Translation of the Minor Capsid Protein of a Calicivirus Is Initiated by a Novel Termination-dependent Reinitiation Mechanism*

Caliciviruses represent a family of positive strand RNA viruses responsible for a variety of syndromes in man and animals. VP10, a minor structural protein of the calicivirus rabbit hemorrhagic disease virus, is encoded in the small 3′-terminal open reading frame (ORF) 2 and is translated with an efficiency of ∼20% of the preceding ORF1. The presence of the ORF1 termination codon is crucial for VP10 expression. Translation of VP10 starts at an AUG codon located at positions –5 to –3 of the ORF1 termination codon. However, VP10 was also expressed in the absence of an AUG initiation codon. The majority of ORF1 could be deleted or replaced by different sequences without significant influence on VP10 expression as long as translation terminated at the given position. The RNA sequence of the 3′-terminal 84 nucleotides of ORF1 but not the encoded peptide was found to be crucial for VP10 expression. In contrast, nearly the entire ORF2 could be replaced by a foreign sequence without abrogation of its translation. Accordingly, VP10 is expressed in a translation termination/reinitiation process that is particular because it is independent of an AUG translational start codon and requires the presence of a sequence element upstream of the initiation site.

preceding ORF1. The presence of the ORF1 termination codon is crucial for VP10 expression. Translation of VP10 starts at an AUG codon located at positions ؊5 to ؊3 of the ORF1 termination codon. However, VP10 was also expressed in the absence of an AUG initiation codon. The majority of ORF1 could be deleted or replaced by different sequences without significant influence on VP10 expression as long as translation terminated at the given position. The RNA sequence of the 3-terminal 84 nucleotides of ORF1 but not the encoded peptide was found to be crucial for VP10 expression. In contrast, nearly the entire ORF2 could be replaced by a foreign sequence without abrogation of its translation. Accordingly, VP10 is expressed in a translation termination/reinitiation process that is particular because it is independent of an AUG translational start codon and requires the presence of a sequence element upstream of the initiation site.
Translation of proteins is a fundamental process in living cells. At the step of translation initiation, the decision is made regarding which RNA is translated. Thus, the initiation process represents a key point for successful gene expression and a potential site for its regulation (recently reviewed in Refs. 1 and 2). In eukaryotic cells most initiation events conform to elements of the ribosomal scanning mechanism proposed by Kozak (3,4). Because the small ribosomal subunit starts scanning at the 5Ј-terminal cap structure of the RNA and then migrates linearly in a 3Ј direction, translation usually starts at the AUG codon closest to the 5Ј end. In higher eukaryotes the AUG has to reside in a favorable sequence context; otherwise, at least some of the ribosomal subunits can pass by and initiate at an AUG further downstream (5,6), which can result in expression of different proteins from one mRNA (7,8).
Viruses are dependent on the biochemistry of the host cell with regard to translation. To cope with the need for a 5Ј cap structure as a signal for translation initiation, viruses have developed a variety of strategies to ensure translation of the viral RNAs. Viruses like picornaviruses, hepatitis C virus, or pestiviruses use an internal ribosome entry site (IRES) 1 within the 5Ј noncoding sequence of their RNA that directly binds ribosomes, circumventing the usual interaction with the 5Јterminal cap (recently reviewed in Ref. 9). The use of alternative strategies for translation initiation allows several viruses to improve the efficiency of their gene expression by interference with the protein synthesis of the host. Moreover, regulation of viral gene expression and efficient usage of the sizelimited viral genome are often achieved by a specific strategy for translation initiation.
The family Caliciviridae comprises a group of rather poorly studied nonenveloped positive strand RNA viruses that are responsible for a variety of diseases in man and animals (reviewed in Ref. 10). Rabbit hemorrhagic disease virus (RHDV) is the causative agent of a contagious disease of high morbidity and mortality that was first described in China in 1984 (11) and then spread to different countries in Europe and Asia. The virus was accidentally transferred to the mainland of Australia in 1995 and two years later also arrived in New Zealand, where it apparently had been released deliberately to kill New Zealand's rabbits (12,13). The caliciviral genomes have a length of ϳ7.5 kb and carry the viral protein VPg (viral protein, genomelinked), which is covalently linked to the RNA 5Ј end. The viral genomic RNAs contain two or three functional open reading frames (ORFs) (reviewed in Ref. 10). In the RHDV genome, one long ORF (ORF1) is present that covers ϳ90% of the RNA and codes for the nonstructural proteins and VP60, the major capsid protein. Translation of ORF1 leads to a polyprotein of ϳ257 kDa that is mainly processed by a viral cysteine protease (14). Close to the 3Ј end of the RNA, a small second ORF (ORF2) is found that overlaps ORF1 by 17 nucleotides (15). A similar 3Ј-terminal reading frame is also present in all other known calicivirus genomes. It was first shown for RHDV and feline calicivirus (FCV) that this frame is expressed (16 -18). The translation product of the RHDV ORF2 represents a minor capsid protein of ϳ10 kDa (VP10) (17). ORF2 is found both in the genomic RHDV RNA and in a 2.2-kb subgenomic viral RNA that is colinear with the 3Ј-terminal part of the genome.
Calicivirus gene expression and replication are very efficient, resulting in high yields of viral RNAs, proteins, and infectious virus particles soon after infection. The underlying mechanisms are not well known. There are indications that VPg plays a crucial role in translation initiation, possibly serving as a substitute for the 5Ј cap structure of regular mRNAs (20,21). Even more obscure is the mechanism leading to expression of the 3Ј-terminal ORF. Additional subgenomic RNAs that might serve as mRNA for the VP10 homologue of FCV were mentioned in older publications but turned out to be artifactual (22). After in vitro translation of FCV RNA and a synthetic RNA similar to the subgenomic viral RNA, expression of the 3Ј-terminal ORF was observed (16). Similarly, heterologous expression of RHDV RNA resulted in translation of VP10 (23). Thus, it is generally accepted now that VP10 in RHDV and its homologues in other caliciviruses are translated from the genomic RNA and/or subgenomic mRNA. The present report describes experiments aiming at elucidation of the mechanism leading to expression of VP10.

EXPERIMENTAL PROCEDURES
Cells and Viruses-BHK21 cells were kindly provided by T. Rü menapf (Justus Liebig Universitä t, Giessen, Germany) and grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (tested for the absence of viruses and mycoplasma) and nonessential amino acids. Vaccinia virus MVA-T7 (modified vaccinia virus Ankara containing the gene coding for the RNA polymerase of bacteriophage T7) (24) was kindly provided by B. Moss (National Institute of Health, Bethesda, MD).
Construction of Recombinant Plasmids-Restriction and subcloning were done according to standard procedures (25). Nucleotide sequencing was done with the Big Dye Terminator cycle sequencing kit and an ABI Prism 377 DNA sequencer (PerkinElmer Life Sciences). Restriction and modifying enzymes were purchased from New England Biolabs (Schwalbach, Germany), Amersham Biosciences, Invitrogen, and Roche Applied Science. Plasmid pCITE-2A was purchased from Angewandte Gentechnik Systeme (Heidelberg, Germany). It contains a T7 RNA polymerase promoter followed by a picornavirus IRES and a multiple cloning site for insertion of sequences that should be expressed. The vector provides the AUG for translation initiation driven by the IRES.
Plasmid pRmRNA was established by inserting the BglII/EagI fragment from pR1228 (23) into pBluescript SK Ϫ (Stratagene, Heidelberg, Germany) cut with BamHI and EagI. From the resulting plasmid an HhaI/StuI fragment was isolated and inserted into pBluescript SK Ϫ EcoRI/SmaI together with oligonucleotides Ol-RmRNA/T7 ϩ and Ol-RmRNA/T7 Ϫ that had been phosphorylated with polynucleotide kinase (25). Into the resulting plasmid cut with BamHI and EagI, the BamHI/EagI fragment from the first described subclone was inserted. The oligonucleotides contained a T7 RNA polymerase promoter fused to the sequence GTGAATGTT, which was determined as the 5Ј end of the RHDV subgenomic mRNA before (15).
The constructs of the series pRmN5d were generated by insertion of the N pro -coding sequence amplified by PCR with oligonucleotides Ol-Npro5/Nco and Ol-Npro3/BamHI and cut with NcoI and BamHI together with the different RHDV cDNA fragments amplified with M13 reverse primer and oligonucleotides Ol-RmN5d1, Ol-RmN5d2, and Ol-RmN5d9, respectively, cut with BamHI and EagI into pCITE-2A cut with NcoI and EagI. The PCR was conducted in 35 cycles (30 s at 94°C, 30 s at 54°C, and 60 s at 72°C) with Vent polymerase (New England Biolabs) using the buffer and the conditions proposed by the producer. pRmN5 was established accordingly with the exception that first the N pro coding region was inserted together with the BamHI/EagI fragment from pR1228 into pCITE-2A and later on a BamHI fragment that covered the first 32 codons of the VP60 ORF derived from a fragment amplified with oligonucleotides Ol-RmN5 and Ol-RHD16 was introduced into the initial construct cut with BamHI and dephosphorylated using calf intestine alkaline phosphatase.
Constructs pRmN5d1FS, pRmN5d3FS, and pRmN5d4FS were generated by cleavage of the corresponding constructs pRmN5d1, pRmN5d3, and pRmN5d4 with BamHI, end filling with Klenow polymerase, and religation. This procedure results in a frameshift that leads to expression of N pro with carboxyl-terminal nonsense extensions of 57, 35, and 7 amino acids, respectively (translation termination occurs at stop codons located in the RHDV sequence at positions 5754 -5756, 6315-6317, and 6939 -6941 for pRmN5d1FS, pRmN5d3FS, and pRmN5d4FS, respectively (numbering according to the published sequence (19)).
Mutant pRmS2D was established by PCR with pRmRNA and primers Ol-RHD39 and Ol-R-S2D. The PCR fragment was cut with AatII and EcoRI and inserted into either pRmRNA or pRmRNA-dM, leading to pRmS2D or pRmS2D-dM. The double frameshift mutant pRm2xFS was generated in two steps of PCR-based mutagenesis. First, the nucleotide preceding AUG2 was deleted by PCR with oligonucleotides Ol-Rd7035r and Ol-RHD39. The PCR fragment was cut with EcoRI and AatII and ligated into pRmRNA restricted with the same enzymes. The resulting plasmid served as a template for two PCRs with primer pairs Ol-R-plus1R/Ol-RHD39 and Ol-R-plus1/M13 reverse. The resulting PCR fragments were cut with AatII/XhoI and XhoI/SacI, respectively, and ligated into pRmRNA cut with AatII and SacI, resulting in pRm2xFS.
To generate constructs pRm3N and pRm3dN, PCR was conducted with pRmRNA and primer pairs Ol-RHD39/Ol-VP10ct or Ol-RHD39/ Ol-VP10nt, respectively. The PCR fragments were cut with AatII and XbaI and ligated with pRmRNA AatII/SacI together with the N procoding sequence amplified with Ol-Npro/nt and Ol-Npro/ct cut with XbaI and SacI. The primers were purchased from Invitrogen. The sequences are deposited as supplemental data online.
Expression, Metabolic Labeling, and Immunoprecipitation of Proteins-Transient expression of plasmids in BHK-21 cells using vaccinia virus MVA-T7 was done as described (27). 3-4 h after transfection the cells were washed two times with label medium containing no cysteine (cysteine labeling) or no cysteine and no methionine (cysteine and methionine labeling) and incubated in this medium for 1 h. Afterward, the medium was replaced by the appropriate labeled medium containing 0.25 mCi/ml [ 35 S]cysteine or 0.25 mCi/ml of Tran 35 S-Label (ICN, Eschwege, Germany), and the cells were incubated for another 6 -20 h at 37°C. Labeled cells were washed two times with phosphate-buffered saline and frozen within the dishes. The cell extracts were prepared under denaturing conditions (17). The extracts were incubated with 5 l of undiluted serum. The precipitates were formed with cross-linked Staphylococcus aureus (28). For quantification of the expression products, a second aliquot of antiserum was added to the supernatant of the first precipitation, and residual proteins were precipitated as described above. The precipitates were combined. Analysis of the immunoprecipitated proteins was done by SDS-PAGE using Tricine-buffered gels (29). Following electrophoresis, the gels were fixed for 1 h with an aqueous solution of 30% methanol and 10% acetic acid, rinsed for 3 h in water containing 20% methanol and 3% glycerol, vacuum-dried at 60°C, and exposed to BioMax x-ray films (Kodak, Stuttgart, Germany). The antisera used for detection of RHDV proteins have been described extensively before (antisera V6 and V1 are equivalent to antisera M and N published in Ref. 17). In vitro transcription of RNA was carried out as described (27), and translation in rabbit reticulocyte lysate was done as recommended (Promega).
Protein Quantification-VP10 expression efficiency was quantified after SDS-PAGE separation of VP60 and VP10 precipitated with antisera V6 and V1, respectively. The gels were analyzed with a Fujifilm BAS-1500 phosphorus imager, and intensities of the signals were determined with TINA 2.0 software (Raytest, Straubenhardt, Germany). For evaluation of the expression efficiency of VP10 relative to VP60, the number of methionines and cysteines within the proteins was determined (12 methionines and one cysteine for VP60, and two methionines and no cysteines for VP10; this counting reflects that the amino-terminal methionine is lost from VP10 but not from VP60). Based on these data the molar ratios of the proteins were calculated from the measured radioactivity. For comparison of expression efficiencies of different constructs, the VP10 expression level of pRmRNA was defined as 100%. The amount of VP10 expression of the other constructs was normalized using the values determined for VP60 or N pro and N pro -VP60 as an internal standard. The corrected value for VP10 was then used for calculation of the expression efficiency given as percentage of the pRmRNA value. The data presented here represent the averages of at least three independent experiments.
RNA Isolation and Northern Blot-Transfection of BHK-21 cells infected with MVA-T7 was done as described above. 5 g of total RNA isolated 6 h post-transfection was analyzed by agarose gel electrophoresis, transfer to Duralon membrane (Stratagene), and hybridization with a DNA probe as described before (30). A 3Ј-terminal 0.5-kb EcoRI/EagI fragment from pRmRNA labeled with [␣-32 P]dCTP (ICN Biochemicals, Eschwege, Germany) by nick translation (nick translation kit; Amersham Biosciences) was used as a probe.

ORF2 Is Expressed in a Transient System in the Absence of Viral Nonstructural
Proteins-Like several other caliciviruses RHDV can so far not be propagated in tissue culture cells. Isolated primary rabbit hepatocytes have been productively infected in culture (18), but the system is not suitable for elaborate studies, and moreover, an RHDV infectious cDNA clone is not available. After having established a cDNA construct encompassing the complete coding region of the RHDV genome, protein expression was investigated in a transient transfection system based on vaccinia virus MVA-T7 (23). In addition to the ORF1 proteins, VP10 could also be detected. To prove that in analogy to FCV the translation of VP10 does not depend on the presence of the nonstructural proteins encoded in ORF1, the expression construct pRmRNA was established, from which an RNA colinear with the viral subgenomic mRNA up to the first nucleotides of the poly(A) tail can be transcribed from a T7 RNA polymerase promoter ( Fig. 1). RNA derived from this construct codes for the viral capsid protein VP60 and also contains the 3Ј-terminal ORF2. Transient expression via MVA-T7 results in a capped transcript because the vaccinia virus provides the enzymatic activity that synthesizes a cap analogue at the RNA transcribed in the cytoplasm (reviewed in Ref. 31). Both VP60 and VP10 were identified in extracts of the transfected cells ( Fig. 2A). Similarly, in vitro translation of a full-length RNA transcribed from pRmRNA resulted in detection of both proteins (Fig. 2B). The same was true when the viral sequence was expressed from another construct via the cell nucleus (not shown). In a Northern blot analysis, a major RHDV-specific RNA species of ϳ3 kb and a minor band of ϳ8 kb was detected in the transfected cells (Fig. 2C). The generation of two bands is probably due to the absence of a specific T7 RNA polymerase termination signal that results in incomplete termination at an as yet undefined sequence within the plasmid and partial readthrough. Importantly, RHDV RNAs smaller than the subgenomic mRNA were not found. Thus, the expression of VP10 represents an intrinsic feature of the viral mRNA and is not due, for example, to the transcription of an as yet unidentified ORF2 mRNA. This seems also true for infected hepatocytes because RNA from RHDV-infected liver tissue analyzed by Northern blots and in vitro translation studies following size fractionation gave no indication of the presence of a third viral RNA species smaller than the known subgenomic mRNA (not shown).
Quantification of the protein bands after transient expression allowed calculation of a molar ratio of the two proteins of ϳ5:1 (VP60:VP10; average of more than 25 experiments). For FCV a ratio of 10:1 was determined after in vitro translation (16). The precipitation of the proteins in our experiments was quantitative, because incubation of the supernatant of the immunoprecipitation with a further aliquot of antibody did not result in recovery of significant amounts of VP60 or VP10 (not shown). Moreover, analysis of the supernatants obtained during washing of the precipitates did not indicate considerable loss of one of the proteins. Accordingly, the result of the phosphorus imager analysis reflects the levels to which the two RHDV proteins have actually accumulated in the cells at the time of cell lysate preparation.
Translation of ORF2 Starts at an AUG Located Upstream of the ORF1 Termination Codon-Two different AUG codons located at positions Ϫ17 to Ϫ15 (AUG1) or Ϫ5 to Ϫ3 (AUG2) with regard to the ORF1 termination codon could be responsible for initiation of VP10 translation (Fig. 3A). Two mutant clones were established with either AUG1 or AUG2 exchanged for AGG or ACG, respectively. After transient expression, immunoprecipitation, and PAGE, the expression products were quantified. The intensity of the VP60 signal served as an internal standard for normalization of transfection efficiency. Calculation of the VP10 yields resulted in values of 87% of the wild type level for the AUG1 mutation and 28% for the AUG2 mutation. This result shows that AUG1 is not important when AUG2 is present. Nevertheless, AUG1 could be responsible for VP10 expression in the absence of AUG2. Such results were obtained for the M2 gene of human respiratory syncytial virus that is expressed by a termination/reinitiation mechanism (32). However, in contrast to the data published for human respiratory syncytial virus, the exchange of both AUGs in the RHDV RNA still resulted in expression of VP10 with 24% of wild type efficiency.
AUG2 could be exchanged for AUC, resulting in similar VP10 expression levels as found for the ACG mutant. In contrast, replacement of AUG2 by UGU reduced VP10 expression to nearly zero (not shown). Taken together, AUG1 has virtually no influence on VP10 expression, whereas AUG2 is important for VP10 expression but can be replaced by alternative initiation codons. This conclusion is in agreement with the fact that AUG2 but not AUG1 is conserved in the genome of European brown hare syndrome virus, a second member of the genus Lagovirus in the family Caliciviridae.
The results obtained with the mutants affecting the putative translational start codons raised the question of whether VP10 starts with the methionine encoded by AUG2. Therefore, the two internal AUG codons of ORF2 were changed by mutation to obtain a VP10-coding sequence with the AUG2 as the only position that could be labeled with a 35 S-labeled amino acid. After translation of this construct, VP10 could not be detected by immunoprecipitation (Fig. 3B), although the protein was expressed as shown by a Western blot analysis (data not shown). To get more data on the amino terminus of VP10, the codon following AUG2 was exchanged for a cysteine codon (mutant pRmS2C). In contrast to wild type VP10 (not shown) and to mutant pRmS2Y, which do not contain cysteine, the protein translated from pRmS2C was detected after [ 35 S]cysteine labeling (Fig. 3B).
If translation of ORF2 started at AUG2, the amino-terminal methionine residue of VP10 should be lost post-translationally, because it is followed by serine. According to the Sherman rules, the Met residue should be retained when the second amino acid was changed to aspartic acid, for example (33). Such a mutant was established in either the wild type sequence context (construct pRmS2D) or the ORF2 sequence depleted of the two internal AUG codons (pRmS2D-dM). After expression of these constructs, radiolabeled VP10 could be detected in both cases (Fig. 3B). Similar mean VP10 expression rates were determined for pRmS2D and pRmS2D-dM when the different numbers of labeled residues in VP10 were considered. Equivalent results were also obtained for mutants expressing proteins with tyrosine or phenylalanine following the methionine encoded by AUG2 (not shown). In comparison with pRmRNA, considerably reduced amounts of VP10 were detected for all of the mutants. This is most likely not due to reduced binding of the antibodies because the changed region was not part of the fusion protein used for antiserum production, and antibody binding was carried out with denatured protein. The reduced recovery of VP10 therefore indicates that the sequence downstream of the AUG has significant influence on ORF2 translation efficiency. In any case, the results show that the translation of ORF2 starts at AUG2. The mature VP10 no longer contains the amino-terminal methionine but starts with the amino acid translated from the first codon downstream of AUG2, which is a serine in the wild type sequence.
The Termination Signal of ORF1 Is Crucial for Expression of ORF2-To analyze the influence of the translational termination codon at the end of ORF1 on the efficiency of ORF2 expression, the UGA termination signal was exchanged for UCA, which resulted in termination of ORF1 at a UAG located 43 codons further downstream. As expected, the respective mutant expressed a slightly larger VP60 (Fig. 4). VP10 was translated from this mutant with only 6% of the wild type level. As demonstrated by RNA and VP60 quantification this result is obviously not due to considerable differences in RNA levels or a general reduction of translation efficiency and therefore results from a change in a sequence element that is specifically important for VP10 expression (RNA data not shown). Abroga-tion of translation termination represents only one possible explanation for the data. However, exchanges of the UGA for UAA or UAG codons reduced VP10 expression efficiency to only 48 or 36%, respectively, (Fig. 4), whereas all changes destroying the translational stop signal resulted in similarly dramatic effects as observed for the UCA mutant (not shown). It therefore can be concluded that the termination signal close to AUG2 is crucial for ORF2 expression.
A Region of 84 Nucleotides Containing the Termination⁄ Initiation Site Is Necessary for VP10 Expression-Translation of an ORF located far downstream of an RNAs 5Ј end can be achieved by internal entry of ribosomes mediated by an IRES in the RNA. To investigate whether in addition to the AUG and UGA codons other sequences from ORF1 or ORF2 are important for translation of the latter ORF, a set of plasmids was established with truncations of the region upstream of the termination/initiation site or most of the VP10-coding sequence replaced by a foreign sequence. To allow controlled quantification and to reduce artificial initiation at the ORF2 AUG codon in consequence of its location close to the 5Ј end of the truncated RNAs, the respective RHDV cDNA fragments were inserted downstream of a foreign sequence, resulting in expression of a fusion protein that could serve as an internal standard similar to full-length VP60 in the cases described above. The N pro -coding sequence of the pestivirus classical swine fever virus was chosen for this purpose. N pro represents a nonstructural protein of ϳ23 kDa that is located at the amino terminus of the polyprotein encoded by the long classical swine fever virus ORF. N pro has autoproteolytic activity and cleaves itself off the nascent polypeptide chain (34). At least partial cleavage was also observed for the fusion proteins expressed from the constructs tested here (Fig. 5). A dramatic drop of VP10 expression was observed for constructs with truncation of ORF1 affecting a sequence element close to the termination/initiation site. The 5Ј end of this sequence element mapped to the region between nucleotides 6958 and 6997 because construct pRmN5d5 (start at nucleotide 6958) showed VP10 expression with a level of efficiency similar to that of wild type, whereas pRmN5d6 (start at nucleotide 6997) was not able to express VP10 at all (Fig. 5A). A second set of constructs with more closely spaced truncations of the VP60 coding sequence allowed precise location of the 5Ј end of the sequence element required for VP10 expression. The difference of only 12 nucleotides between constructs pRmN5d5A and pRmN5d5D reduced VP10 expression from the approximate level of wild type to zero, and even the loss of only 1 codon in pRmN5d5B compared with pRmN5d5A lowered the VP10 yield by ϳ90% (Fig. 5B). It can therefore be concluded that the 3Ј-terminal 84 nucleotides of ORF1 including the stop codon are necessary for optimal expression of VP10. The complete default of ORF2 expression in consequence of truncations beyond nucleotide 6967 shows that this sequence element is not only modulating expression efficiency but is crucial for translation of VP10.
The results of the experiments do not discriminate between IRES-dependent translation initiation and a termination/reinitiation mechanism. Because the activity of IRES elements is not dependent on translation of the respective RNA sequence, a selected set of the constructs of the pRmN5 series was changed in a way that the N pro gene was out of frame with regard to the RHDV sequence. Expression of these constructs results in termination of translation of the protein at the first stop codon following in the new reading frame downstream of the N pro -coding sequence (Fig. 5C). After expression of these constructs, N pro and/or the N pro fusion proteins were detected, but in none of the cases was VP10 visible. The N pro yields were comparable in all cases. RNA analysis conducted for selected In the top panels, the different constructs are shown as schematic drawings (note that the different elements are not drawn to scale). In the bottom panels, autoradiographs of the products expressed from the plasmids are shown. Below the autoradiographs, the efficiency of VP10 expression is given (efficiency calculation based on expression of N pro /N pro -VP60 expression levels as internal standard). T7, T7 RNA polymerase promotor; IRES, IRES of encephalomyocarditis virus. Please note that the N pro protease cleaves at its own carboxyl terminus with different efficiency; this fact was considered when the expression efficiency was calculated for the different constructs. A, constructs expressing fusions of pestivirus amino-terminal protease N pro and amino-terminally truncated VP60. Precipitation was done with an antiserum against N pro (upper part) (34) or with an antiserum against VP10 (lower part) (17). B, similar to A but constructs of the "in frame" and "out of frame" series revealed that the result is not a consequence of different steady state levels of the transiently expressed RNA. It therefore can be concluded that translation down to the Stop/Start signal is essential for VP10 expression, which excludes the possibility that ORF2 translation is initiated by an IRES.
To determine whether the ORF2 sequence is also important for translation initiation, two constructs were established that again contained the classical swine fever virus N pro -coding sequence either fused to the 3Ј end of ORF2 as an extension (pRm3N) or replacing the VP10-coding sequence except for the 5Ј-terminal nine nucleotides (pRm3dN). Expression of ORF2 could be demonstrated for both constructs (Fig. 6). Thus, the ORF2 sequence downstream of the third codon is not essential for VP10 expression.
The RNA Sequence but Not the Peptide Encoded by the Region Upstream of ORF2 Is Essential-The above described experiments raised the question of whether the RNA sequence or the peptide encoded by the last 84 nucleotides of ORF1 are necessary for ORF2 translation. To answer this question, construct pRm2xFS was established that contains a double frameshift. The first change of the reading frame was achieved by insertion of a nucleotide upstream of the 84-nucleotide element (inserted between positions 6955 and 6956). For cloning purposes two additional nucleotide exchanges were introduced immediately downstream of the insertion site. The second frameshift resulted from deletion of residue 7036 just upstream of AUG2. In consequence of these changes, 23 of the carboxylterminal 29 amino acids of VP60 were different, but the RNA within the 84-nucleotide element was preserved except for the deletion of one residue. Importantly, the arrangement of AUG2 and the UGA termination codon in the pRm2xFS-derived RNA was the same as in the wild type sequence. After transient expression, both VP60 and VP10 were detected. VP60 with the changed carboxyl-terminal sequence exhibited a slightly reduced electrophoretic mobility, probably because of the differ-ent amino acid sequence of the protein. VP10 was translated from pRm2xFS with 55% of the wild type level (Fig. 7). It therefore can be concluded that the RNA sequence at the end of ORF1 but not the protein sequence derived from this nucleic acid is crucial for VP10 expression. The reduction in VP10 expression efficiency observed here fits well with the results of the other experiments for which sequence changes in the context of the termination/initiation region resulted in reduced yields of VP10.

DISCUSSION
For RNA viruses the development of virus-specific mechanisms of translation initiation was necessary for survival and offered the opportunity to control viral gene expression, to generate more proteins from a limited amount of genetic material, and to initiate down-regulation of host cellular protein synthesis. As shown in this report, the positive strand RNA virus RHDV has developed an unusual mechanism for the expression of its ORF2 that is located at the 3Ј end of the genomic as well as the single subgenomic viral RNA and codes for the minor capsid protein VP10. The translated region of ORF2 starts with the translational initiation codon AUG2 and overlaps with ORF1 by eight nucleotides (stop codon of ORF1 included). Expression of ORF2 occurs from the known viral RNAs and is dependent on translation of the preceding sequence and the presence of the last 84 nucleotides of ORF1. The ORF1 termination codon has a major influence on ORF2 translation efficiency, so that it also can be regarded as crucial for ORF2 expression. A further component important for the yield of VP10 is the presence of the AUG2, which, however, can be replaced by non AUG initiation codons. These requirements for VP10 expression are somewhat puzzling because they do not fit very well with one of the known mechanisms of translation initiation. The standard ribosome scanning process found for cellular mRNAs (Refs. 3 and 4; recently reviewed in Refs. 1 and 2) can be excluded because the AUG2 of ORF2 represents AUG number 28 in the viral subgenomic RNA and, moreover, is with a closer spacing of the truncations in the relevant region. Letters given below the autoradiograph in the upper part indicate the antiserum (AS) used for precipitation. V6, antiserum against VP60 (17); N, serum against N pro (34). C, similar to A but with truncated VP60-coding sequence out of frame with regard to the N pro gene. The position of the first stop codon following the N pro gene in the new reading frame is indicated. surrounded by a sequence representing a weak context for translation initiation (CUUAUGU). A nonlinear movement of the ribosome from the 5Ј end of the RNA to the ORF2 translation start site according to a "ribosomal shunting" mechanism that was demonstrated for different viral, cellular, and synthetic RNAs (35)(36)(37)(38)(39) can also be excluded on the basis of the arrangement of ORFs and the determined requirement for VP10 expression.
Internal entry of ribosomes at an IRES represents another way to initiate translation far downstream of the RNAs 5Ј end (recently reviewed in Ref. 9). The absolute dependence of RHDV ORF2 expression on the last 84 nucleotides of ORF1 could argue in favor of this mechanism. In addition, the ability to start VP10 translation in the absence of an AUG codon is similar to some types of IRES elements (40 -43). However, IRES function is not linked to translation of upstream sequences. As a matter of fact, the independence of the expression of a downstream ORF from translation of a preceding upstream ORF (uORF) in a dicistronic RNA represents a key feature of an IRES. Because translation of ORF1 is required for VP10 expression, there is no evidence for the presence of an IRES in the RHDV mRNA sequence.
A further possible mechanism for expression of VP10 could rely on ribosomal frameshifting, resulting in a fusion protein composed of VP60 and VP10 that could subsequently be cleaved by a protease into the mature viral proteins. Similar processes lead, for example, to expression of mammalian ornithine decarboxylase antizyme and the Gag-Pol fusion proteins or equivalent proteins of retroviruses or retrotransposons (reviewed in Refs. 44 and 45). Translation of the rat ornithine decarboxylase antizyme via a ϩ1 ribosomal frameshift exhibits interesting similarity with the situation found for RHDV because it is also dependent on an UGA translational stop codon (46 -48). A faint band of 70 kDa showing reactivity with both the antisera directed against VP60 and VP10 was always visible after expression of pRmRNA (data not shown). Compared with VP60 this band was generated with ϳ1% efficiency and was slightly increased after expression of constructs with mutations destroying the ORF1 translational stop codon. Therefore, the sequence surrounding the ORF1 stop codon seems to have some activity in promoting ribosomal frameshifting, but VP10 is not generated by processing of a VP70 fusion protein as demonstrated in pulse-chase experiments (data not shown). Moreover, the presence of a translational start codon is not required within a frameshifting region, but the AUG codon in close vicinity of the termination codon of the major capsid protein coding ORF is conserved among all caliciviruses (Fig.  8).
Reinitiation of translation represents another mechanism that allows ribosomes to start protein synthesis at AUG codons far downstream of the RNAs 5Ј end. In this case post-termina-tion ribosomes are believed to resume scanning and start a new initiation cycle at an appropriate AUG codon (reviewed in Refs. 6 and 49). Examples for protein expression by a termination/ reinitiation mechanism can be divided into two types depending on the length of the uORF. In sequences like the yeast GCN4 gene (reviewed in Ref. 50), the mammalian AdoMetDC gene (51), and the human cytomegalovirus gpUL4 (gp48) gene (52) short uORFs have been found to be involved in the expression of the following large ORFs. Apparently, the presence of the uORFs has mainly regulatory functions in these cases (reviewed in Refs. 49 and 53). The second type of RNAs, for which reinitiation after translation termination is discussed, contain long uORFs that code for functional proteins. Synthetic constructs (54) and the influenza B virus RNA segment 7 (55), some of the downstream ORFs in the cauliflower mosaic virus 35 S RNA (56 -58), and the ORF2 of the M2 gene of human respiratory syncytial virus (32) represent examples of this type. The two groups differ not only with regard to the size of the uORFs but apparently display also different mechanisms of translation initiation, because analyses of RNAs with short uORFs revealed that the efficiency of reinitiation decreased considerably when the length of the uORF increased (50, 59 -60). The favorable length of the uORF was in the range of less than 30 codons, and reinitiation was favored by long intercistronic regions. In contrast, the uORFs of the second group consist of much more than 30 codons, and these RNAs often display only short intercistronic regions or even contain overlaps of the upstream and downstream ORFs. It has been a matter of debate whether ribosomes can scan bidirectionally to reach AUGs located either upstream or downstream of a termination codon or whether the backscanning represents a negligible process. Further factors could be needed to allow efficient reinitiation at AUGs located upstream of the uORF termination codon. It might even be that additional factors are important for reinitiation after termination of translation of large uORFs in general. A well known example is cauliflower mosaic virus, for which reinitiation at ORFs located further downstream in the polycistronic RNA is supported by the viral protein TAV that interacts with the 60 S ribosomal subunit protein L24 and eukaryotic translation factor eIF3 (59,61).
The data reported here for RHDV allow the conclusion that a translational termination/reinitiation process leads to VP10 expression. The necessity for translation down to the end of ORF1 and the effects of mutations affecting the translational stop codon are in agreement with such a mechanism. Because the uORF consists of more than 600 codons, and the ORF2 translation initiation site is located upstream of the ORF1 termination codon, the RHDV RNAs seem most similar to the above described RNAs with long uORFs. There are, however, interesting differences with regard to the published data. First of all, the expression of VP10 is not absolutely dependent on an AUG translational start codon. This result contrasts a general model for translational reinitiation, in which the ribosome or part of it remains bound to the RNA after translation termination and starts a new initiation cycle after having recognized a nearby AUG in strong sequence context. More importantly, expression of ORF2 depends on the presence of the last 84 nucleotides of ORF1. It has never been reported before that a specific sequence element preceding the stop/start site is necessary for translation initiation according to a translation termination/reinitiation mechanism. The most probable explanation for these results is therefore that RHDV uses a novel kind of such a mechanism. This mechanism should rely on two essential elements, namely the UGA codon at the end of ORF1 that promotes translation termination and the upstream sequence element. The role of the latter sequence is difficult to interpret on the basis of the available data, but it might be that the RNA at the end of ORF1 interacts with (part of) the translation machinery in a way that after translation termination the interaction between the RNA and the ribosome or the 40 S subunit is stabilized to allow binding of the necessary factors, recognition of the initiation codon, and reinitiation of translation. The sequence could also be engaged in positioning the ribosome at the new initiation site. Binding and in some cases positioning of a ribosome also represent important functions of an IRES. The RHDV sequence element could therefore be regarded as a "termination codon upstream ribosome-binding site" (TURBS), which in contrast to the IRES elements is not capable of recruiting ribosomes de novo from the cytoplasm.
The arrangement of translation stop/start signals found in RHDV is conserved for different RHDV isolates and other members of the genus Lagovirus. Similar but different arrangements are found in other caliciviruses (Fig. 8). In all cases the 3Ј-terminal ORF overlaps the preceding ORF. However, the number of nucleotides common to both ORFs differs, ranging from eight for lagoviruses and four for vesiviruses and Sapporolike viruses to just one for Norwalk-like viruses. In the RNA of the latter viruses, the AUG of the 3Ј-terminal ORF is located downstream of the termination codon. As outlined above, the conservation of the presence of the two signals represents a strong argument for the termination/reinitiation mechanism. Moreover, the flexibility with regard to the arrangement of the two signals fits well with such a process that can be divided into two rather independent steps, namely termination and reinitiation, for which two separate signals are necessary. As expected, there is no significant homology between the different calicivirus sequences within the last ϳ100 nucleotides preceding the 3Ј-terminal ORF. Even among the different species of the genus Lagovirus, the similarity of the respective sequence is rather low, indicating that the termination codon upstream ribosomal-binding site is more likely a structural than a primary sequence element.
Another open question concerns the function of the specific mechanism leading to expression of VP10. The generation of mature calicivirus proteins relies on two different principles, namely (i) translation of polyproteins with subsequent proteolytic processing and (ii) transcription and translation of a subgenomic mRNA (10,17,23). In theory, both mechanisms could rather easily be adapted to the expression of VP10, and it will be interesting to learn why a third way was developed for this protein.
Despite the apparently unique features observed in studies on translation initiation of viral proteins, it has to be kept in mind that the viral RNAs interact with the same translation machinery that is active on cellular mRNAs. It might turn out in the future that once again the apparently unique translation initiation mechanism leading to expression of the minor capsid protein in caliciviruses is not restricted to some odd RNA viruses but is also found in the repertoire that the host cell uses to cope with the different demands made on this basic mechanism of biology.