Back

Gut Microbiota Modulates and Predicts Disease Severity in Experimental Pemphigoid Disease

Liu, X.; Patzelt, S.; Ma, Y.; Cepic, A.; Dikmen, H.-O.; Bieber, K.; Groussin, M.; Poyet, M.; Ibrahim, S.; Schmidt, E.; Baines, J. F.

2025-10-01 microbiology
10.1101/2025.10.01.679837 bioRxiv
Show abstract

Pemphigoid diseases (PD) are autoimmune blistering diseases with reported alterations in skin and gut microbiota, though their causal contribution to disease pathophysiology remains unclear. Using a passive model of bullous pemphigoid-like epidermolysis bullosa acquisita (BP-like EBA), we compared C57BL/6J mice from two sources (in-house vs. Charles River) that differed in their baseline microbiota. Charles River mice developed significantly less severe disease. Co-housing led to partial homogenization of the gut microbiota, driven by asymmetric transfer of taxa from Charles River to in-house mice, which corresponded with reduced disease severity in the latter. The skin microbiota, however, showed limited exchange. Disease severity was inversely associated with gut microbial alpha diversity, with protective taxa such as Lactobacillus intestinalis and Parabacteroides distasonis enriched in Charles River mice, while pro-inflammatory taxa including Turicimonas muris and Muribaculum intestinale were enriched in in-house mice. A machine learning model further identified a gut taxon most closely matching the candidate genus Scatocola as a strong negative predictor of disease severity. Through experimental exposure and uptake of variable gut microbiota, these findings suggest a direct role of gut microbiota in mediating the severity of the experimental BP-like EBA and highlight the potential of microbiota-based strategies for therapeutic intervention in PD.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.