Intermolecular interactions between AsCpf1 and crRNA-DNA

The crRNA used for crystallization contains a 20-nt direct repeat region [U(−20)-U(−1)] and a 25-nt guide segment (G1-C25) (Figure 1B). We initially introduced one extra base pair (C25-dG1) at the end of the crRNA-DNA heteroduplex to stabilize the nearby cleavage site, which was shown later on to have no effect based on structural results (see below). Most of the nucleotides were well-defined in the electron density, with the exception of U(−20) and C21-C25 of crRNA, as well as dG1-dG5 of the target DNA strand (Figure 1B). Details of the intermolecular interactions in the ternary complex are summarized in Figure 2.

The 5′-direct repeat region of crRNA is bound in the channel formed by OBD and RuvC domains (Figure 3A). Unexpectedly, this part of the crRNA adopts a pseudoknot fold in the complex (Figure 3A and ​and3B),3B), rather than the simple stem-loop as previously predicted¹⁹. The G(−6)-A(−2) segment forms five canonical base pairs with the U(−15)-C(−11) segment, whereas C(−9)-U(−7) segment adopts a loop structure, representing the predicted stem-loop (Figures 1B and ​and3B).3B). U(−10) and A(−18) forms a reverse Hoogsteen base pair (Figure 3B and Supplementary information, Figure S2A). U(−17) forms hydrogen bonds with both A(−12) and U(−13), thereby stabilizing the pseudoknot fold (Figure 3B and Supplementary information, Figure S2B). In addition, U(−1) and U(−16) also form a non-canonical U-U base pair (Figure 3B and Supplementary information, Figure S2C). The pseudoknot fold adopted by the direct repeat segment in the ternary complex was also found in Lachnospiraceae bacterium ND2006 Cpf1 (LbCpf1)-crRNA binary complex²³, indicating a conserved fold of this region.

The PAM-containing DNA duplex is bound within the cleft formed by the Helical-I, OBD, and LHD domains (Figure 3C). Among all the amino acids participating in the interaction with the PAM DNA duplex (Figure 2), Lys607 is the most critical one in that it contributes to base-specific recognition (Figure 3D). The side-chain of Lys607 forms hydrogen bonds with both N3 from dA(+3) and O2 from dT(2^*) (Figure 3D), indicating that base pairing of dT(3^*)-dA(+3) and dT(2^*)-dA(+2) are important for Cpf1 PAM recognition, a result consistent with previous studies indicating that a 5′-TTN-3′ PAM is preferred by Cpf1^19,24.

The crRNA-target DNA heteroduplex is accommodated within the central channel formed by Helical-I, Helical-II, RuvC, and OBD domains (Figure 3E). The PAM-distal and PAM-proximal ends of the heteroduplex are blocked by OBD and Helical-II domain, respectively. To our surprise, the 25-nt crRNA guide and its complementary target DNA strand form a 20-bp, rather than a 25-bp, crRNA-DNA heteroduplex (Figures 1B and ​and3E).3E). The side-chain of Trp382 stacks against C20-dG6 of the heteroduplex, preventing the formation of further base pairs beyond 20 bp (Figure 3F). The unpaired C21-C25 and dG1-dG5 segments are disordered in the structure and cannot be traced from the electron density.

There is good agreement in the positioning of the 5′-direct repeat, PAM-containing DNA duplex and crRNA-target DNA heteroduplex within the Cpf1 protein, as well as the intermolecular contacts in the ternary complex in this study and that reported in ref. 25.

Conformational change of Cpf1 upon target DNA binding

A previous study has shown that LbCpf1 undergoes large conformational changes upon crRNA binding²³. Whether target DNA binding would cause further structural rearrangements remained to be determined. We first compared the individual domains between AsCpf1 and LbCpf1 by superposition and found that all domains from these two species can be aligned very well (Supplementary information, Figure S3). Further comparison between LbCpf1-crRNA binary²³ and AsCpf1-crRNA-DNA ternary (this study; see also ref. 25) structures reveals significant conformational changes upon target DNA recognition (Figure 4A). Unlike the modest shift in helical domains of Cas9¹⁴, the helical recognition lobe of Cpf1 undergoes substantial rearrangements (Figure 4A) on ternary complex formation. The Helical-I domain rotates and moves toward the NUC lobe to form contacts with the bound crRNA-target DNA heteroduplex and PAM-duplex (Figure 3C, ​,3E3E and ​and4B),4B), while the Helical-II domain shifts away from the NUC lobe to generate space for target DNA binding (Figure 4C). In addition, the LHD of Cpf1 also undergoes modest movements toward the Helical-I domain to form interactions with bound PAM-duplex (Figure 4D), in contrast to Cas9 where the PAM-interacting domain is preordered before target DNA binding¹⁴. The rest of NUC lobe of Cpf1, containing OBD, RuvC, and Nuc domains, as well as the direct repeat region of bound crRNA, undergo modest conformation transitions during target DNA binding (Figure 4E), which is again different from Cas9, where the HNH domain in the NUC lobe undergoes a significant displacement toward the target strand¹⁴.

Active sites for DNA and RNA cleavage

Cpf1 generates a 5-nt staggered cut on the target DNA duplex, in contrast to the blunt ends generated by Cas9¹⁹. A recent study demonstrated that the putative nuclease domain Nuc, together with the conserved RuvC domain, contributes to the cleavage of target and non-target DNA strands, respectively²⁵. Compared with the RuvC domain, the Nuc domain is less conserved within Cpf1 family proteins. However, we observed very good structural alignment between LbCpf1-Nuc and AsCpf1-Nuc domains (Supplementary information, Figure S3), indicating a conserved three-dimensional architecture of this domain. Although Cpf1 undergoes large conformational changes during the transition from RNA-bound binary state (Figure 5A) to RNA-DNA-bound ternary state (Figure 5B), the RuvC-Nuc dual domain retains its conformational alignment (Figure 4A and ​and4E).4E). More importantly, the catalytic sites located in the RuvC and Nuc domains also show similar conformational alignments in pre-target-bound binary state (Figure 5C) and target-bound ternary state (Figure 5D; this study and ref. 25), which is different from Cas9 in that the HNH domain undergoes a significant displacement toward the target strand upon DNA binding¹⁴. Interestingly, the published biochemical results show that the R1226A mutation in the Nuc domain will render Cpf1 into a nickase for non-target strand cleavage, while the mutations of the catalytic residues in the RuvC domain will abolish the cleavage activity for both DNA strands, indicative of a prerequisite step of non-target strand cleavage required for target strand cleavage²⁵.

Cpf1 from Francisella novicida has recently been found to function as an RNase to process pre-crRNA into the mature crRNA²⁴. The active site for RNA cleavage is located in the OBD domain (Figure 5A and ​and5B).5B). Similar to the RuvC and Nuc domains, the OBD domain and the bound crRNA direct repeat region also show a good alignment between pre-target-bound and target-bound states (Figure 4A and Supplementary information, Figure S4A). The three conserved catalytic residues in LbCpf1 are co-planar and form interactions with the first nucleotide of the mature crRNA²³ (Figure 5E). Due to the poor electron density of the Arg794-Glu857 region in our AsCpf1-crRNA-DNA ternary structure, we can only observe the Lys860 residue (related to Lys785 in LbCpf1), which is also close to the last nucleotide A(−19) of bound crRNA (Figure 5F). The RNase activity of Cpf1 will contribute to the cleavage of phosphodiester bond between U(−20) and A(−19) of the crRNA, which may explain the lack of electron density of U(−20) in our ternary structure, as well as the first two Gs in the binary LbCpf1-crRNA structure²³. In the recently reported AsCpf1-crRNA-DNA ternary complex²⁵, both Lys809 and Lys860 can be traced and are close to the last nucleotide of bound crRNA (Figure 5F, shown in silver).

Disordered seed sequence in pre-target-bound binary Cpf1 complex

For both the Cas9 and the class 1 Cascade complex, a remarkable feature of the transition from the pre-target-bound binary state to the target-bound ternary state involves the formation of a preordered A-form crRNA, either only within the seed region (Cas9)¹⁴ or throughout the entire guide region (Cascade complex)^31,32,33. This strategy is also employed by the eukaryotic Agronaute complexes during the transition from guide-RNA bound form to target transcript recognition^28,29,30, representing a convergent evolution of this mechanism¹⁴. A seed sequence of the first 5-8 nt at the PAM-proximal side of the protospacer has also been found for Cpf1^19,24. Whether Cpf1 employs a mechanism for seed pre-organization similar to Cas9 and Argonaute is still unclear. The seed region of crRNA is mostly disordered in the LbCpf1-crRNA binary complex²³, with only the first nucleotide traceable from the electron density (in cyan, Figure 6A). Further, the first nucleotide in the seed region adopts two opposite orientations in the pre-target-bound (in cyan, Figure 6A) and DNA target-bound (in magenta, Figure 6B) states. The structural superposition between pre-target-bound binary and target-bound ternary states shows that the direct repeat region of crRNA and its interacting domains (OBD and RuvC) has no notable conformational changes (Supplementary information, Figure S4A). However, the seed nucleotides (G1-C8) interacting region, mostly located in the Helical-I domain, undergoes significant structure rearrangement upon target DNA binding (Figure 4B). The amino acids located in Helical-I domain that are essential for maintaining the A-form structure of the seed region of crRNA are also concomitantly disorganized in the pre-target-bound structure (Figure 6C). In addition, the A-form structure of the seed region (A5-C8) will have extensive steric clashes with the Helical-I domain in the pre-target-bound binary structure (Supplementary information, Figure S4B). Additional conformational changes of Helical-I domain in the target-bound ternary complex result in a proper binding surface to accommodate the A-form seed RNA (Figure 6B). Further supporting the structural findings, previous trypsin limited proteolysis results showed the same digestion patterns of Cpf1 bound either to a full-length crRNA or a crRNA lacking the guide sequence²³, which is distinct from the important role of the seed region for Cas9 in similar experiments¹⁴. Taken together, both the structural and biochemical results indicate that Cpf1 employs a unique disordered structure of the seed region before target DNA binding, rather than the preordered A-form structure found in Cas9 and Argonaute.

Distinct mechanisms adopted by Cpf1 and Cas9 endonucleases for target DNA cleavage

In the CRISPR-Cas9 system, during the transition from RNA-bound binary state to RNA-DNA-bound ternary state, the preordered A-form RNA seed sequence and preformed protein PAM-interacting cleft constitute important landmarks for the Cas9-RNA complex to interact efficiently with potential DNA sequences for target sampling¹⁴. Such a “preordered seed” strategy has also been commonly utilized by class 1 CRISPR Cascade complexes, as well as by the eukaryotic Agonaute system, implying a convergent evolution of this mechanism¹⁴. However, the seed region of the CRISPR-Cpf1 system is disordered in the RNA-bound binary state due to steric hindrance (Figure 6A and Supplementary information, Figure S4B), as well as due to the lack of essential contacts between RNA and protein (Figure 6C). The PAM interacting cleft of Cpf1 also undergoes conformational changes from “open” (RNA-bound binary) to “closed” (RNA-DNA-bound ternary) states (Figure 6D and ​and6E),6E), in contrast to a preformed PAM interacting cleft in Cas9-RNA complex¹⁴. Thus, the target recognition mechanism employed by Cpf1 represents a different evolutionary path from that observed with Cas9 and Argonaute. We propose that the disordered seed sequence and conformational transitions of the PAM interacting cleft of CRISPR-Cpf1 system may contribute to minimization of off-target effects for Cpf1-based genome editing.

Structural basis for conformational change upon DNA binding

The conformational changes within Cpf1 upon target DNA binding involve mainly rigid body movements of the Helical-I, Helical-II, and LHD domains (Figure 4). We propose that the recognition of the PAM DNA duplex is the trigger for the entire structural rearrangement. When the correct PAM sequence is bound, both the LHD and Helical-I domains will rotate and move inward to position the critical amino acids for forming interactions with the bound DNA (Figures 3C, ​,6D6D and ​and6E),6E), representing the “open-to-closed” conformational transition of the PAM interacting cleft. The hinge region between the LHD and OBD domains could provide a rationale for the movement of the LHD domain (Supplementary information, Figure S5A). Similarly, the connection regions between the Helical-I domain with both the OBD (a turn between two helices) and Helical-II (an unstructured loop) domains should also allow for the large movements of the Helical-I domain (Supplementary information, Figure S5B). The structural movements due to PAM DNA binding will further generate a seed-binding surface (Figure 6B) to accommodate the A-form conformation of the seed segment of crRNA and eventually facilitate the pairing between the seed RNA and target DNA. The requirement for base pairing between the crRNA and target DNA will further push the Helical-II domain away from the Helical-I and Nuc domains to its final position in the crRNA-DNA bound ternary complex (Figures 4A, ​,5A5A and ​and5B5B).

Crystallization for AsCpf1-crRNA-DNA ternary complex

The 45-nt crRNA, 33-nt target DNA strand, and 8-nt non-target DNA strand were all synthesized from IDT company. The crRNA was denatured at 95 °C for 5 min, and subsequently annealed by slow cooling in a buffer containing 20 mM Tris, pH 7.5, 5 mM DTT, 150 mM KCl. The two DNA strands were dissolved in H₂O and mixed together with a molar ratio of 1:1.3 (33-nt target strand : 8-nt non-target strand). The mixed DNA sample was then heated at 95 °C for 5 min and annealed by slow cooling to room temperature. The AsCpf1-crRNA binary complex was prepared by first incubating the protein and RNA at a molar ratio of 1:1.1 at 4 °C for 30 min, followed by gel filtration purification in a buffer containing 20 mM Tris, 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, pH 7.5. The purified AsCpf1-crRNA binary complex was then concentrated to ∼8.5 mg/ml. The AsCpf1-crRNA-DNA ternary complex for crystallization was prepared by simply mixing the AsCpf1-crRNA binary complex with the annealed DNA at a molar ratio of 1:1.3, followed by incubating at 4 °C for 30 min before crystallization.

The crystals of AsCpf1-crRNA-DNA were generated by hanging drop vapor diffusion method at 20 °C, from drops mixed from 1 μl of the complex solution and 1 μl of reservoir solution (0.1 M Tris, pH 7.0, 30% PEG600, 0.5 M (NH₄)₂SO₄).

Structure determination

All the diffraction data sets were collected at the Advanced Photo Source (APS) at the Argonne National Laboratory. The diffraction data were indexed, integrated and scaled using the NECAT RAPD online server. The initial phase was calculated by combining the phase contributions from Se-Met and Hg derivative data sets. After density modification process using RESOLVE³⁴, we could build ∼80% sequence of the protein and most of the RNA-DNA into the electron density. Further model building was done by using the structure of LbCpf1-crRNA²³ binary complex as the reference model. The model building was mostly carried out using the program COOT³⁵ and final structural refinement was carried out using the program PHENIX³⁶. The statistics of the data collection and refinement are listed in Supplementary Table S1.

X-ray diffraction studies were conducted at the Advanced Photon Source on the Northeastern Collaborative Access Team beamlines, which are supported by NIGMS grant P41 GM103403 and U.S. Department of Energy grant DE-AC02-06CH11357. The Pilatus 6M detector on 24-ID-C beam line is funded by a NIH-ORIP HEI grant (S10 RR029205). The research was supported by Cancer Research Institute Irvington Postdoctoral Fellowship and start-up funds from the Institute of Biophysics, Beijing, China to PG and NCI grant 1 U19 {"type":"entrez-nucleotide","attrs":{"text":"CA179564","term_id":"35112471","term_text":"CA179564"}}CA179564 to DJP and by the Memorial Sloan-Kettering Cancer Center Core Grant (P30 CA008748).