Envelope stress is a trigger of CRISPR RNA-mediated DNA silencing in Escherichia coli

Authors


E-mail ailong.ke@cornell.edu; Tel. (+1) 607 255 3945; Fax (+1) 607 255 6249;

E-mail md255@cornell.edu; Tel. (+1) 607 254 8560; Fax (+1) 607 255 9166.

Summary

A widespread feature in the genomes of most bacteria and archaea is an array of clustered, regularly interspaced short palindromic repeats (CRISPRs) that, together with a group of CRISPR-associated (Cas) proteins, mediate immunity against invasive nucleic acids such as plasmids and viruses. Here, the CRISPR-Cas system was activated in cells expressing a plasmid-encoded protein that was targeted to the twin-arginine translocation (Tat) pathway. Expression of this Tat substrate resulted in upregulation of the Cas enzymes and subsequent silencing of the encoding plasmid in a manner that required the BaeSR two-component regulatory system, which is known to respond to extracytoplasmic stress. Furthermore, we confirm that the CasCDE enzymes form a stable ternary complex and appear to function as the catalytic core of the Cas system to process CRISPR RNA into its mature form. Taken together, our results indicate that the CRISPR-Cas system targets DNA directly as part of a defence mechanism in bacteria that is overlapping with but not limited to phage infection.

Introduction

The use of small RNAs to regulate gene expression is ubiquitous in all living organisms (Waters and Storz, 2009). In one remarkable instance, bacteria and archaea acquire resistance to bacteriophages and conjugative plasmids by employing an RNA-mediated defence mechanism against these foreign invaders. In this process, short fragments (∼24–48 nucleotides) of the invading DNA are integrated in the genome as spacers between similarly sized clusters of regularly interspaced short palindromic repeats (CRISPR) loci (Bolotin et al., 2004; Lillestol et al., 2006; Makarova et al., 2006; Barrangou et al., 2007; Brouns et al., 2008; Sorek et al., 2008). CRISPR loci have been identified in nearly all archaeal genomes and almost half of eubacterial genomes that have been sequenced to date (Sorek et al., 2008). They are often adjacent to an operon that encodes the CRISPR-associated (Cas) proteins (Fig. 1A), which are predicted RNA binding proteins, endo- and exo-nucleases, helicases and polymerases (Haft et al., 2005; Makarova et al., 2006). Bacteria encoding CRISPR-Cas systems that have been exposed to a virus or conjugative plasmid integrate phage- or plasmid-derived spacer sequences into the leader-proximal end of their CRISPR loci (Barrangou et al., 2007). The appearance of acquired spacer DNA in the genomes of these bacteria correlates with viral or plasmid resistance in a process that is dependent upon several of the cas gene products (Barrangou et al., 2007; Brouns et al., 2008).

Figure 1.

CasE-dependent silencing of ssTorA–GFP.
A. The CRISPR-Cas pathway of E. coli K12. Repeats and spacers are indicated by diamonds and rectangles respectively. Protein family nomenclature is described in elsewhere (Haft et al., 2005). Asterisk indicates the relative location of the mini-Tn10 insertion sites in casE and upstream of cas3 that were isolated by transposon library screening.
B and C. (B) Fluorescence microscopy and (C) flow cytometric screening of E. coli BW25113 (wt), BW25113 ΔdnaK or BW25113 ΔdnaKΔcasE expressing ssTorA–GFP from pTG, as indicated.
D. Northern blot of total RNA isolated from BW25113 (wt), BW25113 ΔdnaK or BW25113 ΔdnaKΔcasE cells carrying no plasmid (−) or expressing ssTorA–GFP from pTG. Blot was probed using GFP-specific digoxigenin-labelled oligonucleotides.

A key feature of CRISPR-encoded immunity is an inhibitory ribonucleoprotein complex comprised of a subset of Cas proteins and a guide RNA. This complex is believed to be responsible for targeting foreign genetic elements through base pairing between the bound RNA guide and either the sense or antisense strands of the target (Brouns et al., 2008). The formation of the guide RNAs begins when the CRISPR repeat-spacer arrays are transcribed from the leader region, producing a CRISPR transcript called the pre-crRNA (Hale et al., 2008; Lillestol et al., 2009). The full-length pre-crRNAs are subsequently processed into small CRISPR RNA (crRNA) molecules that correspond to a spacer flanked by two partial repeats (Tang et al., 2002; Brouns et al., 2008; Hale et al., 2008). How the expression of CRISPR arrays and cas genes is regulated when the cell is threatened by foreign DNA is poorly understood, although in Escherichia coli it appears to involve H-NS repression, which silences transcription from the CRISPR-Cas promoters (Pul et al., 2010) unless relieved by the transcriptional activator LeuO (Westra et al., 2010).

The production of crRNAs is carried out by distinct Cas proteins in different organisms. In Pyrococcus furiosus, Cas6 processes pre-crRNA to release individual invader-targeting RNAs and remains bound to the CRISPR repeat sequences at the 5′ end of the cleavage product (Carte et al., 2008). In E. coli, which lack Cas6, as many as five Cas proteins – CasABCDE – reportedly form a complex called CRISPR-associated complex for antiviral defence (Cascade) that processes CRISPR RNA (Brouns et al., 2008). Of these five enzymes, the catalytic activity of the CasE subunit is essential for pre-crRNA cleavage, which occurs within each repeat of the CRISPR RNA precursor (Brouns et al., 2008; Carte et al., 2008). Processed crRNAs encode the entire spacer unit flanked by the last eight bases of the repeat sequence at the 5′ end and a less well-defined 3′ sequence that ends in the next repeat region (Brouns et al., 2008; Carte et al., 2008). As mature crRNAs are similar at their 5′ but not their 3′ ends implies that this region may serve as a conserved binding site for Cas effector subunits, as suggested previously (Kunin et al., 2007; Carte et al., 2008). However, whether Cas6 or CasE is associated with the crRNA-Cas ribonucleoprotein (crRNP) effector complex, thereby coupling guide RNA biogenesis with target degradation, is currently not known. In fact, there are very few details about the identity and mechanism of the crRNP complex. Finally, while recent observations in E. coli and Staphylococcus epidermidis implicate DNA as the CRISPR-Cas target (Brouns et al., 2008; Marraffini and Sontheimer, 2008), there is little direct evidence to support this. In fact, in P. furiosus the target appears to be RNA (Hale et al., 2009).

Here, we show that the E. coli CRISPR-Cas system is triggered by the expression of a plasmid-encoded protein targeted for export by the twin-arginine translocation (Tat) pathway in cells that lack the DnaK molecular chaperone. Expression of this Tat substrate activated the CRISPR-Cas system, which resulted in specific silencing of the plasmid DNA that encoded the substrate. This silencing required: (i) base complementarity between genomic spacer DNA and the plasmid, (ii) specific targeting of the substrate to the Tat pathway, (iii) the BaeSR two-component signal transduction system and (iv) each of the Cas enzymes (CasABCDE and Cas1–3). At the heart of this silencing is the CasE protein, which formed a stable ternary complex with CasC and CasD, and functioned as the catalytic core of the Cas system to process pre-CRISPR RNA into mature crRNAs.

Results

Identification of CasE as a key component in prokaryotic gene silencing

Previously, we observed that a reporter protein consisting of the Tat signal peptide derived from E. coli trimethylamine N-oxide reductase (ssTorA) fused to the green fluorescent protein (GFP) was undetectable following its expression in E. coli cells that lacked the molecular chaperone DnaK (Fig. 1B and C) (Perez-Rodriguez et al., 2007). We initially suspected that the ssTorA–GFP protein was unstable in the absence of DnaK, rendering it susceptible to proteolytic degradation. However, Northern analysis revealed that the absence of ssTorA–GFP in E. coli BW25113 ΔdnaK cells was due to elimination of the encoding mRNA (Fig. 1D). When plasmid-encoded DnaK was provided in trans to ΔdnaK cells, cell fluorescence and sstorA–gfp mRNA levels returned to near wild-type (wt) levels (Fig. S1A and B).

To identify the factor(s) responsible for disappearance of sstorA–gfp mRNA, we screened a mini-Tn10 transposon library in the ΔdnaK strain background to isolate suppressor mutants that restored ssTorA–GFP expression, and thus cell fluorescence. The isolated suppressors mapped to the cas operon and, in particular, to the endoribonuclease encoded by casE and the promoter region of the HD nuclease-helicase encoded by cas3 (Fig. 1A). A ‘clean’ deletion of the casE gene in strain BW25113 ΔdnaK was created. As expected, cell fluorescence was recovered in ΔdnaKΔcasE cells expressing ssTorA–GFP (Fig. 1B and C). Fluorescence in this strain was attributed to stable accumulation of sstorA–gfp mRNA (Fig. 1D). To ensure this phenotype was due to casE inactivation and not polar effects on surrounding genes, we complemented the ΔdnaKΔcasE strain with a plasmid-encoded copy of CasE. Following expression of ssTorA–GFP, ΔdnaKΔcasE cells that had been genetically complemented with CasE were only weakly fluorescent (Fig. S1A) and accumulated a correspondingly low level of sstorA–gfp mRNA (Fig. S1B), indicating that CasE was responsible for the gene-silencing phenotype. Expression of an H20A variant of CasE (CasEH20A) that is incapable of pre-crRNA cleavage (Brouns et al., 2008) did not confer ssTorA–GFP silencing (Fig. S1A). The E. coli K12 cas system encodes at least eight genes: cas1, cas2, cas3 and five genes designated casABCDE that encode the Cascade complex (Brouns et al., 2008) (Fig. 1A). Individual deletion of each of these genes in a ΔdnaK strain background revealed that all of the other cas pathway components were required for the gene-silencing phenotype (Fig. S2A). It should be noted that the absence of DnaK was obligatory for silencing because wt and single cas mutants in the wt background all emitted strong fluorescence following ssTorA–GFP expression (data not shown). The requirement of the Cas enzymes for silencing was corroborated by stable accumulation of ssTorA–GFP in BL21(DE3) ΔdnaK cells (Fig. S2B), which lack the Cas enzymes (Brouns et al., 2008) but encode two CRISPR loci (Grissa et al., 2007) with several spacers that are complementary to sstorA (Fig. S2C).

We next investigated whether ssTorA–GFP silencing was dependent on complementarity between genomic spacers and plasmid-encoded sstorA. The E. coli K12 genome contains two CRISPR loci, the first of which is adjacent to the cas operon and includes three spacers (spacers 1, 5 and 8) that contain sequence complementarity to different regions of sstorA (Fig. 2A). It is noteworthy that all of these spacers were present in the genome prior to the introduction of plasmid pTG; thorough sequencing of both CRISPR regions prior to and at many time points (ranging from 1 h to as long as 3 days) after the introduction of the pTG plasmid provided no evidence for the insertion of new spacers corresponding to pTG. Nonetheless, when the entire CRISPR locus adjacent to the Cas enzymes (Fig. 1A) was deleted in ΔdnaK cells, silencing of sstorA–gfp was abolished as evidenced by the restoration of fluorescence to a level that was nearly identical to that observed in the ΔdnaKΔcasE mutant (data not shown). This suggested that silencing in ΔdnaK cells was indeed CRISPR-dependent. To determine whether complementarity between sstorA and spacers 1, 5 and/or 8 was responsible for silencing, we modified plasmid-encoded sstorA with synonymous mutations that we predicted would reduce base pairing between the crRNA and complementary target sequence, but would not change the ssTorA sequence at the protein level. Two or six synonymous point mutations were introduced into sstorA yielding the constructs ssS5–GFP or ssS1.5.8–GFP respectively (Fig. 2A). The synonymous mutations did not affect expression or localization of these constructs in wt cells relative to unmodified ssTorA–GFP (Fig. 2B). However, following expression in ΔdnaK cells, only ssTorA–GFP but not ssS5–GFP or ssS1.5.8–GFP was silenced (Fig. 2C and D), indicating that as few as two base pair mismatches was sufficient to block silencing. These results indicate that ssTorA–GFP silencing is specified by sequence identity between the encoding plasmid and a genomic spacer. It should be noted that despite the relatively short sequence matches, the complementary sequences were all GC rich (Fig. 2A).

Figure 2.

Complementarity between spacer DNA and ssTorA sequence is required for silencing activity.
A. Escherichia coli K12 spacers 1, 5 and 8 aligned with the sense strand of sstorA. Bold letters correspond to complementary base pairs. Numbering of sstorA corresponds to the bp position relative to the start codon of native E. coli TorA. Synonymous base changes are indicated below the sstorA sequence. The ssS5–GFP construct was created by mutating the C57 and C60 nucleotides in sstorA to T57 and T60 respectively. The ss1.5.8–GFP construct was created by further mutating ssS5–GFP as follows: G99T, G102A, G126C and G129C.
B. Western blot of the cytoplasmic (cyt) and periplasmic (per) fractions generated from BW25113 wt cells expressing ssTorA–GFP, ssS5–GFP or ssS1.5.8–GFP. Blot was probed with anti-GFP antibodies, stripped and reprobed with anti-GroEL antibodies, where GroEL served as a fractionation marker.
C. Flow cytometric screening of BW25113 wt and ΔdnaK cells expressing ssTorA–GFP, ssS5–GFP or ssS1.5.8–GFP.
D. Northern blot of total RNA from BW25113 ΔdnaK cells expressing ssTorA–GFP, ssS5–GFP or ssS1.5.8–GFP. Blot was probed using GFP-specific digoxigenin-labelled oligonucleotides.

Reconstitution of the core Cas complex

As all of the cas genes tested were required for silencing (Fig. S2A) and as the CasABCDE proteins were previously shown to copurify (Brouns et al., 2008), we next attempted to reconstitute the minimal E. coli Cas complex needed to process the pre-crRNAs into crRNAs. Each Cas protein was purified individually and most behaved as monomers in solution including CasE, which migrated at ∼30 kDa on the sizing column and interacted strongly with nucleic acids based on strong absorption at UV260 (Fig. 3A). CasC, on the other hand, oligomerized into a ∼490 kDa species. Interestingly, CasD was only soluble when coexpressed with CasC and the resulting complex migrated as a 105 kDa species (Fig. 3A) with an estimated molar ratio of 2:1 for CasC and CasD, respectively, based on Coomassie-stained band intensities (Fig. 3B). Strong interactions were also detected between CasC and CasE, which formed a ∼1:1 molar ratio complex migrating at ∼832 kDa (Fig. 3A and C). Different reconstitution schemes confirmed that CasC, CasD and CasE further formed a salt-stable ternary complex that was ∼200 kDa when CasD was fused with an N-terminal SUMO tag (Fig. 3A). Solubility was reduced in the absence of the SUMO tag, and the ternary complex migrated as a > 600 kDa species with an apparent stoichiometry estimated at 6:1:1 for CasC, CasD and CasE respectively (Fig. 3D). CasC appears to be the centre of the complex as no interactions were found between CasD and CasE. Previously, CasA and CasB were found at substoichiometric levels in the affinity purified E. coli Cascade complex (Brouns et al., 2008); however, incubation of separatedly purified CasA with the CasCDE complex did not lead to higher complex formation under ‘physiological’ conditions. Only marginal interactions were observed between CasB and CasC under low-salt conditions, which diminished under stringent conditions. Thus, we conclude that CasA and CasB are not stable components of the core Cas complex. This, however, does not rule out the possibility that CasA and CasB associate with the core Cas complex in a nucleic acid-dependent fashion. To investigate whether the molecular composition of the CasCDE complex is conserved phylogenetically, we carried out similar reconstitution of Thermus thermophilus Cas proteins and observed an identical ternary complex formed by CasCDE (data not shown). Furthermore, T. thermophilus CasE, which is ∼35% identical to E. coli CasE, was found to hetero-assemble with E. coli CasC or CasCD to form salt-stable higher complexes (data not shown). Thus, CasCDE-containing bacteria appear to form a conserved Cas core complex in Cas system subset 2 (Brouns et al., 2008).

Figure 3.

Reconstitution of E. coli Cascade core complexes.
A. Elution profile of CasC and CasE proteins and CasCD, CasCE, and CasCDE complexes on analytical Superdex 200 size-exclusion column. The x- and y-axes depict elution volume and UV absorbance at 280 nm respectively. Nucleic acids have been removed from Cas protein/complexes on prior ion-exchange columns.
B–D. SDS-PAGE analysis of the size-exclusion fractions of the (B) CasCD, (C) CasCE and (D) CasCDE complexes.

CRISPR RNA processing by core Cas complex

To reveal the function of the individual Cas proteins or Cas complexes in CRISPR RNA processing, we carried out in vitro RNA processing assays. Previous studies showed that pre-crRNA processing into unit length in E. coli requires a subset of Cas proteins, especially a catalytically competent CasE (Brouns et al., 2008; Carte et al., 2008). CasE-mediated processing involves an invariable cleavage site at the base of the 5′ hairpin structure to release the mature crRNA and a variable cleavage site at three to four different positions in the vicinity of the hairpin loop that may be due to the non-specific action of cellular ribonucleases (Brouns et al., 2008; Carte et al., 2008). Here, a HEX-labelled CRISPR RNA containing the conserved CRISPR repeat was used as substrate. When individually purified Cas proteins were incubated with this 5′-HEX-labelled CRISPR RNA, only CasE showed strong RNase activity (data not shown). By comparison with the alkaline hydrolysis ladder on sequencing gels, we determined that CasE cleaved at the base of the CRISPR stemloop (Fig. 4), coinciding with the invariable cleavage site that defines the 5′ end of the processed crRNA (Brouns et al., 2008). This activity is specific for CasE as it disappeared when the CRISPR RNA was incubated with the CasEH20A catalytic mutant (Fig. 4). We then carried out a more thorough analysis of the cleavage products produced by CasCE and the CasCDE complexes (in which CasD is N-terminally tagged with SUMO). Both of these complexes were shown to generate the same cleavage pattern on the CRISPR repeat RNA as the CasE protein alone (Fig. 4). The enzymatic activity again came from CasE, as CasCEH20A was unable to cleave the CRISPR repeat RNA. These results suggested that formation of higher order complexes did not alter the enzymatic specificity of CasE protein. We also verified that the cleavage patterns of the SUMO-tagged complex and untagged CasCDE complex, which migrated as a large oligomeric species, were identical in our assay conditions (data not shown) thus eliminating the possibility that the N-terminal SUMO tag in CasD altered the enzymetic specificity of the CasCDE complex. We cannot, however, rule out the possibility that the oligomeric form of the CasCDE complex is more efficient in processing the multimeric form of the pre-CRISPR RNA because of more efficient substrate binding.

Figure 4.

RNase activity of Cas proteins and CasE-containing complexes. Cleavage profile of CasE, CasCE, CasCEDSUMO, CasEH20A, CasCEH20A and CasC alone resolved on 18% sequencing gel. The amount of protein used in the assay was from left to right 2.5, 0.5 and 0.1 µM. Alkaline hydrolysis ladder in lanes 1 and 20 resolves the position of every base on the 18% sequencing gel.

Cas-mediated ssTorA–GFP silencing occurs via targeting of plasmid DNA

We next investigated whether plasmid pTG encoding ssTorA–GFP was targeted by the Cas machinery in ΔdnaK cells. Following induction of ssTorA–GFP, we purified plasmid pTG from an equivalent number of wt, ΔdnaK and ΔdnaKΔcasE cells and performed restriction analysis on the isolated plasmids. Plasmid pTG was stably maintained in wt and ΔdnaKΔcasE cells but was hardly detectable in ΔdnaK cells (Fig. 5A). Silencing of pTG in ΔdnaK cells was dependent on the sstorA–gfp sequence because empty pTrc99A plasmid accumulated at a similar level in all strains including ΔdnaK cells (Fig. 5A). We reasoned that plasmid silencing would cause a measurable decrease in antibiotic resistance conferred by the ampR gene encoded elsewhere on pTG. Indeed, the minimum inhibitory concentration (MIC) on ampicillin (Amp) of wt and ΔdnaKΔcasE cells was eightfold and sixfold greater than that for ΔdnaK cells (Fig. 5B). To verify that the Cas machinery only targets plasmid DNA and not ‘self’ DNA, we determined the viability of wt, ΔdnaK and ΔdnaKΔcasE cells by plating on non-selective medium following ssTorA–GFP induction. The colony-forming units (cfu) of ΔdnaK cells expressing ssTorA–GFP was actually slightly greater than the cfu of wt and ΔdnaKΔcasE cells (Fig. 5C), indicating that E. coli Cas effector complexes specifically target extrachromosomal DNA while avoiding genomic DNA. To test whether the Cas complex itself is capable of plasmid degradation, we measured the DNase activity of CasE, CasCE and CasCEDSUMO against single- and double-stranded DNA oligos, and against supercoiled plasmids including pTG encoding ssTorA–GFP and an unrelated pUC19 plasmid. Under our experimental conditions, neither CasE, nor the CasE-containing complexes had any measurable DNase activity (data not shown).

Figure 5.

Plasmid DNA is the target of CRISPR-Cas interference.
A. RsaI restriction analysis of plasmid DNA isolated from BW25113 wt, ΔdnaK and ΔdnaKΔcasE cells transformed with either pTG expressing ssTorA–GFP or empty pTrc99A (control) as indicated.
B. Selective plating of BW25113 wt, ΔdnaK and ΔdnaKΔcasE cells expressing ssTorA–GFP. An equivalent number of cells was plated on LB agar with no Amp (− Amp) or supplemented with 50, 100 and 200 µg ml−1 Amp.
C. Viability of the same cells as in (B). The number of cfu was determined by plating an equivalent number of induced cells on LB agar with no antibiotics.

Cas-mediated silencing depends on the BaeSR two-component signal transduction system

We next sought to determine how the CRISPR-Cas system was activated in ΔdnaK cells. As GFP expressed without a Tat export signal is not silenced in cells that lack DnaK (Perez-Rodriguez et al., 2007), targeting of GFP to the Tat translocase is a requisite for Cas-mediated silencing in ΔdnaK cells. Indeed, when the twin arginines in the Tat signal peptide of ssTorA–GFP were replaced with twin lysines [TorA(R11K:R12K)], amino acid substitutions that abolish Tat-dependent GFP export (DeLisa et al., 2002), no silencing of ssTorA–GFP was observed in ΔdnaK cells (Fig. 6A and B). Likewise, there was also no measurable silencing when export-competent ssTorA–GFP was expressed in ΔdnaK cells that also lacked the TatC translocase component, which is required for export (Fig. 6A and B).

Figure 6.

CRISPR-Cas silencing is dependent on membrane targeting and the BaeSR two-component stress response system.
A. Flow cytometric screening of BW25113 wt, ΔdnaK and ΔdnaKΔtatC cells expressing either ssTorA–GFP or ssTorA(KK)–GFP as indicated.
B. Northern blot analysis of total RNA from the same cells as in (A). Blot was probed using GFP-specific digoxigenin-labelled oligonucleotides.
C. Flow cytometric screening of BW25113 wt, ΔdnaK, ΔdnaKΔbaeS and ΔdnaKΔbaeR cells expressing ssTorA–GFP.
D. Electrophoretic mobility shift assay for BaeR binding to the Cas promoter region. A total of 0.15 pmol of DNA fragments corresponding to cas3 (415 bp) and casA (1057 bp) promoter regions was incubated without (−BaeR) or with various concentrations of phosphorylated BaeR-His6. The amount of BaeR was from left to right 9.2, 11.5, 16.1, 24.2 and 32.25 µM.

Based on these findings, we hypothesized that membrane localization of the highly expressed ssTorA–GFP fusion in ΔdnaK cells might induce an envelope stress, perhaps because of mislocalization of GFP in the inner membrane in these cells. We further speculated that signal transduction pathways that control the adaptive response to envelope stresses in E. coli might activate the expression of cas genes. If this interpretation is correct, then we would predict little to no cas gene expression, and thus no silencing activity, following ssTorA–GFP expression in cells where envelope stress response systems are compromised. To test this, we investigated the BaeSR two-component regulatory system, which includes a membrane-localized histidine kinase BaeS that senses envelope stress and modulates the phosphorylation state of the cytoplasmic transcription factor BaeR (MacRitchie et al., 2008). We focused our attention on this signalling pathway because overexpression of BaeR was previously shown to activate the promoter of the casA gene (Baranova and Nikaido, 2002). In support of our hypothesis above, we observed stable accumulation of ssTorA–GFP fluorescence in ΔdnaK cells that also lacked BaeS or BaeR (Fig. 6C).

As BaeSR was genetically involved in transmitting the envelope stress caused by ssTorA–GFP expression, we next investigated whether the BaeR regulator was able to bind cas promoter DNA. Recently, DNase I footprinting studies identified the BaeR-box sequence TCTNCANAA, where N is any nucleotide (Yamamoto et al., 2008). Our own inspection of the cas operon revealed a putative BaeR-box sequence of TCTGCATAA within the coding region of casA (nucleotides 298–306) and upstream of casBCDE. To determine whether BaeR was capable of binding this sequence, we performed electrophoretic mobility shift assays using purified BaeR and a 1057 bp DNA fragment covering the BaeR-box sequence. Two additional 358 and 415 bp DNA fragments upstream of casA and cas3 were also tested as candidate BaeR binding regions. As a positive control, a 279 bp fragment including the acrD promoter, a known binding target of BaeR (Hirakawa et al., 2005), was used. BaeR-His6 protein was purified and phosphorylated with acetylphosphate. The electrophoretic mobility of the DNA fragments corresponding to the acrD promoter region (data not shown) and the BaeR-box sequence internal to casA (Fig. 6D) shifted upon the addition of phosphorylated, but not unphosphorylated, BaeR. In contrast, no interaction was observed between phosphorylated BaeR-His6 and the DNA fragments derived from regions upstream of casA (data not shown) or cas3 (Fig. 6D). Thus, complexes formed by BaeR-His6 and the BaeR-box sequence internal to casA detected in this experiment were specific.

Induction of cas genes in ΔdnaK cells expressing ssTorA–GFP

We next determined how different mutations in the chromosome affected cas gene expression when ssTorA–GFP was induced. As expected, upregulation of casD and casE was observed following ssTorAGFP expression in ΔdnaK cells but not wt cells as determined by quantitative reverse transcription polymerase chain reaction (qRT-PCR; Fig. 7). It should be noted that in dnaK mutants, casD and casE expression was induced 15 min after ssTorA–GFP expression (Fig. 7) but returned to uninduced levels by 2 h after ssTorA–GFP expression (data not shown), indicating that cas gene expression is turned off once the envelope stress caused by pTG is eliminated by CRISPR-Cas silencing. However, similar stress adaptation was not observed in a ΔdnaKΔcasE double mutant. In these cells, casD expression remained elevated as long as 2 h after induction of ssTorA–GFP (data not shown), suggesting that the lack of pTG silencing in these cells results in a sustained stress and a corresponding sustained induction of the Cas pathway. Importantly, the induction of both casD and casE was abolished in ΔdnaKΔbaeR or ΔdnaKΔbaeS double mutants (Fig. 7), indicating that the BaeSR pathway is required for induction of the cas genes in response to ssTorA–GFP expression in ΔdnaK cells. This result is entirely consistent with our observation above that BaeS and BaeR are each required for plasmid silencing. As upregulation of cas gene expression depended on BaeSR, we speculated that genes whose induction is mediated by the BaeSR pathway, such as spy (Raffa and Raivio, 2002), might also be upregulated in ΔdnaK cells expressing ssTorA–GFP. In line with this hypothesis, we observed that spy expression was induced in ΔdnaK cells expressing ssTorA–GFP (Fig. 7). As was seen for casD and casE, mutation of baeR or baeS eliminated the elevated spy expression that was seen in the ΔdnaK strain expressing ssTorA–GFP (Fig. 7). However, spy expression was diminished in ΔdnaKΔcasE cells, indicating that the stress caused by expression of ssTorA–GFP in cells where silencing is absent was not a trigger of the BaeSR response. Hence, we conclude that the requirement for the Bae response in cells carrying the induced sstorA–gfp and a ΔdnaK mutation is complex and involves additional factors, beyond a simple sstorA–gfp-mediated upregulation of the Bae response.

Figure 7.

Induction of cas pathway in response to ssTorA–GFP envelope stress. Expression of casD (light gray bars), casE (dark gray bars), spy (black bars) and sstorA–gfp (white bars) as determined by qRT-PCR analysis. Total RNA was isolated 15 min after induction of ssTorA–GFP from each of the strains indicated. All data were normalized to the amount of 16S rRNA measured in each strain. Relative induction was then calculated by the comparative Ct method using the values measured in wt cells expressing ssTorA–GFP as calibrators. It should be noted that the expression levels of sstorA–gfp measured in these qRT-PCR experiments were entirely consistent with our Northern blot and cell fluorescence analysis.

We next examined whether other known Bae inducing cues triggered upregulation of the Cas pathway. Specifically, we tested whether induction of casD and CasE expression was observed in response to spheroplasting and separately, to the formation of misfolded pilin subunits caused by PapG overexpression. Both of these stresses are well known to activate the Bae signal transduction pathway (Raffa and Raivio, 2002). However, we found no evidence for casD or casE induction in response to either of these envelope stresses under the conditions tested here (data not shown).

Functional reconstitution of ssTorA–GFP silencing in cas-deficient E. coli

Finally, we speculated that overexpression of the Cas enzymes from an inducible promoter would bypass the BaeSR-dependent cas regulation and produce silencing in a manner that did not require DnaK. BL21(DE3) cells were used to test this notion because they lack genomic copies of the cas genes but contain CRISPR loci that encode several spacers with complementarity to sstorA. Indeed, coexpression of the entire cas operon from pWUR399 (Brouns et al., 2008) along with ssTorA–GFP in BL21(DE3) led to elimination of ssTorA–GFP in both the presence and absence of DnaK (Fig. S2B). This silencing was abolished when an H20A substitution was introduced to casE in pWUR399 (data not shown). Consistent with their role as a minimal RNA processing unit, coexpression of the CasCDE enzymes from pWUR402 was insufficient to produce silencing in this system (Fig. S2B). This result indicates that additional enzymes, perhaps from the set of CasA, CasB, Cas1, Cas2 and Cas3, are required for silencing and is in line with our single-gene knockout studies in K12 E. coli (Fig. S2A). The ability to functionally reconstitute ssTorA–GFP silencing in BL21(DE3) cells by simple addition of cas genes should enable further dissection of this unique mechanism.

Discussion

Previous studies have firmly established that CRISPR-Cas systems protect bacteria by conferring immunity against bacteriophage infection (Barrangou et al., 2007; Brouns et al., 2008) and limiting plasmid conjugation (Marraffini and Sontheimer, 2008). Here, we made the unexpected discovery that membrane localization of ssTorA–GFP in ΔdnaK cells activated the Cas pathway and led to silencing of the ssTorA–GFP-encoding plasmid. At present, the reason for Cas activation following ssTorA–GFP expression in ΔdnaK cells is poorly understood. The Tat system exports proteins that have undergone folding in the cytoplasm (Sanders et al., 2001; DeLisa et al., 2003) by inserting these folded substrates into the inner membrane (Hou et al., 2006). However, when aberrantly folded proteins are targeted to the Tat system, cells exhibit severe growth defects that are likely due to proton leakage at the membrane (Richter and Bruser, 2005). Hence, we speculate that the absence of the DnaK chaperone, which is known to assist the folding and translocation of certain Tat substrates (Oresnik et al., 2001; Perez-Rodriguez et al., 2007), resulted in misfolding and/or mislocalization of ssTorA–GFP in a manner that disrupted membrane integrity or was otherwise toxic to cells. The subsequent activation of CRISPR-Cas suggests that this system may provide a defence mechanism against defective protein localization that threatens the integrity of the inner cytoplasmic membrane. It is worth mentioning that we have not been able to find any evidence for the acquisition of new genomic spacers derived from plasmid pTG. Instead, the spacers targeting sstorA were encoded in existing genomic spacers. Whether preexisting spacers result in CRISPR-mediated silencing of endogenous torA or any other endogenous proteins is currently unclear. However, it seems unlikely that torA silencing is a conserved mechanism given the enormous variety in CRISPR spacers (Diez-Villasenor et al., 2010).

Nonetheless, our results represent the first evidence for the induction of CRISPR-Cas activity by a cue other than phage infection or plasmid conjugation (Barrangou et al., 2007; Brouns et al., 2008). While these cues are different, they are known to elicit overlapping stress responses (MacRitchie et al., 2008). Thus, we speculate that activation of CRISPR-Cas in response to defective protein export, phage infection and plasmid conjugation may be mediated by the same (or overlapping) stress sensing mechanisms (Fig. 8). In this study, ssTorA–GFP silencing in ΔdnaK cells was dependent upon the BaeSR signal transduction pathway, indicating that an endogenous cellular mechanism may be involved in the CRISPR-Cas defence system. Such regulated control of Cas enzyme expression is probably needed because constitutive expression of these potent RNases and DNases would be detrimental to the fitness of the host. In fact, recent evidence indicates that cas gene expression is strongly repressed by the nucleoid-associated protein H-NS (Pul et al., 2010), a protein known to silence transcription of laterally acquired genes. H-NS downregulates cas expression by binding to a promoter located upstream of the casA gene and its repression is proposed to be released by anti-silencing mechanisms, which allow the cell to express the genes under specifically regulated conditions. One such anti-silencing mechanism involves the transcriptional activator LeuO (Westra et al., 2010). We show here that another possible anti-silencing mechanism involves the BaeSR two-component regulatory system. Following induction of stress by ssTorA–GFP expression, activated BaeR may promote release of H-NS-mediated cas repression by binding to a newly identified promoter sequence (TCTGCATAA) that is located within the coding region of casA and downstream of the cas promoter where H-NS binds. It should be noted that while activation of the cas pathway minimally involves BaeSR, we cannot rule out the involvement of additional envelope stress response pathways that may work to cooperatively combat envelope stress by triggering CRISPR-Cas. Psp response is a likely candidate because it can be induced in E. coli upon filamentous phage infection (specifically phage secretin pIV) and by other membrane-damaging agents (Model et al., 1997), and it has been implicated in mitigating the stress associated with highly expressed Tat substrates (DeLisa et al., 2004).

Figure 8.

Model of CRISPR RNA-mediated DNA silencing in E. coli. Following an attack by a phage or the aberrant expression of a secreted or membrane protein, envelope stress response systems, such as the two-component BaeSR signal transduction pathway, are activated. An activated response regulator, for instance phosphorylated BaeR (BaeR-P), triggers expression of the Cas proteins and perhaps also the transcription of full-length CRISPR RNA. Full-length CRISPR RNA is recognized by the CasCDE core complex and processed to produce small guide RNAs (crRNAs), which have a uniformly processed 5′ end and a variable 3′ end. CasE or perhaps the entire CasCDE ternary complex remain associated with the processed crRNAs and may be guided to the target DNA to initiate DNA degradation. Whether CasCDE is a landmark to attract other effector proteins such as CasA and CasB through transient interactions is currently unclear but additional enzymes are necessary as CasE or CasE-containing complexes do not exhibit any measurable DNase activity. Nonetheless, silencing by the Cas effector complex is achieved by base pairing between the crRNA and the DNA target, which may occur during active replication of phage genome or plasmid DNA. Finally, a small number of bacteria acquire new spacers derived from the foreign DNA (marked by an asterisk), leading to survival by CRISPR RNA-mediated degradation of DNA.

Once activated, the CasABCDE proteins reportedly form a complex called Cascade, where the CasE protein within the complex is responsible for processing full-length CRISPR RNA (Brouns et al., 2008). Here we confirmed that CasE is sufficient for pre-CRISPR RNA processing, and that CasE is involved in the formation of a ternary CasCDE core complex that further oligomerizes to form a larger molecular assembly. A mechanistic model can therefore be generated to summarize our data, in which we hypothesize that by assuming a defined tertiary structure, the oligomeric CasCDE core complex possesses greater affinity for multimeric CRISPR repeat regions than individual Cas proteins. This in turn allows more efficient pre-CRISPR processing by CasE. Structure determination of the CasCDE core architecture and its interactions with the CRISPR repeat sequence will be critical for further understanding the pre-CRISPR processing mechanisms.

In addition to E. coli K12, we have also observed a CasCDE core in T. thermophilus (data not shown), suggesting that the pre-CRISPR processing complex is minimumly conserved in Cas System 2 (CASS2). As CasCD pairs have been identified in five out of the seven categorized CASS subtypes (Makarova et al., 2006), we speculate that these two proteins, especially CasC, likely play an important role in organizing the formation of the Cascade complex. However, it should be noted that the other three Cascade proteins, CasA, CasB and CasE, are not well conserved among CASS subtypes (Makarova et al., 2006). CRISPR repeat sequences and their predicted secondary structures can also differ significantly between organisms (Grissa et al., 2008). Therefore, the pre-CRISPR processing factor in CasE-deficient organisms may differ from that in E. coli. Cas6 is found in many CasE-deficient organisms (Haft et al., 2005). Despite sharing little sequence homology, CasE and Cas6 adopt a common duplicated ferredoxin fold and catalyse metal-independent RNA cleavage (Carte et al., 2008). It remains to be determined whether Cas6 can interact with CasCD in these organisms. Moreover, some bacteria encode neither CasE nor Cas6 in their CRISPR-Cas system. For example, single ferritin fold containing Csy4 protein from Pseudomonas aeruginosa was shown to carry out pre-CRISPR processing in a sequence specific fashion (Haurwitz et al., 2010). Therefore, the pre-CRISPR processing mechanism likely varies among CASS subtypes, underlining the diversity of CRISPR-Cas systems.

While it is now clear that the induction of cas gene expression results in the formation of CasCDE, which promotes the maturation of the crRNAs, the recognition and degradation of the foreign target occurs by mechanisms that remain poorly understood. Once generated, crRNAs use base pairing to presumably guide a crRNP complex to the invasive target. In our studies, disruption of perfect base pairing by as few as two mutations was sufficient to abrogate silencing. This is consistent with earlier studies showing that even a single spacer/target mismatch compromises CRISPR interference (Barrangou et al., 2007; Deveau et al., 2008; Marraffini and Sontheimer, 2008). For instance, of 19 phages that evolved to evade CRISPR targeting in Streptococcus thermophilus, eight contained a single mutation and three contained a double mutation (Deveau et al., 2008). A surprising aspect of our studies was the relatively short (∼8–11 base pairs) spacer/target complementarity that was required for silencing. Given the short lengths of these matches, equilibrium hybridization thermodynamics would seem to be insufficient for discrimination between plasmid DNA and random genomic targets, suggesting that an additional layer of target ‘sensing’ is performed by the CRISPR-Cas machinery. One example of such an additional layer is the observation that differential complementarity outside of the spacer sequence is an in-built feature of CRISPR systems for self (spacer DNA within the encoding CRISPR locus itself) versus non-self discrimination (Marraffini and Sontheimer, 2010). Accumulated evidence points to DNA as the target for CRISPR-mediated degradation in E. coli (Brouns et al., 2008) and S. epidermidis (Marraffini and Sontheimer, 2008); however, recent studies of P. furiosus raise the possibility that some CRISPR-Cas systems target RNA (Hale et al., 2009). This functional difference seen for P. furiosus may be due to the presence of a Cas repeat-associated mysterious protein module, which contains six genes, cmr1–cmr6, that are absent in numerous prokaryotes including E. coli and S. epidermidis (Haft et al., 2005; Makarova et al., 2006). Our own studies suggest that plasmid DNA is the intracellular target of the E. coli CRISPR-Cas system; however, whether the Cas enzymes alternatively target the RNA of a factor required for plasmid replication, segregation and/or maintenance cannot be ruled out. Hence, a key issue moving forward is to definitively prove whether Cas enzymes silence DNA and/or RNA in E. coli. Based on the observation that mature crRNAs remain associated with P. furiosus Cas6 (Carte et al., 2008) and the E. coli CasCDE complex (our unpublished observations), we speculate that the same complex may participate directly in target degradation. One possibility for DNA targets is that active replication of phage genome or plasmid DNA allows a window of opportunity for crRNA to base pair with the single-stranded target DNA. Such a mechanism is reminiscent of eukaryotic RNAi, in which Argonaute2 retains one strand of the small interfering RNA (siRNA) as the guide to base pair and cleave target mRNAs (Liu et al., 2004). While the composition of the in vivo crRNP effector complex remains a mystery, our genetic screen showed that Cas1, Cas2, Cas3 and all five Cascade proteins are required for ssTorA–GFP silencing in a ΔdnaK strain background. Biochemically only the CasCDE enzymes were required in CRISPR processing but they did not exhibit any measurable DNase activity. Thus, given that E. coli Cas1 and Cas2 function upstream of the CRISPR processing and DNA targeting stages (Brouns et al., 2008), it follows that CasA, CasB and/or Cas3 may participate in the downstream effector stage (Fig. 8). We also cannot rule out the participation of other enzymes that may be encoded outside of the cas operon. We are currently investigating whether the CasCDE complex serves as a landmark to attract effector proteins through transient interactions (Fig. 8).

Experimental procedures

Bacterial strains, plasmids and growth conditions

Escherichia coli strain BW25113 (lacIqrrnBT14ΔlacZWJ16hsdR514ΔaraBADAH33ΔrhaBADLD78) and single-gene knockout mutants of BW25113 from the Keio collection (Baba et al., 2006) were used for all experiments unless otherwise noted. BW25113 ΔdnaK double knockout strains were created by P1vir phage transduction. Briefly, kanamycin (Kan)-marked alleles derived from single-gene knockout strains from the Keio collection were transduced in recipient BW25113 ΔdnaK cells that had the Kan resistance marker in dnaK eliminated as described (Datsenko and Wanner, 2000). The CRISPR deletion strain was made by generating a Kan-marked deletion in BW25113 ΔdnaK cells as described (Datsenko and Wanner, 2000). BL21(DE3) [F-ompT hsdSB (inline imageinline image) gal dcmλ(DE3)] (Novagen) or T7 Express lysY (New England Biolabs) cells were used as hosts for expression of the BaeR and Cas enzymes respectively.

Strains were routinely grown aerobically at 37°C in Luria–Bertani (LB) medium, and antibiotics were supplemented at the following concentrations: Amp, 100 µg ml−1; Kan (5 µg ml−1); spectinomycin (Spec, 50 µg ml−1); tetracycline (Tet, 10 µg ml−1). Protein synthesis was induced when the cells reached an absorbance at 600 nm (Abs600) of ∼0.5 by adding 0.2% arabinose and/or 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) to the media. For silencing experiments, protein synthesis was induced for 4–6 h prior to characterizing cellular fluorescence or harvesting plasmid DNA, total mRNA and total proteins.

Plasmids

Plasmid pTG encoded the E. coli ssTorA signal peptide fused to GFP in pTrc99A (Perez-Rodriguez et al., 2007) and pTrc–GFP encoded GFP cloned in pTrc99A (Diao et al., 2006). For dnaK genetic complementation studies, BW25113 Δdnak cells were transformed with pOFXbad-KJ2, which encodes dnaK and dnaJ under control of the pBAD promoter (Castanie et al., 1997). For casE genetic complementation studies, BW25113 ΔdnakΔcasE cells were transformed with pOFXbad–casE, which encodes casE under control of the pBAD promoter. Synonymous mutations in the ssTorA signal peptide were created using site-directed mutagenesis (QuikChange® II Site-Directed Mutagenesis Kit, Stratagene) according to manufacturer's protocols. Plasmid pssS5–GFP was created by mutating the C57 and C60 nucleotides in sstorA to T57 and T60 respectively. Plasmid pss1.5.8–GFP was created by further mutating pssS5–GFP as follows: G99T, G102A, G126C and G129C. Similarly, plasmid pTorA(KK)–GFP was created by mutating the consensus twin-arginine residues in the ssTorA signal peptide to twin-lysine residues using site-directed mutagenesis. For protein purification, CasC, CasE and CasEH20A were PCR amplified from E. coli K12 genomic DNA and cloned into pET-21a(+) (Novagen) with an N-terminal His6-tag. CasA, CasB, CasD, Cas1 and Cas3 were cloned into the pQE-80 vector (Qiagen) with an N-terminal His6-tag. Cas2 was cloned in pGEX-4T-1 (GE Healthcare) to generate a GST fusion. Additional plasmids used in Cas protein coexpression studies included: CasC cloned in the pCDFDuet-1 vector (Novagen) without any affinity tag, CasD cloned into the pSUMO vector with an N-terminal His6-SUMO tag and CasE cloned in pET-29b(+) (Novagen) without any affinity tag. CasEH20A was constructed using Phusion Site-Directed Mutagenesis Kit (New England Biolabs). BaeS and BaeR were cloned in pET-21a(+) (Novagen) with a C-terminal His6-tag. For cas complementation studies, plasmid pWUR399, pWUR402 (Brouns et al., 2008) or pWUR399(CasEH20A) was cotransformed in strain BL21(DE3) along with pTG and induced using IPTG. Plasmid pWUR399(CasEH20A) was created by site-directed mutagenesis of plasmid pWUR399.

Generation of transposon library and mapping of insertions

An initial mini-Tn10 transposon insertion mutant library was generated by tranducing E. coli NLL51, an MC4100 recA− derivative, to tetracycline resistance using a population of lambda phage that carried mini-Tn10 inserts as described previously (Takiff et al., 1989). The initial mini-Tn10 insertion library contained 4.2 × 104 transductants, representing ∼10-fold coverage of the genome. This library was then transduced in ΔdnaK cells carrying pTG by P1vir phage transduction and 3.0 × 104 clones were recovered on LB-Tet/Amp plates, for approximately eightfold coverage of the genome. To isolate highly fluorescent clones, library cells were grown overnight, subcultured and induced for 4–6 h. Induced library cells were diluted 200-fold in 1 ml of phosphate-buffered saline (PBS, pH 7.4) and sorted using a FACSCalibur flow cytometer (Becton Dickinson Biosciences) as described previously (DeLisa et al., 2002). A total of ∼1 × 106 cells were examined in 30 min, and ∼400 events were collected. Individual clones were re-screened via flow cytometry and a fluorescent plate reader (Synergy HT, BioTek Instruments) for verification of fluorescent phenotypes.

To sequence the insertion mutations in the chromosome of positive clones, a nested PCR strategy was used. This entailed a first PCR reaction using genomic DNA from ΔdnaK mini-Tn10 mutants as template, and a reverse primer (CTTTGGTCACCAACGCTTTTCCCG) specific for the 3′ end of the mini-Tn10 insert and a forward arbitrary primer (GGCCACGCGTCGACTAGTACNNNNNNNNNNGCTGG) that annealed to the flanking chromosome. A second PCR was then performed using the PCR product from the first reaction as template, along with a forward primer (GGCCACGCGTCGACTAGTAC) that annealed to the constant portion of the arbitrary primer and a reverse primer (CATATGACAAGATGTGTATCCACC) that annealed to an upstream region of the mini-Tn10 sequence. The resulting amplified DNA was gel purified and sequenced.

Cellular fluorescence analysis

Overnight cultures harbouring GFP-based plasmids were subcultured into fresh LB medium with appropriate antibiotics and protein expression was induced in mid-exponential phase growth. After 4–6 h, cells were harvested by centrifugation and washed once with PBS, pH 7.4. Next, 5 µl of washed cells were diluted into 1 ml of PBS and analysed using a FACSCalibur flow cytometer and CellQuest Pro software. All mean cell fluorescence data from flow cytometric analysis are the average of three replicate experiments (n = 3). Error bars represent the standard error of the mean (s.e.m.). Fluorescence microscopy was performed as described (Kim et al., 2008).

RNA and protein expression analysis

Total RNA was isolated from 10 ml of cells using the Qiagen RNeasy Mini Kit. Northern blots were performed as follows: 5 µg of total RNA was separated on a formaldehyde agarose gel, transferred to a positively charged nylon membrane by capillary elution and immobilized by baking the membrane. Digoxigenin-labelled oligonucleotide probes were synthesized using PCR amplification in the presence of digoxigenin-dUTP according to manufacturer's protocol (Roche Applied Sciences). Hybridizations were performed under stringent conditions (55°C in 50% formamide, 0.25 M sodium phosphate pH 7.2, 0.25 M sodium chloride, 7% SDS, 100 µg ml−1 fragmented salmon sperm DNA and 5 µg ml−1 yeast tRNA). Membranes were washed and exposed to anti-digoxigenin antibody from mouse (diluted 1:10 000) and CDP-Star substrate according to manufacturer's protocol (Roche Applied Sciences). Detection was with X-ray film.

Cytoplasmic and periplasmic protein fractions were generated from 10 ml of cells using the cold osmotic shock procedure (Kim et al., 2005). Subcellular fractions were separated by SDS-PAGE using 12% Tris-HCl gels (Bio-Rad), transferred to Immobilon P (Millipore) and blotted using the following primary antibodies: mouse anti-GFP (Sigma; diluted 1:2 000) and mouse anti-GroEL (Sigma; diluted 1:20 000). Secondary antibodies were goat anti-mouse and goat anti-rabbit horseradish peroxidase conjugates (Promega) diluted 1:2500. Signals were detected using enhanced chemiluminescence (ECL) (Amersham). To verify the fractionation quality, membranes were first probed with primary antibodies and, following development, stripped in Tris-buffered saline/2% (w/v) SDS/0.7 M β-mercaptoethanol. Stripped membranes were re-blocked and probed with anti-GroEL antibodies.

Protein purification

For biochemical studies, Cas protein expression vectors were transformed into T7 Express lysY host cells (New England Biolabs) and expression was induced at 15°C for 20 h with 0.3 mM IPTG. The harvested cells were lysed by sonication in lysis buffer (50 mM Tris-HCl pH 8.5, 500 mM NaCl, 10 mM imidazole, 5 mM β-mercaptoethanol and 1 mM benzamidine chloride). The supernatant after centrifugation was applied onto 5 ml of Ni-sepharose column (Sigma) for each purification. After protein binding, the column was washed thoroughly with 100 column volumes of lysis buffer followed by 10 volumes of lysis buffer supplemented with 40 mM imidazole. The bound protein was eluted from the column using five column volumes of elution buffer (300 mM imidazole pH 7.5, 200 mM NaCl, 5 mM β-mercaptoethanol and 1 mM benzamidine chloride). Each purified protein was concentrated and further purified using a Superdex 200 size-exclusion column followed by an ion-exchange column (MonoQ or SourceS) (GE Healthcare). Cas2 was expressed and purified in an N-terminal GST fusion form. All of the purified proteins were analysed by SDS-PAGE and judged to be greater than 95% purity.

To coexpress the CasCE complex, pCDFDuet-1–CasC and pET-21–CasE were cotransformed into T7 Express lysY cells. Following coexpression, the CasCE complex was purified on a Ni-sepharose column followed by a Superdex 200 sizing column and a SourceQ ion-exchange column on FPLC as described above. The fact that CasC copurified with His6-tagged CasE protein despite stringent high-salt wash suggests that these two proteins form a very stable complex. Following a similar strategy, we confirmed that the coexpressed CasC (from pCDFDuet-1–CasC) and CasD (from pQE80–CasD) proteins form a stable complex. In this case, a single His6-tag was introduced at the N-terminus of the CasD protein.

To test whether a ternary complex formed between CasC, CasD and CasE, we cotransformed pET-29–CasE (encoding CasE without a His6-tag), pCDFDuet-1–CasC and pQE80–CasD (encoding CasD with an N-terminal His6-tag) into the host cell T7 Express lysY. Following coexpression, the CasCDE complex was purified with a Ni-sepharose column followed by a Superdex 200 column and a SourceQ ion-exchange column on FPLC as described above. An apparent stoichiometry of 2:1:2 was observed for the CasC, CasD and CasE proteins in the ternary complex after the above purification (data not shown). However, after extended incubation at 4°C and heparin column separation, the stoichiometry of the ternary complex shifted to 6:1:1 (Fig. 3D), reflecting the tendency of CasC to oligomerize into most likely a helical hexamer in a dynamic fashion. Although only CasD contains a His6-tag, CasE and CasC copurifies with CasD under stringent purification conditions, suggesting that these three proteins indeed form a stable ternary complex. Because the CasCDE complex could not be satisfactorily separated from the CasCD complex by size-exclusion or ion-exchange chromatography, another coexpression system was designed in which CasD was cloned into the pSUMO vector with an N-terminal His6-SUMO tag. The resulting CasCEDSUMO complex again copurified on the Ni-sepharose column under stringent purification conditions, and CasCEDSUMO and CasCD were separated on Superdex 200 column based on their size difference.

RNase activity assay

The HEX-labelled RNA substrate (5′-GAGUUCCCCGCGCCAGCGGGGAUAAACCG-3′) was purchased from Sigma. RNA and proteins were incubated for 30 min in a buffer containing 20 mM Tris pH 7.5, 20 mM NaCl, 1 mM MgAc2 and 10% glycerol. The reaction mixture was combined with loading buffer at a ratio of 1:9 and separated on an 18% sequencing gel. The gel was scanned with a Typhoon 9400.

DNase activity assay

The ds-DNase activity was carried out in 20 µl reactions at 37°C for 2 h in a buffer containing 10 mM Tris-HCl pH 7.9, 50 mM NaCl, 10 mM MgCl2 and 1 mM dithiothreitol using 0.05 g l−1 supercoiled pTG or pUC19 plasmids as substrates. The reaction was then phenol-extracted, separated on a 0.8% agarose gel and stained with ethidium bromide. The ss-DNase assay was carried out in the same buffer. 5′ Fluorescein-labelled 29 nt DNA oligo (fluorescein-TCGCGCATACCCTGCGCGTCGCCGCC) at a concentration of 100 nM encoding part of spacer 5 was used as substrate to allow fluorescent detection. In a parallel set of experiment, this oligo was heat-annealed with equimolar amount of the complementary strand to assay for the ds-DNase activity under the same conditions. The reaction mixture was separated in 18% sequencing gel, and the fluorescent signals were read using a Typhoon 9400 Imager. All reactions were carried out at 37°C with the addition of 200 nM Cas proteins/complexes.

Plasmid DNA restriction analysis

Plasmid DNA was isolated from an equivalent number cells carrying either pTG and or empty pTrc99A (control) using a Qiagen Miniprep kit. An equal volume of plasmid DNA was digested with the non-unique restriction enzyme RsaI (New England Biolabs) for 4 h at 37°C. After removing the enzyme from the reaction using a PCR cleanup kit (Qiagen), 20 µl of each DNA-containing sample was mixed with loading dye (1:1) and separated on a 1% agarose gel pre-stained with ethidium bromide.

Bacterial MIC and viability analysis

The MIC of bacteria grown was determined by selective plating of cells on Amp. Overnight cultures were diluted 105-fold in liquid LB and plated on LB agar supplemented with increasing Amp concentrations and 0.1 mM IPTG. Reported MIC values were the average of three replicate experiments. The number of colony-forming units per millilitre (cfu ml−1) was determined as follows: overnight cultures were subcultured and induced with 1 mM IPTG after reaching an Abs600∼0.5. Following 4 h of induction, an equivalent amount of cells (Abs600∼1.0) was harvested and plated on LB agar with no antibiotics. The number of cfu ml−1 was determined by counting individual colonies on plates after overnight growth. Reported cfu ml−1 values were the average of three replicate experiments.

Electrophoretic mobility shift analysis

Phosphorylated or unphosphorylated BaeR was assayed for its ability to bind putative Cas promoter DNA in the vicinity of the casA and cas3 genes by electrophoretic mobility shift analysis. Putative Cas promoter DNA in the vicinity of the casA and cas3 genes cloned from BW25113 wt genomic DNA by PCR using the following primers: casA forward: 5′TAAACCGCTTTTAAAACCACCACCAAAACCT-3′, casA reverse: 5′GATGGATGGGTCTGGCAGGG-3′; cas3 forward: GGCATATATATTTAAAAGGTTCCATTAATAGCCTCCCTGTTTTTTTAGTA, cas3 reverse: GGTTACACGAAGGGTAAATATTGCCGGACAAATT. Control DNA fragments corresponding to the acrD (−1 to −276 bp) promoter region were similarly prepared by PCR with BW25113 genomic DNA as a template. Separately, BaeR-His6 was expressed in BL21(DE3) cells from pET-21–BaeR and purified using a Ni-sepharose column (Qiagen). Purified BaeR-His6 was phosphorylated using 50 mM acetylphosphate in a buffer containing 100 mM Tris (pH 7.4), 10 mM MgCl2 and 125 mM KCl. The protein was incubated in this buffer for 1 h at 30°C. Next, 0.15 pmol of DNA was mixed with varying amounts (9.2, 11.5, 16.1, 24.2 and 32.25 µM) of the phosphorylated protein in a 20 µl reaction containing 100 mM Tris (pH 7.4), 10 mM MgCl2, 100 mM KCl, 10% glycerol and 2 mM dithiothreitol. The reaction mixture was incubated for 20 min at room temperature and the samples were electrophoresed on a 10% Tris-Borate-EDTA (TBE) gel (Bio-Rad) in TBE buffer at 4°C. The gel was then soaked in a 1:10 000 dilution of SYBR-Gold stain (Invitrogen) and the DNA was imaged under UV light using a Gel Doc 2000 (Bio-Rad).

Analysis of gene expression by qRT-PCR

Quantitative reverse transcription polymerase chain reaction analysis of cas, spy, and sstora-gfp gene transcript abundance was performed similar to that described elsewhere (Westra et al., 2010). Briefly, cDNA was synthesized using Omniscript RT Kit (Qiagen) from RNA extracted using the RNeasy Mini Kit (Qiagen) and DNase I-treated using RNase-free DNase Set (Qiagen). The qRT-PCR reactions were performed using TaqMan Universal PCR Master Mix (Applied Biosystems) according to manufacturer's instructions. The PCR reactions were performed on an ABS Prism 7000 Sequence Detection System (Applied Biosystems) and analysed using ABS Prism 7000 SDS software (Applied Biosystems). Fold change of gene transcription was calculated using the relative quantification method with rrsB endogenous control. The gene transcript abundance of cas, spy and sstorA–gfp in E. coli BW25113 wt cells harbouring pTG were used as calibrators. All PCR reactions were performed in triplicate.

Acknowledgements

We thank Joseph Peters and Adam Parks for assistance with transposon library construction and mapping of insertion mutations. We also thank Fang Ding and Xiaochun Wu for technical assistance. R.P.R. gratefully acknowledges Cornell University for support through a Sloan Fellowship. This material is based upon work supported by the National Science Foundation under Grant CBET 0449080 (to M.P.D.), a NYSTAR James D. Watson Award (to M.P.D.), a NYSTAR Distinguished Faculty Award (to M.P.D.) and the National Institute of Health under Grant GM-086766 (to A.K.).

Ancillary