key: cord-0000031-754nln40
authors: Mäkeläinen, Katri; Mäkinen, Kristiina
title: Factors affecting translation at the programmed −1 ribosomal frameshifting site of Cocksfoot mottle virus RNA in vivo
date: 2005-04-20
journal: Nucleic Acids Res
DOI: 10.1093/nar/gki521
sha: 40e41f0c52b39669dee24e37875d7a9fabc38636
doc_id: 31
cord_uid: 754nln40

The ratio between proteins P27 and replicase of Cocksfoot mottle virus (CfMV) is regulated via a −1 programmed ribosomal frameshift (−1 PRF). A minimal frameshift signal with a slippery U UUA AAC heptamer and a downstream stem–loop structure was inserted into a dual reporter vector and directed −1 PRF with an efficiency of 14.4 ± 1.9% in yeast and 2.4 ± 0.7% in bacteria. P27-encoding CfMV sequence flanking the minimal frameshift signal caused ∼2-fold increase in the −1 PRF efficiencies both in yeast and in bacteria. In addition to the expected fusion proteins, termination products ending putatively at the frameshift site were found in yeast cells. We propose that the amount of premature translation termination from control mRNAs played a role in determining the calculated −1PRF efficiency. Co-expression of CfMV P27 with the dual reporter vector containing the minimal frameshift signal reduced the production of the downstream reporter, whereas replicase co-expression had no pronounced effect. This finding allows us to propose that CfMV protein P27 may influence translation at the frameshift site but the mechanism needs to be elucidated.

The principal mechanism of translation is the accurate decoding of the triplet codon sequences in one reading frame of mRNA. Specific signals built into the mRNA sequences can cause deviations from this rule. Viruses exploit several translational 'recoding' mechanisms, including translational hopping, stop codon readthrough and programmed ribosomal frameshifting (PRF) [reviewed in (1, 2) ], for regulating the amount of proteins produced from their polyproteins. For positive-stranded RNA viruses, À1 PRF is the prevailing recoding mechanism and an essential determinant of the stoichiometry of synthesized viral proteins. Most viral À1 PRF signals are regulating the production of replication-associated proteins. Depending on the virus, the efficiency of À1 PRF can vary between 1 and 40% (3) , and changes in the efficiency can inhibit virus assembly and replication (4) (5) (6) . Therefore, À1 PRF can be regarded as a potential target for antiviral agents (4, 7) . However, the development of efficient antiviral drugs is still hindered, since little is known about the trans-acting factors and the biophysical parameters affecting the À1 PRF efficiencies. Database searches have identified putative frameshift signals from a substantial number of chromosomally encoded eukaryotic mRNAs (8) . Thus, À1 PRF may also have an impact on the complexity of the proteome of several eukaryotic organisms.

Two cis-acting signals, a slippery heptamer X XXY YYZ (the incoming reading frame indicated) and a downstream secondary structure, direct the slippage and are therefore essential for this event (9) . À1 PRF takes place after the accommodation step in the slippery sequence by simultaneous slippage of both tRNAs into the overlapping À1 frame XXX YYY (9, 10) . The sequence of the heptamer allows postslippage base-pairing between the non-wobble bases of the tRNAs and the new À1 frame codons of the mRNA. Downstream RNA secondary structures [reviewed in (11) ] force the ribosomes to pause, and place the ribosomal A-and P-sites correctly over the slippery sequence (12) . However, the pausing of the ribosomes is not sufficient for À1 PRF to occur (13) ; in fact, the duration of the halt does not necessarily correlate with the level of the À1 PRF observed (12) . Crystallographic, molecular, biochemical and genetic studies suggest that a pseudoknot restricts the movement of the mRNA during the tRNA accommodation step of elongation by filling the entrance of the ribosomal mRNA tunnel (14) . This restriction can be eased either by unwinding the pseudoknot, which allows the mRNA to move forward, or by a slippage of the mRNA one nucleotide backwards. Chemical agents such as *To whom correspondence should be addressed. Tel: +358 9 19158342 ; Fax: +358 9 19158633; Email: kristiina.makinen@helsinki.fi ª The Author 2005. Published by Oxford University Press. All rights reserved.

The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oupjournals.org antibiotics, certain mutations in the translation apparatus, and in translation elongation factors that change the translation fidelity and kinetics, have been shown to influence À1 PRF efficiency [(10,15) ; reviewed in (16) ].

The parameters known to contribute to the efficiency of À1 PRF are the sequence of the slippery heptamer, the downstream secondary structure, and the length and sequence of the spacer between the two cis-acting signals. Up-and downstream sequences such as termination codons in the vicinity of the À1 PRF signals, or even several kilobases away from them, can affect the À1 PRF efficiencies (3, (17) (18) (19) (20) (21) (22) . A specific sequence in the Barley yellow dwarf virus (BYDV) 3 0 untranslated region (UTR), 4 kb downstream from the slippage site, is vital for À1 PRF (6, 19) . A stimulating effect is achieved through the formation of a tertiary structure, where complementary nucleotides from the 3 0 UTR base pair with a single-stranded bulge in the cis-acting stem-loop (6). Human immunodeficiency virus (HIV) was also shown to require a more complex secondary structure instead of a simple stemloop for optimal À1 PRF in vivo (21, 22) . These investigations suggest that À1 PRF studies carried out with minimal frameshift signals may lead to inaccurate estimates of the stoichiometry of synthesized viral protein products during infection.

Cocksfoot mottle virus (CfMV; genus Sobemovirus) infects a few monocotyledonous plant species such as barley, oats and wheat. It has a monopartite, single-stranded, 4082 nt long, positive-sense RNA genome (23, 24) . The polyprotein of CfMV is translated from two overlapping open reading frames (ORFs) 2A and 2B by a À1 PRF mechanism (25) . In this study, we wanted to determine the in vivo À1 PRF efficiency guided by the CfMV U UUA AAC heptamer and the stem-loop structure. In addition to the minimal signal (18) , we decided to test the effect of flanking CfMV sequences for their ability to contribute to À1 PRF. We found that the surrounding viral sequences promoted more efficient À1 PRF than the minimal signal sequence in vivo when measured with the dual reporter vector system developed by Stahl et al. (26) . Therefore, we carried out an expression pattern and deletion analysis to understand the molecular basis of the observed upregulation. In addition, we critically analysed the suitability of the implemented experimental system for this type of a recoding study. An interesting possibility is that the viral proteins produced via À1 PRF could regulate À1 PRF. This hypothesis was tested by co-expressing the CfMV proteins P27 and replicase together with the dual reporter vectors.

Three regions from the CfMV polyprotein ORFs ( Figure 1A) were cloned into the NheI and BclI sites between the lacZ and the luc ORFs in pAC74 (26) . This dual reporter vector was a generous gift from Dr J. Rousset of the Universite Paris-Sud, France. The inserted sequences 1602-1720 (A region), 1386-2137 (B region) and 1551-1900 (C region), were amplified by PCR using pAB-21 as a template (18) . Primers were used to introduce NheI and BglII sites to the flanking ends of the inserts. Since NheI digestion removed lacZ ORF, it was reintroduced into the plasmids as a final cloning step. The resulting plasmids were named pAC-A, pAC-B and pAC-C.

Corresponding inframe controls, where one nucleotide was added in front of the slippery heptamer, were generated by PCR-based mutagenesis (Exsite, Stratagene) and named as pAC-Am, pAC-Bm and pAC-Cm, respectively. Deletion plasmids pAC-AB/ABm (1602-2137), pAC-AC/ACm (1602-1900), pAC-BA/BAm (1386-1720) and pAC-CA/ CAm (1551-1720) were also generated. The target sequences are shown in Figure 1B . The base numbering refers to the CfMV genome as in (23) . Transcription was driven from SV40 promoter. Plasmids encoded leucine (LEU2) and b-lactamase (ampicillin resistance) as selective markers. Plasmids were transformed into Saccharomyces cerevisiae H23 [MATa hsp150::URA3 ura3-1 his3-11 15leu2-3 112trp1-1 ade2-1 can100]. Dual reporter plasmid pAC1789 and the inframe control pAC1790 containing a 53 bp sequence from the HIV-1 frameshift region (26) were used as a positive control for monitoring the À1 PRF efficiency.

To analyse the proteins produced during À1 PRF, lacZ-A/ Am/B/Bm/C/Cm-Fluc fragments were cloned inframe with the N-terminal 6xhistidine-tag in pYES2/NT KpnI and XhoI sites (Invitrogen). Reporter fusions were amplified by PCR using pAC-A/Am, pAC-B/Bm or pAC-C/Cm as templates. The resulting plasmids were named pYES2/NT-A/ Am, B/Bm and C/Cm. Protein expression was regulated from GAL1 promoter. Two CfMV encoded proteins, P27 (C-terminal end of ORF2A) and replicase (ORF2B), were cloned into pYES2 (Invitrogen). Translation initiation codons were introduced within the oligonucleotides during PCR. The resulting plasmids were named pYES-P27 and pYES-Rep. Control plasmids, which lacked the translation initiation codons were prepared by PCR-based mutagenesis (pYES-P27DAUG and pYES-RepDAUG) and the resulting plasmids were verified by sequencing. Plasmids encoded auxotrophic marker for uracil (URA3).

All cloning steps were performed using standard protocols. Plasmids were amplified either in Escherichia coli DH5a or JM110, and purified with Qiagen columns. Inserts were verified by sequencing. Yeast transformations were done using the LiAc method (27) , and transformants were selected on a synthetic minimal defined medium (SC) lacking the corresponding auxotrophic marker(s) encoded by the used plasmid(s). Bacteria (E.coli DH5a) were grown in LB-medium containing ampicillin, whereas yeast cells were grown either in YPD, or in an SC medium. Protein expression from GAL1 promoter was repressed during growth at SC medium containing 2% glucose. Expression was induced by replacing glucose with 2% galactose and 1% raffinose.

Reporter fusions were expressed in S.cerevisiae INVSc1 (his3D1/his3D1, leu2/leu2 trp1-289/trp1-289 ura3-52/ura3-52) (Invitrogen) overnight. Protein fusions were purified in denaturing conditions using Ni-NTA agarose (Qiagen), and analysed in 6% SDS-PAGE gels. Proteins were visualized either by Coomassie staining, or by using antisera raised against the CfMV polyprotein region 1386-1724 encoding CfMV VPg (28) . Protein antibody complexes were visualized with horseradish peroxidase-conjugated anti-rabbit antibodies (Sigma) and ECL chemiluminescent reagents (Amersham).

Plasmids pYES-P27, pYES-Rep, pYES-P27DAUG, pYES-RepDAUG, or empty pYES2 were co-expressed with pAC-A or with the corresponding pAC-Am inframe control in S.cerevisiae EGY48 strain (MATa, ura3, trp1, his3, 6lexAop-LEU2) (Invitrogen). Transformants were grown overnight in SC-Leu-Ura media in non-inducing conditions, and used to inoculate induction medium. Cells were harvested at late logarithmic phase. Expression of the CfMV proteins was confirmed by western blotting using polyclonal antisera against the CfMV ORF 2a and 2b proteins (28) . Determining the enzymatic activities as described below monitored the effect of CfMV P27 and replicase on À1 PRF.

For the in vitro analysis, the lacZ-gene of pAC-A/Am, -B/Bm and -C/Cm vectors was replaced with PCR-amplified Renilla luciferase (Rluc) gene from pRLnull vector (Promega). The resulting pACRF plasmids were used as templates for PCR in order to add T7 promoter upstream of the Rluc gene. These PCR products were used for RNA synthesis with RiboMax kit (Promega). Transcripts were treated with RQ1-DNase (Promega), purified with Qiagen RNeasy columns, and quantified spectrophotometrically. The integrity of the transcripts was checked in agarose gels. In vitro translations were carried CfMV À1 PRF test and control sequences were cloned between the b-galactosidase (LacZ) and firefly luciferase (Luc) genes into a dual reporter vector pAC74 (26) . Inframe control constructs had one extra nucleotide inserted in front of the slippery heptamer, which fused the reporters into the same reading frame. Thus, translation of the inframe control results in the production of a b-galactosidase-CfMV-firefly luciferase fusion. Translation of the test constructs in the incoming 0-frame yields a b-galactosidase-CfMV fusion, whereas À1 PRF produces a b-galactosidase-CfMV-firefly luciferase fusions identical to those produced from the inframe controls. À1 PRF efficiencies were calculated from the firefly luciferase activities after b-galactosidase normalization with the given formula. (B) CfMV polyprotein is encoded by two overlapping ORFs, 2A and 2B via À1 PRF. Sequence regions tested in the dual reporter vectors for their activity to promote À1 PRF are indicated. The numbering refers to the CfMV RNA sequence as published in (23) . out in wheat germ extract (WGE) according to the manufacturer's protocols (Promega). Reactions were incubated in room temperature for 60 min, and stopped on ice prior to enzymatic measurements.

Cell cultures were started from at least three independent clones and grown until the late exponential phase. Cells were collected by centrifugation, frozen in liquid nitrogen and stored at À70 C. Bacterial cells were lysed by sonication (3 · 15 s), and yeasts by vortexing with glass beads (0.5 vol) in +4 C for 30 min. Lysates were cleared by centrifugation, and enzymatic activities were determined immediately. Total protein concentrations were measured by using a Bradford protein assay reagent (Bio-Rad). b-Galactosidase (LacZ) and firefly or Renilla luciferase (LUC or RUC) activities were measured with commercial kits from Promega according to the manufacturer's instructions. LacZ activity was determined as the colour intensity at A414 nm. Luciferase activities were measured as relative light units (RLUs) with luminometer (Biohit or ThermoLabsystems). À1 PRF efficiencies were calculated from normalized firefly luciferase activities with the following formula: [(LUC activity from the test construct)/(LacZ or RUC activity from the test construct)]/[(LUC activity from the inframe control)/(LacZ or RUC activity from the inframe control)] · 100%.

In CfMV, the motif for À1 PRF is the slippery heptamer U UUA AAC and a stem-loop structure 7 nt downstream (25) . The efficiency of À1 PRF directed by CfMV cis-acting signals was assayed in vivo using a dual reporter vector system ( Figure 1A ). Since reporters are produced from one single mRNA, factors that affect the stability of the mRNA as well as the rate of translation initiation have a similar influence on the expression of both reporters, and these variations can be monitored as changes in the activity of the upstream reporter. We quantified À1 PRF by comparing the b-galactosidase normalized firefly luciferase activities derived from the test constructs via À1 PRF to those obtained from the inframe controls, in which identical b-galactosidase-CfMV-firefly luciferase fusions are produced without À1 PRF due to the added nucleotide in front of the slippery heptamer (see Figure 1A ). Similar vectors have been shown to detect even small changes in the recoding efficiencies resulting from alterations in the cis-or trans-acting factors (26, (29) (30) (31) (32) .

Three inserts of varied lengths from the CfMV polyproteinencoding region (ORF2A/2B) were introduced between the two reporters ( Figure 1B) . The A-region, which at 119 bp was the shortest, represented approximately the minimal frameshift signal proven to be functional in vitro (18) . The longest region was the B-insert. At 752 bp, it started from the 5 0 -terminus of the 12 kDa viral genome-linked protein (VPg) gene and continued to the end of ORF2A. This region encodes CfMV protein P27 with an unknown function (28) . Since the minimal requirements for the functional frameshift signal in vivo were not known, an intermediate 349 bp C-sequence was also selected for the analysis. A well-characterized 53 bp frameshift cassette derived from HIV-1 RNA was used as a positive control. Our results regarding the HIV À1 PRF efficiency, 0.7 -0.1% in bacteria, and 4.5 -1.1% in yeast (Figure 2) , are corroborated by those published earlier (26, 33, 34) indicating that our dual reporter system was fully functional.

b-Galactosidase has been shown to retain its specific activity well, irrespective of the C-terminal fusions (35) . This is important, since the first reporter serves to control the variations among the abundance and translation rates of the studied mRNAs (26, 30) . In addition to changes in specific activities, heterologous fusions can cause alterations in the solubility and conformation, which can expose cryptic protease target sites and reduce the stability of the proteins (36) . Therefore, for a reliable quantification of À1 PRF, it was important to test that equimolar amounts of fusions produced from the corresponding test and control constructs had similar enzymatic activities. Most inframe controls and the analogous test constructs had equal absolute b-galactosidase activities ( Table 1) . Comparable results were obtained, if activities were normalized with total protein concentration (data not shown). These results indicated that the length of the fusion as such did not affect the specific activities. The b-galactosidase activity from pAC-Am inframe control was also comparable to activity obtained from an empty pAC74, where this enzyme has no fusion (data not shown). This further supported the view that the few observed variations in the b-galactosidase activities more likely resulted from the changes in translatability or stability of the transcripts. In addition to pAC-Cm, two inframe controls pAC-Am and pAC-ACm showed $25% lower b-galactosidase activities when compared to the equivalent test constructs ( Table 1 ), indicating that the productivity from these constructs was reduced. Taken together, b-galactosidase seemed to fit well to be used as the first reporter and thus normalization factor in the in vivo experiments of this study.

CfMV frameshift signals generated significant À1 PRF in yeast. À1 PRF level measured from pAC-A was 3-fold higher than from HIV RNA (Figure 2A and B) . The extent of À1 PRF directed by the minimal region A in yeast, 14.4 -1.9%, was at the same level as that reported for the CfMV minimal frameshift signal in vitro (12.7%) (18) . In contrast to our earlier in vitro observations (18) , the longer CfMV sequences upregulated À1 PRF in vivo. In yeast, the level of upregulation was 2-fold for pAC-B, the À1 PRF frequency being 26.3%, and almost 5-fold for pAC-C resulting in efficiency close to 70% (Figure 2A) , which is an extremely high value, if compared to the other values published earlier (3). CfMV frameshift signals directed À1 PRF at a lower level in bacteria than in yeast ( Figure 2B ). The extent of À1 PRF directed by region A in bacteria was 2.4 -0.7%. As in yeast, the longest B region stimulated À1 PRF 2-fold in bacteria when compared with pAC-A. However, region C did not further improve À1 PRF, but programmed À1 PRF to similar levels as pAC-B, the percentages being 4.7 -1.6% for pAC-C and 5.5 -1.5% for pAC-B.

To identify the sequence(s) responsible for the enhancement of À1 PRF in vivo, a deletion analysis was carried out. The 5 0 -or the 3 0 -sequences flanking the A-region were deleted from pAC-B/Bm or pAC-C/Cm as indicated in Figure 1B , which generated vectors pAC-AB/ABm, pAC-BA/BAm, pAC-AC/ ACm and pAC-CA/CAm. À1 PRF frequencies were determined in yeast ( Figure 2C ). Increased À1 PRF was observed in all deletion constructs in comparison to the À1 PRF directed by the A region. The BA and AB regions promoted À1 PRF as efficiently as the B region, whereas regions CA and AC were better than region A, but not as good as region B. In other words, the presence of nucleotides 1386-1720, or downstream nucleotides 1602-2137, was sufficient to increase À1 PRF to the level directed by the region B. Thus, the deletion analysis did not identify single specific sequence region as being responsible for the increased À1 PRF frequencies.

The expression pattern of the test and control constructs was analysed to understand the basis for the observed upregulation in yeast. Cassettes containing the reporters and the studied intercistronic sequences were expressed and purified as N-terminal histidine fusions. This allowed us to capture all the N-terminally intact products. The affinity-purified proteins were separated in SDS-PAGE gels, and visualized either by Coomassie staining (data not shown), or by western blotting with the CfMV-specific anti-VPg antibodies. The expected b-galactosidase-CfMV fusions terminating at the end of the 0-frame in the test constructs were detected. Also, the longer transframe b-galactosidase-CfMV-firefly luciferase fusion proteins were present in both the test and the inframe constructs ( Figure 3) . Comparison of the Coomassie-stained gels with the western blots revealed that the antisera recognized the products terminating at the CfMV-encoding regions better than the transframe products. Furthermore, the small size of the CfMV-specific region in the pYES2/NT-Am decreased the binding of the antibodies to these inframe control fusions. Thus, this data were not suitable for quantitative analysis of À1 PRF. Interestingly, an additional protein, which reacted with CfMV-specific antisera, was co-purified from the cells expressing pYES2/NT-Bm and pYES2/NT-Cm inframe controls ( Figure 3 ). The size of these fusions suggested that translation had terminated approximately at the site for À1 PRF signals. If such putative termination products were also present in cells expressing the test constructs, the correctly terminated 0-frame products in the western blots masked these products.

A closer look at the absolute b-galactosidase and firefly luciferase activities revealed that firefly luciferase expression from pAC-Cm was clearly reduced (data not shown). In fact, expression from the inframe control was comparable to the corresponding pAC-C test construct. This was also obvious when the firefly luciferase activities were normalized with the total protein amount. After setting the activity from pAC-Am to a relative value of one, the corresponding values from pAC-Bm and pAC-Cm were 0.80 and 0.28. Although the b-galactosidase measurements (Table 1) suggested that the overall translatability of the pAC-Cm mRNA was also reduced to some extent, it explained the decrease in firefly luciferase expression only partially. In the light of these findings, the extremely high À1 PRF frequency estimate calculated for the C-region could be explained with more frequent translation termination at the frameshift signals of the pAC-Cm mRNA, which reduced firefly luciferase activity in relation to b-galactosidase.

À1 PRF was also assayed in vitro in WGE. Although LacZ-encoding gene is suitable for the in vivo studies, it is an unsuitable first reporter for the in vitro determination of À1 PRF efficiencies due to its big size (30) . In good agreement with this, we observed several unexpected products in the in vitro translations programmed with LacZ-CfMV-luc mRNAs (data not shown). Renilla luciferase has been shown to retain its specific activity irrespective of the C-terminal fusions (30) . Therefore, we decided to use Rluc-CfMV-luc transcripts to determine the À1 PRF efficiencies in the cellfree system. First, we verified the suitability of Renilla luciferase for the intended in vitro experiments as described in (30) . Transcripts encoding monocistronic Renilla luciferase and Renilla luciferase fused to firefly luciferase (Rluc-Am/ Cm-luc) were mixed in different ratios and used to program the in vitro translations. Increasing concentrations of transcripts encoding the Rluc-Am-luc fusion resulted in linearly growing firefly luciferase activities. At the same time Renilla luciferase activities remained constant, which showed that its enzymatic activity was not sensitive to the C-terminal fusions ( Figure 4A ). Similar results were obtained with Rluc-Cm-luc mRNA (data not shown). À1 PRF efficiencies were then determined with transcripts that contained CfMV regions A, B and C, and their corresponding inframe controls. In all cases, slightly higher À1 PRF frequencies were obtained than in vivo. In nice correlation with the in vivo results, enhanced À1 PRF was observed with the region B, although the effect was weaker than in vivo. In this context, region C did not differ from the minimal region A in its capacity to program À1 PRF ( Figure 4B) .

The ratio between the CfMV P27 and replicase is regulated by À1 PRF during CfMV infection (28) . We studied whether these proteins could regulate the À1 PRF process. P27, replicase, or an empty expression vector was co-expressed in yeast together with the dual reporter vectors containing the minimal À1 PRF test and inframe control regions as intergenic sequences (pAC-A and -Am). P27 and replicase expression was verified by a western blot analysis ( Figure 5) . A faint band having nearly the same mobility as the replicase was detected in cells grown under repressing conditions. However, due to the small size difference, this protein was not regarded as replicase.

Enzymatic activities were measured from yeast lysates prepared from induced cultures. Measurements showed comparable levels of b-galactosidase in all the samples, indicating that P27 or replicase expression did not affect the stability of the dual reporter mRNA or the translatability of the first reporter ( Table 2 ). The effect of P27 or replicase expression was monitored by comparing the reporter activity ratios to those measured from cells harbouring the empty expression plasmids (Table 2) . Co-expression of CfMV replicase did not affect the normalized firefly luciferase expression (LUC/LacZ) from the inframe control, whereas slightly increased luciferase expression from the test construct was observed. In contrast, P27 expression reduced firefly luciferase expression both from the test and the inframe constructs. The effect was stronger in the presence of inframe control as normalized firefly luciferase levels reached only 54% of expression measured from the empty vector control.

To verify that the observed differences in firefly luciferase production depended on the studied CfMV proteins, we co-expressed the dual reporter vectors with plasmids having the first translation initiation codons of P27 and replicase deleted (pYES-P27DAUG and pYES-RepDAUG). Western blot analysis with antisera against ORF2A or 2B did not detect any proteins produced from these vectors (data not shown). The obtained LUC/LacZ ratios were compared to those measured from cells expressing the CfMV proteins (pYES-P27 or pYES-Rep). LUC/LacZ ratios measured from cells expressing replicase were slightly lower than the ratios calculated from cells harbouring pYES-RepDAUG plasmids, being $90% when co-expressed with pAC-A and $84% when co-expressed with pAC-Am. In the presence of P27, LUC/ LacZ ratio of pAC-A reached $81% of expression measured from cells transformed with pYES-P27DAUG. Again the effect of P27 expression was more evident with pAC-Am inframe control as P27 expression reduced LUC/LacZ ratio to half ($48%) when compared to the corresponding value measured from the cells harbouring pYES-P27DAUG. This verified that CfMV P27 was able to reduce the downstream reporter expression from dual reporter mRNAs. Since CfMV P27 had a proportionally stronger effect to firefly luciferase production from the inframe control mRNAs in comparison to the test mRNAs (Table 2) , the calculated À1 PRF efficiency increased from 14.7 to 22.4%.

Since À1 PRF studies are affected by a huge number of different parameters, it is not an easy task to determine the real ratio between the proteins produced via this mechanism in vivo. However, in viral systems, the efficiency of À1 PRF is an essential determinant of the stoichiometry of synthesized viral protein products, which must be rigidly maintained for efficient propagation of the virus. For example, frameshifting in retroviruses determines the ratio of structural (Gag) to enzymatic (Gag-Pol) proteins, and plays a critical role in viral particle assembly (5) . In this study, the capacity of CfMV frameshift signals to direct efficient À1 PRF was analysed in vivo by using dual reporter vectors. The length of the CfMV sequence clearly affected the actual efficiency percent in vivo. The PRF efficiency was elevated when longer viral sequences were directing the À1 PRF, but the deletion analysis did not identify any specific region as being solely responsible for the enhancement. Up-and downstream sequences nearby or far away from the cis-acting signals have been reported to enhance À1 PRF in other viruses, such as HIV, human T-cell leukaemia virus and BYDV (6, 19, 20) . Also out-of-frame stop codons have been shown to influence À1 PRF frequency in vitro in retroviruses (17) and in CfMV (18) . A study on the spacer sequences located between the cis-acting signals showed that high slippage frequencies were obtained when the first three nucleotides were G/U, G/A and G/A, the first two being the most important (37) . In CfMV, the spacer starts with UAC, which partially explains the capacity of the CfMV sequence to promote high slippage levels. In this study, the observed enhancement of À1 PRF was, however, caused by sequences that were not in the immediate vicinity of the Figure 5 . Co-expression of CfMV P27 or replicase simultaneously with the minimal frameshift signal construct pAC-A or the corresponding inframe control pAC-Am in yeast. Yeast total protein samples were separated in 12% SDS-PAGE gels, transferred onto PVDF membranes, and immunocomplexes detected by ECL chemiluminescent system. CfMV P27 expression was verified by western blotting with antisera raised against ORF2A (A), and CfMV replicase expression was detected with antisera raised against ORF2B (B). Abbreviations: À, repressed; +, induced; C1, pMAL-VPg $53 kDa; and C2, baculovirus expressed CfMV replicase. slippery sequence thus indicating that CfMV sequences further away also have an influence on the level of frameshifting in vivo. We conclude that the most reliable estimates for À1 PRF and consequently for the amount of replicase versus the 0-frame translation product P27 can be obtained only by using the full-length viral sequences. In reality, such a study would however be hampered by the non-quantitative nature of the western blot analysis, the presence of different polyprotein processing intermediates, and the differences in the stabilities of the end products in the infected cells.

The overall competence of CfMV signals to direct À1 PRF was high, when compared to related plant viruses, such as Potato leaf roll virus and BYDV. À1 PRF values of $1% have been reported for these viruses when measured with reporter-based assays (6, 38) . We can hypothesize that one reason for the high efficiency is the slippery tRNA Asn encoding the AAC triplet of the CfMV heptamer. Equal U UUA AAC slippery heptamer has been measured to induce 20-40% of À1 PRF in a diversity of animal viruses [(39); reviewed in (3)]. The low fitness of CfMV À1 PRF signals in bacteria is in agreement with the poor functioning of the eukaryotic slippery heptamers of the order X XXA AAC in prokaryotes (40) (41) (42) . IBV RNA, having an identical shifty heptamer, has been shown to direct À1 PRF at similar 2-3% level in bacteria (41) . A recent study reported that XXXAAAC heptamers dictate À1 PRF to occur via the slippage of two adjacent tRNAs placed over the heptamer, irrespective of whether the host is an eukaryote or a prokaryote (42) . Therefore, the inability of prokaryotic translation systems to direct efficient À1 PRF from this heptamer is not an inherited property of prokaryotic tRNA Asn , but results from differences in the ribosomes (42) .

Paused ribosomes can pass the À1 PRF site by À1 frameshifting, resumption of 0-frame translation, or termination (43) . Transient polypeptide intermediates that result from the pausing of ribosomes in the slippery sequences have been observed during IBV and S.cerevisiae L-A virus polyprotein synthesis (12, 13, 43, 44) . A pseudoknot structure formed by IBV mRNA causes a translational pause at fixed position upstream the secondary structure regardless of whether the slippery heptamer is present or absent (12) . Based on the findings of this study, we propose that also here a certain percent of ribosomes stalled at the secondary structure of the frameshift site in our inframe control and test mRNAs in yeast, and this led to the prematurely terminated products observed with the inframe control constructs pAC-Bm and -Cm. Although not unambiguously proven by this study, high frequency of termination of translation especially at the frameshift site of the pAC-Cm mRNA would nicely explain the extremely high calculated À1 PRF efficiency.

Factors that change the translation fidelity and kinetics have been shown to influence À1 PRF efficiency [ (10, 15) ; reviewed in (16) ]. Autoregulation of +1 frameshifting by mammalian ornithine decarboxylase antizyme has been reported (45) . This mechanism allows modulation of frameshifting frequency according to the cellular concentration of polyamines. One could speculate that such a regulation mechanism could also be useful to adjust the amounts of the replicationassociated proteins to match the requirements of different phases in viral replication cycle. This hypothesis was studied by expressing CfMV proteins P27 and replicase together with pAC-A and pAC-Am in yeast cells. Since b-galactosidase production remained constant regardless of the presence or absence of CfMV proteins, they did not interfere with translation initiation from pAC-A/Am mRNAs per se. However, P27 expression caused a reduction in the firefly luciferase production especially from the inframe control, whereas replicase production only slightly increased the firefly luciferase production from pAC-A, but not from pAC-Am. Since replicase expression had only a faint effect on the normalized firefly luciferase production via À1 PRF, our conclusion is that CfMV replicase had no pronounced effect on translation at the frameshift site. Co-expression of the non-translatable form of P27 with the dual reporter vectors verified that P27 truly affected firefly luciferase expression on the protein level. Therefore, we propose that CfMV protein P27 may influence translation at the frameshift site. If CfMV P27 indeed interferes with viral protein synthesis during CfMV infection, the mechanism, its specificity and the possible biological role needs to be elucidated in the future.

Programmed translational frameshifting

Recoding: dynamic reprogramming of translation

Ribosomal frameshifting viral RNAs

Ribosomal frameshifting efficiency and gag/gag-pol ratio are critical for yeast M1 double-stranded RNA virus propagation

Importance of ribosomal frameshifting for human immunodeficiency virus type 1 particle assembly and replication

A À1 ribosomal frameshift element that requires base pairing across four kilobases suggests a mechanism of regulating ribosome and replicase traffic on a viral RNA

Translating old drugs into new treatments: ribosomal frameshifting as a target for antiviral drugs

Identification of putative programmed À1 ribosomal frameshift signals in large DNA databases

Signals for ribosomal frameshifting in the Rous sarcoma virus gag-pol region

Peptidyl-transferase inhibitors have antiviral properties by altering programmed À1 ribosomal frameshifting efficiencies: development of model systems

Structure, stability and function of RNA pseudoknots involved in stimulating ribosomal frameshifting

Ribosomal pausing at a frameshifter RNA pseudoknot is sensitive to reading phase but shows little correlation with frameshift efficiency

Ribosomal movement impeded at a pseudoknot required for frameshifting

The 9-Å solution: how mRNA pseudoknots promote efficient programmed À1 ribsomal frameshifting

Decreased peptidyltransferase activity correlates with increased programmed À1 ribosomal frameshifting and viral maintenance defects in the yeast Saccharomyces cerevisiae

An 'integrated model' of programmed ribosomal frameshifting

Suppression of translation frameshift by upstream termination codon

Regulation of À1 ribosomal frameshifting directed by Cocksfoot mottle sobemovirus genome

A sequence required for À1 ribosomal frameshifting located four kilobases downstream of the frameshift site

Comparative mutational analysis of cis-acting RNA signals for translational frameshifting in HIV-1 and HTLV-2

The frameshift signal of HIV-1 involves a potential intramolecular triplex RNA structure

Characterization of the frameshift stimulatory signal controlling a programmed À1 ribosomal frameshift in the human immunodeficiency virus type 1

Characterization of cocksfoot mottle sobemovirus genomic RNA and sequence comparison with related viruses

Nucleotide sequence of RNA from the sobemovirus found in infected cocksfoot shows a luteovirus-like arrangement of the putative replicase and protease genes

The putative replicase of the cocksfoot mottle sobemovirus is translated as a part of the polyprotein by À1 ribosomal frameshift

Versatile vectors to study recoding: conservation of rules between yeast and mammalian cells

Improved method for high efficiency transformation of intact yeast cells

Characterization of VPg and the polyprotein processing of Cocksfoot mottle virus

In vivo HIV-1 frameshifting efficiency is directly related to the stability of the stem-loop stimulatory signal

A dual-luciferase reporter system for studying recoding signals

Nonsense-mediated decay mutants do not affect programmed À1 frameshifting

An in vivo dual-luciferase assay system for studying translational recoding in the yeast Saccharomyces cerevisiae

E.coli ribosomes re-phase on retroviral frameshift signals at rates ranging from 2 to 50 percent

The function of a ribosomal frameshifting signal from human immunodeficiency virus À1 in Escherichia coli

Tolerance of Escherichia coli beta-galactosidase C-terminus to different-sized fusions

The position of the heterologous domain can influence the solubility and proteolysis of b-galactosidase fusion proteins in E.coli

Towards a computational model for À1 eukaryotic frameshifting sites

Ribosomal frameshifting in plants: a novel signal directs the À1 frameshift in the synthesis of the putative viral replicase of potato leafroll luteovirus

An efficient ribosomal frame-shifting signal in the polymerase-encoding region of the coronavirus IBV

Differential response to frameshift signals in eukaryotic and prokaryotic translational systems

Expression of coronavirus ribosomal frameshift signal in Escherichia coli: influence of tRNA anticodon modification on frameshifting

Prokaryotic-style frameshifting in a plant translation system: conservation of an unusual single-tRNA slippage event

Kinetics of ribosomal pausing during programmed À1 translational frameshifting

Ribosomal pausing during translation of an RNA pseudoknot

Autoregulatory frameshifting in decoding mammalian ornithine decarboxylase antizyme

We would like to thank Dr Jean-Pierre Rousset of the Université Paris-Sud, France, for the plasmids pAC74, pAC1789 and pAC1790, Dr Deyin Guo for helping with the yeast work, Drs Mikko Frilander and Konstantin Ivanov for critical reading of the manuscript, as well as Jimmy Lucchesi and Prof. Mart Saarma for valuable discussions. This work was supported by the National Technology Agency of Finland (grants 40673/98, 40680/01, 40014/04), the University of Helsinki (grant 2109007) and the Academy of Finland (grants 52265, 206870). Funding to pay the Open Access publication charges for this article was provided by the National Technology Agency of Finland.

Conflict of interest statement. None declared.