Background: In systemic lupus erythematosus (SLE), clinical remission does not necessarily reflect biological quiescence. A substantial proportion of patients with low disease activity subsequently experience flares, suggesting the presence of a subclinical molecular state preceding clinical worsening. High-dimensional proteomic profiling offers an opportunity to uncover latent inflammatory and tissue-damage pathways that anticipate adverse outcomes.

Objectives: To identify molecular endotypes in SLE integrating serum proteomics and clinical data, and to define a baseline proteomic signature associated with subsequent disease worsening in clinically stable patients.

Methods: A longitudinal observational study was conducted in a cohort of 90 patients with SLE (ACR criteria). Serum samples were analysed using Proximity Extension Assay technology (Olink®, Inflammation and Organ Damage panels). Patients with low baseline disease activity were stratified according to 6-month outcome into persistently stable and worsening disease (ΔSLEDAI ≥ 4). A third group included patients in clinical flare at baseline. Unsupervised clustering, principal component analysis and machine-learning approaches were applied to identify molecular profiles associated with clinical profiles and outcome. Differentially expressed proteins were functionally annotated across inflammatory, chemokine, endothelial, angiogenic and tissue-remodelling pathways. The robustness of these inflammatory proteomic profiles was confirmed in an independent SLE cohort (n=90) from the European multicentre cross-sectional PRECISESADS IMI study (NCT02890121).

Results: Unsupervised analysis revealed a high-risk molecular endotype characterized by coordinated activation of inflammatory, chemotactic, endothelial and tissue-remodelling pathways. This profile comprised 90 differentially expressed proteins, including systemic cytokines (TNF, IL-6, IL-17A/C, IL-18, IL-33, IFN-γ, LIF), chemokines driving leukocyte trafficking (IL-8, MCP-1/2/3/4, CCL3/4/11/19/20/23/25, CXCL1/6/9/10/11, CX3CL1), mediators of endothelial activation and angiogenesis (VEGFA, PGF, NOS3, CDCP1, OPG, uPA), and markers of tissue remodelling and organ damage (MMP-1, FGF-5, FGF-21, GDNF, ARTN, NRTN, AGR2, RARRES1, KIM-1). Clinically, this endotype was associated with a more severe phenotype. Patients in cluster 2 were older (47.2 ± 11.8 vs 41.9 ± 10.6 years, p = 0.038) and displayed significantly higher disease activity, as measured by SLEDAI (7.14 ± 7.51 vs 2.83 ± 3.09, p < 0.001). They also exhibited elevated acute-phase reactants (abnormal CRP/ESR: 62.1% vs 14.8%, p < 0.001), a markedly higher prevalence of subclinical atherosclerosis (atheroma plaques in 34.5% vs 13.1%, p = 0.045), and a more adverse cardiovascular profile, with a higher atherogenic index (4.12 ± 1.88 vs 3.25 ± 0.76, p = 0.01). A trend towards increased anti-dsDNA positivity and renal involvement was also observed. The inflammatory endotype proved reproducible in the independent PRECISESADS cohort, where more than a 40% of differentially expressed proteins remained consistently deregulated—confirming the robustness of the underlying molecular profile across cohorts. Importantly, its clinical relevance was validated in parallel, as this cohort revealed a specific patient subgroup with a matching clinical profile: high disease activity (SLEDAI, p=0.0315), renal involvement, and enhanced cardiovascular risk. Machine-learning modelling in the discovery cohort identified a compact baseline protein signature with high predictive performance for subsequent disease worsening (AUC 0.93). The most informative variables (VIP proteins) included OPG, LIF, CCL25 and FGF-5, which converge on pathways of vascular injury and remodelling (OPG), cytokine-driven tissue stress and regeneration (LIF), immune cell trafficking to inflamed tissues (CCL25) and growth factor–mediated tissue remodelling (FGF-5), all previously implicated in chronic inflammation and organ damage in SLE. Among patients with low baseline SLEDAI scores, those who subsequently experienced disease worsening displayed a baseline proteomic profile overlapping with that of patients in clinical flare and distinct from persistently stable patients. Baseline levels of VIP panel proteins—OPG, LIF, CCL25, and FGF-5—differed significantly between patients with subsequent worsening or flare versus those remaining stable. These levels proved comparable between worsening (during follow-up) and flare groups, suggesting a shared subclinical state of inflammation and organ damage that precedes overt clinical deterioration.

Conclusions: This study unveils a high-risk inflammatory endotype in systemic lupus erythematosus, characterized by subclinical molecular activation that persists despite apparent clinical quiescence and foreshadows disease flares. Its reproducibility across independent cohorts and tight linkage to severe phenotypes—encompassing heightened disease activity, renal involvement, and cardiovascular vulnerability—underscore the value of proteomic profiling for identifying at-risk patients and enabling precision medicine strategies beyond traditional clinical metrics.

REFERENCES: NIL.

Acknowledgments: NIL.

Disclosure of Interests: None declared.