SauUSI belongs to the type IV family of restriction endonucleases, enzymes that specialize in attacking DNA carrying chemical modifications rather than specific unmodified sequences. Its target is 5-methylcytosine, or 5mC, and its cousin 5-hydroxymethylcytosine, embedded within a loosely defined recognition motif. These modifications are the camouflage of choice for bacteriophages and plasmids seeking to slip past the classical restriction-modification and CRISPR-Cas defenses of their bacterial hosts. SauUSI turns that camouflage into a liability: any piece of DNA bearing the methyl tag is flagged as nonself and condemned. In S. aureus, the enzyme acts as a major barrier to horizontal gene transfer, shaping the acquisition of virulence factors, pathogenicity islands and antimicrobial resistance genes carried on methylated prophages and plasmids. Homologs are widespread across Gram-positive and Gram-negative bacteria and even some archaea, making the mechanism of broad evolutionary and clinical significance.

The enzyme’s architecture hints at its dual personality. Two SauUSI monomers dimerize through N-terminal phospholipase D-family nuclease domains, forming a single active site capable of cleaving both DNA strands sequentially, much like the well-studied type II enzyme BfiI. Each monomer also carries dual helicase-like ATPase domains and a C-terminal target recognition domain, or TRD, with an SRA fold that presumably flips the modified base into a recognition pocket. Earlier work established that SauUSI consumes ATP to move along DNA and cuts 5 to 20 base pairs from its targets, but the polarity of motion, the number of active motors and the trigger for nuclease activation all remained unresolved. The Bristol team therefore combined ensemble biochemistry, single-molecule manipulation and nanopore sequencing to watch the enzyme at work.

The first surprise came from triplex displacement assays, in which fluorescent triplex-forming oligonucleotides anchored at defined positions report the passage of an incoming motor. Using DNA methylated on one or both strands by site-specific methyltransferases, the researchers showed that SauUSI translocates unidirectionally, moving 3-prime to 5-prime along the methylated strand. On symmetrically methylated DNA, displacement occurred on both sides of the target, a clue that both strands can be engaged. Stopped-flow measurements with varying distances between the methylated site and the triplex yielded a translocation rate of roughly 180 base pairs per second at 25 degrees Celsius, with negligible dissociation during motion. Fitting the data revealed an unexpected detail: the motor does not start immediately adjacent to the methylated base but binds about 22 base pairs upstream, a geometry confirmed by an AlphaFold3 model of the DNA-bound dimer.

ATP-dependence experiments pinned the energetics to a single engine. As ATP concentrations fell, both translocation and ATP hydrolysis rates declined in parallel, giving a coupling ratio of approximately one ATP consumed per base pair traveled. That stoichiometry is the signature of one helicase-like subunit driving unidirectional movement, not two motors pulling in tandem. The picture that emerges is elegant: upon binding 5mC, one subunit’s ATPase initiates translocation while the motor physically pushes toward its own TRD, which cannot be bypassed and is displaced from the modified site. What happens next determines the fate of the DNA molecule.

Single-molecule experiments using dual optical traps and confocal microscopy, performed on a LUMICKS C-trap, exposed that fate in vivid detail. With a hemimethylated DNA tether held between beads and ATP added, the team observed unidirectional translocation in 46 percent of events coinciding with an increase in DNA tension, the hallmark of a growing loop. The displaced TRD, it turns out, can recapture the liberated methylated site in roughly half the events, anchoring the DNA while the motor keeps walking, swelling the loop. In the other half, the motor releases the site entirely and glides freely, sometimes for thousands of base pairs to a DNA end. A nuclease-dead mutant observed in magnetic tweezers, and a topoisomerase supercoiling assay that detects the twisted domains generated by looped translocation, independently confirmed the looping mechanism and its dependence on both the TRDs and the freed 5mC.

When both strands carry the methyl mark, the enzyme becomes even more formidable. Looping frequency on dimethylated DNA rose to 73 percent, and at least half of those loops involved bidirectional translocation, with one motor pulling each way from the anchored methylated site. Bidirectional loop translocation ran at 548 plus or minus 59 base pairs per second in the C-trap, slightly more than double the unidirectional rate, suggesting a degree of allosteric activation between the two ATPase subunits rather than completely independent motors. Force, applied across a range up to 20 piconewtons, did not slow the motors, consistent with the tiny one-base-pair steps they take. The authors also reconciled rate discrepancies between instruments by accounting for laser-induced heating in the optical traps, applying a measured Q10 temperature coefficient to the comparison.

Mapping where the enzyme cuts brought the mechanism full circle. Using ENDO-Pore, a high-throughput nanopore method that resolves double-strand breaks on single DNA molecules, the team showed that primary cleavage occurs stochastically throughout translocation, both near and far from the methylated site, generating mostly one-nucleotide 3-prime overhangs and blunt ends. Cleavage clustered within a 20-base-pair window about 22 base pairs upstream of the 5mC, exactly where a rebinding TRD would park the nuclease, and was strongly enriched at 5-prime-TA-3-prime and 5-prime-TNA-3-prime dinucleotides at distal loci. Crucially, a single dimer proved sufficient for both primary and secondary cleavage, with no evidence that converging motors or protein roadblocks are required for activation. Instead, anything that prolongs the motor’s dwell time, whether 5mC rebinding, motor congestion at high enzyme concentrations, or a dCas9 roadblock deliberately placed on the DNA, raises the probability of a cut. On DNA bearing two methylated sites, single molecules were even seen switching direction, patrolling back and forth between sites in a ping-pong scan that multiplies cleavage opportunities.

The overall strategy, which the authors call iterative translocation-cleavage, explains why methylated DNA is so efficiently fragmented even at low enzyme concentrations: repeated cycles of 5mC recognition, ATP-driven movement and stochastic cutting progressively dice the invader into small pieces, without any need for enzyme-enzyme communication and without releasing a diffusible nuclease that could harm the host. Nuclease activity is constrained in cis, tied to the DNA the enzyme is patrolling, so SauUSI functions as a first-line innate defense that neutralizes threats without suicide, in contrast to abortive systems such as Lamassu. Because N6-methyladenine and N4-methylcytosine, the modifications most common in bacteria, are not recognized, the enzyme spares typical bacterial genomes, though 5mC-modified S. aureus strains could never acquire a SauUSI-like system without self-destruction. The work illuminates how helicase-like motors, ubiquitous in bacterial defense systems, can be repurposed from duplex unwinding into destructive surveillance, and it may guide understanding of related motor-nuclease enzymes such as CglI and NgoAVII. Coupling a nuclease to a processive motor, the authors suggest, is an effective way to deliver multiple, distantly spaced, hard-to-repair breaks, and the resulting fragments might even feed CRISPR-Cas adaptation.

Subject of Research: Mechanism of the ATP-dependent type IV restriction enzyme SauUSI cleaving methylated DNA in Staphylococcus aureus

Article Title: Single Staphylococcus aureus SauUSI restriction enzyme motors fragment methylated DNA

Article References: Single Staphylococcus aureus SauUSI restriction enzyme motors fragment methylated DNA. (n.d.). https://doi.org/10.1038/s41589-026-02353-4

Image Credits: AI Generated

DOI: 10.1038/s41589-026-02353-4

Keywords: SauUSI, Staphylococcus aureus, restriction enzyme, DNA methylation, 5-methylcytosine, helicase, DNA translocation, single-molecule biophysics, nuclease, bacterial immunity, horizontal gene transfer, antimicrobial resistance

The structure shows that the answer lies in a strategy biologists describe as structural mimicry. When Cas12a takes up its DNA guide, the complex does not merely carry the guide as a passive address label. Instead, the enzyme and the guide together fold into a shape that persuades the incoming RNA target that it has met a compatible partner. The guide strand and parts of the protein scaffold arrange themselves into a geometry that resembles the duplex the RNA would normally form, allowing the RNA to thread into the complex and pair with the guide as though it were engaging a conventional nucleic-acid partner. In other words, the system speaks RNA’s language while carrying DNA’s script.

To appreciate why this matters, it helps to recall what makes Cas12a unusual among CRISPR nucleases. Unlike Cas9, which requires two separate RNA molecules to form its active guide and creates blunt cuts in DNA, Cas12a needs only a single short RNA guide, recognizes a distinct class of protospacer-adjacent motifs, and cuts DNA in a staggered fashion that leaves overhanging ends prized by genome engineers. Cas12a also belongs to the subset of CRISPR effectors whose natural guide can be encoded in DNA, and once activated by a matching target it unleashes indiscriminate collateral cutting of nearby single-stranded nucleic acids, a behavior that underpins a growing portfolio of diagnostic tests. Every one of those applications depends on the same underlying event: the correct pairing of the guide with the intended target inside the enzyme’s grip.

The newly determined structure captures that event in remarkable detail. Acidaminococcus sp. Cas12a is seen cradling the pseudo-DNA guide in its central channel, with the guide’s seed region, the stretch of sequence that makes first contact with a prospective target, held in an ordered conformation that pre-organizes it for recognition. When the RNA target arrives, it threads through the complex and pairs with the guide, and the resulting hybrid duplex sits within a pocket lined with positively charged residues that stabilize the intertwined strands. Around this core, the protein domains that had been clamped open in the absence of a target rearrange into a catalytically competent architecture, snapping the enzyme into its cutting mode. The structure therefore presents both the resting and the engaged states of recognition in a single frozen moment.

What stands out most is how the pseudo-DNA guide participates in the deception. In related systems, RNA guides form extensive pairing interactions with the protein that keep them in the correct register. Here, the DNA guide relies on a hybrid strategy: parts of it mimic the conformation that an RNA guide would adopt, while the protein supplies compensatory contacts that read DNA’s distinctive chemical features, including the absence of the 2′-hydroxyl groups that decorate RNA. The researchers show that this arrangement allows the complex to present a target-binding surface that is effectively indistinguishable, in shape and charge distribution, from the surface presented by RNA-guided relatives. The RNA target, encountering this surface, binds and pairs with a partner that is chemically DNA but structurally fluent in RNA.

This mimicry extends to the catalytic heart of the enzyme. Cas12a’s nuclease activity depends on the RuvC domain, a processing module shared with other members of the CRISPR-Cas superfamily. In the new structure, the RuvC active site is positioned relative to the guide-target duplex in a way that mirrors its placement in RNA-guided complexes, confirming that the downstream cutting machinery does not care whether the guide is made of DNA or RNA. What matters is the geometry of the duplex delivered to it. By achieving that geometry through mimicry, Acidaminococcus Cas12a solves a chemical problem that would otherwise seem insurmountable: a DNA guide cannot form the same Watson-Crick interactions with the protein that RNA guides use, yet it must produce the same structural outcome.

The biological logic of such a system is thought to trace back to the evolutionary history of CRISPR effectors. Many researchers believe that the ancestral defense machines were RNA-guided, targeting the genetic material of viruses directly, and that DNA-guided variants emerged as immune systems shifted toward attacking DNA genomes. The new structure offers a snapshot of how that transition could be engineered by evolution without redesigning the whole enzyme: keep the recognition and cutting apparatus intact, and evolve the guide-binding channel so that a DNA guide is chaperoned into an RNA-like conformation. Structural mimicry, in this view, is not a curiosity but an economical evolutionary patch, and the pseudo-DNA guide captured in the structure may itself represent an intermediate stage in that ongoing molecular negotiation.

For technologists, the structure arrives as something closer to a blueprint than a curiosity. Genome editing with Cas12a is already routine in laboratories, and its single-guide simplicity, compact size and staggered cuts have made it a favorite for applications ranging from agriculture to therapeutic development. But rational engineering of Cas12a, whether to alter its motif preferences, improve its specificity, retune its collateral activity for diagnostics, or expand the range of sequences it can target, has been constrained by incomplete knowledge of how the DNA guide and the RNA target actually sit inside the enzyme. By showing precisely which residues cradle the guide, which contacts read the target, and which conformational changes license cutting, the structure gives engineers a map of the interaction surfaces they can mutate deliberately rather than by trial and error.

The diagnostic implications may be especially immediate. Cas12a-based assays, which detect pathogens or disease sequences by coupling target recognition to a fluorescent collateral-cutting reaction, depend critically on the sensitivity and specificity of the initial guide-target pairing. Understanding how a DNA guide presents itself to an RNA target, and how mismatches are sensed within the duplex, opens the door to designing guides and protein variants that discriminate more sharply between true targets and near matches, reducing false positives that have complicated real-world deployment. It also suggests routes to building entirely new guide chemistries: if the enzyme tolerates a pseudo-DNA guide, other modified nucleic acids might be accommodated within the same channel, each tuned for stability or detection chemistry.

Therapeutic engineering stands to gain as well. Cas12a’s relatively compact size makes it deliverable in gene-therapy vehicles that struggle to carry bulkier nucleases, and a validated atomic model of its target-recognition state enables computational screening of variants before any test tube is touched. Researchers seeking to minimize off-target editing can now ask structural questions that were previously unanswerable: which protein contacts relax the specificity of pairing, and which lock the seed region into a demanding standard. The same map can guide the design of anti-CRISPR or regulatory proteins that jam the recognition interface, offering a way to switch editing on or off in living systems.

Like any single structure, the model of Acidaminococcus Cas12a bound to a pseudo-DNA guide and RNA target is one frame in what is certainly a dynamic process. The enzyme undergoes further rearrangements during target cleavage and product release that remain to be visualized, and different Cas12a homologs may solve the DNA-guided recognition problem with variations on the theme revealed here. But the central finding is unlikely to change: the division between RNA-guided and DNA-guided CRISPR systems is thinner than it appears. At the level of three-dimensional architecture, the two families speak the same structural language, and one has been caught in the act of translation. That translation, now legible at near-atomic resolution, is precisely the kind of insight from which the next generation of CRISPR tools will be built.

Subject of Research: Cryo-EM structure of Acidaminococcus sp. Cas12a bound to a pseudo-DNA guide and RNA target

Article Title: Architecture of a DNA-guided Cas12a

Article References: Architecture of a DNA-guided Cas12a. (n.d.). https://doi.org/10.1038/s41594-026-01894-5

Image Credits: AI Generated

DOI: 10.1038/s41594-026-01894-5

Keywords: Cas12a, CRISPR, cryo-electron microscopy, structural mimicry, DNA-guided recognition, RNA target, gene editing, Acidaminococcus, nuclease, molecular structure, biotechnology, diagnostics