Skip to content

Kevin Chao — TimeVault: a vault-protein recorder of the cellular transcriptome

Speaker: Kevin (Yu-Kai) Chao — PhD student, Department of Chemistry, Harvard University / Broad Institute (Fei Chen Lab; Broad Cancer Program). Introduced by Nathan Nakatsuka (Satija Lab). Talk title (official): "TimeVault: A genetically encoded device for transcriptome storage in mammalian cells" Slot: SCGD26, Jun 12, 2026 — 3:20–3:40 PM Method: TimeVault — engineered vault ribonucleoprotein particle that captures + preserves a past transcriptome snapshot in living cells Paper: Chao Y-K, Wu M, Gong Q, Chen F. Science (2026), doi:10.1126/science.adz9353 — plasmids deposited on Addgene (e.g. #242118, full set: article collection) Status: Captured from session transcript

Thesis

To ask why a cell survives a drug, you need its prior (drug-naive) state — but RNA measurement is destructive and drug selection already alters the cell. TimeVault keeps a "time capsule" record of the transcriptome before perturbation, carried through to after, linking past state to future fate (cancer persister cells: drug-tolerant "persister" states pre-exist treatment).

How it works

Built on the vault — a large, abundant (3,000–15,000/cell), non-essential ribonucleoprotein whose major vault protein (MVP) self-assembles (78 MVP copies; native cargo = a poly-ADP-ribose polymerase with an "INT" interaction domain). Fusing the INT domain to a protein packages it into the vault.

TimeVault fuses INT to a poly(A)-binding protein → captures mRNA into the vault particle, isolating it from cytosolic mRNA to preserve a past transcriptome. Recording window is controlled by a Tet-OFF promoter (add doxycycline to stop capture).

Key results

  • Captures RNA: detectable only when both components (MVP + poly(A)-binding) are present; preserves ~1–3% of housekeeping mRNA (GAPDH, ACTB) and protects it from degradation in lysate.
  • Controllable window: doxycycline stops capture (MVP mRNA half-life ~5.4 h); 4sU labeling shows vault-captured RNA stays high over multiple days while the lysate transcriptome turns over in <1 day; RNA-seq shows captured transcriptome correlates well with total lysate (comparable gene counts) → ~1–2 transcripts/particle.
  • Records a past state: heat-shock-then-recover experiment — the recorded fraction shows up-regulated heat-shock response only in the heat-shock-past condition, while the present transcriptome is normal (cell recovered). Clean separation by GO enrichment.
  • Cancer persisters (PC9, EGFR-mutant NSCLC + osimertinib): retrieved the drug-naive transcriptome of cells that later survived. Persisters pre-show up-regulated OXPHOS, reduced division (reduced Warburg), adhesion, detox enzymes, IL-6. Top marker fibronectin (FN1) validated by immunostaining (a rare FN1-high subpopulation pre-exists; majority becomes FN1-high after drug). scRNA on drug-naive PC9 confirms a middle population enriched for FN1/PI3 etc. siRNA against FN1 (and other persister markers) sensitizes cells to osimertinib (combination kills more than drug alone).

Connections to the corpus

  • A third "molecular recorder" on the day — pairs with GEMINI (protein-assembly temporal recorder) and SynapseSeq (barcode recording); the past↔future framing echoes Aviv Regev — lab-in-the-loop.
  • Cancer drug-tolerant-persister biology → AACR clinical / single-cell axes.

Sources

Open questions

  • No separate bioRxiv preprint located — the work appears to have been published directly in Science (2026). (Note: the 2023 bioRxiv "Molecular Time Capsules," 10.1101/2023.10.12.562053, is a different group — Gene-Wei Li lab, MIT — not TimeVault.)
  • Introducer (Nathan Nakatsuka) is from the session transcript, not independently confirmed.