INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

All retroviruses incorporate two identical copies of their RNA genome into each virion. The genomic RNA molecules are linked near their 5' ends by noncovalent interactions to form a stable structure possessing ordered secondary and tertiary structure. Although there are multiple contact points throughout the two parallel RNA molecules, the most stable linkage is called the dimer linkage structure. The dimer linkage structure can be visualized by electron microscopy and appears to be a region about 50 nucleotides (nt) in length near the 5' end of the genome (centered around nucleotide 511 in Rous sarcoma virus [RSV]) (1, 20, 23). Dimerization is required for infectivity, although precisely how it contributes to the replication cycle remains poorly understood. Dimerization is believed to facilitate recombination during reverse transcription by enabling close approximation of the viral RNA molecules, leading to increased genetic diversity and improved viral fitness (15, 16, 26). The dimeric RNA structure has also been implicated in inhibiting the translation of unspliced viral RNA so that genomic RNA is available for packaging; however, there is little experimental evidence in support of this idea (26).

Because the RNA sequences that are important for dimerization overlap those required for RNA incorporation into virus particles, dimerization and packaging were postulated to be functionally associated. In support of this idea, there is evidence that dimerization is required for genomic RNA incorporation in murine leukemia virus and human immunodeficiency virus type 1 (HIV-1) (13, 14). Both of these viruses appear to initially package unstable dimers that mature into more thermostable complexes after protease activation (13, 14). However, for RSV, it is less clear whether dimerization occurs prior to packaging. When isolated by conventional methods, rapid-harvest and protease-defective RSV particles contain primarily monomeric RNA, although a small amount of dimeric RNA has been detected (25, 32; T. Cairns and R. Craven, unpublished results). Incubation of rapid-harvest virus results in conversion of monomers to dimers (2, 3). Thus, it is possible that monomers are initially packaged and dimerization begins later in the assembly process or even after budding. In contrast to these reports, Stoltzfus and Snyder found that the RNA from rapid-harvest B77 avian sarcoma virus particles exists as fragile dimers that readily dissociate upon extraction (33). These authors proposed that the dimers undergo stabilization after budding and that this RNA maturation process might be mediated by a viral protein. From the existing data, it has not been possible to determine with certainty whether RSV packages monomers or dimers.

To elucidate the cis- and trans-acting elements needed for dimer formation, in vitro assays for dimerization have been utilized. Studies of this type with RSV have identified sequences within the 5' region of the genome that allow spontaneous dimer formation in the absence of any viral proteins. Initially, sequences at nt 485 to 530 and 531 to 634 were postulated to initiate dimerization in vitro (19). However, later results disputed this finding, instead reporting that an autocomplementary stem-loop structure (L3) between nt 258 and 274 in avian sarcoma-leukosis virus initiates dimer formation in vitro via a "kissing complex" formed by Watson-Crick base pairing between complementary bases in opposing loops (12). In the latter study, regions downstream of the gag initiation codon (nt 400 to 650) were found to be dispensable for in vitro dimerization. The only trans-acting factor of avian retroviruses that has a demonstrated role in dimer formation is the nucleocapsid (NC) domain of the Gag protein, which promotes in vitro dimerization at low concentrations of RNA (6, 9, 11, 21). The NC region acts as a nucleic acid chaperone to facilitate maturation of the dimer structure and also catalyzes the annealing of the tRNA primer to the primer binding site (5, 6, 11, 29).

The matrix (MA) protein of RSV was recently implicated as another possible factor involved in genomic RNA dimerization (27). In that study, we found that adding 10 codons of the v-src gene as a 5' extension of gag allowed normal virus particle release but led to a loss of infectivity. Mutant virus particles were normal in their levels of incorporation of Env, Gag, and Gag-Pol, and the proteolytic processing of these polyproteins appeared normal. However, genomic RNA packaging was mildly reduced (in the range of two- to fourfold) and the RNA isolated from virus particles was in the form of monomers. It seemed most likely that the loss of dimer formation was due to the change in MA rather than to the alteration in the RNA, because a point mutation, which led to the loss of myristylation of the Src sequence (and thus its membrane binding activity), restored RNA dimerization and infectivity (27). However, because the mutation was located within the portion of the genome that contains important regulatory elements for RNA packaging and dimerization, it was important to determine whether cis-acting elements could be responsible for the defect. In the present study, we constructed new mutants to distinguish between RNA and protein effects. Here we provide definitive evidence that it is indeed the mutant MA protein (rather than the altered RNA sequence) that affects dimerization. Moreover, our experimental results suggest that the intracellular distribution of MA may be linked to genomic RNA dimerization.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Viruses and cells. The proviral RSV constructs utilized in these studies all contain the RSV Prague C gag gene from pATV8. The wild-type, infectious proviral construct pRC.V8 is a derivative of pBH.RCAN.HiSV (pRCAN) (4, 7). Mutants pRC.Myr1E, pRCMyr1E (27), pRC.HB12, and pRC.T14K.B1c (28) have been described previously. The pRCASBP(A)/GFP proviral construct (31), a kind gift from Mark Federspiel, Mayo Foundation, Rochester, Minn. (30), has the gene encoding green fluorescent protein (GFP) in place of the src gene. The chemically transformed QT6 quail fibroblast cell line was utilized for all experiments and was maintained as described previously (4, 22).

Site-directed mutagenesis of the gag gene. Oligonucleotide-directed mutagenesis was performed as previously described (17, 36). The oligonucleotides utilized were 5'-CCCGGTGGATCAAGCATGG(G/A)AAGTAGTAAGTCAAAACCTAAGGACAGCGAAGCCGTCATAAAGGTGAT for myr1e.cc and myr1e.cc and 5'-GGTCGCCCGGTGGATCAACTTAAGGAAGTAGTAAGTCAAAACCTAAGGACAGCATGGAAGCCGTCATAAAG for myr1e.atg. The designation "cc" is an abbreviation for "changed codons." Replicative-form DNAs were digested with SstI (nt 255) and HpaI (nt 2731) and subcloned into pRCASBP(A)/GFP by restriction fragment exchange. A PCR-based strategy was used to make myr1e.MB, using oligonucleotide primers 5'-CTCAGAAGTCGACGAGCTCTACT (USP19.263) and 5'-GGACTAGTGCGGACGAAATCACCTTTATGACGGCTT with pRC.Myr1E as the template. The PCR product was digested with SstI and SpeI and cloned into pRCAN.MA6S (24). All mutations were confirmed by dideoxy sequencing, and two independent clones of each gag allele were used in each assay described below.

Radioimmunoprecipitation. QT6 cells were transfected using the calcium phosphate precipitation method and metabolically labeled with [³⁵S]methionine, and RSV proteins were immunoprecipitated with a polyclonal anti-RSV serum (34), separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and detected by autoradiography as described previously (28).

Immunoblot analysis. Virus particles harvested from culture supernatants of QT6 cell lines were concentrated by ultracentrifugation and normalized according to reverse transcriptase (RT) activity (as above), and viral proteins were separated by SDS-PAGE and subjected to immunoblotting with anti-RSV and anti-MA sera as described previously (4, 27).

Nondenaturing Northern blotting. Virus particles produced from the stable lines or after transient transfection were collected by ultracentrifugation and normalized for RT content, and RNA was isolated, separated by electrophoresis, and blotted to nylon membranes as described previously (14, 27). The blots were hybridized using an antisense gag riboprobe and autoradiography was performed (27).

In vitro dimerization. The sequence from nt 1 to 845 in pRC.V8 was PCR amplified using primers 5'-CAACCGGACGTCGCCATTTGACCATTCACCACA and GTCCGGTACCATAGCAGGATGTGCCAAC (P89). The PCR product was digested with AatII and Asp718 and transferred into pGEM7Zf(+) (Promega). The resulting plasmid, pGEM.RSVLTR/MA, was used to exchange SstI-BspEI fragments to introduce the myr1e, myr1e, myr1e.cc, myr1e.cc, and myr1e.atg gag alleles into the pGEM expression vector. The pGEM.RSVLTR.15-4 deletion was made by digesting pGEM.RSVLTR/MA with SstI followed by Bal 31 endonuclease digestion to create a 77-bp deletion between nt 219 and 296. All constructs were confirmed by dideoxy sequencing. In vitro-transcribed viral RNAs were made using the Ribomax System (Promega) following linearization of each of the pGEM.RSV vectors described above with HindIII. The RNAs were purified using Chromaspin-100 columns (Sigma), and 1.05 µg of RNA was used for each in vitro dimerization reaction following the protocol described by Fosse et al. (12). The products of the reaction were subjected to electrophoresis on a 1.8% Metaphor agarose gel (FMC BioProducts), stained with ethidium bromide, visualized by UV light, and photographed.

RNase protection assay (RPA). Virus particles were collected and concentrated by ultracentrifugation and analyzed for RT activity, and RNA was isolated by the method of Fu and Rein (14, 27). RT activity was used as an estimate of the relative number of virus particles, and each mutant was normalized to the wild type so that equivalent amounts of RNA per particle were used. Serial 1:2 dilutions of RNA were hybridized to an antisense gag riboprobe (approximately 550 bp in length) using the Ambion RPA III kit. The probe had been labeled with [³²P]CTP during in vitro transcription (Promega T7 Riboprobe System). The RNAs were separated on a 5% acrylamide-8 M urea denaturing polyacrylamide gel and analyzed by autoradiography and using a PhosphorImager (Molecular Dynamics).

Virus infectivity assays. QT6 cells were transfected with wild-type or mutant proviral constructs (as above), and growth medium was collected from the cells 1 day after transfection and then once or twice weekly. The medium was pelletted through a 25% sucrose cushion at 100,000 × g for 45 min, resuspended in phosphate-buffered saline, and stored at 60°C. At the end of the collection period, all of the samples were tested for RT activity in triplicate (4).

MA-GFP fusion proteins and confocal microscopy. The pEGFP.N2 plasmid (Clontech) was utilized to make MA-GFP fusion proteins. PCR amplification of the wild-type MA sequence from pRC.V8 was performed using primers USP19.263 and P89. The PCR product and pEGFP.N2 were digested with Asp718, filled in with Klenow, and cut with SstI to create pMA-GFP. Mutations myr1e, myr1e, myr1e.cc, myr1e.cc, and myr1e.atg were cloned into pMA-GFP by restriction fragment exchange utilizing SstI and BspEI and were confirmed by dideoxy sequencing. The resulting plasmids were used to transfect 35-mm-diameter dishes of QT6 cells for 6 to 12 h using the calcium phosphate method. Approximately 6 to 18 h later, the cells were washed twice in Tris-buffered saline and examined using a Ziess confocal LSM 10 BioMed microscope.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

We previously described a mutation in the RSV MA coding sequence that disturbs viral RNA dimer formation, raising the intriguing possibility that the MA protein may play a role in dimerization. This mutant, Myr1E (Fig. 1), has an extension of the 10-amino-acid Src membrane binding domain at the Gag N terminus. We hypothesized that the additional membrane binding domain (that of Src plus the RSV Gag membrane binding domain) on the Myr1E Gag protein could be responsible for inhibiting RNA dimerization (27). However, it was also possible that the dimerization defect was due to changes in RNA-RNA interactions arising from the nucleotides inserted in myr1e, and thus changes at the protein level could be irrelevant. To distinguish between these cis and trans effects, additional MA mutants were examined.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   Schematic diagram of wild-type and mutant MA sequences. The RSV Gag protein is shown at the top, with the MA, p2, p10, CA, NC, and PR domains indicated. Numbers above the protein indicate the amino acid residues for MA and full-length Gag. The N-terminal sequence of wild-type MA is shown below the diagram. In the middle panel, substitutions in the RSV MA sequence are illustrated. The wavy line represents myristic acid attached to the N terminus of the protein. Dashed and dotted lines indicate amino acid residues that are identical to wild type. Myr2/HB12 contains a substitution of basic residues from the HIV-1 M domain (KKKYKLK 28). Myr2/T14K.B1c has a single substitution at position 14 and deletion of residues 74 to 98, as indicated. In the lower panel, N-terminal extension mutants are shown, along with their nucleotide sequences. The boxed sequence (GSSKSKPKD) is derived from the v-Src protein. Blank spaces within the sequence signify a deletion. Nucleotides indicated in bold italic type have been altered from the original myr1e sequence. Ovals indicate the initiation sites for protein synthesis. WT, wild type.

RNA dimerization defects in additional MA mutants. In earlier studies, we described other MA mutants that lacked infectivity despite being competent for particle assembly (28). In mutant Myr2.HB12, a myristic acid addition site was created by an E-to-G change at position 2 and a cluster of basic residues from the HIV-1 MA protein (KKKYKLK) was substituted for RSV MA residues 12 to 18 (Fig. 1). The Myr2.HB12 Gag protein directs particle assembly efficiently and is insensitive to a downstream deletion (residues 74 to 98, designated B1c) within MA that normally inhibits budding (28). This finding suggested that the membrane binding activity of the mutant Gag protein was increased compared to that of the wild type. We also found that a single substitution within MA (T to K at position 14) was sufficient to restore budding to the same downstream deletion (B1c) as long as myristate was present at the N terminus (mutant Myr2.T14K.B1c [Fig. 1]). For both mutants, the virus particles appeared normal with respect to Gag and Gag-Pol content and processing, but viral RNA had not been examined.

View larger version (48K): [in this window] [in a new window]   FIG. 2.   Nondenaturing Northern blot analysis of viral RNA. Viral RNA was extracted from virus particles produced after transfection of QT6 cells with mutant or wild-type (WT) proviral plasmid DNAs. After the RNA was blotted to a nylon membrane, an antisense radiolabeled gag riboprobe was used to detect viral RNA. The positions of genomic RNA dimers (D) and monomers (M) are indicated by arrows.

Alterations in the nucleotide sequences of src-modified gag alleles. Proviral mutants were constructed which altered the RNA coding sequence but maintained the protein sequences of wild-type Gag, Myr1E, and Myr1E (Fig. 1). In Myr1E.ATG, the RNA coding sequence for the Src peptide was inserted just upstream of the gag AUG; additionally, the initiation codon for Src was changed to UAA to prevent translation from beginning there, and instead translation initiated at the gag AUG. Three extra bases (AGC) were inserted immediately preceding the gag initiation codon to match the context of the wild-type translation start site in an attempt to maximize translation initiation from this site. In Myr1E.cc and Myr1E.cc, the RNA sequences were altered but the protein sequences were identical to those of Myr1E and Myr1E, respectively, except for the addition of a serine residue (amino acid 11) resulting from the extra AGC inserted to correspond to the sequence of Myr1E.ATG (Fig. 1).

View larger version (44K): [in this window] [in a new window]   FIG. 3.   Virus particle production in avian cells. (A) Immunoprecipitation of RSV Gag proteins. Following transient transfection of QT6 cells with proviral DNAs bearing the indicated MA mutants, cells were metabolically radiolabeled and lysed, and Gag proteins were immunoprecipitated with polyclonal anti-RSV serum. The full-length Gag protein (Pr76gag), CA, MA, and PR cleavage products are indicated at the left. (B) Immunoblot analysis. Virus particles were concentrated by sedimentation through a sucrose cushion and analyzed by Western blotting using a combination of anti-RSV and anti-MA serum to detect processed p23 (MA + p2) and p19 (MA) proteins, as shown by brackets. WT, wild type.

Dimerization state of the viral RNA. To determine whether the src nucleotide insertion upstream of gag was responsible for disrupting genomic RNA dimerization, viral RNAs from particles synthesized by each mutant were analyzed by nondenaturing Northern blot analysis as previously described (13, 27). If the src mutation had a cis-acting effect, we would expect that mutant Myr1E.ATG, which has the src nucleotide sequence but encodes a wild-type Gag protein, would be defective in RNA dimer formation. However, this was not the case (Fig. 4). Despite the presence of the src sequence, Myr1E.ATG particles contained dimeric RNA just like wild-type particles. Further support for a protein effect was obtained from examining Myr1E.cc and Myr1E.cc. Both of these constructs gave the same results as their unmodified counterparts. That is, Myr1E.cc, like Myr1E, produced particles containing monomers, and Myr1E.cc, like Myr1E, produced dimers. Therefore, these changes in primary nucleotide sequence are not critical for dimer formation, and the src RNA sequence does not prohibit RNA dimerization in vivo.

View larger version (92K): [in this window] [in a new window]   FIG. 4.   Genomic RNA analysis. Genomic RNAs isolated from virus particles produced by transfection of QT6 cells with the indicated mutant and wild-type (WT) DNAs were subjected to electrophoresis under nondenaturing conditions as described in the legend to Fig. 2. The positions of dimers (D) and monomers (M) are shown on the left.

RNA dimerization in vitro. Although it is clear that viral RNAs isolated from Myr1E and Myr1E.cc do not dimerize, we wondered whether these RNAs could form dimers under in vitro conditions. If not, it would be likely that a cis-acting element prevents the RNAs from forming a stable secondary structure. However, if all of the viral RNAs form dimers in vitro, there must be a factor other than the RNA sequence (e.g., MA) that is involved in the defect seen in Myr1E. We generated subviral RNAs (nt 1 to 848) for the wild type and for each mutant by in vitro transcription. Purified RNAs were denatured by heating to 90°C and then incubated at 20°C (monomer conditions) or 50°C (dimer conditions), following an established protocol (12). As a negative control for dimerization, we deleted nt 219 to 296 (pGEM.RSV.15-4), a region shown to be essential for in vitro dimerization (12). This mutation abrogated dimer formation (Fig. 5, upper panel), as expected. Wild-type viral RNA formed a mixture of monomers and dimers at 20°C and was fully dimerized at 50°C (Fig. 5, lower panel), as shown by others (12).

View larger version (75K): [in this window] [in a new window]   FIG. 5.   In vitro dimerization assay. In vitro-transcribed viral RNA was denatured (90°C) or incubated at 20 or 50°C and separated by native agarose gel electrophoresis as described in Materials and Methods. RNA molecular size markers (in kilobases), are indicated on the left. The positions of RNA dimers (D) and monomers (M) are indicated by arrows to the right of the gel. WT, wild type.

myr1e.MB

myr1e.MB

Genomic RNA incorporation into virus particles. In our earlier work, we found that Myr1E, which contains only monomers, had a mild decrease in genome incorporation, at the level of two- to fourfold when measured by Northern slot blot analysis (27). To determine how efficiently genomic RNA was packaged for each of the MA mutants, RPA was performed as a sensitive and quantitative method. The amount of RNA used for each mutant was normalized according to RT activity measured from virus particles prior to RNA extraction (as described in Materials and Methods). As shown in Fig. 6A, serial twofold dilutions of each genomic RNA sample could be detected and the assay was performed in the linear range. Wild-type virus, Myr1E.cc, and Myr1E.ATG, all of which contain dimeric RNA, packaged similar amounts of RNA per virus particle. In contrast, there was a mild decrease in RNA packaging for Myr1E.cc. When we performed RPA using a riboprobe designed to differentiate between spliced and unspliced viral RNAs, we found that the intracellular ratio of unspliced to spliced RNA was about the same for Myr1E, Myr1E, and wild type; furthermore, only unspliced genomic RNA was detected within the mutant and wild-type virus particles (data not shown).

View larger version (56K): [in this window] [in a new window]   FIG. 6.   Levels of viral RNA packaging. (A) RPA. Following transfection of QT6 cells with proviral constructs, RNA was isolated from virus particles. The amount of RNA used for each mutant was normalized compared to the wild type (WT) using RT activity measured prior to RNA extraction. Twofold dilutions of RNA were used for each mutant, as indicated by the triangular box above each set of lanes. After hybridization of the RNA with a 550-nt 32P-radiolabeled antisense riboprobe and digestion of the unprotected fragment with RNase, the RNA was separated by electrophoresis and the amount of radioactivity present in the protected band was quantified using PhosphorImager analysis. (B) Relative packaging efficiencies. The results of replicate RPA experiments were obtained as described in panel A, and the mean for each mutant is shown (two to four independent experiments were performed for each RNA sample). The packaging efficiency for the wild type was assigned a value of 1.0, and the mutants are expressed as a ratio compared to the wild-type value.

Infectivity of the src-containing proviruses. To determine whether the infectivity of the new mutants corresponded to our previous observations for Myr1E and Myr1E, replication in avian cell culture was assessed. Proviral DNAs were transfected into QT6 cells and virus particles were collected at 1 day posttransfection and then once or twice weekly. At the end of the collection period, RT activity was measured for all of the samples. The mean RT activity from two independent experiments is shown in Fig. 7. The time points of particle collection varied somewhat between experiments, so similar time points were combined. On day 1 posttransfection, virus particles were released into the media for all of the constructs, as expected. The wild-type virus and Myr1E.ATG (both of which encode the wild-type Gag protein) quickly spread throughout the culture, and high levels of virus production continued for the duration of the experiment. RT activity for Myr1E and Myr1E.cc increased over time but did not reach full wild-type levels. Thus, Myr1E and Myr1E.cc are both infectious but have a mildly reduced ability to replicate, with Myr1E.cc being more severely affected. As anticipated, Myr1E and Myr1E.cc were not infectious and their levels of RT activity decreased to nearly background after the initial burst of particles produced posttransfection (compare with the mock-infected sample [Fig. 7]). Thus, the src nucleotide sequence in Myr1E.ATG did not impair infectivity and the differences in codon usage between Myr1E and Myr1E.cc had minor effects on their overall patterns of infectivity. The infectivity defects of Myr1E and Myr1E.cc are due to the change in MA, not the RNA sequence.

View larger version (15K): [in this window] [in a new window]   FIG. 7.   Infectivity in avian cells. QT6 cells were transfected with proviral constructs expressing the wild-type or mutant sequences and passaged every 3 to 4 days. Medium samples were collected once or twice weekly, and virus particles were concentrated by ultracentrifugation and stored at 80°C. At the end of the collection period, RT assays were performed in triplicate for each sample. The mean RT activity from two independent experiments is shown in counts per minute (cpm) on a logarithmic scale.

Localization of mutant and wild-type MA proteins. We wondered whether the mutant MA proteins might have different locations within the cell owing to their different membrane-targeting signals. If so, this difference might yield insight into the mechanism by which dimerization is impaired by the MA mutants. Wild-type and mutant MA-GFP fusion proteins were expressed in QT6 cells and examined by confocal microscopy. Representative images are shown in Fig. 8. Expression of GFP alone was diffuse throughout the cytoplasm and nucleus, without any predominant pattern (Fig. 8A). In contrast, the wild-type MA-GFP fusion protein was present in both compartments but there was an enhanced concentration within the nucleus (Fig. 8B). The same intracellular distribution was observed for Myr1E.ATG GFP and Myr1E.cc GFP (Fig. 8C and D, respectively). However, expression of Myr1E GFP and Myr1E.cc GFP proteins, both of which contain an active Src membrane binding domain, revealed that these proteins were very efficiently targeted to the plasma membrane (Fig. 8E and F, respectively). Thus, the mutants that contain dimeric RNA and are infectious localize to the nuclear and cytoplasmic compartments while noninfectious mutants having monomeric RNA are strongly targeted to the plasma membrane. These striking differences in subcellular localization suggest that targeting MA too efficiently to the plasma membrane interferes with the dimerization process.

View larger version (85K): [in this window] [in a new window]   FIG. 8.   Subcellular localization of wild-type and mutant MA-GFP fusion proteins. QT6 cells transfected with plasmid DNAs expressing wild-type or mutant MA-GFP fusion proteins were examined by fluorescent confocal microscopy 6 to 18 h posttransfection. Representative images are shown for each of the indicated proteins in panels A through F.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The involvement of trans-acting factors other than NC in retroviral RNA dimerization has not been previously described. In this study, we have characterized several mutants of the MA protein that are noninfectious and are unable to form stable dimeric RNA complexes within virus particles. Our data convincingly show that the defect in RNA dimer formation is not due to problems with RNA-RNA interactions arising from changes in nucleotide sequence but results from changes in the MA protein sequence itself. Although we have identified four distinct MA mutants with dimerization defects, the Myr1E mutant has been studied most thoroughly. This mutant has a functional Src membrane-targeting domain attached to its N terminus, leading to improper subcellular localization of the MA protein. Taken together, our data suggest that mistargeting of MA (in the case of Myr1E) results in the failure of viral RNA dimerization.

How might the MA protein sequence influence RNA dimer formation? We can envision several possible scenarios. If viral RNA molecules begin to dimerize early in assembly, then sending the assembling RNA-Gag complex along the wrong pathwayperhaps by way of an improper membrane-targeting signal, such as that of Srcmight prevent necessary RNA-RNA interactions from occurring. During the normal assembly process, the wild-type Gag protein moves along an intracellular pathway to a specific plasma membrane site that includes the necessary conditions for dimerization. However, since the mutant Myr1E Gag protein has altered membrane-targeting properties, it might direct the RNA-protein complex to the wrong cellular localization or to the wrong location on the membrane, thereby disrupting dimerization. In this model, any changes that perturb the normal trafficking properties of the MA sequence would be expected to affect genomic RNA dimerization as well.

Alternatively, if dimerization does not occur until after budding in RSV, it is possible that the MA mutants influence the dimerization process within the virus particle. In this case, the Src membrane binding domain of Myr1E might associate too tightly with the viral membrane. This distortion of the usual MA-membrane interaction could disrupt interactions between MA and other structural proteins (CA or NC, for example), ultimately affecting the structure of the virion core and the integrity of the RNA located within it. These structural changes could either prevent dimerization from occurring or allow only very unstable dimers to form, perhaps by interfering with dimer maturation (14). This model does not require any direct contact between the MA protein and the viral genomic RNA, although how MA might interact with components of the core is unknown for RSV. It is important to keep in mind that RSV MA does have RNA binding activity (31), and so a role in promoting dimer formation or stability within the virus particle remains a possibility.

A third possibility is that the MA protein serves to enhance dimer formation in an indirect way. For example, MA might promote the stability of newly synthesized genomic RNA, allowing dimerization to be more efficient. MA could target the Gag-RNA complex to an environment within the cell that protects the RNA from degradation. Recently it was reported that HIV-1 MA appears to play a role in enhancing RNA stability (18), and perhaps RSV MA plays a similar role. Alternatively, MA might interact with another cellular or viral factor (e.g., NC), which in turn promotes RNA dimerization. Additionally, it has been proposed that HIV-1 MA might affect RNA transport out of the nucleus, based on studies of a mutant displaying mislocalization of its viral RNA (8). This MA mutant also has a defect in genomic RNA dimerization, suggesting to us that there may be a link between RNA localization and dimerization. Thus, a potential role for MA in viral RNA dimerization and/or transport of the Gag-RNA complex might be a common feature among retroviruses.

We also found that the Myr1E mutants that contain monomeric RNA package about half as much unspliced viral RNA per particle as wild-type viruses do (Fig. 6). It is feasible that monomers are efficiently packaged into these virus particles. Therefore, the dimer linkage structure might not be required for RNA packaging, at least in RSV. In support of this idea, others have shown that rapid-harvest and immature viral RNA from RSV is largely monomeric (2, 3, 25, 32). However, we still must consider the possibility that some Myr1E particles contain unstable dimers while others contain no viral RNA at all. Future experiments will address this question.

Other questions that remain regarding viral RNA dimerization include when dimerization occurs, early or late in the assembly process. Also, it is not known if passage through a specific cellular compartment or a specific membrane site is required for dimerization in vivo. While we now have evidence that both the MA and NC proteins of RSV have significant effects on RNA dimerization, the molecular mechanisms of their actions need to be further investigated. Finally, it remains to be shown whether additional viral or cellular factors participate in the dimerization process.