Abstract

Reconstitution of the L. shahii C2c2 locus in Escherichia coli confers RNA-guided immunity

We explored whether LshC2c2 could confer immunity to MS2 (25), a lytic single-stranded (ss) RNA phage, without DNA intermediates in its life cycle, that infects E. coli. We constructed a low-copy plasmid carrying the entire LshC2c2 locus (pLshC2c2) to allow for heterologous reconstitution in E. coli (fig. S1A). Because expressed mature crRNAs from the LshC2c2 locus have a maximum spacer length of 28nt (fig. S1A) (19), we tiled all possible 28-nt target sites in the MS2 phage genome (Fig. 1A). This resulted in a library of 3,473 spacer sequences (along with 490 non-targeting guides designed to have a Levenshtein distance of ≥8 with respect to the MS2 and E. coli genomes) which we inserted between pLshC2c2 direct repeats (DRs). After transformation in of this construct into E. coli, we infected cells with varying dilutions of MS2 (10⁻¹, 10⁻³, and 10⁻⁵) and analyzed surviving cells to determine the spacer sequences carried by cells that survived the infection. Cells carrying spacers that confer robust interference against MS2 are expected to proliferate faster than those that lack such sequences. Following growth for 16 hours, we identified a number of spacers that were consistently enriched across three independent infection replicas in both the 10⁻¹ and 10⁻³ dilution conditions, suggesting that they enabled interference against MS2. Specifically, 147 and 150 spacers showed >1.25 log₂-fold enrichment in all three replicates for the 10⁻¹ and 10⁻³ phage dilutions, respectively; of these two groups of top enriched spacers, 84 are shared (Figs. 1B, S2A–G, table S1). Additionally, no non-targeting guides were found to be consistently enriched among the three 10⁻¹, 10⁻³, or 10⁻⁵ phage replicates (fig. S2D, G). We also analyzed the flanking regions of protospacers on the MS2 genome corresponding to the enriched spacers and found that spacers with a G immediately flanking the 3’ end of the protospacer were less fit relative to all other nucleotides at this position (i.e. A, U, or C), suggesting that the 3′ protospacer flanking site (PFS) affects the efficacy of C2c2-mediated targeting (Figs. 1C, S2E–F, S3). Although the PFS is adjacent to the protospacer target, we chose not to use the commonly used protospacer adjacent motif (PAM) nomenclature as it has come to connote a sequence used in self vs. non-self differentiation (38), which is irrelevant in a RNA-targeting system. It is worth noting that the avoidance of G by C2c2 echo the absence of PAMs applicable to other RNA-targeting CRISPR systems and effector proteins (20, 22, 24, 25, 39, 40).

The fact that only ~5% of crRNAs are enriched may reflect other factors influencing interference activity, such as accessibility of the target site that might be affected by RNA binding proteins or secondary structure. In agreement with this hypothesis, the enriched spacers tend to cluster into regions of strong interference where they are closer to each other than one would expect by random chance (fig. S3F–G).

To validate the interference activity of the enriched spacers, we individually cloned four top-enriched spacers into pLshC2c2 CRISPR arrays and observed a 3- to 4-log₁₀ reduction in plaque formation, consistent with the level of enrichment observed in the screen (Figs 1B, S4). We cloned sixteen guides targeting distinct regions of the MS2 mat gene (4 guides per possible single-nucleotide PFS). All 16 crRNAs mediated MS2 interference, although higher levels of resistance were observed for the C, A, and U PFS-targeting guides (Figs. 1D, 1E, S5), indicating that C2c2 can be effectively retargeted in a crRNA-dependent fashion to sites within the MS2 genome.

To further validate the observed PFS preference with an alternate approach, we designed a protospacer site in the pUC19 plasmid at the 5’ end of the β-lactamase mRNA, which encodes ampicillin resistance in E. coli, flanked by five randomized nucleotides at the 3’ end. Significant depletion and enrichment was observed for the LshC2c2 locus (****, p<0.0001) compared to the pACYC184 controls (Fig. S6A). Analysis of the depleted PFS sequences confirmed the presence of a PFS preference of H (Fig. S6B).

C2c2 is a single-effector endoRNase mediating ssRNA cleavage with a single crRNA guide

We purified the LshC2c2 protein (fig. S7) and assayed its ability to cleave an in vitro transcribed 173-nt ssRNA target (Figs. 2A, S8) containing a C PFS (ssRNA target 1 with protospacer 14). Mature LshC2c2 crRNAs contain a 28-nt direct repeat (DR) and a 28 nt spacer (fig. S1A) (19). We therefore generated an in-vitro-transcribed crRNA with a 28-nt spacer complementary to protospacer 14 on ssRNA target 1. LshC2c2 efficiently cleaved ssRNA in a Mg²⁺- and crRNA-dependent manner (Figs. 2B, S9). We then annealed complementary RNA oligos to regions flanking the crRNA target site. This partially double-stranded RNA substrate was not cleaved by LshC2c2, suggesting it is specific for ssRNA (figs. S10A–B).

We tested the sequence constraints of RNA cleavage by LshC2c2 with additional crRNAs complementary to ssRNA target 1 where protospacer 14 is preceded by each PFS variant. The results of this experiment confirmed the preference for C, A, and U PFSs, with little cleavage activity detected for the G PFS target (Fig. 2C). Additionally, we designed 5 crRNAs for each possible PFS (20 total) across the ssRNA target 1 and evaluated cleavage activity for LshC2c2 paired with each of these crRNAs. As expected, we observed less cleavage activity for G PFS-targeting crRNAs compared to other crRNAs tested (Fig. 2D).

We then generated a dsDNA plasmid library with protospacer 14 flanked by 7 random nucleotides to account for any PFS preference. When incubated with LshC2c2 protein and a crRNA complementary to protospacer 14, no cleavage of the dsDNA plasmid library was observed (fig. S10C). We also did not observe cleavage when targeting a ssDNA version of ssRNA target 1 (fig. S10D). To rule out co-transcriptional DNA cleavage, which has been observed in type III CRISPR-Cas systems (22), we recapitulated the E. coli RNA polymerase co-transcriptional cleavage assay (22) (fig. S11A) expressing ssRNA target 1 from a DNA substrate. This assay of purified LshC2c2 and crRNA targeting ssRNA target 1 did not show any DNA cleavage (fig. S11B). Together, these results indicate that C2c2 cleaves specific ssRNA sites directed by the target complementarity encoded in the crRNA, with a H PFS preference.

C2c2 cleavage depends on local target sequence and secondary structure

Given that C2c2 did not efficiently cleave dsRNA substrates and that ssRNA can form complex secondary structures, we reasoned that cleavage by C2c2 might be affected by secondary structure of the ssRNA target. Indeed, after tiling ssRNA target 1 with different crRNAs (Fig. 2D), we observed the same cleavage pattern regardless of the crRNA position along the target RNA. This observation suggests that the crRNA-dependent cleavage pattern was determined by features of the target sequence rather than the distance from the binding site. We hypothesized that the LshC2c2-crRNA complex binds the target and cleaves exposed regions of ssRNA within the secondary structure elements, with potential preference for certain nucleotides.

In agreement with this hypothesis, cleavage of three ssRNA targets with different sequences flanking identical 28-nt protospacers resulted in three distinct patterns of cleavage (Fig. 3A). RNA-sequencing of the cleavage products for the three targets revealed that cleavage sites mainly localized to uracil-rich regions of ssRNA or ssRNA-dsRNA junctions within the in silico predicted co-folds of the target sequence with the crRNA (Figs. 3B–C, S12A–D). To test whether the LshC2c2-crRNA complex prefers cleavage at uracils, we analyzed the cleavage efficiencies of homopolymeric RNA targets (a 28-nt protospacer extended with 120 As or Us regularly interspaced by single bases of G or C to enable oligo synthesis) and found that LshC2c2 preferentially cleaved the uracil target compared to adenine (figs. S12E, S12F). We then tested cleavage of a modified version of ssRNA 4 which had its main site of cleavage, a loop, replaced with each of the four possible homopolymers and found that cleavage only occurred at the uracil homopolymer loop (fig. S12G). To further test whether cleavage was occurring at uracil residues, we mutated single uracil residues in ssRNA 1 that showed cleavage in the RNA-sequencing (Fig. 3B) to adenines. This experiment showed that, by mutating each uracil residue, we could modulate the presence of a single cleavage band, consistent with LshC2c2 cleaving at uracil residues in ssRNA regions (Fig. 3D).

The HEPN domains of C2c2 mediate RNA-guided ssRNA-cleavage

Bioinformatic analysis of C2c2 has suggested that the HEPN domains are likely to be responsible for the observed catalytic activity (19). Each of the two HEPN domains of C2c2 contains a dyad of conserved arginine and histidine residues (Fig. 4A), in agreement with the catalytic mechanism of the HEPN endoRNAse (26–28). We mutated each of these putative catalytic residues separately to alanine (R597A, H602A, R1278A, H1283A) in the LshC2c2 locus plasmids and assayed for MS2 interference. None of the four mutant plasmids were able to protect E. coli from phage infection (Figs. 4B, S13).

We purified the four single-point mutant proteins and assayed their ability to cleave 5’-end-labeled ssRNA target 1 (Fig. 4C). In agreement with our in vivo results, all four mutations abolished cleavage activity. The inability of either of the two wild-type HEPN domains to compensate for inactivation of the other implies cooperation between the two domains. These results agree with observations that several bacterial and eukaryotic single-HEPN proteins function as dimers (27, 28, 41).

Catalytically inactive variants of Cas9 retain target DNA binding, allowing for the creation of programmable DNA-binding proteins (12, 13). Electrophoretic mobility shift assays (EMSA) on both the wild-type (Fig. 4D) and R1278A mutant LshC2c2 (Fig. 4E) in complex with crRNA showed the wild-type LshC2c2 complex binding strongly (K_D ~ 46 nM, fig. S14A) and specifically to 5’-end-labeled ssRNA target 10 but not to the 5’-end-labeled non-target ssRNA (the reverse complement of ssRNA target 10). The R1278A mutant C2c2 complex showed even stronger (K_D ~ 7 nM, fig. S14B) specific binding, indicating that this HEPN mutation results in a catalytically inactive, RNA-programmable RNA-binding protein. The LshC2c2 protein or crRNA alone showed reduced levels of target affinity, as expected (fig. S14C–E). Additionally, no specific binding of LshC2c2-crRNA complex to ssDNA was observed (fig. S15).

These results demonstrate that C2c2 cleaves RNA via a catalytic mechanism distinct from other known CRISPR-associated RNases. In particular, the type III Csm and Cmr multiprotein complexes rely on acidic residues of RRM domains for catalysis, whereas C2c2 achieves RNA cleavage through the conserved basic residues of its two HEPN domains.

Sequence and structural requirements of C2c2 crRNA

Similar to the type V-B (Cpf1) systems (18), the LshC2c2 crRNA contains a single stem loop in the direct repeat (DR), suggesting that the secondary structure of the crRNA could facilitate interaction with LshC2c2. We thus investigated the length requirements of the spacer sequence for ssRNA cleavage and found that LshC2c2 requires spacers of at least 22 nt length to efficiently cleave ssRNA target 1 (fig. S16A). The stem-loop structure of the crRNA is also critical for ssRNA cleavage, because DR truncations that disturbed the stem loop abrogated target cleavage (fig. S16B). Thus, a DR longer than 24 nt is required to maintain the stem loop necessary for LshC2c2 to mediate ssRNA cleavage.

Single base pair inversions in the stem that preserved the stem structure did not affect the activity of the LshC2c2 complex. In contrast, inverting all four G-C pairs in the stem eliminated the cleavage despite maintaining the duplex structure (fig. S17A). Other perturbations, such as those that introduced kinks and reduced or increased base-pairing in the stem, also eliminated or drastically suppressed cleavage. This suggests that the crRNA stem length is important for complex formation and activity (fig. S17A). We also found that loop deletions eliminated cleavage, whereas insertions and substitutions mostly maintained some level of cleavage activity (fig. S17B). In contrast, nearly all substitutions or deletions in the region 3’ to the DR prevented cleavage by LshC2c2 (fig S18). Together, these results demonstrate that LshC2c2 recognizes structural characteristics of its cognate crRNA but is amenable to loop insertions and most tested base substitutions outside of the 3’ DR region. These results have implications for the future application of C2c2-based tools that require guide engineering for recruitment of effectors or modulation of activity (42–44).

C2c2 cleavage is sensitive to double mismatches in the crRNA-target duplex

We tested the sensitivity of the LshC2c2 system to single mismatches between the crRNA guide and target RNA by mutating single bases across the spacer to the respective complementary bases (e.g., A to U). We then quantified plaque formation with these mismatched spacers in the MS2 infection assay and found that C2c2 was fully tolerant to single mismatches across the spacer as such mismatched spacers interfered with phage propagation with similar efficiency as fully matched spacers (figs. S19A, S20). However, when we introduced consecutive double substitutions in the spacer, we found a ~3 log₁₀-fold reduction in the protection for mismatches in the center, but not at the 5’- or 3’-end, of the crRNA (figs. 19B, S20). This observation suggests the presence of a mismatch-sensitive “seed region” in the center of the crRNA-target duplex.

We generated a set of in vitro transcribed crRNAs with mismatches similarly positioned across the spacer region. When incubated with LshC2c2 protein, all single mismatched crRNA supported cleavage (Fig. S19C), in agreement with our in vivo findings. When tested with a set of consecutive and non-consecutive double mutant crRNAs, LshC2c2 was unable to cleave the target RNA if the mismatches were positioned in the center, but not at the 5’- or 3’-end of the crRNA (Fig. S19D, S21A), further supporting the existence of a central seed region. Additionally, no cleavage activity was observed with crRNAs containing consecutive triple mismatches in the seed region (fig. S21B).

C2c2 can be reprogrammed to mediate specific mRNA knockdown in vivo

Given the ability of C2c2 to cleave target ssRNA in a crRNA sequence-specific manner, we tested whether LshC2c2 could be reprogrammed to degrade selected non-phage ssRNA targets, and particularly mRNAs, in vivo. We co-transformed E. coli with a plasmid encoding LshC2c2 and a crRNA targeting the mRNA of red fluorescent protein (RFP) as well as a compatible plasmid expressing RFP (Fig. 5A). For OD-matched samples, we observed an approximately 20% to 92% decrease in RFP positive cells for crRNAs targeting protospacers flanked by C, A, or U PFSs (Fig. 5B, C). As a control, we tested crRNAs containing reverse complements (targeting the dsDNA plasmid) of the top performing RFP mRNA-targeting spacers. As expected, we observed no decrease in RFP fluorescence by these crRNAs (Fig. 5B). We also confirmed that mutation of the catalytic arginine residues in either HEPN domain to alanine precluded RFP knockdown (fig. S22). Thus, C2c2 is capable of general retargeting to arbitrary ssRNA substrates, governed exclusively by predictable nucleic-acid interactions.

When we examined the growth of cells carrying the RFP-targeting spacer with the greatest level of RFP knockdown, we noted that the growth rate of these bacteria was substantially reduced (Fig. 5A, spacer 7). We investigated whether the effect on growth was mediated by the RFP mRNA-targeting activity of LshC2c2 by introducing an inducible-RFP plasmid and an RFP-targeting LshC2c2 locus into E. coli. Upon induction of RFP transcription, cells with RFP knockdown showed substantial growth suppression, not observed in non-targeting controls (Fig. 5D, E). This growth restriction was dependent on the level of the RFP mRNA, as controlled by the concentration of the inducer anhydrotetracycline. In contrast, in the absence of RFP transcription, we did not observe any growth restriction nor did we observe any transcription-dependent DNA targeting in our biochemical experiment (fig. S11). These results indicate that RNA-targeting is likely the primary driver of this growth restriction phenotype. We therefore surmised that, in addition to the cleavage of the target RNA, C2c2 CRISPR systems might prevent virus reproduction also via non-specific cleavage of cellular mRNAs, causing programmed cell death (PCD) or dormancy (45, 46).

C2c2 cleaves collateral RNA in addition to crRNA-targeted ssRNA

Cas9 and Cpf1 cleave DNA within the crRNA-target heteroduplex at defined positions, reverting to an inactive state after cleavage. In contrast, C2c2 cleaves the target RNA outside of the crRNA binding site at varying distances depending on the flanking sequence, presumably within exposed ssRNA loop regions (Figs. 3B, 3C, S12A–D). This observed flexibility with respect to the cleavage distance led us to test whether cleavage of other, non-target ssRNAs also occurs upon C2c2 target binding and activation. Under this model, the C2c2-crRNA complex, once activated by binding to its target RNA, cleaves the target RNA as well as other RNAs non-specifically. We carried out in vitro cleavage reactions that included, in addition to LshC2c2 protein, crRNA and its target RNA, one of four unrelated RNA molecules without any complementarity to the crRNA guide (Fig. 6A). These experiments showed that, whereas the LshC2c2-crRNA complex did not mediate cleavage of any of the four collateral RNAs in the absence of the target RNA, all four were efficiently degraded in the presence of the target RNA (Figs. 6B, S23A). Furthermore, R597A and R1278A HEPN mutants were unable to cleave collateral RNA (Fig. S23B).

To further investigate the collateral cleavage and growth restriction in vivo, we hypothesized that if a PFS preference screen for LshC2c2 was performed in a transcribed region on the transformed plasmid, then we should be able to detect the PFS preference due to growth restriction induced by RNA targeting. We designed a protospacer site flanked by five randomized nucleotides at the 3’ end in either a non-transcribed region or in a region transcribed from the lac promoter (fig. S24A). The analysis of the depleted and enriched PFS sequences identified a H PFS only in the assay with the transcribed sequence but no discernable motif in the non-transcribed sequence (fig. S24B–C).

These results suggest a HEPN-dependent mechanism whereby C2c2 in a complex with crRNA is activated upon binding to target RNA and subsequently cleaves non-specifically other available ssRNA targets. Such promiscuous RNA cleavage could cause cellular toxicity, resulting in the observed growth rate inhibition. These findings imply that, in addition to their likely role in direct suppression of RNA viruses, type VI CRISPR-Cas systems could function as mediators of a distinct variety of PCD or dormancy induction that is specifically triggered by cognate invader genomes (Fig. 7). Under this scenario, dormancy would slow the infection and supply additional time for adaptive immunity. Such a mechanism agrees with the previously proposed coupling of adaptive immunity and PCD during the CRISPR-Cas defensive response (47).

Conclusions

In summary, the Class 2 type VI effector protein C2c2 is an RNA-guided RNase that can be efficiently programmed to degrade any ssRNA by specifying a 28-nt sequence on the crRNA (Fig. 10). C2c2 cleaves RNA through conserved basic residues within its two HEPN domains, in contrast to the catalytic mechanisms of other known RNases found in CRISPR-Cas systems (25, 48). Alanine substitution of any of the four predicted HEPN domain catalytic residues converted C2c2 into an inactive programmable RNA-binding protein (dC2c2, analogous to dCas9). Many different spacer sequences work well in our assays although further screening will likely define properties and rules governing optimal function.

These results suggest a broad range of biotechnology applications and research questions (49–51). For example, the ability of dC2c2 to bind to specified sequences could be used to (i) bring effector modules to specific transcripts to modulate their function or translation, which could be used for large-scale screening, construction of synthetic regulatory circuits and other purposes; (ii) fluorescently tag specific RNAs to visualize their trafficking and/or localization; (iii) alter RNA localization through domains with affinity for specific subcellular compartments; and (iv) capture specific transcripts (through direct pull down of dC2c2) to enrich for proximal molecular partners, including RNAs and proteins.

Active C2c2 also has many potential applications such as targeting a specific transcript for destruction, as performed here with RFP. In addition, C2c2, once primed by the cognate target, can cleave other (non-complementary) RNA molecules in vitro and inhibit cell growth in vivo. Biologically, this promiscuous RNase activity might reflect a PCD/dormancy-based protection mechanism of the type VI CRISPR-Cas systems (Fig. 7). Technologically, it might be used to trigger PCD or dormancy in specific cells such as cancer cells expressing a particular transcript, neurons of a given class, or cells infected by a specific pathogen.

Further experimental study is required to elucidate the mechanisms by which the C2c2 system acquires spacers and the classes of pathogens against which it protects bacteria. The presence of the conserved CRISPR adaptation module consisting of typical Cas1 and Cas2 proteins in the LshC2c2 locus suggests that it is capable of spacer acquisition. Although C2c2 systems lack reverse transcriptases, which mediate acquisition of RNA spacers in some type III systems (52), it is possible that additional host or viral factors could support RNA spacer acquisition. Additionally or alternatively, type VI systems could acquire DNA spacers similar to other CRISPR-Cas variants but then target transcripts of the respective DNA genomes, eliciting PCD and abortive infection (Fig. 7).

The CRISPR-C2c2 system represent a distinct evolutionary path among Class 2 CRISPR-Cas systems. It is likely that other, broadly analogous Class 2 RNA-targeting immune systems exist, and further characterization of the diverse members of Class 2 systems will provide a deeper understanding of bacterial immunity and provide a rich starting point for the development of programmable molecular tools for in vivo RNA manipulation.

Materials and Methods

Expanded materials and methods, including computational analysis, can be found in supplementary materials and methods.

Bacterial phage interference

The C2c2 CRISPR locus was amplified from DNA from Leptotrichia shahii DSM 19757 (ATCC, Manassas, VA) and cloned for heterologous expression in E. coli. For screens, a library of all possible spacers targeting the MS2 genome were cloned into the spacer array; for individual spacers, single specific spacers were cloned into the array. Interference screens were performed in liquid culture and plated; surviving colonies were harvested for DNA and spacer representation was determined by next-generation sequencing. Individual spacers were tested by spotting on top agar.

β–lactamase and transcribed/non-transcribed PFS preference screens

Sequences with randomized nucleotides adjacent to protospacer 1 were cloned into pUC19 in corresponding regions. Libraries were screened by co-transformation with LshC2c2 locus plasmid or pACYC184 plasmid control, harvesting of the surviving colonies, and next-generation sequencing of the resulting regions.

RFP targeting assay

Cells containing an RFP expressing plasmid were transformed with an LshC2c2 locus plasmid with corresponding spacers, grown overnight, and analyzed for RFP fluorescence by flow cytometry. The growth effects of LshC2c2 activity were quantified by titrating inducible RFP levels with dilutions of anhydrotetracycline inducer and then measuring OD₆₀₀.

in vitro nuclease and electrophoretic mobility shift assays

LshC2c2 protein and HEPN mutants were purified for use in in vitro reactions; RNA were synthesized via in vitro transcription. For nuclease assays, protein was co-incubated with crRNA and either 3’ or 5’-labeled targets and analyzed via denaturing gel electrophoresis and imaging or by next-generation sequencing. For electrophoretic mobility shift assays, protein and nucleic acid were co-incubated and then resolved by gel electrophoresis and imaging.

Supplementary Material

Acknowledgments

Footnotes

References and Notes