🧬 Aging Wiki

Targeting Cancer-Specific Mutations with RNA-Triggered Chromatin Shredding

9 min read

Zeng et al. 2026 — Transcript-activated chromatin shredding (CRISPR-Cas12a2)

Nature, advance online 8 June 2026. doi:10.1038/s41586-026-10738-7. Corresponding author Jennifer A. Doudna (Gladstone Institutes / Innovative Genomics Institute, UC Berkeley / HHMI). Open-access preprint: bioRxiv doi:10.64898/2026.05.08.723607.

Access: Nature version paywalled. This page was verified against the openly-licensed (CC-BY) bioRxiv preprint (DOI 10.64898/2026.05.08.723607) full text on 2026-06-13; all quantitative claims confirmed. Minor editorial differences between the preprint and the final Nature version cannot be excluded.

What this paper is

A primary research paper introducing “transcript-activated chromatin shredding”: a programmable way to kill a chosen cell by its RNA signature. It repurposes CRISPR-Cas12a2 — an RNA-guided nuclease from a bacterial abortive-infection system — so that, when its guide RNA recognises a specific target messenger RNA inside a mammalian cell, the enzyme switches into a non-specific (“trans”/collateral) nuclease that shreds that cell’s own chromatin, triggers a DNA-damage response, and kills it. The transcript is the address: only cells expressing the trigger RNA die.

For the molecular technique itself (mechanism, parameters, limitations) see cas12a2-trans-cleavage; for the modality framing (targeted programmable cell ablation as a gene-therapy class) see transcript-activated-cell-ablation. This page records the primary findings.

The paper is a companion to the foundational eukaryotic-cell-killing demonstration Scholz et al. 2026 (Nature, doi:10.1038/s41586-026-10466-y) 1; both build on the original characterisation of Cas12a2 abortive infection, Dmytrenko et al. 2023 2.

Mechanism established

  • Cas12a2 cleaves chromatin in trans on RNA recognition (in vitro). Purified Sulfuricurvum SuCas12a2 ribonucleoproteins (RNPs) degraded fluorescently-labelled RNA and double-stranded DNA, and cleaved supercoiled plasmid in trans, only when the cognate target RNA was present; non-targeting guide RNA controls were inert. A chromatinised 10.6-kb plasmid was degraded more slowly than naked DNA, producing a mono-/di-/tri-nucleosome ladder — i.e. preferential cleavage of internucleosomal linker regions. RNPs also degraded chromatin in nuclei extracted from HEK293 cells (targeting guide only) 3.
  • Activation kills mammalian cells via a DNA-damage response. Nucleofecting SuCas12a2 RNPs targeting GFP mRNA into HEK293-GFP cells (10–100 pmol) halted proliferation, with enlarged/fragmented nuclei by 72 h (mitotic catastrophe) and increased γH2AX (double-strand-break marker) and phospho-KAP1 S824 (heterochromatic-damage marker). In genome-engineered RPE1 cells the DNA-damage reporter rose within 4 h, peaking 24–48 h 3.
  • Killing scales with target-transcript abundance. In HEK293 clones with ~550 (Low), ~1,100 (Mid), or saturating (High) GFP transcripts/cell (single-molecule FISH), a low 5-pmol guide dose killed High cells but only mildly affected Mid/Low; GFP-negative cells were unaffected 3.
  • Activity is sufficient at mammalian Mg²⁺. Although slower at physiological free Mg²⁺ (~0.1–1 mM) than at the 10 mM used in bacterial biochemistry, trans DNA cleavage still occurred at 0.1 mM, and killing efficiency was Mg²⁺-independent across 0.1–10 mM RNP pre-incubation 3.

Selectivity — killing cancer cells by their transcript signature

Three orthogonal selectivity levers were demonstrated, all in human cell lines:

  1. Oncogene over-expression (abundance threshold). In U2OS cells with doxycycline-inducible cyclin E1 (~50 vs ~560 CCNE1 transcripts/cell), some guides killed regardless of level, but two guides killed only the high-expressing (induced) state — a targeting window distinguishing high from low expression of the same transcript. MYC was also targetable 3.
  2. In-frame indel junction. PC9 non-small-cell-lung-cancer cells carry an endogenous EGFR exon-19 deletion (E746_A750del, a 15-bp deletion). A guide spanning the deletion junction (a sequence absent from normal cells) caused robust growth arrest + DNA-damage markers in PC9 but had no effect on EGFR-wild-type RPE1 cells 3.
  3. Single-nucleotide variant via a mutation-created PFS. SuCas12a2 requires an adenine-rich protospacer-flanking site (PFS) 3′ of its RNA target (analogous to a PAM). The TP53 R248Q mutation is a G→A substitution that itself creates the activating “mutA” PFS — so the enzyme fires only on the mutant transcript. Transfecting an R248Q guide into Cas12a2-expressing RPE1 p53^R248Q cells caused robust growth arrest, with no change in wild-type RPE1 3.

Specificity controls (the strong part of the paper):

  • In a 1:1 competitive co-culture of RPE1 p53^R248Q (mCherry) vs RPE1 wild-type (GFP), a non-targeting guide left the mutant outgrowing wild-type (GFP/mCherry ratio ~0.62 at 96 h); the R248Q guide reversed this to ~4.5 (wild-type outgrew mutant) 3.
  • A control mutant, p53^R175H (identical to wild-type at the R248 position), showed no growth change with the R248Q guide — and selectivity held despite R175H cells expressing more TP53 transcript (~500/cell) than R248Q cells (~400/cell) and far more than wild-type (~25/cell). So selectivity tracks the mutation, not expression level. Phospho-KAP1 rose only in R248Q cells; FUCCI imaging showed extended G2 arrest + fragmented nuclei (mitotic catastrophe) 3.
  • Endogenous setting: PC9 cells (endogenous p53^R248Q, ~83 transcripts/cell) treated with the R248Q guide showed growth defects, DNA-damage markers, and propidium-iodide⁺/Annexin-V⁺ death by 96 h 3.
  • Additional TP53 mutations targeted: R280K and E285K (both G→A), and M246I (a G→C single-nucleotide variant) in NCI-H23 cells. A single guide-target mismatch can reduce activity, which the authors exploit for discrimination 3.

Targetability across patients. Analysing TP53 coding mutations in 16,708 patient samples: 25.7% are B-to-A substitutions (B = G/C/T) that could create or enhance an activating PFS, and 73.2% carry an A, ABA, or AA motif within 24 nt of the 3′ end usable as a PFS 3.

In vivo (mouse) — LNP delivery suppresses tumours

  • Delivery as mRNA. Capped, pseudouridylated mRNA encoding nuclear-localised Cas12a2 (in-vitro transcribed), co-delivered with the R248Q guide (transfection or lipid nanoparticles, LNPs), reproduced selective killing of PC9 cells 3.
  • MYC-driven liver tumours (genetically-engineered model). In tumour-prone FVB/NJ mice with transposase-integrated MYC, LNPs co-packaging Cas12a2 mRNA + an anti-MYC guide (treatment from day 6) reduced tumour surface area (% of liver) and liver-to-body-weight ratio versus controls 3.
  • PC9 lung xenografts. SORT (lung-tropic) LNPs delivered cargo to ~7–18% of xenograft cells (modest, hence multi-dosing). Six doses of Cas12a2 mRNA + R248Q-guide LNPs reduced tumour progression (bioluminescence) versus controls. In an advanced model (1 million cells, 21-day pre-treatment growth), the LNPs did not shrink established lung-tumour burden but delayed metastasis formation 3.

Significance for the wiki

This is a cancer paper; it does not test senescence or aging. Its relevance here is as a new programmable cell-ablation modality that is conceptually orthogonal to everything in the current senolytic toolkit, plus one direct p53-aging cameo.

  1. A fourth senolytic-adjacent killing mechanism. Every small-molecule senolytic in the wiki kills by exploiting senescent-cell anti-apoptotic dependencies (the SCAP framework; see senolytics). Transcript-activated shredding kills by RNA identity instead — independent of apoptotic priming. It would sit alongside BH3-mimetics, uPAR-targeted CAR-T cells, and suicide-gene systems (Oisin; INK-ATTAC/p16-3MR proof-of-concept) as a distinct fourth route. See transcript-activated-cell-ablation for that framing and its (currently hypothetical) senescence application.
  2. It plugs into — and is bounded by — the senotype problem. The just-ingested SenNet work (suryadevara-2026-senotypes, anerillas-2026-sencat) shows senescent cells are heterogeneous with no universal marker. A transcript-addressed killer is exactly what “senotype-targeted senotherapeutics” would require — but the same heterogeneity is the obstacle: there is no senescence-exclusive trigger transcript (p16/CDKN2A is not senescence-specific), and abundance-thresholding would need a senescence transcript reliably above the killing threshold only in senescent cells. See cellular-senescence § Senescent-cell heterogeneity and the senotype concept.
  3. A clean p53-aging tradeoff cameo. The authors deliberately kill p53-mutant cells rather than restore/activate p53 — explicitly because unintended p53 activation in healthy cells “can induce senescence and whole-genome duplication.” That is the same antagonistic-pleiotropy tradeoff the wiki documents at tyner-2002-p53-mutant-aging and p53 (tumour-suppression vs accelerated aging).

Limitations and gaps

  • No aging/senescence data. Every selectivity demonstration uses a cancer-defining transcript (mutant TP53, EGFR deletion, amplified oncogene). Transfer to senescent-cell clearance is unproven and non-trivial (point 2 above). gap/no-mechanism (for the aging application specifically)
  • Delivery is the bottleneck. In-vivo efficacy depended on LNPs; tumour-cell delivery was ~7–18%, requiring repeat dosing, and the system failed to reduce established advanced tumour burden (only delayed metastasis). Reaching senescent cells dispersed across many tissues is harder than reaching a tumour mass. gap/long-term-unknown
  • Collateral/off-target by design. Once activated, trans-cleavage is non-specific within the cell — so the entire safety margin rests on trigger-transcript selectivity. Bystander killing of any cell crossing the activation threshold, and innate-immune activation by the DNA-damage/LNP cargo, are the principal hazards. See cas12a2-trans-cleavage § Limitations.
  • Verification scope. Numerics verified against the bioRxiv preprint full text (DOI 10.64898/2026.05.08.723607); all quantitative claims confirmed. The peer-reviewed Nature version (DOI 10.1038/s41586-026-10738-7) remains paywalled — minor editorial changes between preprint and final version cannot be excluded. gap/no-fulltext-access (journal version only)
  • Companion/origin papers cited by DOI only (Scholz 2026; Dmytrenko 2023) — candidates for their own study pages if the platform earns deeper coverage.

Cross-references

Footnotes

Footnotes

  1. doi:10.1038/s41586-026-10466-y · Scholz N, Thompson J, Crosby KT, …, Jackson RN, Beisel CL, Liu Y · Nature 2026 (published 6 May 2026) · companion foundational paper — RNA-triggered, sequence-specific killing of eukaryotic cells by Cas12a2 via genome-wide DNA shredding; basis the Doudna paper applies to cancer-specific transcripts. Closed-access — cited from metadata; candidate for its own study page. gap/no-fulltext-access ↩

  2. doi:10.1038/s41586-022-05559-3 · Dmytrenko O, Neumann GC, Hallmark T, …, Jackson RN, Beisel CL · Nature 2023 · 613(7944):588-594 · origin paper — Cas12a2 elicits abortive infection via RNA-triggered, non-specific destruction of dsDNA/ssDNA/ssRNA after recognising a target RNA with an activating PFS, inducing an SOS DNA-damage response and growth arrest. Establishes the trans-cleavage mechanism repurposed here. Closed-access — cited from metadata. gap/no-fulltext-access ↩

  3. zeng-2026-cas12a2-chromatin-shredding · doi:10.1038/s41586-026-10738-7 (Nature 2026; preprint doi:10.64898/2026.05.08.723607) · Zeng J, …, Doudna JA · in-vitro (human cell lines: HEK293/HEK293T, hTERT-RPE1, U2OS, PC9, NCI-H23) + in-vivo (mus-musculus: MYC-transposon liver GEMM in FVB/NJ; PC9 lung xenografts in immunodeficient mice) · transcript-activated chromatin shredding via SuCas12a2 trans-cleavage; SNV/indel/abundance selectivity; LNP mRNA+gRNA delivery reduced liver-tumour surface area and delayed lung-metastasis. Verified against CC-BY bioRxiv preprint full text (2026-06-13); Nature version paywalled. ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15

Graph View

Backlinks