CRISPR immunity relies on the consecutive binding and degradation of negatively supercoiled invader DNA by Cascade and Cas3 (original) (raw)

Mol Cell. Author manuscript; available in PMC 2013 Jun 8.

Published in final edited form as:

PMCID: PMC3372689

NIHMSID: NIHMS366982

Edze R. Westra,a Paul B. G. van Erp,a Tim Künne,a Shi Pey Wong,a Raymond H. J. Staals,a Christel L. C. Seegers,a Sander Bollen,b Matthijs M. Jore,a Ekaterina Semenova,c Konstantin Severinov,c,d,e Willem M. de Vos,a,f Remus T. Dame,b Renko de Vries,g Stan J. J. Brouns,a,* and John van der Oosta

Edze R. Westra

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Paul B. G. van Erp

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Tim Künne

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Shi Pey Wong

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Raymond H. J. Staals

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Christel L. C. Seegers

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Sander Bollen

bLeiden Institute of Chemistry, Gorlaeus Laboratories, Laboratory of Molecular Genetics and Cell Observatory, Leiden University, P.O. Box 9502, 2300 RA Leiden, The Netherlands

Matthijs M. Jore

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

Ekaterina Semenova

cWaksman Institute, Piscataway, NJ 08854, USA

Konstantin Severinov

cWaksman Institute, Piscataway, NJ 08854, USA

dDepartment of Molecular Biology and Biochemistry, Rutgers, The State University, Piscataway, NJ 08854, USA

eInstitutes of Molecular Genetics and Gene Biology, Russian Academy of Sciences, Moscow 119991, Russia

Willem M. de Vos

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

fDepartment of Bacteriology & Immunology, and Department of Veterinary Biosciences, PO Box 66, University of Helsinki, FIN-00014 Helsinki, Finland

Remus T. Dame

bLeiden Institute of Chemistry, Gorlaeus Laboratories, Laboratory of Molecular Genetics and Cell Observatory, Leiden University, P.O. Box 9502, 2300 RA Leiden, The Netherlands

Renko de Vries

gLaboratory of Physical Chemistry and Colloid Science, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 6, 6703 HB Wageningen, The Netherlands

Stan J. J. Brouns

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

John van der Oost

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

aLaboratory of Microbiology, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 10, 6703 HB Wageningen, The Netherlands

bLeiden Institute of Chemistry, Gorlaeus Laboratories, Laboratory of Molecular Genetics and Cell Observatory, Leiden University, P.O. Box 9502, 2300 RA Leiden, The Netherlands

cWaksman Institute, Piscataway, NJ 08854, USA

dDepartment of Molecular Biology and Biochemistry, Rutgers, The State University, Piscataway, NJ 08854, USA

eInstitutes of Molecular Genetics and Gene Biology, Russian Academy of Sciences, Moscow 119991, Russia

fDepartment of Bacteriology & Immunology, and Department of Veterinary Biosciences, PO Box 66, University of Helsinki, FIN-00014 Helsinki, Finland

gLaboratory of Physical Chemistry and Colloid Science, Department of Agrotechnology and Food Sciences, Wageningen University, Dreijenplein 6, 6703 HB Wageningen, The Netherlands

*To whom correspondence may be addressed. ln.ruw@snuorb.nats Tel. +31-317-483740 Fax. +31-317-483829

Summary

The prokaryotic CRISPR/Cas immune system is based on genomic loci that contain incorporated sequence tags from viruses and plasmids. Using small guide RNA molecules, these sequences act as a memory to reject returning invaders. Both the Cascade ribonucleoprotein complex and the Cas3 nuclease/helicase are required for CRISPR-interference in Escherichia coli, but it is unknown how natural target DNA molecules are recognized and neutralized by their combined action. Here we show that Cascade efficiently locates target sequences in negatively supercoiled DNA, but only if these are flanked by a Protospacer Adjacent Motif (PAM). PAM recognition by Cascade exclusively involves the crRNA-complementary DNA strand. After Cascade-mediated R-loop formation, the Cse1 subunit recruits Cas3, which catalyzes nicking of target DNA through its HD-nuclease domain. The target is then progressively unwound and cleaved by the joint ATP-dependent helicase activity and Mg2+-dependent HD-nuclease activity of Cas3, leading to complete target DNA degradation and invader neutralization.

Introduction

Bacteria and archaea have a variety of defense systems against invasive DNA elements (reviewed by (Labrie et al., 2010)). CRISPR/Cas defense systems provide adaptive immunity by integrating plasmid and viral DNA fragments in loci of clustered regularly interspaced short palindromic repeats (CRISPR) on the host chromosome (Barrangou et al., 2007). The viral- and plasmid-derived sequences, known as spacers, are separated by short host-derived repeat sequences. A re-evaluation of the diversity of CRISPR/Cas systems has resulted in a classification into three distinct types that vary in CRISPR-associated (cas) gene content, and display major differences throughout the CRISPR defense pathway (Makarova et al., 2011). RNA transcripts of CRISPR loci (pre-crRNA) are cleaved specifically in the repeat sequences by Cas endoribonucleases in type I and type III systems (Brouns et al., 2008; Carte et al., 2008; Haurwitz et al., 2010; Przybilski et al., 2011) or by RNase III in type II systems (Deltcheva et al., 2011). The generated crRNAs are bound by a Cas protein complex and utilized as guides to detect complementary DNA or RNA sequences (Brouns et al., 2008; Jore et al., 2011; Lintner et al., 2011; Wiedenheft et al., 2011b; Hale et al., 2009). Cleavage of target nucleic acids has been demonstrated in vitro and in vivo for the Pyrococcus furiosus type III-B system (Hale et al., 2012; Hale et al., 2009), in vitro for the Sulfolobus solfataricus type III-B system (Zhang et al., 2012) and in vivo for the Streptococcus thermophilus type II system (Garneau et al., 2010). For type I systems the mechanism of invader rejection is anticipated to involve DNA cleavage, but, in contrast to type II and type III systems, direct evidence for sequence specific recognition and degradation of invader nucleic acids is still lacking.

Escherichia coli strain K12 encodes a CRISPR/Cas type I-E system containing eight cas genes (cas1, cas2, cas3 and cse1, cse2, cas7, cas5, cas6e) and a downstream CRISPR locus with type-2 repeats (Kunin et al., 2007). Five proteins, Cse1, Cse2, Cas7, Cas5 and Cas6e (previously referred to as CasA, CasB, CasC, CasD and CasE, respectively) form a ribonucleoprotein complex named Cascade (CRISPR associated complex for anti-viral defense) (Brouns et al., 2008). After cleavage of the pre-crRNA by Cas6e, the mature 61 nt crRNA is retained by the complex (Brouns et al., 2008; Jore et al., 2011). The crRNA guides sequence specific binding of Cascade to double stranded (ds) DNA molecules through base pairing between the crRNA spacer and the complementary protospacer (Jore et al., 2011), forming an R-loop. Although target DNA recognition by Cascade involves strand separation, this process is ATP-independent (Jore et al., 2011).

Phage resistance in type I-E systems not only requires Cascade, but also Cas3 (Brouns et al., 2008). Cas3 has an N-terminal HD-nuclease domain and a C-terminal Superfamily 2 (SF2; DExD/H) helicase domain (Makarova et al., 2006). Recently, structural and biochemical analyses of the type I-E Cas3 HD-domain from Thermus thermophilus HB8 (_Tth_Cas3HDdom) have revealed manganese or nickel-dependent endonuclease activity on ssDNA (Mulepati and Bailey, 2011). The type I-A Methanocaldococcus jannaschii Cas3 HD-domain (_Mja_Cas3″) has magnesium-dependent endo- and 3′-5′ exonuclease activity on ssDNA and ssRNA, which is stimulated by the M. jannaschii Cas3 helicase domain (_Mja_Cas3′) in the presence of ATP (Beloglazova et al., 2011). Likewise, the type I-E HD-domain of S. thermophilus Cas3 (_Sth_Cas3) displays magnesium-dependent endonuclease activity on ssDNA, while its DExD/H-box helicase domain has ATP- and magnesium-dependent DNA/DNA and DNA/RNA unwinding activity in the 3′ to 5′ direction (Sinkunas et al., 2011). In addition, Cas3 from E. coli was recently reported to unwind DNA/RNA heteroduplexes (Howard et al., 2011). Based on these biochemical activities it has been speculated that the Cas3 HD-domain may play a role in target DNA cleavage (Jore et al., 2011; Sinkunas et al., 2011), and that the helicase domain may be involved in either protospacer scanning or Cascade recycling (Sinkunas et al., 2011). However, the mechanism of CRISPR-interference through the combined action of Cascade and Cas3 has yet to be demonstrated.

Here we show that Cas3-independent target DNA recognition by Cascade marks DNA for cleavage and ATP-dependent degradation by Cas3. DNA binding by the crRNA-guided Cascade complex is constrained by strict topological requirements of the target DNA. Furthermore, efficient target binding requires that the target sequence is flanked by a Protospacer Adjacent Motif (PAM), with PAM recognition taking place exclusively in the targeted strand of the DNA. Upon binding a protospacer sequence, Cascade recruits Cas3, allowing Cas3 to degrade the target DNA.

Results

Cascade exclusively binds negatively supercoiled target DNA

Recently, the Cascade complex of E. coli K12 was reported to bind unnatural, short dsDNA molecules of 65 or 86 bp sequence-specifically by R-loop formation (Jore et al., 2011; Semenova et al., 2011). It has remained unclear, however, whether Cascade is sufficient to recognize natural dsDNA targets with relevant topologies, or if it requires Cas3 helicase activity to accomplish this.

The 3 kb pUC19-derived plasmid used in this study, denoted pUC-λ, contains a 350 bp DNA fragment corresponding to part of the J gene of phage λ, which is targeted by J3-Cascade (Cascade loaded with crRNA containing spacer J3 (Westra et al., 2010)). Using electrophoretic mobility shift assays we could show that Cascade has high affinity for a nSC target plasmid. At a molar ratio of pUC-λ to J3-Cascade of 1:8 all nSC plasmid containing a CAT PAM was bound by Cascade (Fig. 1A), while a plasmid carrying an escape mutation in the PAM (CGT) (Semenova et al., 2011) was fully bound only at a molar ratio of 1:64 (Fig. 1B). Cascade carrying the non-targeting crRNA R44 (R44-Cascade) displayed nonspecific binding at a molar ratio of 1:128 (Fig. 1C). The dissociation constant (Kd) of nSC pUC-λ was determined to be 13 ± 1.4 nM for J3-Cascade (Fig. S1A), while the Kd of nSC pUC-λ with the mutated CGT PAM was 55 ± 12 nM (Fig. S1B). The Kd of nonspecific pUC-λ binding was determined to be 429 ± 152 nM for R44-Cascade (Fig. S1C). Interestingly, J3-Cascade was unable to bind relaxed target DNA with measurable affinity, such as nicked open circular (OC) (Fig. 1D) or linear pUC-λ (Fig. 1E), showing that Cascade only has high affinity for natural dsDNA substrates that have a nSC topology.

An external file that holds a picture, illustration, etc. Object name is nihms366982f1.jpg

Cascade only binds nSC plasmid DNA with high affinity. A) Gel-shift of nSC plasmid DNA with J3-Cascade, containing a targeting (J3) crRNA. pUC-λ was mixed with 2-fold increasing amounts of J3-Cascade, from a pUC-λ : Cascade molar ratio of 1 : 0.5 up to a 1 : 256 molar ratio. The first and last lane contain only pUC-λ. B) Gel-shift as in (A) with an escape mutant of pUC-λ containing a single point mutation in the PAM (CAT to CGT). C) Gel-shift as in (A) with R44-Cascade containing a non-targeting (R44) crRNA. D) Gel-shift as in (A) with Nt.BspQI nicked pUC-λE) Gel-shift as in (A) with PdmI linearized pUC-λF) Specific binding of Cascade to the protospacer monitored by BsmI footprinting at a pUC-λ : Cascade molar ratio of 10:1. Lane 1 and 5 contain only pUC-λ. Lane 2 and 6 contain pUC-λ mixed with Cascade. Lane 3 and 7 contain pUC-λ mixed with Cascade and subsequent BsmI addition. Lane 4 and 8 contain pUC-λ mixed with BsmI. G) BsmI footprint as in (F) with Nt.BspQI cleavage of one strand of the plasmid subsequent to Cascade binding. Lane 1 and 6 contain only pUC-λ. Lane 2 and 7 contain pUC-λ mixed with Cascade. Lane 3 and 8 contain pUC-λ mixed with Cascade and a subsequent BsmI footprint. Lane 4 and 9 contain pUC-λ mixed with Cascade, followed by nicking with Nt.BspQI and a BsmI footprint. Lane 5 and 10 contain pUC-λ nicked with Nt.BspQI. H) BsmI footprint as in (F) with EcoRI cleavage of both strands of the plasmid subsequent to Cascade binding. Lane 1 and 6 contain only pUC-λ. Lane 2 and 7 contain pUC-λ mixed with Cascade. Lane 3 and 8 contain pUC-λ mixed with Cascade and a subsequent BsmI footprint. Lane 4 and 9 contain pUC-λ mixed with Cascade, followed by cleavage with EcoRI and a BsmI footprint (combined cleavage of BsmI and EcoRI produces a 2.8 kb fragment and a ~200 nt fragment; the latter is not visible on this gel). Lane 5 and 10 contain pUC-λ cleaved with EcoRI.

To distinguish nonspecific binding from specific binding, we made use of the BsmI restriction site that is located within the protospacer to perform a BsmI footprint analysis. While R44-Cascade does not protect the BsmI site from cleavage (Fig. 1F, lane 3), pUC-λ is protected from BsmI cleavage in the presence of J3-Cascade (Fig. 1F, lane 7). This shows that Cascade is able to locate and bind a protospacer sequence in a physiologically relevant target DNA molecule in the absence of Cas3.

Based on the observed helicase activity of Cas3 on RNA/DNA heteroduplexes, Cas3 has been hypothesized to play a role in recycling Cascade, presumably after ssDNA cleavage by its HD-domain (Sinkunas et al., 2011). Since Cascade alone cannot bind a nicked plasmid with measurable affinity, we investigated whether Cascade bound to nSC plasmid remains associated after relaxation of the DNA. To this end, Cascade bound nSC pUC-λ was nicked with Nt.BspQI, followed by a BsmI footprint to monitor Cascade dissociation. Nt.BspQI cleaves pUC-λ ~400 bp downstream of the protospacer, generating pUC-λ with an OC topology. Interestingly, Cascade remains plasmid bound after single strand nicking, as can be seen from the protection of the BsmI site (Fig. 1G, compare lanes 4 and 9). However, in a gel-shift assay, Cascade dissociates from the DNA, indicating that the Cascade-DNA interaction is weakened by the change in topology (Fig. S1D, compare lanes 8 and 10). Similar observations are made when both DNA strands of pUC-λ are cleaved after Cascade binding, both in a BsmI footprint (Fig. 1H, compare lanes 4 and 9) and in a gel-shift assay (Fig. S1E, compare lanes 8 and 10). This shows that, although the Cascade-DNA interaction is weakened upon relaxation of the DNA, Cascade remains bound to the protospacer sequence after introducing a single or double strand break in the nSC target DNA.

Cascade induces bending of bound target DNA

In order to corroborate the binding of Cascade to nSC plasmids, and to investigate the topological consequences of Cascade induced R-loop formation, we employed scanning force microscopy. Using this approach we visualized complexes formed between purified Cascade and pUC-λ. Specific complexes containing a single bound J3-Cascade complex were formed, while unspecific R44-Cascade yielded no DNA-bound complexes in this assay under identical conditions (data not shown). Out of 81 DNA molecules observed, 66 (81%) were found to have J3-Cascade bound (Fig. 2A–P). In most cases (86%) Cascade was found at the apex of a loop, whereas in only a small fraction (14%) Cascade was found at non-apical positions. These data show that Cascade binding causes bending and possibly wrapping of the DNA, probably to facilitate local melting of the DNA duplex.

An external file that holds a picture, illustration, etc. Object name is nihms366982f2.jpg

Cascade induces bending of target DNA upon protospacer binding. A-P) Scanning force microscopy images of nSC plasmid DNA with J3-Cascade containing a targeting (J3) crRNA. pUC-λ was mixed with J3-Cascade at a pUC-λ : Cascade ratio of 1 : 7. Each image shows a 500 × 500 nm surface area. White dots correspond to Cascade.

Negative supercoils provide energy for Cascade-mediated R-loop formation

The preferential binding of plasmids with a nSC topology as compared to OC and linear DNA prompted us to investigate whether this is related to differences in the Gibbs free energy of strand separation associated with these topologies. At 37 °C, the Gibbs free energy for separating strands over the length of a spacer sequence (32 bp) in a long dsDNA molecule is calculated to be approximately 190 kJ/mol (Supplementary Material S1). For DNA binding and strand separation to occur spontaneously, the total change in the Gibbs free energy should be negative. A major part of the approximately 190 kJ/mol will be balanced by the (partial) re-establishement of base pairs between the protospacer and the crRNA. An important contribution in balancing the remainder is the Gibbs free energy stored in the negative supercoils. It can be calculated (Supplementary Material S2 and S3) that the change in the Gibbs free energy of supercoiling associated with unwinding plasmid dsDNA over 32 base pairs contributes approximately −90 kJ/mol, nearly half of the total Gibbs energy required.

Previous Cascade-DNA binding studies were performed with linear dsDNA fragments of 65 and 86 bp (Semenova et al., 2011; Jore et al., 2011), raising the question why short DNA molecules are bound by Cascade while a linear 3 kb DNA molecule is not bound. This can be explained by the lower melting temperature of short DNA as compared to long DNA, which implies that separating strands in shorter DNA fragments requires less energy input. The estimated decrease in the Gibbs free energy of strand separation over 32 bp in 86 bp dsDNA, as compared to long (e.g., 3 kb) dsDNA, is −20 kJ/mol (Supplementary Material S4). Since this effect is much smaller than the described supercoiling effect, other factors may also play a role in causing the observed differences between Cascade binding and strand separation in these targets of different length.

Cascade tolerates four distinct PAMs that are recognized in the base pairing strand

Previously, it has been demonstrated that the presence of a PAM flanking the protospacer is important for Cascade target DNA binding (Semenova et al., 2011). To monitor the plasticity of PAM recognition, we generated mutants of a PAM flanking a previously described phage M13 protospacer (Semenova et al., 2011), and tested CRISPR-interference against these PAM mutants in a phage M13 infection experiment. Although most PAM mutants escape interference, this analysis reveals that Cascade tolerates at least four different PAMs (Fig. 3A). Cascade displays high affinity binding to dsDNA containing the M13 protospacer flanked by either one of these four tolerated PAMs (Fig. 3BCDE). In contrast, dsDNA containing the M13 protospacer flanked by an escape PAM is bound with decreased affinity (Fig. 3F), in good agreement with previous DNA binding studies (Semenova et al., 2011). The distinct binding behavior of Cascade to targets containing a PAM versus those lacking a PAM allowed us to investigate whether PAM recognition takes place in one or in both strands of the target DNA. To this end, we performed gel-shift assays with dsDNA targets containing hybrid PAMs, in which one strand contains a tolerated PAM and the other strand contains an escape PAM (Fig. 3I and Fig. 3J). Comparing Cascade binding to these targets containing hybrid PAMs with Cascade binding to targets either containing or lacking a PAM on both strands demonstrates that PAM recognition takes place exclusively in the crRNA-base pairing strand. Cascade binding to a dsDNA containing the PAM in the target strand (at the 3′ end of the strand that base pairs with the crRNA) (Fig. 3J), is comparable to Cascade binding to a dsDNA substrate containing the PAM in both strands (Fig. 3H). On the other hand, Cascade binding to a dsDNA containing the PAM in the non-target strand (at the 5′ end of the displaced strand) (Fig. 3J), is reminiscent of Cascade binding to a dsDNA substrate lacking a PAM in both strands (Fig. 3G). This demonstrates that PAM recognition exclusively takes place in the target strand.

An external file that holds a picture, illustration, etc. Object name is nihms366982f3.jpg

Cascade tolerates 4 distinct PAMs that are recognized on the target strand. A) Mutagenesis of the PAM flanking the previously described M13 protospacer on the phage M13 genome gives rise to either mutants that escape CRISPR-interference (sequences shown in red) or to mutants that are still subject to CRISPR-interference (sequences shown in black). B–F) Gel-shift assays to monitor M13-Cascade binding to 65 nt dsDNA probes containing the M13 protospacer flanked by the PAM sequences indicated by an asterisk in (A), corresponding to four PAMs (B–E) that are tolerated and a single escape PAM mutant (F). G) Gel-shift assays using 65 nt dsDNA probes containing the M13 protospacer flanked by an escape PAM sequence (GGG/CCC) on both the target and the displaced strand. H) Gel-shift as in (G) with the M13 protospacer being flanked by a tolerated PAM sequence (CTT/AAG) on both the target and the displaced strand. I) Gel-shift assays as in (G) with the M13 protospacer being flanked by an escape PAM sequence (GGG) on the target strand and a tolerated PAM (AAG) on the displaced strand. J) Gel-shift assays as in (G) with the M13 protospacer being flanked by a tolerated PAM sequence (CTT) on the target strand and an escape PAM (CCC) on the displaced strand.

Cascade interacts with Cas3 upon protospacer recognition

Although Cascade alone is sufficient for binding to nSC DNA targets, CRISPR-interference also requires Cas3 (Brouns et al., 2008). Sequence analysis of cas3 genes from organisms containing the type I-E CRISPR/Cas system reveals that Cas3 and Cse1 occur as fusion proteins in Streptomyces sp. SPB78 (Accession Number: ZP_07272643.1), in Streptomyces griseus (Accession Number YP_001825054), and in Catenulispora acidiphila DSM 44928 (Accession Number YP_003114638). The existence of these fusion proteins suggests that stand-alone Cas3 also directly interacts with Cascade in vivo, and that the Cse1 subunit may provide a docking site for such an association.

To investigate this, we designed bimolecular fluorescence complementation (BiFC) experiments to monitor interactions between Cas3 and Cascade in vivo before and after phage λ infection. BiFC experiments rely on the capacity of the non-fluorescent halves of a fluorescent protein (e.g., YFP) to refold and to form a functional fluorescent molecule when the two halves occur in close proximity (reviewed in (Kerppola, 2006)). As such, it provides a tool to reveal protein-protein interactions, since the efficiency of refolding is greatly enhanced if the local concentrations are high, e.g., when the two halves of the fluorescent protein are fused to interaction partners. For our experiments we fused Cse1 C-terminally with the N-terminal 155 amino acids of Venus (Cse1-N155Venus), an improved version of YFP (Nagai et al., 2002). Cas3 was C-terminally fused to the C-terminal 85 amino acids of Venus (Cas3-C85Venus).

The BiFC analysis reveals that Cascade does not interact with Cas3 in the absence of invading DNA (Fig. 4ABC, Fig. 4J and Fig. S2). Upon infection with phage λ, however, cells expressing CascadeΔCse1, Cse1-N155Venus and Cas3-C85Venus are fluorescent if they co-express the anti-λ CRISPR 7Tm (Fig. 4DEF, Fig. 4J and Fig. S2). When co-expressing a non-targeting CRISPR R44 (Fig. 4GHI, Fig. 4J and Fig. S2), the cells remain non-fluorescent. This shows that Cascade and Cas3 specifically interact during infection upon protospacer recognition and that Cse1 and Cas3 are in close proximity of each other in the Cascade-Cas3 binary effector complex.

An external file that holds a picture, illustration, etc. Object name is nihms366982f4.jpg

BiFC analysis reveals that Cascade and Cas3 interact upon target recognition. A) Venus fluorescence of cells expressing CascadeΔCse1 and CRISPR 7Tm, which targets 7 protospacers on the phageλ genome, and Cse1-N155Venus and Cas3-C85Venus fusion proteins. B) Brightfield image of the cells in (A). C) Overlay of (A) and (B). D) Venus fluorescence of phage λ infected cells expressing CascadeΔCse1 and CRISPR 7Tm, and Cse1-N155Venus and Cas3-C85Venus fusion proteins. E) Brightfield image of the cells in (D). F) Overlay of (D) and (E). G) Venus fluorescence of phage λ infected cells expressing CascadeΔCse1 and non-targeting CRISPR R44, and N155Venus and C85Venus proteins. H) Brightfield image of the cells in (G). I) Overlay of (G) and (H). J) Average of the fluorescence intensity of 4–7 individual cells of each strain, as determined using the profile tool of LSM viewer (Carl Zeiss). Error bars represent the standard deviation of the mean.

Nuclease and helicase activities of Cas3 are essential for CRISPR-interference

To address the fate of the invading DNA upon formation of the Cascade-Cas3 effector complex, and to examine the role of the Cas3 nuclease and helicase activities during CRISPR-interference, we made three HD-domain and three helicase domain mutants. Some of these mutants have been shown to abolish _Mja_Cas3′ and T Cas3HDdom th nuclease and _Sth_Cas3 nuclease and helicase activity in vitro (Beloglazova et al., 2011; Mulepati and Bailey, 2011; Sinkunas et al., 2011). The HD-domain mutants that were tested include two conserved residues that are involved in metal ion coordination, Cas3 H74A and Cas3 D75A (corresponding to _Tth_Cas3HDdom H69A and D70A, _Mja_Cas3′ H66A and D67A), and a conserved residue Cas3 K78A (_Tth_Cas3HDdom K73A, M Cas3′ K70A) that may be involved ja in substrate positioning (Mulepati and Bailey, 2011). The helicase domain mutants that were generated are motif I mutant Cas3 K320N (corresponding to the _Sth_Cas3 K316A mutant (Sinkunas et al., 2011)), motif II mutant Cas3 D452N (_Sth_Cas3 D452A (Sinkunas et al., 2011)) and motif III double mutant Cas3 S483A/T485A. While motifs I and II are thought to be involved in ATP binding and hydrolysis, motif III is thought to be involved in coupling ATP hydrolysis to unwinding activity (Tuteja and Tuteja, 2004).

To test whether these mutants are still functional in vivo, resistance against pUC-λ transformation was determined in E. coli BL21-AI strains expressing J3-Cascade and a Cas3 mutant. Upon transformation with a target plasmid, cells expressing J3-Cascade and any one of the three HD-domain mutants of Cas3 have transformation efficiencies that are least 104-fold higher than cells expressing wt Cas3; the obtained efficiencies of the mutants are only slightly below the value observed for cells expressing the non-targeting R44-Cascade and wt Cas3 (Fig. 5A). The helicase domain mutants were also severely compromised in exerting their function, with the exception of the Cas3 S483A/T485A double mutant (Fig. 5A). The finding that Cas3 nuclease and helicase mutants are impaired in CRISPR-interference strongly suggests that both activities are essential for effective invader neutralization.

An external file that holds a picture, illustration, etc. Object name is nihms366982f5.jpg

The role of Cas3 nuclease and helicase activities during CRISPR-interference. A) Competent BL21-AI cells expressing Cascade, a Cas3 mutant and CRISPR J3 were transformed with pUC-λ Colony forming units per microgram pUC-λ (cfu/μg DNA) are depicted for each of the strains expressing a Cas3 mutant. Cells expressing wt Cas3 and CRISPR J3 or CRISPR R44 serve as positive and negative controls, respectively. Experiments were performed in triplicate. Error bars represent the standard deviation of the mean. B) BL21-AI cells carrying Cascade, Cas3 mutant, and CRISPR encoding plasmids as well as pUC-λ are grown under conditions that suppress expression of the cas genes and CRISPR. At t=0 expression of the CRISPR and cas genes is induced. The fraction of cells that retain pUC-λ over time is shown, as determined by the ratio of ampicillin resistant and total cell counts.

Additionally, we analyzed curing of pUC-λ from E. coli BL21-AI strains upon induction of expression of the cas genes and CRISPR. Cells were examined over time for loss of pUC-λ. Five hours after induction, 90% of cells expressing wt Cas3 or the Cas3 S483A/T485A double mutant lost pUC-λ as measured by selective plating (Fig. 5B). Sequencing of pUC-λ from cells that were unable to cure this plasmid (carrying either wt Cas3 or the Cas3 S483A/T485A double mutant) revealed four escape mutants containing either mutations in the seed region or (partial) deletions of the protospacer (Fig. S3) in agreement with previously described escape mutants (Semenova et al., 2011). Cells expressing any of the other Cas3 mutants or cells expressing a non-targeting CRISPR did retain the original pUC-λ plasmid (Fig. 5B). This illustrates that although some mutants still show partial resistance against transformation of a single DNA molecule into the cell (Fig. 5A), they are deficient in curing a high copy number plasmid (Fig. 5B).

The HD-domain of Cas3 in the Cascade-Cas3 effector complex cleaves target DNA

An in vitro characterization of the catalytic features of Cas3 requires its production and purification in an active state. Despite various solubilization strategies, Cas3 overproduced in E. coli BL21 (DE3) was found to be mainly present in inactive aggregates and inclusion bodies (data not shown, and (Howard et al., 2011)). This issue was resolved by producing Cas3 as a Cas3-Cse1 fusion protein, containing a linker identical to that of the aforementioned Cas3-Cse1 fusion protein of S. griseus (Fig. S4). When co-expressed with CascadeΔCse1 and CRISPR J3, a soluble fusion-complex was produced that could be purified to apparent homogeneity with the same apparent stoichiometry as Cascade (Fig. 6A). To address the functionality of this complex we first tested whether it provides resistance against phage λ infection. The efficiency of plaquing (eop) of phage λ on host cells expressing the fusion-complex J3-Cascade-Cas3 was identical to the eop on cells expressing the separate proteins (Fig. 6B).

An external file that holds a picture, illustration, etc. Object name is nihms366982f6.jpg

Cascade-Cas3 fusion complex provides in vivo resistance and has in vitro nuclease activity. A) Coomassie Blue-stained SDS-PAGE of purified Cascade and Cascade-Cas3 fusion complex. B) Efficiency of plaquing of phage λ on cells expressing Cascade-Cas3 fusion complex and a targeting (J3) or non-targeting (R44) CRISPR. Cells expressing non-fused Cascade and Cas3 with a targeting (J3) CRISPR is given as a control. C) Gel-shift (in the absence of divalent metal ions) of nSC target plasmid with J3-Cascade-Cas3 fusion complex. pUC-λ was mixed with 2-fold increasing amounts of J3-Cascade-Cas3, from a pUC-λ : J3-Cascade-Cas3 molar ratio of 1 : 0.5 up to 1 : 128. The first and last lanes contain only pUC-λ. D) Gel-shift (in the absence of divalent metal ions) of nSC non-target plasmid with J3-Cascade-Cas3 fusion complex. pUC-P7 was mixed with 2-fold increasing amounts of J3-Cascade-Cas3, from a pUC-P7 : J3-Cascade-Cas3 molar ratio of 1 : 0.5 up to 1 : 128. The first and last lanes contain only pUC-P7. E) Incubation of nSC target plasmid (pUC-λ, left) or nSC non-target plasmid (pUC-P7, right) with J3-Cascade-Cas3 in the presence of 10 mM MgCl2. Lane 1 and 7 contain only plasmid. F) Assay as in (E) in the presence of 2 mM ATP. G) Assay as in (E) with the mutant J3-Cascade-Cas3K320N complex. H) Assay as in (G) in the presence of 2 mM ATP. (I) Schematic overview of the three DNA substrates used in the in vitro nuclease assay shown in panel (J). The substrates are 89 nt and contain a 39 nt double stranded region at a variable position. Asterisks at the 5′ end indicate the presence of the 32P label. (J) Incubation of Cascade-Cas3 with the 5′ labelled substrates shown in (I) in the presence of 10 mM MgCl2. The endonuclease products that run low in the gel are better visible in the overexposed version shown in Figure S5.

Since the J3-Cascade-Cas3 fusion-complex was functional in vivo, we proceeded with in vitro DNA cleavage assays using this complex. When J3-Cascade-Cas3 was incubated with pUC-λ in the absence of divalent metals, plasmid binding was observed at molar ratios similar to those observed for Cascade (Fig. 6C), while nonspecific binding to a non-target plasmid (pUC-P7, a pUC19-derived plasmid of the same size as pUC-λ, but lacking the J3 protospacer) occurred only at high molar ratios (Fig. 6D), indicating that the nonspecific DNA binding properties of the complex are also similar to that of Cascade alone.

Interestingly, the J3-Cascade-Cas3 complex displays magnesium-dependent endonuclease activity on nSC target plasmids. In the presence of 10 mM Mg2+, J3-Cascade-Cas3 was found to nick nSC pUC-λ (Fig. 6E, lane 2–6), but this increase in plasmid nicking over time is not observed for substrates that do not contain the target sequence (Fig. 6E, lane 8–12). When both magnesium and ATP are added to the reaction, the nSC target plasmid was found to be entirely degraded (Fig. 6F, lane 2–6), in contrast to a non-target plasmid that remains intact (Fig. 6F, lane 8–12). In agreement with the observation that Cascade cannot bind relaxed target plasmids, linear or nicked OC pUC-λ was not degraded by J3-Cascade-Cas3 (data not shown). To test whether the Cas3 helicase domain is required for the exonucleolytic plasmid degradation, we generated a mutant Cascade-Cas3 complex, carrying the Cas3 K320N mutation. As expected, the J3-Cascade-Cas3 K320N complex still shows nicking activity (Fig. 6G), but, despite the presence of ATP, it can no longer degrade target plasmids (Fig. 6H). To determine the direction of Cas3-mediated target degradation, we performed an in vitro exonuclease assay similar to that described for _Mja_Cas3″ (Beloglazova et al., 2011). Incubation of Cascade-Cas3 with any of three different substrates consisting of an 89 nt ssDNA (lacking a protospacer sequence) annealed at a variable position to a 39 nt probe (Fig. 6I) reveals that Cas3 has endonuclease and 3′ to 5′ exonuclease activity on the single stranded region of the DNA substrate (Fig. 6J). While in this assay the products high up in the gel are due to the 3′ to 5′ exonuclease activity of Cas3, the smaller products are due to a combination of single stranded endonuclease activity and exonuclease activity (Fig. 6J and Fig. S5). Altogether, these data demonstrate that during type I CRISPR-interference in E. coli target DNA recognition by Cascade is followed by Cas3-mediated DNA nicking and progressive ATP-dependent degradation of target DNA in the 3′ to 5′ direction.

Discussion

In this study we analyzed the molecular mechanism of the CRISPR-interference pathway of CRISPR/Cas type I-E, from target recognition to target degradation. As depicted in the model presented (Fig. 7), initial binding of a dsDNA target by Cascade leads to DNA bending and recruitment of the nuclease/helicase Cas3. Upon formation of a Cascade-Cas3 binary complex, the HD-domain of Cas3 cleaves the target DNA, which is followed by progressive ATP-dependent unwinding and degradation of the target in the 3′ to 5′ direction (Fig. 7). As Cas3 and Cascade-like complexes are present in all type I CRISPR/Cas systems (Makarova et al., 2011), the model is anticipated to be applicable for all these systems.

An external file that holds a picture, illustration, etc. Object name is nihms366982f7.jpg

Model of the CRISPR-interference type I pathway in E. coli. Steps that are not well-understood are depicted with dashed arrows. (1) Cascade (blue) carrying a crRNA (orange). (2) Cascade associates nonspecifically with the nSC plasmid DNA and scans for a protospacer (red), with protospacer adjacent motif (PAM) (yellow). (3) Sequence specific binding to a protospacer is achieved through base pairing between the crRNA and the complementary strand of the DNA, forming an R-loop. Upon binding, Cascade induces bending of the DNA, and Cascade itself undergoes conformational changes (Jore et al., 2011; Wiedenheft et al., 2011a). (4) The Cse1 subunit of Cascade recruits the nuclease/helicase Cas3 (brown). This may be triggered by the conformational changes of Cascade and the target DNA. (5) The HD-domain (dark brown) of Cas3 catalyzes Mg2+-dependent nicking of the target DNA at an unknown position, possibly within or near to the R-loop. (6) Plasmid nicking alters the topology of the target plasmid from nSC to relaxed OC, causing a reduced affinity of Cascade for the target. Dissociation of Cascade from the target may involve Cas3 helicase activity. Cascade may then remain associated with Cas3 or may be remobilized to locate new targets (7) Cas3 degrades the entire plasmid in an ATP-dependent manner as it progressively moves (in the 3′ to 5′ direction) along, unwinds and cleaves the target dsDNA. Exonucleolytic degradation takes place in the 3′-5′ direction, as was also reported for the combined activities of the helicase _Mja_Cas3′ and the nuclease _Mja_Cas3″ (Beloglazova et al., 2011).

Strikingly, Cascade alone is unable to bind protospacers on relaxed dsDNA. In contrast, it binds with high affinity to DNA with a nSC topology. In mesophiles, all circular dsDNA molecules have a nSC topology in vivo, which is essential for compaction of chromosomal DNA into the nucleoid, and for key cellular processes such as transcription, replication and recombination (reviewed in (Bates, 2005)). The higher affinity of Cascade for nSC targets is related to the energy required for opening the dsDNA duplex, which is DNA topology dependent. Negative supercoiling is considered a high energy DNA conformation and processes that require strand separation are generally supported by an nSC topology. A good example is RecA-mediated base pairing of ssDNA with the complementary strand in a dsDNA duplex during homologous recombination (reviewed in (Bell, 2005; Holthausen et al., 2010)), forming a so-called D-loop (Kasamatsu et al., 1971). Although relaxed dsDNA can also serve as a substrate for a RecA-ssDNA filament during ATP-independent strand exchange, the efficiency of strand exchange is greatly enhanced when the DNA duplex has a negative supercoiled topology (Cai et al., 2001).

Type III-B CRISPR/Cas systems have been demonstrated to target single stranded nucleic acids (RNA) (Hale et al., 2012; Hale et al., 2009; Zhang et al., 2012). It is possible that type I and type II CRISPR/Cas systems can also target ssDNA, such as phage M13 or conjugative plasmid DNA as it enters the cell. While systems that target single stranded nucleic acids can readily access protospacers, systems that require strand opening of dsDNA require an energy source. For denaturing a stretch of 32 bp in a 3 kb plasmid, approximately half of the energy needed (−90 kJ/mol) is supplied by the free energy of supercoiling. The preference for nSC target DNA is therefore likely to be a general characteristic for all mesophilic CRISPR/Cas systems that target double stranded DNA in an ATP-independent manner.

The remaining ~100 kJ/mol of Gibbs free energy that is required for denaturing a stretch of 32 bp may be derived from destabilizing interactions between Cascade and the dsDNA duplex, from stabilizing interactions between Cascade and the displaced strand, and from base pairing between the crRNA and the target strand. Moreover, the recently identified seed sequence in the crRNA (Semenova et al., 2011) may also decrease the energetic barrier, because the duplex first needs to be opened over a shorter stretch of only 8 nucleotides to pair with the crRNA seed.

Upon binding to the protospacer, conformational changes of Cascade (Jore et al., 2011; Wiedenheft et al., 2011a) and bending of the bound DNA take place. Protein-induced or intrinsic DNA bends tend to localize at the apex of supercoiled DNA loops, as this corresponds to the energetically most favorable conformation (Laundon and Griffith, 1988). Similar deformations have also been observed with enzymes that wrap DNA, such as RNA polymerase (Dame et al., 2002; Rivetti et al., 1999) and UvrB (Verhoeven et al., 2001), to facilitate DNA melting.

The conformational changes of Cascade and the target DNA may expose an interaction surface for Cas3 at or near the Cse1 subunit. After recruitment, the Cas3 HD-domain nicks the target DNA, resulting in a decreased affinity of Cascade for the target DNA. The Cas3 helicase domain has previously been reported to unwind R-loops in vitro (Sinkunas et al., 2011), indicating that the Cas3 helicase domain may be involved in recycling of Cascade after relaxation of the nSC target. Moreover, the Cas3 helicase domain has been shown to unwind dsDNA in the 3′ to 5′ direction (Beloglazova et al., 2011; Sinkunas et al., 2011). This dsDNA unwinding activity is essential for progressive 3′ to 5′ exonucleolytic degradation of target DNA by the HD-nuclease domain of Cas3 (this study; Beloglazova et al., 2011). In accordance with this, both the nuclease and the helicase activities of Cas3 are essential for CRISPR-interference. The double mutant of motif III (S483A/T485A), which is thought to couple ATP hydrolysis to unwinding activity, was still functional. Although many SF2 helicases lose their activity when this motif is mutated (e.g. translation initiation factor eIF-4A (Pause and Sonenberg, 1992)), other studies have shown that helicase activity is retained (e.g. the mitochondrial intron splicing helicase Mss116p (Del Campo et al., 2007)). This indicates that motif III mutants may display a phenotype that is not as pronounced as that of motif I and II mutants. The latter interfere with CRISPR-interference in vivo and with Cas3 exonuclease activity in vitro. The exonucleolytic degradation of target DNA by Cas3 in the 3′ to 5′ direction is in line with the reported activities of the helicase _Mja_Cas3′ and the nuclease _Mja_Cas3″ (Beloglazova et al., 2011).

Research over the recent years revealed a huge diversity of CRISPR/Cas defense systems: 10 subtypes (Haft et al., 2005; Makarova et al., 2006) that are grouped into 3 major types (Makarova et al., 2011). CRISPR-dependent invader nucleic acid cleavage has been reported for type II and type III CRISPR/Cas systems. In the S. thermophilus type II system target DNA is cleaved (Garneau et al., 2010). In contrast, the type III-B system of P. furiosus and S. solfataricus cleaves target RNA in vitro and in vivo (Hale et al., 2012; Hale et al., 2009; Zhang et al., 2012), while the type III-A system of Staphylococcus epidermidis targets DNA (Marraffini and Sontheimer, 2008), although cleavage has not yet been demonstrated. Cascade and Cas3 are the key players in the type I CRISPR-interference pathway. This study demonstrates that Cascade carries out ATP-independent DNA surveillance by exploiting the energy stored in nSC target DNA while Cas3 carries out complete ATP-dependent destruction of Cascade-marked invader DNA.

Experimental Procedures

Strains, Gene cloning, Plasmids and Vectors

E. coli BL21-AI and BL21 (DE3) strains were used throughout the study. A description of the plasmids used in this study can be found in the Supplementary Information (S5).

Protein production and purification

Cascade and the Cascade-Cas3 fusion complex were expressed and purified as described (Jore et al., 2011). Throughout purification a buffer containing 20 mM HEPES pH 7.5, 75 mM NaCl, 1 mM DTT, 2 mM EDTA was used for lysis and washing. Protein elution was performed in the same buffer containing 4 mM desthiobiotin.

Electrophoretic Mobility Shift Assay

Purified Cascade was mixed with target DNA (for a list of the oligo’s and plasmids used see Table S2 and S3, respectively) in a buffer containing 20 mM HEPES pH 7.5, 75 mM NaCl, 1 mM DTT, 2 mM EDTA, and incubated at 37 °C for 15 minutes. Samples containing Cascade-plasmid DNA complexes were run overnight at 10 mA on a 0.8 % TAE Agarose gel and post-stained with SybR safe (Invitrogen) 1:10000 dilution in TAE for 30 minutes. Cleavage with BsmI (Fermentas) or Nt.BspQI (New England Biolabs) was performed in the HEPES reaction buffer supplemented with 5 mM MgCl2. Samples containing Cascade bound to short double stranded oligo’s were separated on a polyacrylamide gel as before (Semenova et al., 2011).

Scanning Force Microscopy

Purified Cascade was mixed with pUC-λ (at a ratio of 7:1, 250 nM Cascade, 35 nM DNA) in a buffer containing 20 mM HEPES pH 7.5, 75 mM NaCl, 0.2 mM DTT, 0.3 mM EDTA and incubated at 37 °C for 15 minutes. Subsequently, for AFM sample preparation, the incubation mixture was diluted 10x in double distilled water and MgCl2 was added at a final concentration of 1.2 mM. Deposition of the protein-DNA complexes and imaging was carried out as described before (Dame et al., 2000).

Fluorescence Microscopy

Overnight culture of BL21-AI cells carrying CRISPR and cas gene encoding plasmids, was diluted 1:100 in fresh antibiotic-containing LB, and grown for 1 hour at 37 °C. Expression of cas genes and CRISPR was induced for 1 hour by adding L-arabinose (0.2%) and IPTG (1 mM). Phage Lambda infection was done at a Multiplicity of Infection (MOI) of 4. Cells were applied to poly-L-lysine covered microscope slides, and analyzed using a Zeiss LSM510 confocal laser scanning microscope (excitation at 514 nm and detection at 530–600 nm). For detailed materials see S5.

pUC-λ transformation studies

Transformation of competent cells expressing CRISPR and cas genes with pUC-λ was analyzed by selective plating. Plasmid curing was analyzed by transforming cas gene and CRISPR-containing cells with pUC-λ while suppressing T7-polymerase expression by the addition of 0.5% glucose. Expression of cas genes and CRISPR was induced and cells were plated on non-selective and on pUC-λ-selective LB-agar. After overnight growth the percentage of plasmid loss was calculated from the ratio of colony forming units on the selective and non-selective plates. For detailed materials see S5.

Phage infection studies

Phage Lambda (λvir), and M13 assays were performed as described (Brouns et al., 2008; Semenova et al., 2011). M13 mutants were generated as described previously (Semenova et al., 2011). For detailed methods see S5.

​

Highlights

Supplementary Material

01

Acknowledgments

We thank K. Zhang and H. van der Oost for experimental contributions, K. Peng and P. den Hollander (Laboratory of Virology, Wageningen University) and J.W. Borst (Laboratory of Biochemistry, Wageningen University) for their experimental assistance. We thank N. Goossen for fruitful discussions and D. Swarts for critically reading the manuscript. This work was financially supported by an NWO-TOP grant to JO (854.10.003), an NWO Veni grant to SJJB (863.08.014), the NWO Spinoza award (WMdV), an NWO Vidi grant to RTD (864.08.001), an NIH grant R01 59295 and grants from Russian Foundation for Basic Research and Russian Academy of Sciences Presidium program Molecular and Cellular Biology to KS.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References