Assembly and Translocation of a CRISPR-Cas Primed Acquisition Complex

Abstract

CRISPR-Cas systems confer an adaptive immunity against viruses. Following viral injection, Cas1-Cas2 integrates segments of the viral genome (spacers) into the CRISPR locus. In type I CRISPR-Cas systems, efficient "primed" spacer acquisition and viral degradation (interference) require both the Cascade complex and the Cas3 helicase/nuclease. Here, we present single-molecule characterization of the Thermobifida fusca (Tfu) primed acquisition complex (PAC). We show that TfuCascade rapidly samples non-specific DNA via facilitated one-dimensional diffusion. Cas3 loads at target-bound Cascade and the Cascade/Cas3 complex translocates via a looped DNA intermediate. Cascade/Cas3 complexes stall at diverse protein roadblocks, resulting in a double strand break at the stall site. In contrast, Cas1-Cas2 samples DNA transiently via 3D collisions. Moreover, Cas1-Cas2 associates with Cascade and translocates with Cascade/Cas3, forming the PAC. PACs can displace different protein roadblocks, suggesting a mechanism for long-range spacer acquisition. This work provides a molecular basis for the coordinated steps in CRISPR-based adaptive immunity.

Keywords: CRISPR; Cascade; DNA curtains; fluorescence microscopy; primed acquisition.

Figure 1.. Cse1 promotes facilitated diffusion of Cascade along DNA.

(A) Top: structure of T. fusca (Tfu) Cascade (PDB ID: 5U0A). Star: fluorescent label. Bottom: diverse Cse1 subunits encode a set of eight evolutionarily-conserved positive residues that may interact with DNA. TfuCse1 retains five positive residues (purple). (B) DNA curtains are assembled by immobilizing DNA molecules between microfabricated chrome barriers (B) and pedestals (P). (C) Cascade (magenta) binds non-specifically along the DNA substrate (green). (D) Single-particle traces showing six representative Cascade molecules diffusing on DNA. (E) Cascade diffusion coefficients as a function of ionic strength. Dashed lines: upper boundaries for the theoretical diffusion coefficients based on models either with (helical) or without (non-helical) rotation along the DNA duplex. N > 45 molecules for all conditions. Error bars: S.E.M. The linear fit (red line) estimates 1 ± 0.4 (mean ± 95% C.I.) Coulombic interactions are disrupted at increasing ionic strength. (F) Illustration (top) and kymographs (bottom) of the indicated Cascade variants diffusing on DNA. White and red arrows mark DNA binding and release, respectively. (G) DNA-binding lifetimes of each Cascade variant were fit to a single exponential decay (solid lines). Legend: half-lives ± 95% C.I. (H) Cascade target binding affinities measured via EMSAs. S.D. calculated from at least three replicates. (I) In vivo interference assay and (J) interference efficiency of the indicated Cascade variants. Mean and S.E.M. are calculated from three replicates. Also see Figure S1.

Figure 2.. Cascade transiently samples target sequences via PAM-dependent R-loop propagation and seed-distal complementarity.

(A) Illustration of a DNA substrate with a single Cascade target inserted 21.2 kb away from the cosL DNA end. (B) Top: image of Cascade (magenta) bound to the target sequence on a single-tethered DNA curtain (green). Bottom: histogram of Cascade binding the target site fit to a single Gaussian (center and S.D. are indicated). (C) Top: illustration and kymograph of a diffusing Cascade molecule transiently pausing at the target site. The white and red arrows indicate the beginning and end of a pause, respectively. Bottom: single-molecule tracking indicates that Cascade pauses twice at the target site (dashed line). Gray band: experimental uncertainty in defining the target site. (D) Most encounters with the target sequence result in Cascade pausing (N=27 Cascade molecules; 227 pauses). Error bars are generated via bootstrapping in (B) and (D). (E) Schematic of six DNA substrates containing a second Cascade target 34.5 kb away from the cosL DNA end. Segments of the target DNA are either mismatched (white boxes) or complementary (green boxes) to the crRNA. (F) Pausing probability of Cascade on the six DNA substrates described in (E). Pausing distributions are fit to two Gaussians (red) and recover both target positions (dotted grey lines). N: number of pauses. (G) Cascade pause durations on the substrates shown in (E). In all but two cases, the data required a bi-exponential fit (solid lines). The magnitude of the second population of the two exponentials is reported. N > 95 pauses for all experiments. (H) Model for target recognition by diffusing Cascade surveillance complexes. See also Figure S2.

Figure 3.. Cascade/Cas3 complex translocates via a looped DNA intermediate.

(A) Top: illustration and kymograph of a translocating Cascade/Cas3 complex. Bottom: Cas3 translocating independently of Cascade. White arrows: initiation of translocation; red arrow: Cascade/Cas3 separation. (B) Cas3 initiates translocation after a 30 ± 1s (95% C.I.) pause (N=48). (C) Hydrodynamic force is applied to a 1 µm paramagnetic bead conjugated to the free DNA end. Increasing tension on the DNA ruptures the Cse1 and Cas3 protein-protein contacts, leading to independent Cas3 translocation events. (D) Top: schematic of the dual-labeled Cascade complex. Cse1 and Cas6e always translocated together upon addition of Cas3 and ATP (N=10). (E, F) Translocating Cas3 extrudes a DNA loop due to interactions with target-bound Cascade (top, E,F). Cascade/Cas3 translocates in the 3’ to 5’ direction on the non-target strand, towards the cosR DNA end. Release of the DNA loop (red arrow in panel F) via Cas3-Cse1 rupture or slippage returns the DNA end to its initial position. Scale bars: 3 min (horizontal) and 1 µm (vertical). See also Figures S3 and S4.

Figure 4.. Cas1-Cas2 forms a complex with Cascade and Cas3.

(A) Cas1-Cas2 sampling DNA via 3D collisions. The dashed red line and gray band represent the Cascade target site, as defined in Figure 2. (B) Illustration (top) and kymographs of Cas1-Cas2 (green) recruitment to Cascade (magenta) at the target site. White arrows in (A) & (B): Cas1-Cas2 binding, red arrows: Cas1-Cas2 dissociation. (C) The PAC processively translocates along DNA. Cascade (magenta) and Cas1-Cas2 (green) are fluorescently labeled while the presence of dark Cas3 is observed via translocation of the entire complex. (D) DNA-binding lifetimes of the indicated complexes were each fit to a single exponential decay. A constant was also included in the Cascade/Cas3 and PAC fits. Error: 95% C.I. (E) Representative traces of Cascade (magenta) and Cas1-Cas2 (green) translocating together in the PAC. (F) The mean PAC velocity was statistically indistinguishable from Cascade/Cas3 (N=39; p=0.34). Mean PAC processivity was reduced compared to Cascade/Cas3 (p=0.015). Red diamonds indicate the mean of the PAC distribution. The mean and S.D. of the Cascade/Cas3 distributions are indicated by the solid and dashed gray lines, respectively. (G) The PAC translocates exclusively via a DNA looping mechanism. Error bars generated via bootstrapping. (H) BiFC assay showing the PAC forms in vivo. (I) Cascade interacts with Cas1-Cas2 without a target DNA. Scale bars: 10 µm and 2 µm for the inset. See also Figure S5.

Figure 5.. Differential outcomes of translocating Cascade/Cas3 and the PAC at protein roadblocks.

(A) Top: illustration of four EcoRI binding sites, E1 to E4, upstream of the Cascade target. Bottom: outcomes for collisions between translocating Cascade/Cas3 complexes (magenta) and EcoRI(E111Q) bound at E1 (green). (B) Cascade/Cas3 translocation velocities (left) and processivities (right) on naked DNA or with EcoRI(E111Q) roadblocks. Red diamonds: mean of the distribution. Dashed lines: locations of E1 to E4. Red line: the location of the first roadblock encountered by Cascade/Cas3. N > 25 for all conditions. Cas3 velocity was statistically indistinguishable for all conditions (p=0.08, 0.34 for E1 and E2 relative to naked DNA, respectively), whereas the processivity was significantly reduced in all roadblock experiments (p=5.7×10⁻²⁰, 5.9×10⁻¹⁹ for E1 and E2 relative to naked DNA, respectively). (C) Position of DSBs induced by Cas3 nuclease activity (N ≥ 10) (D) The PAC (magenta) pushes EcoRI(E111Q) (green). (E) Velocities (left) and processivities (right) of the PAC in the absence And presence of EcoRI(E111Q). Both velocities and processivities were reduced with a roadblock compared to naked DNA (p=1.9×10⁻³ and p=4.9×10⁻⁵ for velocity and processivity, respectively). (F) Outcomes of collisions with EcoRI(E111Q). (G) The PAC causes less frequent DSBs on both naked DNA and at a protein roadblock. Error: 95% C.I. of a single exponential fit. Top: Cascade/Cas3 stalls and creates a DSB at roadblocks. Bottom: the PAC can push through roadblocks to acquire additional protospacers. See also Figure S6.

Figure 6.. Stepwise assembly of CRISPR-associated sub-complexes in interference and spacer acquisition.

(I) Cascade surveils foreign DNA via a combination of facilitated 1D diffusion and hopping. (II) Target-bound Cascade can interact with Cas1-Cas2 and Cas3 to assemble the PAC. (III) The PAC samples DNA for possible protospacers during processive translocation. (IV) Alternatively, Cas3 induces a double-stranded DNA break, likely at a protein roadblock. The free DNA ends may be further processed by RecBCD or other host nucleases to generate pre-spacers for adaptive immunity. (V) In naïve acquisition, RecBCD degrades foreign DNA into short oligonucleotide-size fragments. Cas1-Cas2 integrates some of these fragments into the CRISPR locus.