INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The pestiviruses are animal pathogens causing acute, chronic, and persistent infections that can result in substantial economic losses in the livestock industry (22, 31, 41). Members include bovine viral diarrhea virus (BVDV), classical swine fever virus (CSFV), border disease virus, and the newly proposed member, BVDV-2 (4). The pestiviruses are grouped in the Flaviviridae family along with the hepaciviruses (hepatitis C viruses [HCVs]) and classical flaviviruses; however, similarities in genome organization, translation initiation strategy, and polyprotein processing suggest that the pestiviruses and HCVs are more closely related to each other than to the classical flaviviruses.

The Flaviviridae are small, enveloped viruses containing a single-stranded positive-sense RNA genome (31). The genomes are usually between 9.5 and 12.5 kb in length, with 5' and 3' nontranslated regions (NTRs) flanking one long open reading frame (ORF), and lack the 3' poly(A) tract. Translation of the viral RNA yields a single polyprotein that is co- and posttranslationally processed by host and viral proteases. The classical flaviviruses contain a type 1 cap at the 5' end that directs translation initiation in a cap-dependent manner. In contrast, the 5' NTRs of the pestiviruses and HCVs contain structurally and functionally similar internal ribosome entry site (IRES) elements that mediate translation in a cap-independent manner (reviewed in reference 16).

The first protein of the pestivirus ORF, N^pro, is a nonstructural autoprotease unique to the pestiviruses, which autocatalytically generates its own C terminus (34). It is followed by the structural proteins (C, E^rns, E1, and E2) and the nonstructural proteins (p7, NS2-NS3, NS4A, NS4B, NS5A, and NS5B) (reviewed in reference 22). NS3 harbors serine protease, helicase, and NTPase activities (37, 38, 44-46), and NS5B exhibits RNA-dependent RNA polymerase activity (48). The identification of these activities implies that both NS3 and NS5B are directly involved in viral RNA replication.

Two biotypes can be distinguished following infection of cultured cells. Viruses that do not appear to cause any cell damage are considered noncytopathic (ncp), whereas cytopathic (cp) isolates cause cell shrinkage, membrane blebbing, and cell death (22). For BVDV, only unprocessed NS2-NS3 is observed in ncp isolates. In contrast, production of a discrete NS3 protein in addition to unprocessed NS2-NS3 has been associated with the cp phenotype. Considerable data suggest that cp viruses usually develop from ncp viruses by RNA recombination events resulting in genome duplications, rearrangements, deletions, or cellular insertions. However, one report has demonstrated that single-point mutations in NS2 were responsible for cytopathic effects (CPE) in the BVDV strain Oregon (14).

Characterization of several isolates revealed that cp BVDV and CSFV could be composed of RNAs with large internal deletions and an ncp helper virus (15, 20, 21, 23, 24, 40). In these cases, the 5' NTR and initiating methionine codon were intact, but in-frame deletions of the structural (C, E^rns, E1, and E2) and nonstructural (p7 and NS2) genes were noted. For simplicity and continuity, we will refer to these internally deleted genomes as subgenomic (sg) RNAs (5, 19, 40). The most well-characterized cp sg BVDV RNA, DI9, has a deletion encompassing all of the structural proteins as well as p7 and NS2, leaving the C terminus of N^pro fused in frame to NS3 (40). Additional sg BVDV RNAs have been described, in which portions of N^pro and E1 are fused to NS3 (DI13) (15) or, in others, a capsid (or NS2)-ubiquitin fusion is inserted between complete or C terminally truncated N^pro and NS3 (3). However, in most of the sg CSFV RNAs, the initiating AUG codon is fused in frame to the N terminus of NS3 (21, 24).

Following the identification of the cytopathic sg RNAs, several investigators created cDNA copies to study the molecular requirements of pestivirus replication and cytopathogenicity (5, 20, 23, 25, 39). The first was based on the DI9 genome, and the derived RNA replicated and caused CPE following cotransfection with an ncp helper RNA (20). Subsequent studies showed that the structural proteins, as well as p7 and NS2, were dispensable for viral RNA replication and cytopathogenicity since sg DI9-like RNAs replicated autonomously and caused cell lysis (5, 39). It also appeared that the first 39 nucleotides (nt) downstream of the N^pro AUG codon were required for replication of the sg DI9-like RNAs, although the specific N^pro amino acid (aa) sequence was not. However, the specific function of these sequences was not examined. In contrast, the CSFV sg RNAs, which lacked N^pro, replicated efficiently (23, 25).

To create an experimental system to study pestivirus replication, an sg RNA was constructed based on the sg CSFV genome organization (MetNS3) by using the National Animal Disease Laboratory (NADL) strain of BVDV. We present evidence demonstrating that efficient pestivirus translation initiation correlates with autonomous sg RNA replication. Furthermore, the data suggest that binding of 40S ribosomal subunits and formation of 48S ribosomal initiation complexes require a less-structured or less stable RNA structure downstream of the initiating methionine.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cells and viruses. Madin-Darby bovine kidney (MDBK) cells were generously provided by M. Collett (ViroPharma, Inc.) and were propagated in supplemented Dulbecco's modified minimal essential medium (with sodium pyruvate and 10% horse serum) (19). Cells were maintained at 37°C with 5% CO₂.

Plasmids. Plasmids pACNR/NADL and pACNR/cInsNADL, representing the cp and ncp infectious cDNAs, respectively, have been described previously (19). The discistronic reporter plasmid, T7DC1-341, was kindly provided by A. Siddiqui (43). It contained a T7 promoter directing transcription of chloramphenicol acetyltransferase (CAT) and an HCV IRES driving expression of luciferase (LUC).

View larger version (13K): [in this window] [in a new window]   FIG. 1.   Schematic representation of BVDV RNAs transcribed from the cDNA templates, cp NADL (top), ncp cIns-NADL (middle), and sg MetNS3 (bottom), used in these studies. The proposed 5' and 3' NTR structures are shown, with the polyprotein cleavage products indicated by boxes. The numbers in parentheses refer to the AUG start codon (nt 386) and the boundaries of the cellular insert (cIns, nt 4994 to 5263).

In vitro transcription of RNA. Plasmid DNAs containing full-length and sg BVDV NADL sequences were digested with SdaI, and the reporter plasmid, T7DC1-341, was digested with HpaI (LUC expression, transfection studies) or SalI (CAT expression only, in vitro translation). The linearized DNAs were extracted with phenol-chloroform (1:1) and chloroform alone and then were precipitated with ethanol. Pelleted DNAs were resuspended in double-distilled water (ddH₂O) and transcribed in vitro using the T7-MEGAscript kit (Ambion, Austin, Tex.) in the presence of 1 µCi of [³H]UTP (Dupont NEN, Boston, Mass.), as recommended by the manufacturer (19). The in vitro-synthesized RNAs were resuspended in ddH₂O at approximately 1 µg/µl based on [³H]UTP incorporation. Transcript RNAs and standard RNAs of known concentrations were separated on a 1% agarose gel and stained with ethidium bromide. The final RNA concentrations were determined using a Fluorimager (Molecular Dynamics, Sunnyvale, Calif.) in conjunction with the IPLab Gel program (Scantalytics, Inc., Fairfax, Va.). RNA aliquots were stored at 80°C.

Transfection of cultured cells. Transfection of MDBK cells by electroporation was performed essentially as described previously (19). Briefly, MDBK cells (approximately 80% confluent) were detached by trypsin treatment; washed three times in sterile, ice-cold phosphate-buffered saline (PBS); and resuspended at between 1.0 × 10⁷ and 2.0 × 10⁷ cells/ml in PBS. Volumes containing 4 to 5 µg of full-length or sg RNA transcripts were combined with 1 µg of LUC reporter RNA (DCI-341), mixed with 0.4 ml of washed MDBK cells, and pulsed immediately (1.5 kV, 25 µF,  resistance, and 2 pulses) using a Bio-Rad Gene Pulser (Hercules, Calif.). The total RNA transfected per reaction mixture was equalized using uninfected MDBK cellular RNA that had been extracted using TRIzol reagent (Gibco-BRL, Grand Island, N.Y.) as described below. The pulsed samples were diluted to 10 ml with Dulbecco's minimal essential medium (supplemented with 10% horse serum and sodium pyruvate), added to either a 100-mm-diameter dish or split equally into four 35-mm-diameter wells, and incubated at 37°C with 5% CO₂. The specific infectivity of full-length RNA was determined using an infectious center assay as described previously (19).

Protein analysis. Transfection efficiency was monitored in the transfected cells by quantitation of LUC activity at 8 h posttransfection (p.t.). Cell monolayers (35-mm-diameter wells) were washed twice with PBS and resuspended in 250 µl of 1% Triton X-100 lysis buffer (25 mM Tris-HCl, pH 7.5; 25 mM dithiothreitol; 2 mM EGTA; 1% [vol/vol] Triton X-100). The LUC activity in one-tenth of the clarified cell extract was measured using a luminometer (EG&G Berthold, Gaithersburg, Md.).

Viral RNA analysis. Initially, RNA synthesis and accumulation were examined by [³H]uridine (30 µCi/m) (ICN Biomedicals, Inc., Costa Mesa, Calif.) incorporation in the presence of dactinomycin (1 µg/ml) (Fig. 2). Alternatively, 200 µCi of [³²P]orthophosphate (H₃³²PO₄)/ml was added during the exponential phase of sg RNA synthesis in media containing low levels of phosphate at concentrations that yielded NADL viral titers between 50 and 80% of those obtained in complete media (data not shown). RNAs from the transfected, radiolabeled cells were isolated using TRIzol reagent (Gibco-BRL). To ensure recovery of total RNA, detached cells were collected from the supernatants by low-speed centrifugation and the pelleted cells were combined with the corresponding monolayer lysate. RNAs were isolated according to the manufacturer's protocol, except that carrier tRNA (25 µg) was added to each lysate and an additional chloroform extraction was performed.

View larger version (34K): [in this window] [in a new window]   FIG. 2.   Viral RNA accumulation and CPE in cells infected with passage 4 virus. (A) MDBK cells were infected at an MOI of 1 with passage 4 virus (lane 1) or were mock infected (lane 2, media only) and were labeled from 6 to 18 h p.i. with [3H]uridine in the presence of dactinomycin. RNAs were isolated from the cells using TRIzol reagent (Gibco-BRL), denatured with glyoxal, and separated on a 1% agarose gel. The labeled RNAs were visualized by autoradiography. (B) Plaques were detected by crystal violet staining at 3 days p.i. with passage 4 virus as described previously (19).

Assembly and analysis of ribosomal and ribonucleoprotein complexes. Ribosomal subunits, subunit of eukaryotic initiation factor 2 (eIF2) and eIF3 were purified from rabbit reticulocyte lysates (Green Hectares, Oregon, Wis.) as previously described (28, 29). Ribosomal complexes and complexes between BVDV mRNA and eIF3 were assembled on BVDV mRNA, essentially as previously described (27). BVDV plasmids were linearized with BglII, and mRNAs were transcribed in vitro using T7 RNA polymerase. Aliquots of the mRNAs (0.6 µg) were incubated for 5 min at 30°C in a 40-µl reaction volume that contained 0.4 mM GMP-PNP, Met-tRNA_i^Met (6 pmol), 40S subunits (6 pmol), eIF2 (3 µg), and eIF3 (7 µg) as indicated in the text. [³⁵S]methionine-labeled Met-tRNA_i^Met was prepared using rabbit tRNA (Novagen, Madison, Wis.) and aminoacyl-tRNA synthetase purified from Escherichia coli MRE600 (American Type Culture Collection, Manassas, Va.), as described previously (28).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Adaptive mutations are required for replication of sg NADL MetNS3 RNA in cell culture. To study the genetic requirements of pestivirus replication and cytopathogenicity, we created a classical swine fever-like sg cDNA (MetNS3) derived from an infectious NADL cDNA clone. For MetNS3, the complete NADL 5' NTR and initiating methionine codon were fused in frame with the first codon of NS3 (14), which was contiguous with the remainder of the NADL sequence (19). MDBK cells were cotransfected with in vitro-synthesized sg MetNS3 and ncp cInsNADL RNAs (Fig. 1) and observed for the development of CPE (19). Surprisingly, neither CPE nor sg RNA replication was detected in the transfected cells, whereas replication of ncp cInsNADL virus was readily observed (data not shown). Identical results were obtained after transfection of MetNS3 RNA into persistently infected MDBK cells. However, if the extracellular supernatant was serially passaged on persistently infected MDBK cells, significant CPE was eventually observed (data not shown). Similar to results reported for another virus stock containing both sg and full-length RNA (CP13), the number of PFU was 10- to 100-fold lower than the number of focus-forming units in the passaged virus stocks (15). Thus, the range of MOIs that could be reasonably achieved in subsequent infections was limited.

View larger version (38K): [in this window] [in a new window]   FIG. 3.   RNA and protein accumulation in MDBK cells transfected with the wt and reconstructed MetNS3 RNAs, ad3.1, ad3.2, ad3.4, ad3.5, and ad3.6. (A) Sequence changes identified in five individual cDNA clones after replacing the wt MetNS3 sequence with the corresponding RT-PCR product amplified from passage 4 MetNS3 RNA as described in Materials and Methods. The wt MetNS3 nucleotide sequence is shown at the top with the single-letter amino acid code above. The 1st, 13th, 25th, and 50th amino acids are indicated along with the first nucleotide number of the codons. Hash marks () represent breaks in the linear sequence. Mutations are indicated in bold, and any resulting amino acid change is above the codon. Dots indicate homology with the wt MetNS3 sequences. (B and C) Analysis of RNA and protein accumulation, respectively, per ~4 × 106 cells transfected with in vitro-transcribed RNAs. Shown are wt MetNS3 (lanes 2) and the adapted derivatives MetNS3 RNAs (lanes 3 to 7). Mock () transfections (lanes 1) lacked RNA. RNA accumulation was detected by radiolabeling the transfected cells with H332PO4 between 16 and 24 h p.t. in the presence of dactinomycin. Total cellular RNA was glyoxylated, separated on a 1% agarose gel, and visualized by autoradiography. The position of the sg RNA is indicated. The radiolabeled bands were quantitated with a Molecular Imager (Bio-Rad Laboratories) and normalized to transfection efficiency and total 18S rRNA as described in Materials and Methods. The percentages of each sg RNA relative to ad3.4 (100%) are indicated. (B). Transfected cell lysates were prepared at 18 h p.t. and separated by SDS-8% PAGE. NS3 levels were detected with polyclonal rabbit NS3-specific antisera by Western blot analysis. Molecular mass markers are indicated on the right and the migration of NS3 is indicated on the left (C).

Adapted ad3 RNA replicate autonomously. Recent reports demonstrated that an sg BVDV DI9 RNA could replicate autonomously in cultured cells and was able to induce CPE in the absence of an ncp helper virus (5, 39). The sg DI9 or DI9-like RNAs do not encode structural proteins. Thus, viral packaging, release, and spread were absent, and RNA replication and cytopathogenicity were examined as first-cycle events in transfected cells. To determine if the reconstructed sg MetNS3 RNAs were capable of autonomous replication, MDBK cells were transfected with in vitro-synthesized RNA and a LUC reporter RNA as a transfection control (43). Transfected cells were split into 35-mm-diameter wells to examine viral RNA accumulation, protein expression, CPE, and LUC activity.

Translation of wt MetNS3 RNA in vitro is reduced compared to adapted MetNS3 RNAs. To analyze the effect of the wt and mutant NS3 sequences on NADL IRES function, the sg MetNS3 RNAs were translated in vitro in the presence of [³⁵S]methionine to detect the proteins. The level of NS3 translated from each RNA was normalized by comparing the amount of NS3 to the level of CAT expressed from a separate transcript (see Materials and Methods). As shown in Fig. 4, the level of ³⁵S-labeled wt NS3 was between 8- to 10-fold less than the levels of adapted NS3, suggesting that there was a defect in the translation initiation of wt MetNS3 (Fig. 4, compare lane 3 with lanes 4, 5, 6, 7, and 8). In contrast, translation of ad3.6 NS3 appeared similar to that of NS3 expressed from the replicating sg RNAs, suggesting that the inability of ad3.6 to replicate may be due to an impairment in NS3 function and not to reduced translation initiation (Fig. 4, compare lane 8 to lanes 4, 5, 6, and 7). As controls, we confirmed that the in vitro-synthesized NS3 proteins comigrated with authentic NS3 from NADL-infected cells (data not shown) and that no ³⁵S-labeled protein products were observed in the mock reaction mixture (Fig. 4, lane 1).

View larger version (64K): [in this window] [in a new window]   FIG. 4.   In vitro translation of wt and adapted NS3 proteins. The wt and reconstructed MetNS3 RNAs were cotranslated with a capped monocistronic CAT RNA in the presence of [35S]methionine as described in Materials and Methods. Reaction products were diluted 10-fold in Laemmli sample buffer and separated by SDS-9% PAGE, and the labeled proteins were visualized by autoradiography. The mock reaction mixture (m; lane 1) contained ddH2O only, whereas mock plus CAT (m+c), wt MetNS3, ad3.1, ad3.2, ad3.4, ad3.5, and ad3.6 RNAs each contained CAT RNA (lanes 2, 3, 4, 5, 6, 7, and 8, respectively). The radiolabeled bands were quantitated with a Molecular Imager (Bio-Rad Laboratories). The data are representative of two experiments.

Type 2 IRES-mediated translation initiation allows replication of wt MetNS3 RNA. If the lack of detectable wt MetNS3 RNA replication was due to reduced translation, increasing translation efficiency should enhance replication. On the other hand, if wt MetNS3 lacks a cis-acting replication signal or inhibits RNA synthesis directly, increasing the level of translation should not significantly influence sg RNA replication. To distinguish between these possibilities, the wt MetNS3 5' NTR was replaced with a chm BVDV-EMCV 5' NTR that contained the three 5' terminal BVDV hairpins required for BVDV replication (Fig. 5A) fused to the canonical EMCV IRES element (9). Since the EMCV IRES functions independently of sequences downstream of the initiator methionine codon (12), chm MetNS3 RNA would allow us to determine the effect of the wt MetNS3 coding sequences on RNA replication independently of translation.

View larger version (23K): [in this window] [in a new window]   FIG. 5.   Replication of wt MetNS3 RNA containing a type 2 IRES. (A) Schematic representation of the wt BVDV (type 3 IRES) and chm MetNS3 (type 2 IRES) RNAs as described for Fig. 1. (B) MDBK cells were transfected with no RNA (lane 1), wt MetNS3 RNA (lane 2), ad3.1 RNA (lane 3), and chm MetNS3 RNA (lane 4), and RNA synthesis and accumulation were assessed by H332PO4 incorporation as described for Fig. 3. An autoradiogram of the dried gel is pictured, and the results are representative of two experiments.

MetNS3 RNAs with less-stable RNA structures downstream of the AUG codon replicate efficiently. It was possible that the nucleotide sequence or RNA structure downstream of the AUG codon in wt MetNS3 inhibited translation initiation, similar to that reported for HCV IRES-driven translation (10, 11). The 5' NS3 sequence in wt MetNS3 is GC-rich (Fig. 3A) and might predispose this region to form a stable RNA structure. In keeping with this hypothesis, the NS3 sequence downstream of the initiator methionine in the replication-competent sg CSFV MetNS3 RNAs is AU-rich and the nucleotide at position 400 is a U (21).

View larger version (33K): [in this window] [in a new window]   FIG. 6.   Replication of MetNS3 RNAs with less-stable RNA structures downstream of the initiator AUG codon. (A) Diagram of the sequences for the MetNS3 RNAs as described for Fig. 3. ad3.7, ad3.8, and ad3.10 were engineered to contain the indicated mutations. (B) Autoradiogram of RNA accumulation in transfected MDBK cells as described for Fig. 3. The percentages of radiolabeled sg RNAs relative to that of ad3.4 (100%) are indicated. Values are normalized to transfection efficiency and total 18S rRNA.

Adapted MetNS3, but not wt MetNS3, binds 40S ribosomes and forms 48S ribosomal initiation complexes. Translation initiation of BVDV, CSFV, and the related HCV has been evaluated in vitro using purified translation components and in vitro-transcribed RNAs (27, 29). Therefore, to analyze the efficiency of IRES-mediated translation initiation in the wt MetNS3, ad3.7 (AU-rich), and ad3.10 (G400U) RNAs, we reconstituted this process in vitro to the stage of 48S initiation complex formation. The ribosomal complexes were analyzed by toeprinting, which involves primer extension by reverse transcriptase on a template RNA to which a protein or ribosomal complex is bound. Synthesis of the cDNA is arrested at the leading edge of the complex, yielding toeprints that can be accurately mapped on sequencing gels.

View larger version (30K): [in this window] [in a new window]   FIG. 7.   Toeprint analysis of 40S ribosomal and 43S preinitiation complex binding to the BVDV IRES. Ribosomal complexes were assembled under standard reaction conditions on BVDV mRNAs and in the presence of translation components as indicated and were analyzed by primer extension. Full-length cDNA (E) and other cDNA products terminated at sites indicated on the right. The sequence GCCG395-8 refers to nt 395 to 398 in MetNS3 mRNA and to equivalent residues in ad3.7 and ad3.10 mRNA.

View larger version (18K): [in this window] [in a new window]   FIG. 8.   Schematic representation of model secondary and tertiary RNA structures of BVDV as described previously by Pestova et al. (27). The NADL nucleotide sequences are of nt 383 to 406 of the BVDV 5' NTR-Npro sequence (the initiation AUG codon is underlined), and nt 386 to 406 are shown for MetNS3, ad3.7, and ad3.10. Primer extension toeprints of the MetNS3 RNAs caused by binding of 40S ribosomes and 48S preinitiation complexes are shown. Smaller symbols indicate less intense toeprints observed for the RNA.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we created an sg NADL RNA (MetNS3) based on the genome structure of the cp CSFV isolates, in which the 5' NTR and initiating AUG codon were fused to the N terminus of NS3 (Fig. 1; 21, 23). In contrast to earlier reports in the literature for sg CSFV RNA of similar structure, neither replication nor induction of CPE was observed for sg NADL MetNS3. However, following serial passage in the presence of ncp helper virus, detectable sg RNA replication and CPE were detected (Fig. 2). These results suggested that perhaps the N^pro sequence was essential, similar to reports for the BVDV DI9-like RNAs (5, 39). However, efficient replication of the reconstructed RNAs, ad3.1, ad3.2, ad3.4, and ad3.5 (Fig. 3A and B, lanes 3 to 6), showed that a specific N^pro nucleotide or amino acid sequence was not required. Furthermore, comparison of the N^pro and NS3 sequences over the first 114 nt (38 aa) revealed very little homology at the nucleotide (29 to 31% identity) or amino acid (8 to 10.5% identity) level.

The combined results of reduced wt MetNS3 translation in vitro (Fig. 4, lane 3) and autonomous replication of the 5' NTR chimera RNA, chm MetNS3 (Fig. 5B, lane 4), suggested that the sequences downstream of the initiating AUG codon were directly influencing BVDV IRES-directed translation initiation and not RNA synthesis. This conclusion is not unprecedented since translation initiation directed by the BVDV or closely related HCV IRES elements has been reported to require downstream sequences within the 5' portion of the ORFs (6, 11, 17, 30). In contrast, others have reported efficient HCV IRES-directed translation in the absence of HCV coding sequences (32, 42, 43). Honda et al. (10) suggested that the conflicting reports for HCV IRES-mediated translation were due to a stable RNA secondary structure near the initiating AUG codon, because mutations which increased the stability of the stem-loop surrounding the HCV initiator AUG codon were inversely correlated with translation efficiency.

In support of the latter ideas, we observed that sg RNAs (ad3.1, ad3.2, ad3.4, ad3.5, ad3.7, and ad3.10) that were predicted to have less-stable RNA secondary structures near the AUG codon were able to translate (Fig. 4, lanes 4 to 7, and data not shown) and replicate efficiently (Fig. 3, lanes 3 to 6; Fig. 6, lanes 3 to 6, 8, and 10). In contrast, sg RNAs (wt MetNS3 and ad3.8) with secondary structures predicted to be more stable were translated poorly (Fig. 4, lane 3, and data not shown) and did not replicate to detectable levels (Fig. 3B, lane 2; Fig. 6B, lanes 2 and 9). Furthermore, enhanced RNase T₁ cleavage at nt 402 was observed in the replication-competent sg RNAs of ad3.7 and ad3.10, with the prevalent G400U substitution, but not in wt MetNS3 RNA. These data suggest that the presence of a U at nt 400 may promote formation of an RNase T₁-accessible, single-stranded RNA in this region. The inability of ad3.6 to replicate to detectable levels, even though it was translated efficiently (Fig. 4, lane 8), may have been due to the influence of the additional amino acid mutation on NS3 function (R51S; Fig. 3A).

Our data do not preclude the possibility that additional RNA-protein interactions modulate BVDV-IRES-mediated translation initiation in vivo. However, the inability of the 40S ribosomal subunits and 43S complexes to bind wt MetNS3 RNA in vitro provides strong evidence that the wt MetNS3 RNA inhibits translation initiation through an intrinsic property of the RNA (Fig. 7, lanes 2 and 3). Furthermore, base substitutions which are predicted to decrease the thermodynamic stability of RNA structures within this region allowed 40S binding and subsequent 48S complex formation, albeit at different levels for ad3.7 and ad3.10 (Fig. 7, compare lanes 6 and 7 with 10 and 11). Although the ad3.10 RNA was significantly less efficient at binding 40S ribosomal subunits and 43S ribosomal preinitiation complexes in vitro than ad3.7, the levels of sg RNA replication in cultured cells were similar (Fig. 6, lanes 8 and 10).

The binding of 40S ribosomal subunits to pestivirus and HCV IRES elements closely resembles the mechanism of 30S ribosomal subunits binding to prokaryotic mRNAs (13, 27, 29). In prokaryotes, the 30S ribosome binding site is extremely sensitive to the presence of stable RNA secondary structures, such that small increases in stability (2.4 kcal/mol) can decrease binding by an order of magnitude (reviewed in reference 8). Hence, it appears that the stability of the RNA secondary structure at the 30S ribosome binding site is quite close to the threshold for being inhibitory. Our results are consistent with the observation that increased RNA secondary structure can significantly inhibit ribosome binding and that a single G-to-U substitution predicted to decrease the stability of the RNA secondary structure permitted 40S ribosome binding.

It is also possible that either an A or C at position 400 would have allowed translation and sg RNA replication. The translation and autonomous replication of ad3.7 suggest that the identity of the primary sequence, at least at the third base position, is not as important as the resulting effects on secondary structure. This hypothesis is compatible with the observation that changes in RNA secondary structure have a far greater impact on 30S ribosome binding than the identity of the specific nucleotide substitution (8). The predicted stability of RNAs with either an A or C at nt 400 (G between 6.5 and 8.0 kcal/mol) was increased relative to that of ad3.10 and ad3.7 but less than that of wt MetNS3 or ad3.8. Since our results suggested that the thermodynamic threshold (G) for 40S binding to the MetNS3 RNA is between 5.6 and 11.0 kcal/mol, it will interesting to determine the effect of the A and C substitutions at position 400.

Another explanation is that within the cellular milieu, additional host factors influence translation initiation. Several studies have demonstrated interactions between La protein and polypyrimidine tract binding protein (PTB) with the related HCV IRES element (reviewed in reference 16). La protein specifically interacts with the HCV RNA in the context of the initiator methionine codon and may contribute to efficient translation initiation (1, 2). La protein is an RNA binding protein with helicase activity and, hence, may unwind double-stranded RNA in preparation for translation initiation. In addition, La has been shown to associate with small ribosomal subunits in the cytoplasm (26). The possible role(s) of PTB in HCV IRES-mediated translation initiation is somewhat less definitive but may involve PTB acting as an RNA chaperone or perhaps through additional protein-protein interactions (16). For BVDV, recent results have shown that although PTB weakly binds BVDV IRES RNA, it does not influence BVDV IRES-mediated translation in vitro (36). Therefore, it is possible that La or other, as-yet-unidentified cellular proteins influence BVDV IRES-directed translation initiation.

The importance of maintaining wt BVDV 5' N^pro sequence for efficient replication was supported by the observation that at least 28 aa of N^pro were found in 11 sg RNAs isolated from four different animals (3). Our data are also consistent with the results obtained with the D19-like replicons, for which a portion of N^pro was necessary for sg RNA replication (5, 39). Efficient translation initiation, including 40S and 43S binding, was associated with sg RNA replication, and since translation was not examined in the nonreplicating DI9-like RNAs, it is entirely possible that they were defective in protein synthesis.

These data and those of others (10, 11, 13, 13a, 27, 29, 33) led us to propose a model for type 3 IRES-mediated translation initiation (Fig. 9). Effective binding of the 40S ribosomal subunit to the initiator AUG codon requires the surrounding RNA to be relatively unstructured (Fig. 9A, left) since stable RNA secondary structure(s) immediately downstream of the canonical IRES (Fig. 9A, stem-loop, right) appear to inhibit 40S binding (Fig. 9B). In the absence of 40S binding, 48S initiation complex formation and subsequent translation initiation are blocked (Fig. 9C). The ability to influence type 3 IRES-driven translation initiation by increasing RNA secondary structure downstream of the initiator AUG codon may have evolved as a mechanism for controlling viral gene expression and subsequent RNA replication in virus-infected cells. In addition, translation initiation could be modulated by binding of viral or host proteins or of both, such that the stability of the RNA secondary structure is either increased or decreased.

View larger version (24K): [in this window] [in a new window]   FIG. 9.   Model for the mechanism of 48S ribosomal preinitiation complex formation on type 3 IRES elements. (A) A schematic representation of type 3 IRES structures with either less-structured RNA (left) or stable stem-loop RNA structures (right) near the initiator AUG codon. (B) In the absence of stable RNA secondary structures, a single 40S ribosomal subunit binds multiple noncontiguous sites on wt pestivirus and HCV RNAs, including near the initiator AUG codon (11, 13, 27, 29, 30, 32). (C) An eIF2-GTP-tRNA ternary complex binds to a binary (40S RNA) complex forming a 48S complex at the initiation codon on wt RNAs but not on RNAs with stable RNA secondary structure(s) near the initiating AUG codon.

Taken together, these results also suggest that pestivirus translation and RNA replication are coupled processes in vivo. Similarly, Yu et al. (47) reported that the 5' terminal stem loop of the BVDV IRES contains a cis-acting element involved in regulation of both translation and replication. The inapparent replication of wt MetNS3 RNA in the presence of the ncp helper virus also suggests that complementation of RNA replication does not occur in trans, at least to detectable levels. It is likely that the regulation of translation and replication in vivo is tightly regulated and that BVDV modulates various host responses by controlling viral protein and RNA levels.

In summary, using engineered sg BVDV replicons, we have shown that efficient translation initiation likely requires a less-structured RNA surrounding the AUG initiation codon. Moreover, it appears that certain levels of viral protein expression are necessary to support a complete RNA replication cycle. We have demonstrated interactions of the sg RNAs with the 40S and 43S ribosomal complexes and that the formation of the ribosomal preinitiation complexes associates with biological function in vitro.