Enormous thanks to my co-first author Pragun Acharya, the Weaver Lab, and to my mentor Casey. A big thank you to Xi Lin and Richard Blumberg who wrote a generous News & Views alongside it. Always happy to talk about this work, questions welcome.

Does it matter? Mice primed with cytosol-delivered antigen were protected against a heterologous Salmonella carrying the same epitope. A T3SS that evolved for virulence doubles as an antigen delivery route the host exploits.

Programming consequences were clear. MHCII-dependent interactions were mostly required for IFNγ-responsive regulons in colonocytes. TRM formed without epithelial MHCII drifted toward central memory and migratory signatures instead of resident ones.

H2-Ab1 expression peaked in distal colonocytes at d14 and was still up at d90, long after pathogen clearance. Ligand-receptor analysis from scRNA-seq showed enhanced Cxcl16-Cxcr6 and Il7-Il7r interactions in controls, residency and survival cues coming from the epithelium.

Antigen location also set where memory lived. Light-sheet imaging found epithelium-associated cells still there at day 240. Without epithelial MHCII, or with surface antigen, the few left sat in GALT, as you would expect from conventional DC priming.

Does epithelial presentation make TRM or just track with it? Deleting MHCII from the epithelium cost both CD69+CD103− and CD69+CD103+ subsets. Deleting it after clearance cost them too, putting epithelial MHCII in maintenance, not just priming.

Where does it start? Colonization was strikingly focal, confined to the cecal apex beside the cecal patch. SMARTA responses appeared there first, then spread to distal colon and draining nodes.

We fused the LCMV gp66-80 epitope to intimin (stays on the bacterium), tir (translocated, apical membrane), and nleA/espZ (translocated to the cytosol). Same epitope, same TCR, three different sub-cellular locations. The cytosolic pair drove far larger colonic CD4 responses.

Out this week in Nature Immunology: We asked whether where a bacterial antigen ends up inside an IEC determines the CD4 T cell response it generates. Isogenic C. rodentium strains, same LCMV epitope, different destinations. The answer turned out to extend all the way to tissue-resident memory.