Nucleic Acids Research

Incorporation of crosslinking data into secondary structure models suggests that all three gRNA/mRNA pairs interact to form similar structures

The RT analyses indicate that for all three gRNA/mRNA substrate pairs, the gRNA U₁₀ tail interacts with the mRNA in a region just upstream of the anchor duplex (Fig. 7). For gA6-14 and gCYb-558, crosslinking of the terminal uridylate occurred relatively close to the anchor duplex with the preferred sites located from 10-12 and 13-16 nt 5[prime] of the anchor duplex, respectively (Fig. 7B and F). The terminal uridylate of gND7-506 crosslinked farther from the anchor duplex with the preferred sites at 21-28 nt upstream (Fig. 7D). These crosslinking data were incorporated into the computer-predicted secondary structure models by instructing the program to pair the U₁₀ nucleotide with the strongest crosslink site. When this was done, the model secondary structures generated were all very similar and differed substantially from the initial computer predictions (Fig. 7). In all cases, the anchor duplex region is correctly paired and any secondary structure in the immediate editing domain is eliminated (compare Fig. 7A and B for 3[prime]A6U, Fig. 7C and D for 5[prime]ND7UMT and Fig. 7E, F and G for 5[prime]CYbUT). In addition, the gRNA guiding region potentially forms a stem-loop structure positioned across from the first editing site.

DISCUSSION

Guide RNAs are an essential component of the editing process, supplying the information for the multitude of uridylate insertions and deletions that must occur. Little is known, however, about how the gRNAs are used to direct RNA editing. To approach this question, we have begun to analyze the interactions between gRNAs and mRNAs using comparative photoaffinity crosslinking techniques. Placement of the APA group at the 5[prime]-end of the gRNA allowed us to analyze duplex formation between the gRNA anchor and the mRNA. For both gND7-506 and gCYb-558, the two gRNAs for which anchor duplex interactions were analyzed, the crosslink data suggests that the predicted anchor duplexes correctly form. RT mapping of 5[prime] crosslink conjugates indicated that the crosslinks were restricted to 2-3 nt surrounding the anchor duplex 3[prime] (mRNA orientation) border. This may be explained by the fact that the APA group is localized on the 5[prime]-most gRNA nucleotide and can interact with C-H and N-H bonds in its immediate proximity.

In contrast to the 5[prime] crosslinks, analyses of gRNA/mRNA conjugates formed when the APA group was placed on the 3[prime]-end of the gRNA U-tail showed a distinctly different crosslinking pattern. RT mapping of the 3[prime] crosslinked conjugates indicated that the terminal U of the U-tail could interact with a large range of nucleotides located upstream of the first editing site. For all three gRNA/mRNA pairs analyzed, a series of primary crosslinks along with a range of minor crosslinks were detected. In comparing the different gRNA/mRNA interactions, it is interesting that the gRNA U-tails interact in the same relative position, just upstream of the anchor. 3[prime]A6UT is extensively edited throughout most of the message. In the substrate we used, almost no pyrimidine residues are found between 10 and 60 nt upstream of the anchor duplex. However, within the first 10 nt of the anchor, five of the residues are uridylates. Although these five uridylates presumably would not base pair with the gRNA U₁₀ tail, strong crosslink sites were mapped at nt 10-12. This suggests that an interaction of the U-tail does not require the entire tail to base pair to the mRNA. For this substrate pair, the strong RT termination signals found in control reactions makes it difficult to determine if stronger crosslinking nucleotides do in fact occur upstream of nt 12. However, we can say that a sizable proportion of the gRNA molecules are crosslinked at nt 10-12 indicating that the U-tail interaction may involve more constraints than simple purine to U-tail base pairing.

The 5[prime]CYbUT pre-edited substrate differs from 3[prime]A6UT in that the editing domain is quite short, spanning only 19 nt. This region is extremely purine biased with a single pyrimidine found in the 22 nt directly 5[prime] of the anchor duplex region. Using gCYb-558 with a U₁₀ tail produced dominant crosslinks at nt 13-16, just 5[prime] of the center of this purine-rich region. Distinct crosslinks are again found 3[prime] (closer to the anchor), but not farther upstream. Shortening the U-tail to just five U residues shifted the dominant crosslinks to within 1-2 nt of the anchor duplex. It is interesting to note that the crosslinks did not move just five bases, but instead shifted 10 bases closer to the anchor. Incorporation of this crosslink data into the computer-predicted secondary structures suggests a shift in the gRNA stem-loop, with three of the uridylates in the U-tail incorporated into the stem-loop structure (Fig. 7D). This again indicates that the entire U-tail does not necessarily base pair with the mRNA (due to the presence of the anchor duplex) and that the position of the U-tail along the mRNA is most likely not driven solely by base pairing interactions with the purine-rich mRNA. It should be noted, however, that shortening the U-tail did alter the populations of gRNA/mRNA conjugates obtained. In the presence of a U₁₀ tail, a single dominant conjugate (B1) was always observed, with two minor conjugates (B2 and B3) consistently appearing. The drop in tail length to U₅ shifted the distribution of these populations so that the B2 and B3 conjugates were more abundant. This suggests that the drop in tail length destabilized the most common conformation (detectable by our crosslinking technique).

5[prime]ND7UMT differs from the other two substrates in that it contains only one stretch of 13 purines located from 24 to 37 nt upstream of the anchor duplex. While the strongest crosslinks are found within this purine-rich region, a significant number of crosslinks were also mapped much closer to the anchor duplex in a region which is 50% pyrimidine.

Indeed, for all three RNA pairs, minor crosslinks are observed 3[prime] of the major crosslinks at almost all nucleotides down to the anchor duplex. This `ladder' pattern of crosslinks is clearly distinct from that observed in the 5[prime] crosslinking studies. It cannot be explained by reverse transcription read-through nor is it likely that the terminal uridylate bound to a single base while the APA group inserted into a range of neighboring bases. Instead there are two reasonable explanations. (i) In vitro transcription of a gRNA with a U₁₀ tail results in a population of gRNAs with U-tails of varying lengths. Gel purification improved the homogeneity of the gRNAs, however, the populations used did contain U-tails ranging in size from U₅ to U₁₅ (data not shown). This sub-population could be at least partially responsible for the ladder of minor crosslinks. (ii) The interaction of the U-tail with the pre-mRNA may be flexible. The U-tail appears to bind preferentially to a specific region, however, the ability to slide up and down the pre-mRNA sequence could result in the range of minor crosslinks observed. Our data suggests that the heterogeneous populations of gRNAs is unlikely to be the major contributing factor to the minor crosslinks observed. The population of gRNAs showed a Gaussian distribution with respect to the size of the U-tail (data not shown). However, we did not observe a corresponding distribution in the crosslinks. Furthermore, gCYb-558 with a U₅ tail also gave a ladder of termination products which corresponded with those observed with the U₁₀ gRNA. The reduction in the number of U residues significantly decreased the amount of stuttering by T7 polymerase. This indicates that the heterogeneity in the gRNA population is not the cause of the ladder of crosslinks observed. Instead, it appears that although the U-tail shows a preference for a particular region, it is capable of binding to a larger range of upstream sequences. This range is constrained, however, in that we did not find crosslinks to the entire range of purine biased sequences available for interaction with the U-tail.

The initial computer structure predictions did not reveal any secondary structures that were common between the interacting RNAs (Fig. 7A, C and E). However, when the 3[prime] crosslinking data was incorporated into the computer-predicted structures, the structures generated were all very similar (Fig. 7B, D, F and G). In all cases, the anchor duplex region is correctly paired and secondary structure in the mRNA editing domain is eliminated (compare Fig. 7C with D and Fig. 7E with F and G). In addition, the gRNA guiding region forms a stem-loop positioned across from the first few editing sites. These predicted guiding region stem-loop structures are of particular interest as they show similarities to the 3[prime] stem-loop structures identified in gRNAs by structure probing experiments (14). Schmid et al. (14) determined the secondary structures of four different gRNAs from T.brucei using a combination of temperature-dependent UV spectroscopy and chemical and enzymatic probing techniques. Alone, gRNA molecules fold into two hairpin elements separated by a single-stranded region of variable length. The 5[prime]-ends of the four gRNAs investigated were all found to be in a single-stranded conformation followed by a small hairpin that contains the anchor sequence. The second hairpin element (stem-loop II) involves the guiding region of the gRNA and is the more stable of the two hairpin elements. For all four gRNAs, the oligo(U) tail was found to be in a single-stranded conformation. The structure predicted for the gA6-14/A6 mRNA interaction shows two hairpins in the guiding region of the gRNA. The 3[prime]-most stem-loop is identical to the 3[prime] stem-loop observed for gA6-14 alone. The predicted structure for gND7-506/5[prime]ND7UMT also contains a stem-loop in the gRNA that contains many of the same bases as observed in the stem-loop formed by the gRNA alone. Neither of the predicted two stem-loops in gCYb-558 are identical to the identified 3[prime] stem-loop in gCYb-558. This may be due to the fact that the gCYb-558 sequence we generated differs slightly from that utilized by Schmid et al. (14) in their structure probing experiments.

As suggested by Schmid et al. (14), the gRNA 5[prime]-most weak stem-loop involving the anchor region would have to melt out in order for it to form the gRNA/mRNA anchor duplex that initiates the editing events. If the second stem-loop (stem-loop II) is maintained during the initial interaction, the U-tail interaction with the mRNA might be constrained, so that it would tend to interact with relatively close upstream regions. This may be what is limiting the ability of the U-tail to interact with mRNA sequence farther upstream. The positioning of the U-tail near the anchor duplex may also explain the generation of chimeric gRNA/mRNA molecules. Most of the chimeras generated in vitro and characterized in vivo have gRNAs with very short or no U-tails covalently linked to the first few editing sites (7,4,12,29). If the predicted gRNA stem-loop is maintained during the initial editing events, gRNAs missing a U-tail would have their 3[prime]-ends positioned very near the active editing sites possibly allowing the ligation of the gRNA 3[prime]-end to the 3[prime] cleavage product.

We were surprised that the presence of editing-active lysate did not affect the pattern of crosslinks obtained. However, the initial secondary structure studies of gRNAs alone indicated that the presence of mitochondrial proteins did not change the overall structure (30). In detailed studies of gND7-506 complexed with the gRNA binding protein gBP21, Hermann et al. (31) found that the protein binds to the guiding region stem-loop (stem-loop II), with the gRNA structure remaining largely unchanged. This association does appear to increase the stability of the gRNA structure. It may be that the interactions we observed are RNA driven with the proteins possibly reinforcing the preferred interaction. Alternatively, it may be that the molar concentrations of proteins present in our editing lysates were not high enough to affect the structures observed. Currently, we are looking to see if specific editing proteins (such as gBP21) can increase the efficiency of our crosslinking interactions or alter the crosslinking patterns observed.

In this initial study of gRNA/mRNA interactions, photoaffinity crosslinking agents localized to the 5[prime]- and 3[prime]-ends of three different gRNAs were used to map the positions of the gRNA 5[prime] anchor and 3[prime] U-tail along their cognate mRNAs. Taken together, the data suggests that the gRNA 5[prime] anchor does position the gRNA by duplexing with the mRNA just 3[prime] of the editing domain. In addition, our data suggest that the 3[prime] U-tail also interacts with the mRNA possibly base pairing with purine-rich upstream sequences. While this interaction appears to be flexible in that we always saw a range of crosslinks, our data does suggest that the U-tail interaction is somewhat constrained to a region relatively close to the first few editing sites. Computer modeling of the RNA interactions indicates that the stem-loop II structure, present in free gRNAs, may be maintained in the initial gRNA/mRNA interaction. With the gRNA stem-loop II maintained, the 5[prime] anchor and the U-tail mRNA interactions flank the first few editing sites, possibly working together to eliminate mRNA secondary structure in the immediate editing domain. This type of interaction may increase the accessibility of the editing complex to the proper editing sites.

ACKNOWLEDGEMENTS

We thank M. Harris and N. Pace for technical and other advice concerning crosslinking protocols, K. Stuart for help in obtaining editing-active lysate and U. Göringer for supplying us with gRNA plasmids and critical comments on the manuscript. This work was support by NIH grant number AI34155 to D.K.