We are recruiting! If you are passionate about technology development, protein engineering, computational design, directed evolution, chemical biology - please reach out!
(The setting is pretty nice too…)
Co-led by extraordinary trainees Dr. Bo Cai and Andrew Xue, with collaborators Rogelio Hernández-López and postdoc Qian Xue. Xiaojie Qiu and his postdoc Nianping Liu applied Dynamo to analyze CAR T state transitions. Thanks to NSF, CIRM, and Biohub for funding.
www.biorxiv.org/cgi/content/...
To develop scAPEX-seq, we first redesigned bulk APEX-seq. Our updated method, "APEX-seq2", uses a phenol-azide probe + on-bead amplification to achieve 10x higher RNA recovery, 10x fewer input cells, half the processing time—all while preserving spatial specificity. Full protocol in our Methods.
Clinical relevance: Using the Cancer Immunology Data Engine (CIDE; 90 datasets, 8,575 tumors), high CTSW expression—largely in T and NK cells—correlates with improved immunotherapy response and better overall survival across multiple cancers including melanoma and lung cancer.
Overexpressing CTSW in CAR T cells → larger fraction of stem-like memory phenotype (CD62L+CD45RA+), increased proliferation, and dramatically better tumor killing over repeated rounds of antigen challenge!
Next: CAR T + tumor cells. Comparing 2hr vs 21-day cocultures, scAPEX-seq revealed a distinct "late effector" CAR T state missed by whole-transcriptome seq—with higher cytotoxic transcripts and reduced exhaustion markers. One gene enriched in this population: cathepsin W (CTSW).
First test: macrophage-tumor cocultures, scRNA-seq vs scAPEX-seq (APEX targeted to ER membrane). The difference was striking: scAPEX-seq showed greater cluster diversity and clearly separated cocultured from monocultured cells. ~10x more DEGs and far more ligand-receptor interactions detected.
This is especially relevant for cell-cell interactions (CCIs), mediated by surface/secreted proteins translated at the ER. We hypothesized APEX enrichment of ER transcripts would detect CCI-driven RNA localization changes AND boost sensitivity for low-abundance transcripts vs standard scRNA-seq.
Why sequence compartment-specific RNAs instead of whole-cell transcriptomes?
Because RNA localization controls splicing, translation, and degradation. Changes in RNA localization (independent of abundance changes) are undetectable by conventional scRNA-seq.
Today we report single-cell APEX-seq (scAPEX-seq)—a method for unbiased mapping of subcellular transcriptomes at single-cell resolution. It reveals cell states invisible to standard scRNA-seq and identifies regulators of CAR T function that improve solid tumor killing.
tinyurl.com/32pf6b8p
Video introduction to our new “Conformational Biasing” method for computational design of mutations that bias proteins towards desired conformational states
CB part starts at 14:55
Thanks to Peter Cavanagh and Andrew Xue – amazing graduate students who co-led this work
And we had a great editor + helpful/thoughtful reviewers who clearly invested a great deal of effort to help us make this work better – much gratitude to them as well.
Co-led by two exceptionally talented grad students, Peter Cavanagh and Andrew Xue. Co-authors @shizhong-dai.bsky.social, Andrew Qiang, and Tsutomu Matsui also made invaluable contributions to the study.
CB features: ✅ Diverse protein types (receptors, enzymes, synthetic binders) ✅ Monomers & oligomers ✅ Natural or synthetic (no MSA needed) ✅ Input: 2 structures (PDB or AlphaFold) ✅ Fast: ~3k variants in <1 min (RTX 4090) ✅ Scores single & multi-mutants
Code & user-friendly interface here:
Unexpected finding: LplA's conformational equilibrium controls promiscuity.
🔒 Closed-biased variants: Less off-target labeling than WT (better for site-specific tagging).
🔓 Open-biased variants: Highly promiscuous (laying groundwork for new proximity labeling tech).
We then applied CB to lipoic acid ligase (LplA), a conformation-switching metabolic enzyme from E. coli.
We validated that CB-predicted mutations had the intended effect using SEC-SAXS and Trp fluorescence measurements of conformational occupancy.
...stabilize viral epitopes for vaccine development.
We validated CB on 7 deep mutational scanning datasets.
We successfully predicted variants with conformation-specific function (e.g., enhanced binding/activity) for: • K-Ras • SARS-CoV-2 spike • β2AR • 4 enzymes (Src, B-Raf, FabZ, MurA)
Most natural proteins alternate between distinct conformations, each associated with specific functions. If we could design point mutations that stabilize one conformation relative to another, it could enhance the signaling of G proteins or GPCRs, increase the catalytic activity of enzymes, or....
Can we design mutations that bias proteins towards desired conformational states?
Today in @science.org, we introduce Conformational Biasing (CB), a simple and scalable computational method that uses contrastive scoring by inverse folding models to identify conformation-biasing mutations.
P.S. While "protein-only" editors (like those from David Liu) are highly effective, the programmability and versatility of a true CRISPR-based mtDNA system offers huge potential benefits—if we can improve efficiency and translate it to mammalian cells.
This project was a marathon. Kudos to Sifei whose determination and resourcefulness were crucial at every stage, and co-advisor Dan Jarosz who co-led this ambitious project.
Graduate student Sifei Yin obtained these exciting results several years ago, and spent the intervening time understanding how sgRNA import works, and how to make it better. Read the paper to learn what she uncovered!
We started with yeast. We engineered a strain that must repair a STOP codon in mtDNA (an arginine biosynthesis gene) to survive. After screening a library of sgRNA import sequences, we identified one that enables a small degree of functional CRISPR editing in mitochondria.
Yet many organisms have evolved mechanisms to transport RNAs across their mitochondrial membranes. Trypanosomes, for example, do not encode any tRNAs in their mtDNA, and must import all of them from the cytosol to support intra-mitochondrial protein translation.
The CRISPR toolbox has revolutionized the study of nuclear DNA, but the mitochondrial genome (mtDNA) has remained out of reach, mainly because there are no known ways to deliver sgRNAs across both outer and inner mitochondrial membranes.
New work describes our efforts to achieve CRISPR editing of the mitochondrial genome.
www.biorxiv.org/lookup/conte...
I'm thrilled to share our work from @weissmanlab.bsky.social (www.sciencedirect.com/science/arti...). We developed LOCL-TL, an optogenetic approach for monitoring localized translation in mammalian cells. LOCL-TL revealed two distinct strategies for mitochondrially localized translation.
LOV-BirA, light regulated biotin ligase, engineered by Song-Yi Lee
www.sciencedirect.com/science/arti...
Proximity-specific ribosome profiling using LOV-BirA reveals two distinct strategies for mitochondrially-localized translation: one for long coding sequences and one for short. It was a pleasure to contribute to this beautiful work from Jonathan Weissman @weissmanlab.bsky.social and Jingchuan Luo