Background/Purpose: Anti-Ro autoantibody production often precedes the development of Systemic Lupus Erythematosus (SLE) or Sjogren’s syndrome (SS) by years. For anti-Ro+ mothers enrolled in the Research Registry for Neonatal Lupus (RRNL), progression to SS or SLE occurs in about a quarter, while most remain asymptomatic or develop only minor rheumatic symptoms (Asym/UAS). Thus, RRNL mothers uniquely offer a promise to identify genotype-phenotype relationships that are important to preclinical autoimmunity. Since multiple SLE risk alleles from Class II HLA genes are present in anti-Ro+ mothers, we examined interactions of specific microbiome taxa with Class II HLA by independent analytic paths with the goal to identify HLA-related microbiome associations in Anti-Ro+ Mothers of Children with Neonatal Lupus.

Methods: Subjects included 125 RRNL mothers and 23 healthy controls. Stool microbiomes of anti-Ro+ women in RRNL (Asym/UAS, SS/SLE), and healthy controls (HC) were processed using 16S ribosomal RNA sequencing. Sera/plasma were evaluated for cytokines and autoantibody levels. Alleles from HLA Class II genes were genotyped using NextGen sequencing of HLA region or imputed (HIBAG) from GWAS data. Independent analytic paths sought to explore associations of specific taxa and class II HLA included: 1) use of a cumulative logit model to test interactions between FDR significant genera and HLA alleles and 2) assignment of SLE, SS, UAS patients and HC to molecular phenotype clusters by Random Forest (RF), an unsupervised machine learning tool using Z-score transformed cytokine soluble mediators and autoantibody values with settings and the gap statistic that were used to estimate the optimal number of patients and HC within clusters. The overlapping distribution of SS/SLE, HLA alleles and taxa at clusters were then examined.

Results: Findings related to DRB1*15:01 and an interaction with genera of the Ruminococcaceae family were tested. Oscillibacter, with FDR-adjusted significance was shown to exhibit evidence of an interaction (P=0.033 (OR=0.60 (0.37-0.96)). In order to authenticate that SLE HLA risk alleles modify the strength of the association, we examined the molecular phenotype clusters from RF clustering. Radar plots were used to visualize the distribution HLA alleles and the enrichment of microbiome taxa within these clinically relevant phenotypic clusters (Figure 1). DRB1*1501 shows enrichment at cluster 4. Interestingly, the distribution of Oscillibacter, but not Coprococcus 3 was nearly superimposable with the Class II HLA allele with enrichment at cluster 4. However, the distribution of DRB1*1501 was not enriched at cluster profiles representing evaluations of DRB1*0301 and SS/SLE disease classification (Figure 2) demonstrating a limitation of DRB1*1501 to predict risk for transition from benign to pathologic autoimmunity in anti-Ro+ mothers of children with neonatal lupus.

Conclusion: These data support the use of molecular phenotypes that are linked to genetic-environmental interactions to identify HLA-related microbiome associations.

Figure_HLA_USE.jpeg”

SLE_by_cluster.jpeg”

ACR Meeting Abstracts - https://acrabstracts.org/abstract/development-of-biomarker-models-to-identify-hla-related-microbiome-associations-in-anti-ro-mothers-of-children-with-neonatal-lupus/