Structure and mechanism of the CMR complex for CRISPR-mediated antiviral immunity

RNA content of the CMR complex

To determine the specific characteristics of the RNA component of SsoCMR, isolated RNA was cloned and deep-sequenced. 1.88 million sequence reads of 36 nt length were mapped onto the S. solfataricus spacers (Table S1). Analysis of the start and stop positions of the sequence reads for a subset of highly represented spacers from the A and D loci revealed the presence of sequence corresponding to the 5′-tag derived from the CRISPR repeat, with a clear demarcation at the 8^th nucleotide, which corresponds to the site of cleavage by Cas6 (Carte et al., 2008; Lintner et al., 2011) (Figure S2). The 3′ ends of the sequenced RNA were more variable. Some spacers such as A2 and D43 displayed a short 3′ handle whilst others appeared to have very little repeat-derived sequence at the 3′ end. Overall, this fits the suggestion that crRNA is processed by the Cas6 endoribonuclease followed by exonucleolytic digestion of the 3′ end (Hale et al., 2009). By contrast, crRNA isolated from the S. solfataricus CASCADE complex still includes the 3′ repeat derived sequence (Lintner et al., 2011), suggesting that crRNAs are differentially processed depending on their ultimate destination in CASCADE or CMR.

S. solfataricus P2 has six CRISPR loci, denoted A-F, ranging from 8 to 103 repeats in length (Lillestol et al., 2006). Whole-genome transcription data suggested that loci A, B, C and D are all highly transcribed, whilst E and F are very weakly transcribed (Wurtzel et al., 2010). Examples of spacers derived from all six loci were detected in the CMR complex, but the distribution was highly biased. The majority of CMR-bound crRNAs were derived from locus D, followed by locus C, with significant under-representation from the other four loci (Figure 2). Within a locus, spacer representation in the CMR complex was highly variable, with numbers of sequenced crRNAs from adjacent spacers frequently differing by several orders of magnitude (Figure 2). In particular, one spacer (D63) accounted for 45% of the sequence reads, and the top ten spacers accounted for 79 % of the total reads mapped to spacers (Table S2). Reverse transcripts of CRISPR arrays have been detected in Sulfolobus species (Lillestol et al., 2009) but their significance remains unclear. Only 209 RNAs corresponding to reverse transcripts were sequenced, corresponding to 0.01% of the total (Table S1).

Open in a separate window

Figure 2

Distribution of crRNA bound by the S. solfataricus CMR complex

Examples of crRNA’s from all six CRISPR loci were observed, with a clear bias towards locus D, followed by locus C. The individual plots for each locus show that crRNA representation was highly variable, with adjacent spacers represented at levels that often varied by several orders of magnitude. For each graph, the X-axis represents the position of each spacer in the locus and the Y-axis represents the number of sequenced matches to each spacer. 88% of the total sequence reads were derived from locus D, which represents a nearly four-fold over-representation compared to the proportion of crRNAs encoded by that locus. CRISPR loci E and F, which are poorly transcribed, were significantly under-represented, as expected. However, crRNAs from CRISPRs A and B, which are highly transcribed and thought to be actively adding spacers in vivo, are also significantly under-represented in the CMR complex. The table shows the properties and representation in CMR of each CRISPR locus.

The crystal structure of Cmr7

The crystal structure of Cmr7 (Sso1986) was solved to 2.05 Å resolution (PDB 2X5Q) (Figure 3). Sso1986 exhibits a novel fold consisting of six β-sheets and four α-helices, forming a dimer with a buried surface area of 1177 Ų and a concave face (Figure 3, S3A) Homologs of Cmr7 are only detectable in the Sulfolobales, but given the rapid evolution of the CRISPR system it is possible that distant homologs sharing this fold exist more widely. To help determine the most important features of Cmr7, we solved the structure of a second Cmr7 subunit, Sso1725. The sequence similarity between the two proteins is low (19 % identity) and the crystal structure of Sso1725 (solved to 2.08 Å resolution, Table 1) shows limited structural similarity (Cα rmsd of 2.39 Å over 121 atoms for each monomer). The majority of the residues conserved in the six known orthologs of Cmr7 are located in the dimer interface and concave face (Figure 3C and S3). The electrostatic surfaces of the two proteins show broad similarities at the dimer interface, with symmetrical patches of negative charge at the poles and positive charge at the edges (Figure 3D). Given this level of conservation on only one face of the protein, we hypothesize that this region is important for function, possibly as a binding site for other CMR subunits. The stoichiometry of 3 dimers of Cmr7 per CMR complex is most easily accommodated if they form a trimeric (pseudo-hexamer) structure in the context of the complex.

Open in a separate window

Figure 3

Crystal structures of two members of the Cmr7 family, viewed from the concave face of the dimer

A. The Cmr7 proteins Sso1986 (left) and Sso1725 (right) contain a structurally conserved core (yellow) and a variable region (blue and cyan for Sso1986 and Sso1725 respectively). The β13-β14 loop of Sso1986 is disordered and is represented as a dashed, black line. The N- and C-termini are represented as blue and red spheres respectively. B. Sso1986 and Sso1725 both form dimers and the structurally conserved core is located at the dimer interface. The interface itself is also conserved between the two proteins. C. The structurally conserved residues (green) and secondary structure (yellow) are located close to the dimer interface with a significant proportion positioned at the concave face. D. Electrostatic surface images show that the region of the concave face proximal to the dimer interface in both proteins (black box) have broad similarities.

Electron Microscopy reveals the architecture of the CMR complex

To elucidate the three-dimensional architecture of the CMR complex, we performed electron microscopy coupled to single particle analysis experiments for the full complex and additionally for Cmr2/3/7 sub-complex, devoid of RNA (Figure 4 and S4). Projections of the maps matched with 2D averages assigned the same Euler angles (Figure S4), highlighting the validity of the maps. The resolution for both maps was ~25 Å, calculated at 0.5 Fourier Shell Correlation (3 sigma). The CMR complex exhibited cavities, compatible with an RNA threading machine. There was no obvious similarity to the “seahorse” structure of E. coli CASCADE (Jore et al., 2011). The Cmr2/3/7 sub-complex, which lacked bound crRNA, had overall dimensions of 90x120x110 Å, organized in a clamp or “crab claw” structure (Figure 4A). The intact CMR complex with loaded crRNA had overall dimensions of 160x120x110 Å with an upper ‘crab claw’ connected to a protruding region (Figure 4B). These dimensions were compatible with the expected molecular masses of ~290 and ~430 kDa, respectively. The Cmr2/3/7 sub-complex fitted well to the upper region of intact CMR (Figure 4C), consistent with a role as a scaffold for the assembly of the other Cmr subunits around the periphery (black arrows in Figure 4C).

Open in a separate window

Figure 4

3D EM visualization of CMR complex

A. Surface representation of the Cmr2/3/7 sub-complex devoid of crRNA. B. Surface representation of the full CMR complex with bound crRNA. C. Superposition of Cmr2/3/7(blue surface) on CMR/RNA (black mesh). Black arrowheads point to regions of additional density on the full CMR complex with bound crRNA compared to Cmr2/3/7. Grey arrows indicate dimensions in Angstroms.

Ribonuclease activity of SsoCMR

We tested the ability of SsoCMR to recognize and cleave RNA targets corresponding to spacers A1 and D63 in vitro. Both RNA targets were cleaved efficiently when a cognate crRNA (guide RNA) with an 8 nt 5′-tag was present (Figure 5). Manganese ions were essential for this activity and magnesium could not substitute. ATP was not essential but clearly stimulated the cleavage reaction (Figure S5). No crRNA-directed cleavage of DNA targets was observed (data not shown). To rule out the possibility of activity from a contaminating ribonuclease, the CMR complex was immuno-depleted using antibodies raised against the Cmr7 subunit. Immuno-depletion abolished the nuclease activity, suggesting strongly that the activity was associated with the CMR complex (Figure S5A). As a further control, native untagged SsoCMR purified from S. solfataricus cell extract by immuno-precipitation using the anti-Cmr7 antibody had the same activity as the column purified, tagged protein complex (Figure S5B).

Open in a separate window

Figure 5

Characterization of the S. solfataricus CMR complex activity in vitro

A. Radiolabeled target RNA (5 μM) corresponding to spacer A1 was incubated with the Cmr7-tagged SsoCMR complex (0.5 μM) and guide RNA (1 μM). Cleavage of the target RNA was observed at six sites (labeled 1-6) (a). 3′ truncation of the guide RNA did not change the cleavage pattern of the target RNA, ruling out a molecular ruler mechanism (b). The 8 nt 5′-tag was essential for cleavage activity, as either deletion (c) or replacement with an 8A sequence (d) abolished nuclease activity.

B. The D63 target (0.5 μM), was cleaved in the presence of Cmr7-tagged SsoCMR complex (0.5 μM), cognate guide RNA (0.1 μM) (e). Deletion of the unpaired 3′ end of the target RNA, which corresponds to the position of the PAM sequence, abolished activity (f). The presence of an unpaired 6A sequence at this position restored activity (g).

C. Radiolabeled guide RNA corresponding to spacer A1 (3 μM) was incubated with the Cmr7-tagged SsoCMR complex (0.5 μM) and cognate target RNA (1 μM). Cleavage was observed at up to five positions, labeled 2′ to 6′ (h). This activity was dependent on the presence of the 3′ unpaired spacer sequence (i) and was not influenced by the presence of an additional unpaired extension at the 3′ end of the guide RNA (j, k). For each figure part, the 5′ end labeled RNA strand is indicated with an asterisk. Labeled Decade RNA markers (Ambion) are shown.

Features of guide and target RNAs important for cleavage by CMR

The sequence and structural requirements of RNA cleavage by CMR were investigated by constructing a range of target and guide RNA molecules based on the spacer A1 and D63 sequences. To test for a molecular ruler mechanism as observed for PfuCMR, we reduced the length of the guide RNA (Figure 5A, panel b). The cleavage sites did not move in register with the 3′ end of the guide RNA, suggesting that Sulfolobus and Pyrococcus CMR differ fundamentally in this respect. Deletion of the repeat-derived 5′-tag from the guide RNA abolished cleavage activity (Figure 5A, panel c), and could not be rescued by substitution with a 5′ 8A sequence (panel d), showing that this 5′-tag sequence is essential for cleavage, ruling out the possibility of a contaminating nuclease. The presence of an unpaired flap at the 3′ end of the target RNA was also required for activity (panel f). This corresponds to the position of the Protospacer Associated Motif (PAM) that is essential for cleavage of viral DNA targets by CASCADE (Gudbergsdottir et al., 2011; Lintner et al., 2011). However, for target RNA cleavage by CMR the PAM sequence at this position was not essential, as a 6A sequence could substitute (Figure 5B, panel g).

In addition to the target RNA, the guide RNA strand could also be cleaved by SsoCMR (Figure 5C, panel h). Cleavage of the guide RNA was dependent on the presence of a 3′ overhang on the target RNA (Figure 5C, panel I, k). Guide and target were cleaved at approximately equal rates when present at an equimolar ratio, but at ratios of 20:1 or 5:1 excess of target over guide, the guide RNA was cleaved significantly more slowly. Under these conditions, multiple turnover cleavage of the target RNA was observed, suggesting that cleavage of the guide RNA was not essential for catalysis (Figure S5D).

Structure of the CMR complex

The EM envelope of the Cmr2/3/7 sub-complex bears some similarity to the clamp region of RNA polymerase. In particular, the distance between the two sides of the “claw” is ~30 Å wide, and their length is ~40 Å. It is tempting to view this feature as a dsRNA binding cleft, particularly as Cmr2 is assumed to harbor the active site of the CMR complex. In support of this, the cleft is somewhat deeper in the Cmr2/3/7 subcomplex, which contains no bound crRNA, in comparison to the full complex. The lack of crRNA in the Cmr2/3/7 sub-complex fits with a presumed role for the RAMP-containing Cmr subunits (Cmr1, 4 and 6) in RNA binding (Makarova et al., 2011a). This is consistent with the recent prediction that Cmr3 is most closely related to the Cas5 subunit of CASCADE, whilst Cmr1, 4 and 6 are closer matches to the Cas7 subunit (CasC in E. coli), which is known to bind crRNA (Makarova et al., 2011a). Compared to Cmr2/3/7, the additional Cmr subunits are distributed mainly at the front and at the bottom of the complete CMR complex, and may form a structure related to the crRNA binding CasC backbone of CASCADE. Visualization of the path of RNA in the CMR structure remains a key aim to help elucidate the mechanism and molecular organization.

The mechanism of SsoCMR in viral RNA cleavage

We have demonstrated that the SsoCMR complex cleaves target RNA in a sequence specific manner that is dependent on the presence of a guide crRNA with a cognate sequence and an 8 nt repeat-derived tag at the 5′ end. Cleavage occurs at UA dinucleotides, generating products with 3′-OH and 5′-phosphate ends, like those produced by Argonaute and Piwi (Hutvagner and Simard, 2008). The cleavage sequence is extremely common in the Sulfolobales and their viruses. Of the >400 spacers in the CRISPR loci of the S. solfataricus P2 genome, only 11 lack a UA dinucleotide. The palindromic nature of this dinucleotide ensures that an equivalent UA sequence is always present in the guide RNA. Cleavage of both the target and guide RNA at UA sites was observed, with the guide cleaved significantly more slowly at high ratios of target:substrate. Under these conditions, one guide RNA molecule could support the cleavage of several target RNAs, demonstrating that cleavage of the guide is not essential for target RNA destruction.

Although not yet shown directly, the N-terminal permuted HD nuclease domain present in Cmr2 is generally assumed to be the nuclease site of the CMR complex. The distantly related HD domain present in Cas3 has been shown capable of cleaving both RNA and DNA, although they are specific for single stranded nucleic acids and generate products with 3′-cyclic phosphates (Mulepati and Bailey, 2011; Beloglazova et al., 2011). It is also possible that the RNA cleavage activity resides elsewhere in the Cmr2 subunit, or in one or more of the RAMP-containing subunits (Cmr 1,2,4 and 6), which are distantly related to Cas5, 6 and 7 (Makarova et al., 2011a).

We observed no requirement for a PAM sequence adjacent to the protospacer in RNA targets. This is a marked difference from the DNA targeting CASCADE in archaea, which only cleaves substrates containing a PAM (Gudbergsdottir et al., 2011; Manica et al., 2011). The primary role of the PAM may be to maintain a mismatched region between the 5′-tag of the crRNA and the sequence immediately adjacent to the spacer of the target. For DNA targeting systems, this ensures that the chromosomal CRISPR locus is not targeted for cleavage (Marraffini and Sontheimer, 2010b; Mojica et al., 2009). Given that the CMR complex only targets viral RNA, there is no requirement for PAM detection to operate in this case as the host genome will not be a target. Although no PAM is required, CMR-mediated cleavage requires an unpaired RNA region at the 3′ end of the target RNA, downstream of the protospacer. This discrimination may be at the structural rather than sequence level and is consistent with a role in vivo in targeting viral mRNA sequences, which will typically be considerably longer than the guide RNA species.

The observed activity of SsoCMR differs markedly from that reported previously for the enzyme from P. furiosus (Hale et al., 2009). PfuCMR operates by a molecular ruler mechanism without sequence dependent cleavage, generates 3′-cyclic phosphate products, and does not require an extension at the 3′ end of the target RNA, at least in vitro. The long evolutionary distance between the two species may explain these differences. CMR complexes have been classified into five families (A-E) on the basis of the sequence of the large subunit, Cmr2 (Garrett et al., 2011). On this basis, the SsoCMR complex belongs to family B whilst PfuCMR is one of the very few representatives of family C. To put these differences in context, the type IIIA, CMR-like complex from Staphylococcus epidermidis targets DNA rather than RNA (Marraffini and Sontheimer, 2008). This may therefore reflect the plasticity inherent in the CRISPR system.

The RNA targeting functionality of the CMR complex in prokaryotes has parallels with the eukaryal piRNA pathway that uses guide RNA to recognize and cleave the mRNA of mobile genetic elements (Aravin et al., 2007). As in the CRISPR system, in the piRNA pathway the small guide RNAs are generated by cleavage of a long mRNA transcript, loaded into an endoribonuclease (Piwi) and used to target and degrade the mobile mRNA by means of dsRNA cleavage yielding products with 3′-OH termini (reviewed in (Nowotny and Yang, 2009)). There are, though, important differences. Piwi recognizes the 5′ phosphate of the guide RNA specifically (Ma et al., 2005; Parker et al., 2005) whilst the guide RNA generated by Cas6 and utilized by CMR lacks a 5′ phosphate and has an essential 8 nt 5′-tag. More fundamentally, there is no obvious homology between the Piwi and CMR proteins. Piwi uses an RNaseH domain for dsRNA cleavage, which is absent from the CMR complex. The stimulation of CMR cleavage activity by ATP suggests that an ATP-driven conformational change may be utilized to reposition the dsRNA with respect to the active site. In the absence of any detectable Walker A or B motifs, the likely site for ATP binding is the C-terminal polymerase/cyclase domain of Cmr2. This domain has no known function but is expected to bind nucleotide triphosphates. If such an ATP-dependent RNA repositioning mechanism were in operation it would constitute another novel aspect of the CMR complex.

Antibody generation

Sheep polyclonal antibodies were raised against the recombinant Cmr7 (Sso1986) protein and supplied by the Scottish National Blood Transfusion Service, Pentlands Science, Park, Midlothian.

RNA Oligonucleotides used for CMR activity assays

RNA oligonucleotides were chemically synthesized (Integrated DNA Technologies), end labeled with ³²P-ATP and purified by denaturing gel electrophoresis. The sequences used are listed in supplementary experimental procedures.

Purification of the native CMR complex from S. solfataricus

S. solfataricus strain P2 biomass was grown as described previously (Götz et al., 2007). The CMR complex was purified over four column chromatography steps and purification was followed using an antibody raised against subunit Cmr7, as described in the supplementary experimental procedures. This yielded the homogeneous complex shown in Figure 1.

Expression and purification of tagged CMR complex in S. solfataricus

This was carried out as described previously by cloning the relevant gene into entry vector pMZ1 (Zolghadr et al., 2007) followed by subcloning into expression vector pSVA9, expressing the relevant subunit with a C-terminal strep-his tag (Albers et al., 2006), described in detail in the supplementary experimental procedures. The Cmr2/3/7 sub-complex analyzed by electron microscopy was obtained during purification of the CMR complex with a tagged Cmr3 subunit. The sub-complex eluted separately from the full complex on gel filtration and contained no bound crRNA (Figure S5).

RNA isolation and sequencing

RNA was extracted from the purified native CMR complex by the classical phenol/chloroform method followed by ethanol precipitation and vacuum desiccation. Dried RNA was resuspended in 5 μl of water and directly labeled in a 10 μl reaction containing polynucleotide kinase and 2 μCi γ³²P-ATP. Labeled RNAs were analyzed by electrophoresis on a 15% acrylamide, 7M urea, TBE denaturing gel and visualized by phosphorimaging. For crRNA deep sequencing, small RNA sequences were generated by the GenePool at the University of Edinburgh using the Illumina small RNA prep kit v1 and subject to high-throughput sequencing using a Genome Analyzer IIx. This resulted in the addition of the adaptor sequence TCGTATGCCGTCTTCTGCTTG at the 3′ end of each sequence. The adaptor sequence was trimmed away from the reads with a bespoke perl script. Reads were mapped against the S. solfataricus P2 genome with BWA (Li and Durbin, 2009) using default parameters and converted into BAM using SAMtools (Li et al., 2009). Of the 2527217 reads 1997151 were mapped (79 %). The number and strand orientation of the reads mapping to each spacer were quantified. The raw data from the sequencing run is available from the corresponding author on request, and the number of sequences mapping onto each spacer are listed in table S1. The raw sequence data has been uploaded to the Sequence Read Archive with accession number ERP001053 (www.ebi.ac.uk/ena/data/view/ERP001053).

RNA cleavage assays

Purified SsoCMR complex and unlabeled guide RNA were mixed in buffer (20 mM Mes^.HCl pH 6.0, 100 mM potassium glutamate, 10 mM DTT, 10 mM MnCl₂ and the RNase inhibitor SUPERase.In (Ambion)) and pre-incubated at room temperature for 10 min prior to the addition of 5′-³²P-end-labeled synthetic target RNA to the reaction mix. Target and guide RNA and CMR complex concentrations are indicated in the figure legends. The reaction was further incubated at 75 °C for 10 min in standard assays, or for the time indicated in the figure. Reactions were stopped by chilling on ice and addition of formamide loading buffer. Samples were separated on 20 % polyacrylamide, 7 M urea, 1 x TBE gels. Electrophoresis was completed at 90 W, 50 °C for 90 min and the gels visualized by phosphor-imaging. 5′ end-labeled RNA size standards (Decade Markers, Ambion) were used to determine the sizes of the observed products. Cas 6 activity was assayed as described previously (Lintner et al., 2011).

Crystallography of Sso1725

Sso1725 crystals were grown at 20 °C using the sitting drop vapor diffusion method. A reservoir of 0.15 M sodium acetate pH 5.6, 2.0 M ammonium sulfate was used and protein was mixed with the precipitant at a ratio of 2:1. Crystals were cryo-protected by successively soaking in solutions of reservoir containing 8, 16, 20 and 25 % glycerol before freezing in liquid nitrogen. A native dataset at 2.08 Å resolution was collected at Diamond Light Source (Beamline I03). An anomalous SAD dataset was collected at the same beamline using a native crystal briefly soaked in reservoir solution containing 42 mM samarium chloride before cryo-protecting as described above. Datasets were processed and refined using the methods described in the supplemental experimental procedures. The coordinates were deposited in the PDB under the accession code 2XVO.

Thanks to Paul Talbot and the University of St Andrews Mass Spectrometry service for expert technical assistance.

Funding This project was funded by the Biotechnology and Biological Sciences Research Council grant number BB/G011400/1. The Electron Microscopy Facility at Edinburgh is supported by the Scottish Alliance for Life Sciences and the Wellcome Trust (WT087658MA). JR and SVA were supported by intramural funds of the Max Planck Society.