[From fowl plague to influenza pandemic; a reason for taking precaution]. / Van vogelpest tot influenzapandemie; reden tot voorzorgen.

Throughout Eastern Asia, there is currently an epidemic of fowl plague or highly pathogenic avian influenza, on an unprecedented scale. The prospects for rapid containment are poor. The causative virus, influenza A of the H5N1 subtype, is of limited infectivity for humans. If infection occurs, however, then the consequences are serious and even fatal in a majority of cases. In view of the receptor specificity of avian influenza viruses, this may be related to individually increased susceptibility, which does not lead to further spread. However, it is known that influenza A viruses can readily adapt to replication in the human host by the acquisition of specific gene segments or even by mutations of the avian virus. The extreme scale of human contact with influenza virus of the H5N1 subtype at present engenders fear that there is a high risk of such adaptation and a subsequent pandemic spread. Adequate precautions are necessary, not only in terms of an acceleration of vaccine production but primarily in arranging for sufficient availability of the new antiviral drugs.

Sequence requirements for RNA strand transfer during nidovirus discontinuous subgenomic RNA synthesis.

Nidovirus subgenomic mRNAs contain a leader sequence derived from the 5' end of the genome fused to different sequences ('bodies') derived from the 3' end. Their generation involves a unique mechanism of discontinuous subgenomic RNA synthesis that resembles copy-choice RNA recombination. During this process, the nascent RNA strand is transferred from one site in the template to another, during either plus or minus strand synthesis, to yield subgenomic RNA molecules. Central to this process are transcription-regulating sequences (TRSs), which are present at both template sites and ensure the fidelity of strand transfer. Here we present results of a comprehensive co-variation mutagenesis study of equine arteritis virus TRSs, demonstrating that discontinuous RNA synthesis depends not only on base pairing between sense leader TRS and antisense body TRS, but also on the primary sequence of the body TRS. While the leader TRS merely plays a targeting role for strand transfer, the body TRS fulfils multiple functions. The sequences of mRNA leader-body junctions of TRS mutants strongly suggested that the discontinuous step occurs during minus strand synthesis.

Construction of chimeric arteriviruses reveals that the ectodomain of the major glycoprotein is not the main determinant of equine arteritis virus tropism in cell culture.

The recent development of arterivirus full-length cDNA clones makes possible the construction of chimeric arteriviruses for fundamental and applied studies. Using an equine arteritis virus (EAV) infectious cDNA clone, we have engineered chimeras in which the ectodomains of the two major envelope proteins, the glycoprotein GP(5) and the membrane protein M, were replaced by sequences from envelope proteins of related and unrelated RNA viruses. Using immunofluorescence microscopy, we monitored the transport of the hybrid GP(5) and M proteins to the Golgi complex, which depends on their heterodimerization and is a prerequisite for virus assembly. The only viable chimeras were those containing the GP(5) ectodomain from the porcine (PRRSV) or mouse (LDV) arteriviruses, which are both considerably smaller than the corresponding sequence of EAV. Although the two viable GP(5) chimeras were attenuated, they were still able to infect baby hamster kidney (BHK-21) and rabbit kidney (RK-13) cells. These cells can be infected by EAV, but not by either PRRSV or LDV. This implies that the ectodomain of the major glycoprotein GP(5), which has been postulated to be involved in receptor recognition, is not the main determinant of EAV tropism in cell culture.

The influence of downstream protein-coding sequence on internal ribosome entry on hepatitis C virus and other flavivirus RNAs.

Some studies suggest that the hepatitis C virus (HCV) internal ribosome entry site (IRES) requires downstream 5' viral polyprotein-coding sequence for efficient initiation of translation, but the role of this RNA sequence in internal ribosome entry remains unresolved. We confirmed that the inclusion of viral sequence downstream of the AUG initiator codon increased IRES-dependent translation of a reporter RNA encoding secretory alkaline phosphatase, but found that efficient translation of chloramphenicol acetyl transferase (CAT) required no viral sequence downstream of the initiator codon. However, deletion of an adenosine-rich domain near the 5' end of the CAT sequence, or the insertion of a small stable hairpin structure (deltaG = -18 kcal/mol) between the HCV IRES and CAT sequences (hpCAT) substantially reduced IRES-mediated translation. Although translation could be restored to both mutants by the inclusion of 14 nt of the polyprotein-coding sequence downstream of the AUG codon, a mutational analysis of the inserted protein-coding sequence demonstrated no requirement for either a specific nucleotide or amino acid-coding sequence to restore efficient IRES-mediated translation to hpCAT. Similar results were obtained with the structurally and phylogenetically related IRES elements of classical swine fever virus and GB virus B. We conclude that there is no absolute requirement for viral protein-coding sequence with this class of IRES elements, but that there is a requirement for an absence of stable RNA structure immediately downstream of the AUG initiator codon. Stable RNA structure immediately downstream of the initiator codon inhibits internal initiation of translation but, in the case of hpCAT, did not reduce the capacity of the RNA to bind to purified 40S ribosome subunits. Thus, stable RNA structure within the 5' proximal protein-coding sequence does not alter the capacity of the IRES to form initial contacts with the 40S subunit, but appears instead to prevent the formation of subsequent interactions between the 40S subunit and viral RNA in the vicinity of the initiator codon that are essential for efficient internal ribosome entry.

Genetic manipulation of arterivirus alternative mRNA leader-body junction sites reveals tight regulation of structural protein expression.

To express its structural proteins, the arterivirus Equine arteritis virus (EAV) produces a nested set of six subgenomic (sg) RNA species. These RNA molecules are generated by a mechanism of discontinuous transcription, during which a common leader sequence, representing the 5' end of the genomic RNA, is attached to the bodies of the sg RNAs. The connection between the leader and body parts of an mRNA is formed by a short, conserved sequence element termed the transcription-regulating sequence (TRS), which is present at the 3' end of the leader as well as upstream of each of the structural protein genes. With the exception of RNA3, only one body TRS was previously assumed to be used to join the leader and body of each EAV sg RNA. Here we show that for the synthesis of two other sg RNAs, RNA4 and RNA5, alternative leader-body junction sites that differ substantially in transcriptional activity are used. By site-directed mutagenesis of an EAV infectious cDNA clone, the alternative TRSs used to generate RNA3, -4, and -5 were inactivated, which strongly influenced the corresponding RNA levels and the production of infectious progeny virus. The relative amounts of RNA produced from alternative TRSs differed significantly and corresponded to the relative infectivities of the virus mutants. This strongly suggested that the structural proteins that are expressed from these RNAs are limiting factors during the viral life cycle and that the discontinuous step in sg RNA synthesis is crucial for the regulation of their expression. On the basis of a theoretical analysis of the predicted RNA structure of the 3' end of the EAV genome, we propose that the local secondary RNA structure of the body TRS regions is an important factor in the regulation of the discontinuous step in EAV sg mRNA synthesis.

Efficient homologous RNA recombination and requirement for an open reading frame during replication of equine arteritis virus defective interfering RNAs.

Equine arteritis virus (EAV), the prototype arterivirus, is an enveloped plus-strand RNA virus with a genome of approximately 13 kb. Based on similarities in genome organization and protein expression, the arteriviruses have recently been grouped together with the coronaviruses and toroviruses in the newly established order Nidovirales. Previously, we reported the construction of pEDI, a full-length cDNA copy of EAV DI-b, a natural defective interfering (DI) RNA of 5.6 kb (R. Molenkamp et al., J. Virol. 74:3156-3165, 2000). EDI RNA consists of three noncontiguous parts of the EAV genome fused in frame with respect to the replicase gene. As a result, EDI RNA contains a truncated replicase open reading frame (EDI-ORF) and encodes a truncated replicase polyprotein. Since some coronavirus DI RNAs require the presence of an ORF for their efficient propagation, we have analyzed the importance of the EDI-ORF in EDI RNA replication. The EDI-ORF was disrupted at different positions by the introduction of frameshift mutations. These were found either to block DI RNA replication completely or to be removed within one virus passage, probably due to homologous recombination with the helper virus genome. Using recombination assays based on EDI RNA and full-length EAV genomes containing specific mutations, the rates of homologous RNA recombination in the 3'- and 5'-proximal regions of the EAV genome were studied. Remarkably, the recombination frequency in the 5'-proximal region was found to be approximately 100-fold lower than that in the 3'-proximal part of the genome.

Isolation and characterization of an arterivirus defective interfering RNA genome.

Equine arteritis virus (EAV), the type member of the family Arteriviridae, is a single-stranded RNA virus with a positive-stranded genome of approximately 13 kb. EAV uses a discontinuous transcription mechanism to produce a nested set of six subgenomic mRNAs from which its structural genes are expressed. We have generated the first documented arterivirus defective interfering (DI) RNAs by serial undiluted passaging of a wild-type EAV stock in BHK-21 cells. A cDNA copy of the smallest DI RNA (5.6 kb) was cloned. Upon transfection into EAV-infected BHK-21 cells, transcripts derived from this clone (pEDI) were replicated and packaged. Sequencing of pEDI revealed that the DI RNA was composed of three segments of the EAV genome (nucleotides 1 to 1057, 1388 to 1684, and 8530 to 12704) which were fused in frame with respect to the replicase reading frame. Remarkably, this DI RNA has retained all of the sequences encoding the structural proteins. By insertion of the chloramphenicol acetyltransferase reporter gene in the DI RNA genome, we were able to delimitate the sequences required for replication/DI-based transcription and packaging of EAV DI RNAs and to reduce the maximal size of a replication-competent EAV DI RNA to approximately 3 kb.

Arterivirus discontinuous mRNA transcription is guided by base pairing between sense and antisense transcription-regulating sequences.

To generate an extensive set of subgenomic (sg) mRNAs, nidoviruses (arteriviruses and coronaviruses) use a mechanism of discontinuous transcription. During this process, mRNAs are generated that represent the genomic 5' sequence, the so-called leader RNA, fused at specific positions to different 3' regions of the genome. The fusion of the leader to the mRNA bodies occurs at a short, conserved sequence element, the transcription-regulating sequence (TRS), which precedes every transcription unit in the genome and is also present at the 3' end of the leader sequence. Here, we have used site-directed mutagenesis of the infectious cDNA clone of the arterivirus equine arteritis virus to show that sg mRNA synthesis requires a base-pairing interaction between the leader TRS and the complement of a body TRS in the viral negative strand. Mutagenesis of the body TRS of equine arteritis virus RNA7 reduced sg RNA7 transcription severely or abolished it completely. Mutations in the leader TRS dramatically influenced the synthesis of all sg mRNAs. The construction of double mutants in which a mutant leader TRS was combined with the corresponding mutant RNA7 body TRS resulted in the specific restoration of mRNA7 synthesis. The analysis of the mRNA leader-body junctions of a number of mutants with partial transcriptional activity provided support for a mechanism of discontinuous minus-strand transcription that resembles similarity-assisted, copy-choice RNA recombination.

Localization of mouse hepatitis virus nonstructural proteins and RNA synthesis indicates a role for late endosomes in viral replication.

The aim of the present study was to define the site of replication of the coronavirus mouse hepatitis virus (MHV). Antibodies directed against several proteins derived from the gene 1 polyprotein, including the 3C-like protease (3CLpro), the putative polymerase (POL), helicase, and a recently described protein (p22) derived from the C terminus of the open reading frame 1a protein (CT1a), were used to probe MHV-infected cells by indirect immunofluorescence (IF) and electron microscopy (EM). At early times of infection, all of these proteins showed a distinct punctate labeling by IF. Antibodies to the nucleocapsid protein also displayed a punctate labeling that largely colocalized with the replicase proteins. When infected cells were metabolically labeled with 5-bromouridine 5'-triphosphate (BrUTP), the site of viral RNA synthesis was shown by IF to colocalize with CT1a and the 3CLpro. As shown by EM, CT1a localized to LAMP-1 positive late endosomes/lysosomes while POL accumulated predominantly in multilayered structures with the appearance of endocytic carrier vesicles. These latter structures were also labeled to some extent with both anti-CT1a and LAMP-1 antibodies and could be filled with fluid phase endocytic tracers. When EM was used to determine sites of BrUTP incorporation into viral RNA at early times of infection, the viral RNA localized to late endosomal membranes as well. These results demonstrate that MHV replication occurs on late endosomal membranes and that several nonstructural proteins derived from the gene 1 polyprotein may participate in the formation and function of the viral replication complexes.

The putative helicase of the coronavirus mouse hepatitis virus is processed from the replicase gene polyprotein and localizes in complexes that are active in viral RNA synthesis.

The coronavirus mouse hepatitis virus (MHV) translates its replicase gene (gene 1) into two co-amino-terminal polyproteins, polyprotein 1a and polyprotein 1ab. The gene 1 polyproteins are processed by viral proteinases to yield at least 15 mature products, including a putative RNA helicase from polyprotein 1ab that is presumed to be involved in viral RNA synthesis. Antibodies directed against polypeptides encoded by open reading frame 1b were used to characterize the expression and processing of the MHV helicase and to define the relationship of helicase to the viral nucleocapsid protein (N) and to sites of viral RNA synthesis in MHV-infected cells. The antihelicase antibodies detected a 67-kDa protein in MHV-infected cells that was translated and processed throughout the virus life cycle. Processing of the 67-kDa helicase from polyprotein 1ab was abolished by E64d, a known inhibitor of the MHV 3C-like proteinase. When infected cells were probed for helicase by immunofluorescence laser confocal microscopy, the protein was detected in patterns that varied from punctate perinuclear complexes to large structures that occupied much of the cell cytoplasm. Dual-labeling studies of infected cells for helicase and bromo-UTP-labeled RNA demonstrated that the vast majority of helicase-containing complexes were active in viral RNA synthesis. Dual-labeling studies for helicase and the MHV N protein showed that the two proteins almost completely colocalized, indicating that N was associated with the helicase-containing complexes. This study demonstrates that the putative RNA helicase is closely associated with MHV RNA synthesis and suggests that complexes containing helicase, N, and new viral RNA are the viral replication complexes.

Characterization of an equine arteritis virus replicase mutant defective in subgenomic mRNA synthesis.

Equine arteritis virus (EAV) is a positive-stranded RNA virus that synthesizes a 5'- and 3'-coterminal nested set of six subgenomic mRNAs. These mRNAs all contain a common leader sequence which is derived from the 5' end of the genome. Subgenomic mRNA transcription and genome replication are directed by the viral replicase, which is expressed in the form of two polyproteins and subsequently processed into smaller nonstructural proteins (nsps). During the recent construction of an EAV infectious cDNA clone (pEAV030 [L. C. van Dinten, J. A. den Boon, A. L. M. Wassenaar, W. J. M. Spaan, and E. J. Snijder, Proc. Natl. Acad. Sci. USA 94:991-996, 1997]), a mutant cDNA clone (pEAV030F) which carries a single replicase point mutation was obtained. This substitution (Ser-2429-->Pro) is located in the nsp10 subunit and renders the EAV030F virus deficient in subgenomic mRNA synthesis. To obtain more insight into the role of nsp10 in transcription and the nature of the transcriptional defect, we have now analyzed the EAV030F mutant in considerable detail. The Ser-2429-->Pro mutation does not affect the proteolytic processing of the replicase but apparently affects the function of nsp10 in transcription. Furthermore, our study showed that EAV030F still produces subgenomic positive and negative strands, albeit at a very low level. Both subgenomic positive-strand synthesis and negative-strand synthesis are equally affected by the Ser-2429-->Pro mutation, suggesting that nsp10 plays an important role in an early step of EAV mRNA transcription.

Alternative proteolytic processing of the arterivirus replicase ORF1a polyprotein: evidence that NSP2 acts as a cofactor for the NSP4 serine protease.

The C-terminal half of the replicase ORF1a polyprotein of the arterivirus equine arteritis virus is processed by a chymotrypsinlike serine protease (SP) (E. J. Snijder et al., J. Biol. Chem. 271:4864-4871, 1996) located in nonstructural protein 4 (nsp4). Three probable SP cleavage sites had previously been identified in the ORF1a protein. Their proteolysis explained the main processing products generated from the C-terminal part of the ORF1a protein in infected cells (E. J. Snijder et al., J. Virol. 68:5755-5764, 1994). By using sequence comparison, ORF1a expression systems, and site-directed mutagenesis, we have now identified two additional SP cleavage sites: Glu-1430 / Gly and Glu-1452 / Ser. This means that the ORF1a protein can be cleaved into eight processing end products: nsp1 to nsp8. By microsequence analysis of the nsp5 and nsp7 N termini, we have now formally confirmed the specificity of the SP for Glu / (Gly/Ser) substrates. Importantly, our studies revealed that the C-terminal half of the ORF1a protein (nsp3-8) can be processed by the SP following two alternative pathways, which appear to be mutually exclusive. In the majority of the nsp3-8 precursors the SP cleaves the nsp4/5 site, yielding nsp3-4 and nsp5-8. Subsequently, the latter product is cleaved at the nsp7/8 site only, whereas the newly identified nsp5/6 and nsp6/7 sites appear to be inaccessible to the protease. In the alternative proteolytic cascade, which is used at a low but significant level in infected cells, it is the nsp4/5 site which remains uncleaved, while the nsp5/6 and nsp6/7 sites are processed to yield a set of previously unnoticed processing products. Coexpression studies revealed that nsp3-8 has to interact with cleaved nsp2 to allow processing of the nsp4/5 junction, the first step of the major processing pathway. When the nsp2 cofactor is absent, the nsp4/5 site cannot be processed and nsp3-8 is processed following the alternative, minor pathway.

The function of the spike protein of mouse hepatitis virus strain A59 can be studied on virus-like particles: cleavage is not required for infectivity.

The spike protein (S) of the murine coronavirus mouse hepatitis virus strain A59 (MHV-A59) induces both virus-to-cell fusion during infection and syncytium formation. Thus far, only syncytium formation could be studied after transient expression of S. We have recently described a system in which viral infectivity is mimicked by using virus-like particles (VLPs) and reporter defective-interfering (DI) RNAs (E. C. W. Bos, W. Luytjes, H. Van der Meulen, H. K. Koerten, and W. J. M. Spaan, Virology 218:52-60, 1996). Production of VLPs of MHV-A59 was shown to be dependent on the expression of M and E. We now show in several ways that the infectivity of VLPs is dependent on S. Infectivity was lost when spikeless VLPs were produced. Infectivity was blocked upon treatment of the VLPs with MHV-A59-neutralizing anti-S monoclonal antibody (MAb) A2.3 but not with nonneutralizing anti-S MAb A1.4. When the target cells were incubated with antireceptor MAb CC1, which blocks MHV-A59 infection, VLPs did not infect the target cells. Thus, S-mediated VLP infectivity resembles MHV-A59 infectivity. The system can be used to identify domains in S that are essential for infectivity. As a first application, we investigated the requirements of cleavage of S for the infectivity of MHV-A59. We inserted three mutant S proteins that were previously shown to be uncleaved (E. C. W. Bos, L. Heijnen, W. Luytjes, and W. J. M. Spaan, Virology 214:453-463, 1995) into the VLPs. Here we show that cleavage of the spike protein of MHV-A59 is not required for infectivity.

A subgenomic mRNA transcript of the coronavirus mouse hepatitis virus strain A59 defective interfering (DI) RNA is packaged when it contains the DI packaging signal.

In infected cells, only the genomic RNA of the coronavirus mouse hepatitis virus strain A59 (MHV-A59) is packaged into the virions. In this study, we show that a subgenomic (sg) defective interfering (DI) RNA can be packaged into virions when it contains the DI RNA packaging signal (DI RNA-Ps). However, the sg DI RNA is packaged less efficiently than the DI genomic RNA. Thus, while specificity of packaging of RNAs into MHV-A59 virions is determined by the DI RNA-Ps, efficiency of packaging is determined by additional factors.

Recombinant genomic RNA of coronavirus MHV-A59 after coreplication with a DI RNA containing the MHV-RI spike gene.

A strategy for targeted RNA recombination between the spike gene on the genomic RNA of MHV-A59 and a synthetic DI RNA containing the MHV-RI spike gene is described. The MHV-RI spike gene contains several nucleotide differences from the MHV-A59 spike gene that could be used as genetic markers, including a stretch of 156 additional nucleotides starting at nucleotide 1497. The MHV-RI S gene cDNA (from nucleotide 277-termination codon) was inserted in frame into pMIDI, a full-length cDNA clone of an MHV-A59 DI, yielding pDPRIS. Using the vaccinia vTF7.3 system, RNA was transcribed from pDPRIS upon transfection into MHV-A59-infected L cells. DPRIS RNA was shown to be replicated and passaged efficiently. MHV-A59 and the DPRIS DI particle were copassaged several times. Using a highly specific and sensitive RT-PCR, recombinant genomic RNA was detected in intracellular RNA from total lysates of pDPRIS-transfected and MHV-A59-infected cells and among genomic RNA that was agarose gel-purified from these lysates. More significantly, specific PCR products were found in virion RNA from progeny virus. PCR products were absent in control mixes of intracellular RNA from MHV-A59-infected cells and in vitro-transcribed DPRIS RNA. PCR products from intracellular RNA and virion RNA were cloned and 11 independent clones were sequenced. Crossovers between A59 and RI RNA were found upstream of nucleotide 1497 and had occurred between 106 nucleotides from the 5'-border and 73 nucleotides from the 3'-border of sequence homologous between A59 and RI S genes. We conclude that homologous RNA recombination took place between the genomic RNA template and the synthetic DI RNA template at different locations, generating a series of MHV recombinant genomes with chimeric S genes.

An infectious arterivirus cDNA clone: identification of a replicase point mutation that abolishes discontinuous mRNA transcription.

Equine arteritis virus (EAV) is a positive-strand RNA virus that uses a discontinuous transcription mechanism to generate a nested set of six subgenomic mRNAs from which its structural genes are expressed. A stable bacterial plasmid (pEAV030) containing a full-length cDNA copy of the 12.7-kb EAV genome was constructed. After removal of a single point mutation in the replicase gene, RNA transcripts generated in vitro from pEAV030 were shown to be infectious upon electroporation into BHK-21 cells. A genetic marker mutation was introduced at the cDNA level and recovered from the genome of the progeny virus. The potential of pEAV030 as a tool to express foreign genes was demonstrated by the efficient expression of the chloramphenicol acetyltransferase (CAT) reporter gene from two different subgenomic mRNAs. The point mutation that initially rendered the full-length clone noninfectious was found to result in a particularly intriguing phenotype: RNA carrying this mutation can replicate efficiently but does not produce the subgenomic mRNAs required for structural protein expression. To our knowledge, this mutant provides the first evidence that the requirements for arterivirus genome replication and discontinuous mRNA synthesis are, at least partially, different and that these processes may be separated experimentally.

Characterization of two temperature-sensitive mutants of coronavirus mouse hepatitis virus strain A59 with maturation defects in the spike protein.

Two temperature-sensitive (ts) mutants of mouse hepatitis virus strain A59, ts43 and ts379, have been described previously to be ts in infectivity but unaffected in RNA synthesis (M. J. M. Koolen, A. D. M. E. Osterhaus, G. van Steenis, M. C. Horzinek, and B. A. M. van der Zeijst, Virology 125:393-402, 1983). We present a detailed analysis of the protein synthesis of the mutant viruses at the permissive (31 degrees C) and nonpermissive (39.5 degrees C) temperatures. It was found that synthesis of the nucleocapsid protein N and the membrane protein M of both viruses was insensitive to temperature. However, the surface protein S of both viruses was retained in the endoplasmic reticulum at the nonpermissive temperature. This was shown first by analysis of endoglycosidase H-treated and immunoprecipitated labeled S proteins. The mature Golgi form of S was not present at the nonpermissive temperature for the ts viruses, in contrast to wild-type (wt) virus. Second, gradient purification of immunoprecipitated S after pulse-chase labeling showed that only wt virus S was oligomerized. We conclude that the lack of oligomerization causes the retention of the ts S proteins in the endoplasmic reticulum. As a result, ts virus particles that were devoid of S were produced at the nonpermissive temperature. This result could be confirmed by biochemical analysis of purified virus particles and by electron microscopy.

Internal entry of ribosomes is directed by the 5' noncoding region of classical swine fever virus and is dependent on the presence of an RNA pseudoknot upstream of the initiation codon.

Bicistronic RNAs containing the 373-nucleotide-long 5' nontranslated region (NTR) of the classical swine fever virus (CSFV) genome as intercistronic spacer were used to show the presence of an internal ribosome entry site (IRES) in the 5' end of the CSFV genome. By coexpression of the poliovirus 2A protease it was demonstrated that the CSFV 5' NTR-driven translation is independent of the presence of functional eukaryotic initiation factor eIF-4F. Deletion analysis indicated that the 5' border of the IRES is located between nucleotides 28 and 66. The role of a proposed pseudoknot structure at the 3' end of the CSFV 5' NTR in IRES-mediated translation was investigated by site-directed mutagenesis. Mutant RNAs that had lost the ability to base pair in stem II of the pseudoknot were translationally inactive. Translation to wild-type levels could be restored through the introduction of compensatory complementary base changes that repaired base pairing in stem II. In addition, we showed that the AUG codon, which is located 7 nucleotides upstream of the polyprotein initiation site and is conserved in pestiviruses, could not be used to initiate translation. Also, an AUG codon introduced downstream of the polyprotein initiation site was not recognized as an initiation site by ribosomes. These data suggest that after internal entry on the CSFV 5' NTR, ribosomal scanning for the initiation codon is limited to a small region.