Transfer RNA modifications that alter +1 frameshifting in general fail to affect −1 frameshifting

INTRODUCTION

Many retroviruses express Gag-Pol or Gag-Pro-Pol proteins by coupling their translation from overlapping reading frames with −1 ribosomal frameshifts (Farabaugh 2000). The frameshift event occurs on a distinctive heptameric sequence of the general form X-XXY-YYZ, where X is any nucleotide, Y is A or U, and Z can be any nucleotide (Dinman et al. 1991; Jacks et al. 1988). From studies on the Rous sarcoma virus gag-pol frameshifting region, a pretranslocation “simultaneous slippage” model was proposed to explain the mechanism of frameshifting at such sequences (Jacks et al. 1988). According to this model, the frameshifting occurs before translocation and simultaneously by the two tRNAs present in the P- and A-sites. Later, an alternative hypothesis was presented that suggested that the frameshifting event occurs after translocation when the codons in the heptameric sequence occupy the E- and P-sites (Horsfield et al. 1995). Another type of −1 frameshifting at sites not containing this kind of heptameric sequences occurs when the ribosome is stalled at a “hungry codon” induced by, for example, aminoacyl-tRNA limitation (Weiss and Gallant 1983) or at nonsense codons (Weiss et al. 1987, 1990). The peptidyl-tRNA was proposed first to slip in the 5′ direction allowing the amino acid-tRNA in the A-site to decode the codon now at the newly formed frame by cognate interaction (Weiss et al. 1990; Yelverton et al. 1994; Barak et al. 1996).

The most energy-consuming process in the cell is translation of the genetic message transmitted via mRNA. In this process, it is pivotal for the ribosome to maintain the reading frame. Because the message lacks punctuations that would identify the reading frame, the ribosome cannot correct a reading frame error. Therefore, the translation machinery has evolved to reduce such errors to at least a magnitude lower in frequency than missense errors (Kurland et al. 1996). Transfer RNA is the molecule that decodes the message, and therefore it may be an important contributor to the reading frame maintenance. Changes in the tRNA structure, such as those induced by lack of a modified nucleoside, may therefore affect the reading frame maintenance. Modified nucleosides are derivatives of the four major nucleosides U, C, A, and G. At present, 81 different modified nucleosides have been characterized (Rozenski et al. 1999). Although they are present at different positions of the tRNA, most of them are present in the anticodon region and especially at positions 34 (wobble position) and 37 (3′ of and adjacent to the anticodon). Recently, we have demonstrated that the presence of several different modifications in the tRNA prevents +1 frameshifting (Urbonavičius et al. 2001). In that work, we also described a model explaining how frameshifting is promoted by hypomodified tRNA (see Fig. 1). This model has features in common with three of the above described models (Weiss et al. 1990; Horsfield et al. 1995; Barak et al. 1996) and with a recently described model to explain how +1 frameshifts occur (Stahl et al. 2002). We suggested that the same mechanism might also apply for −1 frameshifting errors. In the present work, we tested this hypothesis for several tRNA modifications. Surprisingly, our results demonstrate that in contrast to the +1 frameshifting (Urbonavičius et al. 2001), most tRNA modifications tested have no or in one case a slight stimulatory effect on −1 frameshifting. We suggest that in sharp contrast to the pivotal role of tRNA modification in preventing +1 frameshift errors, tRNA modification has no major role in preventing −1 frameshifting and that intrinsic features of the ribosome prevent this kind of errors.

RESULTS

The working model and the experimental systems used

Modification of tRNA may affect the accuracy of translation either by altering the rate tRNAs bind to the ribosome or by modulating the interaction of the tRNA•mRNA complex with the ribosome. Recruitment of each successive cognate, in-frame tRNA must occur in competition with all other tRNAs, including those that may decode out of frame; altering the rate of recruitment of cognate tRNA could change this competition to allow more frequent errant decoding. Alternatively, modification may modulate the interaction with the ribosome to ensure continued cognate, in-frame decoding by successive tRNAs. Figure 1 shows our model predicting how tRNA modification might influence reading frame maintenance. In all cases discussed below, the frameshift event occurs by peptidyl-tRNA slippage, although the identity of the tRNA that slips differs in each case. In the first case (Fig. 1A), hypomodification of a cognate tRNA might reduce its affinity for the codon in the ribosomal A-site. Because of its decreased rate of binding to the A-site, a wild-type near-cognate tRNA might successfully outcompete it, and be accepted into the A-site. After a normal 3-nt translocation, this near-cognate peptidyl-tRNA might not be able to interact optimally with the ribosomal P-site. Such nonoptimal P-site interaction has been suggested to reduce the efficiency of in-frame decoding leading to +1 frameshift errors (Belcourt and Farabaugh 1990; Urbonavičius et al. 2001). These errors are thought to result from slippage of peptidyl-tRNA prior to A-site recruitment. In principle, slippage in either the +1 or −1 direction might be possible, resulting in +1 or −1 frameshift errors. (P-site effect of near-cognate tRNA). In the second case (Fig. 1B), slow A-site recruitment of an undermodified cognate tRNA might induce frameshifting by causing a translational pause that allows a cognate wild-type peptidyl-tRNA in the P-site to slip either +1 or −1 (A-site effect by undermodified tRNA). Alternatively, in the third case, the defect of the undermodified cognate tRNA may not be in the A-site selection step, but rather undermodification may disrupt its interaction with the P-site after translocation (Fig. 1C). Undermodification might make the anticodon–codon interaction less optimal, as with the interaction by a near-cognate tRNA, resulting in an increased frequency of frameshifting (P-site effect by undermodified tRNA). Of course, hypomodification may cause frameshifting by mediating both A-site and P-site effects. Thus, in contrast to the simultaneous slippage model (Jacks et al. 1988), this model does not require a simultaneous slippage of both A- and P-site-located tRNAs, and the frameshifting event occurs after translocation. Our model has features in common with the model presented by Horsfield et al. (1995) and by Gallant and coworkers (Weiss and Gallant 1986).

Recently, we have demonstrated that a common function of several modified nucleosides in tRNA is to prevent +1 frameshifting (Urbonavičius et al. 2001). In the present work, we have investigated the influence of various modified nucleosides on −1 frameshifting using several assay systems (Table 2, see below). To monitor −1 frameshifting in the bacteria Escherichia coli and Salmonella enterica, the HIV gag-pol programmed frameshifting site (pHIV-1; U-UUU-UUA), the mouse mammary tumor virus (MMTV) gag-pro programmed frameshifting site (A-AAA-AAC), and the nonprogrammed argI gene frameshifting site (G-UUU-UAU) were used. In these sequences, the first triplet denotes the codon in P-site, and the last triplet denotes the codon in A-site before the shift into the −1 frame has occurred according to the simultaneous slippage model (Jacks et al. 1988). In all cases, the lacZ gene is placed downstream from a short frameshifting window in such a way that the β-galactosidase activity is a direct measurement of the frequency by which the ribosome shifts to the −1 frame within this window (Table 1). Previously, both HIV and MMTV sequences were shown to stimulate frameshifting in E. coli (Weiss et al. 1989; Yelverton et al. 1994). Furthermore, in E. coli, a −1 frameshifting site in the argI gene was also demonstrated (Fu and Parker 1994). Therefore, these assay systems are suitable to test the influence of tRNA modifications on −1 frameshifting in bacteria. In the eukaryote Saccharomyces cerevisiae, the HIV gag-pol programmed frameshifting site (U-UUU-UUA) and the L-A virus frameshifting site (G-GGU-UUA) were tested. Both HIV and L-A sequences are prone to frameshift in S. cerevisiae (Stahl et al. 1995; Kurland et al. 1996). In this case, the activity of luciferase was the measurement of frequency by which the ribosome shifts frame.

According to our model (Fig. 1), lack of ms²io⁶A37 of, for example, the tRNA^Leu_cmnm5UmAA may affect the A-site selection step of tRNAs at the HIV frameshifting sites (Table 2). This would stimulate a frameshifting event at the P-site, mediated by tRNA^Phe_GAA (A-site effect by the undermodified tRNA; Urbonavičius et al. 2001). Alternatively, the frameshifting can be stimulated by ms²io⁶A37 of, for example the tRNA^Phe_GAA reading the UUU codon at the P-site, provided that the ribosome makes a pause at the A-site (P-site effect by the undermodified tRNA). Because the modified nucleoside ms²io⁶A37 is present in both of these tRNA, the effect monitored may be either caused by an A- or a P-site effect of the undermodified tRNA or a combined effect. Using the argI site (G-UUU-UAU) and the tgt mutant, which blocks formation of queuosine, we monitored specifically the influence of Q34 deficiency in tRNA^Tyr_QUA on reading the UAU codon in this sequence. Table 2 summarizes the assay systems used and indicates which modification was tested using the various mutants defective in tRNA modification. The assay systems, which are present on plasmids, were introduced into congenic strains of either E. coli, S. enterica, or S. cerevisiae, only differing in the allelic state of genes involved in the synthesis of various modified nucleosides. The frameshifting frequencies were compared with those in the wild type.

Influence on −1 frameshifting by modification in position 34 (the wobble position)

Role of Q34

The hypermodified base 7-(((4.5-cis-dihydroxy-2-cyclopenten-1-yl)-amino) methyl)-7-deazaguanosine or queuosine (Q) is present in position 34 in tRNAs specific for Tyr, His, Asn, and Asp, encoding UAU/C, CAU/C, AAU/C, and GAU/C, respectively. In a tgt mutant, these tRNAs have an unmodified G34 instead of Q34. Of the −1 frameshifting sites available, the presence of G34 instead of Q34 may influence the A-site selection of tRNA^Asn_QUU at the A-AAA-AAC (pMMTV1284) or the A-site selection of tRNA^Tyr_QUA at the G-UUU-UAU (pCF3). To test this hypothesis, we introduced either plasmid pMMTV1284 or plasmid pCFP3 into the wild type and the tgt mutant, respectively, and measured the level of frameshifting. No difference in −1 frameshifting was observed between wild type and the tgt mutant, suggesting that the A-site selection of tRNA^Asn_QUU or tRNA^Tyr_QUA was not affected by Q34 (Fig. 2A).

Role of mnm⁵s²U34

The modified nucleoside 5-methylaminomethyl-2-thio-uridine (mnm⁵s²U) is present in position 34 in tRNAs specific for Gln, Lys, and Glu codons, which are part of the split codon boxes; that is, a codon box in which the four codons encode two amino acids. These tRNAs read codons CAA/G (Gln), AAA/G (Lys), and GAA/G (Glu). To test how the mnm⁵- or the s²- groups influence −1 frameshifting in the P-site by affecting the interaction of tRNA^Lys_mnm5s2UUU with the AAA codon, we introduced plasmid MMTV1284 (A-AAA-AAC) into wild type and mnmE (containing s²U34 instead of mnm⁵s²U34) or mnmA (containing mnm⁵U34) mutants, and measured the level of frameshifting. Lack of the mnm⁵-group, as in the mnmE mutant, reduced the rate of −1 frameshifting 1.7-fold, whereas lack of the s²-group, as in the mnmA mutant, did not affect the frequency of −1 frameshifting (Fig. 2B).

Influence on −1 frameshifting by modification in position 37

Role of ms²io⁶37

The modified nucleoside 2-(methylthio-N⁶-isopentenyl) adenosine ms²i⁶A37 is present in all E. coli tRNAs reading codons starting with U except tRNA_GGA^{Ser I, V} (Grosjean et al. 1985). In S. enterica, the hydroxylated derivative ms²io⁶A37 is present in the corresponding tRNAs. Three S. enterica genes involved in the synthesis of ms²io⁶A37 have been identified: miaA (Ericson and Björk 1986), miaB (Esberg and Björk 1995), and miaE (Persson and Björk 1993). The miaA mutant contains A37 instead of ms²io⁶A37 in its tRNA, whereas the miaB mutant tRNA lacks the ms²-group and therefore contains mainly i⁶A37 and a small amount of io⁶A37 (Esberg and Björk 1995). The miaE mutant lacks the hydroxyl-group and tRNA contains ms²i⁶A37 instead of ms²io⁶A37. We tested how the ms²io⁶- (as in miaA), ms²- (as in miaB), and o⁶- (as in miaE) groups influence the −1 frameshifting of tRNA^Leu_cmnm5UmAA, tRNA^Tyr_QUA, or tRNA^Phe_GAA using the HIV (U-UUU-UUA) or the argI (G-UUU-UAU) frameshifting sites. We introduced the plasmids containing these sites into wild type and the miaA, miaB, and miaE mutants. We also introduced the plasmid containing the argI frameshifting site into wild type and the miaA-miaB double mutant, containing A37 instead of ms²io⁶A37. We measured the level of −1 frameshifting, and no difference between wild type and any of the mutants was observed (Fig. 3A).

Role of m¹G37 and yW37

The Trm5p protein is involved in the formation of modified nucleosides m¹G and wybutosine (yW; nucleoside of the Y-base) at position 37 in a subset of tRNAs in yeast S. cerevisiae. The tricyclic nucleoside yW is present only in tRNA^Phe_GmAA, and the first step in its synthesis is the formation of m¹G37 (Droogmans and Grosjean 1987). We have earlier shown that the trm5 mutant is devoid of yW and has an unmodified G37 in its tRNA^Phe_GAA (Björk et al. 2001). The m¹G in position 37 is present in several tRNA species, including tRNA^Leu_U?AA. The trm5 mutant of S. cerevisiae contains G37 in those tRNAs.

To test how m¹G37 influences reading frame maintenance by affecting A-site selection of tRNA^Leu_U?AA, we introduced the L-A site (G-GGU-UUA) into wild type and the trm5 mutant, and measured the level of frameshifting. We found no difference in frameshifting (Fig. 3B). Thus, lack of m¹G37 in tRNA^Leu_U?AA had no influence on −1 frameshifting compared to the fully modified tRNA. The combined deficiency of yW in tRNA^Phe_GmAA and of m¹G37 in tRNA^Leu_U?AA had no impact on −1 frameshifting when monitored at the HIV-site (U-UUU-UUA) (Fig. 3B).

Influence on −1 frameshifting by other modifications outside the anticodon

Role of Ψ39

The modified nucleoside pseudouridine (Ψ) is present in the anticodon stem and loop of many bacterial tRNAs, including tRNA^Leu_cmnm5UmAA, tRNA^Asn_QUU, tRNA^Phe_GAA, and tRNA^Lys_mnm5s2UUU. The truA (hisT) gene product catalyzes the formation of Ψ in position 39 in those tRNA species. To test how Ψ39 influences reading frame maintenance by affecting A-site selection of tRNA^Leu_cmnm5UmAA and/or P-site codon–anticodon interaction of the tRNA^Phe_GAA, the HIV frameshifting site (U-UUU-UUA) was introduced into wild-type strain and the truA mutant, and the level of frameshifting was measured. We found no difference in −1 frameshifting between the wild type and the truA mutant (Fig. 4A). To test how Ψ39 influences reading frame maintenance by affecting A-site selection of tRNA^Asn_QUU and/or P-site codon–anticodon interaction of tRNA^Lys_mnm5s2UUU, the MMTV frameshifting site (A-AAA-AAC) was introduced into wild type and the truA mutant, and the level of frameshifting was measured. We observed no difference in −1 frameshifting between these strains, as in the case of pHIV-1 (Fig. 4A). Thus, Ψ39 did not influence −1 frameshifting in any case.

Role of Ψ55

The modified nucleoside Ψ55 is part of the TΨC loop and thus is present in all tRNA species of E. coli and S. typhimurium. To test whether Ψ55 influences frameshifting, either the HIV frameshifting site (U-UUU-UUA) or the MMTV frameshifting site (A-AAA-AAC) was introduced into wild type and the truB mutant, and the level of frameshifting was measured. We found no influence of Ψ55 on the frameshifting at both these frameshifting sites (Fig. 4B).

DISCUSSION

We have earlier shown that a common function for several tRNA modifications is to improve the reading frame maintenance by preventing +1 frameshift (Björk et al. 1989; Hagervall et al. 1993; Li et al. 1997; Urbonavičius et al. 2001). Contrary to this observation, here we show that several different modified nucleosides do not influence or only slightly increase the frequency of −1 frameshifting (summarized in Table 3). Taken together, these observations suggest that the mechanism by which tRNA modification exerts its effect on the reading frame maintenance is different for shifts of the reading frame in the 5′ (−1) and the 3′ (+1) direction.

The Q34 deficiency in tRNA^Asn_QUU acting at the slippery sequence A-AAA-AAC did not influence the −1 frameshifting (Fig. 2A), as was observed for +1 frameshifting (Urbonavičius et al. 2001). Similarly, the absence of Q34 in tRNA^Asn did not influence the level of −1 frameshifting at U-UUA-AAC/U sites in the cos-cells and in vitro using the rabbit reticulocyte lysate (RRL) system (Marczinke et al. 2000). Using the same in vitro system, the level of −1 frameshifting at the A-AAA-AAC site only increased 1.5-fold when the Q-deficient tRNA^Asn_GUU from yeast was added to the lysate (Carlson et al. 2001). When the effect by Q34 was analyzed in vivo using a Q34 deficient mutant of E. coli, frameshifting by tRNA^Leu_U?AA decreased 2-fold at the U-UUA-AAC site (i.e., when tRNA^Leu_U?AA interacted with the AAC codon) and increased 2-fold at the U-UUA-AAU site (i.e., when tRNA^Leu_U?AA interacted with the AAU codon; Brierley et al. 1997). The difference between those and our (Fig. 2A) results may be explained by presence of different tRNAs (tRNA^Leu_cmnm5UmAA versus tRNA^Lys_mnm5s2UUU or tRNA^Phe_GAA) in the P-site. Apparently, Q34 in tRNA^Asn_QUU interacting with the AAC codon slightly stimulates −1 frameshift errors, whereas when interacting with the AAU codon, it slightly prevents −1 frameshifts. Therefore, presence of Q34 in the tRNA stimulates −1 frameshifting when it interacts with C, whereas it prevents the same error if interacting with U as the third nucleoside of the codon. Similarly, when monitoring +1 frameshifting, Q34 prevented such errors when interacting with a U-ending codon (Urbonavičius et al. 2001). Q34 deficiency of tRNA^His does not influence −1 frameshifting on the sequence U UUC AUA in vivo in E. coli (Masucci et al. 2002), consistent with our observation and that of Brierly et al. (1997) that Q34 does not influence −1 frameshifting. Because in most cases no effect on −1 frameshift that could be attributed to Q was observed and if an effect was observed, it was small, these results show that Q34 has no major role in preventing −1 frameshifting.

Lack of either the mnm⁵- or the s²-group of mnm⁵s²U34 increases the frequency of +1 frameshifting for the tRNA^Lys_mnm5s2UUU independently if the codons AAA/G are in the A- or in the P-site (Urbonavičius et al. 2001). In contrast, frequency of −1 frameshifting decreased at the A-AAA-AAC site when tRNA^Lys_mnm5s2UUU was lacking the mnm⁵-group, whereas lack of the s²-group did not make any difference (Fig. 2B). Similarly, in E. coli, the s²- or the mnm⁵-group each has very little or no influence on −1 frameshifting at a U-UUA-AAG site, and the s²-group has no influence on −1 frameshifting at the U-UUA-AAA site (Brierley et al. 1997). However, lack of the mnm⁵-group increases the −1 frameshifting at the slippery U-UUA-AAA sites twofold (Brierley et al. 1997). Thus, contrary to the large impact of mnm⁵s²U34 to prevent +1 shifts, its role in −1 frameshifting is minor, and, in fact, in some cases the presence of it increases the −1 frameshift error.

Transfer RNA from all organisms frequently contains modified nucleosides at position 37 (3′ and adjacent to the anticodon). Two such modified nucleosides are ms²io⁶A37 and m¹G37, which both prevent +1 frameshift errors (Björk et al. 1989; Hagervall et al. 1993; Li et al. 1997; Urbonavičius et al. 2001). However, ms²io⁶A37 did not influence −1 frameshifting (Fig. 3A) at any sequence tested. We found also no effect of m¹G37 in yeast tRNA^Leu_U?AA on −1 frameshifts (Fig. 3B). Thus, whereas m¹G37 and ms²io⁶A37 are very important in preventing +1 frameshifts, they apparently have no role in preventing −1 frameshifts.

The highly complex tricyclic modified nucleoside wybutosine (yW) is present at position 37 only in eukaryotic tRNA^Phe_GmAA. Using the rabbit reticulocyte lysate (RRL) in vitro translation system, a threefold increase in −1 frameshifting at the A-AAU-UUU site was attributed to the exchange of yW37 to m¹G37 in tRNA^Phe_GmAA, reading UUU in the A-site (Carlson et al. 1999, 2001). However, the level of −1 frameshifting at U-UUU-UUU/C sites, which have codons read by tRNA^Phe_GmAA both in P- and A-sites, was the same using yW or m¹G37 containing tRNA^Phe_GmAA. Similarly, we observed in vivo no influence by yW37 on −1 frameshifting at the U-UUU-UUA site (Fig. 3B), which has UUU in the P-site, similarly to the U-UUU-UUU/C sequence. Thus, lack of yW37 of tRNA^Phe_GmAA decoding UUU in the P-site did not influence −1 frameshifting either in vivo or in vitro. Because lack of yW37 of tRNA^Phe_GmAA increased −1 frameshifting at A-AAU-UUU but not at U-UUU-UUU/C (Carlson et al. 1999, 2001), the yW37-mediated effect on tRNA^Phe_GmAA decoding UUU in the A-site may be sensitive to which tRNA is occupying the P-site (peptidyl-tRNA^Lys versus peptidyl-tRNA^Phe).

Taken together, whereas modifications such as ms²io⁶A37 and m¹G37 at position 37 have a profound influence on preventing +1 frameshifting, these modifications and yW37 have no or only a minor effect on −1 frameshift errors.

According to our model (Fig. 1; Urbonavičius et al. 2001; Stahl et al. 2002), +1 frameshifts occur after translocation and by slippage of the P-site tRNA when the A-site is empty. According to the simultaneous slippage model (Jacks et al. 1988), −1 frameshifts occur after the A-site is filled, by simultaneous slippage of the tRNAs present in both A- and P-sites, that is prior to translocation. In general, our results (Table 3) and those of others (Hagervall et al. 1993; Brierley et al. 1997; Carlson et al. 1999; Marczinke et al. 2000) suggest that tRNA modifications have no or only a minor effect on −1 frameshifts. In fact, in some cases, the presence of the modified nucleoside actually stimulates −1 frameshifting (Fig. 2B) in sharp contrast to earlier observations that the presence of several modifications in tRNA prevents +1 frameshifts (Urbonavičius et al. 2001). These results suggest that −1 and +1 frameshifting occur by distinct mechanisms. It is known that +1 frameshifting is quite sensitive to the rate of recognition of the codon in the empty A-site, for example, because of slow selection of mutant tRNAs (Li et al. 1997; Qian and Björk 1997; Urbonavičius et al. 2001) or low concentration of cognate tRNA (Belcourt and Farabaugh 1990; Farabaugh et al. 1993). Such slow A-site decoding increases the probability of a forward shift of the reading frame. In contrast, a −1 slippage by the peptidyl-tRNA would depend on a movement of the apparently strongly bound deacylated tRNA in the E-site (for a review, see Ramakrishnan 2002), suggesting that such a peptidyl-tRNA slippage is unlikely. According to the simultaneous slippage model, the E-site is empty at the time of slippage. Still, both tRNAs interact with various parts of the ribosome, making the presence of modified nucleoside of minor importance to inhibit a simultaneous slippage in the 5′ direction. Perhaps interactions between the two tRNAs, the mRNA, and the ribosome during −1 simultaneous slippage frameshifting are epistatic to small structural changes in the tRNA, such as those induced by hypomodification. Such considerations could explain our inability to show an effect of hypomodification on −1 frameshifting and why tRNA modification is more important to prevent +1 than −1 frameshift errors. Our results are therefore consistent with the predictions of the simultaneous slippage model (Jacks et al. 1988).

MATERIALS AND METHODS

Plasmids, strains, and growth conditions

Plasmids and strains, which were constructed by standard genetic methods (Davis et al. 1980), are listed in Table 1. S. enterica serovar Typhimurium strains harboring various plasmids were grown at 37°C in the rich medium NB + AV + ADE, Difco nutrient broth (0.8%; Difco Laboratories) supplemented with the aromatic amino acids (A), aromatic vitamins (V), and adenine (ADE; Davis et al. 1980). Strains of E. coli were grown at 37°C in LB medium (Bertani 1951). Strains of S. cerevisiae were grown at 30°C in SD medium supplemented with appropriate amino acids or uracil (Adams et al. 1997).

Determination of the β-galactosidase, β-lactamase, and luciferase activity

In bacteria, β-galactosidase and β-lactamase activity and the frequency of frameshifting were measured as described earlier (Hagervall et al. 1993; Li et al. 1997). In yeast, β-galactosidase and luciferase activities and the frequency of frameshifting were measured as described previously (Stahl et al. 1995). Statistical variation was calculated using the t test with two tails. A difference in the frequency of frameshifting was considered significant when p < 0.05.