Cas12a2 elicits abortive infection through RNA-triggered destruction of dsDNA

Abstract

Bacterial abortive-infection systems limit the spread of foreign invaders by shutting down or killing infected cells before the invaders can replicate^1,2. Several RNA-targeting CRISPR–Cas systems (that is, types III and VI) cause abortive-infection phenotypes by activating indiscriminate nucleases^3,4,5. However, a CRISPR-mediated abortive mechanism that leverages indiscriminate DNase activity of an RNA-guided single-effector nuclease has yet to be observed. Here we report that RNA targeting by the type V single-effector nuclease Cas12a2 drives abortive infection through non-specific cleavage of double-stranded DNA (dsDNA). After recognizing an RNA target with an activating protospacer-flanking sequence, Cas12a2 efficiently degrades single-stranded RNA (ssRNA), single-stranded DNA (ssDNA) and dsDNA. Within cells, the activation of Cas12a2 induces an SOS DNA-damage response and impairs growth, preventing the dissemination of the invader. Finally, we harnessed the collateral activity of Cas12a2 for direct RNA detection, demonstrating that Cas12a2 can be repurposed as an RNA-guided RNA-targeting tool. These findings expand the known defensive abilities of CRISPR–Cas systems and create additional opportunities for CRISPR technologies.

Main

All domains of life use defence strategies that cause cells to enter dormancy or die to limit the spread of infectious agents¹. In bacteria and archaea, this strategy is called abortive infection, and it is used by a vast variety of bacterial defence systems^1,2. Recently, it was shown that CRISPR RNA (crRNA)-guided adaptive immune systems that target RNA cause abortive-infection phenotypes^3,4,5,6. Type VI systems non-specifically degrade RNA, whereby the Cas13 single-effector nuclease acts as both a crRNA-guided effector and indiscriminate RNase^3,7,8. In type III systems, target RNA binding triggers the production of cyclic oligoadenylate secondary messengers that in turn activate indiscriminate accessory RNases and ssDNases that can drive abortive infection^4,5,9,10,11. Moreover, it has been proposed that abortive infection is mediated by indiscriminate dsDNases (such as NucC) activated through type III secondary messengers^12,13 or by indiscriminate ssDNase activity from type V Cas12a single-effector nucleases¹⁴. However, type III CRISPR-mediated dsDNase activity has yet to be examined in vivo, and the ssDNase activity of Cas12a was recently shown to not cause abortive infection¹⁵.

Here we report that Cas12a2, a type V single-effector CRISPR-associated (Cas) nuclease, induces an abortive-infection phenotype when challenged with plasmids that are complementary to crRNA guides. Biochemical assays using recombinant protein revealed that Cas12a2 recognizes RNA targets, unleashing non-specific dsDNA-, ssDNA- and ssRNA-nuclease activities distinct from those of other single-subunit RNA-targeting (such as Cas13a) and dsDNA-targeting (such as Cas12a) Cas nucleases^8,16,17. Furthermore, we show that the Cas12a2 non-specific nuclease activities damage bacterial DNA, triggering the SOS response and impairing cell growth. Collectively these results suggest that the dsDNase activity of Cas12a2 is instrumental in triggering the abortive-infection phenotype. As a proof-of-principle demonstration, we show that Cas12a2 can detect RNA at a sensitivity that is comparable to that of the RNA-targeting Cas13a nuclease at various temperatures.

Cas12a2 induces abortive infection

Cas12a2 comprises a group of type V effector nucleases that are related to Cas12a¹⁶, with Cas12a2 orthologues previously being classified as Cas12a variants¹⁸. Our analyses similarly place them in a monophyletic clade that shares the last common ancestor with Cas12a nucleases (Fig. 1a and Extended Data Fig. 1). Further analysis revealed that CRISPR–Cas12a2 and CRISPR–Cas12a systems feature CRISPR repeats with a conserved 3′ end, and the nucleases possess homologous RuvC endonuclease domains and a similar predicted secondary structure in the N termini (Fig. 1b and Supplementary Fig. 2). Despite the conserved RuvC-like domains and N termini, Cas12a2 is distinguished from Cas12a by the presence of a large domain of unknown function located in place of the Cas12a bridge helix as well as a zinc-finger domain in place of the Cas12a Nuc domain (Fig. 1b and Supplementary Fig. 2). Considering their original classification combined with our phylogenetic analyses as well as recent structural results¹⁹, we named these distinct type V nucleases Cas12a2.

Notably, some CRISPR–Cas systems contain both cas12a2 and cas12a genes in tandem next to a shared CRISPR array (Fig. 1c). From this observation and the conservation of CRISPR repeats from systems with either of the nucleases (Fig. 1d and Supplementary Fig. 3), we hypothesized that both proteins bind to and process similar crRNA guides. However, as the proteins diverge in other domains, we further hypothesized that Cas12a2 performs a defence function that is distinct from the dsDNA-targeting activity of Cas12a¹⁶.

To test these hypotheses, we encoded the cas12a2 gene from the sulfur-oxidizing epsilonproteobacterium Sulfuricurvum sp. PC08-66 (SuCas12a2) along with a CRISPR array into an expression plasmid, which we introduced into Escherichia coli cells. We next performed a traditional plasmid interference assay that depletes cells by selecting for the plasmid containing the nuclease and crRNA as well as a target plasmid (Fig. 1e). This assay detects broad immune system activity but cannot distinguish between defence activities that only deplete the target from those that activate abortive-infection phenotypes. To test whether Cas12a2 uses an abortive-infection mechanism, we modified the assay by selecting only for the nuclease plasmid (Fig. 1e). Consistent with our hypothesis that Cas12a2 functions differently compared with Cas12a, Cas12a2 depleted cells in both the traditional (about 1,900-fold reduction) and modified (around 1,300-fold reduction) plasmid interference assays, whereas Cas12a from Lachnospiraceae bacterium (LbCas12a) depleted cells only in the traditional assay (Fig. 1f). Similar trends were observed with different targets cloned into the same plasmid location, different Cas12a2 homologues and when comparing SuCas12a2 with the Cas12a homologue from Prevotella bryantii B14 (Pb2Cas12a) (Extended Data Fig. 2a,b). Moreover, mutating predicted active residues within any of the three RuvC motifs in SuCas12a2 impaired immune function (Fig. 1g). Collectively, these results indicate that Cas12a2 relies on a RuvC-nuclease domain and induces abortive infection through a mechanism that is distinct from that of Cas12a.

Cas12a2 targets RNA and degrades dsDNA

CRISPR systems that cause abortive-infection phenotypes (such as types III and VI) rely on indiscriminate nucleases activated by RNA targeting^3,4,5. To determine whether Cas12a2 uses a similar mechanism, we recombinantly expressed and purified SuCas12a2 and tested its enzymatic activities in vitro (Fig. 2 and Supplementary Fig. 4). However, before examining the nucleic-acid-targeting activities, we needed to determine how Cas12a2 crRNAs are processed.

The CRISPR repeats of the Cas12a and Cas12a2 systems are highly conserved at the 3′ end (Fig. 1d and Supplementary Fig. 3), and sequence alignments predict that Cas12a2 shares secondary structure in the region of the Cas12a pre-crRNA-processing active site^18,20 (Supplementary Fig. 5). Consistent with this prediction, RNA-sequencing analysis of pre-crRNAs processed by SuCas12a2 in vitro revealed that processing occurs one nucleotide downstream of the position cleaved by Cas12a (Extended Data Fig. 3a,b). The 3′ end of the spacer also underwent trimming in vivo to form an approximately 24-nucleotide guide (Extended Data Fig. 3b,c), possibly through host ribonucleases as observed for Cas9 crRNAs²¹. Mutating basic amino acids (Lys784 and Arg785) located in the predicted RNA-processing active site abolished activity²² (Extended Data Fig. 3d). Furthermore, plasmid interference assays revealed that Cas12a and Cas12a2 can interchange guides without impairing immunity (Extended Data Fig. 2c). Thus, the Cas12a2 nuclease processes its own crRNA guides like other type V effector nucleases^20,23 and can share crRNAs with Cas12a.

To determine the nucleic-acid target preference of crRNA-guided Cas12a2, complementary ssRNA, ssDNA and dsDNA substrates containing an A/T-rich flanking sequence (paralleling Cas12a substrates)^16,22 were fluorescently labelled with a FAM molecule and combined with crRNA-guided Cas12a2 (Fig. 2a). Similar to CRISPR–Cas systems that cause abortive infection—yet in contrast to the dsDNA-targeting Cas12a—Cas12a2 is activated only in the presence of complementary RNA targets. The potency of plasmid interference with SuCas12a2 (Fig. 1f) was notable given the lack of a defined promoter upstream of the target in this construct. However, we attribute the interference to spurious transcription of the encoding plasmid for two reasons: introducing an upstream terminator significantly reduced plasmid interference in E. coli (Extended Data Fig. 2d,e), and an upstream promoter was required to detect collateral activity in a cell-free transcription–translation assay²⁴ (Extended Data Fig. 2f,g).

As other Cas abortive-infection mechanisms rely on collateral indiscriminate RNase activity, we examined whether specific RNA targeting by Cas12a2 induces indiscriminate nuclease activity. We found that SuCas12a2 robustly and indiscriminately degraded FAM-labelled ssRNA, ssDNA and dsDNA substrates bearing no complementarity to the crRNA guide. By contrast, other Cas nucleases indiscriminately degrade only ssRNA (Cas13a)⁸ or ssRNA and ssDNA (Cas12g)²⁵ after RNA targeting, or only ssDNA after dsDNA targeting (Cas12a)¹⁴ (Fig. 2b and Extended Data Fig. 4a). Of the three collateral substrates, ssRNA and ssDNA appear to be more efficiently cleaved than dsDNA by Cas12a2 (Fig. 2c and Extended Data Fig. 4b). However, this difference could be explained by the presence of twice as many DNA strands in dsDNA substrates compared with ssDNA substrates for the same amount of nuclease. Also, similar to Cas13a⁸, complementary ssDNA and dsDNA do not activate any Cas12a2 non-specific nuclease activity (Extended Data Fig. 4a), and dsRNA is not a primary substrate of collateral cleavage (Extended Data Fig. 4c).

To examine whether Cas12a2 activity is reliant on detecting a ‘non-self’ signal adjacent to the target (called a protospacer-flanking sequence (PFS))⁸, we performed in vitro cleavage assays in which target RNA sequences were flanked on the 3′ side with a ‘self’ sequence complementary to the crRNA repeat (5′-AUCUA-3′), the non-self PFS used in our in vivo assay (5′-GAAAG-3′) or a ‘flankless’ RNA complementary to the guide region of the crRNA, but containing no PFS (Fig. 2b). Notably, only the RNA target containing the non-self PFS activated collateral nuclease activity, demonstrating that specific nucleotides at the 3′ end of the RNA target must be present to activate the collateral activity of Cas12a2. Moreover, introducing disruptive mutations to any of the three RuvC motifs or conserved cysteine residues within the putative zinc-finger domain abolished all non-specific cleavage (Fig. 2d and Extended Data Fig. 4d), consistent with our in vivo plasmid interference results (Fig. 1g).

Our biochemical assays demonstrated that Cas12a2 could quickly remove a FAM label from linear dsDNA substrates, but it was unclear whether Cas12a2 degrades DNA lacking available 5′ or 3′ ends. We therefore challenged crRNA-guided Cas12a2 with an RNA target and a supercoiled pUC19 plasmid. Importantly, pUC19 does not contain any sequence complementary to the Cas12a2 crRNA guide. We observed that SuCas12a2 rapidly nicked, linearized and degraded pUC19 DNA (Fig. 2e), but only in the presence of a cognate target and PFS and with an intact RuvC domain (Extended Data Fig. 4e). This rapid destruction of the supercoiled plasmid contrasts with the slow and incomplete linearization of plasmid DNA by Cas12a nucleases²⁶. These data suggest a mechanism in which activated SuCas12a2 is able to robustly hydrolyse the phosphodiester backbone of non-specific DNA regardless of whether it is supercoiled, nicked or linear. A comparison with Cas12a (dsDNA targeting with collateral ssDNase), Cas13a (ssRNA targeting with collateral ssRNase) and Cas13g (ssRNA targeting with collateral ssRNase and ssDNase) demonstrated that the RNA-targeting ssRNase, ssDNase and dsDNase are unique to SuCas12a2 (Extended Data Fig. 4a). Collectively, these in vitro results reveal that crRNAs guide SuCas12a2 to RNA targets, activating RuvC-dependent cleavage of ssRNA, ssDNA and dsDNA. These activities, in part or in total, may underlie the abortive-infection phenotype.

Cas12a2 exhibits targeting flexibility

Although our in vitro data indicated an underlying mechanism for the Cas12a2 abortive-infection phenotype, we wanted to understand the targeting limitations of these distinct enzymes. In particular, we investigated the stringency of non-self PFS sequence recognition and penalties for mismatches between the crRNA and target. We therefore challenged SuCas12a2 with a library of plasmids encoding all possible 1,024 flanking sequences at the 3′ end of the RNA target to the −5 position (Fig. 3a and Extended Data Fig. 5a,b). We found that SuCas12a2 depleted approximately half of all of the sequences in the library, suggesting a PFS-recognition mechanism that is more stringent than that of Cas13 but still more promiscuous than those of most DNA-targeting systems^8,27. The depleted sequences were generally A rich, consistent with a 5′-GAAAG-3′ PFS, but could not be fully captured by a single consensus motif (Fig. 3a and Extended Data Fig. 5c). We further validated individual depleted sequences, including representatives within five unique motifs recognized by SuCas12a2 but not by Pb2Cas12a—a nuclease that is known for flexible PAM recognition²⁴ (Fig. 3b and Extended Data Fig. 5d). Consistent with its function as an RNA-targeting nuclease, the recognized sequences were broad but did not follow the expected profile if Cas12a2 is principally evaluating tag–anti-tag complementarity. These results further support a mechanism in which PFS recognition by SuCas12a2 operates similar to type III systems that require recognition of a PFS or RNA PAM to activate^28,29,30,31 and distinct from the evaluation of tag–anti-tag complementarity used by RNA-targeting Cas13^8,32 and other type III CRISPR–Cas systems³³….