ABSTRACT
The hammerhead self-cleaving motif occurs in a variety of RNAs that infect plants and consists of three non-conserved helices connected by a highly conserved central core. A variant hammerhead, called the extended hammerhead, is found in satellite 2 transcripts from a variety of caudate amphibians. The extended hammerhead has the same core as the prototypical hammerhead, but has unusually conserved sequence and structural elements in its peripheral helices. Here we present the results of a thiophosphate substitution interference analysis of the pro-Rp phosphate oxygen requirements in the two hammerhead forms. Five pro-Rp phosphate oxygens, all in the central core, were found to be important for self-cleavage by the prototypical hammerhead. A similar set of core positions were important for self-cleavage by the extended hammerhead, but five non-core positions were also found to be important. Thiosubstitution at one of these positions had the most severe effect on self-cleavage observed in this analysis. Mn2+ did not alleviate this negative effect, indicating that this position was not part of a divalent cation binding site. We propose that novel tertiary interactions in the extended hammerhead help form the same catalytic core structure as that used by the prototypical plant virus hammerhead.
One of the smallest and most intensively studied catalytic RNA motifs is the self-cleaving hammerhead domain. Hammerheads were first identified in a comparison of the sequences surrounding the sites of self-cleavage in the virusoid of lucerne transient streak virus, the satellite of tobacco ringspot virus and the avocado sunblotch viroid (1 ) and have since been found in a diverse array of plant viroids, virusoids and viral satellite RNAs (2 ). As originally proposed, the structural model of the consensus hammerhead consists of three base paired stems that emanate from a highly conserved single-stranded core (Fig. 1 A). This model has received strong support from mutational studies (reviewed in 3 ) and is consistent with the three-dimensional structures of several hammerhead variants that were solved by physical techniques such as fluorescence resonance energy transfer (4 ) and X-ray crystallography (5 ,6 ). Despite an abundance of structural data, however, the details of the cleavage mechanism remain to be determined (reviewed in 7 ). The reaction involves nucleophilic attack on the phosphodiester at the cleavage site by the adjacent 2'-OH. The only requirement for the reaction is a divalent metal cation, which is probably involved in extracting the proton from the 2'-OH, but may also help stabilize the pentacoordinated reaction intermediate or the leaving groups. The problem with assigning specific functions to the divalent cations is that it is not clear how many ions are required for catalysis and where the catalytically important cations bind in the active structure. The stringent sequence requirements in the single-stranded central core (8 ), the clustering of important functional groups in this core (9 -14 ) and the absence of sequence specificity in the helical regions (reviewed in 3 ) suggest that the catalytically important features of the hammerhead are all located in the conserved core.
Restriction enzymes were purchased from New England Biolabs, SP6 RNA polymerase was from Epicenter Technologies, T4 polynucleotide kinase was from Promega, [[gamma]-32P]ATP was from New England Nuclear, Sp stereoisomers of nucleoside 5'-([alpha]-thio)-triphosphates (NTP[alpha]S) were from Amersham and iodoethanol was from Fisher Scientific.
EHH-c4 transcripts were transcribed from clone pC4, which was constructed previously and consists of the hammerhead domain from the N.viridescens clone pG47 with three substitutions in the stem III region that mimic the hammerhead domain from Ambystoma talpoideum (23 ). HH-pt transcripts were transcribed from clone pPt, which was constructed by mutation of A10.1 to T in the parent construct, pGES1 SLI (22 ). HH-NecO transcripts were transcribed from pNecO, which was constructed by mutation of the Necturus maculosus genomic clone pNmac7 (18 ) so that the encoded hammerhead had the sequence of the Necturus ovarian satellite 2 transcripts (Y.Zhang, S.R.Coats and L.M.Epstein, unpublished data).
Internally labeled transcripts for kinetic measurements were prepared in 10 [mu]l reactions consisting of 2 [mu]g linearized plasmid DNA, 40 mM Tris-HCl, pH 7.5, 6 mM MgCl2, 5 mM spermidine, 10 mM DTT, 250 [mu]M (each) ATP, GTP and CTP, 5 [mu]M UTP, 10 [mu]Ci [[alpha]-32P]UTP and 25 U SP6 RNA polymerase. After incubation for 60 min at 20oC, radiolabeled transcripts were purified from unincorporated label by chromatography through Sephadex G-50. Full-length transcripts were purified from 10% polyacrylamide-7 M urea gels as described previously (21 ).
For the preparation of 5'-end-labeled, partially thiosubstituted transcripts, the labeled nucleotide was omitted from the transcription mixture, the concentration of each NTP was raised to 250 [mu]M and an amount of one specific NTP[alpha]S was added that resulted in a majority of monosubstituted transcripts as determined by the generation of uniform band intensities after cleavage with iodine (see below). For ATP[alpha]S and GTP[alpha]S, these amounts gave NTP:NTP[alpha]S ratios that were equal to the number of occurrences of the nucleotide in the full-length transcript. This indicates that, as previously reported for transcription by T7 RNA polymerase (9 ,27 ), SP6 RNA polymerase did not differentiate between the normal and the thiosubstituted nucleotides during transcription. However, the optimal concentrations of CTP[alpha]S and UTP[alpha]S were approximately five times greater than those predicted by this ratio. Thus, the concentrations of thiosubstituted nucleotides that were used for the production of EHH-c4 transcripts were 14 [mu]M ATP[alpha]S, 11 [mu]M GTP[alpha]S, 89 [mu]M CTP[alpha]S and 55 [mu]M UTP[alpha]S. For the production of partially substituted HH-pt transcripts, 15 [mu]M ATP[alpha]S, 12.5 [mu]M GTP[alpha]S, 78 [mu]M CTP[alpha]S and 83 [mu]M UTP[alpha]S were optimal. After incubation for 60 min at 23oC, transcription products were dephosphorylated with calf intestinal alkaline phosphatase and 5'-end-labeled with T4 polynucleotide kinase and [[gamma]-32P]ATP as previously described (21 ). Full-length transcripts were separated from unincorporated label and purified as described above.
Transcripts were heated at 80oC for 5 min in 1 mM EDTA, pH 8.0, immediately chilled in ice/water and combined with 2 vol prewarmed buffer to initiate self-cleavage. Self-cleavage reaction mixtures (6 [mu]l) consisted of 0.1-0.2 nM transcripts, 133 mM MES buffer, pH 6.9, 30 mM MgCl2, 10 mM NaCl and 0.3 mM EDTA. After incubation at 42oC for the indicated times, 6 [mu]l gel loading buffer (80 mM EDTA, 75% formamide, 0.05% xylene cyanol, 0.05% bromophenol blue) were added to stop the reactions. Transcripts and cleavage products were separated by electrophoresis through denaturing polyacrylamide gels and dried gels were quantified with a Betascope 603 Blot Analyzer (Betagen Corp.). First order rate constants (k) and reaction end-points (EP) were determined by non-linear least square fits of the data to the equation F = EP (1 - e-kt), where F is equal to the fraction cleaved at time t (28 ).
Gel-purified, 5'-end-labeled, partially thiosubstituted transcripts were pooled and subjected to self-cleavage conditions in the presence of either 30 mM MgCl2 or 30 mM MnCl2 for a time sufficient to give ~50% cleavage. Cleavage products were separated by electrophoresis on 10% acrylamide-7 M urea gels and uncleaved transcripts and 5'-cleavage products were purified from gel slices as described (21 ). After precipitation with ethanol, RNA fractions were resuspended in 1 mM EDTA. Just prior to initiation of the iodoethanol reactions, 2 [mu]l RNA (0.002-0.004 ng) were heated at 80oC for 1 min and chilled on ice. Sulfur-specific cleavage was initiated by the addition of 1 [mu]l of a 25% solution of 2-iodoethanol in 95% ethanol. After incubation for 2 min at 90oC, reactions were stopped by adding an equal volume of stop buffer (80% formamide, 0.05% xylene cyanol, 0.05% bromophenol blue) on ice. Reaction products were separated on 12% polyacrylamide-7 M urea gels. 5'-End-labeled transcripts were also subjected to alkaline hydrolysis to generate standards for identification of the iodine cleavage products by mixing 2 [mu]l RNA (0.002-0.004 ng) with 2 [mu]l 100 mM Na2CO3/NaHCO3, pH 9.0, and boiling for 4 min.
Autoradiographs were scanned on a SM3 scanner with 63.5 [mu]m resolution (Howtek Inc.) and relative band intensities were determined by means of PDI Quantity One software (version 2.2). The relative band intensity is a measure of the intensity of a band relative to the intensity of the entire section of the lane analyzed. For each position, the relative intensity ratio was then calculated as the relative intensity of the band in the uncleaved fraction divided by the relative intensity of the band in the cleaved fraction. The bar graphs show plots of the average relative intensity ratios and their standard deviations, from a minimum of three independent experiments. For each combination of transcript and divalent cation, average relative intensity ratios were used to calculate an overall group mean and a group standard deviation after elimination of outliers from a normal distribution function using SPLUS software (Statistical Sciences Inc.). Potentially important positions were classified as exhibiting relative intensity ratios 1 or 2 SD above the group mean.
EHH-c4 transcripts were chosen to represent the extended hammerhead in the thiophosphate substitution interference analysis (Fig. 1 B). The template for these transcripts was constructed previously and consists of the hammerhead domain from N.viridescens with the A.talpoideum stem III region (23 ). Figure 2 A shows that self-cleavage by these transcripts fits well to the first order rate equation (r2 = 0.98). The rate constant determined from this data, 0.039/min, is 5-fold greater than that of any naturally occurring extended hammerhead and we reasoned that the relatively high activity of EHH-c4 transcripts would facilitate our analysis and provide a greater level of sensitivity for identifying positions that were important but not absolutely critical for self-cleavage.
Phosphorothioate linkages were incorporated into the EHH-c4 and HH-pt transcripts during in vitro transcription by use of commercially available Sp-NTP[alpha]S diastereomers as substrates. Transcription by SP6 RNA polymerase results in an inversion of configuration and the substitution of sulfur atoms in the Rp configuration for non-bridging phosphate oxygens (26 ,29 ). Ratios of NTP and NTP[alpha]S precursors were empirically determined that gave, on average, one phosphorothioate linkage per transcript. If a particular phosphate oxygen is important for cleavage, the substitution of a sulfur atom for that oxygen will inhibit cleavage and that substitution will be enriched in the uncleaved fraction. The occurrence of thiophosphate substitutions at specific positions in uncleaved full-length transcripts and 5'-cleavage fragments was determined by electrophoretic separation of iodoethanol-treated transcripts (26 ). Because the primary purpose of this study was to obtain evidence for fundamental differences between the extended and prototypical hammerheads, the analysis was designed to examine a large sample of pro-Rp phosphate oxygens in the most efficient manner possible. Restriction of the analysis to regions upstream from the cleavage site permitted the use of a single and consistent 5'-end-labeling protocol. Nucleotides near the cleavage site were excluded from the analysis because the heavily labeled 5'-self-cleavage fragment made it difficult to quantify bands representing iodine cleavage at those positions. Nucleotides near the 5'-ends were also excluded because electrophoretic run times that gave the best separation of the larger bands typically caused the smaller bands to migrate off the gels.
Figure 3 shows examples of the analysis of pro-Rp oxygens adjacent to A residues in the prototypical and extended hammerheads. In lanes 2 and 3 a more or less uniform distribution of bands resulting from iodine cleavage 5' to A residues in EHH-c4 transcripts is evident in both the cleaved and uncleaved fractions. Four additional bands occur in these same lanes, corresponding to cleavage at positions C2.4, G2.2, G5 and G8. These are pronounced examples of non-specific cleavages that occurred sporadically throughout our experiments. Cleavage at these particular positions occurred when transcripts were heated in the absence of iodine (data not shown) and bands corresponding to cleavage at these positions were occasionally enriched in the alkaline hydrolysis standards (lane 1). Moreover, the extent of cleavage at these positions was highly variable between experiments and was typically less apparent when self-cleavage reactions were performed in the presence of Mn2+ (lanes 4 and 5). Cleavages at these positions occurred to a lesser extent with transcripts with G substitutions and were undetectable with either C or U substitutions. Although there was no way to control these and similar iodine-independent cleavages, their effect on our calculations was diminished by averaging multiple, independent experiments. In fact, even though these particular bands appeared or didn't appear in unison, one of these positions, G8, was ultimately judged to be important for self-cleavage, while the others were judged to be unimportant (see below). Thus, the sporadic occurrence of iodine-independent cleavage did not affect our ability to differentiate between important and unimportant positions.
If the function of a phosphate oxygen is to coordinate a divalent cation, the negative effect of substituting sulfur for that oxygen should be alleviated if the self-cleavage reaction is performed in the presence of Mn2+, because, unlike hard metal ions such as Mg2+, Mn2+ can be coordinated by sulfur (31 ,32 ). The black bars in Figure 4 show the effects of thiophosphate substitutions in EHH-c4 transcripts when self-cleavage was performed in the presence of Mn2+. Although the overall pattern of relative intensity ratios is very similar to that generated in the presence of Mg2+, the ratios at almost every position differ slightly. Because of this inherent variation, we cannot conclude that the reductions in the presence of Mn2+ of the relatively minor effects at positions UL2.3 and A11.5 indicate divalent cation binding sites, but it is clear that the critically important oxygen adjacent to AL2.2 is not involved in divalent cation binding because the negative effect of substituting sulfur at this position was even more pronounced in the presence of Mn2+. This result differs from the finding at the other critically important position, A9. In the presence of Mn2+, the relative intensity ratio for this position is a little over half of the ratio measured in the presence of Mg2+. Although the absolute value of the ratio is still rather high (4 .6 ), it is common for Mn2+ to rescue substitutions at suspected divalent cation binding sites only partially (10 ,33 ,34 ). Despite this evidence that the pro-Rp phosphate oxygen adjacent to A9 is part of a divalent cation binding site in the extended hammerhead, partial rescue of the A9 position in HH-pt transcripts was not seen in the presence of Mn2+ (data not shown). In fact, Mn2+ had no obvious effect on any position in HH-pt transcripts (data not shown).
The chimeric hammerhead domain in EHH-c4 transcripts resembles the group A extended hammerheads found in N.viridescens and several other species from the family Salamandridae. The importance of pro-Rp oxygens in the peripheral stems of EHH-c4 transcripts prompted us to determine whether similar importance could be assigned to peripheral positions in the group B extended hammerheads. For this purpose we used transcripts that contained the hammerhead domain found in satellite 2 transcripts from N.maculosus (Y.Zhang, S.R.Coats and L.M.Epstein, unpublished data). This hammerhead has the 6 nt stem II loop that is the most distinguishing feature of the group B extended hammerheads, as well as all of the nucleotides in stems I and II that are conserved in this group (Fig. 6 ; 18 ). In our initial experiments we concentrated on stem II and the conserved core of the N.maculosus transcripts because most of the important peripheral phosphate oxygens in the EHH-c4 transcripts were found in these regions. Although no substitutions adjacent to C or U residues in these regions affected cleavage (data not shown), several substitutions adjacent to A and G residues did so (Fig. 6 ). These included positions in the conserved core that were similar to the positions identified in both EHH-c4 and HH-pt transcripts, as well as several positions in the terminal portions of stem II. Thus pro-Rp phosphate oxygens in the external loop of stem II are important for self-cleavage by examples of each of the two extended hammerhead designs.
We used a thiophosphate substitution interference analysis to compare the functional requirements in an extended hammerhead of the type found in caudate amphibian satellite 2 transcripts to those in a hammerhead with features similar to those in the prototypical hammerheads found in infectious plant virus RNAs. Four pro-Rp phosphate oxygens that were important for self-cleavage were found in identical positions in the conserved cores of both hammerhead forms and a fifth was found in similar, although not identical, positions. Three of these positions were identified as important catalytic or structural components in previous studies of prototypical hammerheads. The pro-Rp phosphate oxygen adjacent to A9 was identified in a previous thiophosphate substitution analysis (9 ) and was part of a divalent cation binding site in the crystal structure described by Pley et al. (5 ). This latter result is supported by our finding that the inhibitory effect of thiosubstitution at this position in the extended hammerhead was partially alleviated when self-cleavage took place in the presence of Mn2+. In the substituted RNA, the sulfur atom was unable to coordinate Mg2+ and self-cleavage was inhibited, but because sulfur can coordinate Mn2+, self-cleavage was restored when Mn2+ was provided. However, the pro-Rp phosphate oxygen at A9 was not a component of a similar cation binding site found in the prototypical hammerhead crystallized by Scott et al. (6 ). Although both crystal structures used N7 of G10.1 as a coordination point, coordination to A9 in the latter structure was through the pro-Sp phosphate oxygen. Although the differences in the details of these divalent cation binding sites could have resulted because Pley et al. studied the binding of Mn2+ and Cd2+ to a RNA/DNA hybrid substrate whereas Scott et al. studied the binding of Mg2+ to a predominantly RNA substrate, they might also reflect an inherent flexibility in the groups used for metal ion binding that allows hammerheads to be active with different peripheral sequences and slightly different tertiary structures. This latter explanation would also accommodate our finding that the inhibitory effect of a substitution at this position in the prototypical hammerhead was not alleviated by the presence of Mn2+.
The other two positions we identified as important, in agreement with the results of previous studies, are the pro-Rp phosphate oxygens adjacent to A13 and A14. Both were identified in the previous thiophosphate substitution analysis (9 ) and the phosphate oxygen of A13 was hydrogen bonded to the exocyclic N2 of G8 in the crystal structure of Pley et al. (see fig. 4 of ref. 5 ). This interaction helped to stabilize the non-Watson-Crick base pairing between G8 and A13, which in turn contributed to the coaxial stacking of stems II and III. This indication that the pro-Rp phosphate oxygen adjacent to A13 is involved in a structural interaction and not in metal ion binding is supported by our finding that Mn2+ did not rescue a thiophosphate substitution at this position. It is interesting that in the crystal structure solved by Scott et al., neither of the non-bridging phosphates associated with A13 is positioned to make contact with G8, a further example of the subtle differences in the structures determined in the two crystallographic studies and the subtle effects that peripheral sequences might have on the precise tertiary structure of the conserved core. It seems highly unlikely that all hammerheads use exactly the same functional groups for metal ion binding or for tertiary structural interactions. The slight differences in the positioning of important functional groups in the cores of the extended hammerhead and the various prototypical hammerheads that have been studied can therefore not be taken as proof that these cores fold or are used in dramatically different ways during catalysis. Instead, the similarities in the majority of their important core positions argue that the two hammerhead forms use remarkably similar configurations of the core for catalysis.
Despite the similarities in the core regions of the extended and prototypical hammerheads, the functional groups associated with their peripheral structures significantly differ. No pro-Rp phosphate oxygens in peripheral regions of the prototypical hammerhead proved important for cleavage or structure formation in this or any previous biochemical or physical study (5 ,6 ,9 ). In fact, active prototypical hammerheads have been formed with and without the terminal loops at the ends of the three stems (35 ,36 ) and even when the entire stem II region was replaced with non-helical nucleotide or non-nucleotide linkers (37 ,38 ). Nevertheless, five pro-Rp phosphates were identified as important for cleavage in peripheral regions of the extended hammerhead analyzed here. Although three only slightly exceeded our criterion for importance (A2.7, UL2.3 and A11.5), the other two (AL2.2 and C11.4) were among the most important positions identified in either hammerhead form analyzed in this study. In fact, thiosubstitution of the pro-Rp phosphate oxygen of AL2.2 in the external loop of stem II had a greater inhibitory effect than has been seen for any position in any hammerhead form with the exception of the pro-Rp phosphate oxygen adjacent to the cleavage site (10 ). That this position was not important in the prototypical hammerhead, even though both hammerhead forms used in this analysis had identical sequences in their stem II regions, strongly suggests that this position is involved in a long range rather than a local interaction. The importance of the stem II region in the extended hammerhead was verified when important pro-Rp phosphate oxygens were also found in stem II of the hammerhead domain in satellite 2 transcripts from N.maculosus. This hammerhead, a group B extended hammerhead, has sequences and structural details in its peripheral stems that distinguish it from the group A extended hammerhead used in the bulk of this analysis (18 ).
As has been discussed in a slightly different context (39 ), the different phosphate oxygen requirements in the prototypical and extended hammerheads might simply result from a differential ability to detect similar requirements in reactions that have different rate limiting steps. Although we would have liked to address this issue directly, the stringent requirements in stem-loops II and III of the extended hammerhead preclude the use of a trans-cleavage system to dissect the reaction, as has been done for the prototypical form (40 ,41 ). Nevertheless, we believe that the reactions investigated in this study have similar rate limiting steps. Because each hammerhead investigated has a core region that conforms to the highly conserved consensus and has functional pro-Rp phosphate oxygens in conserved positions, the cleavage step is likely to be similar in these and previously investigated hammerheads. Nevertheless, the first order rate constants determined here are almost two orders of magnitude lower than the Michaelis-Menton rate constants determined for prototypical hammerheads, in which the cleavage step was rate limiting (3 ). Product release cannot be rate limiting in first order reactions where the forward reactions are strongly favored (41 ) and where the products are analyzed under denaturing conditions. Thus, the reactions are probably inhibited before the cleavage step and the differences observed probably represent differences in precleavage assembly. Because the inhibitory effect of thiophosphate substitution at AL2.2 was not reversed by Mn2+, this position does not appear to be involved in metal ion binding. Instead, it is probably a component of one or more tertiary interactions that help form or stabilize the active structure. These interactions are likely to involve the stem I extension, because we previously showed that active extended hammerheads required compatible combinations of these two stems (23 ). The existence of specific interactions between stems I and II is entirely feasible because, in the crystal structures of the prototypical hammerheads, stems I and II are roughly parallel and close together (5 ,6 ). Thus, it is not surprising that an important pro-Rp phosphate oxygen was also found in the internal loop of stem I.
Overall, this study supports our earlier proposal that the extended hammerhead is a variant form of the prototypical hammerhead that uses novel tertiary interactions to promote and stabilize the configuration of the same catalytic central core (23 ). In the prototypical hammerhead, the core is stabilized in part by a rigid backbone formed by coaxial stacking of base pairs in stems II and III (5 ,6 ). In both extended hammerhead forms, stem III is truncated and incapable of contributing to stabilization of the central core. We suggest that in order to compensate for the lack of structural contributions by stem III, interactions have evolved between stems I and II that fix the position of these two stems in the three-dimensional structure and provide a rigid frame for nucleation of the central core. Further studies are necessary to verify this model and to define the nature of these putative interactions. In addition to providing general information about the relationship between structure and function in RNA, these studies could help in the design of maximally stable synthetic hammerheads that combine the interactions between stems I and II used in the extended hammerheads with the interactions between stems II and III used in the prototypical hammerheads.
We thankWilliam G.Scott for providing the coordinates for molecules one and two of the all-RNA hammerhead crystal structure. We also thank Shau-Ming Wu of the Statistical Consulting Center at Florida State University for his assistance with the statistical analysis of the thiophosphate substitution data. Finally, we thank Anne Thistle for her excellent editorial assistance.
REFERENCES