Methylation guide RNAs without box C/D motifs

Methylation guide RNAs without C/D motifs

sR1c is a typical archaeal C/D RNA and has been subjected to extensive mutational analysis (Fig. 1A; Yang et al. 2020). We created a mutant of sR1c, termed bipartite methylation guide 1 (BMG1), by deleting its terminal C/D motifs while keeping a terminal stem of 4 bp. To measure RNA-guided methylation activities, guide RNAs were assembled with the archaeal Sulfolobus solfartaricus (Ss) proteins Nop5, Fib and L7Ae or the eukaryotic Chaetomium thermophilum (Ct) proteins Nop56, Nop58, Fib, and Snu13 and incubated with a cognate substrate paired to the D′ or D spacer (Fig. 1B; Ye et al. 2009; Lin et al. 2011; Yang et al. 2020). Substrates premethylated at target sites were also assayed to evaluate the site specificity of modification. As shown previously, sR1c guided the methylation of both the D′ and D substrates when loaded in both Ss RNPs and Ct RNPs (Yang et al. 2020). Because archaeal sR1c does not fit well into the Ct RNP, Ct RNP-mediated modification was inefficient for the D′ substrate and enhanced when the C/D motifs were removed (Yang et al. 2020). BMG1 efficiently guided the site-specific modification of the D′ substrate in Ss RNPs and Ct RNPs (Fig. 1B, lane 4). This was expected since the C/D and C′/D′ motifs are functional independent (Yang et al. 2020) and BMG1 still contains intact C′/D′ motifs. Surprisingly, the D substrate was also efficiently modified by the BMG1-loaded Ss RNP (Fig. 1B, lane 5). The activity was specific to archaeal Ss RNPs since the D substrate was not methylated by eukaryotic Ct RNPs (Fig. 1B, lane 5). As the results were apparently contradictory to the consensus that C/D motifs are essential for guiding function, we further analyzed the mechanism underlying the unusual activity of BMG1.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

Activity and assembly of bipartite methylation guide RNAs without C/D motifs. (A) Secondary structure models of sR1c, BMG1, BMG2, and BMG3. Box C, D, C′, and D′ motifs are colored in red and spacers in yellow. The D′ and D substrates (sub) are shown in purple with modification sites marked by circles. The sequences deleted in BMG1 mutants are indicated with lines. (B) Methylation assay. Ss RNPs were assembled with guide RNAs (1 µM), Nop5/Fib complex (2 µM), and L7Ae (3 µM). Ct RNPs were assembled with guide RNAs (2 µM), Nop56/Nop58/Fib complex (∼2 µM), and Snu13 (6 µM). The assembled RNPs were incubated with the D or D′ substrates (30 µM) or mixed premethylated substrates (sub-m) in the presence of 30 µM cold SAM and trace amounts of [methyl-³H] SAM for 20 min at 70°C for Ss RNPs or at 50°C for Ct RNPs. RNAs were resolved in denaturing PAGE and visualized by ³H autoradiography. (C) Activity of Ss RNPs in the presence and absence of L7Ae. (D) EMSA of Ss RNPs. Each RNA (1 µM) was assembled with L7Ae (1 or 2 µM, indicated by a small or large plus sign) and Nop5/Fib complex (2 µM) in various combinations, resolved in 5% native PAGE and strained by SYBR Gold. L7Ae*1 and L7Ae*2 RNPs contain one and two copies of L7Ae. RNP1 and RNP2 refer to monomeric and dimeric species. Misfolded RNAs that migrated slower and bound poorly to L7Ae are marked with asterisk.

Deletion of the D spacer of BMG1 eliminated modification of the D substrate (Fig. 1B, lane 8), confirming that the activity was RNA-guided. Modification on the D substrate was greatly reduced when the terminal stem was disrupted by removing 18 nt at the 5′ end or 4 nt at the 3′ end of BMG1 (Fig. 1B, lanes 11 and 14). These suggest that the terminal stem is critical for the function of the D spacer and may replace the structural role of C/D motifs. In light of this idea, we further replaced the internal box C′/D′ motifs of BMG1 with a 4-bp stem, yielding a BMG2 RNA without any C/D motifs (Fig. 1A). Indeed, BMG2 guided site-specific modification of both the D and D′ substrates (Fig. 1C, lanes 7–9). For structural study, we designed a two-pieced, more symmetric C/D-free guide RNA, BMG3, which was composed of two 8-bp stems linked by two identical 13-nt sequences (Fig. 1A). BMG3 also turned out to be a functional methylation guide (Fig. 1C, lanes 10–11), indicating that the terminal stem can be longer than 4-bp.

Since L7Ae specifically binds the K-turn structure formed by C/D motifs (Vidovic et al. 2000; Moore et al. 2004), C/D-free methylation guide RNAs in principle do not require L7Ae for assembly and activity. Indeed, for BMG1, the activity of the D spacer coupled to the terminal stem did not depend on L7Ae, whereas the D′ spacer coupled to the internal C′/D′ motif still required L7Ae (Fig. 1C, lanes 4,5). BMG2 and BMG3 completely lacked C/D motifs and were able to guide specific modification of all substrates in the absence of L7Ae (Fig. 1C, lanes 7–11). sR1c depended on L7Ae for guiding modification of the D′ substrate (Fig. 1C, lane 1), but puzzlingly, it still guided modification of the D substrate in the absence of L7Ae (Fig. 1C, lane 2). Further studies on the unusual activity will be described later.

The assembly of Ss RNPs was examined with electrophoretic mobility shift assay (EMSA) (Fig. 1D). As expected for specific association of L7Ae with K-turn, sR1c with two sets of C/D motifs bound one or two copies of L7Ae, BMG1 with one set of C/D motifs bound only one copy of L7Ae, and BMG2 and BMG3 that contain no C/D motifs only associated nonspecifically with L7Ae. All these guide RNAs can assemble with the Nop5/Fib complex alone, accounting for their activity in the absence of L7Ae. Dimeric RNPs (RNP2) were observed for all RNAs due to self-association of monomeric RNPs (RNP1) (Yang et al. 2020). In conclusion, we have generated a new type of methylation guide that differs from natural C/D RNAs by lacking C/D motifs and requiring no L7Ae for assembly and function.

Structure of a C/D-free RNA-guided methylation RNP

To understand the structural basis of C/D-free methylation guide RNA activities, we determined a crystal structure of BMG3 RNA in complex with Nop5, Fib and substrate (Fig. 2A,B). The structure was solved by molecular replacement with the substrate-bound SL12 (also called CD45) C/D RNP structure minus L7Ae as a searching model (Lin et al. 2011) and refined to 2.2 Å resolution (Supplemental Table S1; Supplemental Fig. S1A). Despite the lack of L7Ae and C/D motifs, the BMG3 RNP structure closely resembles the SL12 RNP structure with a root mean square deviation (RMSD) of 0.709 Å over 1114 Cα atom pairs (Fig. 2C,D; Supplemental Fig. S1B). In the BMG3 RNP structure, Nop5 forms a homodimer by its coiled-coil domain, associates fibrillarin with its NTD and binds the terminal stems of BMG3 with its CTD. Each substrate forms a 10-bp duplex with the spacer of BMG3 and contacts fibrillarin. The active site of fibrillarin is loaded with a copurified S-adenosyl homocysteine (SAH) molecule, which differs from SAM by lacking the methyl donor group and an unmodified ribose at the target site (Supplemental Fig. S1A). These features indicate that the structure represents an active conformation prior to modification.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Crystal structure of substrate-bound BMG3 RNP. (A) Secondary structure diagrams of SL12 and BMG3 RNAs. C/D motifs in SL12 and segments C and D in BMG3 are colored in red, spacers in yellow and substrates in purple. Modification sites are marked by circles. The structurally equivalent Nop5-binding regions are encircled with blue boxes. Nucleotides in guide RNAs and substrates are separately numbered. Spacer sequences are additionally labeled by yellow numbers to indicate the distance to box D/D′ or segment D/D′. (B) Ribbon representation of structure of BMG3 in complex with Nop5, fibrillarin (Fib) and substrate. Two Nop5 molecules are colored in forest green and split pea, Fib in teal, guide RNA in yellow and substrates in purple. SAH bound in the active site of Fib is shown as red spheres. (C,D) RNA–protein interfaces in the BMG3 RNP (C) and SL12 RNP (PDB code: 3PLA) (D) structures. (E) Detailed interactions between BMG3 RNA and Nop5 at segments C/D. Hydrogen bonds are denoted by dashed lines. (F) Schematic of interactions between BMG3 RNA and Nop5.

In the conventional C/D RNP structure (Lin et al. 2011), the C/D RNA is assembled to the complex through two K-turns formed by the C/D motifs and their nearby sequences (Fig. 2A,D). A typical K-turn consists of a noncanonical stem containing tandem sheared GA pairs (GA stem), a 3-nt loop and a canonical stem composed of Watson–Crick pairs (WC stem) (Vidovic et al. 2000; Klein et al. 2001). The WC stem is minorly contacted by Nop5 and L7Ae and could be absent or replaced by a loop to form a K-loop. The K-turn structure is bound by L7Ae and the CTD of Nop5, two proteins that also interact with each other in C/D RNP structures. L7Ae specifically recognizes the GA stem and the internal loop, including a conserved projecting uridine. The CTD of Nop5 binds the GA stem, several adjacent spacer residues of C/D RNA and the loaded substrate. The α9′ helix of Nop5 is a key RNA-binding element that is located at the three-way junction of two guide-substrate duplexes and the GA-stem.

The terminal stem of BMG3 RNA lacks a K-turn and hence cannot bind L7Ae. Nevertheless, the terminal stem still contacts the CTD of Nop5 in a similar manner as the GA stem of C/D RNA does (Fig. 2C,D; Lin et al. 2011). Specifically, the proximal 4-bp duplex of the terminal stem is structurally equivalent to the GA stem of C/D RNA, which is composed of nucleotides 3–6 of box C (RUGAUGA) and the entire box D (CUGA). In addition, the unpaired base (G9 and G38 of BMG3) located at the proximal 3′ end of the terminal stem is structurally equivalent to the seventh base of box C (A16 and A56 of SL12). Both bases stack over the GA stem and bind Nop5. To show structural analogy with C/D RNA, we termed a 5-nt segment C or C′ in the C/D-free guide RNA that corresponds to nucleotides 3–7 of box C or C′ and a 4-nt segment D or D′ that corresponds to box D or D′ (Fig. 2A,C,D). Consequently, BMG3 contains two spacers with an effective length of 12 nt, just like the SL12 C/D RNA. At the side of segments C/D, Nop5 primarily binds the 3′ half of segment C and several spacer residues and barely contacts segment D (Fig. 2E,F), which is more distant to Nop5 compared to box D of C/D RNA (Supplemental Fig. S1B,C). Basic residues R313, K316, R331, K336, and R339 interact with the phosphate groups in segment C and spacers. In addition, residues Q296 and R339 make two hydrogen bonds with the O6 atom of unpaired G9. The protein–RNA interactions are similar at the side of segments C′/D′, but several interactions vary due to the different positions of protein side chains (Supplemental Fig. S1D).

In the BMG3 RNP structure, nucleotides 1 and 12 of the spacers (A10/U21 or A39/U50) do not bind substrates but contact helix α9′ of Nop5. Nucleotides 2–11 (A11–G20 or A40–G49) of the spacer pair with the substrate to form a 10-bp duplex. Lying over the dimerized coiled-coil domain of Nop5, the guide-substrate duplex is capped by two opposing CTDs of Nop5 at both ends and limited in size to 10 bp at the active state (Yang et al. 2016). Although U11/U31 of the substrate was designed to pair with nucleotide 12 of the spacer, it is unpaired in the structure and stacks over H327, as observed previously in a C/D RNP structure (Yang et al. 2016). The conformation of the guide-substrate duplex and its interactions with fibrillarin and Nop5 are nearly identical in both the structures of SL12 C/D RNP and BMG3 C/D-free RNP. Hence, the terminal stem of BMG3 can functionally replace the GA stem of C/D RNA in binding Nop5, ensuring the correct placement of the guide-substrate duplex and the site-specific modification.

Monopartite methylation guide RNAs without C/D motifs

The bipartite architecture has been shown to be dispensable for C/D RNA, and halfmer C/D RNAs with a single set of C/D motifs and a guide are also functional in vitro (Yang et al. 2020). We wondered if C/D-free guide RNAs can be also reduced to a monopartite form. We initially designed a monopartite methylation guide RNA (MMG1) containing a stem–loop and a 12-nt guide at the 5′ end, which has identical sequences as the D spacer of sR1c and BMG2 (Fig. 3A). The modification site was predicted to pair with position 5 of the guide as in bipartite methylation guides.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Activity of monopartite methylation guide RNAs. (A) Secondary structure diagrams of MMG1 to MMG5. These RNAs share the same guide sequence as the D guide of sR1c and BMG2. The pairing of MMG1 with substrates S8–S10 are shown. Modification targets are indicated by circles. (B–H) Activity and specificity of Ss RNPs assembled with sR1c (B), BMG2 (C), MMG1 (D), MMG2 (E), MMG3 (F), MMG4 (G), and MMG5 (H) and without (w/o) or with (w/) L7Ae. Substrates S8–S10 or their premethylated versions (sub-m) were modified at 70°C for 20 min. Reactions from a same gel are boxed. Relative modification levels were quantified against the reference reaction (Ref) present in each gel that showed modification of S10 by sR1c holo-RNP.

Four substrates S8–S11 that formed 8–11 bp duplexes with the guide were tested for modification by Ss RNPs assembled with or without L7Ae. Shorter substrates were commonly modified to less degrees and sometimes with less specificity. All substrates were efficiently modified at the target site by the fully assembled sR1c RNP (Fig. 3B, lanes 9–16). All substrates except for S8 were efficiently and specifically modified by the bipartite C/D-free BMG2 RNP in the L7Ae-independent manner (Fig. 3C). The monopartite MMG1 RNA poorly guided the methylation of these substrates (Fig. 3D). Moreover, the substrates premethylated at predicted target sites were modified to similar degrees as the unmethylated substrates, indicating that the modification occurred mainly at other sites.

In the BMG3 RNP structure, three nucleotides (G9, A10 and A11) in the 3′ overhang of the terminal stem interact extensively with the CTD of Nop5 (Fig. 2E). The three nucleotides contribute to anchoring the guide RNA onto Nop5, but are missing in MMG1. The stem of MMG1 might unwind on the top to create a 3′-overhang to bind Nop5, which would shift the stem's position on Nop5 and alter the specificity of modification. To fix this problem, the 3′ end of MMG1 was extended by AGA, generating an MMG2 RNA (Fig. 3A). In MMG2, the first adenosine of the 3′ overhang could potentially pair with a uridine at position 1 of the guide. To avoid this pairing, the uridine of the guide was further replaced with a cytidine to create an MMG3 RNA (Fig. 3A).

We found that MMG3 can guide modification of substrates S10 and S11 at the expected target but with poor activities (Fig. 3F). The targeting specificity of MMG2 fell between that of MMG1 and MMG3, suggesting that the terminal AU pair was partially formed in MMG2. All the tested monopartite C/D-free guides were much less active than bipartite C/D-free guides. Finally, L7Ae did not affect both the activity and specificity of all C/D-free methylation guides, as expected (Fig. 3C–F).

We also assessed the activity of monopartite C/D RNAs that contain a K-turn (MMG4) or a K-loop (MMG5) (Fig. 3A). In the absence of L7Ae, both RNAs displayed poor activities (Fig. 3G–H, lanes 1–8). MMG4 showed considerable specificity for substrates S10 and S11, but poor specificity for shorter substrates. MMG5 has worse guiding specificity than MMG4. When assembled with L7Ae into holoenzymes, both RNAs guided modification of all substrates with strict specificity and high efficiency (Fig. 3G,H, lanes 9–16).

In sum, monopartite C/D-free guide RNAs require a 3′-overhang for proper assembly and display rather poor activities. Their activity and specificity can be greatly improved by becoming bipartite (BMG2) or incorporating C/D motifs and L7Ae (MMG4 and MMG5).

L7Ae is dispensable for modification of strongly paired substrates in C/D RNP

L7Ae recognizes the K-turn and is considered to be essential for the assembly and function of C/D RNA. The activity assay of sR1c showed that the lack of L7Ae blocked modification of the D′ substrate, but did not affect modification of the D substrate (Fig. 1C, lanes 1–3 and Fig. 4B, lanes 1–3). The D′ and D guides were coupled to the internal C′/D′ motif and the terminal C/D motif, respectively (Fig. 4A). To test whether their positions mattered, the two guide sequences were exchanged in sR1c-ex RNA. The exchanged guide regions both worked in the fully assembled RNPs, but instructed modification of only the original D substrate in the absence of L7Ae (Fig. 4B, lanes 4–6). Hence, the position of these guides is not related to their different activities.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

L7Ae is dispensable for modification of strongly associated substrates by C/D RNPs. (A) Substrate interactions of sR1c, SL12, and Ct snR61 RNAs. Gibbs free energy (ΔG) is indicated for each guide-substrate duplex. The D substrate of sR1c was mutated to change the GC pair at position 3 of the guide-substrate duplex to a GU or GG pair. (B) Activity of sR1c and sR1c-ex where the D and D′ guides were exchanged. The D or D′ substrates and their premethylated versions were modified at 70°C by Ss C/D RNPs assembled with (w/) or without (w/o) L7Ae. Lanes 1–3 are copied from Figure 1C. (C) The D substrate mutants (S) and their premethylated versions (M) were modified at 70°C by sR1c RNPs assembled with (w/) or without (w/o) L7Ae. The relative modification levels were normalized to modification of wild-type (GC) substrate by sR1c holo-RNP. Averages and standard deviations were calculated from three replications. (D) Modification at 70°C and 50°C. sR1c, SL12, and Ct snR61 RNPs were assembled with or without L7Ae and assayed on their cognate D and D′ substrates and mixed premethylated substrates. The relative modification levels were normalized to the reactions catalyzed by holo-RNPs at 70°C. Averages and standard deviations were calculated from two replications.

We noticed that sR1c formed a more stable duplex with the D substrate than the D′ substrate (ΔG = −18.05 vs −14.83 kcal/mol) and hypothesized that L7Ae may be dispensable for strongly paired substrates. To test this idea, the stability of the D guide-substrate duplex was reduced by replacing a GC pair at position 3 with a GU wobble pair (ΔG = −15.87 kcal/mol) or a GG mismatch (ΔG = −14.22 kcal/mol). The GU pair did not affect much modification by the holo-enzyme, but abolished modification by the L7Ae-lacking enzyme (Fig. 4C, lane 3). The GG mismatch greatly reduced modification by the holo-enzyme and totally abolished modification by the L7Ae-lacking enzyme (Fig. 4C, lane 5). Along the same line, short substrates S8 and S9, but not long substrates S10 and S11, showed L7Ae dependence for modification (Fig. 3B). We conclude that L7Ae could be dispensable for modification of strongly bound substrates.

To stabilize pairing interactions between substrates and guide RNAs, modification reactions were conducted at a lower temperature of 50°C (C/D RNPs from thermophilic Ss were commonly assayed at 70°C). Interestingly, the D′ substrate of sR1c was now modified to similar degrees in the presence and absence of L7Ae (Fig. 4D, row 1, lanes 7 vs 10). This indicates that L7Ae became dispensable when the substrate binding was strengthened at lower temperatures. Modification of the D substrate remained independent of L7Ae at 50°C (Fig. 4D, row 1, lanes 8 vs 11), yet was much worse than modification of the D′ substrate. Product release was probably inhibited for stably paired substrates at lower temperatures in the multiple-turnover reactions. We further analyzed how reaction temperature affects the activity of two other C/D RNAs: SL12 and Ct snR61 (Fig. 4D; Lin et al. 2011; Yang et al. 2020). The D′ substrate of SL12 and the D′ substrate of Ct snR61 cannot be modified by L7Ae-free RNPs at 70°C, but can at 50°C (Fig. 4D, rows 2 and 3, lanes 4 vs 10). These data show that low reaction temperature can enhance stability of guide-substrate duplexes and alleviate the requirement of L7Ae in C/D RNA-guided modification.