Mechanism of target site selection by type V-K CRISPR-associated transposases

Abstract

CRISPR-associated transposases (CASTs) repurpose nuclease-deficient CRISPR effectors to catalyze RNA-guided transposition of large genetic payloads. Type V-K CASTs offer potential technology advantages but lack accuracy, and the molecular basis for this drawback has remained elusive. Here, we reveal that type V-K CASTs maintain an RNA-independent, "untargeted" transposition pathway alongside RNA-dependent integration, driven by the local availability of TnsC filaments. Using cryo-electron microscopy, single-molecule experiments, and high-throughput sequencing, we found that a minimal, CRISPR-less transpososome preferentially directs untargeted integration at AT-rich sites, with additional local specificity imparted by TnsB. By exploiting this knowledge, we suppressed untargeted transposition and increased type V-K CAST specificity up to 98.1% in cells without compromising on-target integration efficiency. These findings will inform further engineering of CAST systems for accurate, kilobase-scale genome engineering applications.

Fig. 1.. Type V-K CASTs direct frequent Cas12k- and RNA-independent transposition events.

(A) Schematic of type V-K CAST transposition occurring at on-target sites (RNA-dependent) and untargeted sites (RNA-independent). (B) Experimental TagTn-seq pipeline used for in vitro and genomic samples. (C) Fraction of total genome-mapping integration reads detected at on-target and untargeted sites for the WT pHelper expression plasmid across multiple sgRNAs (top), plotted above on-target transposition efficiencies for the same sgRNAs as measured by Taqman qPCR (bottom). (D) Total genome-mapping reads detected for WT pHelper or pHelper with the indicated deletions, normalized and scaled. (E) Magnified view of integration reads comprising ≤1% of E. coli genome-mapping reads in an experiment performed without Cas12k and guide RNA. (F) Cryo-EM reconstruction of the untargeted transpososome revealing the assembly of TniQ (orange), TnsC (green), and TnsB (purple) in a strand-transfer complex (STC). The target DNA and transposon DNA are represented in light blue and dark blue, respectively. For visualization, a composite map was generated using two local resolution-filtered reconstructions from the focused refinements. Magnified and cutaway views show TnsC forming a helical assembly on the target DNA, positioning residues K103 and T121 (pink) adjacent to one strand of the target DNA (dark blue). The 5′ and 3′ ends of the TnsC-interacting DNA strand are indicated. Two turns of TnsC and TnsB footprint on DNA until TSD cover ~25 and 13 bp, respectively. Only selected TnsC monomers are represented in the cutaway for clarity. (G) Cas12k and the sgRNA were cloned onto a separate vector, and the promoter driving Cas12k expression was varied. Reads detected at on-target and untargeted sites during transposition assays were normalized and scaled. For (C), (D), (E), and (G), the mean is shown from N = 2 independent biological replicates.

Fig. 2.. Biochemical reconstitution of transposition reveals distinct efficiencies at on-target and untargeted sites.

(A) Growth curves upon induction of WT or mutant TnsC with or without TnsB. Data are shown as mean ± SD for N = 2 independent biological replicates inoculated from individual colonies. (B) Assay schematic for probing in vitro plasmid-to-plasmid transposition events using recombinantly expressed CAST components. (C) In vitro integration reads mapping to pTarget from experiments in which TnsC was titrated from 0.1 to 2 μM. Data were normalized and scaled to highlight untargeted integration events relative to on-target insertions. (D) On-target specificity from biochemical transposition assays at varying TnsC concentrations, calculated as the fraction of on-target reads divided by total plasmid-mapping reads (bottom). Total integration activity also decreased as a function of TnsC concentration, as seen by the normalized plasmid-mapping reads (top). (E) Scatter plot showing reproducibility between untargeted integration reads observed in vitro at two high TnsC concentrations; each data point represents transposition events mapping to a single base-pair position within pTarget. The Pearson linear correlation coefficient is shown (two-tailed P < 0.0001); on-target events were masked. (F) Normalized integration reads detected at a representative untargeted site (left) and at the on-target site (right), with 1 μM TnsC and the indicated TnsB concentration. Note the differing y-axis ranges. (G) On-target specificity from biochemical transposition assays at 1 μM TnsC and the indicated TnsB concentration, shown as in (D).

Fig. 3.. RNA-independent integration events occur at preferred sequence motifs.

(A) Schematic for single-molecule DNA curtains assay to visualize TnsC binding. λ-phage DNA substrates are double-tethered between chrome pedestals and visualized used total internal reflection fluorescence microscopy. (B) mNG-labeled TnsC preferentially binds AT-rich sequences on the λ-DNA substrate near the 3′ (pedestal) end (movie S1). (C) Correlation between AT content and mNG-TnsC fluorescence intensity visualized along the length of λ-DNA. The Pearson linear correlation coefficient is shown (two-tailed P < 0.0001). Data are shown as mean ± SD for N = 66 molecules. (D) Binding kinetics for mNG-TnsC at AT-rich and AT-poor regions of the λ-DNA substrate. Apparent k_obs at AT-rich sites ≈0.37 min⁻¹, 95% confidence interval (CI) = 0.35 to 0.39, and at AT-poor sites ≈0.28 min⁻¹, 95% CI = 0.27 to 0.30. Data are shown as mean ± SD for N = 87 molecules (thick line, shaded region). Binding kinetics for AT- and GC-rich sites when compared gave a P value of 0.017 upon bootstrapping. (E) Cumulative frequency distributions for the AT content within a 100-bp window flanking integration events using ShCAST with WT TnsC and sgRNA-1 (N = 5505 unique integration events), compared with random sampling of the E. coli genome (N = 50,000 counts). The distributions were significantly different on the basis of results of a Mann-Whitney U test (P = 1.48 × 10⁻¹³⁵). (F) Cumulative frequency distribution comparison as in (E) but with a K103A TnsC mutant (N = 1932 unique integration events), which revealed a loss of AT bias (P = 0.1349). (G) Meta-analysis of untargeted transposition specificity was performed by extracting sequences from a 140-bp window flanking the integration site and generating a consensus logo. (H) WebLogo from a meta-analysis of untargeted genomic transposition (N = 5855 unique integration events) with a modified pHelper lacking Cas12k and sgRNA. The site of integration is noted with a maroon triangle. An AT-rich sequence spanning ~25 bp likely reflects the footprint of two turns of a TnsC filament (black), whereas motifs within/near the TSD represent TnsB-specific sequence motifs (green). Specific TnsB residues and domains contacting the indicated nucleotides are shown. The magnified inset highlights periodicity in the sequence bound by TnsC. (I) Schematic showing the relative spacing of sequence features bound by Cas12k, TnsC, and TnsB in both on-target (RNA-dependent) and untargeted (RNA-independent) DNA transposition. In both cases, the TnsC footprint covers ~25 bp of DNA and directs polarized, unidirectional integration downstream in a L-R orientation. (J) Magnified view of the ShCAST transpososome structure highlighting sequence-specific contacts between TnsB and the target DNA observed in (H). The Protein Data Bank identification number is 8EA3 (23).

Fig. 4.. Artificial induction of semi-targeted RNA-independent transposition at preferred motifs.

(A) A region on pTarget exhibiting low integration activity (original, blue) was substituted with rationally engineered sequences (colored lines) based on TnsC- and TnsB-binding preferences, generating the indicated pTarget variants (pT-1 to pT6). (B) After performing biochemical transposition assays with the indicated pTarget substrates, integration reads were normalized and mapped to either the forward strand (fwd, red) or reverse strand (rev, black). The intended untargeted integration site based on optimized poly-A and TnsB consensus motifs is marked with a maroon triangle and dotted line; the representative region at right (850 to 900 bp) is shown to highlight consistency in integration events observed elsewhere on pTarget.

Fig. 5.. The fidelity of RNA-guided DNA integration is controlled by TnsC concentration.

(A) Schematic of alternative ShCAST expression strategy in which TnsC was encoded on a separate plasmid (pTnsC) driven by a Lac or T7 promoter. Distinct cellular expression levels were confirmed by Western blot against a 3xFLAG epitope tag fused to TnsC (bottom). (B) Fraction of total genome-mapping integration reads detected at on-target and untargeted sites upon TnsC expression with a Lac or T7 promoter. (C) Genome-wide view of E. coli genome-mapping reads for the original WT ShCAST system compared with a modified ShCAST system with low TnsC expression. The magnified view visualizes reads comprising ≤1% of genome-mapping reads. The target site is marked with a green triangle. (D) Fraction of total genome-mapping integration reads detected at on-target and untargeted sites, with the original ShCAST system or modified ShCAST system with low TnsC expression. Data for five sgRNAs are shown. For (B) and (D), the mean is shown from N = 2 independent biological replicates. (E) Model for target-site selection and transpososome assembly during on-target, RNA-dependent transposition (right) or untargeted, RNA-independent transposition (left) by type V-K CAST systems. Within the untargeted pathway, TnsC preferentially forms filaments at AT-rich regions and is capped by TniQ, leading to the downstream site being selected by TnsB for integration. Cas12k-bound targets may better nucleate TnsC filament formation, and we hypothesize that TnsC filaments loaded at Cas12k-bound targets serve as better substrates for DNA integration, compared with untargeted sites. All structures of TnsC filaments representing untargeted sites (–24, 26), including the BCQ transpososome, reveal K103 residues of the TnsC monomers forming the filament proximal to TnsB, contacting DNA with opposite strand polarity compared with on-target structures (fig. S3B) (22, 23). This could be decisive for the distinct efficiencies observed at these sites.