Nucleic Acids Research

DISCUSSION

All naturally occurring self-cleaving hammerhead structures in plant pathogenic RNAs have a uridine in the central position of the triplet sequence (9). Initial mutational studies to determine the specificity of the hammerhead reaction did not include the central uridine of the triplet until a thorough study was performed (7). The study was based on a complex where ribozyme and substrate oligoribonucleotides had been annealed via helices I and II and which had a 3 bp helix III (Fig. 1, construct III). In this structure changes from uridine to any other nucleotide resulted in a very slow cleavage rate, with a k_cat[prime] of 0.002/min. We, however, determined a k_cat[prime] of 0.085/min for the GAC-containing substrate, a considerably higher rate. It has been reported that it is more difficult to find substrate and ribozyme sequences for the construction of I-II complexes than for the more conventional I-III complexes, as the former tend to form stable alternative conformers (10). Because of this complication the authors pointed out that the particular construct III (Fig. 1) is unsuitable for studies involving mutant or chemically modified hammerheads. Indeed, we found by native gel analysis that substrate and ribozyme with the GAC triplet sequence only formed a complex to ~45%. It might well be that the poor cleavage reported 3[prime] of GAC was caused by complications in formation of a kinetically competent complex (10).

In another study where the GAC triplet was tested, the substrate and ribozyme were part of two transcripts of ~75 nt each (8). No cleavage 3[prime] of the GAC triplet was observed by these authors. However, re-investigating this finding with a ribozyme-substrate complex of the same sequence but with shorter helices I and III (construct IV), the substrate with the GAC triplet was cleaved quite well, with a k_cat[prime] of 0.52/min. It is unclear at present what the origin of this discrepancy is, but presumably it is also a result of unfavourable folding of the transcripts. However, as in the case of construct III, other changes in the triplet reported in that publication comply with the general NUH rule.

Although several subsequent publications have investigated the cleavage efficiency of various substrate permutations of the NUH type, none seems to have questioned that uridine is an invariant nucleotide (11,12). Thus, our observation that a substrate with a GAC triplet could be cleaved in construct I (Fig. 1) came as a surprise. Cleavage was not limited to this particular triplet, but also extended to GAA and GAU, but not to GAG. Substrate cleavage with GAC and GAA triplets were also tested in the sequence context of the well-established ribozyme HH16 (construct II) (6), where they also supported cleavage, indicating that A as the central nucleotide is a more generally tolerated sequence. The time dependence of cleavage with the GAC triplet in construct I indicates that the kinetics cannot be described by a single rate constant beyond 35% cleavage, indicating that more than one conformation must be present, one of which is cleaved more slowly (Fig. 2).

The cleavage rates for the GAH triplets in construct I differed, with GAC > GAA > GAU. The same hierarchy was seen for GUH (Table 1). This is in agreement with observations by others (11-13). Even though the absolute values are not the same from the various studies, it is clear that there is an influence by the third nucleotide and the ranking of cleavage efficiencies for NAH is in general agreement with that found with the NUH triplet. The cleavage rate 3[prime] of the GAC triplet in construct I at pH 7.5 and 25°C is ~7% of that 3[prime] of the GUC triplet. A comparison of rates determined under these conditions is more reliable than those obtained at pH 8.0 and 37°C. In construct II (HH16) the difference is larger, with the rate for GAC cleavage only 3% that of GUC under identical conditions. Thus these rates fall into the same order as those for GUU cleavage, which are ~5% of those for GUC cleavage (7,12).

It is interesting to inspect the X-ray structures of the hammerhead ribozyme to see how the new U^15.1A^16.1 base pair can be accommodated (14). The X-ray structure provides evidence that of the two potential H bonds in the U^16.1A^15.1 base pair, only one, between the 4-keto group of U^16.1 and the 6-amino group of A^15.1, is formed in the conventional type hammerhead. The distance for the second is not commensurate with formation of such a bond. The 2-keto group of U^16.1 is also involved in interactions with the A⁶ ribose. Changing the U-A base pair to an A-U base pair retains the potential for H bonds between the two nucleotides, although with inversion of polarity. However, it is difficult to imagine that the interactions of U^16.1 with the A⁶ ribose could be maintained by A^15.1 or compensated for by similar interactions with U^16.1. Of course, other interactions could be possible which are not seen with the GUC triplet. However, the lower activity for cleavage of GAC-containing substrates than those containing GUC triplets indicates that formation of the transition state for the former is more difficult than for the latter. The change of the U^16.1 A^15.1 base pair to a C^16.1 I^15.1 base pair without loss of activity is more conservative, as the H bond pattern has been retained, except for the inverted polarity, and the pyrimidine nucleoside 2-keto function is still in place (4).

As the exocyclic functional groups of the original U^16.1 A^15.1 base pair are now exchanged without complete loss of activity, it is unlikely that they are active participants in catalysis. Thus the interactions seen in the X-ray structure for U^16.1 might be silent, in that they do not contribute to either binding and/or formation of the transition state. This is another example of the notion that not all interactions seen in the X-ray structure are necessarily important for the reaction. This has been assumed before because of some inconsistencies between chemical modification studies and the X-ray structure (15,16). It has been particularly nicely demonstrated for U⁷, which can be replaced by a pyridin-4-one nucleoside with an increase of catalytic efficiency even though it cannot entertain the H bonds found for this uridine (17).

Now that it has been established that position 16.1 can be either uridine, adenosine or cytidine, the question arises in what way these nucleotides differ from guanosine, which is not tolerated. The cause for the lack of cleavage of NUG triplets has been explained in part by the formation of three H bonds, with C3 locking the molecule in an inactive conformation, and an additional destabilization of the transition state (18). There is at present no data for NGH triplets which would permit an explanation of why G is not tolerated at this position, but it might well be that there is again at least a contribution by the three H bonds to stabilization of the ground state.

Even though the rates for cleavage of NAH-containing substrates are considerably lower than for NUH substrates, we suggest, in conjunction with the data presented for cleavage of NCH triplets (4), reformulation of the NUH triplet specificity of the hammerhead ribozyme to NHH, where H can be any nucleotide except G. This expansion of the cleavage specificity increases the application of the hammerhead ribozyme for inhibition of gene expression, as more sites in the mRNA can be targeted. Even though it might be considered a marginal extension, as only a few sites of a mRNA are generally accessible for annealing of oligonucleotides, this expansion of the triplet rule could be a considerable advantage (19,20).

ACKNOWLEDGEMENTS

This work was supported in part by the Deutsche Forschungsgemeinschaft. We thank E.Westhof (Strasbourg) for discussions.