Machine-learning-based motion tracking revealed the inverse correlation between adhesivity and surface motility of the leptospirosis spirochete

Abstract

Bacterial motility is generally a critical virulence factor for pathogenic species, and thus studies on bacterial motility are significant given that they elucidate the mechanisms of infection and disease. Although fluorescent labeling has been the mainstream approach to detecting individual bacteria in a population or in host tissues, it contains problems related to protein expression stability and interference with bacterial physiology. Here, we applied machine learning to microscopic image analysis to achieve label-free motion tracking of the zoonotic bacterium Leptospira interrogans on cultured animal cells. The label-free method allowed us to measure various leptospiral strains isolated from human and animal patients, and natural reservoirs, showing that fast mobility on kidney cells tends to result in severe symptom. We also analyzed the surface locomotion of mutant strains lacking outer membrane proteins (OMPs) and revealed that the loss of OMPs reduced adherence and facilitated motility on cultured kidney cells. The behavior of clinical isolates and OMP-deficient mutants on cultured cells showed the inverse correlation between adhesion and mobility, which could affect infection outcomes. Thus, our computer vision technique eliminated the restriction on available bacterial strains and provided information that could help in understanding the mechanisms underlying motility-dependent bacterial pathogenicity.