←
Bryson Lab832 posts
Bryson Lab avatar

Bryson Lab

@TheBrysonLab

Engineering innate immunity // Host-pathogen interactions // Systems Biology // Tuberculosis @MIT @MITDeptOfBE

◷ Joined September 2017▧ 84 Photos and videos
Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Caveats we’d flag ourselves: 2 hits is little to train on, and 2/6 is still a lot of misses. Antigens below our LOD (~1-10 copies/cell) may exist, though at that level it’s not obvious they’d make good targets. Preprint: biorxiv.org/content/10.64898…

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Big picture: presentation looks like a series of filters. Not TB-specific. The same logic may apply to other pathogens and to tumor antigens cross-presented by DCs.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Test of the framework: 186 HLA-A*24:02 binders, narrowed to 6 high scorers with favorable C-terminal flanks and 4 low scorers. 2/6 high scorers detected. 0/4 low scorers. PPV from <1% to ~33%, with a panel 1/30th the size.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Expression can’t explain why one EsxA peptide is detected and another isn’t. Looking outside the epitope: Mtb MHC-I peptides are enriched for small residues (A/T/S/G/D) at P1′-P2′. Self peptides show no such bias.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Expression dominated. Binding + expression took AUPRC from 0.25 to 0.92, and the detected antigens sit near the top of the intracellular Mtb transcriptome. But many highly expressed genes yield no detected peptides, so secretion and abundance both seem to be needed.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

So what separates 2 from 284? We scored each peptide on MHC binding, proteolytic processing (flanking sequence), and expression of the source gene in intracellular Mtb (dual host-pathogen RNA-seq), then trained small interpretable classifiers.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Not a sensitivity problem, as far as we can tell. Smaller panels, 10x more cells, and spiked heavy pMHC standards all pointed the same way: LOD ~1-10 copies/cell, the two hits unremarkable. 88% of targeted self peptides were detected in the same runs.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

We took 286 T7SS-derived 9-mers scored as HLA-A*02:01 binders by yeast display and asked by SureQuant targeted MS which were detectable in the MHC-I repertoire of infected macrophages. Two: LLDEGKQSL (EsxA) and LLDAHIPQL (EsxGS).

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Our prior: Mtb peptides detected on MHC-I come overwhelmingly from type VII secretion system (T7SS) substrates, which likely reach the host cytosol. So localization is one filter, MHC binding another. We assumed those two got you most of the way.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Why this matters beyond TB: every T cell vaccine, peptide or full-length protein, rests on antigen selection. If the antigen you pick isn’t presented by the cells you need T cells to recognize, you may not get the protective immunity you were aiming for.

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Oct 23

Antigen discovery and TB vax friends 🧵 We set out to find more Mtb peptides presented on MHC-I. We had a strong biological prior, a high-throughput binding screen, and a very sensitive targeted MS assay. Of 286 candidates, we detected 2. biorxiv.org/content/10.64898…

Bryson Lab avatar
Bryson Lab@TheBrysonLab·May 22

Anyone in my science network have experience with producing soluble TCRs? Better yet affinity maturation of TCRs?

Bryson Lab avatar
Bryson Lab@TheBrysonLab·Jan 28

Anyone in the de novo minibinder design space using their binders as flow cytometry staining reagents? Would love to compare notes if so!

Bryson Lab avatar
Bryson Lab@TheBrysonLab·20 Nov 2025

We're excited! This study supports the concept that there are multiple ways to transfer proteins from the phagosome to the cytosol, and that Mtb specifically takes advantage of this.