Structure-guided engineering of type I-F CASTs for targeted gene insertion in human cells

Abstract

Conventional genome editing tools rely on DNA double-strand breaks (DSBs) and host recombination proteins to achieve large insertions, resulting in heterogeneous mixtures of undesirable outcomes. We recently leveraged a type I-F CRISPR-associated transposase, PseCAST, for DSB-free DNA integration in human cells, albeit at low efficiencies; multiple lines of evidence suggest DNA binding may be a bottleneck for higher efficiencies. Here we report structural determinants of DNA recognition by the PseCAST QCascade complex using single-particle cryogenic electron microscopy (cryoEM), revealing subtype-specific interactions and RNA-DNA heteroduplex features. By combining structural data, library screens, and rationally engineered mutants, we uncover variants with increased integration efficiencies and modified PAM stringencies. We further leverage transpososome structural predictions to build hybrid CASTs that combine orthogonal DNA binding and integration modules. Our work provides unique structural insights into type I-F CASTs and showcases diverse strategies to investigate and engineer RNA-guided transposase architectures for human genome editing applications.

Fig. 1. CryoEM structure of the TniQ-Cascade (QCascade) complex from PseCAST.

a Phylogenetic tree of type I-F CRISPR-associated transposons (CASTs), based on previous work in the lab. Systems with previously solved QCascade structures are marked with red arrows, while PseCAST is marked with a green arrow. Phylogenetic clades are colored. b Experimental design to investigate both DNA binding and overall integration activities for CAST systems in human cells. DNA binding is extrapolated from two different transcriptional activation assays, one in which VP64 is fused to Cas7 and one in which VP64 is fused to TnsC. Overall integration efficiencies are measured via amplicon sequencing. c Comparison of VchCAST and PseCAST across different assays in human cells. Although PseCAST exhibits consistently weak transcriptional activation compared to VchCAST, its absolute DNA integration activity at genomic target sites is approximately two orders of magnitude greater. DNA integration data is adapted from a previous publication. Data are shown as mean for n = 2 biologically independent samples for Cas7 and TnsC activation. Data are shown as mean ± s.d. for n = 3 biologically independent samples for DNA integration. Source data are provided as a Source Data file. d Operonic architecture of PseCAST components from the PseCAST transposon, with genes encoding the QCascade complex labeled accordingly. e Left, dominant reference-free 2D cryoEM class averages. Right, cryoEM densities with colored map regions corresponding to Cas8 (blue), Cas7 monomers 1–6 (light blue), Cas6 (purple), TniQ monomers 1–2 (orange, yellow), crRNA (gray), and target DNA (red) indicated. f Refined model for the Cas8 ɑ-helical domain and its positioning relative to the TniQ dimer interface.

Fig. 2. The role of crRNA in the PAM-distal region of PseQCascade.

a Overall view of the cryoEM reconstruction of the PseCAST QCascade complex. b Magnified view of the dashed region in (a), highlighting the cryoEM density (colored and semi-transparent) for interactions between the indicated crRNA nucleotides and protein subunits. c Magnified view of the dashed regions in (b), highlighting interactions between the crRNA and Cas6 (left), TniQ.1 (middle), and both TniQ.2 and Cas7.6 (right). Key interacting residues are labeled. d Normalized RNA-guided DNA integration efficiency at the genomic AAVS1 locus in HEK293T cells, as measured by amplicon sequencing. The indicated alanine mutations were designed to perturb specific RNA-protein interactions highlighted in (c), and were compared to WT. NT, non-targeting crRNA. Data are shown as mean ± s.d. for n = 3 biologically independent samples. Source data are provided as a Source Data file. e Comparison of the crRNA conformation within the PAM-distal region, adjacent to the site of RNA hairpin stabilization by Cas6, for VchCAST (PDB: 6PIJ) and PseCAST (this study). The region around nucleotide G41 exhibits a distinct configuration for PseCAST, likely affecting the behavior of the adjacent TniQ dimer.

Fig. 3. TniQ recruitment to the Cas6-Cas7.6 interface of Cascade requires hydrophobic and electrostatic interactions.

a Overall view of the PseCAST QCascade complex, oriented to highlight the TniQ dimer (dark/light orange). b Magnified view of the region indicated in (a), showing how TniQ.1 (dark orange) interacts with a hydrophobic cavity on Cas6. The two visual renderings are colored either by Cas6 surface (purple, top) or hydrophobicity (bottom). c Comparison of the hydrophobic interactions between TniQ.1 and Cas6 in PseCAST (left) and VchCAST (right, PDB: 6PIJ), with residues labeled. d Normalized RNA-guided DNA integration efficiency at the genomic AAVS1 locus in HEK293T cells, as measured by amplicon sequencing. The indicated arginine point mutations were designed to perturb TniQ.1-Cas6 hydrophobic interactions. NT, non-targeting crRNA. Source data are provided as a Source Data file. e Magnified views of hydrogen bonding (top) and electrostatic (bottom) interactions between Cas7.6 (blue) and TniQ.2 helix (yellow). f Normalized RNA-guided DNA integration efficiency at the genomic AAVS1 locus in HEK293T cells, as measured by amplicon sequencing. Alanine mutations perturbing Cas7.6-TniQ interactions are generally tolerated. Source data are provided as a Source Data file. Data in (d, f) are shown as mean ± s.d. for n = 3 biologically independent samples.

Fig. 4. Structural and functional consequences of PAM and target DNA recognition by PseQCascade.

a Top, overall view of the PseCAST QCascade complex, oriented to highlight the target DNA recognition. Bottom, Magnified view of the experimental cryoEM density map around Cas7.1 and Cas7.2, showing interactions with the crRNA (gray) and DNA target strand (TS, red). NTS, DNA non-target strand. b Magnified views of the PAM binding pocket, with Cas8 and DNA shown in blue and red, respectively. Residues A243 and A244 lack any base-specific, hydrogen-bonding interactions with the DNA. c Normalized DNA integration efficiency at the genomic AAVS1 locus in HEK293T cells for the indicated Cas8 mutants (top), plotted above the WebLogo for PAM preferences in the –1 and -2 positions (bottom) derived from integration into pTarget. (For additional PAM specificity data, see Supplementary Fig. 7e.) Integration efficiency data are shown as mean ± s.d. for n = 3 biologically independent samples. Source data are provided as a Source Data file. d Overlay of the refined atomic model and cryoEM density (semi-transparent) for the seed region of QCascade bound to the DNA target strand. e Schematic representation showing angles for the first five RNA-DNA base pairs (BP 1–5) within the R-loop. f View of the RNA-DNA heteroduplex at right, highlighting the unfavorable base-pairing surrounding flipped out nucleobases within the first 18 base pairs of the R-loop. g Magnified view of the RNA-DNA heteroduplex segments aligned at the flipped out base pair, revealing consistent unfavorable angles at the adjacent base pairs. h Normalized RNA-guided DNA integration efficiency at the genomic AAVS1 locus in HEK293T cells for the indicated Cas7 mutations, as measured by amplicon sequencing. Data are shown as mean ± s.d. for n = 3 biologically independent samples. Source data are provided as a Source Data file.

Fig. 5. AlphaFold-guided engineering of TnsABC to generate chimeric CAST systems.

a Schematic showing the approach to generate a chimeric CAST system by combining optimal DNA targeting and DNA integration machineries from distinct CAST systems. b AlphaFold-generated structure prediction of the TnsABC co-complex from PseCAST. The C-terminal ‘hook’ region of TnsB that putatively interacts with TnsC is marked. c Visualization of select TnsB graft points within the predicted PseTnsABC structure. Residues where Pse-Vch chimerism was introduced are colored in blue, and the three top performing graft points (V585, S589, Q594; PseTnsB numbering) from panel (e) are labeled. d Experimental workflow to test chimeric TnsAB constructs for RNA-guided DNA integration activity. E. coli BL21(DE3) cells containing a pEffector encoding VchQCascade and VchTnsC were transformed with a plasmid encoding a mini-transposon (mini-Tn) and TnsAB, with TnsAB derived from either VchCAST, PseCAST, or a chimeric combination thereof. Integration efficiency was measured by qPCR (bottom). e E. coli DNA integration efficiencies for each tested TnsAB chimera. The amino acid listed represents the position at which the reading frame was grafted from PseTnsB (red) to VchTnsB (blue). “Custom” denotes a variant in which multiple different VchTnsB sequences were substituted (see Supplementary Data 3 for details). Source data are provided as a Source Data file. Data are shown as mean for n = 2 biologically independent samples.