Introduction

Systemic lupus erythematosus (SLE) is a heterogeneous autoimmune disorder with an elusive etiology, showcasing diverse clinical manifestations that span from mild skin erythema and joint pain to severe complications such as end-stage renal disease and cerebrovascular accidents1. While established indices like SLE disease activity index 2000 (SLEDAI-2K), British Isles Lupus Assessment Group’s disease activity index (BLIAG) and SLE Responder Index 4 (SRI-4) offer reliable assessments of SLE activity, the identification of novel markers remains imperative for a comprehensive understanding of the disease state, early diagnosis, prognostic assessment, and targeted therapeutic intervention2,3,4,5.

Ferritin, recognized as both an iron storage protein and an acute phase response protein (APR), plays a pivotal role in cellular iron homeostasis, the level of which is regulated by intracellular iron status and the serum pro-inflammatory factors like IL-6 and tumor necrosis factor-α (TNF-α). Serum ferritin levels reflect the iron status of macrophages and may be derived primarily from macrophage secretion6. Animal experiments have demonstrated that ferritin induces systemic inflammation and triggers neutrophil extracellular trap (NET) formation, exacerbating systemic inflammation in diseases like adult Still’s disease in a macrophage scavenger receptor 1 (Msr1) -dependent manner7. Several studies have reported significantly elevated serum ferritin levels in active SLE patients, hinting at its potential role as a distinguishing marker that was more specific for SLE activity than conventional marker8,9,10,11,12. Moreover, associations between elevated serum ferritin and adverse outcomes, such as anemia, premature delivery in mid-pregnancy SLE patients, and even macrophage activation syndrome in SLE13,14,15, have been noted, suggesting that the elevated serum ferritin may not only be a consequence of inflammatory activity, but also as a predictor for poor disease prognosis or an indicator of an adverse disease state in SLE patients14,16. However, the precise mechanistic underpinnings of elevated serum ferritin in SLE pathology remain elusive.

The studies of metabolomics in SLE are increasingly intensive and have made great progress. For instance, studies have revealed that dysregulated tryptophan metabolism, elevated ceramides levels, and aberrant glucose metabolism are associated with the onset, lupus manifestations and disease activity prediction17,18,19,20,21. Moreover, metabolomic-based studies have discerned that metabolomic signatures can predict the progression of atherosclerosis in a subset of patients with juvenile-onset SLE (JSLE) with relevance for clinical trial stratification22. Additionally, taurine metabolism has been implicated in SLE pathogenesis by enhancing pDC-mediated type I IFN production23. A multi-omics analysis has pinpointed high ApoB:ApoA1 ratios as potential biomarkers indicative of heightened cardiometabolic risk and worse clinical outcomes in JSLE24. Another multi-omics study presented an imbalance in lipid metabolism, especially in sphingolipid metabolism, accompanied with dysregulated apolipoproteins, which are seen to contribute to SLE disease activity25. These studies demonstrate the great potential of metabolomics in unraveling the complexities of of systemic lupus erythematosus. However, in certain subgroups of SLE, such as those with elevated serum ferritin levels, challenges remain in identifying metabolic abnormalities associated with disease progression, activity, and clinical manifestations. Further research is warranted to elucidate the underlying etiologies and enhance our understanding of the condition.

In this study, we explore the changes of serum metabolic profiles in SLE patients with elevated serum ferritin. By identifying key metabolic pathways and biomarkers, we can gain a better understanding of the complexity of the disease and serve as a foundation for the development of personalized treatment strategies for this subgroup.

Method

Patients

We recruited 52 SLE patients who visited Nanjing Drum Tower Hospital between May 2022 and March 2023. All participants met the 1997 revised diagnostic criteria for SLE set by the American College of Rheumatology (ACR)26. Those with a history of active malignancy, New York Heart Association functional class III-IV, severe infections, liver failure, end-stage renal disease, Adult Still disease, or positive test results for HIV and other pathogenic microorganisms were excluded. Informed consent was obtained from all SLE patients. Demographic and clinical profiles of the patients are summarized in Table 1. Based on serum ferritin level, the patients were divided into 2 groups: the non-high ferritin group (F1 Group, serum ferritin ≤ 322 ng/ml for male or serum ferritin ≤ 291 ng/ml for female) and the high ferritin group (F2 Group, serum ferritin > 322 ng/ml for male or serum ferritin > 291 ng/ml for female). The study was approved by the ethics committee of Nanjing Drum Tower Hospital.

Table 1 Clinical characteristics of participants.
Full size table

Sample preparation

We collected peripheral venous blood samples from the participants after overnight fasting using yellow vacuum blood collection tubes. Around 300 μl of serum was preserved in EP tubes and stored at − 80 °C for further analysis. Thaw the samples at 4 °C, vortex well, take 100 μl of each sample into an EP tube, add 400 μl pre-cooled pure methanol, vortex well, ultrasonic for 20 min in an ice bath, static for 1 h at − 20 °C, centrifuge at 16,000 g for 20 min at 4 °C, and then collect the supernatant to be dried up in a high-speed vacuum concentration centrifuge. For mass spectrometry, 100 μl of methanol–water solution (1:1, v/v) is added for re-dissolution, and the supernatant is centrifuged at 20,000 g for 15 min at 4 °C.

LC–MS/MS analysis

Chromatographic separation

Samples were placed in an autosampler at 4 °C throughout the analysis using a SHIMADZU-LC30 ultra-high performance liquid chromatography (UHPLC) system and an ACQUITY UPLC® HSS T3 (2.1 × 100 mm, 1.8 µm) (Waters, Milford, MA, USA) column. The injection volume was 4 μl, the column temperature was 40 °C and the flow rate was 0.3 mL/min; the chromatographic mobile phases were A: 0.1% formic acid aqueous solution, and B: acetonitrile; and the chromatographic gradient elution program was as follows: 0–2 min, 0 B; 2–6 min, B varied linearly from 0 to 48%; 6–10 min, B varied linearly from 48 to 100%; 10–12 min, B was maintained at 100%; 12–12.1 min, B changed linearly from 100 to 0%, and 12.1-15 min, B was maintained at 0%.

Mass spectrometry acquisition

Each sample was detected by electrospray ionization (ESI) in positive (+) and negative (−) modes, respectively. After the samples were separated by UPLC, we performed mass spectrometry analysis with a QE Plus mass spectrometer (Thermo Scientific), and used a HESI source for ionization with the following ionization conditions: Spray Voltage: 3.8kv (+) and 3.2kv (−); Capillary Temperature: 320 (±); Sheath Gas: 30 (±); Aux Gas: 5 (±) Probe Heater Temp: 350 (±); S-Lens RF Level: 50. The mass spectrometry acquisition settings were as follows:

MS acquisition time: 15 min, parent ion scan range: 70–1050 m/z, primary MS resolution: 70,000 @ m/z 200, AGC target: 3e6, primary Maximum IT: 100 ms. Secondary mass spectrometry analysis was acquired as follows: the 10 highest-intensity parent ions was triggered after each full scan (MS2 scan), secondary mass resolution: 17,500 @ m/z 200, AGC target: 1e5, secondary Maximum IT: 50 ms, MS2 Activation Type: HCD, Isolation window: 2 m/z, normalized collision energy (Stepped): 20, 30, 40.

Data preprocessing

The raw data were processed by MSDIAL software for peak alignment, retention time correction and peak area extraction. We used precise mass number matching (mass tolerance < 10 ppm) and secondary spectrum matching (mass tolerance < 0.01 Da) to identify the metabolite structures, and matched the metabolites with Human Metabolome Database (HMDB) and self-built metabolite standard library (BP-DB).

For the extracted data, ion peaks with > 50% missing values were removed from the group and not involved in the subsequent statistical analysis; the positive and negative ion data were normalized by the total peak area, and the positive and negative ion peaks were integrated and applied to the Python software for pattern recognition, and then pre-processed by Unit variance scaling (UV) for the subsequent data analysis.

Data analysis

For the clinical baseline data, R Studio (version 4.3.2) was used for statistical analysis: the Shapiro–wilk test was used for normality, for continuous variables with normal distribution, mean and standard deviation were used, and t-test was used for comparison between groups; for continuous variables with non-normal distribution, median and quartile were used, and Mann–Whitney U-test was used for comparison between groups; and for categorical variables, number and percentage were used, and Chi-square test or Fisher exact test was used for comparison between groups. Spearman’s rank correlation analysis was used for correlation analysis, and GraphPad Prism 9 was used to draw the heat map of correlation analysis.

For the comparison of metabolite expression differences, t-test was used for univariate analysis, and fold change (FC) was performed to compare the relative expression; PCA and OPLS-DA were used for multivariate analysis; OPLS-DA is a supervised discriminant analysis statistical method based on PCA, which utilizes partial least squares regression to establish the relationship between metabolite expression and sample category, and to realize the prediction of sample category. Variable important in projection (VIP) is the variable weight value of the OPLS-DA model variables, which can be used to measure the influence strength and explanatory power of the accumulation differences of each metabolite on the categorical discrimination of each group of samples. The permutation test was used to assess whether the OPLS-DA model was overfitted. In this study, we determined the differential metabolites based on P < 0.05 and VIP > 1. Also, ROC analysis was performed to assess the diagnostic value of metabolites and the KEGG enrichment analysis of differential metabolites was performed to identify the perturbed biological pathways. KEGG enrichment analyses were carried out with the Fisher’s exact test, and FDR correction for multiple testing was performed. Enriched KEGG pathways were nominally statistically significant at the P < 0.05 level. In this study, P < 0.05 was considered statistically significant.

Result

Clinical characteristics of participations

As depicted in Table 1, the average age of the participants in the two groups was 38.69 ± 14.24 versus 46.75 ± 13.87 years (F1 vs. F2). Compared to the F1 group, the F2 group showed lower levels of C3 and C4 and higher levels of PCT, IL-6, urinary protein/creatinine and immunoglobulin, although not entirely statistically significant. Additionally, the F2 group demonstrated significantly lower serum albumin levels, CD4+ lymphocyte counts, as well as higher levels of ALT and AST. Both groups displayed average hemoglobin (HB) levels below the lower limit of the normal range (female: 115–150 g/L, male: 130–175 g/L) and the F2 group exhibited a greater extent of hematologic involvement than the F1 group. There is no significant difference in SLEDAI and anti-dsDNA antibodies between the two groups. Among the 32 SLE patients with confirmed hematological system involvement, those with hematologic involvement exhibited higher serum ferritin levels (Table 2).

Table 2 Serum ferritin levels and hemocytes counts by hematologic involvement in the SLE patients.
Full size table

Untargeted metabolic alterations in SLE patients with elevated ferritin levels

The PCA analysis (Fig. 1A) showed the quality control (QC) samples were close clustering in both positive and negative mixing modes, indicating excellent reproducibility of the experiment. The OPLS-DA score plot (Fig. 1B) effectively distinguished SLE patients with high ferritin levels from those with non-high ferritin levels, highlighting differences in serum metabolite profiles between the two groups. The permutation test verified that the OPLS-DA model was not overfit, with R2Y = 0.914cum and Q2 = 0.233cum. 128 differential metabolites were identified by t test and OPLS-DA (P < 0.05, VIP > 1). (Fig. 2A–C). These differential metabolites predominantly belonged to categories such as lipids and lipid-like molecules, organic acids and derivatives, organoheterocyclic compounds, benzenoids, organic oxygen compounds (Fig. 2D). Pearson correlation analysis revealed the close associations between most of these differential metabolites (Fig. 2E). Based on further ROC analysis, we selected six significantly differential metabolites with AUC > 0.7 and VIP > 2.5 (Table 3): N-Lactoylvaline, Succinic acid, 3-Hydroxyisovalerylcarnitine, 1-Carboxyethylphenylalanine, N-Lactoylleucine and N-lactoyl-Tyrosine.

Fig. 1
figure 1

Identification of serum metabolic profiles in SLE patients between the F1 group and the F2 group. (A) Unsupervised PCA model; (B) Supervised OPLS-DA model.

Full size image
Fig. 2
figure 2

Identification of differential metabolites in SLE patients between the F1 group and the F2 group. (A) Volcano Plot of metabolites between the two groups. The red dots represent up-regulated metabolites, the blue dots represent down-regulated metabolites. (B) Scatter plot of the mean expression values of metabolites. (C) Hierarchical clustering results of differential metabolites. (D) Pie chart of HMBD Super Class classification of differential metabolites. (E) Heatmap for correlation analysis of differential metabolites.

Full size image
Table 3 Significantly different metabolites between the F1 and F2 groups.
Full size table

Pathways analysis of the differential metabolites

We performed KEGG pathway enrichment analysis on the differential metabolites and identified significant enrichment in 10 second class pathways and 18 third class pathways (Fig. 3A–B), amino acid metabolism was the most significantly enriched second KEGG pathway, and arginine biosynthesis emerged as the most prominent pathway (Fig. 3C–D). In another, we studied the overall metabolic changes using the abundance of differential metabolites (Fig. 3E), found the F2 group exhibited upward trends in those pathways such as tryptophan and carbon metabolism and downward trends in those pathways such as aminoacyl-tRNA biosynthesis, pyrimidine metabolism and protein digestion and absorption.

Fig. 3
figure 3

Pathway analysis. (A) Bubble map of the enriched Significant classify KEGG pathway. (B) Bubble map of the enriched Secondary classify KEGG pathway. Environmental Information Processing: E, Genetic Information Processing: G, Human Diseases: H, Metabolism: M, Organismal Systems: O. (C) Pathway Impact plot of differential metabolites calculated by Out degree centrality method. (D) Pathway Impact plot of differential metabolites calculated by Betweenness centrality method. (E) Differential abundance score plot for differential metabolites.

Full size image

Correlation analysis of clinical features and significantly different metabolites

We conducted Spearman’s rank correlation analysis among the clinical features and the six significantly different metabolites (Fig. 4A–B). In addition to the positive correlations with the six different metabolites, serum ferritin was positively correlated with age, ESR, CRP, ALT, AST and IgG, while negatively correlated with albumin, total lymphocyte count, CD3+ T cell count, CD3+CD4+ T cell count and NK cell count. The six significantly different metabolite showed significantly positive correlations each other, except for N-lactoyl-Tyrosine with succinate acid. They showed significant correlations with specific clinical features, for instance, 1-Carboxyethylphenylalanine and 3-Hydroxyisovalerylcarnitine were positively correlated with PCT, IL-6, ESR and CRP; N-Lactoyl-Tyrosine, 1-Carboxyethylphenylalanine and 3-Hydroxyisovalerylcarnitine were negatively correlated with albumin levels. All of these metabolites, except for N-Lactoyl-Tyrosine, showed a significant positive correlation with ALT or AST. Moreover, 3-Hydroxyisovalerylcarnitine exhibited a positive correlation with serum BUN and Scr levels. These findings highlight the associations between differential metabolites and clinical features, suggesting their potential roles in the progression of SLE with elevated serum ferritin.

Fig. 4
figure 4

Heat maps of correlation analysis of clinical characteristics with significantly differential metabolites. (A) Heat maps of relative coefficient of the correlation analysis; (B) Heat maps of P value of the correlation analysis. ***P < 0.001; **P < 0.01; *P < 0.05; NS, P ≥ 0.05.

Full size image

Discussions

It is noteworthy that approximately 15–30% of SLE patients develop cytokine storm-mediated hyperinflammation, distinct from classic autoantibody-driven pathology27,28. These cases feature inflammasome activation, Toll-Like receptor pathways stimulation, and metabolic reprogramming29,30. Ferritin, which can be produced in large quantities in response to cytokine stimulation, is now receiving increasing attention for its role in SLE12,16. Our study reveals that SLE patients with hyperferritinemia demonstrate distinct metabolic perturbations, particularly in arginine biosynthesis, lipid metabolism pathways and organic acid metabolism pathways, which may underlie ferritin’s potential pathogenic role in SLE beyond being an acute-phase reactant. However, the specific mechanism still needs further exploration.

Among the metabolic pathways analyzed, arginine biosynthesis was found to be the most dominant pathway, in which the high-ferritin group (F2 Group) presented an increase of citrulline. Previous study in MRL mice has shown elevated citrulline levels are associated with preferential propagation and overproduction of IgG in γ-committed B lymphocytes31, which to some extent is proved by our result that a positive correlation between citrulline and IgG (Supplementary Fig. 1).

Previously, most phosphatidylcholines are deregulated in SLE patients32,33, and in our study, three downregulated phosphatidylcholines were identified in the F2 group: LysoPC (18:0), LysoPC (16:0), and LysoPC (P-16:0), which correlated negatively with serum ferritin levels (Supplementary Fig. 2),which aligns with emerging evidence that phospholipid metabolism is crucial to SLE immunoregulation. LysoPCs are precursors of anti-inflammatory mediators and modulate macrophage polarization34,35,36,37,38. Their reduction in high-ferritin SLE patients may favor a pro-inflammatory M1 phenotype, exacerbating tissue damage.

Sixteen differentially expressed fatty acids were identified, three of which were derivatives of the polyunsaturated omega-6-long-chain fatty acid lineolic acid, namely 8,12-Octadecadienoic acid, alpha-Eleostearic acid, and Avenoleic acid, they were all shown to be down-regulated in the F2 group. As omega-6 fatty acids may be involved in immunomodulation and inflammation regulation, which are precursors of both pro-inflammatory and pro-resolving mediators39,40,41,42,43, their depletion may skew the immune response toward chronic inflammation.

Of the 34 differentially expressed organic acids and their derivatives identified, 25 were expressed at elevated levels in the F2 group, and five of these metabolites were significantly different between the two groups, including Succinic acid, N-Lactoylvaline, N-Lactoylleucine and N-lactoyl-Tyrosine. Accumulation of succinate may play a role in SLE patients with elevated serum ferritin, by providing B cell help and promoting pro-inflammatory cytokine production44,45. N-Lactoylvaline, N-Lactoylleucine and N-lactoyl-Tyrosine are classified as N-lactoyl amino acids, synthesized intracellularly via CNDP2-mediated reversed protein hydrolysis of lactate and amino acids, and the plasma levels of which depends on lactate and amino acids46. Additionally, there was a significant increase in the phenylalanine derivative 1-Carboxyethylphenylalanine levels, strongly correlated with the levels of these three N-lactoyl amino acids. However, the corresponding serum valine, leucine, tyrosine, and phenylalanine levels were not significantly altered in these patients (Supplementary Table 1). Since all N-lactate amino acids are synthesized intracellularly, our findings provide a new perspective for studying SLE patients with elevated serum ferritin, suggesting a potential focus on alterations in lactate metabolism and intracellular amino acid metabolism.

3-Hydroxyisovalerylcarnitine increased in the high-ferritin SLE patients significantly, which derived from the leucine alternative catabolic pathway and has been identified as an indicator of biotin deficiency47,48, as well as associated with worse clinical scores in heart failure with preserved ejection fraction49, might also play a role in the pathogenesis of SLE patients with elevated serum ferritin.

In addition to the arginine biosynthesis pathway, our findings suggest correlations between elevated serum ferritin levels in SLE patients and tryptophan metabolism, carbon metabolism, aminoacyl-tRNA biosynthesis and pyrimidine metabolism. Altered tryptophan catabolism contributes to autoimmunity in lupus-susceptible mice and may play an important role in SLE development50,51,52. Carbon metabolism is responsible for redox equilibrium and provides substrates for acetylation and methylation, and regulate epigenetic modifications of immune cells such as macrophage53. Alteration of one-carbon metabolic pathway influences epigenetic of MHC2TA and RFC1, thus contributing to phenotypic heterogeneity of SLE54. Previous study shows dysregulated aminoacyl-tRNA biosynthesis in SLE patients51. Pyrimidine metabolism is a key driver of effector functions in CD4+T cells and Th1+ cells55,56. Together, the disturbances in those metabolism pathways were likely related to the underlying processes of high-ferritin SLE patients.

Limitations of this study are in the following: the sample size is small and needs for more robust evidence. Serum metabolite profiles alone cannot fully represent the body’s metabolic status, emphasizing the need for future investigations at tissue and cellular levels.

In conclusion, we find that SLE patients with elevated serum ferritin exhibit more hematologic involvement, and our results present a unique serum metabolic profile in patients with elevated serum ferritin in SLE, particularly disrupted arginine biosynthesis, lipid pathways and organic acid metabolism pathways. These findings provide valuable insights to further understand the role of elevated ferritin in SLE.