Brome Mosaic Virus Polymerase-Like Protein 2a Is Directed to the Endoplasmic Reticulum by Helicase-Like Viral Protein 1a

Abstract

Brome mosaic virus (BMV), a positive-strand RNA virus in the alphavirus-like superfamily, encodes RNA replication proteins 1a and 2a. 1a contains a C-terminal helicase-like domain and an N-terminal domain implicated in viral RNA capping, and 2a contains a central polymerase-like domain. 1a and 2a colocalize in an endoplasmic reticulum (ER)-associated replication complex that is the site of BMV-specific RNA-dependent RNA synthesis in plant and yeast cells. 1a also localizes to the ER in the absence of 2a or viral RNA replication templates. To investigate the determinants of 2a localization, we fused 2a to the green fluorescent protein (GFP), creating a functional GFP-2a fusion that supported BMV RNA replication and subgenomic mRNA transcription. In the absence of 1a, the GFP-2a fusion was found to be diffused throughout the cytoplasm and in punctate spots not associated with any cytoplasmic organelle so far tested. Formation of these spots was dependent on the C-terminal half of 2a and may represent aggregation of a fraction of 2a. When coexpressed with 1a, GFP-2a colocalized with 1a and ER-resident protein Kar2p in a partial or complete ring around the nucleus. Consistent with these results, cell fractionation showed that both the GFP-2a fusion and wild-type (wt) 2a remained soluble when expressed alone, while in cells coexpressing 1a, most of the GFP-2a fusion or wt 2a cofractionated with 1a in the rapidly sedimenting membrane fraction. Deletion analysis showed that the N-terminal 120-amino-acid segment of 2a, containing one of two 2a regions previously shown to interact with 1a, was necessary and sufficient for 1a-directed localization of GFP-2a derivatives to the ER. These results suggest that 1a, which also interacts independently with the ER and viral RNA, is a key organizer of RNA replication complex assembly.

RNA replication by positive-strand RNA viruses is closely associated with cellular membranes. For all well-studied eukaryotic positive-strand RNA viruses, the viral RNA-dependent RNA replication complex copurifies with membrane extracts from infected cells (8, 9, 14, 18, 43). In vivo and in vitro studies with positive-strand RNA viruses suggest that membrane association is essential for at least some steps of RNA replication (7, 38, 58). In some cases, negative-strand RNA synthesis activity can be solubilized from membranes (24, 43, 57, 58). However, in vivo, both positive- and negative-strand RNA synthesis occurs in membrane-associated complexes (10, 45, 46). The membrane interactions of replication factors from most viruses appear specific in that the replication complexes of different positive-strand RNA viruses associate with different intracellular membranes (18, 19, 41, 51, 52). However, the mechanisms by which such viral replication complexes are targeted to and assembled on specific membrane sites remain poorly understood.

Brome mosaic virus (BMV), the type member of the Bromovirus genus, is a positive-strand RNA virus in the alphavirus-like superfamily (1). The BMV genome is composed of three RNAs. RNA3 encodes the 3a protein, which is required for cell-to-cell movement of infection in plants (3, 37), and the coat protein, which is translated from a subgenomic mRNA (RNA4) and is required for encapsidation and long-range movement in plants (3, 49). RNA1 and RNA2 encode nonstructural proteins 1a and 2a, respectively, which are required for RNA replication (17, 27) and contain three domains conserved with other members of the alphavirus superfamily. The 109-kDa 1a protein contains an N-terminal domain with m⁷G methyltransferase and covalent GTP binding activities implicated in viral RNA capping (2, 32) and a C-terminal domain with similarity to DEAD box RNA helicases (21). The 94-kDa 2a protein has a central domain with similarities to RNA-dependent RNA polymerases (RdRp's) (4, 23). 1a and 2a interact in vitro and in vivo (31, 39), and genetic studies show that compatible 1a-2a interaction is essential for RNA replication in vivo (15, 54).

In addition to its natural plant hosts, BMV directs RNA replication, gene expression, and encapsidation in the yeast Saccharomyces cerevisiae (26, 28, 33). In infected plant cells and in yeast, 1a and 2a colocalize on endoplasmic reticulum (ER) membranes at the sites of viral RNA synthesis, which can be visualized by immunofluorescence of incorporated 5-bromouridine 5′-triphosphate (45, 46). Consistent with these results, membrane-associated RdRp extracts that selectively synthesize BMV negative-strand RNAs have been isolated from BMV-infected plant cells (22, 36, 43, 44) and from yeast expressing 1a and 2a proteins and replicating BMV RNA3 derivatives (42). After detergent solubilization, BMV RdRp activity copurifies with an immunoprecipitable complex of 1a, 2a, and host proteins (43, 44).

BMV replication in yeast parallels that in plant cells in all aspects tested to date, including dependence on 1a, 2a, and defined cis-acting replication and subgenomic mRNA synthesis signals; association of replication complex with the ER membrane; production of a similar excess of positive-strand over negative-strand RNA; and other features (26, 28, 42, 46, 55). Accordingly, yeast is proving to be a tractable model host for studies of viral (27, 55) and cellular (25) functions in BMV replication. The ability to express functional 1a and 2a in yeast from separate plasmids provided the opportunity to study their localization independently, revealing that 1a localizes to the ER in the absence of 2a and viral RNA templates (46). However, to date, 2a localization in the absence of other viral factors has remained obscure, due to weak 2a accumulation and immunofluorescence in the absence of 1a.

To investigate the determinants of 2a localization, we fused 2a to the green fluorescent protein (GFP), creating functional hybrids that support BMV replication. Confocal microscopy of GFP-2a fusion and cell fractionation of wild-type (wt) 2a were used to show that targeting and retention of 2a to ER depends on the helicase-like protein 1a. Deletion analysis showed that sequences near the 2a N terminus were necessary and sufficient for 1a-dependent ER localization.

MATERIALS AND METHODS

Yeast strain and cell growth.

Yeast strain YPH500 (MATα ura3-52 lys2-801 ade2-101 trp1-Δ63 his3-Δ200 leu2-Δ1) was used throughout. Yeast cultures were grown at 30°C in defined synthetic medium containing either 2% glucose or 2% galactose as indicated and lacking relevant amino acids to maintain selection for any DNA plasmids present (5).

Plasmids and plasmid constructions.

Standard procedures were used for all DNA manipulations (50). The sequences of PCR-generated DNA fragments were confirmed by DNA sequencing, and the overall structures of all plasmids were confirmed by restriction analysis.

BMV 1a protein was expressed from pB1CT19 (28), a yeast 2μm plasmid that contains a HIS3-selectable marker. BMV 2a protein was expressed from pB2YT5 (generously provided by M. Ishikawa), which is based on Ycplac111, a yeast CEN4 centromeric plasmid that contains the LEU2 selectable marker gene and the multiple-cloning sites from pUC19 (20). BMV RNA3 was expressed from the galactose-inducible, glucose-repressible GAL1 promoter in pB3RQ39 (26), which is based on Ycplac22, a yeast CEN4 centromeric plasmid containing a TRP1 selectable marker. A yeast CEN4 plasmid expressing a c-myc-tagged version of EMP47 was kindly provided by Sean Munro (53).

The yeast-enhanced version of the GFP gene (12) was fused to the 2a gene in pB2YT5 by PCR-mediated gene fusion. Laboratory designations for plasmids are given in parentheses.

pGFP-2a (pB2YT5-G2), in which GFP was fused to the N terminus of 2a in pB2YT5, was constructed using the following primers: B2-C14, d(AGTCCATGGAATCACCA), which is complementary to nucleotides 871 to 890 of BMV RNA2 and includes the unique NcoI site (underlined) in pB2YT5; B2-GFP4, d(GATCCTGCAGATGTCTAAAGGTGAAGAATT), which corresponds to the first 20 nucleotides with the start codon of the GFP gene (boldface) and includes the unique PstI site (underlined) preceding the 2a gene in pB2YT5 and four extra nucleotides to facilitate PstI digestion; B2-GFP5, d(GTATGGATGAATTGTACAAAATGTCTTCGAAAACCTGGGAT), which contains sequences corresponding to the last 20 nucleotides preceding the stop codon of GFP gene and the first 20 nucleotides with the start codon of the 2a gene (boldface); and B2-GFP6, d(ATCCCAGGTTTTCGAAGACATTTTGTACAATTCATCCATAC), which is complementary to B2-GFP5. PCR was used to amplify the pB2YT5 template with primers B2-GFP5 and B2-C14 and to amplify the GFP gene from yEGFP (12) with primers B2-GFP4 and B2-GFP6. The resulting overlapping PCR products were then combined and reamplified with the two outside primers (B2-GFP4 and B2-C14). The final PCR product containing the GFP-2a fusion was cut with PstI and NheI and used to replace the corresponding PstI-NheI fragment in pB2YT5.

pΔN161 (pB2YT5-D141G), in which GFP was fused to residue 162 from the N terminus of 2a, was similarly constructed by using primers containing sequences overlapping the last 20 nucleotides preceding the GFP stop codon and nucleotides 587 to 607 of the 2a gene.

C-terminally truncated GFP-2a mutants [ΔC428 (pB2YT5-G2D1), ΔC561 (pB2YT5-G2D2), ΔC661 (pB2YT5-G2D3), ΔC682 (pB2YT5-G2D4), ΔC702 (pB2YT5-G2D5), ΔC722 (pB2YT5-G2D6), and ΔC742 (pB2YT5-G2D7)] were constructed by PCR using the GFP-2a plasmid template with primer B2-GFP4 (see above) and one of a set of primers containing sequences complementary to the 20 nucleotides of the 2a gene preceding selected truncation sites (see Fig. 7), followed by an in-frame stop codon and a flanking BamHI site. PCR products were cut with PstI and BamHI and used to replace the corresponding PstI-BamHI fragment in pB2YT5-G2.

FIG. 7.

The N-terminal 120 residues of 2a are required and sufficient for 1a-directed localization of GFP-2a to the ER. (A) Schematic representation of full-length GFP-2a and 2a C-terminal truncations and a 2a N-terminal in-frame deletion derived from it. GFP and 2a segments and the conserved polymerase-like domain of 2a are indicated. + and − indicate the ability of each construct to be directed to perinuclear ER by 1a as determined by direct fluorescence confocal microscopy of live yeast. (B) Representative images of the intracellular distribution of the indicated GFP-2a derivatives in live yeast either coexpressing 1a (+1a) or lacking 1a (−1a). Direct fluorescence confocal microscopy was performed as described in the legend to Fig. 4. (C) Localization of 1a and GFP-2a derivatives in yeast coexpressing 1a and the indicated GFP-2a derivatives. 1a was immunostained as described in the legend to Fig. 3. Two representative images of individual yeast cells are shown for each case. Each image measures 7 μm per side.

p2a-GFP (pB2YT5-G1), in which GFP was fused to the 2a C terminus, was constructed by a strategy similar to that used for GFP-2a. The four primers used were B2-C13, d(CTTTATACTCCGAGAATTTCCTG), which corresponds to nucleotides 2151 to 2173 of BMV RNA2; B2-GFP3, d(GATCGGATCCTTATTTGTACAATTCATCCA), which is complementary to the last 20 nucleotides with the stop codon of the GFP gene and includes the unique BamHI site (underlined) following the 2a gene in pB2YT5 and four extra nucleotides to facilitate BamHI digestion; B2-GFP1, d(TTAAGCCCTCTGATCTGAGATCTAAAGGTGAAGAATTATT), which contains sequences corresponding to the last 20 nucleotides without the stop codon of 2a gene (boldface) and to the first 20 nucleotides excluding the start codon of the GFP gene; and B2-GFP2, which is complementary to B2-GFP1. After PCR-mediated gene fusion, the resulting fragment was cut with SalI and BamHI and used to replace the corresponding SalI-BamHI fragment in pB2YT5.

RNA analysis.

Yeast cells were grown in synthetic galactose medium, harvested in mid-log phase (optical density at 600 nm = 0.4 to 0.6), and frozen on dry ice. Total yeast RNA extraction was performed as previously described (28). For Northern blot analysis, 2.5 μg of total RNA was electrophoresed on a 1% formaldehyde agarose gel and blotted onto Nytran nylon membranes (Schleicher & Schuell) (55). Strand-specific ³²P-labeled RNA probes were generated as previously described (26). Radioactive signals were detected and measured using a Molecular Dynamics PhosphorImager model 425 imaging system.

Protein extraction and cell fractionation.

Yeast cells were grown in synthetic galactose medium, harvested in mid-log phase, and frozen on dry ice prior to protein extraction. Total protein was extracted by boiling for 5 min in 1% sodium dodecyl sulfate (SDS), 10 μg of pepstatin A per ml, 30 mM dithiothreitol, and 45 mM HEPES (pH 7.5). For cell fractionation, yeast cells were grown in synthetic galactose medium to mid-log phase and converted to spheroplasts in 1.0 M sorbitol using Lyticase (Sigma) (48). Spheroplasts were osmotically lysed by pipetting up and down in extraction buffer (50 mM Tris [pH 8.0], 10 mM EDTA, 10 mM dithiothreitol, 5 mM benzamidine, 2 mM phenylmethylsulfonyl fluoride, and 10 μg (each) of aprotinin, leupeptin, and pepstatin A per ml). The resulting lysate was centrifuged for 5 min at 10,000 × g. The supernatant was removed and retained, and the pellet was washed once with extraction buffer and resuspended to the original lysate volume in extraction buffer. The fractions were denatured in Laemmli loading buffer (35) and subjected to immunoblot analysis.

SDS-PAGE and immunoblot analysis.

SDS-polyacrylamide gel electrophoresis (PAGE), immunoblot analysis, and detection by chemiluminescence were performed essentially as described previously (45). Polyclonal anti-1a and monoclonal anti-2a antibodies were used as previously described (45). Monoclonal antibodies against yeast 3-phosphoglycerate kinase (PGK) were used according to the manufacturer's recommendations (Molecular Probes).

Immunofluorescent labeling and microscopy.

Yeast cells were fixed and immunostained as described previously (46). Polyclonal rabbit anti-1a antibodies, anti-Kar2p antibodies, monoclonal mouse anti-c-myc antibodies were as described (46). Mouse monoclonal antibodies against the 60-kDa subunit of yeast vacuole membrane ATPase (29) and Texas red-labelled secondary antibodies were from Molecular Probes. DNA was stained with TO-PRO-3 iodide (Molecular Probes). Images were obtained using a Bio-Rad 1024 confocal microscope at the Keck Neural Imaging Laboratory of the University of Wisconsin—Madison. The fluorescein isothiocyanate channel was used to monitor and record GFP fluorescence. To observe GFP fluorescence in living yeast cells, cells were immobilized onto glass slides by a thin layer of 1% agarose in synthetic galactose medium.

RESULTS

Construction and expression of GFP-fused 2a derivatives.

To overcome prior difficulties in immunofluorescence detection of 2a in the absence of 1a (46), we explored the use of GFP, which allows direct fluorescence microscopy of living cells. A yeast-adapted, fluorescence-enhanced version of GFP (12) has been codon optimized for expression in Candida albicans and includes two amino acid changes causing the protein to excite at 488 nm and to fluoresce more brightly than wt GFP (13). This enhanced GFP was fused in frame to the N or C terminus of 2a to create two fusions, GFP-2a and 2a-GFP, that were expressed from yeast centromeric plasmids by the galactose-inducible GAL1 promoter (Fig. 1A).

FIG. 1.

Expression of BMV 2a and GFP-fused 2a derivatives in yeast. (A) Schematic representation of expression cassettes for 2a and the fusions. The galactose-inducible GAL1 promoter, GAL1 5′ untranslatable region (5′ UTR), 2a- and GFP-coding sequences, and ADH1 polyadenylation signal are indicated. The expression cassettes were assembled into the multiple-cloning sites of Ycplac111, a yeast CEN4 centromeric plasmid, for protein expression in yeast. (B) Immunoblot analysis of 2a and the fusions. Yeast cells were transformed with plasmids expressing wt 2a or the fusions either alone (−1a) or together with a plasmid expressing 1a (+1a). Cells were grown in galactose medium to induce protein expression and harvested at mid-log phase. Total proteins were extracted and subjected to 0.1% SDS–10% PAGE and immunoblot analysis with anti-2a monoclonal antibodies.

2a-GFP and GFP-2a expression in yeast was analyzed by immunoblotting. As shown in Fig. 1B, anti-2a antibodies recognized a single major band in cells expressing wt 2a. In cells expressing 2a-GFP or GFP-2a, the anti-2a antibody-reactive band shifted to a higher position consistent with the expected molecular mass of the fusions. As judged by the intensity of the immunoblot signals, both fusions accumulated to levels similar to wt 2a. In keeping with prior findings for wt 2a (25), 1a coexpression increased accumulation of wt 2a and both 2a fusions. The earlier experiments, with the yeast ADH1 promoter to express 1a and 2a, found an approximately fivefold increase in wt 2a accumulation upon 1a expression (25). Using the stronger GAL1 promoter to express 2a, we found that 1a coexpression increased 2a, 2a-GFP, or GFP-2a accumulation approximately twofold. This lesser increase may be related to a higher starting level of 2a, since in the absence of 1a, the GAL1 promoter used here expressed two- to threefold more wt 2a than the ADH1 promoter used in the earlier experiments.

GFP-fused 2a derivatives support RNA replication and transcription.

To test if the GFP-fused 2a derivatives supported BMV RNA replication and subgenomic mRNA transcription, plasmids expressing wt 2a, GFP-2a, or 2a-GFP were introduced into yeast also expressing 1a and a wt RNA3 replication template (26). RNA3 replication products were analyzed by Northern blotting with strand-specific probes.

As demonstrated previously (27) and shown in Fig. 2, cells expressing 1a but not 2a showed accumulation of plasmid-derived positive-strand RNA3 but no negative-strand RNA3 or positive- or negative-strand RNA4. By contrast, cells coexpressing 1a and wt 2a contained negative-strand RNA3, positive- and negative-strand RNA4, and greatly increased levels of positive-strand RNA3. 2a-GFP and GFP-2a also supported RNA3 replication and subgenomic RNA4 synthesis but to different levels. GFP-2a directed positive-strand RNA3 amplification and RNA4 synthesis to 68 and 77%, respectively, of their levels with wt 2a. Negative-strand RNA3 and RNA4 levels with GFP-2a, however, were only 47 and 54% of the wt. Thus, GFP fusion to the 2a N terminus preferentially decreased negative-strand RNA synthesis. By contrast, positive- and negative-strand RNA3 and RNA4 accumulation with 2a-GFP was only 15 to 26% of the wt (Fig. 2). Since GFP-2a supported higher levels of BMV replication, it was used in subsequent localization experiments.

FIG. 2.

GFP-fused 2a derivatives support BMV RNA3 replication and transcription. 1a- and RNA3-expressing plasmids were cotransformed into yeast with plasmids expressing either wt 2a, the indicated fusion, or the starting plasmid lacking 2a sequences (−2a) as indicated. Equal amounts of total RNA prepared from the resulting galactose-induced yeast were analyzed by Northern blotting with a single-stranded, ³²P-labeled RNA probe complementary to either positive- or negative-strand RNA3 as indicated. (A) Representative Northern blots with the migration positions of each virion RNA indicated. (B) Average relative accumulation and standard deviation of RNA3 and RNA4 replication products, as determined for three independent transformants of each 2a derivative. Values are shown as percentage of RNA3 or RNA4 accumulation in yeast cells expressing wt 2a.

GFP-2a localizes normally to ER in yeast coexpressing 1a and replicating RNA3.

In yeast expressing wt 1a and 2a and actively replicating RNA3, 1a and 2a colocalize with ER markers, predominantly in the perinuclear region (46). To see if GFP-2a localization in the presence of 1a and RNA3 was similar to that of wt 2a, we used triple-channel confocal microscopy to visualize GFP-2a by intrinsic fluorescence, 1a by immunofluorescence, and nuclear DNA by staining with TO-PRO-3 iodide (Molecular Probes). When cells expressing 1a, GFP-2a, and RNA3 were fixed for immunofluorescence, GFP fluorescence was reduced relative to unfixed cells (see below) but was still readily visible.

As shown by representative images of such cells in Fig. 3A, 1a and GFP-2a displayed almost perfect colocalization, usually in the shape of a partial or complete ring. DNA staining (Fig. 3A) showed that these rings bounded the nucleus, suggesting ER localization. To examine this further, we compared the pattern of GFP-2a fluorescence in these cells to the immunofluorescence pattern of a well-characterized ER marker, the Kar2p protein. Kar2p, an ER lumen protein, is the yeast homolog of the mammalian chaperone BiP/GRP78 (47). As found previously (46, 47), Kar2p displayed prominent perinuclear localization with occasional extensions into the cytoplasm and around the cell periphery (Fig. 3B). As for wt 2a (46), the sites of GFP-2a accumulation displayed good colocalization with Kar2p, predominantly colocalizing with Kar2p in the perinuclear region (Fig. 3B). Thus, as expected from GFP-2a activity in BMV replication, GFP-2a localization in cells coexpressing 1a and replicating RNA3 matched that of wt 2a, suggesting that GFP-2a should be a suitable marker for studying the determinants of 2a localization.

FIG. 3.

Colocalization of GFP-2a with 1a and perinuclear ER in yeast coexpressing GFP-2a, 1a, and RNA3. Yeast cotransformed with plasmids expressing GFP-2a, 1a, and RNA3 was grown in galactose medium to induce protein expression and harvested in mid-log phase. Cells were then fixed with formaldehyde, treated with Lyticase to remove the cell wall, and incubated with polyclonal antibodies against 1a (A) or ER-resident protein Kar2p (B) and finally treated with Texas red-labelled secondary antibodies. After secondary antibody treatment, cells were incubated briefly with the DNA stain TO-PRO-3 iodide to visualize the nucleus. For each cell, the differentially fluorescing protein (red, 1a or Kar2p; green, GFP-2a) and DNA (blue) images were gathered simultaneously from the same optical section with a multichannel confocal microscope and appropriate filters. Control experiments omitting 1a or Kar2p antibodies, DNA stain, or GFP-2a confirmed that there was no signal leakage from any channel into the other channels. The three images were digitally superimposed (Merged) to depict the relationship among GFP-2a, 1a, ER, and nucleus. Two representative images of individual yeast cells are shown for each case. Each image measures 7 μm per side.

Localization of GFP-2a to perinuclear ER is 1a dependent.

To determine if 2a localization to the ER was dependent on 1a or the RNA3 replication template, we examined possible effects of 1a, RNA3, or both on GFP-2a distribution in live yeast. Live yeast cells expressing the desired BMV components were immobilized on microscope slides with a thin layer of agarose and examined by direct fluorescence confocal microscopy. As a control, yeast cells expressing free GFP (i.e., not linked to 2a) alone or together with 1a were examined in parallel.

As expected, green fluorescence was observed only in cells expressing either free GFP or GFP-2a but not in cells expressing wt 2a (Fig. 4). Free GFP was distributed uniformly throughout the cytoplasm but was excluded from some large organelles. Neither the strength nor the distribution of free GFP was discernibly affected by 1a.

FIG. 4.

1a-induced redistribution of GFP-2a to perinuclear ER in yeast. A plasmid expressing either wt 2a, free GFP, or GFP-2a was transformed into yeast either alone (−1a) or together with a plasmid expressing 1a (+1a). Cells were grown in galactose medium, harvested at mid-log phase, immobilized on glass slides with a thin layer of 1% agarose gel in synthetic galactose medium, and visualized on a confocal microscope by direct GFP fluorescence. Except for yeast expressing wt 2a, two representative images of individual yeast cells are shown for each case. Each image measures 7 μm per side. Arrowheads indicate spots of green fluorescence.

In the absence of 1a or RNA3, GFP-2a was distributed diffusely throughout the cytoplasm and in localized punctate structures (Fig. 4). Integrating the fluorescence signals throughout typical sections showed that approximately half of the GFP-2a was associated with the punctate structures, and half was distributed diffusely throughout the cytoplasm. Like free GFP, GFP-2a was excluded from some large organelles, including the vacuole (see Fig. 5 below).

FIG. 5.

Relation of GFP-2a localization to cellular organelles in the absence of 1a. Yeast cells expressing GFP-2a alone were fixed with formaldehyde, treated with Lyticase to remove the cell wall, and incubated with antibodies against proteins localizing on the ER, Golgi apparatus, or vacuolar membranes, respectively. DNA stain TO-PRO-3 iodide was used to visualize the mitochondria and the nucleus. For each cell, the differentially fluorescing protein and DNA images were gathered simultaneously with appropriate filters as described in the legend to Fig. 3. The two images were digitally superimposed (Merged) to depict the distribution of GFP-2a relative to cellular organelles. Two representative images of individual yeast cells are shown for each case. Each image measures 7 μm per side.

By contrast, in cells coexpressing 1a, the diffuse cytoplasmic distribution of GFP-2a disappeared and was replaced by concentration of GFP-2a fluorescence in partial or complete rings (Fig. 4) typical of the perinuclear ER localization shown in Fig. 3A and B. GFP-2a localization in such rings was completely 1a dependent, as analogous structures were never found in cells expressing GFP-2a without 1a, even after numerous experiments involving examination of thousands of cells.

Live yeast coexpressing 1a and GFP-2a also occasionally displayed green fluorescence in spots distinct from the primary perinuclear ring structures (Fig. 4, lower rightmost panel). Relative to the punctate GFP-2a fluorescence in cells lacking 1a, the green fluorescent spots in cells coexpressing GFP-2a and 1a were similar in shape and size but were found at much lower frequency and displayed weaker fluorescence. In keeping with their weaker fluorescence, such spots were rarely observed after cells coexpressing GFP-2a and 1a were fixed for immunofluorescence (Fig. 3), likely because such fixation reduced the intrinsic fluorescence of GFP-2a as noted above.

Expressing RNA3 in cells either coexpressing GFP-2a and 1a or expressing GFP-2a alone did not affect the localization of GFP-2a (data not shown), indicating that RNA3 neither was required for nor affected 1a-dependent localization of GFP-2a to the perinuclear ER.

GFP-2a spots are not associated with the ER, Golgi apparatus, or other tested organelles.

In the absence of 1a (Fig. 4), general cytoplasmic fluorescence suggested that much of GFP-2a was a soluble, cytoplasmic protein, while brighter, localized spots suggested that some GFP-2a was either aggregated or associated with specific organelles. Among other possibilities, intrinsic localization of a portion of 2a to ER sites might assist replication complex assembly. Alternatively, the yeast Golgi apparatus has a punctate distribution (53), and retrograde Golgi-to-ER transport might similarly target 2a for interaction with 1a. To test these and other possibilities, we compared the immunofluorescence distribution of relevant cell markers and direct GFP-2a fluorescence in yeast expressing GFP-2a without 1a or RNA3. To visualize the ER and Golgi apparatus, respectively, we used antibodies to Kar2p (see also Fig. 3) and Emp47p (53). In addition, we visualized yeast vacuoles with antibodies against the 60-kDa subunit of yeast vacuole membrane ATPase (29) and visualized mitochondria by staining mitochondrial (and nuclear) DNA. As illustrated by the representative images shown in Fig. 5, the punctate sites of GFP-2a fluorescence in the absence of 1a were not consistently associated with any of these organelles.

1a induces soluble GFP-2a and wt 2a to associate with membrane.

To further explore the 1a-induced ER association of GFP-2a, to test for similar 1a-dependence in ER association of wt 2a, and to examine possible membrane association of the punctate fraction of GFP-2a in the absence of 1a, we used cell fractionation. Yeast expressing GFP-2a or wt 2a in the presence or absence of 1a was spheroplasted and osmotically lysed, and membranes were pelleted at 10,000 × g. Western blotting was then used to examine the distribution between membrane and supernatant fractions of GFP-2a, wt 2a, 1a, and, as a control, PGK, a soluble cytoplasmic protein (6). As shown in Fig. 6, GFP-2a and wt 2a showed parallel distribution patterns throughout the experiment. In the absence of 1a, GFP-2a and wt 2a behaved like the soluble, cytoplasmic protein PGK and were found almost completely in the postmembrane supernatant fraction. By contrast, virtually all of 1a was found in the membrane fraction. When 1a was coexpressed, PGK fractionation into the supernatant was unaltered, but approximately 50% of GFP-2a and wt 2a was found in the membrane pellet. Thus, as for GFP-2a, membrane association of wt 2a was 1a dependent. Possible reasons for the incomplete membrane association of wt 2a and GFP-2a in cells coexpressing 1a are considered in the Discussion.

FIG. 6.

Effects of 1a on GFP-2a and wt 2a distribution by cell fractionation. Yeast cells expressing either GFP-2a or wt 2a alone (−1a) or coexpressing 1a (+1a) were grown in galactose medium and harvested at mid-log phase. The cells were then treated with Lyticase to remove the cell wall, and the resulting spheroplasts were lysed osmotically to yield a total protein fraction (Tot.). A portion of the lysate was then subjected to low-speed centrifugation to yield pellet (Pell.) and supernatant (Sup.) fractions. Equal percentages of each fraction were subjected to 0.1% SDS–10% PAGE and immunoblot analysis with antibodies against 2a, 1a, or PGK, a cytoplasmic soluble protein.

N-terminal 120 residues of 2a direct 1a-dependent ER localization.

BMV 2a protein consists of a central polymerase-like domain flanked by N- and C-terminal extensions of approximately 200 and 125 amino acids, respectively (Fig. 7A). Immunoprecipitation and yeast two-hybrid experiments show that a region within 2a amino acids 25 to 140 interacts with the helicase-like domain of 1a (30, 39). However, genetic experiments show that the 2a polymerase-like domain also is involved in selective interactions with 1a that are essential for RNA replication (54).

To identify the 2a region(s) required in vivo for 1a-directed ER localization, a series of C-terminal truncations was made in the 2a sequence of GFP-2a (Fig. 7A). The resulting proteins retained varying lengths of an N-terminal 2a amino acid sequence, i.e., 394 (ΔC428), 261 (ΔC561), 161 (ΔC661), 140 (ΔC682), 120 (ΔC702), 100 (ΔC722), or 80 (ΔC742) residues.

In the absence of 1a, as shown for ΔC702 and ΔC722 as examples (Fig. 7B), all C-terminal truncations were distributed diffusely over much of the cell but excluded from some large organelles, presumably including the vacuole as for full-length GFP-2a (Fig. 5). This distribution was similar to the diffuse cytoplasmic part of the full-length GFP-2a distribution in the absence of 1a. However, none of the truncated derivatives (Fig. 7B) formed the localized punctate structures observed for full-length GFP-2a (Fig. 4).

In cells coexpressing 1a, the intracellular distribution of the C-terminal truncation derivatives varied with the length of the truncation. Direct fluorescence microscopy of live cells showed that GFP-2a truncations retaining 120 or more 2a N-terminal residues formed partial or complete rings typical of the 1a-dependent perinuclear ER localization of wt 2a (46) or GFP-2a (Fig. 3 and 4). This is illustrated in Fig. 7B for ΔC702, and equivalent results were obtained for all longer GFP-2a derivatives (ΔC428, ΔC561, ΔC661, and ΔC682). After fixation of such cells, immunofluorescence confirmed that these GFP-2a derivatives colocalized with 1a (i.e., ΔC702 [Fig. 7C]). In contrast, truncations containing 100 or fewer 2a N-terminal residues showed no change in intracellular localization upon 1a expression, as illustrated in Fig. 7B and C for ΔC722. Equivalent results were obtained for ΔC742. Since the diffused GFP-2a fluorescence was much weaker than the localized punctate and the perinuclear GFP-2a fluorescence, such diffused fluorescence was rarely observed after cells were fixed for immunofluorescence (Fig. 7C), likely because such fixation reduced the intrinsic fluorescence of GFP-2a as noted earlier.

Thus, the first 120 amino acids of 2a were sufficient for 1a-directed ER localization in vivo. However, Smirnyagina et al. showed that a 2a derivative lacking the first 161 amino acids can support RNA replication when expressed from a plasmid (54). Moreover, although this 2a derivative lacks the 1a-interactive N-terminal region mapped in vivo (30, 39), it shows strain-specific compatibility requirements for 1a (54). To determine whether this central portion of 2a might also contain sequences able to direct 1a-dependent ER localization in vivo, we constructed ΔN161, a GFP-2a derivative with an in-frame deletion of the first 161 residues of 2a (Fig. 7A). As shown in Fig. 7B, ΔN161 localization in the absence of 1a was similar to full-length GFP-2a (Fig. 4B), showing both diffuse localization throughout the cytoplasm and some punctate spots. However, 1a coexpression did not alter the ΔN161 distribution (Fig. 7B) and produced no significant colocalization of ΔN161 with 1a (Fig. 7C).

DISCUSSION

Association of the RNA replication complex with intracellular membranes appears to be a universal feature of positive-strand RNA viruses. We are using the BMV-yeast system as a model to investigate such fundamental aspects of RNA replication. In this report, we constructed functional GFP-fused 2a proteins to study the determinants of 2a localization to the ER sites of BMV RNA synthesis. The results from both confocal microscopy and cell fractionation show that localization of the polymerase-like 2a protein to the ER depends on the multifunctional viral 1a protein. In keeping with this, prior work shows that 1a can interact directly with 2a in vitro (30) (see also below). In vivo, 1a also interacts independently with the ER (46) and with viral RNA replication templates (55). Together, these findings imply that 1a is a key organizer of the assembly of the BMV replication complex and a major determinant for its ER localization and retention. Thus 1a, which also has RNA-capping activities (2) and a DEAD box helicase domain, plays major roles in the assembly and function of the BMV RNA replication complex.

GFP-2a and wt 2a distribution in the absence of 1a.

The strong fluorescence of GFP-2a in live or fixed cells greatly facilitated determining its intracellular distribution. In the absence of 1a, GFP-2a was present in two forms in the cytoplasm. Approximately 50% of GFP-2a fluorescence was distributed diffusely through the cytoplasm like a soluble protein, while the other 50% was found in brighter, punctate structures (Fig. 4). While low accumulation and weak immunofluorescence of wt 2a in the absence of 1a did not allow comprehensive or conclusive definition of its intracellular distribution (46), such diffuse distribution and punctate spots were also observed for wt 2a in the absence of 1a (M. Restrepo-Hartwig and P. Ahlquist, unpublished results). In keeping with the association of spots with both GFP-2a and wt 2a, free GFP did not form such spots (Fig. 4), and GFP-2a formation of such spots required the C-terminal half of 2a (Fig. 7).

The punctate localization did not appear to result from association with subcellular organelles, since no association of GFP-2a with ER, Golgi apparatus, mitochondria, or vacuole was found by confocal microscopy (Fig. 5), and GFP-2a and wt 2a both remained in the supernatant after centrifugation under conditions that precipitate membrane-bounded organelles (Fig. 6). Alternatively, the punctate spots may represent 2a aggregation or polymerization. 2a and other positive-strand RNA virus polymerases have shown a tendency to aggregate when expressed to significant levels (16; R. Hershberger and P. Ahlquist, unpublished results). Interestingly, for poliovirus 3D RNA polymerase, oligomerization may be required for efficient binding of template RNA and RNA synthesis (40). Further studies will be needed to determine whether punctate localization of GFP-2a represents aggregation and, if so, whether the interactions involved have any functional role in replication. 2a aggregation would be consistent with the cell fractionation results (Fig. 6), since protein aggregates may not be sedimented by the relatively low-speed centrifugation (34) or 2a aggregates might have dissociated during cell fractionation.

ER localization of 2a depends on 1a and N-proximal 2a sequences.

Genetic studies show that both 1a and 2a are required in vivo for each form of BMV RNA synthesis: negative-strand and positive-strand genomic RNA synthesis and subgenomic mRNA transcription (15, 27). Moreover, RNA replication in vivo requires compatible 1a-2a interaction (15). Accordingly, the observed colocalization of 1a and 2a at the ER sites of viral RNA synthesis is essential for BMV replication. While 1a localizes to the ER in the absence of any other viral factors (46), the confocal microscopy (Fig. 4 and 7) and cell fractionation (Fig. 6) results presented here revealed that GFP-2a and wt 2a localization to the ER depends on coexpression of 1a.

Deletion analysis of GFP-2a further showed that sequences within the first 120 amino acids of 2a were necessary and sufficient for 1a-directed localization to the ER and that the first 100 amino acids of 2a were insufficient for this function (Fig. 7). These results fit well with previous findings that sequences within 2a amino acids 25 to 140 interact directly with the C-terminal helicase-like domain of 1a in vitro or in yeast two-hybrid assays (30, 39). The strength of this interaction between the 2a N terminus and 1a and its role in 1a-dependent ER localization of 2a imply that this interaction provides a positively selectable function for the virus. Surprisingly, however, prior experiments showed that these N-terminal 2a sequences were dispensable for BMV RNA replication under at least some circumstances (54). Specifically, when 2a was constitutively expressed in plant protoplasts from a DNA plasmid, deletion of the first 161 amino acids of 2a resulted in less than a twofold decrease in RNA3 replication. In the absence of these N-terminal sequences and some C-terminal sequences, gene reassortments with a related bromovirus showed that the 2a core retained selectivity for functioning with its cognate 1a (54), revealing that the polymerase-like 2a core also interacts with 1a in selective, essential ways.

As initially suggested by Smirnyagina et al. (54), the N-terminal, 1a-interactive segment of 2a may be dispensable only under certain modes of 2a protein expression. In the deletion experiments of Smirnyagina et al. (54), 1a and 2a derivatives were expressed from DNA plasmids that supported BMV RNA3 replication to higher levels than wt BMV infection (15). Under high-level, plasmid-directed, constitutive expression, higher concentrations of 1a and 2a may facilitate sufficient 1a-2a colocalization for replication through the lower affinity interaction of the 2a polymerase-like core with 1a. By contrast, in wt BMV infection, 1a and 2a are expressed from replicating RNA1 and RNA2 introduced at low multiplicity per cell, and 1a and 2a concentrations early in infection are low. Under these conditions, the high-affinity interaction between 1a and the 2a N terminus, which is required for efficient 2a targeting to the ER (Fig. 7), may be crucial to promote rapid replication complex assembly before the inoculum RNAs are degraded, which can occur within minutes (27). Direct testing of this hypothesis has been complicated by cis-acting RNA2 replication signals in the region encoding the N-terminal 1a-interactive segment of 2a (56). Nevertheless, even after accounting for the cis-acting effects of these deletions on RNA2 replication and thus on 2a protein expression, deletions in the N-terminal 1a-interactive segment of 2a show significant trans-acting inhibitory effects on RNA replication, consistent with an important role for the 2a N terminus in natural infection (15, 56).

Incomplete 2a localization to ER.

Upon cell fractionation (Fig. 6), virtually all detectable 1a protein was found in the rapidly sedimenting membrane fraction. However, in cell populations expressing both 1a and 2a, only about 50% of 2a was found with 1a in this membrane fraction, while the other 50% remained in the supernatant (Fig. 6). Several factors may have contributed to these observations. First, because of imperfect segregation to daughter cells, a typical yeast 2μm plasmid is missing from 10 to 30% of yeast cells in liquid culture, even under selective conditions (11). Similarly, immunofluorescence microscopy showed that approximately 30% of cells in 1a-expressing yeast populations lack 1a. Thus, since 30% of yeast cells expressing 2a lack 1a expression, 2a should not be membrane associated in these cells. Additionally, 2a may have a lower affinity for membrane association than 1a, and some 2a associated with membranes in vivo may have become dissociated during cell fractionation. Nevertheless, close inspection of GFP-2a fluorescence in cells coexpressing 1a showed that, while most GFP-2a was localized in a partial or complete perinuclear ring, some was distributed diffusely in the cytoplasm, similar to GFP-2a localization in the absence of 1a (Fig. 4 and data not shown). Thus, in vivo, a fraction of 2a was not localized to the ER by 1a. Whether this was caused by 2a overexpression relative to 1a is not yet known.