Session Type: Abstract Submissions (ACR)
Background/Purpose: The heterogeneous nature of SLE and a lack of reliable biomarkers make SLE a challenge to classify in patients. A differential diagnosis of SLE from other rheumatic and autoimmune diseases can be difficult due to lack of definitive tests with sufficient sensitivity and specificity. Genome-wide DNA methylation analysis of SLE patients’ PBMCs was performed on Illumina HumanMethyl450 BeadChips to assess whether SLE has a DNA methylation signature that outperforms current diagnostic tests.
Methods: PBMC Genomic DNA was isolated from 252 patients and evaluated with the Illumina HumanMethylation450chip. The 147 patient sample training set consisted of 15 SLE, 16 rheumatoid arthritis, (RA), 13 osteoarthritis (OA), 28 healthy control (HC), and 75 with undifferentiated arthritis (UA), but not SLE after one year of follow up. Differentially methylated loci (DML) were identified using a non-parametric statistical test to rank CpG loci that were differentially methylated in SLE samples compared to non-SLE samples. An independent, blinded 105 patient PBMC sample test set: 17 SLE, 28 RA, 12 HC, 17 other rheumatic (ankylosing spondylitis, psoriatic arthritis, reactive arthritis, spondyloarthopathy), and 31 other autoimmune disease (Crohn’s, Diabetes Type I, multiple sclerosis, psoriasis, ulcerative colitis) patients was used to confirm the model.
Results: Of 456,355 autosomal CpGs whose methylation frequencies were assessed, the top 25 DML were identified. These 25 DML were associated with 18 genes (10 kb upstream to 10 kb downstream of gene). 10 of these genes had previously been found to be differentially methylated or differentially expressed in published SLE studies. DML selection and model training were repeated on 258 CpGs associated with these 10 genes; a panel of the top 25-DML from the resultant list was selected. This 10 gene-based 25 DML panel was used to train a support vector machine model (radial basis function kernel, cost = 1). The model cut off for SLE/non-SLE classification was determined using leave one out cross validation on the training set. In a prospective study, our 25-DML SLE diagnostic model correctly classified 14 of 17 independent SLE samples (82% sensitivity) and 86 of 88 non-SLE samples (98% specificity), achieving overall accuracy of 95%. Disease severity and medication information did not explain the misclassification of 3 SLE samples. The 18 genes associated with the initial 25 DML were enriched in Ingenuity’s pathways for 1) Interferon Signaling and 2) Activation of interferon regulatory factors by Cytosolic Pattern Recognition Receptors (Benjamini-Hochberg p-value < 0.01 for each). The Interferon Signaling pathway was also enriched in SLE patients for the selected 25 DML panel validated on the 105 prospective samples.
Conclusion: This study confirms that SLE can be distinguished with high accuracy from other rheumatic diseases using DNA methylation biomarkers from PBMCs. Expanded studies are warranted to improve the diagnosis of SLE based on DNA methylation signatures. Furthermore, these biomarkers could identify the underlying molecular pathways associated with SLE. This understanding might be used to select therapies for patients or to identify novel therapeutic targets.
D. W. Anderson,
J. E. Lim,
« Back to 2013 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/a-systemic-lupus-erythematosus-sle-diagnostic-test-based-on-dna-methylation-signatures-from-peripheral-blood-mononuclear-cells/