Nucleic Acids Research

No non-hammerheads were selected

After four rounds of selection of the N18 population under non-stringent conditions there was no further enrichment in cleavage activity. Sequence analysis of this population revealed that the population was very largely composed of hammerhead-like sequences. The 6 nt linker joining A9 and G12 was essentially random. The random distribution (0.25) of identities at these positions in g4 was compromised to some extent by a high occurrence of U at position 11.1 and A at position L.3. There is also a lower than expected occurrence of G at positions 10.1, L.2, L.3 and 11.1. However, this effect was slight. The same sequence was not sampled more than once. It was therefore reasonable to maintain that the diversity of g4 approached a substantial part of the 4⁶ maximum (with some additional variation being contributed by position 7). Presumably this diversity reflected the lack of highly stringent demand on the phenotype.

None of the non-hammerhead-like sequences exhibited any cleavage activity when examined individually or as a pool. The selection was conducted in trans allowing the possibility of isolating molecules exhibiting dimer-dependent activity. Non-hammerhead sequences sampled could therefore plausibly represent truncations of potentially active structures whose additionally required components have not been observed in the cloned and sequenced sample. Alternatively, these may represent inactive molecules or `noise'. Notable also in g4 is molecule 4.31 (Fig. 3a), which incorporates an extraneous U, presumably as a bulge in helix III. An additional nucleotide can be incorporated between positions 11.1 and 12 (22) and also between 10.1 and 9 (23). In these instances, the additional nucleotide perturbs the structure of helix II, causing a reduction of, but not a complete loss of, activity. In our study, molecule 4.31 exhibited moderate activity at 1 mM Mg²⁺ (Table 1), indicating that a bulge can be accepted here. Whilst 4.31 did not exhibit a high cleavage rate constant, it is not known whether this is attributable to the extraneous nucleotide or the linker identity (10.1-11.1). Molecule 4.31 and several others have only 5 nt in the linker and generally these cleaved poorly.

Highly active miniribozymes are selected

A series of higher stringency selections yielded the sixth generation of the N18 population. In the most stringently evolved populations (g6c and g6d) the dramatic enrichment of molecules bearing the sequence motif G(10.1)NNNNC(11.1) was correlated with significantly improved cleavage rate constants for those populations (data not shown). The individual reactions (single turnover conditions) were mostly biphasic with an initial phase that was highly variable in rate constant (k₁) and extent (P₁). The relatively low extent of many of the reactions in the first phase is likely to be due to the 29mer biotinylated substrate that was used in the selection experiments. This ribozyme-substrate combination has substantial overhangs at each end which may have contributed to the propensity to form a higher than usual proportion of inactive conformers. Kinetic analysis of individual members of these populations revealed that the G(10.1)NNNNC(11.1) motif, of itself, is insufficient to confer a robust cleavage phenotype (k₁ determined at 1 mM ranged from 0.04 to 1.4 min^-1). Rather, it predisposed molecules towards having moderate to high cleavage kinetics (12 out of 17 molecules bearing this motif displayed first phase k₁ values [ge]0.5 min^-1, with only four having values [le]0.1, at 1 mM Mg²⁺, pH 8.0, 37°C). The data indicated that only a subset of linker permutations satisfy the requirement for very high cleavage activity. It is possible that the linker exerts influence on the equilibrium position between variably active conformations and that some linkers favour a more active conformation resulting in high k₁ and P₁. The rate constants and the extents of the first phase were generally reduced when Mg²⁺ was lowered from 10 to 1 mM. However, the reduction in cleavage activity was not uniform. For some molecules k₁ dropped dramatically (20- to 50-fold). In several instances this reduction involved complete extinction of the biphasic character of the reaction. At 1 mM these molecules exhibited slow monophasic kinetics. Other molecules experienced only a moderate reduction in extent and rate of the initial cleavage reaction (i.e. in the range 2- to 5-fold) when Mg²⁺ was lowered to 1 mM. A subset of linkers could therefore be identified which imparted tolerance to low concentrations of Mg²⁺ (shown above the dotted line in Table 1). This group was distinguished on the basis of exhibiting k₁ [ge] 0.5 min^-1 and P₁ > 10% at 1 mM MgCl₂, pH 8.0, 37°C.

The results of the high stringency selections clearly demonstrated that a G-C pair was required to close the loop in the linker sequence for high cleavage efficiency. That population had been arrived at through a series of selections with varying stringency and it was therefore decided to perform a highly stringent selection on a virgin population randomised only in the linker sequence between G10.1 and C11.1. The third generation (N4g3) was cloned and sequenced (Fig. 4); all sequences conformed to the 5[prime]-NNHH-3[prime] consensus observed for the most stringently selected sixth generation molecules. The set includes some molecules which were previously observed plus additional representatives. Sequence data obtained from N4g3 allowed us to recognise three distinct classes of molecule within the NNHH family. The high activity group identified among the N18g6 molecules, shown above the dotted line in Table 1, can also be ascribed to these three classes. The data contained within the low Mg²⁺ evolved N4 populations therefore supports the finding from the low Mg²⁺ evolved N18g6 populations, that there is a discrete composition of linker which is required for high activity at low Mg²⁺ concentration.

Novel classes of RNA tetraloop?

These miniribozymes are acting as monomers, as demonstrated by the lack of dependence of cleavage rate on ribozyme concentration in the range 5-100 nM and the absence of a correlation between predicted ability to dimerise and observed cleavage efficiency. Therefore, linkers supporting improved kinetic activity presumably act by providing a scaffold which will promote correct folding of the conserved G-A base pair stack which constitutes domain II in the hammerhead crystal structures (13). Our data suggest that a G(10.1).C(11.1) base pair provides a key component of this scaffold. We propose that another key component of an optimal scaffold in a miniribozyme is a linker sequence able to form a stable tetraloop between positions 10.1 and 11.1, i.e. a structure which is able to turn around a narrow enough radius without creating torsion on the complex base paired structure which stacks above it. This hypothesis is supported by the statistically high occurrence of the known tetraloop GNRA in the g6 populations (0.15 as compared to 0.03 expected random occurrence). The GNRA motif forms a sheared G-A base pair which enables it to reverse the direction of the RNA backbone within an extremely narrow radius (24). However, there is evidence that some other known stable tetraloops do not satisfy the requirement for high cleavage rates in a miniribozyme. For instance, CUUG, when incorporated into a G(10.1).C(11.1) miniribozyme, exhibited poor cleavage kinetics (11). It seems therefore that only a subset of tetraloops are able to satisfy the requirement for both stable reversal of strand direction and a scaffold which favours an active conformation of the conserved domains. The most effective molecules characterised here were GUUUUC (6.21), GGUAAC (6c10) and GUCUAC (6.14). Related molecules occurred at high frequency in the low Mg²⁺ selected N4 populations. Three classes of molecule were delineated in the N4g3 sequence data (Fig. 4). The Class I consensus (5[prime]-YRHH) includes UGAA, a known tetraloop (25). Class III has significant homology with the GNRA family. Class II does not resemble any known tetraloops and therefore may represent a novel and highly represented classification.

Generality

The effect of the evolved linkers on miniribozyme activity (high cleavage rate constants especially at low magnesium concentrations) was fully transportable to another context, as demonstrated by the activity of KrMRc10 and the comparison of its activity with that of the standard hammerhead at both 10 and 1 mM MgCl₂ (Table 2). Cleavage rate constants for conventional hammerheads on short substrates vary markedly in response to the length of helix I and appear to be highest when helix I is 5-6 bp (15). On the other hand, rate constants for miniribozymes appear to be optimal at longer arm lengths (16). Thus the superiority of the miniribozymes over the standard hammerhead even at 10 mM MgCl₂ is not necessarily in conflict with apparently contrary observations made in the context of shorter hybridising arms. Nonetheless, the difference between miniribozymes and hammerheads becomes greater when Mg²⁺ concentration is lowered.

Relation to other work

The most recent communication involving hexanucleotide replacement of helix II reported G(10.1)CGNGC(11.1) as a highly active set of motifs identified using in vitro selection (26). Rate constants for the most active motif (GCGUGC) were 0.64 min^-1 (0°C, 1 mM MgCl₂, pH 8.0) and 0.23 min^-1 (37°C, 1 mM MgCl₂, pH 8.0). Zillmann et al. (26) clearly indicate that C(L.1).G(L.4) is a preferred identity according to their scheme of selection. The differences between the Zillmann result and those reported here are probably due to the selection protocol. The selection was conducted at 0°C and with a very long helix I. Not unexpectedly, the selected ribozyme was most active at low temperatures. This illustrates the need to formulate conditions of selection which are consistent with the intended use of the molecule. At 37°C the molecules we have identified in this study cleave between two and 10 times more rapidly than those identified by Zillmann et al. Previous studies have demonstrated that the replacement of stem-loop II in the hammerhead with a 6 nt linker resulted in a reduction in catalytic activity. Tuschl and Eckstein (11) report an order of magnitude reduction in k_cat (i.e. 3.1 to 0.3 min^-1) when a 4 bp helix II was replaced by the linker sequence G(10.1)CUUGC(11.1). Similarly, Long and Uhlenbeck (12) report a 10-fold decrease in k_cis (1.0 to 0.09 min^-1) and k_cat (1.5 to 0.12 min^-1) when they compare a 4 bp helix II molecule to G(10.1)UUUGC(11.1). In both of these cases there is a G nucleotide in loop position L3 or L4, where our selection found they were not well tolerated. Additionally, their standard ribozymes had only 5 bp in each of helices I and III, a situation which favours the ribozyme over the miniribozyme (27).

Conclusion

The results demonstrate the existence of highly active miniribozymes and that there is not necessarily a reduction in activity associated with the minimisation of helix II to a single Watson-Crick base pair. The activity of a miniribozyme is dependent on linker composition. Rate constants in excess of those observed for conventional hammerhead ribozymes can be achieved by appropriately engineered linker motifs. This high activity can be maintained at physiologically relevant Mg²⁺ ion concentrations. The composition of the linker region has been thoroughly assayed in this study by means of in vitro evolution. Three classes of linker design have been elucidated via in vitro selection using low Mg²⁺ concentrations. The high activities displayed by some of these miniribozymes and the transferability of these observations into different substrate backgrounds has resulted in generalised designs for oligonucleotide reagents that retain high catalytic activity at physiological concentrations of Mg²⁺ ion. Whilst the relative activities of miniribozymes and conventional hammerheads will almost certainly vary with the length of helix I, the superiority of the miniribozyme at a physiological concentration of Mg²⁺, using helix lengths devised for optimum results on gene-length substrates, suggests that miniribozymes offer a first choice design for in vivo studies.