Background/Purpose: Inflammatory arthritis occurs in up to 90% of patients and is a main cause of SLE work-related disability. SLE arthritis is difficult to evaluate because of the evanescent nature of the symptoms and limitations of physical exams and imaging. Although advanced imaging modalities have been used to characterize SLE arthritis insights into the pathophysiology and predictive biomarkers to address optimal surveillance, and clinical trial design are lacking. We leveraged serum collected at the time of arthritis evaluations to agnostically survey potential soluble mediators of arthritis in SLE patients using the Olink platform.

Methods: Sera from 15 subjects with clinical lupus arthritis were evaluated. All samples were collected in proximity to clinical and imaging evaluations. Swollen and tender joint counts were performed, MSK US used multiplanar grey-scale (GS) to evaluate synovitis/effusion and power Doppler (PD) for synovial hypertrophy/inflammation, and was scored by OMERACT definitions to generate a composite PDGS score on a scale of 0-3 (2-3 representing moderate/severe arthritis) at bilateral PIP joints. Serum proteins were assessed by the Olink Explore HT assay to evaluate >5400 proteins. All included samples passed standard Olink quality control metrics. K-means clustering agnostically defined groups of interest. Principal component analysis (PCA) was used for data visualization and a t-test was employed to compare clusters with candidate proteins considered at adjusted p-value by Benjamin-Hochberg method < 0.01 and log2fc |1.5|.

Results: Clinical and imaging variables for the 15 patients are described in Table 1. Olink data identified 2 clusters of interest, cluster A (5 subjects) and cluster B (10 subjects) (Fig 1A). Differential protein expression revealed 238 proteins significantly higher in cluster A vs cluster B (Fig 1B). Gene ontology (GO) analysis showed significant enrichment of pathways related to innate immune signaling in cluster A (Figure 1C). The top 10 proteins higher in cluster A compared to cluster B are shown in Figure 2. Clinically, patients in cluster A had significantly higher levels of active synovitis measured by GS compared to those in cluster B (PDGS >/=2) (p=0.05) (Table 1) despite having similar tender/swollen joint counts on physical exam (Table 1).

Conclusion: In this serum proteomic evaluation, an unsupervised machine learning algorithm was able to agnostically identify patients with more severe arthritis as measured by MSK US but not physical exam. These data suggest significant aberrations in the serum proteome likely contribute to arthritis severity in SLE.

Table 1. Demographic and clinical characteristics of the lupus arthritis patients of interest.

Figure 1: (A) PCA showing the two clusters identified by k-means clustering. (B) Volcano plot showing the differentially abundant proteins between cluster A and cluster B with adjusted p-value < 0.01 and log2FC |1.5|. (C) Gene ontology pathway enrichment analysis of the 238 proteins that were significantly higher in cluster A compared to cluster B.

Figure 2: Boxplots showing the top 10 differentially expressed proteins by log2FC, all met adjusted pvalues < 0.01.

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/an-agnostic-evaluation-of-serum-proteins-discriminates-the-severity-of-ultrasound-arthritis-in-sle-patients/