The ribonuclease activity of Csm6 is required for anti-plasmid immunity by Type III-A CRISPR-Cas systems

ABSTRACT

CRISPR-Cas systems provide prokaryotes with RNA-based adaptive immunity against viruses and plasmids. A unique feature of Type III CRISPR-Cas systems is that they selectively target transcriptionally-active invader DNA, and can cleave both the expressed RNA transcripts and source DNA. The Type III-A effector crRNP (CRISPR RNA-Cas protein complex), which contains Cas proteins Csm1-5, recognizes and degrades invader RNA and DNA in a crRNA-guided, manner. Interestingly, Type III-A systems also employ Csm6, an HEPN family ribonuclease that does not stably associate with the Type III-A effector crRNP, but nevertheless contributes to defense via mechanistic details that are still being determined. Here, we investigated the mechanism of action of Csm6 in Type III-A CRISPR-Cas systems from Lactococcus lactis, Staphylococcus epidermidis, and Streptococcus thermophilus expressed in Escherichia coli. We found that L. lactis and S. epidermidis Csm6 cleave RNA specifically after purines in vitro, similar to the activity reported for S. thermophilus Csm6. Moreover, L. lactis Csm6 functions as a divalent metal-independent, single strand-specific endoribonuclease that depends on the conserved HEPN domain. In vivo, we show that deletion of csm6 or expression of an RNase-defective form of Csm6 disrupts crRNA-dependent loss of plasmid DNA in all three systems expressed in E. coli. Mutations in the Csm1 palm domain, which are known to deactivate Csm6 ribonuclease activity, also prevent plasmid loss in the three systems. The results indicate that Csm6 ribonuclease activity rather than Csm1-mediated DNase activity effects anti-plasmid immunity by the three Type III-A systems investigated.

Introduction

The multiple diverse CRISPR-Cas systems function to protect bacteria and archaea from viruses, plasmids, and other mobile genetic elements [1-3]. Each system employs distinct CRISPR RNA species and Cas proteins to form effector crRNPs that engage in crRNA-guided, Cas nuclease-mediated destruction of foreign nucleic acids. Of the six major types of CRISPR-Cas systems [2,4], Types I, II, and V selectively destroy invader DNA, while Type VI is specific for invader RNA [1]. How Type IV CRISPR-Cas systems function remains to be determined. The Type III CRISPR-Cas systems are notable in that they eliminate both the RNA and DNA components of invasive mobile genetic elements [5–17].

Type III-A and the related Type III-B CRISPR-Cas effector crRNPs include five (Type III-A; Csm1-5) and six (Type III-B; Cmr1-6) protein subunits, respectively [18–24]. Both Type III systems include an additional protein – Csm6 (Type III-A) or Csx1 (Type III-B) [2] – that is not observed to be a stable component of the associated effector complex [6,7,14,17,21]. It is unclear if Csm6 and Csx1 function in trans or are recruited to the effector crRNPs transiently during function.

DNA targeting by Type III systems requires that the invasive DNA be actively expressed [8–12,15,17]. Interaction between the crRNA and a newly transcribed target RNA activates the DNase activity of the complex resulting in co-transcriptional destruction of the non-template strand of invader DNA as well as target RNA destruction [11]. Interestingly, Type III CRISPR-Cas system-mediated protection from viruses and plasmids relies on a DNase and two separate RNase activities [7,8,13,15,25–29].

The DNase activity of Type III CRISPR-Cas systems maps to the Csm1 (III-A) and Cmr2 (III-B) protein subunits [9–13,30–32]. The location of the Csm1 and Cmr2 DNase active site(s) is the subject of seemingly conflicting evidence: different mutational studies support a role for the conserved HD domain [9,10,12,17,31–33], the GGDD motif of the conserved palm domain [11,17,34], or both [9,10,17]. Importantly, reconsideration of the conclusions of early mutational analysis is warranted by recent findings that indicate that the palm domain of Csm1/Cmr2 functions in the synthesis of cyclic oligoadenylates [35,36], which have been shown to activate the RNase activity of Csm6 (III-A) [35,36]. The same signaling mechanism presumably also operates to activate the RNase activity of the related Csx1 (III-B) protein and recently it was determined that short linear oligoadenylates greatly stimulate Csx1 RNase activity in vitro [37]. The palm domain mutations may be preventing Csm6/Csx1 RNase activity rather than preventing Csm1/Cmr2 DNase activity.

One of the RNase activities of the Type III effector crRNP complexes is associated with the Csm3 protein (III-A) [5,14,38] or Cmr4 (III-B) [6,18,39–41]. Csm3 and Cmr4 subunits bind along the length of the integral crRNA and cleave bound target RNAs at regular six-nucleotide intervals [14,21,38,40,41]. The second RNase activity is provided by Csm6 (III-A) or Csx1 (III-B) proteins, which cleave a variety of RNA substrates in vitro [13,17,35–37,42,43]. These proteins share N-terminal CARF (Cas-associated Rossman fold) and C-terminal HEPN (Higher eukaryotes and prokaryotes nucleotide binding) domains. Evidence indicates that the catalytic activity resides in an RNase active site located within the HEPN domain [13,37,42–45]. The RNase activities of Csm6 and presumably Csx1 are stimulated by the short cyclic oligoadenylate signaling molecules generated by the Csm1/Cmr2 subunits of the Type III effector crRNP complexes upon target RNA recognition [35–37]. The conserved N-terminal CARF domains of Csm6 and Csx1 bind the oligoadenylate ligands and allosterically activate the RNase activity [35–37,45].

The role of Csm6 in Type III-A mediated immunity is of particular interest because, while it appears to be dispensable for RNA and DNA targeting activities of the crRNP effector complex in vitro [5,11,12,14], it is critical for immunity to plasmids or phages in vivo [7,13,15,25], at least under certain conditions [13]. Here, we have investigated Type III-A systems from three different bacteria – Lactococcus lactis, Streptococcus thermophilus, and Staphylococcus epidermidis – expressed as modules of component crRNAs and Csm1-6 proteins in E. coli [28]. All three modules confer resistance to plasmids that express transcripts containing sequences targeted by the crRNAs [28]. Our results reveal a conserved requirement for the RNase activity of Csm6 and the GGDD motif in the palm domain of Csm1 (known to activate Csm6 ribonuclease function) rather than DNase activity, in Type III-A CRISPR-Cas system-mediated immunity to plasmids.

Results

L. lactis Csm6 is an HEPN-dependent ribonuclease

Previous biochemical characterization of purified S. thermophilus and S. epidermidis Csm6 proteins indicated that these proteins are single-stranded ribonucleases [13,35,43]. The RNase activity of the proteins was disrupted by mutation of the conserved histidine of the R-X_4-6-H RNase active site within the C-terminal HEPN domain [13,35,43]. To determine if L. lactis Csm6 functions as a ribonuclease, wildtype and HEPN active site mutant (H360A) versions of the protein expressed in and purified from E. coli (Figure 1(a)) were incubated with 5ʹ-radiolabeled single- and double-stranded RNAs and DNAs, and DNA/RNA hybrid molecules (Figure 1(b,c)). Wildtype Csm6, but not the HEPN mutant protein, selectively cleaved single-stranded RNAs. Cleavage by L. lactis Csm6 resulted in accumulation of distinct RNA breakdown products, suggesting possible endonucleolytic activity (Figure 1(b,c)). The results demonstrate that L. lactis Csm6 is an HEPN-dependent ribonuclease that exhibits specificity for cleavage of single-stranded RNA.

Figure 1.

L. lactis Csm6 is a single-strand-specific ribonuclease that employs an HEPN active site. (a) The upper diagram highlights the conserved domains of Csm6 which include an N-terminal CARF domain and a C-terminal HEPN domain containing the R-X_4-6-H RNase active site motif. The lower panel shows purified wildtype and HEPN active site mutant (H360A) L. lactis 6x-His tagged Csm6 proteins following SDS-PAGE and Coomassie staining. Protein size standards (M) in kDa. (b) Wildtype and HEPN mutant L. lactis Csm6 were tested for nuclease activity using ³²P-labeled RNA (RNA1) compared to no protein (-) control. Reactions were incubated for 60 mins. and the RNA was resolved by denaturing polyacrylamide gel electrophoresis. The arrow indicates full length RNA and the bracket indicates Csm6-specific cleavage products. RNA size standards (M) in nucleotides. Results shown are representative of seven repeats of this experiment. (c) The ability of wildtype and HEPN Mutant L. lactis Csm6 to degrade a range of single- and double-stranded RNA and DNA substrates was tested (RNA2, RNA2+RNA3, DNA1, DNA1+DNA2, DNA1+RNA2). See Table S1 for sequences of the RNA and DNA substrates. Results shown are representative of two repeats of this experiment.

L. lactis Csm6 is a divalent metal-independent endoribonuclease

Further investigation revealed that L. lactis Csm6 can function as an endoribonuclease. The ability of the purified protein to cleave linear and circular forms of an RNA substrate was assessed (Figure 2(a)). Wildtype L. lactis Csm6 cleaved both forms of the RNA. The slightly shifted pattern of 5ʹ radiolabeled Csm6 RNA cleavage products observed between the linear vs. circular forms of substrate RNA apparently reflects that the radiolabel is internalized in the circular RNA and indicates that RNA cleavage by Csm6 occurred at specific sites within the substrate RNA. The Csm6 HEPN mutant was unable to cleave either form. The ability of wildtype Csm6 to cleave circular RNAs shows that the protein is an endoribonuclease and does not require free RNA ends as is the case for obligate exoribonucleases.

Figure 2.

L. lactis Csm6 is a divalent metal-independent endoribonuclease and cleaves on the 5ʹ side of the phosphodiester bond. (a) Endoribonuclease activity of L. lactis Csm6 was tested by incubating wildtype (WT), HEPN mutant (H360A) or no (-) Csm6 protein with linear or circular forms of a 5ʹ radiolabeled RNA (RNA4; Table S1) as described in Figure 1. Results shown are representative of three repeats of this experiment. (b) Divalent metal-dependency of Csm6 RNA cleavage activity was tested by incubating wildtype L. lactis Csm6 with radiolabeled RNA (RNA5; Table S1) in the absence (-) or presence of increasing amounts of EDTA (0.1, 0.5, 1, 5, and 10 mM) and compared to a no Csm6 protein (-) control. Results shown are representative of two repeats of this experiment. (c) The enzyme poly(A) polymerase (PAP) was utilized to determine the 3ʹ terminal chemical end group of Csm6-specific RNA degradation products. The RNA (RNA6; Table S1) was incubated in the presence or absence (-) of Csm6 followed by treatment with or without (-) PAP. Results shown are representative of four repeats of this experiment. (d) The enzyme Terminator 5ʹ-Phosphate-Dependent Exonuclease (TEX) was utilized to determine the 5ʹ terminal chemical end groups of Csm6 RNA degradation products. The RNA (Circularized or linear RNA6; Table S1) was incubated in the presence of absence (-) of Csm6 followed by treatment with or without (-) TEX. In panels A-D, the RNAs were resolved by denaturing polyacrylamide gel electrophoresis and full-length RNA is indicated with an arrow, while Csm6-specific RNA cleavage products are indicated with a bracket. In panel C, RNAs elongated by poly(A) polymerase are indicated with stars while in panel D, a star indicates a TEX-specific, minor RNA breakdown product. Results shown are representative of three repeats of this experiment.

Other HEPN family member ribonucleases have been found to cleave RNA in a divalent metal ion-independent manner [37,42–44]. We tested the hypothesis that Csm6 also cleaves RNAs without a requirement for divalent metal-ions (Figure 2(b)). We found that RNA cleavage by Csm6 did not require addition of divalent metal ions and furthermore, the cleavage reaction was not affected by addition of the divalent metal ion chelator ethylenediaminetetraacetic acid (EDTA) even at high (10 mM) concentrations. A hallmark feature of divalent metal-independent ribonucleases is that they cleave on the 5ʹ side of phosphodiester bonds such that a 5ʹ hydroxyl group, and a 2ʹ-3ʹ cyclic phosphate or 3ʹ phosphate groups are left on the RNA degradation products [37,42,43,46]. Using assays previously employed to map the chemical end groups of RNA cleavage products [42 and see Methods section], we determined that digestion of RNA by L. lactis Csm6 resulted in products with 5ʹ hydroxyl, and 2ʹ-3ʹ cyclic phosphate or 3ʹ phosphate termini (Figure 2(c,d)). A control RNA containing a 3ʹ hydroxyl end group was elongated by E. coli poly(A) polymerase (PAP), which only adds non-templated adenosine residues to RNAs with 3ʹ hydroxyl groups, but Csm6 RNA cleavage products were not elongated (Figure 2(c)). Moreover, the Csm6 RNA cleavage products were resistant to cleavage by Terminator 5´-Phosphate-Dependent Exonuclease (TEX), an enzyme that degrades RNAs containing a 5ʹ phosphate group (Figure 2(d)).

Csm6 demonstrates a preference for cleavage at purines

Next, we investigated whether Csm6 ribonuclease activity exhibits base specificity. Previous work found that S. thermophilus Csm6 showed a preference for cleaving after GA and AA dinucleotides [35]. We incubated L. lactis Csm6 (Figure 3) and S. epidermidis Csm6 (Figure 4) with three 5ʹ radiolabeled RNA substrates having different sequences. The RNA products were separated on sequencing gels and mapped at nucleotide resolution. To assess the specificity of any sites of Csm6 cleavage, alkaline hydrolysis and RNase T1 ladders of each substrate RNA were used in parallel along with RNA standards differing by ten nucleotide size intervals. As specificity controls, we omitted Csm6 protein or included RNase-defective mutant forms of each Csm6 (H360A or H369A) protein. Degradation products observed in control lanes were considered non-specific and ignored for analysis. The mapping results indicate that both L. lactis and S. epidermidis Csm6 exhibit a clear preference for cleavage after purines; major cleavages were only observed after guanosine and adenosine residues (Figures 3 and 4). L. lactis and S. epidermidis Csm6 produced nearly identical cleavage patterns for each of the tested RNA substrates, however minor differences in the particular purines recognized can be observed. Very weak cleavage activity after certain uridine residues was detected with S. epidermidis but not L. lactis Csm6. The results indicate that Csm6 proteins from three different bacterial species (L. lactis (Figure 3), S. epidermidis (Figure 4), and S. thermophilus [35]) cleave RNA preferentially after purines.

Figure 3.

L. lactis Csm6 cleaves RNA with a preference for adenosines and guanosines. (a) Three different RNA sequences (19mer (RNA1), 30mer (RNA5), and 39mer (RNA7); sequences given in panel B and Table S1) were incubated with no protein (-) or wildtype or H360A mutant L. lactis Csm6 for 5, 15, 30, and 60 minutes (a 60-minute incubation time was chosen for mutant Csm6 and no protein control). The RNAs were resolved by denaturing sequencing gel electrophoresis alongside 5ʹ radiolabeled RNA markers (M), RNase T1 ladders (T1) and alkaline hydrolysis ladders (OH). Vertical bars show cleavage sites. Nucleotides at major cleavage sites are shown in red bold and with red arrows in panel A. Nucleotides at minor cleavage sites are shown in black bold and with black dots in panel A. Results shown are representative of multiple repeats of this experiment. RNA1 was tested six times and RNA7 was tested two times.

Figure 4.

S. epidermidis Csm6 cleaves RNA with a preference for adenosines and guanosines. Three different RNA sequences (19mer (RNA1), 30mer (RNA5), and 39mer (RNA7); sequences given in panel B and Table S1) were incubated with no protein (-) or wildtype or H369A mutant S. epidermidis Csm6 for 5, 15, 30, and 60 minutes (a 60-minute incubation time was chosen for mutant Csm6 and no protein control). The RNAs were resolved by denaturing sequencing gel electrophoresis alongside 5ʹ radiolabeled RNA markers (M), RNase T1 ladders (T1) and alkaline hydrolysis ladders (OH). Red arrows indicate major and black dots indicate minor Csm6-specific RNA cleavage products. (b) Cleavage products were mapped back to each RNA. Vertical bars show cleavage sites. Nucleotides at major cleavage sites are shown in red bold and with red arrows in panel A. Nucleotides at minor cleavage sites are shown in black bold and with black dots in panel A. Results shown are representative of multiple repeats of this experiment. RNA1 and RNA5 were tested three times and RNA7 was tested one time.

The RNase activity of Csm6 is critical for anti-plasmid immunity

L. lactis, S. epidermidis, and S. thermophilus Type III-A systems including the Csm6 protein selectively carry out anti-plasmid immunity when expressed as modules in E. coli [28]. Immunity was tested using two types of plasmids for each system: a Csm module plasmid and a target plasmid (Figure 5(a)). As described previously [28], the Csm module plasmids contain all components necessary for producing Csm crRNPs and the target plasmids contain sequences with crRNA sequence homology flanked by promoter and terminator elements to confer expression of the target RNA. To determine if Csm6 is important for Type III-A immunity in vivo, we tested for crRNA-dependent anti-plasmid immunity in strains lacking Csm6 or expressing RNase-defective Csm6 (Figure 5). The E. coli strains were challenged with either a plasmid that contained a sequence that is complementary to the crRNA (target) or one that lacks crRNA complementarity (non-target). The plasmids encode ampicillin resistance. Plasmid challenge results in a high number of ampicillin resistant colonies in strains lacking a functional III-A CRISPR-Cas system (Figure 5(a)). Strains with an active Type III-A CRISPR-Cas system yield a low number of ampicillin resistant colonies at the plated dilutions (Figure 5(a)). The wildtype L. lactis, S. epidermidis, and S. thermophilus Type III-A systems successfully conferred anti-plasmid immunity against plasmids containing the crRNA target sequences, but not those that lacked target sequences, as was also previously demonstrated [28],(Figure 5(b–d), wildtype Csm6, target and non-target plasmids).

Figure 5.

Csm6 RNase activity is required to support Type III-A anti-plasmid immunity. (a) In vivo plasmid interference assay. E. coli strains harboring plasmids (orange; chloramphenicol selectable) expressing Csm crRNPs from L. lactis, S. epidermidis, or S. thermophilus were transformed with target and non-target plasmids (blue; ampicillin selectable) as previously described [28]. The target plasmid contains a sequence (yellow) that is complementary to expressed crRNA while the non-target plasmid has a sequence (green) that lacks crRNA homology. Serial ten-fold dilutions of transformed cells are spotted onto plates containing chloramphenicol and ampicillin and CRISPR-Cas mediated plasmid loss (defense) is indicated by a reduction in colonies. (b–d) Plasmid interference assays were performed with Csm modules containing or lacking (ΔCsm6) wildtype Csm6 or an RNase-defective form of Csm6 (HEPN mutant) in L. lactis (b), S. epidermidis (c), and S. thermophilus (d) Type III-A systems. Results shown are representative of three different experiments per system.

For all three Type III-A systems tested, anti-plasmid immunity was lost when the Csm6 protein was not expressed or if an RNase-defective HEPN mutant (H to A) protein was substituted (Figure 5(b–d)). Strains in which the Csm6 protein was omitted or mutated typically resulted in 4–5 orders of magnitude more colonies than wildtype strains containing Csm6 when plasmids with crRNA targets were assayed. In the absence of Csm6 (∆Csm6), there appears to be a slight (one order of magnitude or less) reduction in colony number when comparing the data of plasmids with crRNA targets vs. those lacking targets. Wildtype and HEPN mutant L. lactis Csm6 proteins were recovered at comparable levels from the strains (Figure S1), suggesting that the phenotype of the Csm6 HEPN mutant was due to disruption of ribonuclease activity rather than simply to loss of stability. Furthermore, biochemical analysis of T. thermophilus Csm6 showed that the same HEPN point mutation did not hinder Csm6 homodimerization further indicating that Csm6 structure is not obviously disrupted by the mutation [43]. To determine whether or not Csm6 was functioning as an integral component of the Type III-A effector crRNPs, the effector complex was isolated and evaluated for the presence of any associated Csm6 protein (Figure S1). Affinity purification of the 6x histidine-tagged Csm3 subunit confirmed that Csm1-5 but not Csm6 are stable components of the L. lactis effector crRNP complexes (Figure S1), in agreement with findings with other Type III-A systems [7,14]. Our results show that anti-plasmid activity of the Type III-A systems from L. lactis, S. epidermidis, and S. thermophilus critically depends upon the Csm6 protein. Furthermore, the evidence indicates that anti-plasmid immunity depends on the HEPN domain and the ability of Csm6 to catalyze RNA cleavage.

The palm domain of Csm1 is needed for anti-plasmid immunity

The Csm1 proteins of Type III-A systems cleave DNA in vitro [11,12,30–33]. It is not clear whether the DNase active site resides in the HD domain and/or the palm domain [7,11,12,32,33] or whether the DNase activity is vital for CRISPR-Cas immunity. Moreover, mutation of the palm domain of Csm1 has recently been shown to disrupt Csm6 ribonuclease activity by blocking cyclic oligoadenylate production [35,36], complicating the interpretation of previous studies where reduction of Csm6 activity may have contributed to phenotypes being examined [7,11].

To assess whether the two conserved domains (HD and palm) of the Csm1 subunit are important for anti-plasmid immunity in L. lactis, S. epidermidis, and S. thermophilus Csm systems including Csm6 in vivo, we examined the activity of E. coli strains expressing Csm1 HD domain (HD to AA) or palm domain (GGDD to GGAA) mutants (Figure 6). For all three Type III-A systems, mutations in the palm domain significantly reduced the anti-plasmid immunity compared to strains with wildtype Csm1. In contrast, anti-plasmid immunity was not observably affected by mutations in the HD domain. The results reveal that the GGDD motif of the palm domain but not the HD domain of Csm1 is critical for plasmid resistance in all three Type III-A systems investigated (Figure 6).

Figure 6.

Mutation of the palm domain of Csm1 prevents anti-plasmid immunity. Plasmid interference assays were carried out as described in Figure 5 except that the effects of HD and palm domain mutations of the Csm1 protein were compared to that of wildtype Csm1. The effects of Csm1 mutations on plasmid defense were tested for L. lactis (a), S. epidermidis (b) and S. thermophilus (c) Type III-A systems. Results shown are representative of three different experiments per system.

Discussion

Type III-A CRISPR-Cas systems are selectively activated when a crRNA target sequence on a phage or plasmid is transcribed [7,11,12,15,28,32]. Interestingly, Type III-A systems include three distinct nucleases that can function to target the destruction of both RNA and DNA molecules of the invaders. Target RNA binding by the Type III-A effector crRNP complex (consisting of Csm1-5 and crRNA) is the central trigger that prompts the DNase (Csm1) and ribonuclease (Csm3 and Csm6) activities [11–13,35,36]. It is not yet clear how these three nucleases are deployed to eliminate invaders or the hierarchical importance of each nuclease for CRISPR-Cas immunity. In this study, we focused on dissecting the role of the Csm6 protein in anti-plasmid immunity using three different Type III-A systems comprised of components from L. lactis, S. epidermidis, and S. thermophilus. Although Csm6 is apparently not stably associated with the Type III-A crRNP effector complex (Figure S1 [7,14], we found that it is vital for anti-plasmid immunity in each of the three bacterial systems tested (Figure 5). Our biochemical studies confirmed that Csm6 proteins function as divalent metal-independent endoribonucleases that employ the HEPN RNase active site to cleave single-stranded RNA molecules (Figure 1 and 2). In vivo analysis clearly established that Csm6 RNase activity is important for anti-plasmid immunity; the ability of the Type III-A systems to execute anti-plasmid immunity was nearly eliminated by mutations that disrupt RNase activity of Csm6 (Figure 5). In addition, we show that mutations in the Csm1 palm domain that have been shown to inhibit activation of Csm6 by cyclic oligoadenylate signaling [35,36] also disrupted plasmid defense (Figure 6).

Csm6 enzymes cleave RNAs after purines

Our detailed biochemical characterization of the L. lactis Csm6 protein showed that this protein exhibits all of the hallmark features of other previously characterized HEPN superfamily RNases [44]. Indeed, Csm6 and related Csx1 (Type III-B) proteins from various bacterial and archaeal species including L. lactis (Figures 1–3), S. epidermidis (Figure 4 and [13,43]), S. thermophilus [35], T. thermophilus [43], P. furiosus [42], P. horikoshii [43] and S. islandicus [37] all appear to function as endoribonucleases that specifically cleave single-stranded RNA using an HEPN RNase active site. Like other divalent metal-independent ribonucleases, Csm6 and Csx1 cleave on the 5ʹ side of the phosphodiester bond to produce RNA degradation products with 5ʹ hydroxyl and a 2ʹ-3ʹ cyclic phosphate or 3ʹ phosphate group end groups (Figure 2(b) and [17,36,42].

Our detailed mapping revealed that both L. lactis and S. epidermidis Csm6 RNases selectively cleave after purines (Figure 3 and 4). This specificity for cleavage after certain guanine and adenosine residues was also reported for the S. thermophilus Csm6 protein [35]. No base specificity was reported in an earlier study that characterized RNA cleavage by S. epidermidis Csm6 [13,36], however, it is clear that discrete RNA cleavage products accumulated in the previous study, and our results with the same enzyme predict that the observed Csm6 RNA cleavage products would map to purines if high resolution mapping studies were performed [13,36 and Figure 4]. Lack of base specificity was also reported for RNA cleavage by T. thermophilus Csm6 [43], but was based only on testing of homo-polymer RNA substrates and not mixed-sequence RNA substrates [43]. In contrast, the Csx1 protein of the P. furiosus Type III-B system cleaves RNAs with a strict specificity for adenosine residues [42]. It is not clear whether the observed purine specificity of Csm6 or Csx1 proteins has any physiological significance other than to provide an effective means to cleave a wide range of possible viral, plasmid or host RNAs encountered during invader defense. Our findings that Csm6 proteins cleave after certain purines but not others in single-stranded RNA substrates (Figures 3 and 4) and do not cleave double-stranded RNA (Figure 1), suggests that the enzyme may cleave at exposed single-stranded regions within RNA secondary structures.

Role of Csm6 RNase activity in anti-plasmid immunity

The importance of Csm6 and the Type III-B-associated HEPN ribonuclease Csx1 in immunity against plasmids and phages appears to vary with a number of biological conditions. For example, in S. epidermidis it appears that Csm6 is not required for phage immunity if the phage target sequence is expressed early in the phage infection cycle [13]. However, Csm6 is critical if the target DNA is expressed later in the cycle or if the target sequence includes mutations expected to weaken the immunity afforded by the DNase (Csm1) and RNase (Csm3) activities of the Type III-A effector crRNPs [13]. Deletion of S. epidermidis csm6 results in disruption of immunity against conjugative plasmids [7], but not transformed plasmids [13]. In contrast, deletion of Staphylococcus aureus csm6 partially reduces immunity against transformed plasmids [47]. Similarly, for unclear reasons, anti-plasmid immunity by Type III-B systems requires Csx1 in S. sulfolobus [8] but not P. furious [9]. Some of these findings may reflect a conditional dependence of Csm6 (and Csx1) in defense wherein these RNases function as auxiliary systems to the Type III crRNP effector complexes and are essential when the effector crRNP complexes fail to efficiently eliminate the invader DNA [13]. For three Type III-A systems tested with transformed plasmids, we found that Csm6 was essential for anti-plasmid immunity and that a functional HEPN RNase domain was integral for this function of Csm6 (Figure 5).

Importantly, our finding that deletion or mutation of Csm6 severely disrupts anti-plasmid immunity (Figure 5) indicates that DNA destruction by the Type III-A effector crRNP is not sufficient for efficient anti-plasmid immunity in our assay. However, the mechanism by which Csm6 affects immunity against plasmids will require further investigation. The plasmid transcript that is targeted in our system (an artificial transcript engineered to contained a crRNA target sequence [28] is not expected to be essential for plasmid maintenance. However, there is evidence that Csm6 cleavage activity is not strictly confined to the region of crRNA interaction in vivo, but spreads several kilobases in either direction from the crRNA interaction site [13]. Thus, it is possible that Csm6-mediated RNA cleavage spreads to encompass mRNAs essential for plasmid replication resulting in plasmid loss, or to host mRNAs resulting in death of bacterial cells containing a plasmid. A similar mechanism has been proposed to explain E. coli growth defects observed during anti-plasmid immunity by the exclusively RNA targeting Cas13a (Type VI) CRISPR-Cas system [48].

Key function of the palm domain of Csm1 in anti-plasmid immunity

There is considerable evidence that the Csm1 (Type III-A) and Cmr2 (Type III-B) proteins (members of the Cas10 superfamily) function both as target RNA-activated DNases [7,9–13,15,17,32,36] and cyclic oligoadenylate synthases (that activate the RNase activities of associated Csm6 and Csx1 proteins) [35–37]. We found that anti-plasmid immunity by three bacterial Type III-A systems requires an intact Csm1 palm motif, which is required for the cyclic oligoadenylate synthesis and possibly also for DNase activity [7,11,30],(Figure 6, Mutation of the HD domain, which has also been implicated in DNase activity in some systems [31,32,33,35], did not disrupt anti-plasmid activity (Figure 6).

It is unclear whether the disruption in anti-plasmid immunity observed with mutation of the Csm1 palm domain in our study is the result of disruption of DNase or oligoadenylate polymerase activity, or both. Similar palm domain mutations have been shown to be critical for DNase activity of S. epidermidis Csm1 [11,30], but not S. thermophilus Csm1 [12], in vitro. This suggests that the importance of the Csm1 palm domain observed in our work (Figure 6) is via the activation of Csm6 RNase activity, at least in the case of the S. thermophilus system. The lack of effective anti-plasmid immunity in the absence of Csm6 indicates that DNA destruction by the Type III-A effector crRNP is not sufficient for anti-plasmid immunity by any of the 3 systems investigated (Figure 5). A straightforward explanation for the impact of the Csm1 palm domain mutations in our work (Figure 6) is that oligoadenylate signaling and Csm6 RNase activation (rather than DNA cleavage) is required for anti-plasmid immunity.

Comprehensive genome sequence analyses reveal that Type III CRISPR-Cas systems represent one of the most common, as well as highly diverse and variable types of CRISPR-Cas systems [2]. In addition to the Type III-A and III-B systems, there are also uncharacterized Type III-C and III-D systems that are predicted to encode equivalents to Csm1 or Cmr2 subunits with catalytically inactive HD or palm domains, respectively [2]. The diversity in Type III components predicts natural variation in the roles of the protein homologs in protection against plasmid and viral invaders. Detailed characterization of members of each of the four subtypes of Type III systems is needed to tease out both common features as well as the predicted mechanistic differences.

Materials and methods

Cloning and mutagenesis

The construction of the wildtype L. lactis, S. epidermidis, and S. thermophilus Csm module plasmids (encoding Csm1-6, Cas6 and CRISPR array; pACYC), L. lactis csm6 gene deletion plasmid, wildtype Csm6-only plasmids (pT7), and the crRNA target and non-target plasmids (pTrcHis), were reported [28]. The subsequent csm6 or csm1 gene deletions or mutations of the components of each Csm module plasmid as well as plasmids encoding just Csm6 were generated using either Quikchange site directed mutagenesis (Stratagene), splicing by overlap extension PCR, or inverse PCR, as described next. Quikchange site directed mutagenesis was performed on csm6 genes of the Csm module plasmids (S. epidermidis H369A and S. thermophilus H347A) or Csm6-only plasmids (L. lactis H360A and S. epidermidis H369A) to create the desired HEPN domain mutation. Inverse PCR was used to delete the csm6 genes from the S. epidermidis and S. thermophilus systems. Splicing overlap extension PCR was used to make constructs with the following wildtype or mutant csm6 and csm1 genes: L. lactis wildtype and HEPN mutant (H360A) (6x histidine-tagged or untagged) csm6, L. lactis HD (H13A/D14A) and palm domain (D576A/D577A) csm1 mutations, and S. thermophilus HD (H15A/D16A) and palm domain (D575A/D576A) csm1 mutations. To generate Csm1 HD domain (H14A/D15A) and palm domain (D586A/D587A) mutants in the S. epidermidis module, the wildtype csm1 gene was first subcloned into Blunt II TOPO vector with the Zero Blunt II TOPO kit (Thermofisher) prior to inverse PCR mutagenesis and resubcloning (using BamHI and NdeI restriction enzymes) to replace the wildtype csm1 gene in the Csm module plasmid with the mutant csm1 genes. All DNA oligonucleotides were from Eurofins Genomics and sequences used for cloning can be found in Table S2 of the supplemental materials. Successful mutagenesis for all plasmids was confirmed via DNA sequencing (Eurofins Genomics).

Affinity purification of histidine-tagged Csm proteins

N-terminal 6x-histidine tagged Csm6 protein was expressed using pT7 transformed E. coli BL21-Star strain (DE3, Thermo Scientific) grown at 37°C with 50 μg/ml of kanamycin to an OD600 of 0.6. Protein expression was induced at room temperature overnight using Isopropyl β-D-1-thiogalactopyranoside (IPTG) at a final concentration of 0.5 mM for L. lactis Csm6 containing cells or 0.3 mM for S. epidermidis Csm6 containing cells. Cells were resuspended in lysis buffer (40 mM Tris-HCl (pH 7.5), 200 mM NaCl, 10 mM Imidazole) and lysed via sonication. Insoluble material was removed by centrifugation at 14,000 rpm for 30 minutes and filtered through a 0.8 μm syringe filter (Corning Incorporated) prior to purification of Csm6 proteins using gravity affinity chromatography and Ni-NTA resin (Thermo Scientific). The lysate was incubated with the resin (pre-equilibrated with lysis buffer) for 1 hr at 4°C then washed with the lysis buffer, wash buffer A (40 mM Tris-HCl (pH 7.5), 200 mM NaCl, 20 mM Imidazole), and wash buffer B (40 mM Tris-HCl (pH 7.5), 200 mM NaCl, 30 mM Imidazole). Csm6 was eluted using four elution buffers containing increasing amounts of imidazole (75, 150, 250, 500 mM). Eluted S. epidermidis Csm6 was dialyzed (40 mM Tris-HCl (pH 7.5), 200 mM NaCl) and purified a second time using gravity affinity chromatography in the same manner. Wildtype and mutant Csm6 proteins were purified under identical conditions. For purification of either the 6x histidine-tagged, L. lactis Csm6 or Csm3 proteins expressed from the L. lactis Csm module (Figure S1), proteins were affinity purified using a batch method that involved incubating the soluble lysate with Ni-NTA resin for 1 hour at 4°C prior to washing and elution with imidazole as described above. The purity of proteins was assessed using (10 % and 12.5% polyacrylamide) SDS-PAGE and Coomassie blue staining.

Preparation of RNA and DNA substrates

Synthetic RNAs were purchased from Integrated DNA Technologies and DNAs from Eurofins Genomics. The RNA and DNA sequences are given in Table S1. The RNA and DNA were 5ʹ end labeled as previously described [42]. Double-stranded oligonucleotides and circularized RNAs were generated as previously described [42].

Csm6 nuclease cleavage assay

To determine the nuclease activity of L. lactis Csm6 proteins against ssRNA, a 20 μl reaction consisting of 500 nM of Csm6, ~0.5–1.5 nM of radiolabeled RNA, 1x assay buffer (20 mM Tris-HCl (pH 7.5), 100 mM NaCl) was incubated at 30°C for 1 hour. In order to test substrate specificity for L. lactis Csm6, ssRNA was used in conjunction with dsRNA, ssDNA, dsDNA, and a dsDNA-RNA hybrid. The substrates were created as previously described [42]. Substrate specificity reactions were performed in a 20 μl reaction consisting of 500 nM of Csm6, ~0.5–1.5 nM of radiolabeled RNA or DNA, 10 mM EDTA, and 1x assay buffer and incubated at 30°C for 1 hour. Reactions were stopped by adding Gel Loading Buffer II (Life Technologies) and visualized by using denaturing 7M urea 15% polyacrylamide gels followed by autoradiography.

RNA cleavage product end group analysis

5ʹ and 3ʹ chemical end group analysis of Csm6 RNA cleavage products was conducted by using a combination of linear and circular substrates created as described above. First, wildtype L. lactis Csm6 was incubated with linear and circular RNA in standard cleavage assay conditions described above. PCI extraction and ethanol precipitation was then performed on the reactions and the resulting RNA was normalized based on cpm. Analysis of the 5ʹ end group was conducted by incubating linear and circular RNAs with 1 U Terminator 5ʹ-Phosphate-Dependent Exonuclease (TEX) (Epicentre), 2,500 cpm of RNA, and 1x terminator reaction buffer A (Epicentre). Reactions were incubated at 30°C for 1 hour. Analysis of the 3ʹ end group was conducted by treating linear RNA with 5 U E. coli poly (A) polymerase (NEB), 15,000 cpm of RNA, 1 mM ATP, and 1x PAP reaction buffer A (NEB). Reactions were incubated at 37°C for 15 minutes. Enzyme activity was halted by adding Gel Loading Buffer II (Life Technologies). The RNAs were visualized after electrophoresis on denaturing 7 M urea 15% polyacrylamide gels for TEX reactions and denaturing 8.3 M urea 20% polyacrylamide gels for PAP reactions followed by autoradiography.

Mapping Csm6 RNA cleavage sites

RNA1, RNA5, and RNA7 substrates (sequences given in Table S1) were incubated with Csm6 under reaction conditions supporting cleavage as described above with the following differences: 250 nM S. epidermidis Csm6 was used with 15 mM EDTA, 150 nM L. lactis Csm6 and 5 nM EDTA was used for RNA1 and RNA5 while 250 nM L. lactis Csm6 and 15 mM EDTA was used for RNA7. All reactions were incubated at 30°C and samples were removed at four time points over the course of 1 hour. Reactions were stopped by a combination of Gel Loading Buffer II (Life Technologies) and placing the reactions on ice. Three ladders were used for mapping purposes: a Decade Marker (Life Technologies), a partial alkaline hydrolysis ladder (Ambion), and an RNase T1 ladder (Ambion) as described [42]. Reactions and ladders were separated on denaturing 15% polyacrylamide sequencing gels and the RNAs were visualized by autoradiography.

Plasmid interference assay

Plasmid interference assays were performed similarly to previously described experiments [28]. Electrocompetent E. coli BL21-Ai (Thermo Scientific) strains containing a Csm module (pACYC; chloramphenicol resistance) were transformed with 100 ng of either target or non-target plasmid (pTrcHis vector; ampicillin resistance). Electroporation was performed using a Gene Pulser II Electroporation System (Bio-Rad); 950 μl of SOC medium was immediately added to the 50 μl of the electrocompetent host cells and shaken at 200 rpm for 90 mins at 37°C. A series of six sequential 1:10 dilutions were made for each transformation and 5 μl of each dilution was spotted onto LB agar plates containing 34 μg/ml of chloramphenicol, 50 μg/ml of ampicillin, and 0.2% arabinose. Arabinose was omitted from the LB agar plates for the L. lactis and S. epidermidis systems in experiments shown in Figure 6 since these systems were previously determined to induce plasmid resistance in the absence of arabinose induction of Csm module components [28]. Plates were incubated overnight at 37°C and images of the colonies were captured using white light in Gel Doc XR (Bio-Rad). Assays were done in triplicate.

Funding Statement

This work was supported by National Institutes of Health grants R35GM118160 (M.P.T.), R01GM54682 (M.P.T. and R.M.T.), and 1F31GM125365 (K.F.).

Acknowledgments

We thank members of the Terns laboratory and Claiborne Glover for helpful discussions.

Disclosure statement

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed here