Sequence-dependent co-condensation of Lsr2 with DNA elucidates the mechanism of genome compaction in Mycobacterium tuberculosis

The xenogeneic silencer protein Lsr2 from Mycobacterium tuberculosis plays a critical role in its survival and pathogenesis. Lsr2 is a nucleoid-associated protein (NAP) that interacts with DNA in vivo and regulates many genes. Purified Lsr2 forms nucleoprotein filaments with DNA molecules, leading to highly compacted DNA conformations. However, the physical mechanism underlying Lsr2-mediated DNA compaction, resulting in gene regulation, remains elusive. We employed a combination of biochemical assay, single-molecule imaging, and molecular dynamics simulations to investigate the governing principles of Lsr2-mediated DNA compaction. We show that, while Lsr2 alone undergoes phase separation, addition of DNA substantially lowers the required concentration for its phase separation. Strikingly, our single-molecule and simulation data establish that Lsr2 forms condensates with long stretches of AT-rich DNA, providing strong evidence for sequence-dependent co-condensation. We further validate our findings by carrying out in vivo imaging of endogenously expressing Lsr2 tagged with eGFP in Mtb cells. This observation is contrary to the classical view of sequence-dependent binding of individual protein molecules to DNA; our findings rather suggest that protein-DNA co-condensates "sense" the average binding energy landscape. We present a physical model for Lsr2-mediated DNA compaction and gmycene regulation, describing a novel mechanism for NAP-mediated genome organization in bacteria.