p33-Independent Activation of a Truncated p92 RNA-Dependent RNA Polymerase of Tomato Bushy Stunt Virus in Yeast Cell-Free Extract

Abstract

Plus-stranded RNA viruses replicate in membrane-bound structures containing the viral replicase complex (VRC). A key component of the VRC is the virally encoded RNA-dependent RNA polymerase (RdRp), which should be activated and incorporated into the VRC after its translation. To study the activation of the RdRp of Tomato bushy stunt virus (TBSV), a small tombusvirus of plants, we used N-terminal truncated recombinant RdRp, which supported RNA synthesis in a cell-free yeast extract-based assay. The truncated RdRp required a cis-acting RNA replication element and soluble host factors, while unlike the full-length TBSV RdRp, the truncated RdRp did not need the viral p33 replication cofactor or cellular membranes for RNA synthesis. Interestingly, the truncated RdRp used 3′-terminal extension for initiation and terminated prematurely at an internal cis-acting element. However, the truncated RdRp could perform de novo initiation on a TBSV plus-strand RNA template in the presence of the p33 replication cofactor, cellular membranes, and soluble host proteins. Altogether, the data obtained with the truncated RdRp indicate that this RdRp still requires activation, but with the participation of fewer components than with the full-length RdRp, making it suitable for future studies on dissection of the RdRp activation mechanism.

INTRODUCTION

The genomes of plus-strand RNA [(+)RNA] viruses are replicated via a multistep process in the infected cells. Translation of the viral (+)RNA leads to the production of the viral replication proteins, which then facilitate the selection/recruitment of viral RNA into replication. In addition, viral RNA replication also requires the assembly of the viral replicase, which performs complementary [minus-strand (−)] synthesis, on subcellular membrane surfaces. The (−)RNA intermediate is then used by the viral replicase to synthesize an excess amount of new (+)RNA progeny, which are released from the site of replication to the cytosol (1, 5, 35, 56, 63). Interestingly, the viral replicase is proposed to function only when bound to the subcellular membrane, which likely prevents collision between the ribosome (progressing in a 5′-to-3′ direction on the viral RNA) and the replicase in the cytosol (16, 32, 50, 56). Also, replicase inactivity in the cytosol prevents the formation of double-stranded RNAs (dsRNAs), whose presence would induce efficient gene silencing and RNA-induced innate immunity (21, 24, 57, 59). Therefore, the assembly and activation of the viral replicase seem to be critical regulatory steps during the infection cycle.

Assembly of the viral replicase complex on the cytosolic surfaces of intracellular membranes is poorly understood (4, 56). The viral replicase complex (VRC) consists of virally encoded RNA-dependent RNA polymerase (RdRp) and viral auxiliary replication proteins, as well as a not yet fully defined number of coopted host proteins (12, 14, 26, 31–33, 36, 38, 56, 62). For example, the tombusvirus replicase complex contains the heat shock protein 70 molecular chaperone (Hsp70, encoded by SSA1/2 in Saccharomyces cerevisiae), glyceraldehyde-3-phosphate dehydrogenase (GAPDH, encoded by TDH2/3 in yeast), which is also an RNA binding protein, Pdc1p (pyruvate decarboxylase), Cdc34p E2 ubiquitin conjugating enzyme, eukaryotic translation elongation factor 1A (eEF1A), eEF1Bγ, and the Ded1p DEAD box helicase (2, 3, 17, 23, 25, 27, 28, 50, 58, 60, 68–70). While Hsp70 and eEF1A are involved in the assembly of the viral VRC, the functions of host RNA-binding proteins eEF1A, eEF1Bγ, GAPDH, and Ded1p are to facilitate viral RNA synthesis by the tombusviral VRC (33, 50, 68–70). In addition, two temporary resident proteins, the Pex19p shuttle protein (47) and the Vps23p adaptor ESCRT protein, also contribute to the tombusvirus VRC assembly (2, 3, 27). Similarly, a purified Tomato mosaic virus (ToMV) replicase preparation contained at least five host proteins, including an RNA-binding protein (GCD10; one of the subunits of the 10-component eIF-3 complex) (39), eEF1A, HSP70, Tom1p (a seven-path membrane protein), and the ARL8 small GTP-binding protein (22, 37, 75). The highly purified Brome mosaic virus (BMV) replicase contained ∼10 host proteins, including reticulons and the p41 subunit of the eIF-3 complex (15, 53).

An important feature of RdRps of several (+)RNA viruses, such as Tomato bushy stunt virus (TBSV) p92^pol, BMV 2a^pol, P2 of Alfalfa mosaic virus (AMV), ToMV 180K, and Hepatitis C virus (HCV) NS5B, is that they have to become activated after the assembly of the viral replicase in membranous spherules or vesicles (44, 45, 52, 67). In spite of intensive efforts, the assembly mechanism of the VRC and the mechanism of RdRp activation are currently unknown.

Characterization of the activity of VRCs has been performed with several viruses by obtaining solubilized template-dependent viral replicases by using nonionic detergents with or without RNase treatment (to remove the endogenous viral RNA) (19). These preparations and purified recombinant viral RdRps have been used to dissect cis-acting RNA elements and trans-acting protein factors involved in initiation, elongation, and termination of viral RNA synthesis or in RNA recombination. These replicase preparations, however, have limitations in addressing the assembly and activation of the VRC.

To circumvent some of the above problems, in the current study we tested the activities of truncated tombusvirus RdRps, which showed unusual characteristics in template recognition and RNA synthesis and exhibited premature termination. Also, these truncated RdRps required fewer factors than the full-length RdRp for activation to synthesize RNA. The truncated tombusvirus RdRps with new features will be helpful for increasing our understanding of the regulatory factors needed for the assembly and activation of VRCs in cells.

MATERIALS AND METHODS

Yeast and Escherichia coli expression plasmids.

The E. coli-based expression plasmid pMalTBSV92Δ167AAΔRPR was constructed in 3 steps. In the first step, the TBSV p92 open reading frame (ORF) was PCR amplified using pMalTBSV92 (54) and primers 2079 (GTCGTCTAGAATGGAGACCATCAAGAGAATG) and 3529 (CCAGCTGCAGTCAAGCTACGGCGGAGTCGAGG), digested with XbaI and PstI, and cloned into XbaI-PstI-digested pMalc-2X plasmid (NEB) to create the pMalTBSV92XP plasmid, with restriction sites suitable for further cloning steps. In the second step, the RPR motif (54) was deleted from pMalTBSV92XP, creating pMalTBSV92ΔRPR: the N-terminal portion was PCR amplified with primers 2079 and 3522 (GGAGGCTAGCCCCAGTGGACGCGATCACC), while the C-terminal portion was amplified using primers 3521 (GGAGGCTAGCTATGCGGCAAAGATCGCAC) and 3529. The resulting PCR products were digested with NheI and ligated together. The final PCR product was PCR amplified using the previous ligation mix and primers 2079 and 3692 (CCAGTTCGAACCATCTTCCAACCGCCTTG) and then digested with XbaI and Bsp119I, followed by ligation into a similarly digested pMalTBSV92XP plasmid, thus producing pMalTBSV92ΔRPR. The N-terminally truncated version of p92ΔRPR was produced in the third step: the PCR product obtained with primers 183 (GAGGAATTCGCACGAGCACACATGGAG) and 3529 and template pMalTBSV92ΔRPR was digested with EcoRI and PstI and cloned into pMalc-2X digested with EcoRI and PstI. This resulted in the final clone, pMalTBSV92Δ167AAΔRPR.

Plasmids pMalTBSV92Δ167AAΔS1 and pMalTBSV92Δ167AAΔS2 were produced in a similar fashion, except that in the second step we used primers 2079 and 4832 (GGCAGCTAGCATTCTCTGGACTGTTCTTAAG) and primers 4833 (GGCAGCTAGCATCATGGACAAAGACTGCGTC) and 3529 for PCRs to introduce the S1 deletion, with primers 2079 and 4834 (GGCAGCTAGCAATAGCCAAAGGCAATATGAC) and primers 4835 (GGCAGCTAGCTCGGCGGCACTATGGGGCTCAC) and 3529 used to introduce the S2 deletion within the N-terminally truncated p92 ORF. The cloning of pMalTBSV92Δ83-153AA was accomplished in a similar manner, using PCRs with primers 2079 and 4830 (GGCAGCTAGCATCACCCCTCTGCCGTCTCTTC) and primers 4831 (GGCAGCTAGCCCTAGGGAAAAACTGTCGGTATTTAAG) and 3529, but the third cloning step was omitted.

pMalTBSV92Δ80AA was PCR amplified using primers 4827 (GGCATCTAGAGGTGATTACATTGTCCCTCTATC) and 3529 and the pMalTBSV92 template (54), digested with XbaI and PstI, and cloned into XbaI- and PstI-digested pMalc-2X. pMalTBSV92Δ98AA was cloned using a similar approach, based on PCR with primers 4828 (GGCATCTAGAAGTTATGCCACTAGGGTACGC) and 3529.

To express Cucumber necrosis virus (CNV) p92 (C92) with a deletion of the N-terminal 150 amino acids (aa), pGADHis92delta150 (40) was digested with BamHI and XhoI, and then the released insert was cloned into pESC-His/Cup/HF (61) digested with BamHI and XhoI. To express CNV p92 with an N-terminal deletion of 167 aa in yeast, we used primers 4914 (CCAGGGATCCGCCCGAGCACATATGGAGGATG) and 952 (CCCGCTCGAGTCATGCTACGGCGGAGTCAAGGA) and a pCupHisCNV92 template (unpublished data) for PCR. The resulting PCR product was digested with BamHI and XhoI and cloned into pEsc-Leu/Cup/Flag/ssa (50) digested with BamHI and XhoI. Note that all of the CNV replication proteins were from yeast, while the TBSV replication proteins, expressed as MBP fusion proteins, were purified from E. coli throughout this work.

Preparation of RNA templates for in vitro RdRp/replicase assays.

To study the effects of various RNA templates on the activity of the truncated p92 protein, we produced the following set of RNA templates by using T7 polymerase-based transcription on PCR templates all carrying the T7 promoter: 5′Δ43, 5′Δ69, and ΔRI, obtained using primers 415 and 1190, primers 122 and 1190, and primers 17 and 1190, with pDI72SXP (71) as the template; and ΔR2, ΔR3, and DI-73, obtained using primers 359 and 1190 and pDI72SXPΔR2, pDI72SXPΔR3 (42), or pDI73 (71) as the template. DI-72 (−)RNA was produced from the PCR product obtained with primers 20 and 22 and the pDI72SXP template (71). DI634 (+)RNA was produced from the PCR product obtained with primers 3508 and 3509 and the pFHV634 template (11). T7-based in vitro transcription reactions were done after the PCR amplification and purification of PCR products (41).

To test the role of cis-acting sequences in the activation of tombusvirus polymerase, we utilized RNA templates already characterized for the ability to assemble active viral replication complexes (44, 46). The list included DI-mini (construct 42; wild-type [wt] minimal template [44]), DI-mini (RII*) [construct 57; minimal template with a C-to-G mutation in RII(+)-SL], DI-mini (SL3*) (construct 59; minimal template with a C-to-G mutation in the SL3 hairpin), and DI-mini (SL1*) (construct 212; minimal template with a G-to-C mutation in the SL1 hairpin) obtained with PCR using primers 1300 and 1190 or primers 1300 and 343.

To produce RNA templates containing duplicated RII(+)-SL sequences, the minimal wt template (construct 42 [44]), and the minimal template with a C-to-G mutation in RII(+)-SL (construct 57 [44]) were recloned into pUC19, downstream of a T7 promoter and upstream of a SmaI restriction site to facilitate the production of linearized templates for the subsequent T7-based RNA transcription. We used primers 4887 and 4888 to amplify constructs 42 and 57 (44). The PCR products were digested with EcoRI and HindIII and cloned into similarly digested pUC19, resulting in constructs pUC/181 and pUC/210, respectively. In order to introduce several restriction sites into these constructs, we amplified the satC sequence of Turnip crinkle virus (TCV) (6) by a PCR using primers 1764 (CGAGGCATGCAGATCTGGGATAACTAAGGGTTTCATAC) and 1766 (CGAGGCATGCGTCGACGGGCAGGCCCCCCGTCCGAG), digested the product with SphI, and cloned it into the SphI sites present in pUC/181 and pUC/210, located between RII(+)-SL and the SL3 hairpin, resulting in intermediate plasmids pUC/181/satC and pUC/210/satC. This cloning step introduced BglII and SalI restriction sites, which were utilized in the following steps. The wt and mutant RII(+)-SL constructs were PCR amplified using primers 1781 (CGAGAGATCTTCAGGAAAGCGGTTTGTGAGAAGGTTGGGGT) and 1782 (CGAGGTCGACGTGATATGCAGACTCTCCACGGCTC) and templates pDI72 and pDI72CtoG (51). The PCR products were then digested with BglII and SalI and ligated into similarly digested pUC/181/satC and pUC/210/satC, resulting in the constructs DI-mini/2xRII-C/G, DI-mini/RII+RII-C/G, DI-mini/RII-C/G+RII, and DI-mini/2xRII.

To produce an artificial RNA template, called mini-art-stem, which could support 3′-terminal extension (3′TEX) in the in vitro RdRp assay, first we obtained a PCR product containing an artificial sequence (GCATGCATAAATGAATGAATGAATGAAATGAATGAATGAATGAATGAAAGGACTTCGGTCC; underlined sequences are an SphI site and the 3′-terminal hairpin, respectively) between the RII(+)-SL and SL3 hairpin sequences. For this purpose, we used primers 1831 (GGCGGCATGCAAATGAATGAATGAATGAATGAAACGGTTGATCTCACCCT) and 1069 and construct 42 (44) as the template for PCR. A second PCR was then performed with primers 1832 (GGCGGCATGCATAAATGAATGAATGAATGAAATGAATGAATGAATGAATGAAACGG) and 1069 and the first PCR product as the template. The resulting PCR product was digested with SphI and SacI and cloned into similarly digested construct 42. This construct then served as the template for a final PCR with primers 1300 and 4911 (GGACCGAAGTCCTTTCATTCATTCATTCATTCATTTC) to create a product that contained the RII(+)-SL sequence at the 5′ end, followed by an SphI restriction site, the artificial sequence, and a very stable heterologous hairpin with a tetraloop (10). To make mini-art-AA and mini-art-AAA, with two and three non-base-paired nucleotides at the 3′ end, we used primers 1300 and 4926 (TTGGACCGAAGTCCTTTCATTCATTCATTCATTCATTTC) and primers 1300 and 4927 (TTTGGACCGAAGTCCTTTCATTCATTCATTCATTCATTTC), respectively.

Preparation of cell-free yeast extract.

Cell-free yeast extracts (CFE) were prepared as described previously (49, 50), using yeast strain BY4147 after inducing the expression of CNV p33 and p92 replication proteins from the CUP1 promoter for 40 min (using 50 μM copper sulfate in the growth medium). The second type of CFE was prepared from BY4741 yeast not expressing tombusvirus proteins and RNA (50).

Expression and purification of recombinant tombusvirus replicase proteins from E. coli.

Tombusvirus replicase proteins were expressed and purified as previously described (55), except that the Tris buffer was replaced by 20 mM HEPES buffer during purification and the elution buffer also contained 1 mM dithiothreitol (DTT) instead of 10 mM β-mercaptoethanol.

Replication assay based on CFE.

The CFE-based replication assay was done as described previously (49, 50), except that the preincubation step was omitted. Instead, the complete reaction mixture was incubated on ice for 5 min before being switched to the reaction temperature, usually room temperature. The reaction buffer contained 50 mM instead of 150 mM potassium acetate. The reaction mix also included 0.5 μg each of recombinant replicase proteins, such as TBSV p33 and/or TBSV p92 and TBSV p92 derivatives, as specified. The “soluble” (supernatant) fraction of yeast CFE was prepared by centrifugation for 20 min at 42,000 × g at 4°C.

The S1/S2 peptide (RLIYQRVMIEIMDKDCVRYVDRDVILPLAIGCCFVYPDGVEES) and the RPR peptide (TKVIASTGRPRRRPYAAKIAQ) were synthesized by Peptide2.0. These peptides were dissolved in RdRp buffer (final concentration, 10 mg/ml). The peptides were added at the beginning of the replication assays, to reach final concentrations of 0.05, 0.17, and 0.5 μg/μl.

FHV replicase purification from yeast and in vitro replicase assay.

To obtain flock house virus (FHV) replicase, yeast strain BY4741 was transformed with plasmids pGAD/Cup/FHV/proteinA/C-term/HA/FLAG and pESC-His-GAL1::FHVRNA1frameshift (23). FLAG-based affinity purification of the FHV replicase and the subsequent replicase assay were performed as described previously (23). The S1/S2 and RPR peptides were applied in the FHV replicase assay as described above.

Micrococcal nuclease treatment.

The replication/RdRp assays were conducted as described above, except that the samples were treated with micrococcal nuclease (final concentration of 0.04 U/μl in the presence of 1 mM CaCl₂) at the 50-min time point for 15 min at room temperature. After the digestion step, 2.5 mM EGTA was added to the micrococcal nuclease-treated samples to inactivate the nuclease. The reaction mix was further incubated for a total of 3 h (counted from the beginning of the RdRp assay) at room temperature before the products were purified by phenol-chloroform extraction and precipitated, followed by analysis in denaturing PAGE gels (34).

RNase I digestion of in vitro RdRp assay products.

In vitro RdRp assays were conducted as described above. The ³²P-labeled RNA products were purified by phenol-chloroform extraction and precipitated. Following a 70% ethanol wash and air drying, they were treated with RNase I as described previously (34).

RNase H digestion of in vitro RdRp assay products.

In vitro RdRp assays were conducted as described above. The ³²P-labeled RNA products were purified by phenol-chloroform extraction and precipitated as described previously (34). Following a wash of the pellet with 70% ethanol, we air dried the samples and dissolved them in 10 μl of sterile distilled water (dH₂O). Ten microliters of 2× STE buffer (20 mM Tris, pH 8.0, 4 mM EDTA, pH 8.0, and 100 mM NaCl) and 100 pmol oligonucleotide were added to each reaction mix. The samples were heated to 94°C in a PCR machine and gradually cooled to room temperature over 15 min. RNase H digestion was carried out in a 100-μl final volume in the presence of 20 mM Tris, pH 8.0, 50 mM NaCl, and 10 mM MgCl₂ with 1 U of RNase H at 30°C for 15 min. Following digestion, the products were purified by phenol-chloroform extraction and precipitated, followed by analysis in denaturing PAGE gels (34).

RESULTS

N-terminally truncated p92^pol replication protein is an active RdRp in the absence of p33 replication protein in a cell-free yeast extract.

To test the functional roles of the N-terminal region of the tombusviral p92^pol replication protein, which is identical to the p33 replication protein due to the readthrough expression strategy of the tombusvirus genomic RNA (73), we made N-terminal truncations in p92^pol. Unlike full-length p92^pol, which is inactive in the absence of the p33 replication cofactor in CFE (Fig. 1A, lanes 11 and 12) (49, 50), we found that the recombinant Δ167 and Δ150 mutants of CNV p92^pol and the Δ167 mutant of TBSV p92^pol (Fig. 1A, lanes 2 to 4 and 13) were active RdRps in the absence of p33. However, the RdRp products generated by the recombinant Δ167 and Δ150 mutants of CNV p92^pol and the Δ167 mutant of TBSV p92^pol moved slower in the denaturing PAGE gel than the de novo-generated full-length replication product generated by recombinant full-length CNV p92^pol in the presence of the p33 replication cofactor (termed p92^pol/p33 replicase) in yeast CFE (Fig. 1A, lanes 9 and 10).

Fig 1.

RNA synthesis is supported by an N-terminally truncated p92^pol replication protein in cell-free yeast extracts. (A) Denaturing PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with recombinant N-terminally truncated p92^pol replication proteins (lanes 2 to 4 and 13) or full-length p92^pol (lanes 1 and 7 to 12) in the absence or presence of the p33 replication protein when the 621-nucleotide (nt) DI-72(+) repRNA was also added to the cell extract. The cell extract (CFE) was obtained from yeast cells coexpressing CNV p33 and full-length or truncated CNV p92^pol replication proteins, as shown (lanes 1 to 10). Lanes 11 to 14 represent in vitro assays obtained with CFE prepared from yeast cells lacking viral proteins. Purified recombinant viral proteins and viral (+)repRNA were added to the assay mixture. The standard de novo-initiated repRNA replication products are marked with solid arrowheads, while the 3′TEX-like products are shown with open arrowheads. The amounts of 3′TEX and de novo products were estimated separately using ImageQuant software. Note that the CNV replication proteins are labeled C33 and C92 (expressed in yeast), while those of TBSV are named T33 and T92 (purified from E. coli). The standard (lane 14) is ³²P-labeled DI-72 (+)RNA made by in vitro T7 transcription. Each experiment was repeated two or three times. (B) Micrococcal nuclease sensitivity of ³²P-labeled RNA products made using the purified recombinant Δ167 or T92-plus-T33 proteins from E. coli in an in vitro assay based on CFE prepared from yeast cells lacking viral replication proteins. Note that the micrococcal nuclease treatment, performed at the 1-h time point for 15 min, reduced the amount of de novo-generated RNA products with T92 plus T33 (lanes 7 and 8), since a large portion of the newly made DI-72 (+)RNAs was released from the membrane-bound replicase complex into the buffer during the in vitro reaction, which made the free (+)RNAs nuclease sensitive. Note that micrococcal nuclease treatment removes RNAs which are not protected by the membrane-bound replicase. See additional details in the description for panel A.

To study the nature of the RNA product generated by the recombinant TBSV Δ167 RdRp and to test if the template RNA is protected in a membranous compartment, we used micrococcal nuclease, which preferentially cleaves single-stranded RNAs (ssRNAs) not protected by a membrane compartment (49, 50). Interestingly, the RNA product of TBSV Δ167 RdRp was micrococcal nuclease sensitive (Fig. 1B, lanes 3 and 4), in contrast with the RNA product of the TBSV p92^pol/p33 replicase (lanes 7 and 8), which was partially (∼20%) micrococcal nuclease resistant due to the presence of the replicating viral (+)RNA and (−)RNA in membranous compartments in the yeast CFE (49, 50). The reduced signal for the RNA product of TBSV p92^pol/p33 replicase based on micrococcal nuclease treatment is due to the release of the majority of the newly made (+)RNA progeny into the buffer from the membranous compartments, which makes the released fraction of (+)RNA micrococcal nuclease sensitive (49, 50). Based on the sizes and micrococcal nuclease-sensitive nature of the RNA products of TBSV Δ167 RdRp, they likely represent 3′TEX products (due to self-priming by the 3′ end of the template RNA [see below]), which are not protected by the membranous compartments.

To further test the features of the TBSV Δ167 RdRp, we analyzed its activity under different conditions. For example, TBSV Δ167 RdRp was almost as active in the presence of only the soluble fraction of CFE as in whole CFE (Fig. 2A, compare lanes 15 and 16 with lanes 7 and 8), suggesting that TBSV Δ167 RdRp does not require membranes for its activity. This is in contrast with the TBSV p92^pol/p33 replicase (Fig. 2A, lanes 9 and 10 versus lanes 1 and 2), which requires cellular membranes for replicase activity (49, 50). Interestingly, the activity of TBSV Δ167 RdRp was greatly enhanced by soluble factors present in the CFE (Fig. 2B, lanes 3 and 4), with only minimal activity in the RdRp buffer (Fig. 2B, lanes 1 and 2). However, the identities of the critical host factors in the soluble fraction of CFE are currently unknown.

Fig 2.

Roles of host proteins and the p33 replication cofactor in in vitro RNA synthesis supported by the N-terminally truncated p92^pol replication protein. (A) PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with recombinant TBSV proteins. The purified recombinant T33 replication cofactor (0.12 μg [lanes 3 to 4, 11, and 12] or 0.5 μg [lanes 5, 6, 13, and 14]) was added to an in vitro assay mixture containing either Δ167(T92) (0.25 μg) or T92 (0.25 μg) RdRp protein and yeast CFE (lanes 1 to 8) or only the soluble fraction of CFE (lanes 9 to 16). Note that DI-72 represents the de novo-generated product. See further details in the legend to Fig. 1. Each experiment was repeated two or three times. (B) Denaturing PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with recombinant Δ167 RdRp protein in the presence of buffer only (lanes 1 and 2) or the soluble fraction of CFE (lanes 3 and 4).

Another interesting observation from these experiments is that the TBSV Δ167 RdRp was able to produce some full-length viral RNAs in the presence of the p33 replication cofactor in CFE (Fig. 2A, lanes 3 to 6). However, this activity required the cellular membrane fraction, since the TBSV Δ167 RdRp could not produce full-length viral RNAs in the presence of the p33 replication cofactor and the soluble fraction of CFE (Fig. 2A, lanes 11 to 14). Furthermore, the p33 cofactor reduced the putative 3′TEX activity of TBSV Δ167 RdRp in the soluble fraction of CFE (Fig. 2A, compare lanes 13 and 14 to lanes 15 and 16). Altogether, these data suggest that the p33 replication cofactor and cellular membranes are both critical for production of de novo full-length viral RNAs by the TBSV Δ167 RdRp in vitro.

The N-terminal domain is an inhibitor of the RdRp function of the p92^pol replication protein in cell-free yeast extract.

Since the TBSV Δ167 protein is a functional RdRp in CFE, while full-length TBSV p92^pol alone is inactive, we wanted to define which sequences within the N-terminal sequence of TBSV p92^pol are involved in the inhibitory effect on RdRp activity. We generated a set of deletion mutants of TBSV p92^pol (Fig. 3A) and determined their RdRp activities in CFE with or without the p33 replication cofactor. Deletions of the two transmembrane sequences (construct Δ83-153), the very N-terminal region (construct Δ80), or one transmembrane region in combination with the very N-terminal region (construct Δ98) did not result in functional RdRps in the absence of p33 in CFE (Fig. 3B, lanes 1 to 3). Since constructs with deletions of both transmembrane sequences in combination with the very N-terminal region (Δ150 [Fig. 1A, lanes 3 and 4] and Δ167 [Fig. 1A, lane 13, and Fig. 3B, lane 4]) showed RdRp activity, we concluded that the transmembrane and very N-terminal sequences play inhibitory roles, likely preventing TBSV p92^pol from being active in the cytosol of infected cells prior to its translocation to membranous compartments (see Discussion). Interestingly, deletion of one of the RNA binding sequences (Δ167/ΔRPR) (Fig. 3B, lane 5) only reduced the RdRp activity, possibly due to the presence of additional RNA binding regions in TBSV p92^pol (Fig. 3A) (54). Deletions of either one of the p33-p33/p92 interaction subdomains (Δ167/ΔS1 and Δ167/ΔS2) (Fig. 3B, lanes 1 to 4) rendered the RdRp inactive, suggesting that interaction between TBSV p92^pol molecules is likely critical for RdRp activity in vitro.

Fig 3.

Inhibitory role of the N-terminal sequence of the p92^pol replication protein in RdRp activity in vitro. (A) Schematic representation of the known functions of T33 and T92 replication proteins and truncated T92 proteins. The N-terminal segment in TBSV p92^pol contains the same sequence as p33 due to the overlapping expression strategy of the TBSV genome, while the C-terminal region of p92^pol carries the RdRp domain and two RNA binding sequences. The various domains in the shared sequences are indicated as follows: mPTS, peroxisomal membrane targeting sequences; ub, monoubiquitinated sequence; TMD, transmembrane domains; late domain, sequence recognized by the ESCRT factors; P, phosphorylation sites; and RPR, arginine-proline-rich RNA binding domain. S1 and S2 are subdomains of the p33-p33/p92 interaction domain. (B and C) Denaturing PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with truncated recombinant T92 proteins. The purified T92 derivatives (0.25 μg) were added to the yeast CFE in the absence (lanes 1 to 5) or presence (lanes 6 to 10) of purified T33 replication cofactor (0.5 μg). See further details in the legend to Fig. 1. Each experiment was repeated two or three times. (D) Inhibition of Δ167 RdRp activity by the S1/S2 peptide. Denaturing PAGE analysis was performed on the ³²P-labeled RNA products obtained in an in vitro assay with truncated recombinant Δ167 RdRp. The purified Δ167 RdRp (0.25 μg) was added to yeast CFE containing 0.05, 0.17, or 0.5 μg/μl (final concentration) of either the S1/S2 (lanes 1 to 3) or RPR (lanes 4 to 6) peptide. (E) Lack of inhibition of in vitro activity of FHV replicase by TBSV peptides. Denaturing PAGE analysis of ³²P-labeled RNA products shows the activity of the template-dependent affinity-purified FHV replicase, which was tested in the presence of 0.17 or 0.5 μg/μl (final concentration) of S1/S2 (lanes 2 and 3) or 0.5 μg/μl RPR (lane 4) peptide. The replicase assay was programmed with the TBSV DI-72 repRNA.

To further test the requirement for interaction between Δ167 RdRp molecules to form functional RdRps, we measured the inhibitory effects of peptides on the in vitro activity of TBSV Δ167 RdRp. Interestingly, the S1/S2 peptide, representing the p33-p33/p92 interaction domain, completely inhibited the in vitro activity of Δ167 RdRp (Fig. 3D, lanes 1 to 3), while the RPR peptide, representing the RNA binding domain in p33/p92, did not inhibit this activity (lanes 4 to 6). The activity of the unrelated flock house virus replicase was not inhibited by these peptides (Fig. 3E), confirming that the inhibitory effect of the S1/S2 peptide on the TBSV Δ167 RdRp was specific. Based on these and the above data with the Δ167/ΔS1 and Δ167/ΔS2 mutants, we propose that interactions between Δ167 RdRp molecules are required for RdRp activity. Thus, the Δ167 RdRp likely functions as a homodimer or a multimer in vitro.

The RdRp activities of the above constructs in the presence of the p33 replication cofactor were different. All constructs with deletions of the very N-terminal domain, the two transmembrane domains, or the RPR sequence were capable of producing full-length RNA progeny in the presence of p33 in CFE (Fig. 3B, lanes 6 to 10), albeit with various efficiencies. Thus, the p33 replication cofactor could partially complement the N-terminal deletions in TBSV p92^pol in CFE. However, deletions of either one of the p33-p33/p92 interaction subdomains (Δ167/ΔS1 and Δ167/ΔS2) (Fig. 3B, lanes 6 to 9) made the RdRp inactive even in the presence of p33, suggesting that p33 could not complement this defect in vitro. This finding indicates that interaction between TBSV p92^pol molecules or between p92^pol and p33 molecules is likely critical for de novo-initiated RdRp activity in vitro.

The N-terminally truncated Δ167 RdRp uses only (+)RNA templates carrying the RII(+)-SL cis-acting element in vitro.

To examine the template specificity of the TBSV Δ167 RdRp protein, we first tested various (+)RNA templates in the CFE-based assay. DI-72 (+)RNA, the longer DI-73 (+)RNA (Fig. 4A), and the short DI-mini construct, carrying the previously defined critical cis-acting elements required for template recognition [RII(+)-SL] and replicase assembly (SL1-SL2-SL3) (44, 46), were all used efficiently by the TBSV Δ167 RdRp in the CFE-based assay (Fig. 4B, lanes 5, 7, and 9). The RNA products generated with these constructs were all RNase I sensitive (resulting in short dsRNA products after RNase I digestion), suggesting that they represent 3′TEX-like products. Interestingly, the DI-72/ΔRII construct, which was missing the template recognition element, was poorly used by the TBSV Δ167 RdRp in the CFE-based assay (Fig. 4B, lane 1).

Fig 4.

The recombinant Δ167 RdRp protein shows high template specificity and depends on a cis-acting RNA element. (A) Schematic representation of DI-73(+) and DI-72(+) repRNAs and a derivative RNA. The three known cis-acting RNA elements affecting the assembly and activation of the full-length T92 RdRp protein, i.e., RII(+)-SL, UL-DL, and RIV (SL1-SL2-SL3), are encircled. The critical C-C mismatch in RII(+)-SL required for template recognition by the TBSV replication proteins is highlighted. The (−)RNA initiation promoter (termed gPR) is shown with a dark arrowhead. (B) Denaturing PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with Δ167 RdRp protein in yeast CFE in the absence of the T33 replication cofactor. The samples were either treated with ssRNA-specific RNase I or not treated, as shown. The RNase I digestions were conducted to confirm whether the products were produced de novo or by 3′TEX. The original (untreated with RNase I) 3′TEX-like RNA products are marked with white arrowheads, while partially digested 3′TEX-like products are shown with gray arrowheads. Note that the original 3′TEX-like RNA products can run aberrantly under these conditions due to their heat-resistant secondary structures (highly stable hairpin structures [see Fig. 6]). See further details in the legend to Fig. 1. Each experiment was repeated two or three times. (C and D) Denaturing PAGE analysis of ³²P-labeled RNA products obtained with various RNAs in an in vitro assay with Δ167 RdRp protein in yeast CFE. RII* indicates a single C-to-G point mutation within the C-C mismatch in RII(+)-SL, SL1* represents a point mutation within the −-strand RNA initiation promoter (termed gPR), and SL3* shows a point mutation within the RSE (replication silencer element) in the SL3 stem-loop structure. The mutations within the gPR and RSE interrupted the essential base pairing between these elements (48). Note that the de novo product and the 3′TEX-like product of the DI-mini-derived RNAs ran similarly under these conditions. DI-634 represents (+)RNA derived from FHV. Note that lanes 9 and 10 and lanes 11 and 12 in panel D represent samples from in vitro assays without and with the ATP-regenerating components, respectively.

The TBSV Δ167 RdRp did not use the minus-strand DI-72 RNA or the heterologous FHV-based DI-634 (+)RNA in the CFE-based assay (Fig. 4D, lanes 1 and 2 and lanes 5 and 6). This is in contrast to the case with the purified tombusvirus replicase from yeast or the partially purified CNV replicase from plants, both of which could efficiently utilize these templates in vitro (23, 34, 45).

Additional experiments based on DI-mini and an RII(+)-SL mutant (RII*; carries a G-to-C mutation in the stem-loop) demonstrated that the RII(+)-SL template recognition element is indeed critical for the template activity of the TBSV Δ167 RdRp in the CFE-based assay (Fig. 4C, lanes 1 and 9). On the other hand, the second critical cis-acting element, i.e., the replicase assembly element (SL1-SL2-SL3), was not needed for template activity by the TBSV Δ167 RdRp in the CFE-based assay (Fig. 4C, lanes 3 and 5), while these sequences were needed for the activity of the TBSV p92^pol/p33 replicase (Fig. 4C, lanes 4, 6, and 8), as shown previously (44, 46). Altogether, the data suggest that the TBSV Δ167 RdRp shows template specificity, which is, however, less complex [requiring only the RII(+)-SL template recognition element] than that observed with the TBSV p92^pol/p33 replicase in vitro.

The observation that the N-terminally truncated TBSV Δ167 RdRp can be activated in the CFE-based assay, while full-length p92^pol cannot be activated, could be interpreted as showing that the very N-terminal sequence serves as a negative regulator of the TBSV RdRp activity. The inhibitory activity of the N-terminal sequence might be due to “folding back” of the N terminus to the “body” of the RdRp, thus inhibiting the key RNA binding function of the TBSV RdRp protein. Therefore, we tested if TBSV Δ167 RdRp could bind to the RII(+)-SL cis-acting sequence required for RdRp activation. Indeed, electrophoretic gel mobility shift assays (EMSAs) revealed that TBSV Δ167 RdRp could bind to the RII(+)-SL cis-acting sequence, while full-length p92^pol could not (Fig. 5). Thus, these data support the model that the very N-terminal sequence of full-length p92^pol functions as an inhibitor of RNA binding (see Discussion).

Fig 5.

TBSV Δ167 RdRp binds to the RII(+)-SL cis-acting RNA sequence. (A) Schematic representation of wt TBSV RNA-derived RII(+)-SL and the mutated RII(+)-SL* element used in the binding assay. Note that RII(+)-SL*, used as a negative control here, does not bind to the TBSV p33 replication protein in vitro. (B) In vitro binding assay with purified TBSV Δ167 RdRp and full-length (FL) p92^pol, using ³²P-labeled wt RII(+)-SL and mutated RII(+)-SL* as templates. The assay mixtures contained the RNA (∼0.1 pmol) and increasing amounts (0.4 μg [lanes 1, 3, 6, and 8] and 0.8 μg [lanes 2, 4, 7, and 9]) of purified recombinant proteins. The free and RdRp-bound ssRNAs were separated in nondenaturing 5% acrylamide gels. Each experiment was repeated at least two times.

The N-terminally truncated TBSV Δ167 RdRp produces a 3′-terminally extended and prematurely terminated RNA in vitro.

To determine the RNA product produced by the TBSV Δ167 RdRp, we first used DI-72 (+)RNA with increasingly long 5′ truncations, along with the TBSV Δ167 protein, in the CFE-based assay (Fig. 6A and B). We observed that the larger the deletion in the template RNA was, the faster the undigested RNA product moved in the denaturing PAGE gel (Fig. 6B, white arrowheads). However, RNase I digestion resulted in a faster-migrating but same-sized RdRp product, regardless of the length of the 5′ truncation (Fig. 6B, lanes 3, 7, 11, and 15, gray arrowheads). In contrast, the sizes of the de novo-generated RdRp products obtained with the same set of templates with the TBSV p92^pol/p33 replicase in the CFE-based assay did not change after RNase I digestion (Fig. 6B, lanes 1, 5, 9, and 13, black arrowheads). All of these data are consistent with the prediction that the TBSV Δ167 RdRp produces a 3′TEX product that is prematurely terminated at the RII(+)-SL cis-acting element (based on size estimation of the RNase I-digested product and shown schematically in Fig. 6C).

Fig 6.

Characterization of the RNA product made by the recombinant Δ167 RdRp protein in vitro. (A) Schematic representation of DI-72(+) derivative RNAs used as templates. These RNAs lack either the entire 169-nt region I (RI) or a portion of RI of the DI-72(+) repRNA, as shown. The presence of the important cis-acting RNA elements is shown in Fig. 4A. (B) Denaturing PAGE analysis of ³²P-labeled RNA products obtained in an in vitro assay with Δ167 RdRp protein or T92 plus T33 in yeast CFE. The samples were either treated with ssRNA-specific RNase I or not treated, as shown. The de novo-generated products are shown with black arrowheads, the original (untreated with RNase I) 3′TEX-like RNA products are marked with white arrowheads, and partially digested 3′TEX-like products are shown with gray arrowheads. Note that the original 3′TEX-like RNA products can run aberrantly under these conditions due to their heat-resistant secondary structures. See further details in the legend to Fig. 1. Each experiment was repeated two or three times. (C) Schematic representation of the 3′TEX product of the DI-72 (+)RNA obtained with Δ167 RdRp, based on size estimation of RNase I and RNase H products. The solid line represents the template-dependent newly synthesized (and thus ³²P-labeled) portion of the 3′TEX product. The site of predicted premature termination relative to the RII(+)-SL element is indicated. (D) Denaturing PAGE analysis of RNase H-treated ³²P-labeled RNA products obtained in an in vitro assay with Δ167 RdRp protein or T92 plus T33 in yeast CFE. The oligonucleotides used are shown schematically in panel E. The oligonucleotides were annealed to the ³²P-labeled RNA products prior to the RNase H treatment. Note that oligonucleotide 1164 hybridizes to minus-strand sequences, while the other oligonucleotides form base pairs with (+)RNA sequences. The RNase H cleavage products in the Δ167 RdRp assay are marked with gray arrowheads. Note that oligonucleotides 1160 and 313 did not promote efficient cleavage of the RdRp product in the RNase H assay, likely due to the inefficient hybridization of these oligonucleotides to the structured portion of the 3′TEX product (see panel E). (E) Schematic representation of 3′TEX and de novo RNA products and the expected products after RNase H treatment. The locations of the annealed oligonucleotides and the predicted RNase H cleavage sites are indicated. Gray arrowheads indicate expected cleavage sites, while white arrowheads indicate a lack of cleavage. Note that only the newly synthesized portion of the 3′ TEX product (the area indicated by a solid black line) becomes ³²P labeled, while both (+)RNA and (−)RNA products of de novo initiation in the CFE-based assay become ³²P labeled.

This model was further supported by RNase H digestion of the RdRp products of the DI-72 template RNA with a set of DNA oligonucleotides annealing to different portions of the DI-72 RNA (Fig. 6E). As expected, the 3′TEX products made by TBSV Δ167 RdRp (Fig. 6D) showed a different RNase H digestion pattern from that of the de novo RdRp products generated by the TBSV p92^pol/p33 replicase. Altogether, the results from RNase H digestion confirmed that the TBSV Δ167 RdRp product was a 3′TEX product that was prematurely terminated at the RII(+)-SL cis-acting element.

To further define the template requirement and the need for 3′TEX initiation of RNA synthesis for the TBSV Δ167 RdRp, we generated an artificial RNA template carrying the RII(+)-SL cis-acting element and an artificial stem-loop sequence to facilitate 3′TEX (Fig. 7A, mini-art-stem construct) (7, 9). As expected, mini-art-stem RNA supported primer extension with the TBSV Δ167 RdRp in the CFE-based assay (Fig. 7B, lanes 11, 12, 23, and 24), while it was not used as a template by the TBSV p92^pol/p33 replicase in the CFE-based assay (Fig. 7B, lanes 5, 6, 17, and 18). Overall, the fact that mini-art-stem RNA was able to activate the TBSV Δ167 RdRp demonstrates that RII(+)-SL is the only cis-acting element required for activation, while the SL3 and SL1 hairpins are not necessary for this function.

Fig 7.

Minimal template sequence requirement for recombinant Δ167 RdRp protein in vitro. (A) Schematic representation of minimal RNAs used as templates. These RNAs contained the RII(+)-SL cis-acting element and an artificial stem (art-stem), as shown. The 3′-end sequence of the art-stem either forms a perfect stem-loop or contains two or three non-base-paired A's, as shown. (B) Denaturing PAGE analysis of ³²P-labeled RNA products obtained with the minimal mini-art-stem RNA in an in vitro assay with Δ167 RdRp protein or with T92-plus-T33 replicase in yeast CFE (lanes 1 to 12) or in the soluble fraction of CFE (lanes 13 to 24). The 3′TEX RNA products (which are the result of premature termination of 3′TEX) of the expected size are marked with white arrowheads, while the putative full-length 3′TEX products are indicated with an arrow. See further details in the legend to Fig. 1. Each experiment was repeated two or three times. (C) Denaturing PAGE analysis of ³²P-labeled RNA products obtained with the templates shown in panel A in an in vitro assay with Δ167 RdRp protein in yeast CFE.

Another interesting observation is that mini-art-stem RNA did not efficiently support the fully complementary 3′TEX product but generated mostly the prematurely terminated 3′TEX product (Fig. 7B). We also noted that the p33 replication cofactor did not significantly alter the production of the 3′TEX RNAs obtained with the TBSV Δ167 RdRp (Fig. 7B, lanes 7 to 10).

To demonstrate that initiation by the TBSV Δ167 RdRp is due to 3′TEX, we added extra AA or AAA nucleotides to the 3′ end of mini-art-stem (these extra nucleotides will not be able to form base pairs, as shown in Fig. 7A), which is known to greatly inhibit 3′TEX (7, 9). Indeed, mini-art-stem with extra AA or AAA nucleotides at the 3′ end produced much smaller amounts of RNA products with the TBSV Δ167 RdRp (Fig. 7C, compare lanes 2 and 3 with lane 1), which is consistent with 3′TEX. Altogether, the presented data support the hypothesis that the TBSV Δ167 RdRp produces new RNA products via 3′TEX followed by premature termination at the base of the RII(+)-SL sequence.

The RII(+)-SL cis-acting sequence serves as a premature stop site for the TBSV Δ167 RdRp in vitro.

To study if the extended secondary structure of RII(+)-SL serves as a premature stop site for the TBSV Δ167 RdRp in the CFE-based assay, we constructed RNA templates with duplicated functional and/or nonfunctional (due to the presence of a debilitating G-to-C mutation replacing the critical C-C mismatch) (Fig. 8A) RII(+)-SL cis-acting elements. As expected, DI-mini templates carrying one or two copies of the nonfunctional G-C mutant RII(+)-SL cis-acting element supported inefficient 3′TEX by the TBSV Δ167 RdRp in the CFE-based assay (Fig. 8B, lanes 1 to 4 and 9 to 12).

Fig 8.

Effect of duplicated cis-acting sequences on termination of RNA synthesis by recombinant Δ167 RdRp protein in vitro. (A) Schematic representation of the minimal RNA and its derivatives used as templates. These RNAs contained one or two copies of the RII(+)-SL cis-acting element, with either a wt or mutated (the inactive G-to-C mutation within the C-C mismatch) sequence, as shown. (B) Denaturing PAGE analysis of ³²P-labeled RNA products obtained with the minimal RNA and its derivatives in an in vitro assay with Δ167 RdRp protein. The samples were treated with RNase I or left untreated, as shown. Note that the untreated 3′TEX products migrated aberrantly under these conditions due to their heat-resistant structure, as shown schematically on the right.

On the other hand, DI-mini templates carrying one or two copies of a functional RII(+)-SL cis-acting element with the critical C-C mismatch generated 3′TEX products efficiently (Fig. 8B, lanes 5 to 8 and 13 to 25). Interestingly, the sizes of the 3′TEX products were largely determined by the location of the RII(+)-SL element carrying the critical C-C mismatch mutation (representing the wt RNA sequence). For example, the DI-mini/RII-C/G+RII construct (Fig. 8A), which carried a duplicated RII(+)-SL element with a 3′-proximal wt RII(+)-SL and a 5′-proximal mutated RII(+)-SL, produced 3′TEX products showing a similar RNase I digestion pattern to that of DI-mini with a single functional RII(+)-SL element (Fig. 8B, compare lanes 18 and 20 with lanes 6 and 8). This pattern of RdRp products suggests that the premature termination occurred almost exclusively at the base of the functional 3′-proximal (internal) wt RII(+)-SL element, not at the base of the nonfunctional 5′-proximal mutated RII(+)-SL element. On the other hand, the DI-mini/RII+RII-C/G construct (Fig. 8A), which carried a duplicated RII(+)-SL element with a 3′-proximal mutated RII(+)-SL and a 5′-proximal wt RII(+)-SL, produced only a small amount of RNase I product, with a similar migration pattern to that of DI-mini (Fig. 8B, compare lanes 14 and 16 with lanes 6 and 8). This suggests that only a fraction of the TBSV Δ167 RdRp terminates 3′TEX at the internal RII(+)-SL element in the CFE-based assay. Thus, the nonfunctional RII(+)-SL* sequence can also terminate RNA synthesis, but far less effectively than the wt RII(+)-SL element. Accordingly, a large fraction of the TBSV Δ167 RdRp terminates 3′TEX at the 5′-proximal RII(+)-SL element with construct DI-mini/RII+RII-C/G.

The picture was a bit more complex with the DI-mini/2xRII construct, carrying duplicated wt RII(+)-SL sequences. In this case, the 3′TEX products obtained in the TBSV Δ167 RdRp assay showed a similar RNase I digestion pattern (Fig. 8B, lanes 22 and 24) to that of DI-mini/RII+RII-C/G (lanes 14 and 16), with a high termination rate at the 5′-proximal RII(+)-SL and a lower termination rate at the 3′-proximal RII(+)-SL. We do not know why termination favored the 5′-proximal wt RII(+)-SL sequence.

Altogether, these data suggest that the functional RII(+)-SL element is a better RdRp termination site than the mutated RII(+)-SL element, although both the wt and mutated RII(+)-SL elements did serve as RdRp termination sites, with various efficiencies. Altogether, these data are the first direct evidence of the ability of the TBSV RdRp (albeit its truncated version) to terminate at a cis-acting sequence in vitro, which could be important during subgenomic RNA synthesis, DI RNA formation, and RNA recombination (see Discussion). Indeed, the RII cis-acting sequence in the genomic TBSV RNA is a recombination hot spot and is involved in DI RNA formation (71, 72).

DISCUSSION

Activation of the truncated tombusvirus RdRp is less complex than that of the full-length RdRp in cell extracts.

(+)RNA virus replicases are multisubunit enzymes that are membrane associated in eukaryotic cells (4, 32). The catalytic subunit, called RdRp, is virally encoded and is translated in the cytosol as a single protein or part of a polyprotein, and then the RdRp is translocated to the site of replication, which usually consists of deformed membranes and/or vesicular structures (13, 32). The freshly translated RdRp is likely nonfunctional in the cytosol to prevent the generation of viral dsRNA molecules, which would easily be recognized by the host RNA surveillance system. Thus, the activation of the viral RdRp possibly takes place after its membrane association and during the full assembly of the functional replicase.

The assembly and activation of the tombusvirus replicase are rather complex processes which require the TBSV p92^pol RdRp, the p33 replication cofactor, cis-acting elements in the viral (+)RNA, subcellular membranes, Hsp70, ESCRT, eEF1A, and an additional, undefined number of host proteins (2, 3, 26, 28, 32, 50). Surprisingly, we found that N-terminally truncated CNV or TBSV p92^pol RdRp requires fewer components than the complete p92^pol/p33 replicase for activation of its polymerization function. For example, the TBSV Δ167 RdRp does not require the p33 replication cofactor or the cellular membranes to produce RNA products on the added viral RNA template in the CFE assay (Fig. 1 and 2). Full-length TBSV p92^pol is not an active RdRp under these conditions (50). However, the activation of TBSV Δ167 RdRp still needs a soluble host factor(s) and the RII(+)-SL cis-acting RNA element. Altogether, the E. coli-expressed and purified TBSV Δ167 RdRp showed little RdRp activity in the absence of the soluble fraction of yeast CFE (Fig. 2B), suggesting that the host factor(s) is critical for activation of the truncated TBSV RdRp. However, unlike the complete tombusvirus replicase, containing p92^pol and p33 replication proteins, the TBSV Δ167 RdRp did not need the 3′-terminal SL1-SL2-SL3 cis-acting RNA sequences for its activation (Fig. 4C) (46). The 3′-terminal SL3 sequence has been shown to bind to cellular eEF1A, while the SL1 sequence interacts with host eEF1Bγ (26, 28, 58). These RNA-host protein interactions are proposed to be important to change the 3′-terminal structure in the (+)RNA to render it more suitable for initiation of minus-strand synthesis. Since these 3′ cis-acting RNA sequences are not needed for the activation of the truncated TBSV RdRp, it is possible that eEF1A and host eEF1Bγ might not be involved in the RdRp activation process.

Unique features of the TBSV Δ167 RdRp in RNA synthesis.

In addition to its high template selectivity, the TBSV Δ167 RdRp shows other unique features (Fig. 9). For example, the active TBSV Δ167 RdRp shows template selectivity by using RNA templates that contain the RII(+)-SL cis-acting element and discriminates against mutated TBSV RNAs, the TBSV (−)RNA, or the heterologous FHV RNA (Fig. 4). Moreover, the TBSV Δ167 RdRp, in the absence of p33 and subcellular membranes, can initiate RNA synthesis only via 3′TEX, while the de novo-initiated full-length (−)RNA or (+)RNA products were not detected in vitro.

Fig 9.

Summary of the activities and requirements of the various replicase and RdRp preparations of tombusviruses. (A) Complete replicase; (B) Δ167 RdRp; (C) Δ167 RdRp plus p33. We propose that the tombusvirus RdRp becomes activated and competent for de novo initiation in the presence of the p33 replication cofactor, subcellular membranes, soluble host factors, and the viral (+)RNA. In contrast, the activation of Δ167 RdRp is less complex, requiring only soluble host factors and the RII(+)-SL recruitment element in the viral (+)RNA.

Another feature is that the TBSV Δ167 RdRp requires the p33-p33/p92 interaction domain for its function (Fig. 3C), suggesting that TBSV Δ167 RdRp must form a homodimer or multimer in order to become an active RdRp. This is further supported by the strong inhibitory effect of the S1/S2 peptide (carrying the interaction sequence), which should prevent the efficient formation of Δ167 RdRp homodimers, on the activity of the Δ167 RdRp. Since the p33-p33/p92 interaction domain has been shown to also be important for selective RNA recognition (51), it is possible that homodimeric TBSV Δ167 RdRp, but not monomeric Δ167 RdRp, forms a structure competent for RNA binding and RNA synthesis.

The complete tombusvirus replicase containing p92^pol and p33 replication proteins is a processive polymerase that can synthesize long cRNA products, even on fully double-stranded RNA templates (43, 64). Thus, it is interesting that the TBSV Δ167 RdRp terminates RNA synthesis at the internal RII(+)-SL cis-acting element at a high frequency (Fig. 8). Using duplicated RII(+)-SL sequences in the templates, we obtained data supporting the hypothesis that the functional RII(+)-SL sequence with the critical C-C mismatch, required for selective recognition by the tombusvirus replication proteins (51), is a more efficient termination signal for the TBSV Δ167 RdRp than a mutated, nonfunctional RII(+)-SL sequence carrying a G-to-C mutation (instead of the internal C-C mismatch) (Fig. 8). The structure of the mutated RII(+)-SL hairpin is predicted to be more stable than that of the corresponding wt hairpin, yet the wt hairpin is the stronger termination site. Therefore, we propose not only that the strength and stability of the hairpin are important but also that the ability of the RII(+)-SL hairpin to bind to the TBSV replication proteins is possibly important as well (29, 51). It is possible that extra, wt RII(+)-SL hairpin-bound Δ167 RdRp homodimeric proteins could promote more efficient pausing for the advancing TBSV Δ167 RdRp on the template RNA.

Identification of termination sites in the viral (+)RNA for the TBSV Δ167 RdRp could open up the possibility to further study transcriptional pausing and termination by tombusvirus RdRps in vitro. These studies could be important for understanding subgenomic RNA synthesis, viral RNA recombination, and DI RNA formation, all of which are predicted to require RdRp pausing or termination at various sites within the viral genome (7–9, 18, 30, 73, 74).

The N-terminal sequence in the tombusvirus RdRp plays a regulatory role in replicase activation.

Using various N-terminally truncated TBSV p92^pol constructs, we showed that the presence of either the N-terminal domain (predicted to face the cytosolic compartment) or the transmembrane domains of TBSV p92^pol inhibits the RdRp activity in the absence of the p33 replication cofactor. Based on these new data, we suggest that the N-terminal sequence of p92^pol RdRp plays an important regulatory role in the replicase assembly/activation process. We propose that the N-terminal putative cytosolic sequence serves as a negative regulator, possibly by “folding back” to the “body” of the RdRp, thus inhibiting the key RNA binding function of the full-length TBSV p92 RdRp protein (Fig. 5).

On the other hand, the transmembrane domains are likely involved in membrane association of the p92^pol RdRp prior to the activation of the RdRp function. Interestingly, these proposed functions of p92^pol could be complemented in the CFE assay by the p33 replication cofactor, which carries the same sequences (Fig. 2A). Also, the replication incompatibility of the N-terminally truncated p92^pol protein can be partially complemented by wt p33 in yeast (40), possibly due to “piggybacking” of the truncated p92^pol protein with wt p33 to the site of replicase assembly. In spite of the new insights into p92^pol activities obtained by use of the above RdRp mutants, we still do not understand why full-length p92^pol is an inactive RdRp when the p33 replication cofactor and cellular membranes are absent.

The p33 replication cofactor facilitates de novo RNA synthesis and (+)RNA production and increases the processivity of the tombusvirus Δ167 RdRp in vitro.

One of the interesting observations in this study is the critical role of the p33 replication cofactor in making the TBSV Δ167 RdRp competent for de novo RNA synthesis and plus-strand synthesis, but only in the presence of cellular membranes, in the CFE assay (Fig. 2). Moreover, the presence of the p33 replication cofactor makes the TBSV Δ167 RdRp processive, thus producing full-length RNA products instead of the premature RdRp products characteristic for the TBSV Δ167 RdRp in the CFE assay (Fig. 2). Altogether, in the presence of subcellular membranes, the p33 replication cofactor makes the truncated tombusvirus RdRp almost as good as full-length p92^pol in the CFE assay and in yeast (40). Determining the actual functions of the p33 replication cofactor and cellular membranes in the above processes will need further studies.

Altogether, many features of the TBSV Δ167 RdRp are different from those of other recombinant viral RdRps, such as the TCV p88 RdRp. This is because the TBSV Δ167 RdRp shows high template selectivity and cannot utilize (−)RNA or heterologous templates, while TCV p88 can (55). Also, the norovirus and HCV RdRps show less template specificity and can utilize cellular RNAs as a template (20, 65). The BMV 2a RdRp, expressed in human cells in the absence of the BMV 1a replication protein, also shows template selectivity toward its own 2a-encoding RNA, but this RdRp may initiate RNA synthesis de novo (66).

Since the TBSV Δ167 RdRp is an active polymerase and can be “rescued” to become a fully functional RdRp in the presence of the p33 replication cofactor and cellular membranes, the TBSV Δ167 RdRp will be useful in future studies to advance our mechanistic understanding of many steps in and the roles of some host factors during the TBSV replication process.