) 0.0553 IFN-α (pg/mL) 1.515 ± 0.6934 (n = 6) 1.893 ± 1.522 (n = 9) 0.5813 IFN-γ (pg/mL) 1.602 ± 0.8940 (n = 6) 2.577 ± 1.882 (n = 9) 0.2618 KL-6 (U/mL)^* 142.5 (125.00, 161.25) (n = 8) 173.00 (165.53, 298.00) (n = 11) 0.0112 CD3 (%) 63.53 ± 5.707 (n = 7) 67.37 ± 10.19 (n = 9) 0.3882 CD4 (%) 44.46 ± 3.506 (n = 7) 39.81 ± 8.003 (n = 9) 0.1751 CD8 (%) 16.49 ± 5.455 (n = 7) 23.24 ± 9.139 (n = 9) 0.1070 CD4/CD8 3.043 ± 1.241 (n = 7) 2.142 ± 1.382 (n = 9) 0.1984 IgG (g/L) 12.02 ± 2.098 (n = 11) 13.78 ± 3.005 (n = 13) 0.1166 IgA (g/L)^* 1.03 (0.91, 1.39) (n = 11) 1.14 (0.84, 1.35) (n = 13) 0.9547 IgM (g/L)^* 2.29 (1.59, 2.71) (n = 11) 2.34 (1.48, 3.68) (n = 13) 0.7646 IgE (IU/mL)^* 29.5 (18.45, 75.75) (n = 9) 19.40 (16.70, 22.30) (n = 13) 0.0986 [69]Open in a new tab SACE serum angiotensin converting enzyme, ESR erythrocyte sedimentation rate, TG Triglyceride, TC Total cholesterol, HDL-C High-Density Lipoprotein Cholesterol, LDL-C Low-Density Lipoprotein Cholesterol, Lp(a) Lipoprotein(a), Hcy homocysteine, KL-6 Krebs von den Lungen-6 *Data are presented as median (25th, 75th percentile) Statistical analysis of Serum metabolic profiles in sarcoidosis patients Serum samples from each group were analyzed using UPLC-Q Exactive Orbitrap MS to obtain metabolic profiles. PCA was performed to assess the initial group classifications. As shown in Fig. [70]1A, QC samples clustered tightly, indicating system stability and minimal operational variability. A noticeable trend toward differentiation between the Case and HC groups was observed, suggesting metabolic disruptions in sarcoidosis patients. When considering prognostic differences, a trend toward distinct metabolic profiles was seen between self-healing patients and those with disease progression (Fig. [71]1B). To further investigate the metabolic differences across groups, supervised PLS-DA was conducted. The PLS-DA model demonstrated robust and reliable performance, supported by strong statistical indicators (R² = 0.998, Q² = 0.552) (Fig. [72]1C). Fig. 1. [73]Fig. 1 [74]Open in a new tab Statistical analysis of serum metabolic profiles. A, B PCA score plots illustrating the distribution of different groups with and without prognostic information. C PLS-DA score plots for various groups. HC: Healthy control; SG: Self-healing group; PG: Progressive group; QC: Quality control Characterizing metabolic alterations in patients compared to healthy controls To better capture intergroup differences, OPLS-DA was applied to maximize metabolic separation and identify significantly altered metabolites between the sarcoidosis and control groups (Fig. [75]2A). The model’s validity was confirmed by permutation testing, with all permuted R^2 and Q^2 values falling below the original values and the Q^2 (cum) intercept at − 0.257 (Fig. [76]2B). These results demonstrate that the OPLS-DA model was both reliable and effective in distinguishing sarcoidosis patients from healthy controls. Fig. 2. [77]Fig. 2 [78]Open in a new tab Serum metabolomic analysis of sarcoidosis patients and healthy controls. A OPLS-DA score plot comparing sarcoidosis patients and healthy controls. B Permutation test for model validation. C Volcano plot of differential metabolites. D Heatmap of differential metabolites To investigate the differences in metabolites between the case and HC groups, metabolites with p < 0.05,|log[2]FC| >1, and VIP > 1.0 were considered as differential metabolites (Fig. [79]2C). A total of 209 metabolites were annotated (Fig. [80]2D, Table [81]S1), with 10 metabolites showing a significant increase in the case group, including 1-stearoylglycerophosphoserine, 5-hydroxyuracil, and LysoPC(14:0/0:0). In contrast, 199 metabolites exhibited a notable decrease, such as histidine, 25-hydroxycholesterol, and uric acid (UA). To explore the aberrant metabolic pathways in sarcoidosis patients, candidate metabolites were imported into MetaboAnalyst 6.0 for pathway enrichment analysis, with the goal of identifying relevant metabolic pathways and potential biological functions (Fig. S2). The results revealed that the differentially expressed metabolites were mainly associated with amino acid metabolism, particularly histidine and β-alanine pathways. Dysregulation in lipid metabolism was also observed, including alterations in bile acid and sphingolipid metabolism. Additionally, several pathways linked to immune responses—such as purine, pyrimidine, and arachidonic acid metabolism—were implicated. These metabolic disturbances underscore a complex interplay between energy homeostasis, immune signaling, and chronic inflammation in sarcoidosis. Comparative analysis of metabolic profiles in progressive and self-healing sarcoidosis patients To explore the serum metabolic characteristics of sarcoidosis patients with different prognoses, we performed OPLS-DA on metabolomics data from the PG and SG groups. The robustness and predictive performance of the OPLS-DA model were evaluated through 200 permutation tests, which confirmed strong predictive capability with R^2Y and Q^2 values of 0.975 and 0.65, respectively (Fig. [82]3A and B). Significant differential metabolites between the PG and SG groups were identified (p < 0.05,|log[2]FC| >1, and VIP > 1.0), revealing 8 upregulated and 17 downregulated metabolites (Fig. [83]3C and D, Table S2). Notably, most of these metabolites were lipids, suggesting that different lipid species may play distinct biological roles in sarcoidosis. Fig. 3. [84]Fig. 3 [85]Open in a new tab Serum metabolomics analysis of the progressive group and self-healing group in sarcoidosis. OPLS-DA score plot (A), permutation test (B), and volcano plot (C) comparing the progressive and self-healing groups. D Heatmap of differential metabolites Key metabolites for differentiating sarcoidosis and its prognoses Based on the differential metabolites identified between sarcoidosis patients and healthy controls, as well as between the progressive and self-healing groups, we identified four key metabolites: uric acid (UA), testosterone sulfate (TS), allopregnanolone sulfate (AS), and 24,25-dihydroxyvitamin D[3] (24R,25(OH)[2]D[3]) (Fig. [86]4A and B). To assess their predictive value for disease progression, we performed ROC curve analysis and compared their respective area under the curve (AUC) values, which were 0.848, 0.899, 0.818, and 0.879, respectively—comparable to that of KL-6 (Fig. [87]4C and S3). Furthermore, logistic regression analysis was conducted, with sarcoidosis set as the dependent variable and the four identified metabolites as independent variables (Fig. [88]4D). The ROC curve analysis of the resulting model demonstrated a robust predictive performance, achieving an AUC of 0.975, highlighting their potential as valuable diagnostic and prognostic biomarkers. Fig. 4. [89]Fig. 4 [90]Open in a new tab Analysis of key metabolites in sarcoidosis patients. A Venn diagram showing the identification of key metabolites. B Expression levels of key metabolites across different groups. C ROC analysis of four key metabolites or a panel of metabolites (D) for distinguishing SG and PG groups. E Correlation analysis between metabolites and clinical parameters. UA: Uric acid; TS: Testosterone sulfate; AS: Allopregnanolone sulfate; 24R,25(OH)[2]D[3]: 24,25-Dihydroxyvitamin D[3], TPR, true positive rate; FPR, false positive rate. * P < 0.05. *** P < 0.001 To explore the relationship between serum metabolites and clinical parameters, correlation analyses were conducted on key metabolites that showed significant changes in sarcoidosis patients. Several clinical parameters exhibited notable correlations with these metabolites (Fig. [91]4E). For instance, KL-6 and CD8 ^+ were negatively correlated with UA, while IL-2 showed positive correlations with both TS and AS. Additionally, IgG and IgM were positively and negatively correlated with TS, AS, and 24R,25(OH)[2]D[3], respectively. These findings suggest potential associations between serum metabolites and clinical parameters in sarcoidosis patients. Discussion Serum is widely used in clinical metabolomics because it reflects systemic metabolic changes. Although research on sarcoidosis-related metabolic alterations remains limited, emerging evidence points to distinct disruptions, particularly in lipid metabolism. For example, Lim et al. reported enrichment of cholesterol metabolism pathways in monocytes and macrophages from chronic sarcoidosis patients [[92]22]. Other studies have linked sarcoidosis with an increased risk of atherosclerosis [[93]23], and alterations in polyunsaturated fatty acid profiles have been proposed as potential predictors of disease involvement [[94]24]. Consistent with these findings, our study analyzed serum samples from sarcoidosis patients at initial diagnosis and identified significant dysregulation in amino acid, lipid, and immune-related metabolic pathways. Notably, we observed reduced bile acid metabolism, which is essential for cholesterol elimination and lipid homeostasis [[95]25]. This disruption may contribute to lipid accumulation and heightened atherosclerotic risk, potentially exacerbated by systemic inflammation and immune dysregulation—hallmarks of sarcoidosis [[96]26]. To minimize the confounding effects of glucocorticoids, which are known to affect lipid metabolism, we only included patients who had never received steroids or had discontinued them at least six months prior to sampling. Thus, the lipid abnormalities observed are more likely attributable to intrinsic disease mechanisms. Recent studies further support the link between lipid metabolism and immune activation in sarcoidosis. For instance, extracellular heat shock protein 90α (eHSP90α), secreted by macrophages, has been shown to drive inflammation and fibrosis in interstitial lung diseases, including sarcoidosis, through interference with LRP1, STAT-3, and PI3K/AKT pathways [[97]27]. In line with this, our study found reduced levels of 25-hydroxycholesterol (25HC), an oxysterol produced by activated macrophages. While 25HC can promote pro-inflammatory cytokines like IL-1 and IL-8 [[98]28], it also has anti-inflammatory functions by suppressing NLRP3 inflammasome activation [[99]29]. Its reduction may thus reflect impaired anti-inflammatory regulation and contribute to chronic inflammation in sarcoidosis. Additionally, we detected elevated LysoPC(14:0/0:0), a subtype of lysophosphatidylcholine (LPC), which is known to activate oxidative stress and inflammatory pathways, stimulate IFN-γ secretion, and promote B cell activity [[100]30, [101]31]. Interestingly, LPC can also enhance regulatory T cell function via G2A receptor signaling, promoting Foxp3 expression and immunosuppressive activity [[102]32]. This dual role underscores the immunometabolic complexity of sarcoidosis, where lipid mediators may simultaneously drive immune activation and regulation. Pulmonary sarcoidosis exhibits a highly variable clinical course, ranging from asymptomatic cases and spontaneous remission to chronic progressive disease involving multiple organs and requiring systemic therapy. Given that many patients remit spontaneously, treatment benefits may sometimes reflect the natural disease trajectory, while exposing patients to corticosteroid-related side effects [[103]33]. Currently, no reliable tools exist to predict prognosis, and the mechanisms underlying this variability remain unclear. The preferred clinical strategy is to closely monitor newly diagnosed patients and initiate treatment only upon signs of progression. In our study, serum metabolomics effectively distinguished between remitting and progressive sarcoidosis. Notably, four metabolites—UA, TS, AS, and 24R,25(OH)[2]D[3]—emerged as potentially relevant prognostic markers. Uric acid, the end product of purine metabolism, is primarily excreted by the kidneys, and up to one-third of sarcoidosis patients may present with renal involvement [[104]34]. In this study, a significant decrease in serum UA levels was observed, aligning with previous findings [[105]35]. Emerging evidence suggests a potential bidirectional regulatory relationship between UA metabolism and vitamin D metabolism [[106]36–[107]38]. We observed decreased serum levels of 24R,25(OH)[2]D[3]—a metabolite generated in the kidney from 25-hydroxyvitamin D[3] via 24-hydroxylation—indicating disrupted vitamin D metabolism. Previous studies indicate that sarcoidosis patients often have elevated levels of 1,25-dihydroxyvitamin D, primarily due to the excessive activation of 1α-hydroxylase in granuloma macrophages, this is also one of the reasons for hypercalcemia in patients with sarcoidosis [[108]39, [109]40], and 25-hydroxyvitamin D, as a precursor form, is deficient or normal in most patients with sarcoidosis [[110]41]. The reduction in 24R,25(OH)[2]D[3], particularly in progressive cases, may result from negative feedback by elevated 1,25-dihydroxyvitamin D or impaired synthesis due to renal tubular injury. Given the shared metabolic link between cholesterol and vitamin D synthesis, further studies are warranted to explore the interplay between vitamin D dysregulation, lipid metabolism, and renal function in sarcoidosis. TS and AS are steroid hormones that serve as precursors or metabolites of adrenal corticosteroids. Previous studies have shown that serum testosterone levels are significantly decreased in patients with chronic lung disease [[111]42, [112]43], and a notable proportion of male outpatients with sarcoidosis (46.7%) exhibit low circulating testosterone concentrations [[113]44]. Sarcoidosis, a disease characterized by immune system dysregulation, may impair specific hormone synthesis pathways due to persistent immune activation. This ongoing immune response can negatively impact adrenal function, leading to decreased synthesis and secretion of these hormones. Furthermore, sarcoidosis patients often face challenges such as inadequate nutrient intake or increased metabolic demands, which can result in insufficient levels of precursor substances essential for hormone synthesis, such as vitamin D and zinc. This, in turn, further disrupts the synthesis of adrenal corticosteroids. KL-6 is a mucin-like glycoprotein widely used as a biomarker for sarcoidosis and other interstitial lung diseases, with elevated levels linked to poorer prognosis and higher mortality risk [[114]45–[115]49]. In our study, KL-6 levels were significantly higher in the progressive group than in the self-healing group (p = 0.0278), likely reflecting greater and sustained alveolar epithelial injury. Notably, KL-6 levels showed a negative correlation with UA, suggesting interplay between immune activation and metabolic disturbance. Given UA’s antioxidant and immunomodulatory roles, its reduction may be influenced by chronic inflammation or altered renal handling. To rule out confounding by renal function, we conducted multivariate logistic regression adjusting for age, gender, BMI, and eGFR. Serum UA remained significantly associated with disease progression (OR = 0.99996, p = 0.030; Table S3), indicating that UA reduction likely reflects disease-related metabolic and inflammatory alterations rather than impaired kidney function alone. Our previous research found that IL-2R can serve as a serum biomarker for assessing multi-organ involvement and prognosis in sarcoidosis [[116]50]. IL-2 is an important cytokine primarily produced by activated T cells, and it plays a crucial role in maintaining immune homeostasis. IL-2R can bind to IL-2, thus regulating its activity. In this study, we found that IL-2 was positively correlated with two decreased steroid hormones, which may reflect the complex interplay between immune system activation and steroid hormone synthesis. This correlation might also suggest a potential link between immune dysregulation and hormonal imbalance in sarcoidosis patients. Additionally, the negative correlation between CD8^+ T cells and UA levels indicates a potential enhancement of peripheral CD8^+ cytolytic T cell responses in these patients [[117]51]. The observed correlations between multiple key metabolites and IgG and IgE in sarcoidosis patients could be attributed to various factors, including chronic inflammatory responses, immune system activation, shifts in Th1/Th2 balance, and changes in the cytokine environment. However, a more comprehensive understanding of the disease course requires considering other clinical and laboratory parameters. A major strength of this study lies in its comprehensive approach, employing UPLC-MS/MS to profile serum metabolites not only between sarcoidosis patients and healthy controls but also across prognostic subgroups defined through long-term clinical follow-up. Our findings suggest that metabolomic subtyping at diagnosis may help predict disease progression. Supporting this, a recent preprint by Dr. Drake’s group using NMR-based metabolomics similarly identified distinct serum metabolic profiles between regressing and progressive sarcoidosis [[118]17]. While their study reported elevated trimethylamine N-oxide and taurine and reduced glycerate, alanine, and proline in progressors, our cohort showed a similar increase in taurine but no significant changes in alanine or proline (Fig. S4)—possibly reflecting population differences. Despite methodological differences, both studies converge on the presence of distinct metabolic phenotypes, highlighting the promise of metabolomics in sarcoidosis prognosis. This study has several limitations. First, the relatively small sample size and absence of a disease control group limit the interpretation of disease-specific metabolic alterations. Second, as a retrospective study, some clinical data extracted from medical records may be incomplete, potentially affecting the accuracy of correlation analyses. Third, although we excluded patients with recent corticosteroid use to reduce medication-related confounding, other factors such as dietary habits were not systematically assessed due to a lack of standardized data—this represents a methodological limitation, as diet can influence metabolomic profiles. Finally, this study focused on a single omics layer. Future research should incorporate multi-omics approaches and prospectively control for variables such as medications, comorbidities, and diet to better elucidate the metabolic landscape of sarcoidosis. Despite these limitations, our findings provide meaningful insights into disease-related metabolic heterogeneity and lay the groundwork for further prognostic studies. Conclusions This study expands our understanding of serum metabolomic alterations in patients with sarcoidosis. The identified metabolic signatures closely reflect the prognostic differences between progressive and self-healing cases. Several metabolites showed associations with key clinical parameters, underscoring the potential of serum metabolomics as a valuable tool in the clinical management of sarcoidosis. Beyond aiding in diagnosis, these metabolic markers may serve as indicators for monitoring disease progression and evaluating treatment response. Supplementary Information [119]12890_2025_3863_MOESM1_ESM.pdf^ (880.9KB, pdf) Supplementary Material 1: Appendix A. Supplementary data. Supplemental figures and tables. Acknowledgements