Disease diagnostics using machine learning of immune receptors.

Clinical diagnosis typically incorporates physical examination, patient history, and various laboratory tests and imaging studies, but makes limited use of the human system's own record of antigen exposures encoded by receptors on B cells and T cells. We analyzed immune receptor datasets from 593 individuals to develop MAchine Learning for Immunological Diagnosis (Mal-ID) , an interpretive framework to screen for multiple illnesses simultaneously or precisely test for one condition. This approach detects specific infections, autoimmune disorders, vaccine responses, and disease severity differences. Human-interpretable features of the model recapitulate known immune responses to SARS-CoV-2, Influenza, and HIV, highlight antigen-specific receptors, and reveal distinct characteristics of Systemic Lupus Erythematosus and Type-1 Diabetes autoreactivity. This analysis framework has broad potential for scientific and clinical interpretation of human immune responses.

Ancestry-based differences in the immune phenotype are associated with lupus activity.

Systemic lupus erythematosus (SLE) affects 1 in 537 Black women, which is >2-fold more than White women. Black patients develop the disease at a younger age, have more severe symptoms, and have a greater chance of early mortality. We used a multiomics approach to uncover ancestry-associated immune alterations in patients with SLE and healthy controls that may contribute biologically to disease disparities. Cell composition, signaling, epigenetics, and proteomics were evaluated by mass cytometry; droplet-based single-cell transcriptomics and proteomics; and bead-based multiplex soluble mediator levels in plasma. We observed altered whole blood frequencies and enhanced activity in CD8+ T cells, B cells, monocytes, and DCs in Black patients with more active disease. Epigenetic modifications in CD8+ T cells (H3K27ac) could distinguish disease activity level in Black patients and differentiate Black from White patient samples. TLR3/4/7/8/9-related gene expression was elevated in immune cells from Black patients with SLE, and TLR7/8/9 and IFN-α phospho-signaling and cytokine responses were heightened even in immune cells from healthy Black control patients compared with White individuals. TLR stimulation of healthy immune cells recapitulated the ancestry-associated SLE immunophenotypes. This multiomic resource defines ancestry-associated immune phenotypes that differ between Black and White patients with SLE, which may influence the course and severity of SLE and other diseases.

Genetic load in incomplete lupus erythematosus.

OBJECTIVE: Patients with incomplete lupus erythematosus (ILE) have lupus features but insufficient criteria for SLE classification. Some patients with ILE transition to SLE, but most avoid major organ involvement. This study tested whether the milder disease course in ILE is influenced by reduced SLE risk allele genetic load. METHODS: We calculated the genetic load based on 99 SLE-associated risk alleles in European American patients with SLE (≥4 American College of Rheumatology (ACR) 1997 criteria, n=170), patients with ILE (3 ACR 1997 criteria, n=169), a subset of patients with ILE not meeting Systemic Lupus International Collaborating Clinics (SLICC) classification (ILESLICC, n=119) and healthy controls (n=133). Unweighted genetic loads were calculated as the total sum of risk alleles for each individual, while weighted genetic loads were defined as the sum of risk alleles multiplied by the natural logarithm of the previously published OR of each risk allele for SLE susceptibility. RESULTS: The median unweighted and weighted SLE risk allele genetic load was significantly greater in patients with ILE (unweighted: 81, p value=0.01; weighted: 16.3, p value=0.001) and patients with SLE (80, p value=0.02; 16.29, p value=0.0006) compared with healthy controls (78, 15.76). Patients with ILESLICC trended towards an increased genetic load, although not statistically significant (unweighted: 80, p value=0.14; weighted: 16.05, p value=0.07). However, the median genetic load did not significantly differ between ILE and SLE, and genetic load did not differentiate patients with ILE and SLE (area under the curve=0.51, p=0.78) by receiver operator characteristic analysis. CONCLUSIONS: Patients with ILE and SLE have comparable genetic loads of SLE risk loci, suggesting similar genetic predispositions between these conditions. Phenotypical differences between SLE and ILE may instead be influenced by ILE-specific variants and gene-environment interactions.

A Flare Risk Index Informed by Select Immune Mediators in Systemic Lupus Erythematosus.

OBJECTIVE: Systemic lupus erythematosus (SLE) is marked by immune dysregulation linked to varied clinical disease activity. Using a unique longitudinal cohort of SLE patients, this study sought to identify optimal immune mediators informing an empirically refined flare risk index (FRI) reflecting altered immunity prior to clinical disease flare. METHODS: Thirty-seven SLE-associated plasma mediators were evaluated by microfluidic immunoassay in 46 samples obtained in SLE patients with an imminent clinical disease flare (preflare) and 53 samples obtained in SLE patients without a flare over a corresponding period (pre-nonflare). SLE patients were selected from a unique longitudinal cohort of 106 patients with classified SLE (meeting the American College of Rheumatology 1997 revised criteria for SLE or the Systemic Lupus International Collaborating Clinics 2012 revised criteria for SLE). Autoantibody specificities, hybrid SLE Disease Activity Index (hSLEDAI) scores, clinical features, and medication usage were also compared at preflare (mean ± SD 111 ± 47 days prior to flare) versus pre-nonflare (99 ± 21 days prior to nonflare) time points. Variable importance was determined by random forest analysis with logistic regression subsequently applied to determine the optimal number and type of analytes informing a refined FRI. RESULTS: Preflare versus pre-nonflare differences were not associated with demographics, autoantibody specificities, hSLEDAI scores, clinical features, nor medication usage. Forward selection and backward elimination of mediators ranked by variable importance resulted in 17 plasma mediator candidates differentiating preflare from pre-nonflare visits. A final combination of 11 mediators best informed a newly refined FRI, which achieved a maximum sensitivity of 97% and maximum specificity of 98% after applying decision curve analysis to define low, medium, and high FRI scores. CONCLUSION: We verified altered immune mediators associated with imminent disease flare, and a subset of these mediators improved the FRI to identify SLE patients at risk of imminent flare. This molecularly informed, proactive management approach could be critical in prospective clinical trials and the clinical management of lupus.

Biomarker panel increases accuracy for identification of an MS relapse beyond sNfL.

BACKGROUND: For relapsing-remitting multiple sclerosis (RRMS), there is a need for biomarker development beyond clinical manifestations and MRI. Soluble neurofilament light chain (sNfL) has emerged as a biomarker for inflammatory activity in RRMS. However, there are limitations to the accuracy of sNfL in identifying relapses. Here, we sought to identify a panel of biomarkers that would increase the precision of distinguishing patients in relapse compared to sNfL alone. METHODS: We used a multiplex approach to measure levels of 724 blood proteins in two distinct RRMS cohorts. Multiple t-tests with covariate correction determined biomarkers that were differentially regulated in relapse and remission. Logistic regression models determined the accuracy of biomarkers to distinguish relapses from remission. RESULTS: The discovery cohort identified 37 proteins differentially abundant in active RRMS relapse compared to remission. The verification cohort confirmed four proteins, including sNfL, were altered in active RRMS relapse compared to remission. Logistic regression showed that the 4-protein panel identified active relapse with higher accuracy (AUC = 0.87) than sNfL alone (AUC = 0.69). CONCLUSION: Our studies confirmed that sNfL is elevated during relapses in RRMS patients. Furthermore, we identified three other blood proteins, uPA, hK8 and DSG3 that were altered during relapse. Together, these four biomarkers could be used to monitor disease activity in RRMS patients.

Relationship Between a Vitamin D Genetic Risk Score and Autoantibodies Among First-Degree Relatives of Probands With Rheumatoid Arthritis and Systemic Lupus Erythematosus.

Objective: Higher 25-hydroxyvitamin D (25(OH)D) levels have been associated with reduced risk for autoimmune diseases and are influenced by vitamin D metabolism genes. We estimated genetically-determined vitamin D levels by calculating a genetic risk score (GRS) and investigated whether the vitamin D GRS was associated with the presence of autoantibodies related to rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) in those at increased risk for developing RA and SLE, respectively. Methods: In this cross-sectional study, we selected autoantibody positive (aAb+) and autoantibody negative (aAb-) individuals from the Studies of the Etiologies of Rheumatoid Arthritis (SERA), a cohort study of first-degree relatives (FDRs) of individuals with RA (189 RA aAb+, 181 RA aAb-), and the Lupus Family Registry and Repository (LFRR), a cohort study of FDRs of individuals with SLE (157 SLE aAb+, 185 SLE aAb-). Five SNPs known to be associated with serum 25(OH)D levels were analyzed individually as well as in a GRS: rs4588 (GC), rs12785878 (NADSYN1), rs10741657 (CYP2R1), rs6538691 (AMDHD1), and rs8018720 (SEC23A). Results: Both cohorts had similar demographic characteristics, with significantly older and a higher proportion of males in the aAb+ FDRs. The vitamin D GRS was inversely associated with RA aAb+ (OR = 0.85, 95% CI = 0.74-0.99), suggesting a possible protective factor for RA aAb positivity in FDRs of RA probands. The vitamin D GRS was not associated with SLE aAb+ in the LFRR (OR = 1.09, 95% CI = 0.94-1.27). The SEC23A SNP was associated with RA aAb+ in SERA (OR = 0.65, 95% CI = 0.43-0.99); this SNP was not associated with SLE aAb+ in LFRR (OR = 1.41, 95% CI = 0.90 - 2.19). Conclusion: Genes associated with vitamin D levels may play a protective role in the development of RA aAbs in FDRs of RA probands, perhaps through affecting lifelong vitamin D status. The GRS and the SEC23A SNP may be of interest for future investigation in pre-clinical RA. In contrast, these results do not support a similar association in SLE FDRs, suggesting other mechanisms involved in the relationship between vitamin D and SLE aAbs not assessed in this study.

High incidence of proliferative and membranous nephritis in SLE patients with low proteinuria in the Accelerating Medicines Partnership.

OBJECTIVE: Delayed detection of LN associates with worse outcomes. There are conflicting recommendations regarding a threshold level of proteinuria at which biopsy will likely yield actionable management. This study addressed the association of urine protein:creatinine ratios (UPCR) with clinical characteristics and investigated the incidence of proliferative and membranous histology in patients with a UPCR between 0.5 and 1. METHODS: A total of 275 SLE patients (113 first biopsy, 162 repeat) were enrolled in the multicentre multi-ethnic/racial Accelerating Medicines Partnership across 15 US sites at the time of a clinically indicated renal biopsy. Patients were followed for 1 year. RESULTS: At biopsy, 54 patients had UPCR <1 and 221 had UPCR ≥1. Independent of UPCR or biopsy number, a majority (92%) of patients had class III, IV, V or mixed histology. Moreover, patients with UPCR <1 and class III, IV, V, or mixed had a median activity index of 4.5 and chronicity index of 3, yet 39% of these patients had an inactive sediment. Neither anti-dsDNA nor low complement distinguished class I or II from III, IV, V or mixed in patients with UPCR <1. Of 29 patients with baseline UPCR <1 and class III, IV, V or mixed, 23 (79%) had a UPCR <0.5 at 1 year. CONCLUSION: In this prospective study, three-quarters of patients with UPCR <1 had histology showing class III, IV, V or mixed with accompanying activity and chronicity despite an inactive sediment or normal serologies. These data support renal biopsy at thresholds lower than a UPCR of 1.

Serologic markers of Epstein-Barr virus reactivation are associated with increased disease activity, inflammation, and interferon pathway activation in patients with systemic lupus erythematosus.

SLE is a clinically heterogeneous disease characterized by an unpredictable relapsing-remitting disease course. Although the etiology and mechanisms of SLE flares remain elusive, Epstein-Barr virus (EBV) reactivation is implicated in SLE pathogenesis. This study examined the relationships between serological measures of EBV reactivation, disease activity, and interferon (IFN)-associated immune pathways in SLE patients. Sera from adult SLE patients (n = 175) and matched unaffected controls (n = 47) were collected and tested for antibodies against EBV-viral capsid antigen (EBV-VCA; IgG and IgA), EBV-early antigen (EBV-EA; IgG), cytomegalovirus (CMV; IgG), and herpes simplex virus (HSV-1; IgG). Serological evidence of EBV reactivation was more common in SLE patients compared to controls as demonstrated by seropositivity to EBV-EA IgG (39% vs 13%; p = 0.0011) and EBV-VCA IgA (37% vs 17%; p = 0.018). EBV-VCA, CMV1, and HSV-1 IgG seropositivity rates did not differ between SLE patients and controls. Furthermore, concentrations of EBV-VCA (IgG and IgA) and EBV-EA (IgG) were higher in SLE patients. SLE patients with high disease activity had increased concentrations of EBV-VCA IgA (mean ISR 1.34 vs. 0.97; p = 0.041) and EBV-EA IgG levels (mean ISR 1.38 vs. 0.90; p = 0.007) compared with those with lower disease activity. EBV reactivation was associated with enhanced levels of the IFN-associated molecule IP-10 (p < 0.001) and the soluble mediators BLyS (p < 0.001) and IL-10 (p = 0.0011). In addition, EBV-EA IgG responses were enriched in two previously defined patient clusters with robust expression of IFN and inflammatory or lymphoid and monocyte responses. Patients in these clusters were also more likely to have major organ involvement, such as renal disease. This study supports a possible role for EBV reactivation in SLE disease activity.

Adults with systemic lupus exhibit distinct molecular phenotypes in a cross-sectional study.

BACKGROUND: The clinical and pathologic diversity of systemic lupus erythematosus (SLE) hinders diagnosis, management, and treatment development. This study addresses heterogeneity in SLE through comprehensive molecular phenotyping and machine learning clustering. METHODS: Adult SLE patients (n = 198) provided plasma, serum, and RNA. Disease activity was scored by modified SELENA-SLEDAI. Twenty-nine co-expression module scores were calculated from microarray gene-expression data. Plasma soluble mediators (n = 23) and autoantibodies (n = 13) were assessed by multiplex bead-based assays and ELISAs. Patient clusters were identified by machine learning combining K-means clustering and random forest analysis of co-expression module scores and soluble mediators. FINDINGS: SLEDAI scores correlated with interferon, plasma cell, and select cell cycle modules, and with circulating IFN-α, IP10, and IL-1α levels. Co-expression modules and soluble mediators differentiated seven clusters of SLE patients with unique molecular phenotypes. Inflammation and interferon modules were elevated in Clusters 1 (moderately) and 4 (strongly), with decreased T cell modules in Cluster 4. Monocyte, neutrophil, plasmablast, B cell, and T cell modules distinguished the remaining clusters. Active clinical features were similar across clusters. Clinical SLEDAI trended highest in Clusters 3 and 4, though Cluster 3 lacked strong interferon and inflammation signatures. Renal activity was more frequent in Cluster 4, and rare in Clusters 2, 5, and 7. Serology findings were lowest in Clusters 2 and 5. Musculoskeletal and mucocutaneous activity were common in all clusters. INTERPRETATION: Molecular profiles distinguish SLE subsets that are not apparent from clinical information. Prospective longitudinal studies of these profiles may help improve prognostic evaluation, clinical trial design, and precision medicine approaches. FUNDING: US National Institutes of Health.