Structural basis of PAM-dependent target DNA recognition by the Cas9 endonuclease

Abstract

The CRISPR-associated protein Cas9 is an RNA-guided endonuclease that cleaves double-stranded DNA bearing sequences complementary to a 20-nucleotide segment in the guide RNA. Cas9 has emerged as a versatile molecular tool for genome editing and gene expression control. RNA-guided DNA recognition and cleavage strictly require the presence of a protospacer adjacent motif (PAM) in the target DNA. Here we report a crystal structure of Streptococcus pyogenes Cas9 in complex with a single-molecule guide RNA and a target DNA containing a canonical 5'-NGG-3' PAM. The structure reveals that the PAM motif resides in a base-paired DNA duplex. The non-complementary strand GG dinucleotide is read out via major-groove interactions with conserved arginine residues from the carboxy-terminal domain of Cas9. Interactions with the minor groove of the PAM duplex and the phosphodiester group at the +1 position in the target DNA strand contribute to local strand separation immediately upstream of the PAM. These observations suggest a mechanism for PAM-dependent target DNA melting and RNA-DNA hybrid formation. Furthermore, this study establishes a framework for the rational engineering of Cas9 enzymes with novel PAM specificities.

Extended Data Figure 1. The PAM duplex binds in a positively charged cleft on the C-terminal PAM-interacting domain

a, Zoomed-in view of the PAM binding site in Cas9. Nucleic acids are shown in stick format, coloured according to the scheme in Fig. 1a and overlaid with refined experimentally phased, solvent flattened electron density map (grey mesh, contoured at 1σ). b, PAM binding site in Cas9, shown in the same orientation as in panel a. The molecular surface of Cas9 is coloured according to electrostatic potential.

Extended Data Figure 2. The PAM binding site is preordered in the Cas9–RNA complex

a, Comparison of the structures of Cas9–sgRNA bound to a PAM-containing target DNA duplex (left) and single-stranded DNA target (right). The target DNA strands of the complexes were superimposed using a least-squares algorithm in Coot and the complexes are shown in identical orientations. Bound nucleic acids are shown in stick format and coloured according to the scheme in Fig. 1a. b, Superimposed Cas9 molecules from the PAM-containing and ssDNA bound complexes. The colour scheme is the same as in panel a. In both complexes, the HNH domain is in an inactive conformation, with the active site located approximately 40 Å away from the scissile phosphate in the target DNA strand, suggesting that the domain undergoes a conformational rearrangement upon target strand cleavage. c, Superimposed nucleic acid ligands. sgRNA and target DNA from the single-stranded target complex are coloured grey. d, Detailed view of the PAM binding site in the superimposed complexes, indicating a slight tightening of the PAM binding cleft.

Extended Data Figure 3. Endonuclease activities of Cas9 proteins containing mutations in the PAM binding motif

a, Endonuclease activity assay of WT and mutant Cas9 proteins using supercoiled circular (SC) plasmid DNA containing a target sequence fully complementary to the sgRNA in Fig. 1a. Nucleotide sequences of target sites are provided in Extended Data Table 2. b, Endonuclease activity assay of WT and mutant Cas9 proteins using an oligonucleotide duplex containing a target sequence fully complementary to the sgRNA in Fig. 1a.

Extended Data Figure 4. PAM binding motifs in Cas9 orthologs

a, Cas9 orthologs with known PAM sequences^,,,. The PAM of Cas9 from Lactobacillus buchneri has been inferred from known protospacer sequences, but has not been experimentally validated. b, Alignment of the amino acid sequences of the major groove interacting regions of Cas9 orthologs. Primary sequences of type II-A Cas9 proteins from Streptococcus pyogenes (GI 15675041), Listeria innocua Clip 11262 (GI 16801805), Streptococcus mutans UA159 (GI 24379809), Streptococcus thermophilus LMD-9 (S.thermophilus A, GI 11662823; S.thermophilus B, GI 116627542), Lactobacillus buchneri NRRL B-30929 (GI 331702228), Treponema denticola ATCC 35405 (GI 42525843), type II-B Cas9 from Francisella novicida U112 (GI 118497352), and type II-C Cas9 proteins from Campylobacter jejuni subsp. Jejuni NCTC 11168 (GI 218563121), Pasteurella multocida subsp. multocida str. Pm70 (GI 218767588) and Neisseria meningitidis Zs491 (GI 15602992) were aligned using MAFFT. Amino acids are coloured in shades of blue according to their degree of conservation. The red boxes denote amino acid residues inferred to be involved in PAM recognition in type II-A and type II-B Cas9 proteins based on the sequence alignment and the crystal structure of the Cas9-sgRNA-DNA complex elucidated in this study.

Extended Data Figure 5. Glutamine substitution of Arg1333 and Arg1335 in S. pyogenes Cas9

a, Endonuclease activity assay of WT and mutant Cas9 proteins using a linearized plasmid containing a target sequence fully complementary to sgRNA 2 and a 5′-TGG-3′ PAM (Extended Data Table 2). 2014 and 598 bp bands correspond to Cas9 cleavage products. b, Endonuclease activity assay as in panel a using linearized plasmid containing a 5′-TAA-3′ PAM. c, Endonuclease activity assay as in panel a using linearized plasmid containing an extended 5′-TAAAA-3′ PAM.

Extended Data Figure 6. PAM-dependent interaction of the +1 phosphate with the phosphate lock loop

a, Comparison of the bound target DNA (left) and the modelled B-form DNA (right). Docking of the ideal B-form duplex yields a steric clash with the phosphate lock loop. The arrow indicates the rotation of the +1 phosphate group (+1P) needed for interaction with the phosphate lock loop. b, Comparison of the phosphate lock loop and the +1 phosphate positions in the Cas9-sgRNA-DNA complex containing a PAM (left) and the Cas9-sgRNA-ssDNA target complex (right). Molecule A from the crystallographic asymmetric unit of the Cas9-sgRNA–ssDNA complex is shown. In molecule B, the nucleotides upstream of the +1 phosphate are structurally ordered due to crystal packing interactions, and the +1 phosphate is positioned within hydrogen-bonding distance as a result. Numbers indicate interatomic distances in Å. c, Superposition of the two structures shown in panel b.

Extended Data Figure 7. Endonuclease activity of phosphate lock loop Cas9 mutants against mismatch- and bubble-containing DNA substrates

a, Endonuclease activity assay of WT and mutant Cas9 proteins using double-stranded oligonucleotide DNA containing a target sequence fully complementary to sgRNA used in Fig. 1a. Samples were taken after 15 sec, 30 sec, 1 min, 2 min, 5 min, 15 min, 1 h, and 2 h. b, Endonuclease activity assay as in panel a but using an oligonucleotide duplex containing mismatches to the sgRNA at positions 1–2. c, Endonuclease activity assay as in panel a but using an oligonucleotide duplex consisting of a target sequence with mismatches to the sgRNA at positions 1–2 and a non-target sequence mismatched to the target strand at positions 1–2. d, Quantification of cleavage defects observed with mismatch- and bubble-containing substrates from a–c. For each protein, the amount of cleaved product obtained after 2 h was normalized to the amount of product obtained from perfectly complementary DNA substrate.

Figure 1. Crystal structure of Cas9 in complex with a sgRNA and a PAM-containing target DNA

a, Schematic diagram of guide and target nucleic acids. Guide RNA is coloured orange, target DNA strand in light blue, and non-target DNA strand in black. The 5′-NGG-3′ PAM trinucleotide in the non-target strand is highlighted in yellow. The colour code is used throughout. Empty ovals denote nucleotides not observed in electron density. b, Orthogonal views of the sgRNA–target DNA four way junction. c, Front and rear views of the Cas9–sgRNA–DNA complex.

Figure 2. The GG dinucleotide of the PAM is read out by major groove interactions

a, Zoomed-in view of the PAM binding region in Cas9. b, Schematic of Cas9 interactions with the PAM duplex. Red circles denote bridging water molecules. c, Detailed view of the major groove. Sequence-specific hydrogen-bonding interactions with the GG PAM dinucleotide are indicated with dashed lines. d, Electrophoretic mobility shift assay using catalytically inactive dCas9–sgRNA complexes and fluorophore-labelled target DNA duplex. e, Endonuclease activity assay of wild type (WT) and mutant Cas9 proteins using a linearized plasmid DNA containing a target sequence fully complementary to the sgRNA in Fig. 1a. 2104 and 598 bp bands correspond to Cas9 cleavage products.

Figure 3. Interactions with the +1 phosphodiester group orient the target strand for guide RNA binding

a, Detailed view of the minor groove of the PAM region. b, Hydrogen-bonding interactions (dashed lines) of the +1 phosphate (+1P) with the Lys1107-Ser1109 (phosphate lock) loop. c, Superposition of the unwound target DNA strand with an ideal B-form DNA duplex (green) d, Endonuclease activity assays using linearized plasmid DNA containing a fully complementary target sequence (top) or a target sequence mismatched to the sgRNA at positions 1–2 (bottom). e, Crystal structures of dCas9-sgRNA bound to DNA substrates containing mismatches to the sgRNA at positions 1–2 (top) and 1–3 (bottom), overlaid with refined 2mF_o-DF_c electron density maps (grey mesh, contoured at 1σ). The sgRNA is identical to that in Fig. 1a. In both structures, the target DNA strand is provided in two fragments, as indicated in the schematics. Residual electron density corresponding to the +1 base pair is indicated with a red arrowhead.

Figure 4. Model for PAM-dependent target DNA unwinding and recognition by Cas9

Guide RNA binding to Cas9 results in the formation of the PAM binding site. Cas9-RNA engages the PAM GG dinucleotide using Arg1333 and Arg1335, and positions the target DNA duplex such that the +1 phosphate (orange circle) interacts with the phosphate lock loop, resulting in local strand separation immediately upstream of the PAM. Base-pairing between displaced target DNA strand and the seed region of the guide RNA promotes further stepwise strand displacement and propagation of the guide-target heteroduplex.