To Örebro University

Background: Systemic lupus erythematosus (SLE) is characterised by the production of a multitude of autoantibodies (Abs), yet the antigen specificities remain elusive for most of these Abs. The diagnosis of SLE is supported by the presence of certain Abs; however, SLE-specific Abs currently used for diagnosis or surveillance, e.g., anti-dsDNA and anti-Smith, lack sensitivity.

Objectives: To perform a broad explorative screen of IgG and IgA antibodies against autoantigen specificities in patients with SLE vs healthy controls (HC) to gain insight into disease pathogenesis and identify novel Abs that could potentially serve as diagnostic biomarkers.

Methods: We analysed plasma samples from 289 patients with SLE and 196 age- and sex-matched HC from the European PRECISESADS project (NTC02890121). Samples were screened for IgG and IgA seroreactivity against a panel of >1,600 protein autoantigens using KREX-based i-Ome arrays (Sengenics). Differential expression analysis was performed with the limma R package adjusting for e.g., polyspecific antibody reactivity. The level of significance for differentially expressed Abs (DEAbs) was set at p<0.05 and |fold change (FC)| >1.5. Machine learning (ML) was applied to identify the ten most discriminative Abs. ROC analysis was carried out for each individual DEAb and for the combined top ten ML-selected DEAb model, producing discriminative performance metrics including sensitivity, specificity, area under the curve (AUC), and accuracy. Functional analysis was performed with gene ontology (GO) enrichment and signalling pathway impact analysis using the corresponding antigens of the DEAbs (p<0.1; |FC| >1.1).

Results: DEAb analysis revealed 14 IgG and 2 IgA upregulated DEAbs in patients with SLE vs HC. Of IgG DEAbs, anti-LIN28A (sen=0.77, spe=0.69, AUC=0.80), anti-HBGB2 (sen=0.65, spe=0.81, AUC=0.79), anti-HMG20B (sen=0.75, spe=0.71, AUC=0.70), and anti-NRF1 (sen=0.63, spe=0.80, AUC=0.77) demonstrated the best discriminative capacity. ML identified the top ten IgG Abs distinguishing SLE from HC (SSB, TROVE2, LIN28A, TFCP2, HNRNPA2B1, KLF12, RAD51, CCNB1, EEF1D, PSMD4) with a good whole-panel performance (sen=0.84, spe=0.72, AUC=0.78). Functional analysis of IgG DEAbs revealed 2 GO-term enrichments at the cellular component level: extracellular region and cell surface, suggesting that autoantigens eliciting IgG in SLE are accessible at the cell surface or released by cell lysis and may be altered in an immunogenic manner by SLE-associated pathophysiological processes. Of IgA DEAbs, anti-LIN28A (sen=0.65, spe=0.83, AUC=0.80) and anti-SSB (sen=0.64, spe=0.82, AUC=0.78) demonstrated the best discriminative capacity. ML identified the top ten IgA Abs discriminating SLE from HC (APOBEC3G, SUB1, LIN28A, SSB, TROVE2, PCBP2, TFCP2, PRKRA, ZMAT3, UBE2I) with reasonable whole-panel performance (sen=0.84, spe=0.69, AUC=0.76). Functional analysis of IgA DEAbs revealed 17 GO-term enrichments related to DNA-binding transcription repressor activity, chromatin binding, transcription factor binding, and regulation of transcription by RNA polymerase II, suggesting that autoantigens eliciting IgA in SLE are enriched for nucleic acid binding.

Conclusion: This study corroborated both classical IgG Ab specificities (e.g., anti-SSB, anti-TROVE2) and previously reported specificities (anti-LIN28A, anti-HBGB2, anti-HMG20B, anti-HNRNPA2B1)1 and identified novel IgG (anti-NRF1, anti-PCBP2, anti-TGIF2, HMGN5, anti-SUB1, anti-PSIP1, anti-CCNB1) and novel IgA (anti-LIN28A) Abs described for the first time in SLE, with robust accuracy in distinguishing SLE from HC. Importantly, these novel Abs outperformed the accuracy of currently used Abs, e.g., anti-dsDNA (sen=0.36, spe=0.95, AUC=0.61)1. ML further enhanced the diagnostic accuracy, heralding promise for the early diagnosis of SLE. Functional analysis revealed distinct mechanisms with suggested differential roles in the production of IgG and IgA Abs in SLE.