Sequence- and structure-specific RNA processing by a CRISPR endonuclease

In bacteria and archaea, a specific class of short RNAs participates in an adaptive immune pathway that targets viruses and plasmids (1–5). Fragments of foreign DNA are integrated into clustered regularly interspaced short palindromic repeat (CRISPR) loci (6), from which they are transcribed as parts of long RNAs that include a repetitive sequence element characteristic of the host organism. These precursor CRISPR RNAs (pre-crRNAs) are post-transcriptionally processed into 60-nucleotide crRNAs that serve as homing oligonucleotides to prevent propagation of invading viruses or plasmids harboring cognate sequences (1, 7). Each of the eight known CRISPR subtypes (8–12) also includes a set of open reading frames encoding CRISPR-associated (Cas) proteins, which can be difficult to identify or classify due to extreme primary sequence divergence (3, 5, 11). Pseudomonas aeruginosa UCBPP-PA14 (hereafter Pa14), a Gram-negative opportunistic pathogen harboring a CRISPR/Cas system of the Yersinia subtype, contains six Cas genes flanked by two CRISPR elements (Fig. 1A). While Cas1 is found universally among CRISPR-containing organisms, and Cas3 is evident in most subtypes, Csy1-4 genes are unique to the Yersinia subtype. Both CRISPR elements comprise a characteristic arrangement of 28-nucleotide near-identical repeats interspersed with unique ^~32-nucleotide spacers, some of which match sequences found in bacteriophage or plasmids (13). Pre-crRNA transcripts are post-transcriptionally cleaved to yield crRNAs that each contain one unique sequence flanked by sequences derived from the repeat element (1–2, 14–17).

To identify the protein(s) responsible for producing crRNAs from pre-crRNAs in the Yersinia CRISPR subtype, we recombinantly expressed each of the six Cas proteins from Pa14 and tested them for endoribonuclease activity using an in vitro transcribed pre-crRNA (data not shown). Based on the observation of sequence-specific pre-crRNA processing when pre-crRNA was incubated with Csy4, we concluded that Csy4 is the endoribonuclease responsible for crRNA biogenesis (Fig. 1B). CRISPR transcript cleavage is a rapid, metal ion-independent reaction, as observed for crRNA processing within two other CRISPR/Cas subtypes (1–2). Csy4 cleaves Pa14 pre-crRNA within the repeat element at the base of a predicted stem-loop structure, generating 60-nucleotide crRNAs consisting of a 32-nucleotide unique (phage-derived) sequence flanked at the 5’ and 3’ ends by eight and 20 nucleotides of the repeat sequence, respectively (Fig. 1A). Csy4 does not cleave pre-crRNA from Streptococcus thermophilus, which has a repeat stem-loop of a distinct sequence from Pa14 (Fig. 1B).

For Csy4 to be effective, we hypothesized that its RNA recognition mechanism must be highly specific in order to target only CRISPR-derived transcripts and not other cellular RNAs. To test this, Csy4 was expressed in Escherichia coli either alone or together with a synthetic Pa14 CRISPR RNA consisting of eight repeat sequences (derived from the CRISPR locus proximal to the Cas1 ORF) and seven identical spacer sequences. In spite of its highly basic composition (pI=10.2), Csy4 did not associate with endogenous cellular nucleic acids. However, when co-expressed with a Pa14 CRISPR transcript, the protein co-purified with a protected ^~19-nucleotide crRNA fragment (Fig. 1C). These observations underscored the specificity of Csy4 recognition, leading us to explore the protein/RNA interactions required for Csy4 substrate recognition and cleavage. Csy4 binding and activity assays were performed in vitro using RNA oligonucleotides corresponding to different regions of the 28-nucleotide Pa14 CRISPR repeat sequence (data not shown). Using this approach, we identified a 16-nucleotide minimal RNA fragment, consisting of the repeat-derived stem-loop and one downstream nucleotide, that is sufficient for Csy4-catalyzed cleavage. Cleavage assays utilizing this minimal RNA as a substrate showed that Csy4 activity requires the presence of a 2’-hydroxyl group on the ribose immediately upstream of the cleavage site. A 2’-deoxyribonucleotide substitution at this position completely abrogates cleavage, but does not disrupt Csy4 binding (fig. S1, A and B).

To obtain structural insights into pre-crRNA recognition and cleavage, we co-crystallized Csy4 in complex with the non-cleavable 16-nucleotide minimal RNA substrate described above (fig. S2). Three unique crystal forms of the complex were obtained; one contained wild-type Csy4 and two contained a catalytically active point mutant (S22C) of Csy4 (fig. S3). The crystal structures of the Csy4-RNA complex in the three crystal forms were solved to a resolution of 2.3 Å, 2.6 Å and 1.8 Å, respectively (table S1). In all three crystal forms, the RNA binds to Csy4 in an almost identical manner in which the protein makes extensive interactions with the ssRNA-dsRNA junction at the base of the crRNA stem as well as with the major groove of the RNA hairpin (Fig. 2A). The RNA hairpin is clamped into a highly basic groove between the main body of the protein and an arginine-rich helix (α3, residues 108–120) that inserts into the RNA major groove (Fig. 2B).

In the complex, Csy4 adopts a two-domain architecture consisting of an N-terminal ferredoxin-like domain (residues 1–94) and a C-terminal domain (residues 95–187) that mediates most of the interactions with the RNA (fig. S4A). At the sequence level, Csy4 shares less than 10% identity with the two other known endoribonucleases involved in crRNA biogenesis, CasE from Thermus thermophilus (18) and Cas6 from Pyrococcus furiosus (2). The crystal structures of CasE and Cas6 in their nucleic acid-free states showed that these proteins possess a duplicated ferredoxin fold. The N-terminal ferredoxin fold is preserved in Csy4; structural superpositions using the DALI server (19) indicate that Csy4 in its RNA-binding conformation superposes with CasE and Cas6 with root-mean-square deviation (rmsd) of 3.8 Å (over the N-terminal 111 Cα atoms) and 3.9 Å (over 104 Cα atoms), respectively. Although the C-terminal domain of Csy4 (residues 95–187) shares the same secondary structure connectivity as a ferredoxin-like fold, its conformation is markedly different due to the position of the bound RNA (fig. S4B).

The crRNA substrate forms a stem-loop structure, as predicted for this subclass of crRNA repeats (20). Nucleotides 6–10 and 16–20 base pair to produce a regular A-form helical stem. The GUAUA pentaloop contains a sheared G11-A15 base pair and an extruded nucleotide U14. This closely resembles the structures adopted by GNR(N)A pentaloops found in the yeast U6 small nuclear RNA intramolecular stem-loop (21) and in bacteriophage lambda boxB RNA (22). In the Csy4-RNA complex, the RNA stem-loop straddles the β-hairpin formed by strands β6-β7 of Csy4, with the C6-G20 base-pair directly stacking onto the aromatic side chain of Phe 155 (Fig. 2C). Thus, in the context of the full-length crRNA repeat, the role of Phe 155 is to recognize the ssRNA-dsRNA junctions in the pre-crRNA substrate. This anchors the RNA stem and orients it at the proper angle to permit sequence-specific interactions in the major groove.

Arg 102 and Gln 104, located in a linker segment connecting the body of Csy4 to the arginine-rich helix, make sequence-specific hydrogen bonding contacts in the major groove of the RNA stem to nucleotides G20 and A19, respectively (Fig. 2B). The Csy4-crRNA interaction is further stabilized by the insertion of the arginine-rich helix into the major groove of the RNA hairpin in the proximity of the extruded nucleotide U14 (Fig. 2C). The side chains of Arg 114, Arg 115, Arg 118, Arg 119 and His 120 contact the phosphate groups of nucleotides 7–12. Additionally, the sidechain of Arg 115 hydrogen-bonds to the base of G6 as the only sequence-specific interaction between the arginine-rich helix and the RNA hairpin. Interestingly, the binding of the Arg-rich helix to the major groove of the crRNA hairpin is reminiscent of the recognition mechanism employed by certain highly basic viral peptides for dsRNA binding, as observed in the N-peptide/boxB RNA interaction in lambdoid phages (23) (fig. S5, A and B) and in lentiviral Rev-RRE and Tat-TAR complexes (24–25).

Csy4 recognizes the hairpin element of the CRISPR repeat sequence and cleaves immediately downstream of it. In the Csy4-RNA complex structure, where RNA cleavage is abrogated by the introduction of a 2’-deoxy modification in nucleotide G20, ordered electron density is only evident for the scissile phosphate between G20 and C21. The ribose and cytosine moieties of C21 are not resolved and presumably disordered. The scissile phosphate binds in a pocket located between the β6-β7 hairpin turn on one side and helix α1 on the other (Fig. 3A), hydrogen-bonding to the side chain of His 29 and the backbone amide of Gln 149. Ser 148 is adjacent to the 2’ ribose carbon atom of nucleotide G20 (4.6 Å) and may make a hydrogen-bonding interaction with the 2’-hydroxyl group of G20 in a bona fide pre-crRNA substrate.

Both His 29 and Ser 148 are invariant among Csy4 homologs (fig. S6). In addition, a strongly conserved tyrosine (Tyr 176) is also positioned near the scissile phosphate. To confirm the functional roles of these residues in pre-crRNA processing, we introduced point mutations at each of these residues and tested the mutant proteins for cleavage activity in vitro (Fig. 3B). Mutations of His 29 or Ser 148 (to alanine and cysteine, respectively) completely abolished cleavage activity without disrupting RNA binding (Fig. 3B, data not shown). By contrast, mutation of Tyr 176 to phenylalanine did not disrupt activity, indicating that Tyr 176 may play a role in orienting His 29 but does not directly participate in catalysis. Additional mutants were generated to probe the requirements for individual sequence- and structure-specific interactions in pre-crRNA processing. Alanine substitution of Arg 102 abolished pre-crRNA processing in vitro, whereas mutation of Gln 104 to alanine did not significantly disrupt activity (Fig. 3B). This suggests that Arg 102, which recognizes the terminal C6–G20 base pair, is key to properly orienting the RNA substrate but that Gln 104 is dispensable for crRNA processing. An alanine mutation at Phe 155 severely impaired crRNA biogenesis, suggesting that this residue also plays an important role in substrate recognition.

These mutational data suggest that the interaction between Csy4 and the closing base pair of the RNA stem is crucial for pre-crRNA processing, but that the sequence-specific interaction between Csy4 and the penultimate base pair in the stem is less important. To probe the Csy4-RNA interactions required for cleavage, we incubated Csy4 with a panel of short RNA oligonucleotides containing a variety of mutations in the CRISPR repeat stem-loop sequence (fig. S7). These data confirm that Csy4 requires a C6–G20 base pair closing the RNA stem and that Csy4 can accommodate different nucleotides at the penultimate RNA base pair.

Csy4-catalyzed pre-crRNA cleavage requires the presence of the 2’-hydroxyl group in the nucleotide immediately upstream of the cleavage site. This suggests that the catalytic mechanism of Csy4 proceeds through a 2’-3’ cyclic intermediate or yields a 2’-3’ cyclic product. In this context, the observation of a strictly conserved serine residue (Ser 148) adjacent to the 2’ position is unprecedented and points to Ser 148 playing a role in activating or positioning the 2’-hydroxyl for a nucleophilic attack on the scissile phosphate. How the serine sidechain may function in this context remains to be determined. Histidine, on the other hand, is commonly found in metal ion-independent active sites, such as in RNase A (26). Although mutation of His 29 to alanine leads to loss of activity in Csy4, mutation to lysine partially restores activity (fig. S8), suggesting that His 29 acts as a proton donor in stabilizing the 5’ oxygen leaving group.

Phylogenetic analysis of CRISPR loci suggests that CRISPR repeat sequences and structures have co-evolved with the Cas genes (20). The similarity of Csy4 at the fold level to the CRISPR-processing endonucleases CasE and Cas6 suggests that collectively they are likely to have descended from a single ancestral endoribonuclease enzyme that has diverged throughout evolution. The structure described here reveals how Csy4 and related endonucleases from the same subfamily utilize an exquisite recognition mechanism to discriminate crRNA substrates from other cellular RNAs and illustrates the importance of co-evolution in shaping molecular recognition mechanisms in the CRISPR pathway.