Topology of C2c1

The structure of C2c1 in the ternary complex adopts an overall “Crab Claw” fold consisting of two lobes: an α-helical recognition (REC) lobe including Helical-I and Helical-II, and a NUC lobe including OBD, RuvC, and Nuc domains (Figure 1B). The bridge helix (BH) motif is inserted into the Helical-II domain in close proximity to the PAM-distal region of the guide:target heteroduplex.

Helical-I domain of C2c1 (Figure 1B and S1A) resembles a “dumbbell” shape and consists of 14 α-helices and 2 short β-strands forming a small antiparallel β-sheet between helix α1 and helix α2. Helices α1-α7 and helices α9-α14 form two separate regions that are located at the PAM-proximal and PAM-distal regions of the guide:target heteroduplex respectively. The long helix α8, which connects the two helical bundles, is located in the minor groove of PAM-proximal region, and extends to the PAM-distal region. The Helical-II domain of C2c1 (Figure 1B and S1A) consists of 6 α-helices, and interacts with PAM-distal regions of the guide:target heteroduplex at the opposite side to Helical-I domain. Unlike the BH motif in Cpf1, which forms a separate region and connects the REC and NUC lobes from the middle of the whole complex (Gao et al., 2016; Yamano et al., 2016), the BH motif in C2c1 is inserted between helix α1 and helix α2 of Helical-II and forms a helix bundle together with the Helical-II domain (Figure 1B and S1).

The OBD domain of C2c1 consists of a 9-stranded β-sheet and 2 α-helices (Figure S1B), which closely resembles the core region of the OBD domain of Cpf1 (Figure S1E). The OBD and Helical-I domains contact the PAM duplex. The RuvC domain of C2c1 (Figure S1C) is composed of a 8-stranded β-sheet that is surrounded by 6 α-helices and a β-hairpin formed by β7 and β8, which shows a closely similar conformation to the RuvC domain in Cpf1 (Figure S1F). The Nuc domain of C2c1 contains 2 β-sheets and 6 α-helices (Figure S1C), showing low structural similarity to the Nuc domain in Cpf1 (Figure S1F).

Topologies of sgRNA and Target DNA

The 111-nt sgRNA was derived by connecting the 3′-end of 78-nt tracrRNA to the 5′-end of 42-nt crRNA based on the predicted repeat: anti-repeat (R:AR) duplex as described previously (Shmakov et al., 2015). The crRNA sequence (in magenta) contains a 20-nt guide region and a 14-nt repeat region, whereas the tracrRNA sequence (in cyan) contains a 16-nt anti-repeat region and three stem loops (Figure 1C). The sgRNA and target DNA (in gray) form a T-shaped architecture, consisting of the guide:target heteroduplex, the R:AR duplex, three stem loops, and the PAM duplex (Figure 1C, D). The nucleotides dG(20) to dG(1) segment (all sequences are labeled from 5′ to 3′ ends) of the target DNA strand and C1 to C20 segment of the crRNA form the guide:target heteroduplex. The dG(-1) to dA(-8) segment of the target DNA strand and the dT(-8*) to dC(-1*) segment of the non-target DNA strand (PAM in red) form the PAM duplex (Figure 1C, D).

In the crystal structure, the R:AR duplex comprises two regions: the predicted R:AR duplex-1 formed by nucleotides U(-13) to U(-6) of crRNA and G(-25) to A(18) of tracrRNA (Shmakov et al., 2015), and an unanticipated R:AR duplex-2 formed by G(-5) to C(-1) of crRNA and G(-63) to C(-59) of tracrRNA (Figure 1C, D). The segment of the R:AR duplex-1 formed by U(-13) to A(-11) of crRNA and U(-20) to A(-18) of tracrRNA and the tetraloop A(-17) to C(-14) cannot be traced in the structure (dotted segment in Figure 1C, D). In addition, we are unable to trace nucleotides U(-75) to U(-73) between stem 1 and stem 2.

Recognition of Repeat:Anti-Repeat Duplex

The sgRNA is recognized by OBD, Helical-II, RuvC and Nuc domains via extensive interactions, including several sequence-dependent intermolecular contacts (Figure 2A–C, S2 and S3). In contrast to the R:AR duplex-2, which forms extensive interactions with C2c1 (details of the intermolecular contacts are listed under Supplementary Text), the R:AR duplex-1 protrudes away from the whole complex and has almost no contacts with the protein. One end of R:AR duplex-1 and the adjacent tetraloop cannot be traced in the structure, indicating that these regions may be flexible due to lack of interactions with C2c1. Similarly, no contacts were observed between the end of R:AR duplex with the protein in the Cas9 system (Deltcheva et al., 2011; Jinek et al., 2012).

Recognition of tracrRNA

Stem 1 of the tracrRNA interacts with Helical-II, RuvC, and Nuc domains primarily by non-sequence-specific hydrogen bonds (Figure 2C and S2) (details of the intermolecular contacts are listed under Supplementary Text). No interaction was observed between stem 2 and the protein in the structure.

The linker region A(-58) to G(-55) between R:AR duplex-2 and stem 3 of the tracrRNA is recognized by the Helical-II domain involving both sequence-dependent and backbone-binding interactions (Figure S2 and S3B) (details of the intermolecular contacts are listed under Supplementary Text).

Recognition of Guide:Target Heteroduplex

The 20-bp guide:target heteroduplex is recognized primarily within a channel formed by REC lobe, BH motif, OBD domain and RuvC domain in a sequence-independent way (Figure 2D–E and S4). The PAM-proximal region of the heteroduplex interacts with the OBD, Helical-I, Helical-II, and RuvC domains (Figure 2E and S2). The +1 phosphate group between dG(-1) and dG1 of the target DNA strand is stabilized by the side chain of Arg507 and the main-chain amide group of Gly478 in the OBD domain, which facilitates the unwinding of +1 phosphate group and base pairing between dG1 of target DNA and C1 of crRNA. Arg507 is conserved in some species and Gly478 is highly conserved in the C2c1 family. Such ‘phosphate lock’ type interactions have also been observed in Cas9 (Anders et al., 2014; Jiang et al., 2016; Nishimasu et al., 2014) and Cpf1 (Gao et al. 2016; Yamano et al., 2016). Notably, the R507A and G478P mutants show reduced cleavage activities, indicating the importance of the ‘phosphate lock’ (Figure 2F). The PAM-distal region of the heteroduplex is recognized mainly by Helical-I and Helical-II domains (Figure S2 and S4B–D). Two conserved residues Arg653 and Arg657 in the BH motif interact with the phosphate groups of C14 to A17 of crRNA (Figure S4C).

Recognition of the 5′-TTN-3′ PAM Segment

The PAM duplex is buried within the cleft formed by the OBD and Helical-I domains (Figure 3A–F and S2). The loop between helices α4 and α5 in Helical-I inserts into the minor groove of the PAM duplex and interacts with the dC(-4) to dA(-8) region of the target DNA strand via backbone interactions (Figure 3E). The head of helix α8 and the loop between helices α7 and α8 in the Helical-I domain also interact with the region dT(-8*) to dG(-4*) of the non-target DNA strand through the minor groove in a sequence-independent manner (Figure 3F). The recognition of the 5′-TTC-3′ PAM segment involved nucleobases on both strands of the duplex (Figure 3B–D), a feature also observed in Cpf1 (Gao et al. 2016; Yamano et al., 2016). The nucleobases of the dA(-3):dT(-3*) and dA(-2):dT(-2*) base pairs in the PAM duplex are recognized by several hydrogen bonds (Figure 3B, 3C and S2). The O4 and O2 of dT(-3*) hydrogen bond with the side chain of Arg122 and the main-chain amide group of Gly143, respectively. In addition, the N6 and N7 of dA(-3) hydrogen bond with the side chain of Asn400 (Figure 3B). We modeled the dG(-3):dC(-3*) base pair and found steric clashes between the N4 of dC(-3*) with the side chain of Arg122 and between the N2 of dG(-3) and the backbone of Gly143 (Figure S5A). The O2 of dT(-2*) and N7 of dA(-2) hydrogen bond with the main-chain amide group of Asn144 and the side chain of Asn400 (Figure 3C). We also modeled the dG(-2):dC(-2*) base pair and found steric clashes between the N2 of dG(-2) with the side chain of Asn144 (Figure S5B). Our findings explain the requirement for recognition of the two A-T base pairs in the PAM. The dG(-1):dC(-1*) base pair in the PAM duplex stacks with Gln118 and Gln119 without hydrogen bond formation (Figure 3D), consistent with the absence of sequence specificity at the 5′ position of the PAM (Shmakov et al., 2015). The observed stacking involving these two Gln residues may contribute to D-loop formation adjacent to the PAM sequence involving separation of target and non-target DNA strands. Asn400 and Gly143 are highly conserved in C2c1 homologs, whereas Gln118, Gln119, and Arg122 are conserved in some species. The mutations of these residues exhibit reduced cleavage activity (Figure 3G), with the Q118A/Q119A mutation potentially destabilizing the junction between adjacent PAM and guide:target heteroduplex..

Recognition of Substrate DNA in the RuvC Catalytic Pocket

In the structure of the ternary complex, we observed a 8-nt ssDNA (reflecting excess added 8-nt non-target DNA strand, colored in silver) in the RuvC catalytic pocket, which represents the substrate DNA bound state of the RuvC domain (Figure 4A). The ssDNA is positioned on the binding surface formed by the RuvC and Nuc domains, with complex formation stabilized mainly by sequence-independent interactions (Figure 4B, C and S2). The dT1 to dT3 segment is recognized by the Nuc domain, while the dG4 to dT7 is recognized by both RuvC and Nuc domains (Figure 4A, B). The nucleotide dT7 is located at the edge of the RuvC domain and points towards the cavity formed by the RuvC and Helical-II domains, which is close to the guide:target heteroduplex. The last nucleotide dC8 cannot be traced in the structure, indicating that it is flexible and likely forms few interactions with the protein.

The three Ala-mutated catalytic residues Asp570, Glu848, and Asp977 of the RuvC domain are positioned close to the phosphate group between dG4 and dG5, indicating this as the scissile phosphate position. We then modeled the side chains of the three catalytic residues and found they could be positioned for hydrogen bond formation with the scissile phosphate group of dG5 (Figure 4B–D). In addition, the side chains of Ser899 and Arg911 of the Nuc domain also hydrogen bond with this scissile phosphate. It is also worth noting that the side chain of Tyr853 inserts between dG4 and dG5, thereby stacking with the nucleobase of dG4. This π-π interaction breaks the stacking between dG4 and dG5 and makes the nucleobase of dG5 twist by ~90° and point in an alternate direction (Figure 4C, D). Notably, the related aromatic ring of Phe916 in SpyCas9 was also found to stack with the corresponding nucleobase at the scissile phosphate site (Jiang et al., 2016). All the above residues are highly conserved in the C2c1 family, indicating that they play essential roles in the cleavage reaction. We propose that these residues participate in the divalent metal-ion mediated catalytic cleavage of the phosphodiester bond between dG4 and dG5. Mutations of these acidic catalytic resides (Asp570, Glu848 and Asp997) exhibit complete loss of cleavage activity (top panel, Figure 4E), while others lining the catalytic pocket (Tyr853, Ser899, and Arg911) exhibit either very low or abolished cleavage activity (bottom panel, Figure 4E), which is consistent with our proposal.

The phosphate groups adjacent to scissile phosphate are also stabilized by several intermolecular hydrogen bonds (Figure 4B, C). Notably, the phosphate oxygens between dT3 and dG4 are hydrogen-bonded with the main-chain amide groups of Phe897, Ser898, Ser899, and the side chain of Arg900. The phosphate oxygens between dG5 and dT6 are hydrogen- bonded with Arg574 and the main-chain amide group of Leu573. These interactions are likely to fix and stabilize the single-stranded DNA in the catalytic pocket. The nucleobase of dT7 flips out the catalytic pocket and stacks with the side chain of Trp930, which could potentially stabilize the kink associated with this nucleotide.

We found a sulfate group bound adjacent to the 5′-end of target DNA strand (Figure 5A). The sulfate group is stabilized by coordination to Arg766 in the Helical-II domain and Arg643 and Arg646 in the BH motif. In addition, we also found that the sulfate group bridges the nucleotide of dT7 and the 5′-end of the target DNA strand (Figure 5A). It is likely that this sulfate group mimics a phosphate group and that the target DNA strand could be connected to the ssDNA positioned in the catalytic pocket.

AacC2c1 Generates a Staggered Cleavage Site

It was reported previously that AacC2c1 was able to cleave both supercoiled and linear target DNA in presence of crRNA, tracrRNA, and Mg²⁺ in vitro (Shmakov et al., 2015). When supplied with an engineered sgRNA, AacC2c1 also showed comparable cleavage activity to both supercoiled and linear target DNA. We therefore investigated the optimal cleavage temperature for AacC2c1 in presence of sgRNA. We purified wild-type AacC2c1 and assayed its DNA cleavage activity in vitro at different temperatures. We generated in-vitro-transcribed sgRNA consisting of a 20-nt spacer targeting a sequence from the human EMX1 locus as described previously (Shmakov et al., 2015) and used PCR to amplify a 646 bp linear fragment containing the same DNA target site. We found that AacC2c1 along with sgRNA was able to cleave the target DNA between 37°C and 60°C and showed temperature-dependent cleavage activity (Figure S5C). The cleavage ability at 37°C is lower than when compared to that at higher temperature. Our results at higher temperatures are similar to those reported previously for the system containing crRNA and tracrRNA (Shmakov et al., 2015).

We also mapped the cleavage site of the target DNA using Sanger sequencing. We found that AacC2c1 generated a staggered double-strand break resulting in 7-nt 5′ overhangs (Figure 4F). The cleavage site in the non-target DNA strand is located after the 17^th base, while the cleavage site in the target DNA strand is located after the 24^th base, with the latter position close to the guide:target heteroduplex (Figure 4F).

Recognition of an Extended Target DNA Strand in the RuvC Catalytic Pocket

We therefore designed an extended 38-nt target DNA strand containing a 10-nt extension beyond the guide:target heteroduplex (sequence in Figure 5C) and determined the crystal structure of the ternary complex of AacC2c1, sgRNA and extended target DNA at 2.9 Å resolution (x-ray statistics in Table S1). The overall structure of the extended target DNA ternary complex (Figure 5B) is similar to the non-extended target DNA ternary complex reported above. As we anticipated, the 5′ of the extended target DNA strand kinks and inserts into the RuvC catalytic pocket (Figure 5C), with the scissile phosphate positioned between dT24 and dG25, such that the cleavage site in the target DNA strand is located after the 24^th base. The interactions between the catalytic pocket and extended target DNA strand are nearly identical to those observed in the substrate DNA strand ternary complex (Figure 5A), with the sulfate group replaced by a phosphate group between dG20 and dG21 (Figure 5C). The backbone of dG22 in the extended DNA strand (Figure 5C) is located in a similar position to dT7 in the substrate strand (Figure 5A), while the two bases point in different directions. The side chain of Trp930 also stacks with the nucleobase of dG21, while the nucleobase of dT22 hydrogen bonds with Arg331 in the Helical-I domain and Gln866 in the RuvC domain (Figure 5D). All these intermolecular interactions are likely to help in unwinding the target DNA strand so as to precisely position it in the catalytic pocket and also stabilize the kink of the target DNA strand.

We carried out mutagenesis assays for residues interacting with the substrate strand (Figure S5D) and found that single mutations of Arg574 and Arg900 reduced catalytic activity, while other single mutations had minimal impact.

Recognition of an Extended Non-Target (dT)_n DNA Strand in the RuvC Catalytic Pocket

We also designed an extended non-target strand consisting of (dT)₂₀ linked to the PAM (Figure 5E) and determined the crystal structure of the ternary complex of AacC2c1, sgRNA and extended non-target DNA to 3.8 Å resolution (x-ray statistics in Table S1). Given the modest resolution, we could trace the PAM and an adjacent d(T-T) segment, as well as a d(T-T-T-T) segment positioned in the RuvC catalytic pocket and spanning the cleavage site (Figure 5E). The intervening (dT)_n segment was disordered in the structure of the ternary complex. Given the sequence redundancy associated with the (dT)₂₀ used in this study, the structural data do not identify the position of the cleavage site on the non-target strand.

Structure Determination of the AacC2c1-sgRNA Binary Complex

To clarify how AacC2c1 assembles with sgRNA prior to the target DNA binding, we have also determined the crystal structure of the binary complex of AacC2c1(E848A) bound to sgRNA at 3.3 Å resolution (x-ray statistics summarized in Table S1). The overall structure of the AacC2c1-sgRNA binary complex (two views in Figure 6A) also adopts the bilobed architecture observed for the AacC2c1-sgRNA-DNA ternary complex (Figure 1B).

Modest Domain Rearrangements Following DNA Recognition

A structural comparison between the AacC2c1-sgRNA binary and AacC2c1-sgRNA-DNA ternary complexes reveals a modest “rigid body” shift in the REC lobe (helical-I and helical-II) and BH motif following DNA recognition (Figure 6B and S6A, B). The helical α1-α7 segment in the “dumbbell” shaped Helical-I domain that interacts with the PAM duplex exhibits small conformational changes (Figure 6B and S6A). The helical α9-α14 segment and the long helix α8 in the Helical-I domain that interacts with the guide:target heteroduplex rotates towards the NUC lobe (Figure 6B and S6A). In addition, the Helical-II domain and the BH motif move away from the NUC lobe (Figure 6B and S6B). These movements create enough space to accommodate the guide:target heteroduplex (Figure 6C, D). The OBD, RuvC, and Nuc domains in the NUC lobe are nearly identical between the RNA-bound and RNA-DNA-bound complexes (Figure S6C).

Pre-organized A-form crRNA backbone in the Binary Complex

Previous structural data revealed formation of a pre-organized A-form crRNA in the seed region of the Cas9-sgRNA binary complex (Jiang et al., 2015), but not in Cpf1-crRNA binary complex (Dong et al. 2016). In the AacC2c1-sgRNA binary complex, only the first five nucleotides of the guide segment of sgRNA could be traced and located in the head of guide:target heteroduplex binding channel (Figure 6E). This region maintains a nearly A-form conformation along the sugar-phosphate backbone, with the first two nucleotides adopting a nearly identical conformation as those in AacC2c1-sgRNA-DNA ternary complex (Figure 6F). Notably, stacking is discontinuous within the 2–3–4 segment in the binary complex (Figure 6C). The side chain of conserved Trp234 in the Helical-I domain occupies the position otherwise adopted by the A6-C7 step, thereby disrupting formation of a pre-ordered A-form crRNA in the remaining A6 to C20 segment. Following DNA recognition on ternary complex formation, the Helical-I domain moves away, resulting in the rotation and movement of Trp234, thereby facilitating formation of an A-form helix along the length of the crRNA (Figure 6G, H). In addition, the Watson-Crick edges of bases 1 to 5 in the binary complex are exposed towards solvent (Figure 6H), facilitating access to and pairing with target DNA. Formation of a pre-organized A-form backbone for the crRNA over the first five nucleotide segment in the C2c1 binary complex may facilitate nucleation of guide:target duplex formation and allow the target DNA strand to be easily accommodated in the heteroduplex channel.

Target DNA Recognition

In Cas9, the +1 phosphate group is recognized by the “phosphate lock” loop between the OBD and Helical-I domains (Anders et al., 2014; Jiang et al., 2016; Nishimasu et al., 2014), whereas both C2c1 and Cpf1 (Gao et al., 2016; Yamano et al., 2016) recognize the +1 phosphate group by residues in the OBD domain. The anticipated 25-bp guide:target heteroduplex in Cpf1 ternary complex is disrupted by the side chain of an inserted Trp, thereby instead forming a 20-bp heteroduplex (Gao et al. 2016; Yamano et al. 2016), which allows the target DNA strand be cleaved in the resulting single-stranded region adjacent to the shortened heteroduplex (Zetsche et al., 2015). Despite lacking an inserted Trp in the C2c1 ternary complex, 20-bp of guide:target heteroduplex are recognized and accommodated in the binding channel, with site-specific cleavage observed in the adjacent single-stranded target DNA segment (Figure 4F).

PAM Duplex Recognition

Both Cpf1 (Gao et al. 2016; Yamano et al. 2016) and C2c1 (Figure 3B–D) recognize both target and non-target DNA strands of the PAM duplex through both minor and major grooves. In both ternary complexes, the 5′-TTN-3′ PAM duplex segment is recognized through sequence-specific interactions with the first two A-T pair bases, while the last base pair is stabilized by sequence-independent stacking interactions in C2c1 or backbone-interactions in Cpf1. In Cpf1, the LHD domain interacts with the OBD domain, and together with the Helical-I domain, forms a relatively large channel in the binary complex (Dong et al. 2016), thereby providing access to the PAM duplex. Upon ternary complex formation (Gao et al. 2016; Yamano et al., 2016), PAM duplex binding induces movements in both LHD and Helical-I domains of Cpf1, that results in an “open-to-closed” conformational transition (Gao et al., 2016). In C2c1, the PAM duplex cleft is pre-organized in the binary complex (Figure 6I), and utilizes an alternate mechanism to achieve complex formation. Helices α5, α6 and the flexible loop between helices α4 and α5 in C2c1 occupy a similar position to that of the LHD domain in Cpf1. This region cannot be traced in the C2c1 pre-target-bound state (Figure 6I), indicative of flexibility and potential for a conformational transition on addition of target DNA. Upon achieving sequence-specific base recognition, the PAM duplex-binding loop segment of C2c1 becomes ordered on ternary complex formation (Figure 6J). This leads to additional sequence-independent sugar-phosphate backbone recognition in minor groove of the PAM duplex (Figure 3B–F), thereby further stabilizing and fixing the alignment of the PAM duplex within the PAM duplex-binding cleft. It is likely that the disordered to ordered (“locked”) transition in the PAM duplex-binding cleft of C2c1 has the potential for reducing off-target effects during target DNA recognition, binding and cleavage.

Pre-orangized crRNA Seed Segment

A critical feature, namely the pre-ordered guide RNA seed segment, has been found in Class 1 CRISPR Cascade complexes (Jackson et al., 2014; Mulepati et al., 2014; Zhao et al., 2014), eukaryotic Argonaute binary complexes (Elkayam et al., 2012; Nakanishi et al., 2012; Schirle and MacRae, 2012), as well as for Cas9 in the pre-target-bound state (Jiang et al., 2015). By contrast, the seed segment cannot be traced in the Cpf1 pre-target bound state (Dong et al., 2016). In the pre-target-bound state of C2c1, the first 5 nucleotides of the guide RNA segment form a partially stacked pre-ordered A-form sugar-phosphate backbone conformation (Figure 6E, G), with a similar observation reported for Cas9 in the pre-target bound state. The intermolecular backbone interactions between the first 5 nucleotides and C2c1 are nearly identical in both pre-target-bound and target-bound states, resulting in the anchoring of these nucleotides in both states. The side chain of Trp234 disrupts formation of the A-form conformation beyond nucleotide 5 in the C2c1 binary complex (Figure 6G, H). On binding the target DNA, the Trp side chain reorients and moves away, so as to provide enough space for heteroduplex accommodation along the entire length of the guide segment (Figure 6G, H) (see also, Jiang et al., 2015).

Taken together, it appears that the first five nucleotides in guide segment adopt a pre-organized conformation, with the Watson-Crick edges of their bases exposed to solvent, and available for nucleation with target DNA. In addition, such an alignment allows the target DNA ready access to the guide:target heteroduplex-binding channel, compatible with high cleavage efficiency of C2c1.

Cleavage of Target DNA Strand

One challenging question relates to the precise mechanism of type V endonuclease cleavage of target DNA strand. For Cas9, the target strand is cleaved by the HNH domain, while the non-target strand is cleaved by the RuvC domain (Jiang et al., 2015). By contrast, both Cpf1 and C2c1 contain a RuvC domain but lack the HNH domain. It has been proposed for Cpf1 that the acidic catalytic pocket residues in the RuvC domain are key to cleavage of the non-target DNA strand, while conserved Arg1226 in the adjacent Nuc domain is the primary contributor to cleavage of the target DNA strand (Yamano et al., 2016). In the absence of structures of DNA in either proposed catalytic pocket of the Cpf1 ternary complex, these conclusions were solely based on the impact of mutational studies of putative catalytic residues.

Having observed DNA in the catalytic pocket of the C2c1 ternary complex for both substrate DNA (excess 8-nt DNA) (Figure 5A) and an extended target DNA strand (Figure 5C), our ternary complex structures provide direct insights into the cleavage mechanism of the target DNA strand. In our structures, the RuvC and Nuc domains together form a large flat catalytic pocket that can accommodate around a 6-nt ssDNA. The scissile phosphate group is in a position to hydrogen bond with the three modeled catalytic Asp570, Glu848, and Asp977 residues of the RuvC domain, with additional hydrogen bonding contributions from Ser899 and Arg911 on the adjacent Nuc domain (Figure 4C, D). These five residues have their corresponding counterparts in AsCpf1 (Asp908, Glu993, Asp1263, Ser1071 and Arg1226) (Gao et al. 2016; Yamano et al. 2016) (Figure 7B, E). Another important residue in C2c1 is Tyr853, which disrupts the stacking of the nucleobases adjacent the scissile phosphate group (Figure 4C, D), while this region (aa 996-1009; contains conserved Phe999) is disordered in all reported AsCpf1 ternary complexes (Gao et al., 2016; Yamano et al., 2016). It appears that the above residues most likely play a similar role in the cleavage and function of C2c1 and Cpf1 endonucleases.

Our mutagenesis assays establish that all the three acidic catalytic residues are critical for the cleavage of both strands in C2c1 (Figure 4E, top panel), consistent with a previous mutagenesis study of acidic resides in the RuvC domain in AsCpf1 (Yamano et al., 2016). Of additional interest, mutations of Ser899, Arg911, and Tyr853 in C2c1 also impact strongly on cleavage activity of both DNA strands (Figure 4E, bottom panel). These observations involving alignment of an extended target DNA strand in the RuvC pocket, together with identification of residues critical for cleavage chemistry (Figure 4E), definitively identifies the RuvC pocket as the site for cleavage of the target DNA strand.

Positioning a Bent Target DNA Strand in the Catalytic Pocket

In our structure of the C2c1 ternary complex, the extended single-strand segment of the target DNA strand folds back to access the catalytic pocket. This kinked conformation of the target strand is stabilized through backbone- and stacking-interactions with residues from the BH motif, Nuc, Helical-I, and Helical-II domains (Figure 5D). In Cpf1, the PAM-distal end of the guide:target heteroduplex mainly interacts with the Helical-II domain. The residues involving in the DNA bending interactions in C2c1 are absent at the corresponding positions in Cpf1. Interestingly, the cleavage site of the target DNA strand generated by both Cpf1 (Zetsche et al., 2015) and C2c1 (Figure 4F) is between the 24^th and 25^th nucleotides. In addition, the length of the guide:target heteroduplexes is 20-bp for both C2c1 and Cpf1 (Gao et al., 2016; Yamano et al., 2016). Notably, in Cpf1, the distance between the last nucleotide of the guide:target heteroduplex and the catalytic pocket in the RuvC domain is considerably longer than that in C2c1 (Figure 7C, F). This may imply that the catalytic RuvC and Nuc domains in Cpf1 will have to undergo additional movements towards the target DNA strand to facilitate cleavage.

Although the catalytic residues that interact with the scissile phosphate in C2c1 adopt similar conformations to their counterparts in Cpf1, additional residues in RuvC and Nuc domains that interact with the backbones of nucleotides flanking the cleavage site share low sequence similarity between these two proteins. In addition, the Nuc domains in C2c1 (Figure S1C) and Cpf1 (Figure S1F) show low structural similarity, indicative of differences between C2c1 and Cpf1 for recognition and stabilization of target DNA strand in the catalytic pocket.

Cleavage of Non-Target DNA Strand

Another challenging question relates to the precise mechanism of type V endonuclease relates to the cleavage of the non-target target DNA strand. To address this question, in the first instance, we have solved the structure of a ternary complex containing an extended (dT)₂₀ sequence projecting from the PAM sequence. To this end, we observed a d(T-T) segment adjacent to the PAM and a d(T-T-T-T) segment positioned in the RuvC catalytic pocket (Figure 5E), together with a disordered looped out intervening (dT)_n segment. To address the issue of target site cleavage on the non-target DNA strand, we are currently substituting a few purines at the putative cleavage site within the (dT)₂₀ overhang on the non-target strand (see schematic in Figure 4F) and in turn attempt to solve the structure.

A Single RuvC Pocket for Cleavage of Target and Non-target DNA Strands

A major conclusion of our study is that both target (Figure 5C) and non-target (Figure 5E) DNA strands are independently positioned in the same RuvC catalytic pocket lined by three invariant acidic and flanking catalytic residues in the C2c1 system. This result based on structures of ternary complexes of C2c1 containing either target or non-target DNA strand in the RuvC catalytic pocket reported in the current study contrasts with earlier conclusions on the related Cpf1 system, where separate binding pockets were proposed based solely on cleavage assays on catalytic mutants (Yamano et al. 2016). Specifically, it was proposed that the non-target DNA strand was cleaved within the RuvC catalytic pocket, while the target strand was cleaved in an adjacent pocket on the Nuc domain with contributions from acidic residues lining the RuvC pocket (Yamano et al. 2016). Clearly, further insights into cleavage by Cpf1 would benefit from determination of ternary complex structures that trap the target and non-target DNA strands in their respective catalytic pocket(s).

It has not escaped our attention, that if the same RuvC pocket in C2c1 is involved in independent cleavage of both target and non-target DNA strands, it will be necessary in the future to attempt elucidation of the sequential order of cleavage of DNA strands leading to double-strand breaks of target DNA.

Bound sgRNA Reinforces the Pre-organized Conformation of C2c1

In Cpf1, which binds crRNA, the transition from the pre-target-bound state (Dong et al. 2016) to the target-bound state (Gao et al. 2016; Yamano et al. 2016) involved large conformational changes (Figure 7D, highlighted for helical-I and helical-II domains) (Gao et al., 2016). By contrast, Cas9, which binds sgRNA (crRNA and tracrRNA), undergoes a modest conformational change (Jiang et al., 2016), presumably due to extensive interactions with the bound sgRNA. In our structure, the sgRNA also forms extensive interactions with C2c1, resulting in modest conformational changes (Figure 6B and ​and7A)7A) for the same transition. We propose that these intermolecular interactions with sgRNA in C2c1 reinforce maintenance of the pre-organized conformation of C2c1 for target binding and cleavage. The R:AR duplex of sgRNA interacts with the OBD, Helical-II, and RuvC domains, and the adjacent stem 3 forms extensive contacts with the Helical-II domain (Figure 2B and S3B). These interactions may stabilize the OBD, Helical-II and RucvC domains in a pre-organized conformation in the binary complex. Stem 1 interacts with the surface formed by the RuvC and Nuc domains (Figure 2C), which together may help in maintenance of the conformation of the catalytic pocket. The RuvC and Nuc domains of C2c1 show minor conformational transitions on proceeding from the pre-target-bound state to the target-bound state, likely implying that the intermolecular interactions between these domains and stem 1 maintain the catalytic pocket in a “pre-organized” state for target DNA strand recognition. The C2c1 ternary complex was generated using an excess 8-nt DNA, conditions under which this excess substrate DNA is positioned in the catalytic pocket (Figure 5A), where it adopts an identical conformation to that observed in the ternary complex for an extended target DNA strand (Figure 5C), further implying that the RuvC and Nuc domains are “pre-organized” for access of the target DNA strand.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Plasmid DNA for in vitro transcription was amplified in Escherichia coli DH5α strain in Lysogeny broth (LB) medium at 37 °C overnight.

Recombinant proteins were overexpressed in Escherichia coli BL21 (DE3) strain in LB medium or M9 for seleonomethionine-derivatized proteins. The cells were grown at 37 °C until OD600 reached 0.8 and then induced with 0.25 mM isopropyl β-D-1-thiogalactopyranoside (GoldBio) at 18 °C for 20 hr.

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

In vitro cleavage and Sanger sequencing experiments were repeated three times, and representative results were shown.

We thank Satoko Ishibi-Murakami for technical assistance in generation of C2c1 mutants. 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 NIHORIP HEI grant (S10 RR029205). The research was supported by NIH grant GM104962 to D.J.P. and by the Memorial Sloan-Kettering Cancer Center Core Grant (P30 CA008748), and Cancer Research Institute Irvington Postdoctoral Fellowship and start-up funds from the Institute of Biophysics, Beijing, China to P.G.