The role of Cas8 in type I CRISPR interference

CRISPR (clustered regularly interspaced short palindromic repeat) systems were discovered in Streptococcus thermophilus [1], by providing adaptive immunity to invasive genetic elements, recently reviewed in [2,3]. Immunity arises from base pairing of host encoded CRISPR RNA (‘crRNA’) with invader DNA/RNA, promoting nucleolytic degradation of the invader, processes called ‘interference’. DNA sequences that are targeted by crRNA are called ‘protospacers’ and can be identified from an archive of previously encountered protospacers arrayed in a CRISPR locus as ‘spacers’, separated from one another by repeat sequences. ‘Adaptation’ or ‘spacer acquisition’ processes furnish CRISPR with new spacer-repeat units requiring two highly conserved CRISPR-associated (‘Cas’) proteins, Cas1 and Cas2. Cas1–Cas2 adaptation may be functionally linked to interference (‘primed’) or not (‘naïve’), in each case by mechanisms unclear, reviewed in [3–5].

Cas proteins that catalyse interference show substantial diversity, with current classification into three major groups, types I, II and III [6–8], characterized by distinct effector complexes that manoeuvre crRNA to base pair with target DNA or RNA. Type I, II and IIIA systems target DNA catalysed by respectively, ‘Cascade’ (Cas complex for antiviral defence) [9,10], Cas9 [11,12] or CSM complex [13,14]. In contrast, CMR complexes target RNA in type IIIB CRISPR systems [15–17].

Cascades catalyse interference in type I CRISPR immunity. They are nucleoprotein assemblies of crRNA and Cas proteins (To help follow the variable nomenclature used in the literature for Cas proteins within Cascade complexes: CasA=Cse1/Cas8; CasC=Cas7; CasD=Cas5; CasE=Cas6e.). Diversity of amino acid sequence and gene synteny between Cascade components prompted categorization of type I CRISPR systems into sub-types A–F [6], with further refinement into types A–G to accommodate variants in types IA, ID and IG [18]. Common structural features are observed in Cascades from the different CRISPR sub-types in bacteria and archaea [9,10,19–22]. crRNA is delivered into Cascade as a single spacer element sequence after truncation of crRNA transcripts by Cas5 and Cas6 nucleases [10,19,20,23–28], reviewed in [29,30]. Sulfolobus solfataricus [19] and Thermoproteus tenax [21] Cascades (both type IA) and Escherichia coli (type IE) [10] use multiple copies of Cas7 protein to form a backbone filament with crRNA, functionally analogous to the Csy3-crRNA backbone of Pseudomonas aeruginosa Cascade (Type IF) [20]. Another variation of Cascade subunit type with common function is observed in type IC systems that use Csd2 to form Cas7-like crRNA filaments [31]. Interference is established by Cascade base pairing crRNA with protospacer (invader) DNA through ‘seeding’ [20,32–35] and further into an R-loop [9], reviewed in [36] that promotes nucleolytic degradation of invader DNA probably by interaction of Cascade with Cas3 helicase-nuclease. Recent atomic structures of E. coli Cascade complex have provided detailed insight into the arrangement of protein subunits relative to one another and to crRNA to provide a mechanism for interference in type IE CRISPR [37–39].

Cascade can identify invader DNA by interaction with PAM (protospacer adjacent motif) sequences. PAMs are short (2–5 nt) sequences located on invader DNA upstream of the protospacer that trigger Cascade–Cas3 interference [40]. A mechanism for Cascade–PAM recognition described in E. coli involves a CasA ‘large subunit’ contacting target DNA via its ‘L1 loop’ [41,42], as part of multiple interactions with Cas5 [37–39]. CasA is the ‘signature’ protein of type IE CRISPR systems, essential for Cascade function [6]. Atomic resolution structures of Cascade [37–39] show CasA interlocking with CasD, contributing to binding of the crRNA 5′-handle. CasA also contacts PAM as part of the tight association with CasD. In other type I CRISPR systems Cas8 is predicted to be functionally analogous to CasA [18], as a ‘signature’ protein for type IA, IB and IC systems, referred to respectively, as Cas8a2, Cas8b and Cas8c [6]. A more recent analysis highlighted diversity of Cas8 proteins leading to their renaming as Cas8, Cas8′ and Cas8″ proteins and creating new type-I CRISPR variants based on Cas8 protein sequences and positioning of cas8 genes relative to other cas genes [18]. An important role for Cas8 in interference was previously demonstrated in the euryarchaeon Haloferax volcanii (Hvo) [43]. Here we report genetic and biochemical analyses of Cas8 homologues from Hvo and Methanothermobacter thermautotrophicus (Mth).

H. volcanii strains H119 (ΔleuB, ΔpyrE2, ΔtrpA) [44] and Δcas8 [22] were grown aerobically at 45°C in Hv-YPC medium [45]. H. volcanii strains Δcas8 containing plasmids with mutated cas8 genes were cultivated in Hv-Ca medium supplemented with 0.25 mM tryptophan (Trp). E. coli strains DH5α (Invitrogen) and GM121 were grown aerobically at 37°C in 2YT medium [46].

For transformation of H. volcanii, plasmids were passaged through methylase deficient E. coli GM121 cells and introduced into H. volcanii by the PEG method [44]. Transformants were plated on selective media. Gene deletion in H. volcanii was performed as described previously [47]. Briefly, an integrative plasmid carrying flanking regions of the gene to be deleted and the pyrE2 gene as an auxotrophic marker, was incorporated into the genome by homologous recombination. Removal of this plasmid was forced by supplementing the media using 5-fluoroorotic acid (5-FOA, final concentration 50 μg/ml) that is converted to toxic 5-fluorouracil by orotate phosphoribosyltransferase encoded by pyrE2 gene. Positive clones were selected by colony PCR and gene deletion was subsequently confirmed by Southern blot hybridization. Using this method cas8 was deleted resulting in strain Δcas8.

Primers and plasmids are detailed in Supplementary Information. For generation of the integrative plasmid for cas8 deletion, a cas8 fragment with up- and downstream flanking regions (546 and 501 bp respectively) was amplified from genomic DNA by PCR using the oligonucleotides Csh1KOup, Csh1KOdo and Phusion DNA polymerase (Biozym). This fragment was ligated in pTA131 digested with EcoRV. With this plasmid, an inverse PCR was performed using the oligonucleotides IPCsh1KOup and IPCsh1KOdo, followed by ligation of the PCR product to obtain the integrative plasmid pTA131-cas8updo with cas8 flanking regions only [48].

For complementation of cas8, pTA927-N-FLAG-cas6 [22] was digested with HindIII and BamHI to remove cas6 and to subsequently insert the cas8 gene. The insert was generated by a PCR on genomic DNA with oligonucleotides 5-HindIII-cas8 and 3-cas8-NcoI-BamHI and subsequent digestion of the PCR product with HindIII and BamHI. Mutations were introduced into the cas8 gene using the QuikChange® II-Site Directed Mutagenesis Kit (Agilent Technologies). Both pTA927-N-FLAG-cas8 and pTA927-cas8-mutX were used for complementation of deletion strain Δcas8. For interference tests, the four PAM sequences that have been identified for H. volcanii TAA (PAM25), TAT (PAM26), TAG (PAM27) and CAC (PAM54) [43] were subcloned in pTA352 [49] together with sequence of spacer 1 from locus P1. Therefore, pTA409-PAM25, pTA409-PAM26, pTA409-PAM27 and pTA409-PAM54 as well as pTA352 (without any insert) were digested each with XhoI and BamHI according to previous studies [35] and the PAM-spacer fragments were ligated each with pTA352 to obtain pTA352-PAM25, pTA352-PAM26, pTA352-PAM27 and pTA352-PAM54.

From H. volcanii cultures grown to exponential growth phase, total RNA was isolated using TRIzol® Reagent (Life Technologies) and remaining DNA was digested with RQ1 RNase-Free DNase (Promega). Ten micrograms RNA was separated on 8% urea-polyacrylamide gels and transferred to a nylon membrane (Hybond-N+, GE Healthcare). For detection of crRNA, the oligonucleotide probes against spacer 1 from locus P1 and 5S RNA (control) were labelled radioactively using γ-³²P-ATP and T4 polynucleotide kinase (Thermo Scientific). Signals were detected with a radiosensitive photofilm (GE Healthcare).

A plasmid based invader assay [43] was performed to test functionality of Δcas8 x pTA927-N-FLAG-cas8 and Δcas8 x pTA927-cas8-mutX in the interference reaction. Plasmid invaders pTA352-PAM3 [43], pTA352-PAM9 [43], pTA352-PAM25 (present study), pTA352-PAM26 (present study), pTA352-PAM27 (present study) and pTA352-PAM54 (present study) were used and the vector pTA352 (without any insert) served as control. Transformants were plated on Hv-Min+Trp medium without leucine and uracil. Interference tests were performed at least three times to obtain statistically relevant data and activity in interference was defined for minimum 100-fold reduction in transformation rate.

H119 was transformed with pTA927-N-FLAG-cas7 and cells grown to a D₆₅₀ of 0.6 in medium containing 0.25 mM Trp to induce protein expression. To further induce protein expression additional Trp was added to a final concentration of 3 mM. The culture was incubated for further 3 h, cells were pelleted and washed once with salt-enriched PBS buffer [2.5 M NaCl, 150 mM MgCl₂, 1 × PBS (137 mM NaCl, 2.7 mM KCl, 8 mM Na₂HPO₄, 2 mM K₂HPO₄, pH 7.4)]. Cells were resuspended in lysis buffer [100 mM Tris/HCl, pH 7.5, 10 mM EDTA, 10 mM MgCl₂, 1 mM CaCl₂, 8 units/μl DNase RQ1 (Promega), 13 μl/ml protease inhibitor cocktail (Sigma)], incubated for 30 min at 4°C and subsequently lysed by sonication. Cell lysate clarified by ultracentrifugation (15 min at 100,000 g) and 0.03 volume of 5 M NaCl was added to the resulting supernatant. For subsequent FLAG-affinity purification, the supernatant was incubated over night at 4°C with anti-FLAG M2 affinity gel (Sigma) equilibrated with precooled washing buffer (0.2 M Tris/HCl, pH 7.4, 0.5 M NaCl). After washing, FLAG-tagged Cas7 was eluted by adding 3× FLAG peptide (150 ng/μl in washing buffer; Sigma). Proteins of the elution fraction were separated by SDS/PAGE (8% polyacrylamide) which was subsequently stained with Coomassie. The proteins were in-gel digested with trypsin as described in [50]. Peptides extracted from the in-gel digestion were analysed by LC–MS/MS on an Orbitrap XL instrument (Thermo Fischer Scientific) under standard conditions. The fragment spectra obtained for peptides were searched against H. volcanii database (www.halolex.mpg.de) [51] using MASCOT as a search engine. Peptides with the peptide score lower than 20 were considered unspecific.

DNA primer sequences for cloning and mutagenesis are listed in Supplementary Material. Cas8′ (ORF Mth 1090) was amplified by PCR from M. thermautotrophicus ΔH genomic DNA. The gene fragment cloned into pET14b facilitated expression of N-terminal hexahistidine-tagged Mth Cas8′ (His₆–Cas8′). Site-directed mutagenesis of Cas8′ was based on the Quick-change protocol, with mutations verified by DNA sequencing. His₆–Cas8′ protein was expressed in Escherichia coli strain BL21 Codon Plus, at 37°C with expression induced by IPTG (0.5 mM) at D₆₀₀ 0.6 for 2–4 h at 30°C. Harvested cells were resuspended in buffer B (20 mM Tris/HCl, pH 8.0, 500 mM NaCl, 5 mM imidazole) containing PMSF (0.1 mM) and freeze-thawed prior to lysis by sonication, followed by centrifugation at 39000 g for 20 min. His₆–Cas8′ was purified on an AKTA–FPLC with each step followed by SDS/PAGE. Soluble proteins were loaded on to a 5 ml of His Trap FF column charged with nickel chloride and equilibrated in buffer B. His₆–Cas8′ eluted into fractions within a gradient of 5—500 mM imidazole in buffer B were pooled and loaded on to a HI Load Superdex 200 26/60 column equilibrated in buffer C (20 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM DTT and 0.1mM PMSF) followed by elution in the same buffer in one column volume. His₆–Cas8′ fractions were pooled and loaded on to 5 ml of heparin HP column equilibrated in buffer C. His₆–Cas8′ eluted in a gradient of 150–1500 mM NaCl and fractions containing His₆–Cas8′ were pooled and dialysed into buffer D [20 mM Tris/HCl, pH 8.0, 500 mM KoAc, 1 mM DTT, 0.1 mM PMSF and 40% (w/v) glycerol] for storage in aliquots at −80°C. Mutant Cas8′ proteins were purified using the same methods.

Oligonucleotides were purchased from MWG and are listed in Supplementary Materials. Labelling of oligonucleotides and their annealing into substrates followed standard methods, summarized briefly: oligonucleotide (300 ng) was 5′-end-labelled with ³²P from γ³²P-ATP using T4 polynucleotide kinase (NEB). Labelled oligonucleotide was purified from unincorporated γ³²P-ATP in BioSpin6 columns (Bio-Rad) followed by annealing of appropriate oligonucleotide mixtures in sodium citrate buffer. Substrates were purified by electrophoresis through 10% polyacrylamide/1× Tris-Borate-EDTA (TBE) electrophoresis, for 3 h at 120 V and then excision of the appropriate band, detected on photographic film and elution of DNA by diffusion into Tris (10 mM)–NaCl (50 mM) buffer, pH 7.5, at 4°C.

EMSAs mixed protein(s) with substrate in buffer HB [100 mM Tris/HCl, pH 7.5, 10 mM DTT, 500 μg/ml BSA and 30% (v/v) glycerol], typically incubated at 44.8°C for 10 min. Reactions were then mixed by pipetting and loaded directly into wells of a gel comprising 7% polyacrylamide in 1× TBE buffer. Protein–nucleic acid complexes were separated by electrophoresis at 105 V for approximately 170 min in 1× TBE running buffer and detected by gel drying and phosphorimaging. Protein–nucleic acid complex formation was quantified compared with a no-protein control, using AIDA software to calculate the percentage of substrate bound and plotting in Prism to determine binding affinity expressed as K_D.

His₆–Cas8′ proteins were mixed with substrates (2 nM) in HB buffer supplemented with either 10 mM MgCl₂, 5 mM EDTA or nothing and incubated at 44.8°C for 10 min. Reactions were terminated by addition of 3 μl of stop solution [2.5% (w/v) SDS, 200 mM EDTA and 10 mg/ml proteinase K] and loaded into 10% TBE non-denaturing gels or 15% polyacrylamide/urea denaturing gels. Gels were dried, imaged and analysed as for EMSAs.

The gene encoding cas5 (ORF Mth1087) was amplified from M. thermautotrophicus (Mth) ΔH genomic DNA by PCR and the gene fragment cloned into pMal-C2x for expression of Mth Cas5 fused at its N-terminus to E. coli maltose-binding protein (MBP–Cas5). MBP tagging of Mth Cas5 greatly improved its solubility and stability for expression in E. coli. cas7 (ORF Mth1088) was amplified similarly to cas5, for cloning into pCDF-1b generating a non-tagged Cas7 protein. Co-expression of MBP-Cas5 and Cas7 in E. coli strain BL21 Codon Plus was in broth containing additional glucose (0.2% w/v), protein expression being induced by addition of IPTG (1 mM) at D₆₀₀ between 0.4–0.5. Cas5–Cas7 was purified as a complex through multiple steps on an AKTA–FPLC, followed using SDS/PAGE. Clarified soluble proteins were loaded into a column containing 5 ml of amylose sepharose resin and equilibrated in amylose column-binding buffer (ACBB; 20mM Tris/HCl, pH 8.0, 100 mM NaCl, 1 mm DTT and 0.1 mM PMSF). MBP–Cas5 and Cas7 co-eluted within a gradient of 0–5 mM maltose in ACBB and fractions containing MBP–Cas5–Cas7 were pooled and loaded on to 5 ml of Heparin HP column equilibrated in buffer C. MBP–Cas5–Cas7 co-eluted in a gradient of 150–1500 mM NaCl and fractions containing MBP–Cas5–Cas7 were pooled and dialysed into buffer D [20 mM Tris/HCl, pH 8.0, 500 mM KoAc, 1 mM DTT and 40% (w/v) glycerol] for storage in aliquots at −80°C.

MBP–Cas5–Cas7 was used to test for physical interaction with Cas8′. Fifty microlitres of amylose resin slurry was equilibrated in 100 μl of wash buffer (W; 20 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA and 1% Tween) and centrifuged at 700 g for 30 s, supernatant removed and washing repeated five times. Twenty micrograms of MBP–Cas5–Cas7, His₆–Cas8′ or MBP–Cas5 and–Cas7 and His₆–Cas8b′ we added to the resin to a final volume of 500 μl and end-to-end mixed for 2–4 h at 4°C. Resin was pelleted as before and washed three times as previously. SDS/PAGE disruption buffer was added to resin pellet and boiled. First wash and pellet analysed via SDS/PAGE. Two identically loaded SDS/PAGE gels were used for electroblotting on to PVDF and western blotting to detect the presence of MBP–Cas5 or His₆–Cas8′ proteins via their affinity tags. Membranes were incubated overnight at 4°C in western blocking buffer (WBB; 50 mM Tris/HCl, pH 7.6, 150 mM NaCl and 0.1% Tween, supplemented with 5% milk powder), before probing each separately with monoclonal antibodies against MBP (NEB) or His₆ (Sigma). Washed membranes were then probed with HRP-conjugated anti-mouse antibody (against His₆) or anti-goat antibody (against MBP) to develop using an ECL detection kit and imaged using FujiFilm LAS300 machine.

Plasmid protection assays in H. volcanii (Hvo) identified multiple PAM sequences and that disruption of cas8 abolished interference [43]. To investigate Cas8 further, Δcas8 Hvo cells were analysed in plasmid protection assays when expressing mutant or wild-type Cas8 from a second plasmid (Figure 1A). Single amino acid substitutions were introduced into Hvo Cas8 based on the alignment with Cas8 homologues from archaea and a bacterium (summarized in Figures 1B and 1C and a full alignment in Supplementary Figure S1). Cas8 proteins are diverse [18] with low overall sequence identity, but conserved amino acids were identified to investigate Cas8 function (Figure 1C). Mutation in Hvo Cas8 Asp²³⁰, Asn⁶²⁵ and Leu⁶²⁷ abolished interference, equivalent to cells lacking cas8 (Figure 1D). Mutated Asn²³² showed reduced interference, by ~20%, toward plasmids with PAM 5′-TTC, but had little effect on interference toward plasmid with PAM 5′-ACT (Figure 1D). Additional assays on Δcas8 Hvo pN232A-Cas8 using six Hvo PAMs highlighted a PAM bias, with interference reduced when PAM 5′-TTC or 5′-CAC was used, but with no effect of N232A on the other PAMs (Figure 1E). These genetic assays identified regions of Cas8 that are essential for interference and suggest that Cas8 is sensitive to PAM sequences.

As noted in Figure 1, homologues of Hvo Cas8 are present in other archaea and bacteria. Hvo is an extreme halophile, creating problems for analysis in native conditions of Hvo Cas8 binding to DNA/RNA substrates in EMSAs. Therefore, we purified the Cas8 homologue from M. thermautotrophicus (Mth) (Supplementary Figure S2), an organism amenable for analysis of its DNA-binding proteins. Mth Cas8 is called Cas8′ [18] because it is 119 amino acids shorter than Hvo Cas8, but it has the conserved amino acid residues required for CRISPR interference by Hvo Cas8, summarized in Figures 1C–1E. We constructed nucleic acid substrates for Cas8′ binding that centred on duplex DNA or R-loop shown in Figure 2(A). Substrates were either+or − PAM and contained a 5′-crRNA handle known to be important for interference in type IB CRISPR systems, whereas the 3′-handle is dispensable for interference in the same systems [52]. To predict Mth PAM, we analysed 123 spacers in Mth CRISPR-1, identifying protospacers from seven mobile genetic elements to deduce a PAM of 5′-CCN-3′, detailed in Supplementary Results and Supplementary Figure S3. CC dinucleotide PAM for Mth had been identified in a previous analysis [53], although reported as GG from the reverse complement of the Mth genome. Therefore we incorporated 5′-CCC into substrates for +PAM or 5′-AAA for −PAM.

Results of Cas8′ EMSAs are in Figures 2(B) and 2(C). Cas8′ bound to a +PAM R-loop or duplex DNA with highest affinity (Figure 2B respectively, K_d 5.3±0.7 nM and K_d 9.1 ±0.2 nM), compared with the same substrates −PAM (Figure 2B respectively, K_d 40.7±1.0 nM and K_d 18.3±0.8 nM). It was significant that Cas8′ bound to R-loops,+or −PAM, as distinct in-gel protein–DNA complexes, compared with in well aggregates of protein–DNA observed for Cas8′ mixed with duplex DNA (Figure 2C). These EMSAs suggested that Cas8′ in isolation can recognize PAM sequence and may have structural preference for binding stably to branched DNA or R-loops compared with duplex DNA.

Cas5 and Cas7 are integral to bacterial and archaeal Cascades [9,19] and are predicted to function with Cas8 during CRISPR interference. Previous studies identified a Cas5–Cas7 complex in Hvo [22] and physical association of Cas8′ with Cas7 during fractionation of Methanothermobacter cell biomass [54]. In the present work, FLAG-tagged Cas7 was expressed in Hvo cells to detect protein interactors, identifying Cas8 when FLAG–Cas7 affinity enriched cell extracts were analysed by MS (Figure 3A). To test for physical interaction of Cas8′ with Cas5–Cas7, we first co-purified Mth Cas7 with Cas5 (Supplementary Figure S4A), the latter an N-terminal fusion to E. coli MBP, giving soluble MBP–Cas5–Cas7 that bound crRNA1. MBP–Cas5–Cas7 interacted physically with Cas8′ (Figure 3B), with the same results observed either with or without pre-incubation of Cas5–Cas7 with crRNA1. Cas8′ bound to MBP–Cas5–Cas7 pre-incubated with amylose resin (Figure 3B, top panel, lane 8), but did not bind to control amylose resin in BSA (Figure 3B, top panel, lane 6). MBP–Cas5–Cas7 was detected, as expected, in Figure 3(B), lane 8, but not in lane 6 containing BSA. The reciprocal reaction (MBP–Cas5–Cas7 to Ni²⁺-NTA bound Cas8′) was not effective because MBP–Cas5–Cas7 bound Ni²⁺-NTA even when Cas8′ was absent. These assays indicated physical interaction of Cas8′ with Cas5–Cas7, although a maximum of only 10% of Cas8′ input could be detected as bound to MBP–Cas5–Cas7 in these conditions.

Cas8′ was tested for any influence effect on binding by Cas5–Cas7 to duplex ± PAM in EMSAs. Cas8′ stimulated total substrate binding when +PAM but had little effect on binding to −PAM (Figure 3C and representative gels in Supplementary Figure S4B). Significantly, EMSAs mixing Cas5–Cas7+Cas8′ showed a novel complex that was not present when either Cas5–Cas7 or Cas8 were alone (Figure 3D): Cas5–Cas7 at 10 nM or 100 nM formed in-well protein–DNA aggregates with, respectively, 8% and 67% of substrate (labelled A, Figure 3D, lanes 1 and 2). Cas8′ alone (5 nM) bound 55% of the substrate in distinct complexes (labelled B in Figure 3D, lane 3). A new complex (complex C) was defined when pre-mixing Cas8′ (5 nM) with Cas5–Cas7 (10 or 100 nM) and 90%–100% of substrate was bound (Figure 3D, lanes 4 and 5). Western blotting of identical EMSA using antibodies against MBP identified MBP–Cas5–Cas7 in complex C (Figure 3D, lanes 8 and 9), confirming that Cas5–Cas7 can form a distinct complex in EMSAs that is not an aggregate but dependent on Cas8′.

Mutant Mth Cas8′ proteins D151G, N153A and N536A were purified (Supplementary Figure S2), corresponding to mutations in Hvo Cas8 that had reduced or abolished interference activity in plasmid protection assays (Asp²³⁰, Asn²³² and Asn⁶²⁵ Table 1A). Cas8′ mutant proteins were proficient in binding to R-loop +PAM (Supplementary Figure S5) and other substrates (result not shown) and interacted with MBP–Cas5-Cas7 (Supplementary Figure S6). We investigated N153A substrate binding in more detail because the corresponding Hvo mutation (Asn²³²) caused reduced interference to PAMs 5′-CAC or -TTC (Table 1B). Duplex and R-loop substrates containing PAM 5′-CCC were not bound by N153A appreciably better than substrates −PAM in EMSAs (Figure 4A), contrasting with the binding behaviour of wild-type Cas8′ (Figure 2). It is possible that this observation might account for subtly reduced interference, discussed later.

To test for Cas8′ catalytic activity correlating to lack of interference from the aspartic acid/asparagine mutants we re-visited previous work that had identified Mth Cas8′ nuclease activity targeting ssDNA flaps [54]. We now compared this DNase activity to equivalent RNase activity. Cas8′ was much more efficient as a nuclease, measured as a function of time, on an ssRNA flap of the same sequence as ssDNA (Figure 4Bi). In these reactions, RNA was present in a RNA–DNA hybrid, but RNA nuclease activity was not detected in RNA–RNA duplex, even though Cas8′ binding to DNA–DNA, RNA–DNA and RNA–RNA substrates was similar (Supplementary Figure S7). Cas8′ nuclease activity was detected on RNA with 3′- and 5′-ends (Figure 4Aii), but only on the strand with ssRNA overhang. Cas8′ D151G and N536A had negligible RNase activity (Figure 4C, with further examples of nuclease assay gels in Supplementary Figure S8A), but binding was intact (Supplementary Figure S8B). N153A had intermediate nuclease activity (Figure 4C). The catalytic activity of Cas8′ and inactivating mutations in conserved residues that are required for interference in Hvo, is evidence that Cas8′ is an RNA nuclease in cells, discussed more below.

Cas8 proteins are candidates for guiding Cascade to invader DNA in some type I CRISPR systems [18,21]. Mutation of Cas8 had been implicated in loss of interference in Haloferax (Hvo) [43] and we investigated this using genetic analysis of Hvo Cas8 in CRISPR interference and nucleic acid binding and processing by Cas8′ from Methanothermobacter (Mth). We propose that Cas8/Cas8′ is part of Cascade, contributing to PAM and structure specific nucleic acid binding, influencing interaction of Cas5–Cas7 with nucleic acids. An interesting ssRNA nuclease activity of Cas8′ was detected in vitro, requiring conserved amino acids that are essential for interference in vivo.

Three lines of evidence indicated nucleic acid binding and PAM sensing by Cas8′, in isolation and when mixed with Cas5–Cas7. Firstly, isolated Cas8′ formed distinct complexes with R-loop substrates in EMSAs and predicted PAM 5′-CCN stimulated its binding to duplex and R-loop substrates (Figure 2). Second, the subtle PAM induced behaviour of Cas8 or Cas8′ mutated in respectively, Asn²³² (Tables 1A and 1B) or Asn¹⁵³ (Figure 4A) supports interaction with PAM either alone or when bound with Cas5–Cas7 in Cascade, an interaction that is perturbed by the asparagine mutation. Third, we observed enhanced substrate binding from Cas8′–Cas5–Cas7 +PAM, compared with either Cas5–Cas7 or Cas8′ alone, which was not observed when −PAM (Figure 3). Cas8′ in these assays converted Cas5–Cas7 protein aggregates into a distinct binding complex suggesting that Cas8′ modulates how Cas5–Cas7 can precisely assemble on the substrate, thereby controlling its aggregation. Based on E. coli Cascade structures detailing precise positioning for CasA relative to CasD (Cas5) [37–39], it is likely that interaction of Cas8 with Cas5–Cas7 is important for PAM sensing and for the choreography of Cascade binding to nucleic acids that leads to stable R-loop formation.

Mutation of Hvo Asp²³⁰ and Asn⁶²⁵ abolished Cas8 interference (Table 1A) and mutation of the corresponding residues in Mth Cas8′, Asp¹⁵¹ and Asn⁵³⁶, abolished ssRNA nuclease activity (Figure 4). Cas8′ degraded ssRNA with either 3′- or 5′-ends. However, we were unable to detect RNase activity in vitro from Hvo Cas8, despite using a wide range of high and low salt assay conditions, possibly because of instability of the purified protein that appeared during storage. Therefore we cannot conclude that abolished nuclease activity of Cas8′ D151G and N536A mutants explains the loss of genetic interference by Hvo Cas8 D230A and N625A mutants, but the correlation is interesting. Northern blotting for crRNA in Hvo Δcas8 cells showed that Cas8 is not needed for processing crRNA into pre-crRNA or crRNA in cells (Supplementary Figure S9). Also, Hvo cells do contain a nuclease that removes nts from the 3′-end of crRNA after processing from pre-crRNA but this RNase function is not altered if cells lack Cas8 (Anita Marchfelder, personal communication). 5′-crRNA handles are essential for interference in Hvo cells and are therefore not processed after crRNA formation [52]. Therefore the role of Cas8 RNase activity, if any, in Cascade-mediated CRISPR interference is undetermined. We cannot exclude the possibility that Cas8 RNase activity may be needed for some other aspect of RNA metabolism and processing in these organisms that has an indirectly important role for some type I CRISPR systems.

Simon Cass, Karina Hass and Britta Stoll did experiments. Omer S. Alkhnbashi and Rolf Backofen analysed and determined the CRISPR-Cas types of Haloferax volcanii and Methanothermobacter thermautotrophicus using bioinformatics, determining the Mth Cas8 as Cas8′ and the Hvo as Cas8. Edward Bolt and Anita Marchfelder organized the project and wrote the manuscript. Kundan Sharma and Henning Urlaubdid mass spectrometry analyses to identify the proteins co-purifying with the FLAG-tagged Cas7.

We thank Thorsten Allers (Nottingham, U.K.) for Haloferax plasmids.

This work was supported by the Biotechnology and Biological Sciences Research Council PhD studentship and the German Research Council (Deutsche Forschungsgemeinschaft) [grant numbers DFG MA1538/16-1, UR225/1-1 and BA2168/5-1].

ACBB

amylose column-binding buffer

Cas

CRISPR-associated

Cascade

CRISPR-associated complex for antiviral defence

CRISPR

clustered regularly interspaced short palindromic repeat

crRNA

CRISPR RNA

His₆–Cas8′

N-terminal hexahistidine-tagged Mth Cas8′

MBP

maltose-binding protein

PAM

protospacer adjacent motif

TBE

Tris-Borate-EDTA

Trp

tryptophan