A final thank you to all the authors: Laurel, @amardeeep.bsky.social, Nathan, Lindsay, Toni, Layla, @kevincorbett.bsky.social, and @aaronwhiteley.bsky.social. This work also used many shared facilities at CU, particularly the incredible folks at CU Research Computing and the Alpine HPC cluster.

This work has allowed us to understand how a bacterial NLR-related protein can sense multiple unrelated phage proteins. We’re excited for the other insights to come as we and others continue to apply this strategy to understand host-pathogen interactions throughout the tree of life!

Finally, we wanted to know whether we could use this strategy to find activators of other defense systems. We screened a total of 15 (for now 😉) bNACHT proteins and identified putative activators for many of them

Some questions that remain: What are these phage proteins doing? How could bNACHT11 have evolved to bind to all of them? These areas of active investigation.

However, we found that a subtle mutation in the bNACHT11 sensor domain enabled phage mutants to now escape detection by this defense system. This suggested that bNACHT11 has many activators during infection! In doing so, we imagine it would be difficult for a phage to evade this system.

We next wanted to understand whether these activators were important for bNACHT11 function during phage infection. We knocked out several of these phage proteins in various combinations, but nothing we did allowed the phage to escape detection by wild-type bNACHT11!

We further confirmed that these activators directly bind bNACHT11 and cause oligomerization of this system. Similar to the mammalian NLRs that form inflammasomes, this oligomerization requires ATP binding. We’re not sure how oligomerization leads to phage protection, but we’re working on it!

We next wondered: is AlphaFold right? Are there actually this many proteins that bind to bNACHT11? Excitingly, we found that indeed, several of these phage proteins activate bNACHT11 in coexpression and cotransformation assays.

We ranked predicted interactions by their weighted pTM score as calculated by AlphaFold-multimer and found that several phage proteins were predicted to interact with this system.

We first clustered the proteins in 91 related phage genomes to build a nonredundant dataset, then screened each of those proteins for interactions with a defense protein called bNACHT11 using AlphaFold-multimer. (Shoutout to AlphaPulldown which helped us be more efficient with our computation.)

Recent work has shown that NLR-related proteins are antiphage systems that encode a variety of sensor domains. Some of the triggers of these systems have been identified, but many more remain unknown. We wondered: can we build a computational pipeline to identify activators of these defense system?