Mature clustered, regularly interspaced, short palindromic repeats RNA (crRNA) length is measured by a ruler mechanism anchored at the precursor processing site

Upstream Repeat Elements Direct crRNA Primary Processing.

S. epidermidis RP62a (21) contains a Type III CRISPR/Cas system with four direct repeats, three spacers, and nine cas genes (Fig. 1A and Fig. S1). crRNAs derived from the first spacer (spc1 crRNAs) are generated by cleavage of the precursor within the first and second direct repeat (DR) sequences at the base of a putative stem loop structure (7) (Fig. S2A). Spc1 crRNAs mediate CRISPR interference against conjugation by targeting a region of the nickase gene present in staphylococcal conjugative plasmids (7). We studied crRNA processing in vivo in S. epidermidis Δcrispr (7) by introducing plasmids carrying different mutations of the CRISPR array. We used primer extension analysis and conjugation assays to measure primary processing and CRISPR function, respectively (details about mutations and conjugation efficiencies of all constructs used in this study can be found in Tables S1 and S2, respectively).

To define the repeat sequences required for CRISPR function, we generated constructs carrying different deletions within repeats flanking spc1 (Fig. 1 and Fig. S1). These deletions showed that primary processing requires upstream, but not downstream, repeat sequences. Deletions of the 3′ half of DR1 and DR2 [mutants DR1 (3′del) and DR2 (3′del)] eliminated processing of spc1 and spc2 crRNAs, respectively (Fig. 1 B and C and Fig. S1), but not of others, indicating that primary processing of each repeat spacer unit is independent. Deletions of the 5′ half of DR1 and DR2 [mutants DR1 (5′del) and DR2 (5′del)] allowed processing but at a reduced efficiency that is not sufficient to provide CRISPR interference (Fig. 1 B–D and Fig. S1). We attribute this diminished processing to the elimination of the stem loop structure in the crRNA precursors derived from these constructs (see below). A construct containing only DR1-spc1 sequences [DR2–4 (del)] produced WT levels of crRNA and complemented the CRISPR deficiency of the Δcrispr strain, indicating that this unit is the minimal unit required to support CRISPR interference (additional deletions in DR1 abrogated processing and function in this minimal construct) (Fig. S1 and Tables S1 and S2).

To determine the precise sequence requirements for primary processing, we mutated each of the 36 nt of DR1 to their Watson–Crick complement (Fig. 1 E and F). We found that mutations of the bases surrounding the cleavage site (nucleotides 26–31, GGGACG) eliminated processing and accumulation of spc1 crRNA as well as interference against the conjugative plasmid pG0400. In the crRNA precursor, bases G25–G28 would pair with C13–C16 in the putative stem loop structure of S. epidermidis repeats (Fig. S2A). Interestingly, mutation of C13 resulted in lower levels of spc1 crRNA that do not support interference in vivo (Fig. 1 E and F and Table S1), and the same phenotype was obtained for the combined mutation of C14–16G (Fig. S2), suggesting that formation of an RNA hairpin is important for processing. This finding was corroborated by compensatory mutations in positions 13 and 28; although the G28A mutation eliminated primary processing, the double mutant C13T-G28A partially restored it (Fig. S2). The double mutant C13G-G28C, however, showed no processing, suggesting that purines are better tolerated than pyrimidines at position 28. Altogether, these data indicate that recognition of both the GGGACG sequence and the stem loop structure is needed for efficient primary processing and CRISPR interference.

Maturation Is Independent of the Sequence, Structure, or Length of the Intermediate crRNA.

In S. epidermidis, primary processing produces an intermediate crRNA of 71 nt, and additional nucleolytic cleavage of the 3′ end generates smaller mature crRNAs (Fig. 2A and Fig. S3). To determine whether 3′-end maturation relies on a specific sequence or hairpin structure present downstream of spc1, we performed Northern blots of crRNAs containing different mutations in this region (Figs. 1B and 2B and Fig. S4). We analyzed three classes of mutants: (i) containing different sequences immediately downstream of spc1 [DR2 (5′del), DR2–4 (del), and DR2 (1–11)], (ii) lacking the hairpin (HP) structure in DR2 [DR2 (5′del), DR2–4 (del), and DR2 (no HP)], and (iii) producing longer intermediate crRNAs [DR2 (5′del), DR2 (3′del), DR2–4 (del), and DR2(G28C)]. Strikingly, all mutants produced two mature crRNA species of WT length and abundance that supported CRISPR interference against plasmid conjugation (Fig. 2C and Table S2). Although the intermediates generated by primary processing vary in length accordingly with the introduced mutations in DR2 (Fig. S4D), the mature crRNAs maintained a constant size.

To precisely measure the length of the mature species, we captured crRNAs with a biotinylated oligonucleotide complementary to spc1 (Fig. 2D). All intermediate crRNAs were reduced to mature crRNAs with precise lengths of 43 and 37 nt compared with ladders generated by alkaline hydrolysis or RNase T1 cleavage of a synthetic spc1 crRNA. Together, these results indicate that 3′-end maturation does not depend on the sequence, structure, or length of the intermediate crRNA. Interestingly, the ladder RNAs required 3′-end dephosphorylation to match the electrophoretic mobility of the mature crRNAs (Fig. S5B), indicating that mature crRNAs likely contain 3′-hydroxyl groups as opposed to intermediate species that display 3′-phosphate or 2′-3′-cyclic phosphate ends (9, 10, 12, 13). To confirm a 3′-OH group on mature crRNAs, we showed that both the 43- and 37-nt species are suitable substrates for T4 RNA ligase-mediated addition of [5′-³²P]Cp to the 3′ end (Fig. S5C). The difference in the chemical nature of the 3′ end between intermediate and mature crRNAs suggests that primary processing and final maturation are achieved by distinct catalytic mechanisms.

Length of Mature crRNAs Is Measured from the 5′ Primary Processing Site.

In the above experiments, the sequence, length, and/or secondary structure of the 3′ end of the crRNA intermediate was modified, whereas the 5′ end of the crRNA intermediate was maintained constant. The mature crRNA length also remained constant; therefore, we hypothesized that the length of mature crRNA could be determined by a mechanism that measures the distance from the 5′-end primary processing site. To test this hypothesis, we moved the primary processing sequence within DR1 by deleting or inserting nucleotides in the 8-nt crRNA tag (Fig. 3A). In this way, we modified the sequence and length of the 5′ end of the intermediate crRNA but kept its 3′ end constant. The engineered crRNAs exhibited corresponding shortened or lengthened 5′ tags in primer extension assays (Fig. 3B). In addition, Northern blots showed intermediate crRNAs of different sizes that correlated with the deletions or insertions introduced in the crRNA tag region (Fig. 3C). In striking contrast, the length of the mature crRNAs remained constant at 43 and 37 nt (Fig. 3 C and D). For reasons that are still unclear, the mature species in these mutants exist in different proportions. Importantly, sequence analysis of both mature species confirms that extension or reduction of the 5′ tag results in a corresponding reduction or extension of the 3′ end that maintains the crRNA length constant (Fig. S6 and Table S3). Altogether, these results show that maturation is achieved by a mechanism that measures the overall crRNA length using the 5′-end primary processing site as the reference point.

Primary Processing and Maturation Are Mediated by Different cas Genes.

To study the enzymatic machinery responsible for crRNA maturation, we integrated the CRISPR/Cas locus of S. epidermidis RP62a (21) (Fig. 1A) into the lipase gene (geh) of S. aureus RN4220 (22) using pCL55 (23) (Fig. S7). The resulting strain produced mature crRNAs of sizes similar to those crRNAs observed in S. epidermidis but with reduced levels that cannot support CRISPR interference against pG0400 (Fig. 4 and Fig. S7).

We created in-frame deletions of each cas gene and measured the resulting accumulation and maturation of crRNAs by primer extension, Northern blot analysis, and isolation using an oligonucleotide antisense to spc1 (Fig. 4 A and B and Fig. S7). Mature crRNAs were detected in cas1, cas2, and csm6 mutants, indicating that the proteins encoded by these genes do not participate in crRNA biogenesis. In contrast, crRNAs were not detected in cas10, csm4, or cas6 mutants. Cas6 is responsible for primary processing in Pyrococcus furiosus (8, 24), and therefore, the S. epidermidis homolog most likely has the same function. However, the presence of cas6 alone is insufficient for crRNA accumulation (Fig. S7E). Cas10 and Csm4 may be required either to activate or assist Cas6 or for the stability of crRNAs. Interestingly, only the 71-nt intermediate crRNA was detected in strains with single deletions of csm2, csm3, or csm5 or strains containing only cas10, csm4, and cas6 (Fig. 4B, Δcas1/2Δcsm2/3/5/6 lane). These results reveal that Csm2, Csm3, and Csm5 mediate crRNA maturation and that maturation and primary processing are different stages of crRNA biogenesis.

Bacterial Strains and Growth Conditions.

S. epidermidis [RP62a (21) and Δcrispr (7)] and S. aureus [RN4220 (22) and OS2 (30)] strains were grown in brain heart infusion media (Difco). When required, the medium was supplemented with antibiotics: neomycin (15 μg/mL) for selection of S. epidermidis Δcrispr, chloramphenicol (10 μg/mL) for selection of pCRISPR- and pCL55-based plasmids, and mupirocin (5 μg/mL) for selection of pG0400.

Plasmid Construction.

pCRISPR-based plasmids were generated using PCR to amplify pCRISPR (19) using primers containing the desired mutation. PCR products were 5′-phosphorylated using T4 Polynucleotide Kinase (New England Biolabs) and then circularized using T4 DNA ligase (New England Biolabs). The reaction mix was transformed into S. aureus OS2 (30). Plasmids were isolated, and their mutations were corroborated by sequencing and transformed into S. epidermidis Δcrispr as described previously (7). Primer sequences and their uses are shown in Tables S4 and S5, respectively.

Strain Construction.

Construction of S. aureus RN4220::CRISPR/cas strain was performed by integrating pCL55 (23) derivatives containing WT or cas/csm mutant versions of the S. epidermidis CRISPR/cas locus (Table S6) into the geh locus of S. aureus RN4220 (22). Full details are in SI Materials and Methods.

Conjugation.

Conjugation was carried out by filter mating as previously described elsewhere (7). Corroboration of the presence of the desired plasmid in transconjugants and recipients was achieved by extracting DNA of at least two colonies, performing PCR with suitable primers (L6/L50, L70/L71, and L86/L87; specific for S. epidermidis RP62a, pG0400, and pCRISPR, respectively), and sequencing the resulting PCR product.

Primer Extension.

Page-purified primers P1, P2, P3, and PrrfA (for spc1, spc2, and spc3 crRNAs and 5S rRNA, respectively) (Table S4) were used in primer extension reactions that were performed as described in ref. 7.

Northern Blot.

Oligonucleotide probes antisense to spc1, spc2, and spc3 crRNAs and 5S rRNA (Probe1, Probe2, Probe3, and PrrfA, respectively) (Table S4) were used. Northern blots were performed as described in ref. 19.

CRISPR RNA Capture.

50 μg of total RNA and 2 ng of 5′-biotinylated PAGE-purified oligonucleotide antisense to spc1 (5′-CTTTGTACTGATGATTTATATACTTCGGCATACGT) were combined in annealing buffer (5 mM Tris⋅HCl, pH 8.3, 75 mM KCl, 1 mM EDTA) and heated to 95 °C for 1 min and 48 °C for 45 min. Samples were then mixed with 20 μL streptavidin-coated magnetic beads (Thermo) and incubated at room temperature for 1 h. Beads were separated from the supernatant, washed three times in annealing buffer, and resuspended in 30 μL annealing buffer. The samples were heated to 95 °C for 2 min and then placed on ice to release the bound RNA. The RNA was labeled using γ[³²P]ATP and T4 polynucleotide kinase (New England Biolabs) and resolved on a 10% denaturing polyacrylamide gel. The gel was exposed to a storage phosphor screen and visualized using a Typhoon 9400 PhosphorImager.