CRISPR's New 'Kill Switch': How Cas12a2 Could Change Cancer Treatment

In a series of cell-culture dishes at the University of Utah last year, a CRISPR-based system called Cas12a2 reduced the number of human cells infected with HPV by 94% in a single treatment. Uninfected cells sitting next to them in identical culture media were untouched. The result, published on 6 May 2026 in Nature, is part of a small but specific body of evidence that the CRISPR system most people know as a gene editor can be turned into something quite different: a programmable, RNA-guided "kill switch" for diseased cells [1].

A paper shredder instead of a scalpel

When the public thinks of CRISPR, it usually thinks of Cas9: the enzyme that, with a guide RNA, makes a single precise cut at a chosen spot in the genome. That is the technology behind the 2020 Nobel Prize in Chemistry, awarded jointly to Jennifer Doudna and Emmanuelle Charpentier, and behind CASGEVY (exagamglogene autotemcel), the first CRISPR-based therapy approved in the United States, which the FDA cleared on 8 December 2023 for severe sickle cell disease [2][6]. In that therapy, Cas9 is used to edit a patient's own blood stem cells outside the body before they are returned. The point is to fix something.

Cas12a2 has no interest in fixing anything. As Yang Liu, an assistant professor of biochemistry at the University of Utah Health and one of the paper's co-senior authors, puts it in plain terms: "Its goal is not to correct anything. Instead, it's to destroy anything it sees" [3]. The University of Utah Health press release uses a complementary image: Cas9 is "molecular scissors"; Cas12a2 is a "paper shredder" [3]. Both descriptions are doing the same work. Cas9 cuts once, and the cell's repair machinery usually tidies up. Cas12a2, once activated, cuts everywhere, and the cell cannot keep up.

The switch is a guide RNA

The mechanism is what makes this more than a blunt cytotoxic agent. Cas12a2 is an RNA-guided nuclease. The guide RNA is designed to base-pair with a specific RNA transcript inside the cell. If the match is good enough, the enzyme unleashes an indiscriminate double-stranded DNase activity that shreds the cell's own DNA. The damage is overwhelming. The authors tracked 53BP1 foci, a marker of double-strand DNA breaks, and found at least 5.2-fold more damage in cells where Cas12a2 had been activated than in non-targeting controls, on a par with the damage caused by the chemotherapy drugs cisplatin and etoposide [1]. In yeast, the same logic played out even more starkly: GeCas12a2 reduced transformants 134-fold compared with non-targeting controls, whereas the conventional DNA-cutting FnCas12a reduced them only 4-fold [1]. In human cells, with a guide targeting GFP, GeCas12a2 ribonucleoproteins depleted HeLa-GFP cells by 86% after five days, compared with 37% for the RNA-targeting enzyme LbuCas13a pointed at the same transcript [1].

The other half of the mechanism is the part that matters for safety. Cas12a2 only activates on near-perfect RNA complementarity. A single nucleotide mismatch is usually enough to keep it dormant [1]. That single-nucleotide discrimination is what makes a mutation-specific therapy conceivable in principle: the gRNA is designed to recognise a transcript that carries the disease-specific letter, not the healthy version of the same transcript. Cas12a2 listens for RNA, not DNA, which means the guide targets the cell's current molecular state, not its inherited genome. That is a meaningful shift from Cas9, which is generally directed at DNA.

Headline results in cancer cells and virus-infected cells

The clearest demonstration is in cells carrying the KRAS G12C mutation, one of the most common oncogenic drivers in lung cancer. The mutation locks the K-Ras protein in its active, growth-promoting state. In cell culture, Cas12a2 programmed against the mutant KRAS transcript depleted a G12C-overexpressing cell line by 62%, with no measurable depletion of otherwise identical cells carrying wild-type KRAS [4]. In human lung cancer cells that naturally harbour KRAS G12C, growth was reduced by 50%, "working about as well as established anticancer drugs like cisplatin", according to the University of Utah Health press release [3].

The combination data is what most oncologists will want to read twice. Sotorasib, an FDA-approved KRAS G12C inhibitor, depleted cancer cells by 65% on its own. Adding KRAS G12C-targeting Cas12a2 lifted that to 85% [4]. In cells that had become resistant to sotorasib, Cas12a2 alone still reduced growth by more than 50% [4]. The interpretation worth keeping in mind is that Cas12a2 and sotorasib work through different molecular routes, so resistance to one does not necessarily mean resistance to the other.

The HPV result reads most cleanly. Guides targeting the viral E6 or E7 transcripts, both of which are oncogenes HPV uses to disable the host cell's tumour suppressors, reduced HPV-infected cell numbers by 94% with no measurable effect on uninfected cells [1][4]. In mice bearing HPV-positive tumours, a single treatment with Cas12a2 reduced tumour volume by about 50% [1][5]. The University of Utah Health press release describes the same in vivo result more cautiously, as "slowed tumour growth" [3], which is consistent: a 50% volume reduction is a slowed tumour, not a cured one. The framing here matters. HPV-driven cancers, including cervical, anal, and a subset of head-and-neck cancers, remain a continuing challenge in oncology, where prevention works well via vaccination but established disease is harder to clear. A tool that listens for viral RNA and then destroys the infected cell is, in principle, a different kind of antiviral than a drug that targets the virus's own enzymes.

The team and the stakes

The work is a multi-institutional collaboration. The co-senior authors are Yang Liu at the University of Utah Health and the Huntsman Cancer Institute, Ryan Jackson at Utah State University, and Chase Beisel at the Helmholtz Institute for RNA-based Infection Research and the Botnar Institute [1][5]. The co-first authors are Paul Scholz, who is also co-founder and head of R&D at Akribion Therapeutics, Jared Thompson, a graduate researcher at the University of Utah, and Kadin Crosby, a doctoral candidate at USU [3][5]. Akribion Therapeutics, BRAIN Biotech AG, the University of Würzburg, and the Helmholtz Institute are all listed as collaborating institutions [3]. Funding came from the R. Gaurth Hansen Family endowment, the European Research Council, and the NIH, including the NCI's P30 cancer centre support grant [3].

Several authors have a financial stake in the outcome. Scholz, Zurek, Krohn, Dmytrenko, Jackson, and Beisel have pending or awarded patents on the work; Scholz and others are Akribion employees; Krohn is co-founder and co-CEO of Akribion; and Beisel is co-founder and scientific advisor of Locus Biosciences and Leopard Biosciences [3]. None of that changes the underlying biology, but it is worth knowing when weighing how the work is being positioned for a general audience.

What it doesn't yet do

Cas12a2 is not a treatment. Every efficacy number in the Nature paper comes from cell culture or from mouse models [1][5]. There are no human clinical trials underway, and the authors and the press materials are explicit that more animal work is needed before any can be considered [3][4]. Two technical problems sit on top of that. The first is delivery: getting the Cas12a2 protein, or the mRNA encoding it plus a guide, into enough of the right cells inside a living patient is unsolved, although the team has shown that mRNA plus guide RNA packaged in lipid nanoparticles can work in HEK293T cells in culture [1]. The second is off-target validation: each new guide design will need exhaustive testing for unintended activity in healthy tissues, and Yang Liu has said as much directly [4]. There is also the question of immune responses to a bacterial-derived nuclease delivered repeatedly, which has not been addressed.

The interesting strategic question is not whether Cas12a2 will turn into a cancer drug on its own. It probably will not, in the near term, on the strength of a 50% reduction in cell growth and tumour volume. The more interesting question is whether the field is now closer to a class of "elimination therapies": tools that identify a diseased cell by its RNA signature and then instruct that cell to remove itself. Cancer is the obvious starting point. Liu has also mentioned neurodegenerative disease and blood disorders as places where distinct pathogenic cell populations might be defined by their transcripts [4]. The Nature paper is a proof of principle. What it is not, yet, is a treatment.