CRISPR systems come in two classes.
Deleting the Viper protein or perturbing the GGY motifs abolished silencing.
The Viper protein forms a right-handed filament along its guide RNA, about six subunits per turn.
But the Viper RNA runs 25 nucleotides per turn, while both DNA strands run 15.
"There are likely already patent applications in process for the Viper system's genome editing and DNA targeting applications," Dove predicted.
For decades, researchers have treated the bacterial immune system's bacterial defense systems as a curiosity. That curiosity became CRISPR, the gene-editing technology now reshaping everything from agriculture to FDA-approved therapies. The same lab that helped bring CRISPR to the world has now found something older, smaller, and potentially more subtle hiding in plain sight. Jennifer Doudna's group at UC Berkeley has unearthed a system they call Viper, which silences DNA without ever cutting it. Speaking on the podcast This Week in Virology, virologist Vincent Racaniello laid out the appeal: "You only need one protein, less than 200 amino acids, and a single guide RNA which is less than 100 nucleotides. And you can reprogram them." The discovery, published in two back-to-back Science papers, could redefine what a "minimal" genome-editing tool looks like.
A CRISPR ancestor found not by sequence, but by structure
The hunt for Viper began with a puzzle. CRISPR systems come in two classes. The first to evolve—Class 1—uses "repeat-associated mysterious proteins," or RAMPs, which share so little genetic sequence that standard homology searches fail. The Doudna team instead extracted the structural features of RAMPs and ran similarity searches against AlphaFold's database of predicted protein structures. That search surfaced 2,664 RAMP homologs in 0.7% of prokaryotic genomes—a family no one had catalogued.
The method matters as much as the result. "The sequences were in public databases for years," Racaniello said. "Nobody knew this was hiding in the corner." Alan Dove, a co-host and science journalist, reinforced the point: the breakthrough came from asking the right question with the right tool. In this case, two tools. To align the newly discovered guide RNAs, the team had to turn to Evo 2, a nucleic-acid language model, because—as the hosts put it—every conventional alignment method failed.
The scale of that shift is visible in the numbers:
Feature CRISPR-Cas Viper Protein machinery Multiple Cas genes, multiple RAMPs One protein (Viper) plus one accessory protein (VAP) Guide RNA CRISPR arrays with spacer-repeat architecture 9–16 tandem GGY-NN repeats plus a pseudoknot tail Base pairing Perfect, contiguous Imperfect, non-contiguous Nuclease activity Cuts both DNA strands None; DNA stays intact Mechanism Base-pair invasion displaces the duplex Mechanical force peels the duplex Genomic location Mostly cellular genomes Mostly phages and archaeal viruses Effect Gene disruption via double-strand break Transcriptional silencing, potentially reversible
The quorum of recognition: how Viper avoids random noise
CRISPR recognizes a target by perfect base pairing. Viper cannot. Its guide RNA is built from an alternating motif: GGY repeats that are bound by the protein, and NN positions that actually base-pair with DNA. A two-base match is meaningless—it appears everywhere in a genome. The system solves this with cooperative assembly.
Dove explained the logic: a single protein-RNA unit binds too weakly to persist. "It's too weak to stay there until others come along and pair next to it. And then when you get up to some level, some quorum, they're going to say, 'Oh yeah, this is a legit match.'" That quorum-sensing-like behavior is what gives Viper its specificity. It also elegantly solves the self-versus-non-self problem: the locus encoding the system itself contains the same GGY-NN-GGY motif, so it will never be targeted.
The experimental validation was unambiguous. When the team placed target sites upstream of a GFP reporter gene, only one configuration caused repression. Deleting the Viper protein or perturbing the GGY motifs abolished silencing. Targeting the lambda phage's cro gene reduced plaquing efficiency by five orders of magnitude. Mutating any one of twelve NN nucleotides in the guide sharply reduced protection. And targeting the cI repressor forced a silent lambda prophage into the lytic cycle—proof the system could control a gene's expression, not merely destroy it.
A strained, three-stranded R-loop with no cutting involved
The companion structural paper—the first authored by Peter Yoon, with co-first authors Trevor Doctor and Zuan Terry Jiang—solved 21 cryo-EM structures. What emerges is a geometry that is itself the mechanism.
The Viper protein forms a right-handed filament along its guide RNA, about six subunits per turn. Each subunit binds one GGY unit, cooperatively displacing those bases so only the NN nucleotides remain available for pairing. The complex forms a three-stranded R-loop—unlike the classic two-stranded R-loop of CRISPR. The target DNA strand pairs with the Viper RNA, while the non-target strand is threaded through the center of the protein filament, held by direct protein contacts rather than base pairing.
The structural strain is quantified in a single number the hosts highlighted. All three strands share a helical pitch of about 66 angstroms. But the Viper RNA runs 25 nucleotides per turn, while both DNA strands run 15. Brianne Barker, a co-host and immunologist, described the consequence: "The geometric strain distorts and destabilizes the duplex and then the target strand is handed off to pair with VR RNA. So the result is the duplex is peeled apart by this protein-driven mechanical force rather than by the base pairing."
The broad lesson, Racaniello said, is mechanical: CRISPR unwinds DNA through thermodynamically favored base-pair invasion. Viper peels it apart with protein force, leaving the DNA intact. "CRISPR-Cas9 cuts both DNA strands," he noted. "The DNA is broken, the gene is disrupted—disruptive mechanism. Viper has no nuclease activity... The filament blocks RNA polymerase from accessing the DNA... The DNA is intact."
The design space is now open. "There are likely already patent applications in process for the Viper system's genome editing and DNA targeting applications," Dove predicted. The University of California, Berkeley is named in connection with the research. Viper's architecture—one small protein, one short guide, no permanent cut—suggests applications in genome editing, DNA locus imaging, and epigenetic modification. Dove drew a comparison to catalytically dead Cas9, which sits on DNA and blocks transcription rather than cutting. Viper achieves that with far less machinery.
The minimalism is quantified: Viper requires one protein under 200 amino acids and a guide RNA under 100 nucleotides—both reprogrammable by altering the NN positions.
