MainNatural CRISPR–Cas adaptive immune systems are described widely as RNA-guided machineries that recognize and degrade invading mobile genetic elements1,2. In well-characterized Class 2 systems, including Type II Cas9 and Type V Cas12a, short CRISPR-derived RNAs provide sequence information for target recognition, whereas the Cas proteins execute nucleic acid cleavage3,4. The functional guide RNA is composed of a CRISPR RNA (crRNA) base-paired with a trans-activating crRNA (tracrRNA), which together form a structured RNA scaffold recognized by Cas94. By contrast, Type V Cas12a requires only a single crRNA, in which the processed repeat-derived handle folds into a conserved pseudoknot structure that engages the Cas12a protein directly3. In these systems, the guide serves a dual role: encoding target sequence information and stabilizing a surveillance-competent Cas conformation5,6. Mechanistically, Cas proteins assemble with their guides into inactive binary complexes that become licensed for target interrogation only after recognition of a protospacer adjacent motif (PAM) within foreign DNA7,8,9. Structural and biochemical studies have shown that PAM engagement precedes extensive guide–target pairing and induces conserved conformational rearrangements that activate nuclease function3,7,8,10. These observations suggest that PAM recognition constitutes an initial checkpoint in CRISPR activation, independent of guide–target complementarity. Efforts to broaden targeting, through PAM engineering11,12, RNA guides reprogramming13,14 and protein mutagenesis15,16, have improved flexibility, but have implicitly preserved the assumption that an RNA guide is a fundamental biochemical requirement for Cas activity. Bulk biochemical measurements indicate that Cas–guide RNA binary complexes can nonspecifically associate with PAM-rich DNA, revealing an intrinsic DNA-binding propensity often invoked to explain off-target effects5,17,18,19. Here we asked whether this strategy could be inverted, and RNA-guided effectors could be repurposed into DNA-guided, RNA-targeting platforms. We engineer synthetic CRISPR DNA (crDNA) that exploits PAM-dependent activation to assemble a functional deoxyribonucleoprotein (DNP) complex, and combine structural, biochemical and cellular analyses to define the molecular basis of this alternative activation pathway. This work establishes a previously unrecognized mode of CRISPR activation and expands the design space for programmable RNA manipulation.Programmable RNA recognition and cleavage with DNA-guided CRISPR–Cas12aThis reasoning prompted us to ask whether the informational role of the guide and the structural role of PAM engagement could be decoupled physically, such that PAM-mediated activation is provided by a DNA element while sequence specificity is supplied by an RNA substrate. Conformational studies of Cas12a trans‐cleavage on heteroduplex substrates revealed that a 3′–5′ crDNA, bearing a stem–loop inverted relative to conventional crRNA, paired with its complementary RNA unexpectedly activates the trans-cleavage activity of Cas12a (Supplementary Fig. 1). To understand this phenomenon, we performed AlphaFold‐guided modeling and molecular docking20, which showed that crDNA occupies the canonical double-stranded DNA (dsDNA)‐binding groove; its duplex region, including the PAM, interacts with the groove formed by the wedge (WED), recognition 1 (REC1) and PAM-interacting (PI) domains rather than the WED and endonuclease RuvC (homology domain of UV-sensitive gene product C activity for resolving Holliday junction) domains (regarded as canonical crRNA repeat-derived handle binding domain)21,22 (Fig. 1a). We therefore hypothesized that a synthetic DNA molecule mimicking key features of PAM-containing dsDNA could form a DNP complex with Cas12a and redirect the effector toward RNA substrates. Guided by this structural framework, we redesigned and optimized the crDNA architecture, enabling efficient reprogramming of Cas12a into a DNA-guided, RNA-targeting effector (Supplementary Fig. 2).Fig. 1: DNA-guided CRISPR–Cas12a cleaves ssRNA targets.The alternative text for this image may have been generated using AI.Full size imagea, Structural comparison of DNA-guided and RNA-guided CRISPR–Cas12a systems. Left: AlphaFold3-predicted model of a DNA-guided Cas12a ternary complex, in which the crDNA (black), containing a PAM sequence, basepairs with the target RNA (red). The resulting crDNA–RNA heteroduplex is positioned within a binding groove formed primarily by the WED (yellow), REC1 (light gray) and PI (mint green) domains. Right: structure of an RNA-guided Cas12a complex (PDB 5B43)21, showing crRNA (red) engagement within the canonical RNA-binding channel. BH, bridge helix; NTS, nontarget strand; TS, target strand. b, Representative cis-cleavage assay of a 450-nt ssRNA target by Cas12a–crDNA. Representative gel image from three independent experiments with similar results. c, Sequence of the crDNA and its complementary ssRNA target, with the cis-cleavage site indicated. The PAM sequence is highlighted in pink. d, Mapping of Cas12a–crDNA cis-cleavage sites. Lanes labeled Control, OH− and Cas12a indicate the 24-nt RNA control, alkaline hydrolysis ladder and Cas12a–crDNA cleavage products, respectively. Cleavage products corresponding to distinct sites are numbered. Representative gel image from three independent experiments with similar results. e, Trans-cleavage activity of DNA-guided Cas12a across eight different RNA targets: 10 nM target RNA, condition containing 10 nM input target RNA; NTC, no-target control. Bars: mean ± s.d.; n = 3 technical replicates. f, Comparison of DNA-guided Cas12a trans-cleavage activity across three Cas12a orthologs. Bars: mean ± s.d.; n = 4 technical replicates; 10 nM target RNA and NTC as in e. g, Kinetic analysis of trans-cleavage activity for DNA-guided and RNA-guided Cas12a systems. Datapoints are the mean of technical replicates. RFU, relative fluorescence units; kcat,crRNA, apparent turnover number (min−1) of Cas12a trans-cleavage activity when programmed with a CRISPR RNA (crRNA) guide; kcat,crDNA, apparent turnover number (min−1) of Cas12a trans-cleavage activity when programmed with a CRISPR DNA (crDNA) guide.Source dataWe performed cleavage experiments in vitro using a 5′-carboxyfluorescein (5′-FAM) labeled RNA targets (Supplementary Table 1), confirming that DNA‐guided Cas12a binds and cleaves target RNA specifically (Fig. 1b and Supplementary Fig. 3). Further experiments indicate that RNA recognition and binding by Cas–crDNA complexes is Mg2+‐independent, whereas cleavage strictly requires Mg2+ (Supplementary Fig. 4), consistent with metal-dependent catalysis by the RuvC domain. High-resolution cleavage site mapping shows that DNA‐guided Cas12a cuts at two, five and six bases downstream of the spacer (Fig. 1c,d). Notably, the cleavage site of DNA-guided Cas12a is different in the RNA-guided system, as here the RNA targets occupy canonical crRNA binding region, suggesting possible resemblance to 3′ pre-crRNA cleavage feature23,24,25. Consistent with this model, targeted mutagenesis of Cas12a demonstrated that DNA-guided RNA cleavage depends on the same RuvC catalytic residues required for dsDNA cleavage in canonical RNA-guided Cas12a, whereas residues implicated in 5′ crRNA array processing are dispensable (Supplementary Fig. 5). Together, these results establish that DNA-guided Cas12a engages RNA substrates through a mechanistically distinct yet catalytically conserved pathway.We next assessed whether DNA-guided Cas12a supports robust target-induced trans-cleavage using a fluorophore-quencher-labeled single-stranded DNA (ssDNA) reporter. Across eight distinct RNA targets, DNA-guided Cas12a exhibited strong and reproducible trans-cleavage activity (Fig. 1e and Supplementary Fig. 6), indicating that activation is not restricted to specific sequences. Comparable trans-cleavage was observed across several Cas12a orthologs guided by crDNA (Fig. 1f), supporting the generality of this DNA-guided strategy. Kinetic analysis using Michaelis–Menten modeling revealed that, at equivalent concentrations of activating nucleic acids, the maximal trans-cleavage rate of DNA-guided Cas12a is approximately twofold lower than that of the canonical RNA-guided system (Fig. 1g). This reduction probably reflects the absence of stabilizing interactions provided by the repeat-derived RNA pseudoknot. Nevertheless, sensitivity assays demonstrated that DNA-guided Cas12a enables direct detection of RNA targets at picomolar concentrations without preamplification (Supplementary Fig. 7), comparable to established RNA-guided Cas12a diagnostic platforms26,27.Collectively, these results demonstrate that Cas12a can be reprogrammed efficiently into a DNA-guided effector capable of sequence-specific RNA recognition and robust cis- and trans-cleavage activities.Structural basis for DNA-guided reprogramming of CRISPR–Cas12aAlthough these biochemical and biophysical assays establish PAM-dependent activation by crDNA, they do not by themselves distinguish whether Cas12a is engaged through the canonical PAM-recognition pathway or through an alternative, noncanonical nucleic-acid-mediated mechanism. To define the structural basis of DNA-guided CRISPR–Cas12a activity, we determined the cryogenic electron microscopy (cryo-EM) structure of a ternary complex comprising AsCas12a, a single-stranded crDNA guide and a complementary RNA target. To reduce preferred orientation during data processing, particle orientations were rebalanced, yielding a final map at overall 3.17 Å resolution (Fig. 2a). In the resulting map, 41 of the 43 nucleotides (nt) of the crDNA and the first 23 nt of the RNA target were well resolved and modeled, forming a 20-bp DNA–RNA heteroduplex stabilized within the AsCas12a binding channel (Fig. 2e). As expected, the PAM sequence embedded within the single-stranded crDNA adopts a stem–loop conformation and is recognized specifically by the PI domain of Cas12a (Fig. 2b–d). Notably, the WEDIII domain exhibits high flexibility, consistent with the absence of a repeat-derived RNA pseudoknot in the DNA-guided system. Overall, the architecture of the crDNA-guided Cas12a complex closely resembles that of the canonical 20-bp R-loop formed with crRNA-mediated DNA targeting (PDB 8SFO)28 (Supplementary Fig. 8a,c). In particular, the positioning of the PAM and the key PI residues are nearly superimposable (Supplementary Fig. 8k,l), underscoring the capacity of Cas12a to accommodate a PAM-bearing ssDNA guide while preserving its canonical target-recognition framework.Fig. 2: Structural basis for DNA-guided reprogramming of CRISPR–Cas12a.The alternative text for this image may have been generated using AI.Full size imagea, Left: top view of cryo-EM density of AsCas12a in complex with crDNA and target RNA. The density of AsCas12a is colored according to the scheme in Supplementary Fig. 9. Right: side view with densities of Rec1/2 domains omitted to show bound DNA–RNA duplex in the middle. b, Model of AsCas12a–crDNA–RNA fitted into the map in a. c, Density of crDNA (blue) and RNA (red) duplex. PAM sequence is highlighted in a pink box. d, Interaction between the PAM sequence of crDNA (pink) and AsCas12a. e, Schematic showing full protein–nucleic acid interactions between AsCas12a and visible DNA–RNA duplex. f, Top: an example class exhibiting extra density downstream of 3′ end of target RNA is observed at the opening of C-shaped AsCas12a and proximal to the nuclease site of AsCas12a, after focused 3D classification to resolve the heterogeneity. Important residue E993 at active site of AsCas12a nuclease domain is shown. Bottom: extra density of downstream RNA is found docking onto a basic patch of Nuc domain surface.Close inspection of the nuclease (Nuc) domain revealed an additional density adjacent to its surface basic patch. Given that the Nuc domain sequence and the crDNA are modeled almost entirely with only a single unresolvable nucleotide of crDNA at each terminus, we speculate that this extra density corresponds to the 3′-downstream sequence of the RNA target. Focused three-dimensional (3D) classification around the Cas12a opening captured several conformational states of this density, which appear to extend continuously from the 3′ end of the RNA target in the DNA–RNA heteroduplex (Fig. 2f and Supplementary Fig. 8). This observation suggests that the downstream RNA segment folds back toward the catalytic cleft formed between the RuvC and Nuc domains, positioning it for cleavage.Together, these structural data establish that crDNA can direct formation of a catalytically competent Cas12a ternary complex and provide a mechanistic framework for understanding how PAM-dependent DNA engagement enables subsequent RNA recognition and cleavage in this DNA-guided system.A PAM-dependent model for crDNA engagement by Cas12aCryo-EM analyses of the crDNA–RNA–Cas12a ternary complex revealed that, despite the absence of the canonical repeat-derived RNA pseudoknot, the PAM sequence of crDNA occupies the PI domain of Cas12a in a geometry indistinguishable from that observed in RNA-guided ternary complexes and is associated with conformational rearrangements consistent with catalytic activation. In contrast, electrophoretic mobility shift assay (EMSA) experiments showed that, in the absence of a structured RNA guide, direct binding of RNA to Cas12a is weak (Supplementary Fig. 10), suggesting that RNA alone is insufficient to initiate productive complex formation.Based on these observations, we propose a model in which crDNA associates initially with apoCas12a to form a DNP complex through PAM-mediated interactions that subsequently recruits the complementary RNA target. In this framework, formation of the RNA–DNA duplex within the Cas12a binding channel induces further conformational rearrangements that activate both cis- and trans-cleavage activities. AlphaFold3 predictions support the feasibility of direct crDNA engagement by apoCas12a (Fig. 3a). Consistently, binary complex EMSA experiments, including competition assays with a fivefold excess of yeast RNA, demonstrate stable and specific binding between crDNA and apoCas12a, whereas random single-stranded DNA fails to form detectable complexes (Fig. 3b,c and Supplementary Fig. 11). Notably, the presence of yeast RNA increased modestly the fraction of unbound crDNA, suggesting that background RNA can compete partially with crDNA binding (Fig. 3c).Fig. 3: crDNA preferentially engages apoCas12a through PAM-dependent interactions.The alternative text for this image may have been generated using AI.Full size imagea, Model of the binary Cas12a–crDNA DNP complex, predicted by AlphaFold3. b, EMSA shows crDNA with PAM sequence binds with increased Cas12a. Representative EMSA image from three independent experiments with similar results. c, Cas12a–crDNA competing experiment with more than five times concentration of yeast RNA. d, Limited trypsin proteolysis suggests structural differences between apoCas12a and crDNA–Cas12a binary complex. Representative gel image from three independent experiments with similar results. e, Correlation between number of PAM mismatches and Kd increase. Kd was measured by fluorescence polarization assay. f, Correlation between number of PAM mismatches and trans-cleavage activity decrease; n = 3 technical replicates. Bars: mean ± s.d. g, Proposed model for DNA-guided Cas12a activation. h, Fluorescence polarization measurement for binary complex. i, Fluorescence polarization measurement for ternary complex. Datapoints are the mean of six independent experiments (n = 6), and the solid line represents a logistic fit to the average data.Source dataLimited trypsin proteolysis further revealed that crDNA binding induces a distinct conformational state of Cas12a, characterized by the appearance of a pronounced cleavage fragment at ~110 kDa and the disappearance of a fragment at ~130 kDa relative to apoCas12a (Fig. 3d). This pattern is indicative of an ordered conformational rearrangement rather than nonspecific association. Systematic mutation of the PAM sequence resulted in progressive increases in the dissociation constant (Kd) of the crDNA–Cas12a binary complex, accompanied by corresponding reductions in trans-cleavage activity (Fig. 3e,f and Supplementary Figs. 12 and 13). Complementary protein mutagenesis experiments further demonstrated that residues involved in PAM recognition are essential for maintaining efficient DNA-guided trans-cleavage activity (Supplementary Fig. 14).Together, these data support a two-step binding and activation model for DNA-guided CRISPR–Cas12a (Fig. 3g). In this framework, Cas12a first associates with crDNA to form a PAM-dependent DNP complex characterized by dissociation constant Kd1, which subsequently binds the RNA target to generate the catalytically active ternary complex with dissociation constant Kd2. Fluorescence polarization measurements of Kd1 and Kd2 (Fig. 3h,i) revealed that binding affinities comparable in magnitude to those of canonical RNA-guided Cas12a systems (Supplementary Fig. 15). Notably, previously reported affinities for RNA-guided Cas12a span several orders of magnitude depending on the measurement approach5,29,30; when compared specifically with fluorescence polarization–based studies31, the values obtained here fall within the same order of magnitude, underscoring the functional viability of this alternative activation pathway.We also evaluated alternative binding scenarios, including initial association of apoCas12a with RNA or prehybridization of RNA with crDNA (Supplementary Fig. 16). Quantitative binding analyses showed that the affinity of apoCas12a for RNA is approximately an order of magnitude weaker than that for crDNA, whereas the affinity between crDNA and RNA is comparable to that of the DNP complex. In practical applications, Cas12a and crDNA are supplied as reagents, whereas RNA represents the analyte or target. Under these conditions, the crDNA–apoCas12a engagement pathway predominates and is therefore the operationally relevant route in our system.DNA-guided CRISPR–Cas system is RNA-specific targetingConsistent with their established role as RNA-guided, DNA-targeting nucleases, RNA-guided Cas12a ribonucleoprotein (RNP) complexes readily bind single-stranded RNA (ssRNA), ssDNA and dsDNA (Fig. 4a and Supplementary Fig. 17), in agreement with previous reports of Watson–Crick-mediated target recognition11,12,26. By analogy, we initially anticipated that DNA-guided Cas12a, where crDNA bridges Cas12a and its RNA target through Watson–Crick base pairing, might similarly engage ssDNA substrates. EMSAs revealed that DNA-guided Cas12a binds exclusively to ssRNA, but not to either ssDNA or dsDNA (Supplementary Fig. 17). The absence of dsDNA binding is explained readily by pre-occupation of the PI groove by crDNA, which sterically precludes dsDNA engagement and unwinding. In contrast, the inability of DNA-guided Cas12a to bind ssDNA is striking, given the DNA-targeting ancestry of Cas12a. Fluorescence-based trans-cleavage assays corroborated these binding data: whereas RNA-guided Cas12a exhibits robust collateral cleavage activity against ssRNA, ssDNA and dsDNA, the DNA-guided system is selectively activated only by ssRNA (Fig. 4b and Supplementary Fig. 18). We propose that, in the DNA-guided complex, the canonical guide-binding channel of Cas12a preferentially accommodates RNA bases while disfavoring DNA, thereby enforcing RNA selectivity at the level of heteroduplex formation. In this model, crDNA–ssDNA hybrids fail to be retained stably within the binding channel and are displaced competitively, preventing productive complex assembly. Molecular simulations support this interpretation, demonstrating stable ternary complex formation between Cas12a, crDNA and RNA, whereas crDNA–ssDNA duplexes dissociate from Cas12a during equilibration (Supplementary Fig. 19).Fig. 4: DNA-guided CRISPR–Cas system is RNA-specific targeting.The alternative text for this image may have been generated using AI.Full size imagea, Target-selectivity schematic of DNA-guided and RNA-guided CRISPR–Cas system. b, Trans-cleavage activity tests showing the target specificity of the DNA-guided Cas12a in comparison with the RNA-guided version of the same protein; n = 3 technical replicates. c, Orthogonal cleavage assay results of DNA-guided Cas12a cis-cleavage activity across eight crDNA–RNA pairs. Representative gel image from three independent experiments with similar results. d, Orthogonal heat map of DNA-guided Cas12a trans-cleavage activity across eight crDNA–RNA pairs. Each cell represents the mean fluorescence signal from three technical replicates (n = 3), with color intensity indicating average trans-cleavage activity. e, Schematic of single-nucleotide specificity test sequence. f, Single-nucleotide specificity of DNA-guided CRISPR–Cas12a system with 20-nt spacer; n = 3 technical replicates. Statistical significance was assessed by ordinary one-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparisons test (each group compared with the NTC group; two-sided). g, Single-nucleotide specificity of DNA-guided CRISPR–Cas12a system with 16-nt spacer; n = 3 technical replicates. Statistical significance was assessed by ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test (each group compared with the NTC group; two-sided). Exact P values are provided in Supplementary Table 2; bars: mean ± s.d.Source dataTo evaluate sequence requirements systematically, we tested eight orthogonal crDNA guides and their corresponding RNA targets. Both cis- and trans-cleavage activities were activated robustly only when the RNA target was fully complementary to the crDNA spacer (Fig. 4c,d and Supplementary Fig. 20), demonstrating stringent RNA-sequence dependence of DNA-guided Cas12a activation.It is notable that the DNA-guided Cas12a system discriminates single‐nucleotide mismatches in the spacer; compared with the fully complementary target (wild type (WT)), most point mutations significantly reduced trans-cleavage-induced fluorescence (P