Serum autoantibody profiling in systemic lupus erythematosus reveals associations with low disease activity and remission
- Laura E. Eades1
- William A. Figgett2
- Katie Lennard3
- Hossein Valipour Kahrood4
- Alberta Y. Hoi1
- Rangi Kandane-Rathnayake1
- Eric F. Morand1
- Artemis Papadaki5
- Fabien B. Vincent1*corresponding author
- 1Centre for Inflammatory Diseases, School of Clinical Sciences at Monash Health, Monash University, Clayton, VIC, Australia
- 2Data Science Platform, Garvan Institute of Medical Research, Darlinghurst, NSW, Australia
- 3Proteomics Bioinformatics Department, Standard Biotools Inc, South San Francisco, CA, United States
- 4Monash Proteomics and Metabolomics Platform, Monash Biomedicine Discovery Institute and Department of Biochemistry and Molecular Biology, Monash University, Clayton, VIC, Australia
- 5National Heart and Lung Institute, Faculty of Medicine, Imperial College London, South Kensington, London, United Kingdom
Abstract
Knowledge of autoantibody specificities and their clinical associations in systemic lupus erythematosus (SLE) is largely limited to panels testing a finite number of autoantibodies. We aimed to explore high-plex serum autoantibody profiles in SLE and examine associations with socio-demographics, phenotypes and outcomes using an unbiased approach. In this exploratory, biomarker-discovery study, 1,613 serum IgG autoantibodies were quantified using an i-Ome® Discovery Protein Array in 60 patients with SLE. The array was customised to detect antibodies against three interferons. Autoantibody clusters were defined using unsupervised clustering and associations between clusters and clinical features were assessed using multinomial logistic regression. Associations between individual autoantibodies and clinical features were assessed using regression models and supervised learning. Four patient clusters were identified; Cluster 4, characterised by anti-La and anti-Ro60 autoantibodies, included patients significantly more likely to be of Asian ethnicity and with higher ESR levels than those in Cluster 1. In analysis of individual autoantibody specificities, anti-LRR binding FLII-interacting protein 2 (anti-LRRFIP2) was associated with remission and lupus low disease activity state (LLDAS). Anti-gephyrin (GPHN) was associated with CNS lupus. Anti-adaptor protein-2 associated kinase 1 (anti-AAK1), anti-ribosomal protein S6 kinase B1 (anti-RPS6KB1) and anti-ribosomal protein S6 kinase B2 (anti-RSP6KB2) were negatively associated with mucocutaneous lupus. Unsupervised clustering of autoantibody profiles revealed a patient subset with anti-La and anti-Ro60 profiles, associated with Asian ethnicity and higher ESR. Anti-LRRFIP2 was associated with remission and LLDAS, and anti-GPHN, previously described in autoimmune CNS disease, was associated with CNS lupus. While these exploratory associations warrant validation in larger independent cohorts, the identification of biomarker candidates highlights the potential value of comprehensive IgG autoantibody profiling for biomarker discovery in SLE.
Keywords
What is known on this topic
Routinely measured autoantibodies, including antinuclear antibodies and autoantibodies to extractable nuclear antigens, play a role in the diagnosis of systemic lupus erythematosus (SLE) and can predict disease phenotype and outcomes. Very few studies, however, have explored high-plex serum autoantibody profiling in patients with SLE, and no studies have investigated associations between detailed longitudinal clinical phenotypes and unbiased serum autoantibody profiles.
What this study adds
This study generated high-plex serum IgG autoantibody profiles for 60 patients with SLE, and described associations between autoantibody profiles and detailed longitudinal clinical phenotypes. We identified several novel autoantibody biomarker candidates, including anti-LRR binding FLII interacting protein 2 (anti-LRRFIP2), associated with remission and lupus low disease activity state, and anti-gephyrin (GPHN), associated with CNS lupus, which could potentially serve as useful biomarkers in the future.
How this study might affect research, practice or policy
The identification of autoantibody biomarker candidates associated with clinical phenotypes and outcomes in patients with SLE may contribute novel biomarkers for use in diagnosis and prognostication, and/or as therapeutic targets.
Introduction
Systemic lupus erythematosus (SLE) is a multi-system autoimmune disease, predominantly affecting women of childbearing age (1). SLE is more prevalent and severe in people of non-European ethnicity, particularly those of Asian, African and First Nations ethnicity (2). The presence of circulating autoantibodies is the immunological hallmark of SLE (3). Conventionally measured autoantibody panels, such as antinuclear antibodies (ANA), antibodies to double-stranded (ds) DNA and extractable nuclear antigen (ENA) autoantibodies, are widely used as diagnostic biomarkers in clinical practice and for classification criteria in SLE (3, 4). Some of these autoantibodies can predict organ involvement and prognosis, and some are used to monitor disease activity (5, 6).
Non-routinely measured autoantibodies targeting components of pathogenic pathways operative in SLE are increasingly garnering interest as potential clinically informative biomarkers. Interferon (IFN) pathways play a pathogenic role in SLE, with evidence for both Type I and Type II IFN pathways being associated with increased disease activity and increased rates of lupus nephritis (LN) (7, 8). Furthermore, the Type I IFN receptor is now being used successfully as a therapeutic target in SLE (9). Autoantibodies targeting various components of the IFN pathways have been identified in patients with SLE (10, 11), with some associated with increased disease activity (8), and others associated with quiescent disease (12), although associations with more detailed clinical phenotypes have not been described. Such observations raise the possibility of undiscovered autoreactivities with the potential to be useful clinical biomarkers and/or reveal therapeutic targets in SLE.
In clinical practice, autoantibody measurement is limited to a small number of autoantibodies (13). The emergence of high-throughput, high-plex autoantibody profiling platforms, enabling simultaneous measurement of thousands of autoantibodies, offers new avenues for biomarker discovery in SLE (12). Wang et al. reported increased abundance of autoantibodies targeting proteins known to play a role in SLE pathogenesis in patients with SLE compared to healthy controls (HC) (12). Lewis et al. grouped SLE patients into clusters based on unbiased autoantibody profiles, and noted differences in SLE organ involvement between clusters (14). Recently, Lindblom et al. found a number of IgG and IgA autoantibodies which were more highly abundant in SLE patients compared to HC (15). Some of these autoantibodies, including anti-LIN28A IgG, anti-HMGN5 IgG, and both isotypes for anti-IRF5 and anti-TGIF1, were positively associated with high disease activity, and negatively associated with lupus low disease activity state (LLDAS) (15). To date, however, no studies have examined relationships between high-plex serum autoantibody profiles and validated longitudinal clinical outcome measures, or used supervised machine learning to determine whether autoantibody profiles may be able to predict clinical outcomes.
Here, we aimed to identify serum autoantibody profiles in patients with SLE and to examine their associations with socio-demographics, SLE phenotypes and outcomes, using an unbiased high-plex autoantibody profiling approach, in a well-characterised multi-ethnic SLE cohort. In addition, we aimed to explore associations between autoantibodies targeting components of IFN pathways and clinical features using a targeted approach.
Materials & methods
The patient cohort and methodological approach used in this study have been described previously in our published study of serum IgA autoantibody profiling in SLE (16).
Study participants
This retrospective, single-centre, hypothesis-generating exploratory study included the selection of 60 patients with SLE, enrolled in the Australian Lupus Registry and Biobank (ALRB) (17), and recruited at the Monash Health Lupus Clinic (Melbourne, Australia) between July 2019 and November 2020. All participants enrolled in the ALRB fulfil SLE classification criteria, either the 1997 American College of Rheumatology (ACR) revised criteria or the Systemic Lupus International Collaborating Clinics (SLICC) criteria (18, 19). The participant selection strategy included equal proportions of individuals of self-identified European and Asian ethnicity, balanced in terms of age, sex and disease activity, according to SLE Disease Activity Index-2000 (SLEDAI-2K) (20).
All participants provided informed written consent to their stored clinical data and biosamples being used in research related to SLE. Ethics approval was granted by the Monash Health Human Research Ethics Committee (Ref: 13019A). This study was performed in accordance with the National Statement on Ethical Conduct in Human Research (2007).
Data collection & clinical assessment
Clinical data collection and assessment most pertinent to study analyses is summarised below; for a complete description of clinical data extracted, please see Supplementary Table 1. Baseline socio-demographics were extracted from the ALRB, including date of birth, sex, self-reported ethnicity, employment, household income level, education level, smoking status and alcohol consumption. Cross-sectional data including disease activity, flare, routine laboratory tests and medication use (prednisolone, hydroxychloroquine and/or immunosuppressants), were obtained from the ALRB clinic visit closest to the date of study serum sample collection (<30 days), or from the most recent annual visit for data pertaining to organ damage. Disease activity was determined using the SLEDAI-2K (20), with active disease defined as SLEDAI-2K > 4 as previously described (21). Renal disease activity was defined as the presence of SLEDAI-2K proteinuria. Organ damage was assessed annually using the SLICC-ACR Damage Index (SDI) (22), with presence of organ damage defined as SDI > 0. Low disease activity was defined as achieving LLDAS (23). Remission was defined according to the Definition of Remission in SLE (DORIS) score (24). A more stringent definition of remission, similar to DORIS but not allowing glucocorticoid use, was also studied. Flare was defined using the SELENA-SLEDAI flare index (25). Routine laboratory data included erythrocyte sedimentation rate (ESR) and estimated glomerular filtration rate (eGFR) on the day of study serum sample collection. Renal impairment was defined as an eGFR <60 mL/min/1.73 m2. The results of any previous renal biopsy were also extracted.
Longitudinal clinical data were retrieved from each patient visit, spanning from ALRB enrolment to the end of the study period (October 2023). Organ-specific involvement was defined as the presence of a positive SLEDAI-2K score at least once across the study period for each clinical SLEDAI-2K organ domain. Adjusted mean SLEDAI (AMS) was calculated using the methods described by Ibanez et al. (26), with active disease across the study period defined as AMS > 4, as previously described (21). Percentage time spent in LLDAS or DORIS remission across the study period were derived, and binarised to <50% vs. ≥50% of cumulative time in LLDAS or DORIS remission (27).
Biological sample collection, processing & storage
Venous blood was collected via venepuncture into a serum-separating tube. Tubes were stored at 4 °C for up to 24 h, prior to serum extraction and freezing at −80 °C for long-term storage, as previously described (28). No freeze-thaw cycles were performed.
High-plex autoantibody profiling
High-plex serum autoantibody profiling was performed by Sengenics (Kuala Lumpur, Malaysia, a subsidiary of Standard BioTools), using a customised i-Ome® Discovery Protein Array. This microarray is based on Sengenics patented KREX™ protein folding technology, enabling highly multiplexed dual detection and relative quantification of IgG and IgA autoantibodies (16). Only IgG autoantibody data were analysed in this study. The i-Ome® Discovery Protein Array consists of 1,613 full length and correctly folded antigens that represent different classes and various types of antigens selected on the basis of their involvement in the immune response, such as cancer antigens, cytokines, interleukins and transcription factors. The array was customised to include detection of autoantibodies to three interferons; IFNα1 (Uniprot ID: P01562), IFNγ (Uniprot ID: P01579) and IFNω (Uniprot ID: P05000). Array intensities are represented in relative fluorescence units (RFUs).
Statistical & bioinformatics analyses
Data analyses and visualisation were performed using R version 4.4.0 and R Studio version 2024.12.1 software. A list of R packages and functions used for individual analyses is displayed in Supplementary Table 2. A p-value of <0.05 was considered statistically significant.
Data transformation, normalisation and filtering
As there were no missing values for autoantibody abundance, no imputation was required. As part of the loess normalisation approach, log2 transformation of raw mean net intensity (NetI) values was first performed. Then, the normalizeCyclicLoess function from the limma R package (version 3.58.1), which normalises the columns of a matrix, was used to cyclically apply loess normalisation to normalise to the “average” (in silicon average) array (29), thus removing any systematic experimental variation in the dataset. Batch effects were assessed by visual inspection of loess-normalised heatmaps, with a particular attention to pooled “normal” plasma clustering, as well as by permutational multivariate analysis of variance (PERMANOVA), to assess the contribution of technical variables (most notably slide batch and assay number) on the data structure. ComBat normalisation was then performed to adjust for batch effects.
Autoantibodies retained for downstream analyses were identified by negative control filtering (NCF) of net intensity data (pre-normalisation), which excludes autoantibodies most strongly correlated with non-specific (negative) binding control spots on the array, which are therefore least likely to be of clinical significance; approximately one third of the dataset was retained for downstream use.
Heatmaps and outlier assessment
Heatmaps of scaled and centered autoantibody abundance across the patient cohort with hierarchical clustering were created. Principal component analysis (PCA) was performed to assess potential patient outliers. Outliers were determined via visual inspection of PCA plots. Heatmaps were reviewed to determine whether outlier status may be related to technical error, and clinical characteristics of outliers were reviewed to determine whether they shared similar clinical features.
Descriptive statistics
Normally and non-normally distributed continuous variables are reported as mean [standard deviation (SD)] or median [interquartile range (IQR)], respectively. Categorical variables are summarised as frequency (percentage). Where clinical data were missing for a particular clinical variable, the corresponding patient was excluded from analyses of that clinical variable.
Unsupervised clustering analysis
Poly-specific autoantibody (PSA) signal denotes the non-specific presence of many autoantibodies in an individual, but is not known to be associated with SLE or other autoimmune rheumatic diseases, and hence may confound results if not adjusted for. All samples, including those PSA-positive, were included in the clustering analysis. When clustering on the full dataset (PSA+ and PSA− samples together), PSA status accounted for a dominant source of variation in the autoantibody matrix, resulting in clusters that primarily reflected PSA reactivity, rather that broader autoantibody patterns. To reduce variation attributable to PSA status while simultaneously correcting for slide batch, ComBat was applied using a composite PSA status × slide batch variable. Prior to clustering, the normalised data were row-centered (per-autoantibody centering) and a global threshold (defined as the median of all row-centered values plus three standard deviations calculated across the same pooled distribution) was applied to the NCF-selected set as a secondary filter to select autoantibodies eligible for unsupervised clustering; autoantibodies with >5% of samples exceeding this threshold were selected for clustering analysis.
Partitioning Around Medoids (PAM) clustering was performed using pairwise Pearson correlation computed on the retained autoantibody features. Maximum silhouette width provided the initial optimal number of clusters (k) candidate, while the final choice was determined after applying prespecified bootstrap cluster stability and minimum cluster-size criteria, using fpc::clusterboot() [R package fpc v2.2-14; Hennig 2007 (30)] with a custom CBI implementing PAM on the same matrix, over 500 bootstrap resamples with replacement (k = 2–10). Cluster-level Jaccard similarity was computed between each original cluster and its best-matching bootstrap cluster per the Hennig formulation (30).
Cluster solutions producing any cluster smaller than max(5, 5% of n) samples were excluded from evaluation. k values with any cluster mean bootstrap Jaccard <0.50 were excluded as insufficiently reproducible; among the remainder, the k maximising the proportion of clusters with mean Jaccard ≥0.60 was identified, with overall mean Jaccard as tie-breaker. This approach acknowledges that at modest sample sizes, uniform high stability across all clusters is not expected, and selects the partition in which the largest fraction of subtypes are reproducibly recovered.
Per-sample stability was derived from the same bootstrap partitions: for each sample, the proportion of resamples in which it co-clustered with the majority of its original cluster-mates was computed using the same best-Jaccard label matching as the cluster-level analysis. Samples were classified as highly stable (≥0.80), stable (≥0.60 and <0.80), or unstable (<0.60), noting that these thresholds are an interpretive extension of Hennig's cluster-level conventions to the individual sample level. Per-cluster and per-sample stability metrics are shown in Supplementary Figures 1, 2 and Table 3.
Univariable and multivariable multinomial logistic regression were performed to assess associations between clusters (outcome) and clinical features (exposure). In the multinomial logistic regression models, each cluster was used in turn as the reference cluster, meaning that all clusters were directly compared to each other. Covariates adjusted for in the multivariable models included age, sex and ethnicity.
Differential abundance analysis
Following scaling and centering of the autoantibody abundance matrix (not batch-corrected for PSA signal), differential abundance analysis was performed using multivariable linear regression, to examine relationships between clinical features (exposure) and autoantibodies (outcome). Two sets of multivariable models were performed; one adjusting for PSA signal only (autoantibody abundance ∼ clinical feature + PSA), and one adjusting for PSA signal, age, sex and ethnicity (autoantibody abundance ∼ clinical feature + PSA + age + sex + ethnicity). The Benjamini–Hochberg method was used to adjust for multiple testing, and adjusted p-values were used to determine statistical significance, with only these values reported. For all multivariable models, a minimum of n = 12 patients were permitted in a subgroup, with the exception of SLEDAI-2K organ specific involvement, in which subgroups as small as n = 4 were allowed for multivariable analyses adjusting for PSA signal only, to allow the assessment of rarer organ domains. For these domains, no multivariable model adjusting for socio-demographic covariates is reported.
Least absolute shrinkage and selection operator (LASSO)
LASSO was performed using the train() and trainControl() functions from the caret package in R (version 7.0-1), using the ROC metric and glmnet method. The optimal lambda (λ) value was selected based on the area under the receiver operating characteristic curve (ROC AUC), using a repeated (n = 100) k-fold (k = 6) cross-validation method within train() function. A separate LASSO model was performed to predict each categorical clinical feature of interest. Patient age, sex, ethnicity and PSA signal were included as covariates alongside all autoantibodies (exposure). Scaling and centering of autoantibody abundance as well as covariates was performed to allow comparison between all potential predictors in the model. A LASSO model was deemed of poor quality if the overall model accuracy was <0.7, and corresponding results were not reported.
To assess the robustness of LASSO-selected autoantibodies in the context of a modest sample size, we performed a repeated stratified hold-out resampling analysis for each binary clinical outcome [100 resamples; 80/20 train/test split, stratified by outcome class; within each training set, LASSO logistic regression was fitted using glmnet with 5-fold cross validation (cv.glmnet, alpha = 1, λ = lambda.1se)]. For each resample the predictive performance was evaluated on the held-out test set by calculating the area under the receiver operating characteristic curve (AUC) and balanced accuracy. Feature-selection stability was calculated as the percentage of resamples in which each autoantibody was retained (assigned non-zero coefficient). Statistical significance of model performance was assessed by permutation testing (1,000 permutations of outcome labels per model), with an empirical p-value defined as the proportion of permuted mean AUCs equal to or exceeding the observed mean AUC. Models with fewer than three patients in either class after an 80/20 split were considered not evaluable under this resampling method. Age, sex, ethnicity and PSA status were included as covariates alongside autoantibodies in the design matrix for each model, matching the predictor set used for the original Table 1 models, with all predictors (covariates and autoantibodies) subject to the same L1 penalty.
Table 1. IgG autoantibodies predictive of clinical outcomes via LASSO modelling in SLE.
| Outcome of interest | Lambda | Accuracy | Balanced accuracy | Predictive autoantibodies |
|---|---|---|---|---|
| Cross-sectional clinical data (at sample collection) | ||||
| Asian ethnicity | 1.00 × 1010 | 0.50 | 0.57 | # |
| Active disease (SLEDAI > 4) | 1.00 × 1010 | 0.63 | 0.63 | # |
| Active renal disease (SLEDAI proteinuria score > 0) | 0.043 | 0.68 | 0.55 | # |
| Organ damage (SDI > 0) | 0.148 | 0.62 | 0.54 | # |
| LLDAS | 0.042 | 0.62 | 0.55 | # |
| DORIS remission § | 0.129 | 0.80 | 0.63 | Positive predictors: LRRFIP2, RRAS, AURKA, STK26, SMAD4, SH2D1A |
| Remission off prednisolone† | 0.049 | 0.78 | 0.60 | Positive predictors: COPS6, ZNRD1, RRAS, MAP2K3, AURKA, ARHGDIA, MEF2A, BAG3, SLA, SMAD4, KAT2A, SYCE1, CLK3, SH2D1A Negative predictors: HNRNPA2B1, CERS2, HEXIM1, MAPKAPK5 |
| Prednisolone use | 0.002 | 0.51 | 0.50 | # |
| Immunosuppression use | 0.074 | 0.69 | 0.50 | # |
| Longitudinal clinical data | ||||
| AMS > 4§ | 0.107 | 0.71 | 0.65 | Positive predictors: TRIM21, MED22, CDK16, SLA, GNGT2, ILK Negative predictor: USF1 |
| ≥50% time in LLDAS† | 0.141 | 0.75 | 0.53 | Positive predictors: LRRFIP2, STK26, LMNA, CHN1 |
| ≥50% time in remission | 0.074 | 0.68 | 0.52 | # |
| Renal disease ever (biopsy proven) | 0.002 | 0.52 | 0.50 | # |
| Active SLEDAI-2K domain ever | ||||
| Renal | 0.006 | 0.69 | 0.58 | # |
| CNS† | 0.123 | 0.90 | 0.50 | Positive predictor: GPHN |
| Vasculitis‡ | 0.002 | 0.90 | 0.48 | Positive predictors: HCLS1, HNRNPA2B1, PRC1, SPDEF, PTK7_int, BANK1, MPP5 Negative predictors: Female sex, SOX10, ROPN1, CHIC2 |
| Musculoskeletal | 0.155 | 0.59 | 0.56 | # |
| Mucocutaneous‡ | 0.077 | 0.932 | 0.50 | Negative predictors: AAK1, RPS6KB1, RPS6KB2 |
| Serositis† | 0.775 | 0.78 | 0.51 | Positive predictors: NFYB, H2AFY, ZNF354A, SGK3, CPA4, KIF9, SYCE1, BRD3 Negative predictors: PDPK1, LMNA |
| Haematological | 1.00 × 1010 | 0.93 | 0.50 | — |
AMS, adjusted mean SLEDAI-2K; CNS, central nervous system; DORIS, Definition of Remission in SLE; LLDAS, lupus low disease activity state; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology SLE Damage Index; SLEDAI-2K, SLE Disease Activity Index-2000.
See Supplementary Table 7 for class balance, held-out AUC, balanced accuracy, feature-selection stability and permutation testing results for all 20 models.
Deemed poor model.
—No predictive autoantibody selected by model.
Model exceeded chance-level discrimination under repeated hold-out resampling and permutation testing (permutation p < 0.05).
Model did not exceed chance-level discrimination under this robustness analysis.
Model could not be robustly evaluated owing to an insufficient number of affected patients (n ≤ 4) for reliable resampling.
Targeted analysis of IFN pathway-associated autoantibodies
In addition to above-described unbiased analyses, targeted analysis of autoantibodies associated with IFN pathways was performed. This included analysis of autoantibodies to IFNα1, IFNγ and IFNω, as well as to interferon-induced protein with tetratricopeptide repeats 1 (IFIT1), gamma-interferon-inducible protein 16 (IFI16), interferon-induced 35 kDa protein (IFI35) and IFN regulatory factor 5 (IRF5). Multivariable linear regression was performed to examine relationships between clinical features (exposure), and IFN-associated autoantibody (outcome), using the same covariates as for the differential abundance analysis models. No adjustment for multiple testing was performed, and nominal p-values are reported.
Measurement of LRRFIP2 gene expression
Multivariable linear regression adjusting for age and sex was performed in a publicly available SLE patient RNAseq dataset [ID: GSE88884], to examine the relationships between LRRFIP2 gene expression (outcome) and disease activity as measured by SLEDAI-2K (exposure), and other pro-inflammatory genes including toll like receptor 4 (TLR4) (exposure).
Results
Cohort characteristics
The cohort included 60 patients with SLE (Table 2), with a median (IQR) age of 42 (34–52) years. 90% of patients were female, with 50% each being of European and Asian ethnicity. Median (IQR) disease duration was 12.2 (5.1–18.2) years. 50% of patients had active disease at the time of blood sample collection, as assessed by SLEDAI-2K (SLEDAI-2K breakdown scores are shown in Supplementary Table 4), and 62% had an AMS > 4 across the study period. At the time of blood sample collection, 61% of patients had organ damage (SDI breakdown scores are shown in Supplementary Table 5), 45% were treated with glucocorticoids, 85% were treated with hydroxychloroquine and 73% were treated with immunosuppressive agents.
Table 2. SLE patient cohort characteristics.
| Patient characteristics | SLE (N = 60) |
|---|---|
| Cross-sectional data | |
| Age (years), median (IQR) | 41.5 (34, 52) |
| Female sex, n (%) | 54 (90%) |
| Asian ethnicity*, n (%) | 30 (50%) |
| Paid employment, n (%) | 24 (40%) |
| Disability pension, n (%) | 9 (15%) |
| Household income (<$35K per annum), n (%) | 16 (39%) |
| Education (<secondary level), n (%) | 8 (13.6%) |
| Current smoker, n (%) | 5 (8.3%) |
| Alcohol (>10 standard drinks/week), n (%) | 1 (1.8%) |
| Disease duration (years), median (IQR) | 12.2 (5.1, 18.2) |
| Active disease (SLEDAI-2K > 4), n (%) | 30 (50%) |
| Highly active disease (SLEDAI-2K ≥ 10), n (%) | 9 (15%) |
| Low disease activity (LLDAS)#, n (%) | 21 (35.6%) |
| DORIS remission, n (%) | 18 (30%) |
| Flare, n (%) | 9 (15%) |
| Organ damage (SDI > 0)^, n (%) | 34 (60.7%) |
| Prednisolone use, n (%) | 27 (45%) |
| Hydroxychloroquine use, n (%) | 51 (85%) |
| Immunosuppression use, n (%) | 44 (73.3%) |
| Longitudinal data | |
| Follow-up (years), median (IQR) | 9.5 (6.5, 13.1) |
| AMS > 4, n (%) | 23 (38.3%) |
| ≥50% of time in LLDAS, n (%) | 15 (25%) |
| ≥50% of time in DORIS remission, n (%) | 17 (28.3%) |
| Active SLEDAI-2K domain ever | |
| Renal | 43 (71.7%) |
| CNS | 6 (10%) |
| Vasculitis | 4 (6.7%) |
| Musculoskeletal | 27 (45%) |
| Mucocutaneous | 56 (93.3%) |
| Serositis | 13 (21.7%) |
| Haematological | 4 (6.7%) |
AMS, adjusted mean SLEDAI-2K; CNS, central nervous system; DORIS, Definition of Remission in SLE; ESR, erythrocyte sediment rate; IQR, interquartile range; LLDAS, lupus low disease activity state; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology SLE Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index-2000.
All patients not of Asian ethnicity were of European ethnicity.
Data missing for 1 patient.
Data missing for 4 patients.
Serum IgG autoantibody profiles in SLE
After data filtering via NCF, 533 serum IgG autoantibodies, including all three IFN targets (IFNα1, IFNγ and IFNω), were retained in the final dataset. Autoantibody profiles showed marked heterogeneity across the cohort (Figure 1A). Four potential outliers were identified via PCA (Figure 1B); review of their clinical data did not show any clear patterns in either socio-demographics or disease characteristics, and no indication of a technical issue was identified on the heatmap (Figure 1A). Hence, these potential outliers were included in all downstream analyses. Of note, all potential outliers demonstrated positive PSA signal, which was adjusted for on downstream analyses.
Unsupervised clustering analysis
Using a data-driven approach, we first sought to determine whether clusters of patients existed according to serum IgG autoantibody profiles, using unsupervised PAM clustering. Maximum silhouette width initially identified k = 4 as the optimal cluster solution (Supplementary Figure 3). Bootstrap Jaccard stability analysis (500 resamples; mean Jaccard = 0.63; Supplementary Figures 1–S2 and Table 3) confirmed that k = 4 satisfied the prespecified cluster stability criteria and was retained as the final solution. A total of 37 autoantibodies had a positive signal in more than 5% of samples and were retained for determining patient clusters. Patients in Cluster 1 had autoantibody profiles chiefly characterised by anti-PDCL3, compared to profiles characterised by anti-HCLS1 in Cluster 2, anti-NPM1, anti-PSME3 and anti-HK1 in Cluster 3 and anti-La, anti-Ro60 and anti-PSME3 in Cluster 4 (Figure 2). Patients in Cluster 4 were significantly more likely to be of Asian ethnicity than those in Cluster 1 (adjusted OR: 14.6; 95% CI: 1.51, 140; p = 0.020) and had higher ESR levels than those in Cluster 1 (adjusted β: 1.13; 95% CI: 1.03, 1.24; p = 0.01), and those in Cluster 3 on multivariable regression (adjusted β: 1.23; 95% CI: 1.06, 1.41; p = 0.005). There were no significant associations between clusters and other clinical features, including disease activity, LLDAS, remission or organ damage. Descriptive statistics for each cluster are shown in Supplementary Table 6.
Table 3. IgG autoantibodies associated with clinical features on differential abundance analysis in SLE.
| Exposure of interest | Multivariable models (adjusted for PSA signal) | Multivariable models (adjusted for PSA signal + socio-demographics) |
|---|---|---|
| Cross-sectional clinical data (at sample collection) | ||
| Asian ethnicity | — | — |
| Active disease (SLEDAI > 4) | — | — |
| Active renal disease (SLEDAI proteinuria score > 0) | — | — |
| Organ damage (SDI > 0) | — | — |
| LLDAS | — | — |
| DORIS Remission | LRRFIP2 (β: 1.13; 95% CI: 0.65,1.62; p = 0.009) | LRRFIP2 (β: 1.18; 95% CI: 0.66, 1.69; p = 0.015) PSIP1 (β: 0.95; 95% CI: 0.49, 1.41; p = 0.034) |
| Remission off prednisolone | — | — |
| ESR | SSB (β: 0.04; 95% CI: 0.02, 0.06; p = 0.01) PTK7_int (β: 0.04; 95% CI: 0.02, 0.06; p = 0.013) | — |
| Prednisolone use | — | — |
| Immunosuppression use | — | — |
| Longitudinal clinical data | ||
| AMS | TGIF2 (β: 0.18; 95% CI: 0.10, 0.26; p = 0.020) | — |
| AMS > 4 | — | — |
| Percentage time in LLDAS | LRRFIP2 (β: 0.02; 95% CI: 0.01, 0.03; p = 0.039) | — |
| ≥50% time in LLDAS | LRRFIP2 (β: 1.18; 95% CI: 0.66, 1.71; p = 0.018) | LRRFIP2 (β: 1.20; 95% CI: 0.64, 1.76; p = 0.042) |
| Percentage time in remission | LRRFIP2 (β: 0.02; 95% CI: 0.01, 0.03; p = 0.011) | LRRFIP2 (β: 0.02; 95% CI: 0.01, 0.03; p = 0.022) |
| ≥50% time in remission | — | — |
| Renal disease ever (biopsy proven) | — | — |
| Active SLEDAI-2K domain ever | ||
| Renal | — | — |
| CNS | GPHN (β: 1.69; 95% CI: 0.93, 2.46; p = 0.022) | N/A |
| Vasculitis | — | N/A |
| Musculoskeletal | — | — |
| Mucocutaneous | AAK1 (β: −2.21; 95% CI: −3.08, −1.34; p = 0.002) RPS6KB1 (β: −2.00; 95% CI: −2.91, −1.09; p = 0.013) RPS6KB2 (β: −1.96; 95% CI: −2.88, −1.04; p = 0.013) | N/A |
| Serositis | — | — |
| Haematological | — | N/A |
—No significant results on differential abundance analysis.
N/A, insufficient subgroup size to perform multivariable models adjusting for covariates other than PSA signal.
AAK1, adaptor-associated protein kinase 1; AMS, adjusted mean SLEDAI-2K; CI, confidence interval; CNS, central nervous system; DORIS, Definition of Remission in SLE; ESR, erythrocyte sediment rate; GPHN, gephyrin; LLDAS, lupus low disease activity state; LRRFIP2, leucine rich repeat flightless interacting protein 2; PSA, poly-specific autoantibody; PSIP1, PC4 and SFRS1-interacting protein 1; PTK7-int, inactive tyrosine-protein kinase 7 intracellular domain; RPS6KB1, ribosomal protein S6 kinase B1; RPS6KB2, ribosomal protein S6 kinase B2; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology SLE Damage Index; SLEDAI-2K, SLE Disease Activity Index-2000; SSB, Sjögren's Syndrome Type B; TGIF2, transforming growth factor beta induced factor homeobox 2.
Associations between clinical features and autoantibodies
We next explored associations between each of the 533 individual autoantibodies and each cross-sectional clinical parameter of interest (including overall and organ-specific disease activity, organ damage, remission and medication use) using differential abundance analysis. We also performed supervised machine learning using LASSO, to develop parsimonious autoantibody models which could potentially be used to predict the presence of each clinical feature of interest in other SLE patient cohorts. In addition to assessing associations between autoantibodies and cross-sectional clinical data at the time of study sample collection, we examined associations between autoantibody abundance and longitudinal clinical parameters. The clinical variables relating to the presence of organ-specific SLEDAI-2K disease activity across the study period are intended to reflect general SLE disease phenotypes, rather than linking organ activity across a period in time to the presence of autoantibodies measured at a single point in time. Whilst this facilitates the exploration of autoantibody associations with organ-specific SLE phenotypes in general, it does not allow more detailed exploration of whether any of the autoantibody associations noted are reflective of prior disease activity, or potentially predictive of future disease activity Therefore, these analyses should be interpreted as demonstrating associations between autoantibody profiles measured at a single time point and longitudinal phenotypes, rather than true prediction of future disease course.
Low disease activity and remission
Using differential abundance analysis of the cross-sectional data, DORIS remission was significantly positively associated with autoantibodies to LRR binding FLII interacting protein 2 (LRRFIP2) (adjusted β: 1.18; 95% CI: 0.66, 1.69; p = 0.015) (visualised using Mann–Whitney U test in Figure 3A). When analysing associations between autoantibodies and longitudinal clinical data, percentage time in LLDAS was positively associated with anti-LRRFIP2 abundance when adjusting for PSA signal only (adjusted β: 0.02; 95% CI: 0.01, 0.03; p = 0.039) (visualised using Spearman Rho test in Figure 3B), and patients who spent at least 50% of the study period in LLDAS had higher levels of anti-LRRFIP2 than those who did not (adjusted β: 1.20; 95% CI: 0.64, 1.76; p = 0.042) (Figure 3C). Similarly, percentage time in DORIS remission was positively associated with anti-LRRFIP2 abundance (adjusted β: 0.02; 95% CI: 0.01, 0.03; p = 0.022) (Figure 3D), but there was no significant difference in anti-LRRFIP2 abundance between patients who spent at least 50% of the study period in DORIS remission and those who did not (Table 3).
In LASSO analysis of the cross-sectional data, two predictive models were obtained; one for DORIS remission, which comprised six autoantibodies, including anti-LRRFIP2, and one for remission off glucocorticoids which comprised 18 autoantibodies (Table 1). In analysis of the longitudinal data, a LASSO model was obtained for patients spending ≥50% time in LLDAS, which comprised four autoantibodies, including anti-LRRFIP2 (Table 1), No predictive model could be generated for patients spending ≥50% time in DORIS remission. To assess the robustness of LASSO-based feature selection and predictive performance given the modest sample size, we performed a repeated stratified hold-out resampling analysis (100 resamples, 80/20 split, class-stratified) with permutation testing (1000 label permutations) for these LASSO models (Supplementary Table 7). Of these, only the DORIS remission model [mean held-out AUC: 0.61 (95% empirical interval across resamples range: 0.35–0.92); permutation p = 0.003] demonstrated predictive performance exceeding chance under this resampling scheme; anti-LRRFIP2 was the most consistently selected feature for DORIS remission (retained in 55% of resamples), consistent with the differential abundance findings above. The remission-off-glucocorticoids and ≥50%-time-in-LLDAS models did not demonstrate predictive performance beyond chance under this resampling scheme (permutation p = 0.792 and p = 0.481 respectively) and should be interpreted as exploratory hypothesis-generating feature-selection results rather than validated predictive models.
Overall and organ-specific disease activity
Time-adjusted disease activity as assessed by AMS was significantly positively associated with anti-TGF-β-induced factor homeobox 2 (anti-TGIF2) abundance when adjusting for PSA signal only (adjusted β: 0.18; 95% CI: 0.10, 0.26; p = 0.020) (Supplementary Figure 4). With regards to organ-specific disease activity ever occurring over the study period, when adjusting for PSA signal only, CNS SLE was significantly associated with anti-GPHN (adjusted β: 1.69; 95% CI: 0.93, 2.46; p = 0.022) (Supplementary Figure 5). when adjusting for PSA signal only, mucocutaneous disease was negatively associated with anti-adaptor protein-2 associated kinase 1 (anti-AAK1) (adjusted β: −2.21; 95% CI: −3.08, −1.34; p = 0.002), anti-ribosomal protein S6 kinase B1 (anti-RPS6KB1) (adjusted β: −2.00; 95% CI: −2.91, −1.09; p = 0.013) and anti-RPS6KB2 (adjusted β: −1.96; 95% CI: −2.88, −1.04; p = 0.013) (Supplementary Figure 6). There were no significant associations between autoantibodies and other clinical SLEDAI-2K domains (Table 1), and active disease or active renal disease at the time of study blood sample collection.
Four predictive LASSO models for organ-specific disease activity were obtained, including one autoantibody nominally associated with CNS disease (anti-GPHN), 10 autoantibodies predictive of vasculitis, three autoantibodies predictive of mucocutaneous involvement (anti-AAK1, anti-RPS6KB1, anti-RPS6KB2) and 10 autoantibodies predictive of serositis (Table 1). LASSO models could not be generated for active disease or active renal disease at the time of study blood sample collection. A LASSO model was obtained for time-adjusted disease activity (AMS > 4) across the study period, comprising seven autoantibodies (Table 1). When these five LASSO models were evaluated using the repeated stratified hold-out resampling and permutation testing described above (Supplementary Table 7), only AMS > 4 model showed predictive performance exceeding chance under this resampling scheme [mean held-out AUC 0.56 (95% range: 0.45–0.86); permutation p = 0.021]. In contrast, the CNS/anti-GPHN model did not demonstrate predictive performance exceeding chance-level (mean held-out AUC: 0.50) and anti-GPHN was not consistently selected in any of the 98 valid resamples; the vasculitis and mucocutaneous models could not be robustly evaluated because of the small number of affected patients (n = 4 each); and the serositis model likewise did not demonstrate predictive performance beyond chance (permutation p = 0.599). These four organ-domain models should therefore be interpreted as exploratory hypothesis-generating feature-selection results rather than validated predictive models.
Erythrocyte sedimentation rate
Using differential abundance analysis, ESR levels were positively associated with anti-intracellular PTK7 autoantibody (adjusted β: 0.05; 95% CI: 0.03, 0.07; p = 0.004) (Supplementary Figure 7A), and with anti-La when adjusting for PSA signal only (adjusted β: 0.04; 95% CI: 0.02, 0.06; p = 0.01) (Supplementary Figure 7B).
Of note, there were no significant associations between autoantibodies and socio-demographics, glucocorticoid use, immunosuppression use or organ damage using differential abundance analysis, and no LASSO models for these features could be generated.
LRRFIP2 gene expression in SLE
To further explore the findings of an association between anti-LRRFIP2 and DORIS remission and LLDAS, we examined the association between disease activity and LRRFIP2 at the gene expression level using a publicly available RNAseq dataset of 1,760 patients with SLE. We identified a positive association between disease activity as assessed by SLEDAI-2K and LRRFIP2 gene expression (adjusted β: 0.002; 95% CI: 0.001, 0.004; p < 0.001). LRRFIP2 gene expression was also positively associated with TLR4 gene expression (adjusted β: 0.031; 95% CI: 0.110, 0.136; p < 0.001).
Targeted analysis of IFN pathway-associated autoantibodies in SLE
Given the pathogenic role of IFN pathways in SLE, we performed a targeted analysis to investigate relationships between autoantibodies targeting components of IFN pathways and clinical features. In cross-sectional analysis, using multivariable regression modelling (Table 4), organ damage was negatively associated with anti-IFNω (adjusted β: −0.27; 95% CI: −0.52, −0.02; p = 0.035) (Figure 4A) and anti-IFI35 (adjusted β: −0.52, 95% CI: −1.01, −0.41, p = 0.034) (Figure 4B). Active LN was positively associated with autoantibodies to IRF5 (adjusted β: 0.25; 95% CI: 0.003, 0.497; p = 0.047) (Figure 4C).
Table 4. IFN pathway autoantibodies associated with clinical features on differential abundance analysis in SLE.
| Exposure of interest | Multivariable models (adjusted for PSA signal) | Multivariable models (adjusted for PSA signal + socio-demographics) |
|---|---|---|
| Cross-sectional clinical data (at sample collection) | ||
| Asian ethnicity | — | — |
| Active disease (SLEDAI > 4) | — | — |
| Active renal disease (SLEDAI proteinuria score > 0) | IRF5 (β: 0.2; 95% CI: 0.01, 0.47; p = 0.039) | IRF5 (β: 0.25; 95% CI: 0.003, 0.497; p = 0.047) |
| Organ damage (SDI > 0) | *IFNω*** (β: −0.30; 95% CI: −0.54, −0.06; p = 0.016) | *IFNω*** (β: −0.27; 95% CI: −0.52, −0.02; p = 0.035)IFI35 (β: −0.52; 95% CI: −1.01, −0.04; p = 0.034) |
| LLDAS | — | — |
| DORIS Remission | — | — |
| Remission off prednisolone | — | — |
| ESR | — | — |
| Prednisolone use | — | — |
| Immunosuppression use | — | — |
| AMS | IFNω (β: −0.05; 95% CI: −0.095, −0.002; p = 0.042) | — |
| AMS > 4 | — | — |
| Percentage time in LLDAS | — | — |
| ≥50% time in LLDAS | — | — |
| Percentage time in remission | ||
| ≥50% time in remission | — | IRF5 (β: 0.25; 95% CI: 0.004, 0.500; p = 0.047) |
| Renal disease ever (biopsy-proven) | IFNω (β: −0.23; 95% CI: −0.452, −0.001; p = 0.049)- | — |
| Active SLEDAI-2K domain ever | ||
| Renal | — | — |
| CNS | — | N/A |
| Vasculitis | — | N/A |
| Musculoskeletal | *IFNα<sub>1</sub>* (β: 0.75; 95% CI: 0.10, 1.40; p = 0.025) | *IFNα<sub>1</sub>* (β: 0.79; 95% CI: 0.10, 1.48; p = 0.026) |
| Mucocutaneous | *IFNγ*** (β: 0.37; 95% CI: 0.001, 0.736; p = 0.0495) IFNω (β: 0.48; 95% CI: 0.04, 0.93; p = 0.034) | N/A |
| Serositis | — | — |
| Haematological | — | N/A |
AMS, adjusted mean SLEDAI; CI, confidence interval; CNS, central nervous system; DORIS, Definition of Remission in SLE; ESR, erythrocyte sediment rate; IFN, interferon; LLDAS, lupus low disease activity state; PSA, poly-specific autoantibody; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology SLE Damage Index; SLEDAI-2K, SLE Disease Activity Index-2000.
—, No significant results on differential abundance analysis; N/A, insufficient subgroup size to perform multivariable models adjusting for covariates other than PSA signal.
In analysis of the longitudinal clinical data (Table 4), AMS was negatively associated with anti-IFNω when adjusting for PSA signal only (adjusted β: −0.05; 95% CI: −0.095, −0.002; p = 0.042) (Figure 5A). A history of biopsy-proven LN was also negatively associated with anti-IFNω when adjusting for PSA signal only (adjusted β: −0.23; 95% CI: −0.452, −0.001; p = 0.049) (Figure 5B). Musculoskeletal involvement during the study period was positively associated with anti-IFNα1 (adjusted β: 0.79, 95% CI: 0.10–1.48, p = 0.026) (Figure 5C). Mucocutaneous involvement was positively associated with anti-IFNω (adjusted β: 0.48; 95% CI: 0.04, 0.93; p = 0.034) (Figure 5D) and anti-IFNγ (adjusted β: 0.37; 95% CI: 0.001, 0.736; p = 0.0495) (Figure 5E). Patients who spent at least 50% of the study period in DORIS remission had increased abundance of anti-IRF5 compared to those who did not (adjusted β: 0.25; 95% CI: 0.004, 0.500; p = 0.047) (Figure 5F).
There were no significant associations between other SLE phenotypic features and IFN pathway-associated autoantibodies.
Discussion
Knowledge on autoantibody specificity in SLE chiefly relies on ANA, anti-dsDNA and anti-ENA autoantibodies, routinely used in clinical practice as diagnostic and disease-monitoring biomarkers. Few studies have, however, explored high-plex serum autoantibody profiling in SLE, and studies investigating associations between unbiased autoantibody profiles and validated longitudinal outcome measures, such as time in LLDAS and DORIS remission are lacking. In this exploratory study, we examined associations between high-plex serum IgG autoantibody profiles and both cross-sectional and longitudinal SLE phenotypes and outcomes in a multi-ethnic cohort of 60 patients. Autoantibody abundance was highly heterogenous in this SLE cohort. Using a data-driven approach, unsupervised clustering revealed a subgroup of patients with anti-La and anti-TROVE2 (Ro60) autoantibody profiles, associated with Asian ethnicity and higher ESR, but unbiased clusters did not have any other meaningful clinical associations. In contrast, using differential abundance analysis and a supervised machine learning approach, we identified novel autoantibody associations with DORIS remission and LLDAS attainment (anti-LRRFIP2), disease activity (anti-TGIF2), CNS SLE (anti-GPHN) and cutaneous SLE (anti-AAK1, anti-RPS6KB1 and anti-RPS6KB2). In targeted analysis, we found associations between autoantibodies to IFNs and IFN-associated proteins and organ-specific disease phenotype, disease activity and outcomes.
In clustering analysis, patients in Cluster 4, driven by anti-La and anti-TROVE2 (Ro60) were more likely to be of Asian ethnicity, and to have higher ESR than those in Cluster 1. This is in keeping with known associations between anti-La and anti-Ro60 and Asian ethnicity as well as ESR in SLE (31, 32). Four SLE patient clusters based on high-plex serum autoantibody profiles have previously been reported, including one cluster driven by anti-Ro and anti-La autoantibodies (14). Although this study found differences in SLE organ involvement between clusters, it did not report on differences in ethnicity or ESR level between clusters (14). Altogether, although unsupervised clustering identified structure within the IgG autoantibody dataset, this did not translate into clearly distinct clinical phenotypes in this cohort, beyond associations with ethnicity and ESR. This may reflect the modest cohort size, the disease heterogeneity, and/or the possibility that broad IgG autoantibody profiles do not map directly onto conventional clinical outcome categories, suggesting that clinically relevant information may be better captured by individual IgG autoantibody specificities than by global autoantibody signatures.
In differential abundance analysis, anti-LRRFIP2 was positively associated with DORIS remission cross-sectionally, as well as with greater percentage of time spent in DORIS remission and LLDAS across the study period, and was also predictive of these outcomes using LASSO modelling. LRRFIP2 is a pro-inflammatory protein which activates NF-кB by binding to the cytosolic part of TLR4 protein and to MyD88 (33, 34). While neither anti-LRRFIP2 or LRRFIP2 protein has previously been described in patients with SLE or other autoimmune diseases, the known pro-inflammatory functions of LRRFIP2 protein provide biological plausibility for the observed association between anti-LRRFIP2 and low disease activity and remission. Although speculative, one possible explanation is that anti-LRRFIP2 may exert a regulatory or neutralising effect on its target antigen. Naturally occurring non-pathogenic autoantibodies are known to play a role in immune homeostasis (35). More specifically, anti-cytokine autoantibodies with regulatory and modulatory functions have been reported in SLE, such as those targeting Type I IFN, which are associated with lower disease activity, supporting the concept that some autoantibodies may have immunomodulatory rather than pathogenic effects (10, 11). Further research is warranted to externally validate this candidate biomarker, to determine whether anti-LRRFIP2 has any functional effects on LRRFIP2 activity, and to explore its biological relevance through functional studies.
We report for the first time a positive association between anti-gephyrin (GPHN) and CNS SLE using differential abundance analysis and LASSO modelling. GPHN protein plays a role in organising the inhibitory neurotransmitter response at γ-aminobutyric acid (GABA)-ergic synapses (36). Anti-GPHN autoantibodies have been described in a patient with Stiff-Person Syndrome, a rare autoimmune CNS disease characterised by autoantibodies to inhibitory synaptic antigens (37). It is noteworthy that anti-GPHN autoantibodies have been identified in patients with systemic sclerosis, and are associated with gastrointestinal dysmotility, thought possibly due to binding to gephyrin in the myenteric ganglia in the enteric nervous system (38). Altogether, these findings raise the possibility of a role for anti-GPHN in CNS lupus. However, the LASSO-based predictive model did not demonstrate predictive performance beyond chance in repeated stratified hold-out resampling and permutation testing (mean held-out AUC: 0.50; anti-GPHN was not consistently re-selected across resamples). Consequently, this LASSO analysis should not be regarded as independent predictive evidence beyond the differential abundance association described above. Given the small number of patients with CNS lupus in our cohort (n = 6), these findings should be interpreted with caution and regarded as preliminary, pending validation in larger, independent cohorts. Further research, including functional studies, is warranted to validate these findings and to explore the potential of anti-GPHN as a diagnostic biomarker and/or therapeutic target in CNS SLE.
Anti-AAK1, anti-RPS6KB1 and anti-RPS6KB2 were negatively associated with mucocutaneous involvement using differential abundance analysis and LASSO. AAK1 protein plays a pathogenic role in susceptibility to viral infections, neurodegenerative diseases, neuropathic pain and cancers including melanoma (39). RPS6KB1 and RPS6KB2 are components of the insulin signalling pathway, playing a role in insulin resistance (40), and in melanogenesis in the skin (41). RPS6KB1 is also associated with cellular inflammation in obese patients (40), being activated by inflammatory cytokines TNFα and IL-1β and positively correlated with other inflammatory proteins (40). RPS6KB1 RNA expression is also up-regulated in patients with multiple sclerosis (42), but down-regulated in lesional skin of psoriasis patients (41). RPS6KB2 interacts with pro-inflammatory genes, including NF-кB1 (43). AAK1, RPS6KB1 and RPSkKB2 have not previously been described in autoimmune connective tissue diseases, including SLE, and the mechanism by which autoantibodies to these proteins may play a role in mucocutaneous lupus requires further research. The LASSO component of this finding could not be robustly evaluated by repeated hold-out stratified resampling, as only four patients had mucocutaneous involvement over the study period. Accordingly, this autoantibody panel should be regarded as a candidate association identified from differential abundance analysis pending validation in a larger cohort.
The pathogenic role of the Type I IFN pathway in SLE is well described (44). SLE patients with high Type I IFN expression have increased disease activity, with increased rates of arthritis, cutaneous lupus and renal involvement (7). In our IFN targeted analysis, anti-IFNω was negatively associated with AMS, organ damage and biopsy-proven LN and positively associated with mucocutaneous disease. Our findings support the literature that anti-IFNω autoantibodies are detectable in SLE patients (10, 11). Associations of these autoantibodies with clinical phenotype have not previously been described, although our finding of decreased disease activity in patients with high anti-IFNω abundance likely reflects the pro-inflammatory role of the Type I IFN pathway in SLE. Autoantibodies targeting IFNα17 and IFNα8 have been reported in patients with SLE (10–12), and are associated with reduced disease activity, and increased rates of remission (10). Our finding of a positive association between anti-IFNα1 and musculoskeletal disease may support this, reflecting a less severe, non-organ threatening SLE disease phenotype.
IRF5 plays a role in the induction of Type I IFN, and IRF5 mRNA is elevated in patients with SLE (45). Upregulation of the IRF5 gene is associated with increased risk of SLE development, and inhibition of IRF5 hyperactivation has been shown to reduce SLE disease activity in a lupus-prone mouse model (46). Anti-IRF5 autoantibodies have recently been reported to be more highly expressed in SLE patients compared to HC, and to be positively associated with high disease activity, and negatively associated with LLDAS (15). Our finding of a positive association between anti-IRF5 and renal disease activity is in line with this. However, our finding of a positive association between anti-IRF5 and spending >50% time in remission contradicts these results, and suggests the need for further studies of anti-IRF5 in SLE. Additionally, we found a negative association between anti-IFI35 and organ damage. To our knowledge, anti-IFI35 has not previously been described in SLE.
The Type II IFN pathway may also play a role in SLE pathogenesis, with IFNγ levels increased several years prior to SLE diagnosis in patients compared to HC, and correlated with IFNα levels (47). Furthermore, IFNγ expression is increased in LN and is positively associated with disease activity (48). It is also overexpressed in keratinocytes in SLE skin lesions (49). Autoantibodies targeting IFNγ have been previously identified in patients with SLE and are associated with increased susceptibility to severe infections and increased disease activity (8). Although we found a positive association between anti-IFNγ and mucocutaneous disease, we did not find any associations between anti-IFNγ and overall disease activity.
Our study has several limitations. Although our multi-ethnic cohort was highly annotated with comprehensive longitudinal clinical data, and well balanced in terms of ethnicity and disease activity, the sample size of this single-centre study was moderate, and thus patient subgroups for some clinical parameters were small, rendering our study under-powered to demonstrate further significant associations on unsupervised clustering (particularly with regards to organ-specific disease activity) and differential abundance analysis, and potentially contributing to low accuracy and/or balanced accuracy on some LASSO models. The targeted IFN-pathway analyses were pre-specified exploratory analyses and were thus not corrected for multiple testing; these results should therefore be interpreted cautiously and should be considered hypothesis-generating, given the increased risk of false-positive findings. Given the exploratory, hypothesis-generating nature of this study, our findings require external validation. Further investigation of the autoantibody associations with SLE phenotypes in larger, independent cohorts using independent orthogonal autoantibody profiling technologies is warranted. Highly heterogenous autoantibody profiles may also have affected the revelation of a strong signal in the clustering analysis. Finally, associations with longitudinal clinical data were explored using autoantibody profiling performed at a single time point. While this provides an indication that overall SLE disease phenotypes may be associated with certain autoantibodies, it does not allow us to infer whether these associations are a result of past disease activity, or whether they may predict future disease activity. Accordingly, analyses involving longitudinal clinical phenotypes presented in this study should be interpreted as measures of association rather than prediction. Further longitudinal studies of autoantibody profiling performed at multiple time points with repeat sampling are warranted to determine the temporal stability and predictive value of these associations. Specifically, LASSO models for vasculitis and mucocutaneous involvement could not be robustly evaluated by repeated stratified hold-out resampling owing to the very small number of affected patients (n = 4 for each outcome), and 4 of the remaining 6 reported and evaluable LASSO models, including the CNS/anti-GPHN model, did not demonstrate predictive performance beyond chance. These LASSO findings should therefore be interpreted as exploratory and hypothesis-generating, pending validation in larger independent cohorts.
In conclusion, this is the first study to report associations between high-plex serum IgG autoantibody profiles and comprehensive longitudinal disease outcomes in patients with SLE. Our findings highlight the biological heterogeneity of SLE at the IgG autoantibody level. The identification of anti-GPHN and anti-LRRFIP2 as candidate biomarkers for CNS phenotype and LLDAS/remission attainment, respectively, may reflect underlying immunopathological pathways. Work to examine these associations in larger cohorts is underway, but these preliminary findings may point to new therapeutic targets. Further research is warranted to investigate the role of these autoantibodies in functional studies.
Previous presentations of work
Part of this work was presented as a poster at the Australian Rheumatology Association Conference in May 2025. Part of this work was presented as a publication only abstract (i.e., no oral presentation or poster) at the EULAR Rheumatology Congress in June 2025.
Generative AI statement
The author(s) declared that generative AI was used in the creation of this manuscript. Artificial intelligence (AI) tools were used to assist with drafting, language refinement and structuring of the manuscript text and reviewer responses. AI was not used for study design, data analysis or interpretation of results.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/flupu.2026.1892328/full#supplementary-material
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Supplementary material Supplementary material at the source
Footnotes
Edited by: Luís Pedro Sousa Inês, Coimbra Hospital and University Center, Portugal
Reviewed by: Kleio-Maria Verrou, New York University, United States
Shengfang Bao, Shanghai Children’s Medical Center, China
Statements
- Conflict of interest
This study received in-kind support from Sengenics (a subsidiary of Standard BioTools) for sample processing and bioinformatic support on the Sengenics platform. FV reports financial support was provided by National Health and Medical Research Council (Australia). FV reports financial support was provided by Standard BioTools Inc. LE reports financial support was provided by National Health and Medical Research Council (Australia). FV reports financial support was provided by Arthritis Australia & Arthritis and Osteoporosis Western Australia. LE reports a relationship with Janssen-Cilag Ltd that includes: funding grants. FV reports a relationship with Janssen-Cilag Ltd that includes: funding grants. FV reports a relationship with AstraZeneca Pharmaceuticals that includes: funding grants. FV reports a relationship with Bristol Myers Squibb Co that includes: funding grants. FV reports a relationship with CSL Limited that includes: funding grant. AH reports a relationship with AstraZeneca Pharmaceuticals that includes: funding grants and speaking and lecture fees. AH reports a relationship with Novartis that includes: speaking and lecture fees. AH reports a relationship with Seqirus Inc that includes: speaking and lecture fees. AH reports a relationship with Janssen Pharmaceuticals Inc that includes: consulting or advisory. AH reports a relationship with Roche that includes: consulting or advisory. AH reports a relationship with GlaxoSmithKline Inc that includes: consulting or advisory and funding grants. AH reports a relationship with UCB Pharma that includes: consulting or advisory. AH reports a relationship with Recordati that includes: consulting or advisory. AH reports a relationship with Merck Serono that includes: funding grants. AH reports a relationship with Bristol Myers Squibb Co that includes: funding grants. RK-R reports a relationship with Bristol Myers Squibb Co that includes: funding grants. RK-R reports a relationship with GlaxoSmithKline Inc that includes: funding grants. RK-R reports a relationship with Janssen Pharmaceuticals Inc that includes: funding grants. RK-R reports a relationship with Novartis that includes: funding grants. Katie Lennard reports a relationship with Standard BioTools Inc that includes: employment. EM reports a relationship with AbbVie Inc that includes: consulting or advisory and funding grants. EM reports a relationship with Amgen Inc that includes: funding grants. EM reports a relationship with AstraZeneca Pharmaceuticals that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with Biogen that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with Bristol Myers Squibb Co that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with Dragonfly Therapeutics, Inc. that includes: board membership, consulting or advisory, and speaking and lecture fees. EM reports a relationship with Eli Lilly and Company that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with EMD Serono Inc that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with Genentech-Hoffmann La Roche that includes: funding grants. EM reports a relationship with GlaxoSmithKline Inc that includes: board membership, consulting or advisory, funding grants, and speaking and lecture fees. EM reports a relationship with Janssen Pharmaceuticals Inc that includes: funding grants. EM reports a relationship with Novartis that includes: board membership, consulting or advisory, and funding grants. EM reports a relationship with Quell Therapeutics Limited that includes: board membership, consulting or advisory, and speaking and lecture fees. EM reports a relationship with RemeGen that includes: board membership, consulting or advisory, and speaking and lecture fees. EM reports a relationship with Takeda that includes: funding grants. EM reports a relationship with Union Chimique Belge that includes: board membership, consulting or advisory, and funding grants. EM reports a relationship with Zenas that includes: board membership, consulting or advisory, and speaking and lecture fees.
The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
- Funding
The author(s) declared that financial support was received for this work and/or its publication. This study received in-kind support from Sengenics (a subsidiary of Standard BioTools) for sample processing and bioinformatic support on the Sengenics platform. FBV was supported by funding from National Health and Medical Research Council (Emerging Leadership 1, grant 1196112) and is supported by funding from Arthritis Australia & Arthritis and Osteoporosis Western Australia (Grant 26-AA/AWA-F). LEE is supported by funding from the National Health and Medical Research Council (PhD scholarship 2014036).
- Data availability
The study findings and supporting materials are included in the article/Supplementary Material. Access to the underlying data is restricted due to ethical and patient consent requirements. Further inquiries can be directed to the corresponding author.
- Ethics approval
The studies involving humans were approved by Monash Health Human Research Ethics Committee, as part of the Mechanisms of Inflammation in Rheumatic Diseases (MIRD) study (Ref: 13019A). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.
- Author contributions
LE: Data curation, Formal analysis, Investigation, Visualization, Writing – original draft, Writing – review & editing. WF: Methodology, Writing – review & editing. KL: Data curation, Formal analysis, Methodology, Visualization, Writing – review & editing. HV: Formal analysis, Writing – review & editing. AH: Data curation, Supervision, Writing – review & editing. RK-R: Formal analysis, Supervision, Writing – review & editing. EM: Methodology, Supervision, Writing – review & editing. AP: Methodology, Writing – review & editing. FV: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing.
- Acknowledgements
The authors wish to acknowledge the Traditional Owners of the lands on which they live and work, and pay respects to their Elders past, present, and emerging. The authors also wish to acknowledge the study participants who contributed their blood samples and clinical data. In addition, the authors which to acknowledge the ALRB staff who contributed to this project, in particular, Ms Rachel Koelmeyer and Ms Lihini De Silva, as well as the Standard BioTools staff who contributed, in particular Ms Bermet Abylova.
References
- 1. Morand EF, Fernandez-Ruiz R, Blazer A, Niewold TB. Advances in the management of systemic lupus erythematosus. Br Med J. (2023) 383: e073980. doi:10.1136/bmj-2022-073980
- 2. Eades LE, Hoi AY, Liddle R, Sines J, Kandane-Rathnayake R, Khetan S. Systemic lupus erythematosus in Aboriginal and Torres Strait Islander peoples in Australia: addressing disparities and barriers to optimising patient care. Lancet Rheumatol. (2024) 6 (10): e713 – e26. doi:10.1016/S2665-9913(24)00095-X
- 3. Vincent F, Figgett W, Hibbs M. Hallmark of systemic lupus erythematosus: role of B cell hyperactivity. In: Hoi A, editor. Pathogenesis of Systemic Lupus Erythematosus. Cham: Springer (2021). p. 9 – 36.
- 4. Aringer M, Leuchten N, Johnson SR. New criteria for lupus. Curr Rheumatol Rep. (2020) 22 (6): 18. doi:10.1007/s11926-020-00896-6
- 5. Narayanan K, Marwaha V, Shanmuganandan K, Shankar S. Correlation between systemic lupus erythematosus disease activity index, C3, C4 and anti-dsDNA antibodies. Med J Armed Forces India. (2010) 66 (2): 102 – 7. doi:10.1016/S0377-1237(10)80118-2
- 6. Arroyo-Ávila M, Santiago-Casas Y, McGwin G, Cantor RS, Petri M, Ramsey-Goldman R. Clinical associations of anti-smith antibodies in PROFILE: a multi-ethnic lupus cohort. Clin Rheumatol. (2015) 34 (7): 1217 – 23. doi:10.1007/s10067-015-2941-y
- 7. Northcott M, Jones S, Koelmeyer R, Bonin J, Vincent F, Kandane-Rathnayake R. Type 1 interferon status in systemic lupus erythematosus: a longitudinal analysis. Lupus Sci Med. (2022) 9 (1): e000625. doi:10.1136/lupus-2021-000625
- 8. Chen L, Chi H, Teng J, Meng J, Zhang H, Su Y. Neutralizing anti-IFN- γ IgG was increased in patients with systemic lupus erythematosus and associated with susceptibility to infection. Clin Rheumatol. (2024) 43 (1): 189 – 98. doi:10.1007/s10067-023-06758-7
- 9. Kalunian KC, Furie R, Morand EF, Bruce IN, Manzi S, Tanaka Y. A randomized, placebo-controlled phase III extension trial of the long-term safety and tolerability of anifrolumab in active systemic lupus erythematosus. Arthritis Rheumatol. (2023) 75 (2): 253 – 65. doi:10.1002/art.42392
- 10. Mathian A, Breillat P, Dorgham K, Bastard P, Charre C, Lhote R. Lower disease activity but higher risk of severe COVID-19 and herpes zoster in patients with systemic lupus erythematosus with pre-existing autoantibodies neutralising IFN-α. Ann Rheum Dis. (2022) 81 (12): 1695 – 703. doi:10.1136/ard-2022-222549
- 11. Slavikova M, Schmeisser H, Kontsekova E, Mateicka F, Borecky L, Kontsek P. Incidence of autoantibodies against type I and type II interferons in a cohort of systemic lupus erythematosus patients in Slovakia. J Interferon Cytokine Res. (2003) 23 (3): 143 – 7. doi:10.1089/107999003321532475
- 12. Wang EY, Dai Y, Rosen CE, Schmitt MM, Dong MX, Ferré EMN. High-throughput identification of autoantibodies that target the human exoproteome. Cell Rep Methods. (2022) 2 (2): 100172. doi:10.1016/j.crmeth.2022.100172
- 13. Kurien BT, Scofield RH. Autoantibody determination in the diagnosis of systemic lupus erythematosus. Scand J Immunol. (2006) 64 (3): 227 – 35. doi:10.1111/j.1365-3083.2006.01819.x
- 14. Lewis MJ, McAndrew MB, Wheeler C, Workman N, Agashe P, Koopmann J. Autoantibodies targeting TLR and SMAD pathways define new subgroups in systemic lupus erythematosus. J Autoimmun. (2018) 91: 1 – 12. doi:10.1016/j.jaut.2018.02.009
- 15. Lindblom J, Lagutkin D, Sherina N, Idborg H, Beretta L, Borghi MO. Novel IgG and IgA autoantibodies validated in two independent cohorts are associated with disease activity and determine organ manifestations in systemic lupus erythematosus: implications for anti-LIN28A, anti-HMGN5, anti-IRF5, and anti-TGIF1. Ann Rheum Dis. (2025) 84 (7): 1164 – 79. doi:10.1016/j.ard.2025.04.008
- 16. Eades LE, Papadaki A, Lennard K, Hoi AY, Kandane-Rathnayake R, Morand EF. Serum IgA autoantibody profiling in systemic lupus erythematosus reveals associations with disease phenotypes. RMD Open. (2026) 12 (3): e006811. doi:10.1136/rmdopen-2026-006811
- 17. O'Neill S, Morand EF, Hoi A. The Australian lupus registry and biobank: a timely initiative. Med J Aust. (2017) 206 (5): 194 – 5. doi:10.5694/mja16.01282
- 18. Petri M, Orbai A-M, Alarcón GS, Gordon C, Merrill JT, Fortin PR. Derivation and validation of the systemic lupus international collaborating clinics classification criteria for systemic lupus erythematosus. Arthritis Rheum. (2012) 64 (8): 2677 – 86. doi:10.1002/art.34473
- 19. Hochberg MC. Updating the American college of rheumatology revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum. (1997) 40 (9): 1725. doi:10.1002/art.1780400928
- 20. Gladman DD, Ibañez D, Urowitz MB. Systemic lupus erythematosus disease activity index 2000. J Rheumatol. (2002) 29 (2): 288 – 91.
- 21. Vincent FB, Kandane-Rathnayake R, Koelmeyer R, Harris J, Hoi AY, Mackay F. Associations of serum soluble fas and fas ligand (FasL) with outcomes in systemic lupus erythematosus. Lupus Sci Med. (2020) 7 (1): e000375. doi:10.1136/lupus-2019-000375
- 22. Stoll T, Seifert B, Isenberg DA. SLICC/ACR damage Index is valid, and renal and pulmonary organ scores are predictors of severe outcome in patients with systemic lupus erythematosus. Br J Rheumatol. (1996) 35 (3): 248 – 54. doi:10.1093/rheumatology/35.3.248
- 23. Franklyn K, Lau CS, Navarra SV, Louthrenoo W, Lateef A, Hamijoyo L. Definition and initial validation of a lupus low disease activity state (LLDAS). Ann Rheum Dis. (2016) 75 (9): 1615. doi:10.1136/annrheumdis-2015-207726
- 24. van Vollenhoven RF, Bertsias G, Doria A, Isenberg D, Morand E, Petri MA. 2021 DORIS definition of remission in SLE: final recommendations from an international task force. Lupus Sci Med. (2021) 8 (1): e000538. doi:10.1136/lupus-2021-000538
- 25. Buyon JP, Petri MA, Kim MY, Kalunian KC, Grossman J, Hahn BH. The effect of combined estrogen and progesterone hormone replacement therapy on disease activity in systemic lupus erythematosus: a randomized trial. Ann Intern Med. (2005) 142 (12 Pt 1): 953 – 62. doi:10.7326/0003-4819-142-12_Part_1-200506210-00004
- 26. Ibañez D, Urowitz MB, Gladman DD. Summarizing disease features over time: i. Adjusted mean SLEDAI derivation and application to an index of disease activity in lupus. J Rheumatol. (2003) 30 (9): 1977 – 82.
- 27. Kandane-Rathnayake R, Golder V, Louthrenoo W, Chen Y-H, Cho J, Lateef A. Lupus low disease activity state and remission and risk of mortality in patients with systemic lupus erythematosus: a prospective, multinational, longitudinal cohort study. Lancet Rheumatol. (2022) 4 (12): e822 – e30. doi:10.1016/S2665-9913(22)00304-6
- 28. Vincent FB, Nim HT, Lee JPW, Morand EF, Harris J. Effect of storage duration on cytokine stability in human serum and plasma. Cytokine. (2019) 113: 453 – 7. doi:10.1016/j.cyto.2018.06.009
- 29. Ballman KV, Grill DE, Oberg AL, Therneau TM. Faster cyclic loess: normalizing RNA arrays via linear models. Bioinformatics. (2004) 20 (16): 2778 – 86. doi:10.1093/bioinformatics/bth327
- 30. Hennig C. Cluster-wise assessment of cluster stability. Comput Stat Data Anal. (2007) 52 (1): 258 – 71. doi:10.1016/j.csda.2006.11.025
- 31. Denvir B, Carlucci PM, Corbitt K, Buyon JP, Belmont HM, Gold HT. Prevalence of concomitant rheumatologic diseases and autoantibody specificities among racial and ethnic groups in SLE patients. Front Epidemiol. (2024) 4: 1334859. doi:10.3389/fepid.2024.1334859
- 32. Gong M, Dai L, Xie Z, Hong D, Li N, Fan X. Serological and clinical associations of autoantibodies in Chinese patients with new-onset systemic lupus erythematosus. Sci Rep. (2023) 13 (1): 10101. doi:10.1038/s41598-023-37100-5
- 33. Gunawardena HP, Huang Y, Kenjale R, Wang H, Xie L, Chen X. Unambiguous characterization of site-specific phosphorylation of leucine-rich repeat fli-I-interacting protein 2 (LRRFIP2) in toll-like receptor 4 (TLR4)-mediated signaling *. J Biol Chem. (2011) 286 (13): 10897 – 910. doi:10.1074/jbc.M110.168179
- 34. Dai P, Jeong SY, Yu Y, Leng T, Wu W, Xie L. Modulation of TLR signaling by multiple MyD88-interacting partners including leucine-rich repeat fli-I-interacting proteins. J Immunol. (2009) 182 (6): 3450 – 60. doi:10.4049/jimmunol.0802260
- 35. Elkon K, Casali P. Nature and functions of autoantibodies. Nat Clin Pract Rheumatol. (2008) 4 (9): 491 – 8. doi:10.1038/ncprheum0895
- 36. Tyagarajan SK, Ghosh H, Yévenes GE, Nikonenko I, Ebeling C, Schwerdel C. Regulation of GABAergic synapse formation and plasticity by GSK3 β -dependent phosphorylation of gephyrin. Proc Natl Acad Sci USA. (2011) 108 (1): 379 – 84. doi:10.1073/pnas.1011824108
- 37. Butler MH, Hayashi A, Ohkoshi N, Villmann C, Becker C-M, Feng G. Autoimmunity to gephyrin in stiff-man syndrome. Neuron. (2000) 26 (2): 307 – 12. doi:10.1016/S0896-6273(00)81165-4
- 38. McMahan ZH, Kulkarni S, Andrade F, Perin J, Zhang C, Hooper JE. Anti-Gephyrin antibodies: a novel specificity in patients with systemic sclerosis and lower bowel dysfunction. Arthritis Rheumatol. (2024) 76 (1): 92 – 9. doi:10.1002/art.42667
- 39. Yoshida A, Ohtsuka S, Matsumoto F, Miyagawa T, Okino R, Ikeda Y. Development of a novel AAK1 inhibitor via kinobeads-based screening. Sci Rep. (2024) 14 (1): 6723. doi:10.1038/s41598-024-57051-9
- 40. Catalán V, Gómez-Ambrosi J, Rodríguez A, Ramírez B, Andrada P, Rotellar F. Expression of S6K1 in human visceral adipose tissue is upregulated in obesity and related to insulin resistance and inflammation. Acta Diabetol. (2015) 52 (2): 257 – 66. doi:10.1007/s00592-014-0632-9
- 41. Loite U, Raam L, Reimann E, Reemann P, Prans E, Traks T. The expression pattern of genes related to melanogenesis and endogenous opioids in psoriasis. Int J Mol Sci. (2021) 22 (23): 13056. doi:10.3390/ijms222313056
- 42. Akbarian F, Tabatabaiefar MA, Shaygannejad V, Shahpouri MM, Badihian N, Sajjadi R. Upregulation of MTOR, RPS6KB1, and EIF4EBP1 in the whole blood samples of Iranian patients with multiple sclerosis compared to healthy controls. Metab Brain Dis. (2020) 35 (8): 1309 – 16. doi:10.1007/s11011-020-00590-7
- 43. Slattery ML, Lundgreen A, Herrick JS, Wolff RK. Genetic variation in RPS6KA1, RPS6KA2, RPS6KB1, RPS6KB2, and PDK1 and risk of colon or rectal cancer. Mutat Res. (2011) 706 (1–2): 13 – 20. doi:10.1016/j.mrfmmm.2010.10.005
- 44. Rönnblom L. Potential role of IFN α in adult lupus. Arthritis Res Ther. (2010) 12 (Suppl 1): S3. doi:10.1186/ar2884
- 45. Shu J, Li L, Zhou L-B, Qian J, Fan Z-D, Zhuang L-L. IRF5 Is elevated in childhood-onset SLE and regulated by histone acetyltransferase and histone deacetylase inhibitors. Oncotarget. (2017) 8 (29): 47184 – 94. doi:10.18632/oncotarget.17586
- 46. Song S, De S, Nelson V, Chopra S, LaPan M, Kampta K. Inhibition of IRF5 hyperactivation protects from lupus onset and severity. J Clin Invest. (2020) 130 (12): 6700 – 17. doi:10.1172/JCI120288
- 47. Munroe ME, Lu R, Zhao YD, Fife DA, Robertson JM, Guthridge JM. Altered type II interferon precedes autoantibody accrual and elevated type I interferon activity prior to systemic lupus erythematosus classification. Ann Rheum Dis. (2016) 75 (11): 2014. doi:10.1136/annrheumdis-2015-208140
- 48. Wen S, He F, Zhu X, Yuan S, Liu H, Sun L. IFN- γ, CXCL16, uPAR: potential biomarkers for systemic lupus erythematosus. Clin Exp Rheumatol. (2018) 36 (1): 36 – 43.
- 49. Yerram KV, Baisya R, Kumar P, Mylavarapu R, Rajasekhar L. Serum interferon-alpha predicts in-hospital mortality in patients hospitalised with acute severe lupus. Lupus Sci Med. (2023) 10 (2): e000933. doi:10.1136/lupus-2023-000933
Citation
Eades, L. E., et al. (2026). Serum autoantibody profiling in systemic lupus erythematosus reveals associations with low disease activity and remission. Frontiers in Lupus, Volume 4. https://doi.org/
- Received
- Accepted
- Published
- Volume
- 4 - 2026
Copyright
© 2026 Eades, Figgett, Lennard, Valipour Kahrood, Hoi, Kandane-Rathnayake, Morand, Papadaki and Vincent.
License: https://creativecommons.org/licenses/by/4.0/
*Correspondence: Fabien B. Vincent
Article metrics
- Views
- 0
- Downloads
- 0
- Citations
- 0
- Shares
- 0