Protein determinants of high-affinity crRNA repeat binding and cleavage

The high-affinity interaction between Csy4 and the crRNA repeat substrate is tighter than many protein:RNA complexes studied to date. We were therefore interested in gaining a detailed understanding of the primary sources of binding energy, as informed by interactions identified from our crystal structure. We began by focusing on the bottom of the RNA stem, where the side-chains of Arg102 and Gln104 are each involved in two sequence-specific hydrogen bonds with the major groove faces of G20 and A19, respectively. Using a synthetic, noncleavable substrate that is bound indistinguishably from the WT-crRNA repeat (Supplemental Fig. 1B), EMSAs with Csy4-R102A and Csy4-Q104A mutants revealed that the binding energies contributed by these amino acids are quite distinct. The crRNA repeat binds >2000-fold more weakly to Csy4-R102A, representing a ΔΔG of 4.6 kcal/mol, whereas RNA binding by Csy4-Q104A is destabilized by only 1.4 kcal/mol relative to WT (Fig. 2A; Supplemental Table 2). This difference may be explained in part by the expected +1 charge on the arginine's guanidinium group at physiological pH. Whereas deletion of an uncharged hydrogen bond typically weakens binding between enzyme and substrate by 0.5–1.8 kcal/mol, charged hydrogen bonds generally contribute some 3–6 kcal/mol binding energy (Fersht 1987), in good agreement with our data.

Open in a separate window

FIGURE 2.

Amino acid contributions to binding energy and cleavage kinetics. (A) Csy4 residues involved in base-pair recognition and phosphate backbone contacts were mutated to alanine in order to assess their energetic contributions to binding. EMSAs were performed with a noncleavable crRNA repeat substrate containing a deoxyribonucleotide substitution at G20, and binding defects relative to Csy4-H29A were determined and converted to ΔΔG values (T = 298 K). Plotted are the average and SEM from at least three independent experiments. (B) First-order rate constants (k_obs) for WT-crRNA repeat cleavage by each Csy4 mutant were determined. Cleavage data for R118A/R115A and R115A/R119A mutants showed biphasic kinetics and were fit with a double exponential decay to yield two rate constants (Supplemental Table 2), the faster of which is shown. Plotted are the average fold defects (relative to WT-Csy4) and SEM from three independent experiments. Average K_d, k_obs, and SEM values are reported in Supplemental Table 2.

In addition to its interaction with Arg102, G20 of the crRNA repeat stacks onto the aromatic side-chain of Phe155. Stacking interactions between aromatic amino acids and nucleotides can contribute up to 5.5 kcal/mol of binding energy (Nolan et al. 1999; Auweter et al. 2006), but we were surprised to observe a negligible 1.5-fold binding defect (ΔΔG = 0.2 kcal/mol) with a Csy4-F155A mutant (Fig. 2A). Given the pre-crRNA processing defects we observed previously with Csy4-F155A (Haurwitz et al. 2010), these data suggest that Phe155 instead plays a role in achieving rapid cleavage kinetics. Indeed, under single-turnover conditions with saturating enzyme concentrations (see Materials and Methods), the F155A mutant led to an ~50-fold reduction in the observed cleavage rate constant (Fig. 2B). Csy4-R102A also exhibited an ~20-fold defect in cleavage kinetics, whereas the rate of cleavage by Csy4-Q104A was within 2.5-fold of WT (Fig. 2B). Collectively, these data suggest that, independent of their effects on binding energy, Phe155 and Arg102 are important for anchoring the G20 guanine in the active site and may thereby assist in positioning the ribose for subsequent activation of its 2′-OH nucleophile.

Moving up the crRNA repeat stem, we next focused on interactions observed in the crystal structure between the RNA and residues found in the α-helix that inserts into the major groove of the double-stranded stem (Fig. 1B). The guanidinium groups of Arg114, Arg115, Arg118, and Arg119 each present ≥2 hydrogen-bond donors within 3 Å of acceptors in the RNA phosphate backbone, yet their contributions to overall binding energy differ widely, as assessed through double R→A mutations. In particular, Arg114 and Arg118, which contact adjacent phosphates, contribute only 0.7 kcal/mol of binding energy, whereas alanine mutations at Arg115 and Arg119 led to a >15,000-fold binding defect (ΔΔG = 5.8 kcal/mol) (Fig. 2A). While all four residues are positioned to act as arginine forks, in that each side-chain contacts adjacent phosphates (Calnan et al. 1991), only Arg115 and Arg119 may simultaneously utilize all three nitrogen atoms of the guanidinium group as hydrogen bond donors. Arg115 hydrogen bonds to two phosphates in addition to the major groove face of G11, which forms part of the G·A sheared base pair at the bottom of the GUAUA pentaloop, and Arg119 is situated in a unique pocket of the loop where it interacts with phosphates separated by two nucleotides. His120 also interacts with a phosphate at the apex of the loop and contributes 0.8 kcal/mol of binding energy (Fig. 2A). The specific network of multi-dentate contacts between the arginine-rich helix and the RNA stem–loop suggests that high-affinity binding to the crRNA repeat is highly shape-specific, especially with regard to the tertiary structure of the loop. The large magnitude of the binding energy contributed by this protein helix enables Csy4 to maintain a tight grip on the substrate and product, but this interaction is not required for catalytic activity. Cleavage rates for the H120A and R→A mutants under saturating conditions were within fivefold of WT-Csy4 (Fig. 2B).

High-affinity crRNA repeat binding is sensitive to the loop structure

The direction of CRISPR loci transcription in P. aeruginosa has not been directly analyzed, and a recent report that detected mature crRNAs by Northern blot analysis used dsDNA probes that were not strand-specific (Cady and O'Toole 2011). Transcription in a direction opposite to that of our own predictions would generate pre-crRNAs containing the reverse complement of the crRNA repeat sequence. To determine whether Csy4 also recognizes and cleaves this potential substrate, we generated the reverse complement crRNA (rc-crRNA) repeat by in vitro transcription and tested its affinity for Csy4-H29A. We found that the rc-crRNA repeat binds Csy4 >10⁵-fold more weakly than the WT-crRNA repeat (Fig. 3A) and is cleaved >750-fold more slowly (Supplemental Fig. 3A), strongly suggesting that the genuine Csy4 substrate in vivo is pre-crRNA transcribed in an orientation consistent with our previous work (Haurwitz et al. 2010). Northern blot analysis using single-stranded probes indeed confirmed the presence of crRNAs in P. aeruginosa UCBPP-PA14 with the repeat sequence we define in Figure 1A, but failed to detect transcripts from the opposite strand (Supplemental Fig. 4).

Open in a separate window

FIGURE 3.

Importance of loop sequence for high-affinity RNA binding. (A) EMSAs demonstrate that Csy4 binds the reverse complement of the crRNA repeat (rc) >10⁵-fold more weakly than the WT-crRNA repeat. (B) Mutant RNA substrates were generated by changing the WT loop sequence (GUAUA) to a quintuple mutant (UAUAC), the highly stable UUCG tetraloop, or a poly(A) pentaloop, or by removing the loop through use of a substrate nicked between U12 and A13. EMSAs reveal substantial defects associated with binding these mutant RNAs.

When comparing the two RNA sequences, the rc-crRNA repeat contains the identical five-base-pair stem sequence as the WT-crRNA repeat but with an additional predicted G–U wobble base pair below and different loop and flanking ssRNA sequences, indicating that one or more of these regions are specifically recognized by Csy4. Having already demonstrated the negligible binding defects resulting from deletion of flanking ssRNA nucleotides, we suspected that destabilized binding of the rc-crRNA repeat resulted primarily from the inability of Csy4 to interact productively with the UAUAC loop sequence and/or the unique tertiary structure it would impose on the RNA substrate. The GUAUA loop encoded by CRISPR locus 2 in P. aeruginosa UCBPP-PA14 forms a GNR(N)A pentaloop structure (Legault et al. 1998), in which U14 flips out of the loop to enable a GNRA tetraloop fold that involves sequential stacking of U12, A13, and A15 on the 3′ strand of the stem (Haurwitz et al. 2010). The CRISPR 3 locus encodes a GUGUA loop in the repeat sequence that is predicted to form the same pentaloop structure, and this crRNA structure is bound and cleaved indistinguishably from the substrate with a GUAUA loop (Supplemental Fig. 5). We hypothesized that Csy4 specifically recognizes this loop motif, and that other loop sequences unable to conform to a GNRA tetraloop fold would bind much more weakly.

To test this, we generated a panel of RNA substrates containing mutated loop sequences and tested their affinity for Csy4-H29A. In agreement with our hypothesis, Csy4 bound to each RNA at least 7000-fold more weakly than WT (Fig. 3B). Even a nicked RNA substrate formed from two oligonucleotides annealed in trans interacted more favorably with Csy4 than those containing a non-GNRA-like loop (Fig. 3B; Supplemental Fig. 6). These experiments confirm that high-affinity Csy4 binding relies in part on a precise substrate tertiary structure in the loop region, independently of base-specific contacts, and that the absence of a loop altogether is less detrimental to binding than the presence of a nonnative loop. It is interesting to note that, despite their weakened binding, RNAs with mutated loops were cleaved at rates within 2.5-fold of the WT-crRNA repeat at saturating Csy4 concentrations (Supplemental Fig. 3B). This was true even for a substrate containing the same loop (UAUAC) as the rc-crRNA repeat, which had a >750-fold defect in k_obs. Since the stacking interaction between the terminal C–G base pair and the aromatic side chain of Phe155 is important for cleavage (Fig. 2B), we suspected that the additional base pair below the WT stem in the rc-crRNA repeat might impede Csy4 activity (see below).

Specificity within the crRNA repeat stem sequence during binding and cleavage

We were particularly interested in investigating the ability of Csy4 to discriminate between substrates containing the cognate five base pairs in the stem and those with similar but noncognate sequences. We therefore made all individual Watson-Crick base-pair substitutions at each position in the double-stranded stem and determined the energetic costs associated with binding each mutant RNA substrate relative to the WT-crRNA repeat using EMSAs (Fig. 4A). The data reveal that base-pair changes throughout the stem result in varying degrees of Csy4:RNA complex destabilization, ranging from 0.4 to 4.2 kcal/mol. The largest defects result from G–C and C–G substitutions at the ultimate and penultimate base pair, respectively, where Arg102 and Gln104 provide a direct readout mechanism of recognition and confer similar degrees of discrimination in spite of their unequal contributions to binding energy. To confirm this, we repeated binding experiments with RNA substrates containing substitutions at the bottom two base pairs using either Csy4-R102A or Csy4-Q104A (Fig. 4A). As expected, the overall specificity for particular base pairs at either position is lost when the amino acid specificity determinant is absent. The Csy4:RNA co-crystal structure did not reveal sequence-specific contacts with Watson-Crick base pairs in the upper part of the double-stranded stem (Haurwitz et al. 2010), but we observed substantial energetic penalties for binding substrates with base-pair substitutions in this region (Fig. 4A). Furthermore, the magnitude of these binding defects was highly sequence-dependent; when multiple base-pair substitutions were made in the top three base pairs simultaneously, binding defects ranged from seven- to almost 5000-fold (Supplemental Fig. 7), with the largest destabilization occurring when each C–G pair was mutated to its complement. These results reveal that substrate sequence specificity is mediated by Csy4 via a mechanism that does not rely exclusively on base-specific interactions.

Open in a separate window

FIGURE 4.

Substrate specificity within the crRNA repeat stem. (A) A library of mutated crRNA repeat substrates was generated containing all possible Watson-Crick base-pair substitutions at each position in the double-stranded stem. EMSAs were performed with Csy4-H29A and these RNA substrates, and the resulting binding defects relative to WT-crRNA repeat were determined and converted to ΔΔG values (T = 298 K). The WT stem sequence is shown at the left, with data for base-pair substitutions at each position color-coded similarly. Binding experiments with RNAs mutated at the bottom C–G or U–A base pair were repeated with Csy4-R102A (middle) or Csy4-Q104A (right), respectively; ΔΔG values were calculated relative to WT-crRNA repeat binding by each Csy4 mutant. Shown above are chemical structures of the interactions made by Arg102 and Gln104 with the WT base pairs. (B) Single-turnover cleavage assays were performed with the same library of RNA mutants as in A, and the resulting defects in k_obs relative to WT-crRNA repeat were determined. The data are plotted as in A. (C) To investigate the importance of the terminal C–G base pair during cleavage, mismatched substrates were generated by mutating C6 or G20 individually and single-turnover cleavage assays were performed.

We next investigated whether these specificity determinants also influence the chemical cleavage reaction. To test this, we conducted single-turnover cleavage experiments with WT-Csy4 at saturating concentrations using the same library of RNAs as in Figure 4A and determined the first-order rate constants for RNA cleavage (k_obs) relative to WT. In stark contrast to the observed binding specificity, rate constants governing the cleavage of RNA substrates with base-pair substitutions at any position other than the terminal position were within fourfold of WT (Fig. 4B). However, any mutation of the terminal C–G base pair in the stem–loop was detrimental for cleavage of the crRNA repeat, with kinetic defects ranging from ~100- to 7500-fold. To further dissect the importance of the terminal C–G base pair, we generated a series of RNA substrates containing mismatches at this position by mutating either C6 or G20 independently. Cleavage time courses with these substrates (Fig. 4C) clearly demonstrate the importance of G20, regardless of whether or not a base pair can form at the terminal position. RNA substrates containing C6A or C6G mutations were cleaved at rates within 40-fold of WT, whereas mutation of G20 to either adenosine or cytosine led to >10,000-fold defects.

Northern blot analysis

Total RNA was extracted from cultures of P. aeruginosa PAO1, P. aeruginosa UCBPP-PA14, and a csy4 deletion strain of P. aeruginosa UCBPP-PA14 (SMC3894) (Zegans et al. 2009) grown to exponential phase using the mirVana kit (Ambion). Duplicate samples of each RNA preparation (6 μg) were separated on adjacent lanes of a 15% denaturing polyacrylamide gel and subsequently transferred to a nylon membrane (Hybond-N+, GE Healthcare) using a semi-dry transfer cell (Bio-Rad). The single membrane was then cut in half to yield two membranes with identical samples. The membranes were pretreated with ULTRAHyb-Oligo Hybridization Buffer (Ambion) and probed with 5′-[³²P]-radiolabeled DNA oligonucleotides corresponding to either the crRNA repeat sequence (5′-GTTCACTGCCGTATAGGCAGCTAAGAAA-3′) or the reverse complement of the crRNA repeat (5′-TTTCTTAGCTGCCTATACGGCAGTGAAC-3′). Membranes were washed twice with 2× saline-sodium citrate (SSC) buffer containing 0.5% SDS and visualized by phosphorimaging.

RNA transcription, purification, and 5′ radiolabeling

The following RNAs were synthesized by Integrated DNA Technologies: the noncleavable substrate, product RNA (Δ21–28), 5′ truncation constructs (Δ1–5, Δ1–4), the 5′-strand (nucleotides 1–12) and 3′-strand (nucleotides 13–28) used to generate the nicked substrate, the G20A mismatched substrate, and three substrates containing base-pair substitutions at the bottom of the stem (C6U/G20A, C6G/G20C, U7A/A19U). All other RNAs were transcribed in vitro using T7 polymerase and purified using denaturating polyacrylamide gel electrophoresis, according to the following protocol. Synthetic single-stranded DNA templates (Integrated DNA Technologies) containing the reverse complement of the desired crRNA repeat construct were annealed to a 1.5-fold molar excess of an oligonucleotide corresponding to the T7 promoter sequence (5′-TAATACGACTCACTATA-3′). Templates encoded an extra guanosine at the 5′ end of all constructs in order to ensure optimal transcription by T7 polymerase. This had no effect on binding affinities but did lead to a slight (~20%) increase in k_obs for cleavage of the WT-crRNA repeat substrate. Transcription reactions (100 μL) were incubated at 37°C for 3–5 h and contained 1 μM template DNA, 100 μg/mL T7 polymerase, 1 μg/mL pyrophosphatase (Roche), 5 mM NTPs, 30 mM Tris-Cl (pH_RT 8.1), 25 mM MgCl₂, 10 mM dithiothreitol (DTT), 2 mM spermidine, and 0.01% Triton X-100. Reactions were then treated with 5 units of DNase (Promega) and incubated for an additional 30 min at 37°C before being loaded on a 15% urea-polyacrylamide gel. RNAs were excised from the gel and eluted into DEPC H₂O overnight at 4°C. 5′ triphosphates were removed by incubating RNAs at 37°C for 1 h with 10 units of calf intestinal phosphate (New England Biolabs) in 1× NEBuffer 3, followed by phenol-chloroform extraction and ethanol precipitation. RNAs were resuspended in DEPC H₂O and stored at −20°C.

For biochemical experiments, 10 pmol RNA were 5′-radiolabeled by incubating with 5 units T4 polynucleotide kinase (New England Biolabs) and ~3–6 pmol (~20–40 μCi) [γ-³²P]-ATP (Promega) in 1× T4 polynucleotide kinase reaction buffer at 37°C for 30 min, in a 25 μL reaction. After heat inactivation (65°C for 20 min), reactions were spun through an illustra MicroSpin G-25 column (GE Healthcare) to remove ATP. Radiolabeled RNAs were diluted to ~100 nM stock concentrations with DEPC H₂O and stored at −20°C.

Electrophoretic mobility shift assays

Protein concentrations were determined by taking multiple absorbance spectra using a NanoDrop spectrophotometer (Thermo Scientific), averaging A_280nm values and converting to molar concentrations using the calculated Csy4 extinction coefficient (15,470 M⁻¹ cm⁻¹). Spectra were also recorded under denaturing conditions (6 M guanidine hydrochloride, 20 mM potassium phosphate buffer, pH 6.5), and absorbance values were within error of those taken under native conditions. Binding experiments were conducted in the following buffer: 20 mM HEPES (pH_RT 7.5), 100 mM KCl, 5% glycerol, 0.01% Igepal-630, 1 mM DTT, and 0.1 mg/mL yeast tRNA (Sigma-Aldrich) to prevent nonspecific binding. After diluting concentrated 5′-[³²P]-labeled RNA and Csy4 stock solutions into 1× binding buffer, trace amounts of RNA (≤0.05–0.2 nM, depending on construct and specific activity) were incubated with increasing concentrations of Csy4 in a 15 μL reaction at room temperature (~24°C) for one hour. Twelve microliters of each reaction were then loaded on a 10% native polyacrylamide gel containing 0.5× TBE buffer and resolved by running at 12 W for 90–120 min at 4°C in 0.5× TBE running buffer. Phosphor screens were exposed to dried gels and scanned with a Storm imager (GE Healthcare), and the intensities of unbound and Csy4-bound RNA were quantified using ImageQuant (GE Healthcare). The fraction of RNA bound at each Csy4 concentration was plotted as a function of Csy4 concentration, and binding data were fit with a standard binding isotherm using Kaleidagraph (Synergy Software), according to the equation:

where A is the amplitude of the binding curve.

Binding experiments with the substrate nicked between U12 and A13 contained ~1 nM radiolabeled 3′-strand (nucleotides 13–28) and a 1000-fold excess (1 μM) of cold 5′-strand (nucleotides 1–12). For experiments with K_d values in the low pM range, binding data were also fit with the solution of a quadratic equation describing a bimolecular dissociation reaction, as described previously (Maag and Lorsch 2003), out of concern that [RNA] in these experiments was not sufficiently below the K_d to approximate [Csy4]_total = [Csy4]_free. This analysis returned values that agreed well with equilibrium dissociation constants determined from the standard binding isotherm equation, so these original values are reported. When fitting binding data with the rc-crRNA repeat, the amplitude was set equal to one because saturation could not be reached. Binding data with the RNA substrate containing a five G–C base-pair insertion showed apparent cooperativity and were fit with a modified binding equation using a variable Hill coefficient (n ≈ 1.5) and an amplitude fixed at one.

At least one binding experiment for each RNA or Csy4 mutant titrated Csy4 across a concentration range of three orders of magnitude centered around the K_d. Additional replicates typically tested five concentration points centered around the K_d and returned values in excellent agreement with those derived from a more complete titration. K_d values presented in the text and in Supplemental Tables 1 and 2 represent the average and standard error of the mean from at least three independent experiments. The average percent error for all reported K_d values is 10%. ΔΔG values for Csy4 or RNA mutants were calculated according to the equation:

where R is the gas constant, T is temperature (set to 298 K), and K_d,WT/K_d,mutant is the ratio of K_d values for the WT and mutant construct.

The P. aeruginosa UCBPP-PA14 Δcsy4 strain was kindly provided by the G. O'Toole laboratory (Dartmouth Medical School). We thank K. Berry (Harvard Medical School), D. Sashital (The Scripps Research Institute), and other members of the Doudna laboratory for helpful discussions and critical reading of the manuscript. S.H.S. acknowledges support from the National Science Foundation and National Defense Science & Engineering Graduate Research Fellowship programs. J.A.D. is an Investigator of the Howard Hughes Medical Institute.