INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Rubella virus (RUB) is a significant human pathogen which causes rubella, or German measles. RUB is the sole member of the Rubivirus genus within the Togaviridae family. Like genomes of the members of Alphavirus, the other genus in the togavirus family, the genome of RUB consists of a single-stranded, plus-sense RNA genome (9,762 nucleotides [nt] in length) which is 5' capped and 3' polyadenylated and contains two open reading frames (ORFs): the 5'-proximal ORF encodes nonstructural proteins including the RNA-dependent RNA polymerase (RdRp), which functions in viral RNA replication, and the 3'-proximal ORF encodes the viral structural proteins, the capsid protein (C), and two envelope glycoproteins, E1 and E2. RUB RNA replication initiates with the synthesis of a genomic RNA complement of minus polarity by the viral RdRp. This minus-strand RNA is then used as a template for the synthesis of both genomic RNA and the subgenomic RNA from which the structural proteins are translated (reviewed in reference 14).

The 3' terminus of the genomes of plus-strand RNA viruses is postulated to contain the cis-acting elements required for the synthesis of the minus-strand RNA. In alphaviruses, a conserved 19-nt stretch preceding the poly(A) tract was found critical for viral viability (26); further analysis showed that some of the nucleotides in this stretch were more important than others (24). However, there is no sequence homology between RUB and alphaviruses at the 3' end of the genome (reviewed in reference 14). A prominent stem-and-loop (SL) structure 59 nt from the 3'-terminal poly(A) tract of the RUB genome has been noticed by several groups (13, 36). This SL consists of a UAUA loop and a 13-nt GC-rich stem which contains a U-U bulge. Analysis of the interaction between host factors and the 3' end of the RUB genome demonstrated that three cellular proteins with molecular masses of 68, 63, and 61 kDa, respectively, were able to specifically bind to this SL (37); the 61-kDa protein was later identified as calreticulin (CAL) (38, 50), a ubiquitous calcium binding protein found in most eukaryotic cells. CAL is located primarily in the endoplasmic reticulum and has been found to function as a molecular chaperone in the maturation of the cytomegalovirus glycoprotein B (60), the influenza virus hemagglutinin (17, 18, 41), hepatitis C virus E2 and E1 (9), and the human immunodeficiency virus gp160 envelope glycoprotein (39). In all three cases, CAL binds to the nascent polypeptide. Thus, the finding that CAL binds with the RUB RNA was surprising. Although the CAL binding site was not mapped, the interaction of CAL with the U-U bulge appears to be important since deletion of the loop dramatically reduced the binding of all three proteins in cell lysates (37). The phosphorylation state of CAL was found to be altered in RUB-infected cells, and this change correlated with increased binding of CAL in cell lysates to the SL (38, 50). Using a bacterially expressed maltose binding protein (MBP)-CAL fusion, it was later found that MBP-CAL can be autophosphorylated and both the autophosphorylation sites and RNA-binding activities of MBP-CAL were localized to the N terminus of the CAL sequences (2), which is the most conserved domain among CALs from different species (reviewed in reference 23). The binding of CAL to the RUB SL implied a role in regulation of RUB RNA replication. A number of studies have shown specific interactions between viral genomic RNAs and cellular proteins and have indicated a role of such binding in viral RNA replication or translation (1, 5, 6, 10, 15, 16, 21, 25, 27, 35, 37, 38, 40, 42, 43, 50, 52, 53).

The interaction between the RUB genome and CAL has also been proposed to be related to viral pathogenesis, particularly arthritis, which is a frequent complication of rubella in adults (up to 25% of female vaccines suffer from transient arthalgia or arthritis). CAL was classified as an autoantigen as a result of several studies that have shown that the antisera from patients suffering from autoimmune disease, such as onchocerciasis (47, 48), can cross-react with CAL, and several autoantigens, such as DR2Dw4/DR3 (33, 56) and Ro/SS-A (29, 30), have been reported to show sequence homology with CAL. It was later found that CAL can interact with hY RNA and might function as a molecular chaperone in the formation of a Ro/SS-A RNP complex (8). Interestingly, CAL must be dephosphorylated to bind with this RNA species. Two other host factors with molecular masses of 52 and 59 kDa were found to interact with an SL at the 5' end of the RUB genome (42), and one of these was later identified as the La autoantigen (43). Significantly, rubella patients were found to have increased levels of anti-La antibodies (43). A mutagenic study of the 5' SL element using Robo302, a RUB infectious clone, showed that this structure functions in translation rather than in RNA replication (45).

Previous studies of RUB defective interfering (DI) RNAs generated from serial undiluted passages found that the 3'-terminal 305 nt were retained in all such DI RNAs (7, 11). Therefore, this region must contain the 3' cis-acting elements required for RUB genome replication. In this study, we used Robo302 to begin analysis of this region to map the 3' cis-acting elements required for RUB RNA replication, including the SL structure that binds CAL.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

RNA secondary structure predictions. Optimal RNA secondary structures were predicted by the method of Zuker and Steigler (61), using the RNAFOLD program in the Wisconsin Package (Genetics Computer Group, Madison, Wis.).

Viruses, cell lines, and transfection. The F-Therien (fTH; originally obtained from J. Chantler) and HPV77 (obtained from Viral Antigens Inc., Memphis, Tenn.) strains of RUB were propagated and titered by plaque assay described previously (57).

Molecular cloning, in vitro transcription, and reverse transcription. Standard recombinant DNA techniques were used, with minor modifications. Restriction enzymes were purchased from New England BioLabs (Beverly, Mass.) or Boehringer Mannheim Biochemicals (Indianapolis, Ind.), while T4 DNA ligase was obtained from New England BioLabs.

Enzymatic probing of RNA secondary structure. Two plasmids containing the T7 RNA polymerase promoter followed by the 3' 110 nt of either the fTH (pUCRUB3'110-fTH) or HPV77 (pUCRUB3'110-HPV) strain, a 20-nt poly(A) tract, and an EcoRI site for linearization were constructed. Intracellular RNA from fTH- or HPV77-infected Vero cells was extracted at 48 h postinfection by using Tri-reagent (Molecular Research Center, Cincinnati, Ohio) according to the manufacturer's protocols. First-strand cDNA was primed with primer 105 (Table 1) and then amplified by PCR using primer 105 as the reverse primer and forward primer 337 (Table 1). PCR amplification using Ex-taq DNA polymerase (Panvera, Madison, Wis.) was done as described previously (44). The HindIII-EcoRI-restricted PCR fragment was ligated with HindIII-EcoRI-restricted pUC18 (Gibco/BRL).

Site-directed mutagenesis. Mutations were created by PCR using primers containing the desired mutations, Ex-taq polymerase, and EcoRI-linearized Robo302 plasmid as a template under conditions optimized for amplification of the high-G+C RUB sequences (39). Primers for constructing mutants using PCR amplification are listed in Table 1. Generally, the PCR products were purified by phenol-chloroform extraction and ethanol precipitation with sodium acetate at a final concentration of 0.3 M. The DNA was then pelleted by centrifugation, dissolved in water, and digested with appropriate restriction enzymes for cloning. Following cloning, mutations were confirmed by DNA sequencing using a dideoxy-chain termination sequencing kit from United States Biochemicals (Cleveland, Ohio).

Confirmation of mutation by RT-PCR and automated sequencing. Single plaques were picked from transfection plates and amplified by one passage in Vero cells (in the case of occasional mutants that produced no plaques, medium from the DMEM plate was amplified). Once the CPE became obvious, intracellular RNA was extracted with Tri-reagent and used as a template for reverse transcription-PCR (RT-PCR). First-strand cDNA was synthesized with primer 105 and amplified by PCR with primers 105 and 52 (Table 1). The PCR fragments were examined on a 1% agarose gel and purified with a Wizard PCR Prep kit (Promega). The 3'-terminal 188 nt from these PCR fragments were amplified asymmetrically by using primer 451 (5'-GGGCTGCTGCCCATTGG-3'; colinear to RUB nt 9577 to 9599) and an ABI PRISM Dye Cycle Sequencing Ready Reaction kit (PE Applied Biosystems) and purified on Centrisep columns (Princeton Separations, Adelphia, N.J.), and the nucleotide sequences were determined by cycle sequencing using an ABI 373 sequencer (The Perkin-Elmer Corporation, Foster City, Calif.).

Purification of MBP-CAL, in vitro kinase reaction, and dephosphorylation reaction. A pMAL-c2 vector construct encoding an MBP-CAL fusion protein was provided by C. D. Atreya (Center for Biologies Evaluation and Research, Food and Drug Administration, Bethesda, Md.). The fusion protein was expressed in Escherichia coli DH5 and purified by maltose chromatography according to the manufacturer's protocol followed by dialysis against 10 mM Tris-HCl (pH 7.4)-10 mM KCl-5% glycerol buffer at 4°C overnight and storage at 20°C. The amount of protein was determined with a Bio-Rad (Hercules, Calif.) protein assay. The in vitro kinase reaction was carried out by incubating 2 µg of purified MBP-CAL in kinase reaction buffer at 30°C for 30 min as described previously (50).

Preparation of cell lysates. Cell lysates were prepared as previously described (37). Briefly, confluent monolayers of Vero cells in 75-cm² T flasks were washed with ice-cold phosphate-buffered saline twice, scraped with a rubber policeman, and pelleted by centrifugation at 400 × g for 10 min. Pelleted cells were resuspended in 200 µl of cytolysis buffer as described above and allowed to incubate on ice for 30 min. The lysate was clarified by centrifugation at 11,000 × g for 15 min, and the supernatant was collected. The amount of protein was determined by a Bio-Rad protein assay.

In vitro synthesis of RNA probe. Construction of plasmids containing the T7 RNA polymerase promoter followed by the 3' 110 nt of each SL2 mutant, a 20-nt poly(A) tract, and an EcoRI site was done by PCR amplification using primers 420 and 105 (Table 1), with EcoRI-linearized SL2 mutant DNA (435, 436, 444, 461, and 430-GAG) as a template. The HindIII-EcoRI-restricted PCR fragments were cloned into pUC18. Construction of plasmids containing the T7 RNA polymerase promoter followed by the 3' untranslated region (UTR) was also done by PCR using primers 498 (Table 1) and 105 [with poly(A) tract and an EcoRI site] or 509 [Table 1; complementary to RUB genome nt 9739 to 9757 to produce a 3' UTR without the poly(A) and the 3' terminal 5 nt but with an EcoRI site], and the HindIII-EcoRI-restricted fragment was cloned into pUC18 (resulting in 498 and 498-A, respectively).

Gel mobility shift assay. Various amounts of cell lysate or purified MBP-CAL were incubated with 10,000 cpm of the different ³²P-labeled probes (~0.6 to 0.9 ng) in cytolysis buffer containing 100 ng of poly(I)-poly(C) (P-L Biochemicals, Inc., Milwaukee, Wis.) and 10 U of RNasin in a final volume of 10 µl for 15 min on ice or 30 min at room temperature. In competition assays, various amounts of unlabeled competitor RNAs were preincubated with the cell lysates prior to addition of the probe. RNA-protein complexes were electrophoresed in nondenaturing 4% polyacrylamide gels (50:1, polyacrylamide-to-bisacrylamide ratio) in Tris-borate-EDTA buffer (90 mM Tris, 90 mM boric acid, 2 mM EDTA) at room temperature. The gel was then dried, and the RNA-protein complexes were visualized by autoradiography.

UV-induced cross-linking. Binding reactions were set up as described above with ³²P-labeled RNA probe (~70,000 cpm) and 20 to 40 µg of cell lysate in each reaction. The binding reactions were then transferred to an ice-water bath and irradiated with a 254-nm UV lamp (UV Crosslinker; Fisher Scientific, Pittsburgh, Pa.) placed 3 cm from the reaction for 30 min. After irradiation, 1 U of RNase A was added to the reaction mixture and incubated for another 15 min at room temperature to remove unprotected RNA probe. The UV cross-linked products were boiled in Laemmli sample buffer for 2 min and resolved on a discontinuous SDS-10% acrylamide gel. The complexes were then visualized by autoradiography (5, 37).

Nucleotide sequence accession number. The sequence shown in Fig. 1A has been assigned GenBank accession no. M15240.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Thermodynamically predicted secondary structure of the 3' cis-acting elements. The secondary structure of the 3'-terminal 305 nt of the RUB genome (the region conserved in DI RNAs) was analyzed by using the sequence of our standard lab strain, fTH, and four SLs were predicted (Fig. 1A). SL1 (G ~ 63.2 kcal/mol), the largest, was between nt 9527 and 9647. SL2 (G ~ 20.3 kcal/mol), which has been shown to interact with CAL, is in the coding sequences at the exact C terminus of the E1 protein coding sequence (nt 9671 through 9702). In the 3' UTR were SL3 (nt 9703 through 9730; G ~ 7.9 kcal/mol) and a very short SL, SL4 (nt 9742 through 9753; G ~ 0.8 kcal/mol). SL4 is connected with SL3 by an 11-nt single-stranded hinge region and followed by a 7-nt single-stranded leader sequence preceding the poly(A) tract.

View larger version (28K): [in this window] [in a new window]   FIG. 1.   (A) Thermodynamically predicted secondary structure of the 3'-terminal 240 nt of the RUB genome (W-Therien strain). A structure with four prominent SLs (SL1 to SL4) and a G of ~99.8 kcal/mol was predicted; SL2 is the structure shown to interact with CAL. Both SL1 and SL2 are within the E1 coding region; the UAG stop codon, located at the base of SL2, is underlined. SL3 and SL4 are within the 3' UTR and are connected by a single-stranded hinge region. SL4 is followed by a 7-nt leader and the poly(A) tract. Nucleotide variations found in other RUB strains are indicated by arrows. fTH, F-Therien (clear-plaque variant derived from W-Therien) (accession no. D00156); M33, M33 (X05259); HPV, HPV77 (vaccine strain derived from M33) (M30776); JUD, Judith (L19420); THM, Thomas (L19421); RA, RA27/3 (vaccine strain currently used in the United States) (L78917). (B) Alternative structures of SL2 and SL3. Structure I (top) was predicted for the W- and F-Therien strains (G ~30 kcal/mol); structure II (bottom) was predicted for the M33/HPV77, Judith, RA27/3, and Thomas strains (both are drawn by reference to the W-Therien sequence) (G ~25 Kcal/mol in the RA27/3 strain). The complementary UAGU and ACUA stretches which pair to form the longer SL3 in the second structure are underlined.

Analysis of the secondary structure of SL2 and SL3 by RNase probing. Considering the apparent importance of the unpaired U loop in CAL binding (37), it was important to confirm the secondary structure of SL2 and SL3. Therefore, we performed RNase probing using RNA transcripts from pUCRUB3'110-fTH and -HPV, plasmids containing the 3' 90 nt plus a 20-nt poly(A) tract derived from the fTh and HPV77 strains, which are predicted to have the first and the second SL2-SL3 conformations, respectively. RNA transcripts from these two constructs were subjected to digestion by several RNases, and the results of RNase digestion were resolved by primer extension. A representative gel is shown in Fig. 2A, and the digestion observed is summarized in Fig. 2B (single-stranded RNases) and C (double-stranded RNase V₁).

View larger version (130K): [in this window] [in a new window]   FIG. 2.   Analysis of SL2/SL3 conformations by RNase probing. RNA probes consisting of the 3'-terminal 90 nt plus a poly(A) tract from the fTH or HPV77 (HPV) strain were digested with a battery of single- or double-stranded RNases, and the digestion pattern was resolved by primer extension. (A) Results of primer extension from pUC3'RUB110-fTH transcripts digested in a 20-µl reaction with no RNase (; lanes 1 and 19); mung bean nuclease (MB; cleaves single-stranded RNA with no nucleotide specificity), 10 (lane 2) and 5 (lane 3) U; RNase T2 (cleaves single-stranded RNA with no nucleotide specificity), 0.5 (lane 4), 0.25 (lane 5), 0.1 (lane 6), and 0.05 (lane 7) U; RNase V1 (cleaves double-stranded RNA with no nucleotide specificity), 0.1 (lane 12), 0.05 (lane 13), and 0.025 (lane 14) U; and RNase A (cleaves single-stranded RNA with preference for C and U), 0.1 (lane 20), 0.05 (lane 21), 0.025 (lane 22), and 0.01 (lane 23) U. In lanes 8 to 11 and 15 to 18 are the sequencing ladders for orientation produced by using the primer extension primer and plasmid pUCRUB3'110-fTH as a template. Structural regions within the probe are shown on both margins. Digestion landmarks highlighted include the loops of SL2 and SL3 (>), both of which are sensitive to the single-stranded RNases but not RNase V1 (the SL3 loop is not sensitive to RNase A because it contains a GAAA sequence), and the single efficient RNase V1 digestion site in SL2 and the digestion of the 5' side of the SL4 stem by RNase V1 (). The overall results of RNase probing on both probes are summarized in panels B (single-stranded RNases) and C (double-stranded RNase V1). Because the primer was complementary to the 11 nt preceding the poly(A) tract (italics), digestion within these nucleotides could not be resolved. Single-stranded RNases: T1, RNase T1 (G residue); T2, RNase T2 (no specificity); A, RNase A (C and U residues); MB, mung bean nuclease (no specificity); U2, RNase U2 (A residue).

Site-directed mutagenesis of 3'-terminal 90 nt. To analyze the significance of the sequences within the 3'-terminal 90 nt of the RUB genome on replication, site-directed mutagenesis was performed with Robo302, a RUB infectious clone. For each mutant construct, RNA transcripts from at least four individual plasmids were used for transfection. For those mutants which did not produce CPE within three passages, mutagenesis, cloning, and transfection were repeated one more time to confirm the observation, while for those mutations which yielded viable virus, the nucleotide sequence was confirmed after one or two passages of the mutant virus in Vero cells by RT-PCR amplification of the 3'-terminal 600 nt and sequencing of the amplification product.

(i) Site-directed mutagenesis of the 3' UTR. Deletion mutagenesis was used to probe the 3' UTR; the mutants are listed in Fig. 3. Large deletions, including one encompassing the complete UTR (315) as well as most of a number of smaller deletions, were nonviable, indicating that most of the 3' UTR is critical for RUB replication. An exception was the 3'-terminal 5 nt as deletions of 1 (328), 2 (334), 3 (335), 4 (359) or all 5 (341) of these nt were viable. These mutated transcripts had transfection efficiencies similar to those of Robo302 transcripts; the plaque morphology and titer produced by each of these mutant viruses were also similar to those for Robo302 virus (~10⁶ to 10⁷ PFU/ml). Viruses were also recovered from mutated transcripts with the 3'-terminal 6 (340) and 7 (323) nt deleted; however, the transfection efficiency of these mutated transcripts was lower (indicative of reversions occurring), and sequence analysis revealed that the viruses recovered had an AG added between the deletion and the poly(A) (Fig. 4). The titer produced by these viruses was about 10- to 100-fold less than that of the WT Robo302 virus (~10⁵ PFU/ml). Deletion of nt 6 to 10 (393) abolished infectivity completely, indicating that the specific sequence is critical for viability rather than the length of this region. Further mutagenesis demonstrated the importance of nt 6 and 7, since a dramatic reduction in transfection efficiency was observed with mutant transcripts lacking these two nucleotides (392). This mutant virus also formed tiny plaques and grew to a 10-fold-lower titer than did Robo302. Virus recovered from this mutant in one transfection preserved the mutated sequence, while virus recovered from a second independent transfection had an insertion and a rearrangement of the downstream 5 nt (UAGUGU).

View larger version (88K): [in this window] [in a new window]   FIG. 3.   Site-directed mutagenesis of the 3' UTR. The Robo302 sequence of the 3' UTR beginning with the UAG stop codon of the E1 gene is given on the top line, and the structural features of the UTR are delineated [in addition to those shown in Fig. 1, a UG(U/C) triplet is marked in bold]. Among the mutations created, deletions are indicated by dashes and substitutions are indicated by lowercase letters. Viability is indicated as (nonviable), + (viable with plaques or CPE apparent in transfected cultures), +1 (CPE apparent after one passage), and +3 (CPE after three passages). Sequences were confirmed on all viable mutants after one amplification in Vero cells; those designated +a contained alterations in the mutant sequence (Fig. 4), while those designated +r contained regenerated poly(A) tracts and formed opaque plaques in transfected cells but clear plaques in the subsequent passages. Mutants which formed plaques similar to those of Robo302 on transfection plaques are inicated as "WT," while those that formed tiny plaques on transfection plaques or no plaques are indicated as "tiny" and "N.P.," respectively. Plaque morphologies indicated in parentheses are those formed after subsequent passaging for amplification.

View larger version (55K): [in this window] [in a new window]   FIG. 4.   Characterization of viruses with mutations in SL4, the leader, and the poly(A) tract. Sequences determined from 15 viable mutants are given (a and b indicate results of independent transfections); altered nucleotides not present in the original transcripts are in lowercase. For comparison, sequences of six nonviable mutants are also given at the bottom. To determine replication ability, mutant viruses isolated from transfection plaques (p) or medium (m) were amplified by one passage in Vero cells, and the titers of the amplified stocks were determined. These stocks were then used for infection of Vero cells (in the first experiment [superscript "1"], which compared the replication of Robo302 and eight mutants, the multiplicity of infection [MOI] was 5; in the second experiment [superscript "2"], which compared Robo302 and six additional mutants, the MOI was 0.01). The infected culture media were harvested at 2 (MOI = 5) or 3 (MOI = 0.01) days postinfection, and the titer (PFU per milliliter) was determined. The relative titer produced by each mutant is shown in comparison with the Robo302 titer (set at 100%) produced in the same experiment (N.D., not determined). The proposed critical stretch of UG(U/C) triplets is underlined.

View larger version (22K): [in this window] [in a new window]   FIG. 5.   Thermodynamically predicted secondary structure of SL3 in virus transcripts with mutations in this region. The native SL3 structure (WT) is at the top. Point mutations are underlined. The entropy of each mutant is also given.

(ii) Mutagenesis in SL2. Because SL2 is within the coding region of E1, mutagenesis was done by creation of point mutations. The point mutations constructed in SL2 included mutations which destabilized the GC-rich stem (436 and 461) or disrupted the U-U bulge by either opening it (435) or base pairing it (444 and 430) (Fig. 6A). Two control constructs were also made: 330, which had the same sequence as the HPV strain at nt 9699; and 419, which had a mutation upstream from SL2 which introduced an NsiI site for cloning. Most of these mutations in SL2 using Robo302 were made at third codons to conserve the coding of E1. The exceptions were 444, in which a leucine (UUG) was replaced with a tryptophan (UGG), 430, in which the stop codon (UAG) at the end of the E1 was changed to lysine (AAG) or glutamic acid (GAG), and 419, in which a leucine (UUG) was changed to a methionine (AUG) (Fig. 6A). All of the mutant transcripts which retained the amino acid sequence (435, 436, 461, and 330) had specific infectivities similar to those of Robo302 transcripts and produced similar plaques; however, two mutants with changes in amino acid sequences (444 and 430-GAG) had lower specific infectivities and formed tiny plaques (Fig. 6A, insert). 419 had a specific infectivity similar to that of Robo302. Additionally, the 430-AAG (Lys) mutant did not form plaques, and the sequence of virus recovered from transfection fluid was found to have changed to CAG (glutamine). These results indicated that the SL2 structure could accommodate substantial changes in structure without affecting viability. The growth curves of the stable SL2 mutants were also characterized (Fig. 6B). Mutants 435, 436, and 461 (mutations which destabilized the SL2 structure) had growth curves similar to that of Robo302, while 444 and 430-GAG (mutants which base paired the entire structure) produced approximately 10-fold-lower titers after two passages.

View larger version (26K): [in this window] [in a new window]   FIG. 6.   (A) Mutagenesis of SL2. Because SL2 is located in the E1 coding region, analysis was done by creating point mutations that resulted in silent changes (436) or synonymous codons (435 and 461). However, to eliminate the U-U bulge, we created two missense mutations (444 and 430-AAG/GAG) which led to changes in coding as indicated. Mutation 419 was made to create an NsiI site for cloning, and 330 was made to conform to HPV77 sequence. The bold characters are the stop codon ending the E1 gene; the nucleotide variations found in other RUB strains are in lowercase. Transfection efficiencies of mutant transcripts are shown as +++ (equivalent to that of Robo302); ++ (about 30 to 50% of that of Robo302) +, (about 5 to 10% of that of Robo302), and N.D., not determinable, but transfected cells inoculated with medium showed CPE. The plaque morphology described is that produced following transfection. Following amplification of 430-AAG from transfection medium, the sequence was found to have changed to CAG. (B) Growth curves of SL2 mutants. Vero cells were infected with viruses amplified from transfection plaques at an MOI of 0.01 PFU/cell. At indicated times postinfection, aliquots of the infected culture media were harvested and titered on Vero cells.

Interaction between CAL and mutated SL2 RNA. To investigate the significance of the SL2 mutations on the interaction between CAL and SL2, binding between a purified MBP-CAL fusion protein and gel-purified ³²P-labeled SL2 RNA probe from each of the individual SL2 mutants was analyzed by gel mobility shift assay. The probes contained additional 10 nt (430-GAG, 435, 436, 444, and 461) or 8 nt (pUCRUB3'110-fTH and -HPV) at their 5' ends but had the exact 3' end of SL2 (due to the presence of a Bsp120I site immediately following SL2) and thus were similar to those used by Nakhasi et al. in their binding assays (2, 37). Binding of MBP-CAL to SL2 RNA from Robo302 is shown in Fig. 7A; as can be seen, increasing amounts of MBP-CAL-SL2 RNA complexes were formed with increasing input concentrations of protein. It was reported that MBP-CAL is able to be autophosphorylated in vitro and that in vitro phosphorylation was required for the interaction between MBP-CAL and SL2 RNA (2). However, in our hands, in vitro phosphorylation was not required for binding, as shown in Fig. 7A. Interestingly, we found that in vitro phosphorylation was extremely difficult to detect and efficient phosphorylation of MBP-CAL in vitro required addition of a Vero cell lysate (incubation of MBP-CAL in a phosphorylation reaction, either with or without cell lysate, made no difference on its ability to bind SL2). In one experiment, autophosphorylation was detectable; however, this reaction used a preparation MBP-CAL which had been stored frozen for a long period of time (~6 months), and prolonged autoradiography was required for detection (in comparison with the companion reaction containing a Vero cell lysate, the intensity of the MBP-CAL band produced by autophosphorylation was only about 3.5% of the band produced in the presence of cell lysate) (data not shown). To test the possibility that MBP-CAL was phosphorylated in the bacteria, either by autophosphorylation or by bacterial kinases, MBP-CAL was dephosphorylated with CIAP. The efficiency of CIAP dephosphorylation of MBP-CAL phosphorylated in vitro in the presence of [³²P]ATP and Vero cell lysate was about 85% (data not shown). To examine whether CIAP treatment would affect the binding activity of MBP-CAL, untreated and CIAP-treated MBP-CAL fusion proteins were incubated with SL2 probe and analyzed by gel mobility shift assay; however, no significant difference was observed between treated and untreated proteins (data not shown). Thus, although SL2 binding by MBP-CAL appeared to be largely independent of phosphorylation, because dephosphorylation may not have been complete, we cannot rule out that phosphorylation of key sites is required for SL2 binding.

View larger version (69K): [in this window] [in a new window]   FIG. 7.   Binding of MBP-CAL to SL2 probes. (A) Robo302 SL2 probe (<1 ng) was incubated in a binding reaction with various amounts of MBP-CAL prior to resolution of binding by electrophoresis in a nondenaturing polyacrylamide gel. Lane 1, control (0 µM MBP-CAL); lanes 2 to 7, 0.125, 0.25, 0.5, 1, 2, and 4 µM MBP-CAL. (B) SL2 probes containing sequences of WT Robo302 (lane 1 to 3) and mutations 430-GAG (lanes 4 to 6), 444 (lanes 7 to 9), 435 (lanes 10 to 12), 436 (lanes 13 to 15), and 461 (lanes 16 to 18) were incubated with 0 µM (lanes 1, 4, 7, 10, 13, and 16), 0.25 µM (lanes 2, 5, 8, 11, 14, and 17), or 1 µM (lane 3, 6, 9, 12, 15, and 18) MBP-CAL. After binding, formation of RNP complexes was determined by electrophoresis in polyacrylamide gels under nondenaturing conditions. P, probe only.

Interaction of MBP-CAL with SL2 is dependent on SL2 structure. To compare the binding activities of MBP-CAL to the SL2 mutants, ³²P-labeled SL2 probe containing each of the mutations was incubated in a binding reaction with increasing amounts of MBP-CAL (Fig. 7B). Interestingly, mutations in which the U-U bulge was base paired (444 and 430-GAG) had a binding activity similar to that of Robo302 SL2 (Fig. 7B, left panel) while mutations with destabilization in the U-U bulge (435) or the GC-rich stem (436 and 461) had a lower binding activity (Fig. 7B, right panel), indicating that maintenance of the entire SL2 stem structure is critical in CAL binding. The interaction between MBP-CAL and mutant SL2s was also quantified as the percentage of total radioactivity in the MBP-CAL-SL2 RNP over an extensive range of input MBP-CAL concentrations (Fig. 8); an SL2 probe with the HPV77 sequence was also included in these assays. 444 and 430-GAG had high activity similar to that of Robo302, reaching 50% binding between 2 and 3 µM MBP-CAL, while 435, 436, 461, and HPV77 never attained over 5 to 10% binding. Considering that viruses containing these mutations were viable and most replicated to similar titers as did Robo302 virus, the interaction between SL2 and CAL does not appear to be critical for RUB replication.

View larger version (18K): [in this window] [in a new window]   FIG. 8.   Comparison of binding activities of Robo302 and mutated SL2s to MBP-CAL. Robo302 and mutant SL2 probes were incubated with increasing amounts of MBP-CAL, and formation of RNP complexes was determined by gel mobility shift (an example of binding of MBP-CAL and Robo302 SL2 is shown in Fig. 6A). The binding was quantitated as the percentage of total radioactivity (bound plus unbound) in RNP complexes.

Investigation of interaction between cellular factors and 3' UTR. We wished to know whether the finding that the 3' UTR is critical in RUB replication is related to its interaction with host factors. As shown in Fig. 9A, three RNP complexes were detected by gel mobility shift assay when a ³²P-labeled 3' UTR probe was incubated with a Vero cell lysate, indicated as RNP I, RNP II, and RNP III. RNP II was detectable when incubated with a lower amount of cell lysate (~0.75 µg), while detection of RNP I and III required higher amounts of cell lysate, indicating that the host factors composing RNP II either had a higher activity for the RUB 3' UTR or were more abundant in the cell lysate preparations. The specificities of these three RNPs were examined by competition assay using several competitors. The specific competitor, 3' UTR, was able to outcompete the probe for binding at the lowest concentration used (20× molar excess), while nonspecific competitors SL2 RNA and poly(I)-poly(C) did not outcompete for binding even at the highest concentration used. However, tRNA was able to outcompete the 3' UTR probe at the 20× molar excess for RNP I and 150× excess for RNPs II and III. Six proteins with molecular masses of 120, 80, 66, 55, 48, and 36 kDa were detected after UV-induced cross-linking; of these, the 120-, 66-, and 55-kDa proteins were the three major species (Fig. 9C).

View larger version (105K): [in this window] [in a new window]   FIG. 9.   Binding of cellular factors to the 3' UTR. (A) 32P-labeled UTR probe (lacking both the poly(A) tract and the 3'-terminal 5 nt was incubated with 0 µg (lane 1), 0.2 µg (lane 2), 0.375 µg (lane 3), 0.75 µg (lane 4), 1.5 µg (lane 5), 3 µg (lane 6), 6 µg (lane 7), and 12 µg (lane 8) of cytoplasmic lysate from uninfected Vero cells, and formation of complexes was detected by electrophoresis in nondenaturing polyacrylamide gels. The three RNP complexes detected are indicated on the right as RNP I, II, and III. (B) Increasing amounts of different competitor RNAs, including both nonspecific [SL2, poly(I)-poly(C), and yeast tRNA] and specific (3' UTR RNA transcripts) competitors, were incubated with 4 µg of Vero cell lysate prior to the addition of the 32P-labeled probe. Lane 5: 1 (P), probe alone; 2 (C), positive control without any competitor; 3 to 5, SL2 at 20× (lane 3), 50× (lane 4), or 150× (lane 5) molar excess; 6 to 8, poly(I)-poly(C) at 50 ng (lane 6), 100 ng (lane 7), or 200 ng (lane 8) [the standard binding assay buffer containing contains 100 ng of poly(I)-poly(C)]; 9 to 11, yeast tRNA at 20× (lane 9), 50× (lane 10), or 150× (lane 11) molar excess; 12 to 14, 3' UTR RNA at 20× (lane 12), 50× (lane 13), or 150× (lane 14) molar excess. (C) 3' UTR probe in the absence (lane 2, P [probe only]) or presence (lane 3) of 30 µg of Vero cell lysates was exposed to UV light and digested with RNase. Cross-linked proteins were revealed by SDS-polyacrylamide gel electrophoresis. Approximate positions of size markers (lane M) are given on the right.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The elements in the 3'-terminal 110 nt in the RUB genome, including the poly(A) tract, critical in virus replication were analyzed in this study by site-directed mutagenesis using a RUB infectious clone, Robo302. Previous studies on RUB DI RNAs found that the 3'-terminal 305 nt were retained in all DI RNAs generated during serial undiluted passage, and it was concluded that these sequences contained the 3' cis-acting elements required for replication (7, 11). Thermodynamically predicted secondary structures of this 305-nt region revealed four prominent SL structures, including two SLs (SL1 and SL2) in the E1 coding region and two in the 59-nt 3' UTR (SL3 and SL4). Because of its size and location in the E1 coding region, SL1 was not analyzed in this study.

Two alternative structures containing SL2 and SL3 were predicted, depending on the virus strain. Because previous studies indicated that the interaction of SL2 with CAL was dependent on the presence of the U-U bulge (37), which did not exist in the second structure, RNase mapping was done to analyze the SL2/SL3 structure in both the fTh and HPV77 strains, in which the first and second structures, respectively, were predicted. The RNase mapping data were consistent with the formation of the first structure in both strains. It is likely, however, that the two structures achieve equilibrium in both strains since both single- and double-stranded RNase digestion of the ACUA in the hinge region instrumental in forming the second structure was observed in both strains. However, this could also be due to nonspecific digestion by RNase V₁, since it has been noticed that RNase V₁ may cleave within a single-stranded region if base stacking occurs (28). RNase mapping also confirmed the loops of SL2 and SL3 and the predominantly single-stranded nature of the hinge region between SL3 and SL4. The bulge and upper stem of SL3 were digested by single- and double-stranded RNases in a pattern that was generally, but not completely, consistent with the predicted structure. Single-stranded RNases failed to cleave the unpaired U loop in SL2, and the double-stranded RNase did not cut the lower stem of SL3 and cleaved at only one major site in the GC-rich SL2 stem. These data possibly indicate the existence of a tertiary structure between SL2 and SL3 that could explain the inaccessibility of SL2 and unexpected aspects of the cleavage pattern of SL3. The existence of SL4 is tentative, as it is a thermodynamically weak structure and varies to some extent among strains according to the thermodynamic prediction. However, RNase mapping confirmed its existence.

Our mutagenesis data revealed that most of the 3' UTR is required for replication, in contrast to the case for both alphaviruses (24) and picornaviruses (54), in which large deletions within the 3' UTR can be accommodated. Recently, mutagenesis resulting in complete removal of the 3' UTR from poliovirus type 1 and human rhinovirus type 14 infectious clones with recovery of viable virus was reported (54). However, these mutant transcripts had much lower transfection efficiency than WT transcripts with the 3' UTR, and the 3' UTR minus-strand virus was impaired in replication (onset of CPE was observed at 18 to 24 h posttransfection with the latter mutants, compared with 8 h posttransfection with WT virus). The only region in the RUB 3' UTR that could be deleted is the single-stranded leader region in which nt 1 to 5 could be deleted, and the resulting virus was phenotypically WT. This is also in contrast to alphaviruses, in which the nucleotide preceding the poly(A) tract was necessary for viability, although the insertion of nucleotides between this nucleotide (a C) and the poly(A) could be tolerated (24). Addition of nonviral sequence in the RUB single-stranded leader region was tolerated since mutants having poly(A) replaced with 10 U's (508T) or with an EcoRI restriction site (507 and 509) were viable and retained these sequences along with a regenerated poly(A) tract; these viruses were also phenotypically WT.

The only two nucleotides in the single-stranded leader sequence required for viability were nt 6 and 7. Deletions of these nucleotides either by themselves or in the context of the entire leader resulted in a dramatic reduction in transfection efficiency, and addition of nucleotides downstream prior to the poly(A) tract occurred in the viruses which were recovered; the replication of these viruses was impaired to some extent. Addition of nonviral sequences at a similar location by viral RdRp has been described for Sindbis virus (SIN) (46), although an AU-rich motif was added instead of the G residue usually added in RUB 3' terminus [although nonviral AAT or AATT also exists in some poly(A) mutants, these sequences are actually copied from an EcoRI protruding end used for runoff transcription]. Despite SL4 not being a particularly stable structure according to thermodynamic prediction, RNase mapping data indicated that SL4 existed in the native sequence and that the nucleotides within this small SL are critical in viral replication. A number of deletions within the SL4 sequence were lethal (only clone 524, in which nt 8 to 10 were deleted, was viable) as were most of the mutations which changed the sequence (only clone 575, in which the UGU in the loop was changed to ACA, was viable). However, the virus recovered from 575 restored the SL4 sequence, while the virus recovered from 524 restored all but the last nucleotide. From available data, we cannot conclude whether the primary sequence or the SL4 structure is the important feature. While the SL4 sequences were maintained in all viable mutants and revertants, the SL4 structure is not predicted thermodynamically in some of them when the complete 3' 110-nt segment is folded.

Deletions in the hinge region were also lethal. Interestingly, one of these deletions of the ACUA, which formed part of the second SL2/SL3 structure (344C), was lethal, which implied that formation of this alternative structure might be important in viral replication. While the sequences in the upper stem of SL3 could be switched, mutations which changed the structure were lethal. An exception was mutant 391, in which the SL3 GAAAC loop was deleted. However, the mutant was highly attenuated. Given these findings, it is tempting to speculate that the significance of SL3 is in specific interactions with the RUB RdRp and/or cellular factors since bulges or loops in an RNA secondary structure have been implicated as specific recognition sites for RNA binding proteins because they expose RNA backbones and bases to interaction while the regular helical RNA stems are less useful (reviewed in reference 32). The requirement of a secondary structure is in contrast to alphaviruses, which share a consensus 19-nt stretch immediately preceding the poly(A) tract which contains no significant secondary structure.

Like SIN replication (19), RUB replication can apparently occur independent of the poly(A) tract; however, also as for SIN virus, the poly(A) tract was rapidly regenerated. The role of poly(A) in eukaryotic mRNAs is proposed to be in mRNA stability and translational efficiency (reviewed in reference 59). Opaque plaques were formed after transfection by mutants with defects in poly(A) tract; after one passage when poly(A) regeneration was detected, these viruses produced large, clear plaques and grew to titers similar to Robo302 (Fig. 4). It is not clear how the nontemplated polyadenylation occurs. Unlike the case for coronavirus (20) or vesicular stomatitis virus (3), which might utilize a short poly(U) as a template for polyadenylation, polyadenylation of RUB genomic RNA is not due to polymerase reiterating at a poly(U) stretch since there are no such elements found in the 3' UTR or its minus-strand complement in the RUB genome. The cellular cytoplasmic polyadenylation machinery, cellular terminal transcriptase-like enzymes, and the viral RdRp were implicated in the regeneration of the poly(A) tract in SIN (46).

Because of its location within the E1 coding sequences, substitution mutagenesis was used to investigate the interaction of SL2 with CAL and its effect on virus replication. The CAL binding site on SL2 has never been precisely mapped, although it has been shown that deletion of the U-U bulge abolished binding of cellular proteins (37), including the one later identified as CAL, and the binding of MBP-CAL to SL2 was abolished when the U-U bulge was replaced with A-A, C-C, or G-G bulges (1a). Because of the constraints imposed by E1 coding, we were unable to make these mutations. Therefore, we constructed mutations that either base paired the U-U bulge or created other bulges in this SL structure. We found that the mutations with the U-U bulge base paired had as great an affinity for MBP-CAL as did the native SL2 with the U-U bulge. This finding does not necessarily contradict the findings of Nakhasi et al., since all mutants maintained at least one of the two U residues, and it has been shown that the members of a U-U bulge in an SL structure can be hydrogen bonded (58), which would also explain why no cleavage was observed at the U-U bulge with any single-stranded RNase. In contrast, a mutant in which the SL in the region of the U-U pair was destabilized by the introduction of noncomplementary nucleotides had the lowest activity for MBP-CAL. In addition to the U-U bulge, maintenance of the GC-rich stem in SL2 was also critical in CAL binding since reduction in binding activity was also observed in mutants which destabilized the GC-rich SL2 stem, including 436, which was based on the sequence of the M33/HPV77 strain. (Although binding with CAL was first performed with an HPV77 probe, the unpaired CA bulge in the upper stem of SL2 of HPV77 [Fig. 2B] was changed to a CG base pair in subsequent studies [37]). All of these findings indicate that maintenance of a long stem is the primary requirement for MBP-CAL binding. Interestingly, CAL was also found to bind with hY RNA (8), whose structure also exhibits a long stem which was confirmed by enzymatic structure probing (55). Therefore, it is possible that SL2 mimics the structure feature of hY RNA which is recognized by CAL.

Although many attempts have been made to determine the significance of the host factor-viral RNA interaction on virus replication, most of the studies were done with a reporter gene system. Ours is the first direct study of the effects of a host factor's interaction with a viral cis-acting element in virus replication using a virus infectious clone. Interestingly, despite the differences observed in the binding activity between SL2 mutants and MBP-CAL, maintenance of the SL2 structure required for CAL binding was not required for viral replication. Mutants having lower binding affinities (435, 436, and 461) without changes in amino acid sequence had similar transfection efficiencies, formed similar plaques, and grew to titers that were similar to or somewhat lower than those for Robo302. Mutants having U-U bulge base paired exhibited lower transfection efficiencies, tiny plaques, and lower titers, and some of these mutants (444 and 430-AAG) reverted to the WT Robo302 sequence within a few passages. This is likely due to the changes in these mutants of the amino acid sequence in the C terminus of E1, which is proposed to function as a cytoplasmic tail of the E1 proteins that interacts with the capsid in virions. In one case, a CAG revertant was recovered from the 430-AAG mutant, which, interestingly, resulted in glutamine residue; 430-GAG which was stable encoded a glutamic acid at that position.

Altogether, these data indicate that binding of CAL to SL2 is independent of RUB replication. However, despite the decreased MBP-CAL binding affinities of some of the viable SL2 mutants, some binding was still observed. Thus, it is still possible that CAL and SL2 in the RUB genome interact in infected cells, particularly if local CAL concentrations are high. In this regard, RUB RNA replication was recently confirmed to occur within subcellular vesicles of lysosomal origin (31). Since CAL is found primarily associated with endoplasmic reticulum membranes, it would have to be relocalized to interact with the RUB RNA during replication. Relocalization of cellular proteins that bind viral RNA has been found upon viral infection (10, 27; reviewed in reference 4) although relocalization of La was not observed during SIN replication (40). It would be of interest to determine if CAL relocalizes in RUB-infected cells.

In contrast to SL2, our mutagenesis data clearly demonstrated the importance of the 3' UTR in RUB replication; however, the function could be in translational efficiency as well as serving as a promoter for minus-strand RNA synthesis. Nakhasi et al. showed that the RUB 3' sequences enhanced translation of a reporter gene expressed in context with the 5' SL structure (38, 42). However, binding of cellular factors has been shown to a number of putative cis-acting replicational elements, and thus binding of cellular proteins to the 3' UTR was investigated. Three RNA complexes were resolved by gel mobility shift assay using a 3' UTR probe lacking the 3'-terminal 5 nt, and poly(A) and UV-induced cross-linking revealed that six protein species bound to the 3' UTR probe. The interaction was not outcompeted by nonspecific competitors such as SL2 RNA or poly(I)-poly(C) but was inhibited by 150× molar excess amounts of yeast tRNA. However, we also found that the interaction between SL2 and MBP-CAL can be outcompeted by a 30× molar excess of yeast tRNA (data not shown). Thus, the interaction between these cellular factors and the 3' UTR appears to be specific. However, further characterization of these proteins and these interactions with the 3' UTR is necessary to define their role(s) in viral replication.