Competition between the Sendai Virus N mRNA Start Site and the Genome 3'-End Promoter for Viral RNA Polymerase

JOURNAL OF VIROLOGY, Sept. 2003, p. 9147–9155 0022-538X/03/$08.00⫹0 DOI: 10.1128/JVI.77.17.9147–9155.2003 Copyright © 2003, American Society for Microbiology. All Rights Reserved. Vol. 77, No. 17 Competition between the Sendai Virus N mRNA Start Site and the Genome 3⬘-End Promoter for Viral RNA Polymerase Philippe Le Mercier, Dominique Garcin, Eduardo Garcia, and Daniel Kolakofsky* Department of Genetics and Microbiology, University of Geneva School of Medicine, CH1211 Geneva, Switzerland Received 3 March 2003/Accepted 30 May 2003 The genomic and antigenomic 3ⴕ-end replication promoters of Sendai virus are bipartite in nature and symmetrical, composed of le or tr sequences; a gene start or gene end site, respectively; and a simple hexameric repeat. The relative strengths of these 3ⴕ-end promoters determines the ratios of genomes and antigenomes formed during infection and whether model mini-genomes can be rescued from DNA by nondefective helper viruses. Using these tests of promoter strength, we have confirmed that tr is stronger than le in this respect. We have also found that the presence of a gene start site within either 3ⴕ-end promoter strongly reduces 3ⴕ-end promoter strength. The negative effects of the gene start site on the 3ⴕ-end promoter suggest that these closely spaced RNA start sites compete with each other for a common pool of viral RNA polymerase. The manner in which this competition could occur for polymerase off the template (in trans) and polymerase on the template (in cis) adds insight into how the viral RNA polymerase switches between its dual functions as transcriptase and replicase. Sendai virus (SeV), a Respirovirus of the Paramyxovirinae subfamily, is a model nonsegmented negative-strand ([⫺]) RNA virus. The first step in the SeV replication cycle is the production of mRNA from a helical subviral structure, the nucleocapsid (NC), in which the genome RNA is tightly and stoichiometrically associated with the viral NC protein (N protein). This N-subunit assembly, together with multiple attached viral polymerases (a complex of the P and L proteins) is the minimum subviral unit that is thought to retain infectivity (16). The synthesis of [⫺] RNA genomes (or [⫹] antigenomes) and their assembly with N protein is coupled, and these viral RNAs are only found as nucleocapsids (10). Electron micrographs of SeV nucleocapsids show a flexible helical assembly with 13 N subunits per turn and variable pitch, in which each N subunit binds 6 nucleotides (nt) (5) (Fig. 1). For viruses of the Paramyxovirinae, efficient replication of model mini-genomes in transfected cells requires that their total length be a multiple of six, and viruses of this group whose genome has been entirely sequenced mostly have genome lengths that are multiples of six (2, 15). Inefficient replication of non-hexamer-length mini-genomes was not due to the lack of encapsidation of the mini-genome but apparently was due to the inability of viral RNA-dependent RNA polymerase (vRdRP) to initiate at the mini-nucleocapsid 3⬘ end (2). It has been suggested that nucleocapsid assembly begins with the first 6 nt at the 5⬘ end of the nascent chain and continues by assembling 6 nt at a time until the 3⬘ end is reached. The efficiency of the 3⬘-end promoter then presumably depends on the position of these cisacting sequences relative to the N subunits, and this hexamer or N-subunit “phase” is determined by the precise length of the genome chain (15, 26). The requirement for hexameric ge- nome length of the Paramyxovirinae has recently been underscored for human parainfluenza virus type 2 (hPIV2), a rubulavirus, and measles virus, a morbillivirus. hPIV2 recovered from non-hexameric-length cDNAs were found to contain a biased distribution of mutations that all restored hexameric genome length (15, 22). Similarly, an obligatorily diploid measles virus was recovered from DNA that encodes a defective H protein, again via mutations that restore hexameric genome length (21). The genomic and antigenomic promoters (G/Pr and AG/Pr) of the Paramyxovirinae are found within the terminal 96 nt or 16 N subunits of each RNA and are composed of leader (le) or trailer (tr) sequences; a gene start or gene end site, respectively; and a simple hexameric repeat (Fig. 1; see also Fig. 5). Although G/Pr and AG/Pr are similar in overall structure, they carry out separate functions. G/Pr is associated with le RNA, antigenome RNA, and mRNA synthesis, while AG/Pr is associated with tr RNA and genome RNA synthesis. The relative abundance of the various viral RNAs throughout the infection is regulated in large part via G/Pr and AG/Pr, and this ensures as well that a preponderance of [⫺] genomes are produced during infection. These 3⬘-end promoters are (at least) bipartite in nature (11, 18, 20, 24). There is both an end element comprising ca. the first 30 nt at the 3⬘ ends (in which the first 12 nt are conserved between G/Pr and AG/Pr and across each genus), and a downstream element within the 5⬘ untranslated region of the N gene ([⫺] nucleocapsids) or the 3⬘ untranslated region of the L gene ([⫹] nucleocapsids). For SeV and hPIV3 and apparently all respiro- and morbilliviruses, the downstream element is a simple but phased hexameric repeat (3⬘ [C1N2N3N4N5N6]3, bound to the 14th, 15th, and 16th N protein subunits [11, 24]). For rubulaviruses like SV5, [N1N2N3N4G5C6]3 is repeated in subunits 13, 14, and 15, such that all these conserved nucleotides are adjacent to the first 12 nt in the helical nucleocapsid with 13 subunits per turn (19) (Fig. 1). This common surface, comprising the conserved sequences of * Corresponding author. Mailing address: Department of Genetics and Microbiology, University of Geneva School of Medicine, CMU, 9 Ave. de Champel, CH1211 Geneva, Switzerland. Phone: 41 22 379 5657. Fax: 41 22 379 5702. E-mail: Daniel.Kolakofsky@medecine .unige.ch. 9147 9148 LE MERCIER ET AL. J. VIROL. FIG. 1. Genome 3⬘-end promoter (G/Pr) in resting nucleocapsids. A model of the helical N-subunit assembly of the SeV resting genome nucleocapsid is shown, based on the electron micrographic reconstruction of Egelman et al. (5), with a pitch of 5.3 nm and 13 subunits per turn. The position of the genome RNA within this structure is unknown; the RNA line is placed on the outside surface of the N assembly for clarity and because it is just long enough when extended to occupy this position. The N subunits (and the 6 nt positions within each subunit) are numbered from the genome 3⬘ end (lower right). An enlargement of the bipartite 3⬘-end promoter (dark subunits on the left) is shown on the right, where the 3⬘-terminal 96 nt are placed within each subunit in groups of six. The 3⬘-terminal 12 nt and the [C1N2N3N4N5N6]3 repeat conserved in all genomes and antigenomes of respiro-and morbilliviruses are highlighted. The position of the N gene start site (gs-1) is also indicated. both elements of the bipartite promoter on two turns of the helix, is thought to mediate the initial interaction of vRdRP with resting nucleocapsids to initiate RNA synthesis at the genome 3⬘ end (see below). The remarkable sequence simplicity of the downstream promoter element is presumably possible because its hexamer phase also contributes to its activity (19, 24). This phase effect is thought to stem from the different chemical environments of the 6 nt bases associated with each N subunit, as revealed by chemical attack studies of resting SeV nucleocapsids (12). Adenosines in any hexamer phase are largely protected from dimethyl sulfate, whereas cytosine reactivity is highly variable and strongly depends on hexamer phase. Cytosines are highly reactive only in hexamer positions 1 and 6, precisely the positions of the conserved cytosines in the downstream element of G/Pr and AG/Pr (Fig. 1). If this correlation is not coincidental, this suggests that the conservation of hexamer phase of this cis-acting sequence serves to make these cytosines accessible to vRdRP in resting NCs. Another feature strictly conserved among respiro-, morbilli-, and rubulaviruses is that the first (N) mRNA starts precisely opposite U56 (underlined) within the decameric gene start signal 3⬘ 55A/UCCCA NUUUC/N66 (for SeV, hexamer phase indicated). We refer to this mRNA initiation signal as gene start 1 (gs-1). Curiously, none of the conserved cytosines of gs-1 are in hexamer positions 1 and 6. However, once vRdRP has initiated RNA synthesis at the genome 3⬘ end and begun to elongate, we assume that conformational changes occur within N:RNA that alter vRdRP’s interaction with the nucleotide bases, including gs-1. Since SeV RdRP is minimally a coiled-coil tetramer of P (4, 25) and a single L protein (with an aggregate molecular mass equivalent to 8 N-subunits) and this polymerase initiates RNA chains at two such closely spaced sites at the genome 3⬘ end (only 56 nt apart), one would expect these two events to interfere with each other, at least under some conditions. However, detailed studies involving modified DI mini-genomes did not uncover any indication that gs-1 affected G/Pr strength; only the first 30 nt of le/tr regions were found to influence this property. We now report studies that show that in some cases exactly the same situation applies, namely, that only the le/tr sequences appear to influence promoter strength. More importantly, we find other cases in which gs-1 clearly decreases 3⬘-end promoter strength. The manner in which gs-1 negatively influences G/Pr strength is discussed. MATERIALS AND METHODS Generation of recombinant SeV (rSeV). Briefly, one 9-cm-diameter dish of BSR-T7 cells that endogenously express T7 RNA polymerase (1) were trans- VOL. 77, 2003 PROMOTER COMPETITION FOR SENDAI VIRUS RNA POLYMERASE fected with 1.5 ␮g of pGEM-L, 5 ␮g of pGEM-N, 5 ␮g of pGEM-PHA (which does not express C proteins), and 15 ␮g of the various pSeV infectious constructs. After 24 h the cells were scraped into their medium and injected directly into the allantoic cavity of 9-day-old embryonated chicken eggs. Three days later, the allantoic fluids were harvested and reinjected undiluted into eggs. For further passages, the viruses were diluted 1/500 before injection. The presence of viruses was determined by pelleting allantoic fluids through a TNE (10 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1 mM EDTA)–25% glycerol cushion for 20 min at 14,000 rpm in an Eppendorf 5417C centrifuge. Virus pellets were resuspended in sample buffer, and the proteins were separated by sodium dodecyl sulfate–10% polyacrylamide gel electrophoresis and stained with Coomassie brilliant blue. Virus infection and passage. A549 or BSR T7 cells were infected at a multiplicity of infection (MOI) of 10 in Dulbecco’s modified Eagle medium. After 1 to 2 h of absorption, the inoculum was removed and replaced with fresh medium containing 10% fetal bovine serum. Virus supernatant from 48-h-infected cells was cleared by filtration, treated for 1 h with trypsin (1.2 ␮g/ml), and used to reinfect fresh A549 or BSR T7 cells. Real-time PCR. A 9-cm-diameter dish of A549 cells was infected with SeV at an MOI of 10. After 24 h, medium was removed, the cells were scraped, and total RNA was extracted using Trizol. One-tenth of this RNA was mixed with 0.5 ␮g of specific primer and converted to cDNA with Moloney murine leukemia virus reverse transcriptase for 1 h at 37°C. One-tenth of the cDNA was used for real-time PCR, using TaqMan Universal Master mix and the ABI Prism 7700 sequence detector. Several dilutions of cDNA were tested to construct a standard curve that was used to obtain relative numbers for the samples. DI particle rescue with helper virus. Dishes of BSR-T7 cells (diameter, 9 cm) were infected with either wild-type SeV (SeV-wt) or SeV-AGP55 at an MOI of 10. One hour later, the medium was removed and the cells were transfected with 5 ␮g of the DI-encoding plasmid. Twelve hours later the medium was replaced with fresh Dulbecco’s modified Eagle medium containing 10% fetal bovine serum. The ND and DI particles present in the culture medium were then passaged on fresh cells as described above. DI particle rescue with plasmids. Dishes of BSR-T7 cells (diameter, 9 cm) were transfected with 1.5 ␮g of pGEM-L, 5 ␮g of pGEM-N, 5 ␮g of pGEM-PHA (which does not express C proteins), and 5 ␮g of the various DI-encoding plasmids. RESULTS The ratio of genomes to antigenomes was previously determined by (32)P-primer extension analysis. This ratio can also be determined by real-time reverse transcriptase PCR (RTPCR) analysis, as follows (Fig. 2A) (and as described in Materials and Methods). Four pairs of primers that span gene junctions are used; one primer of each pair specifically extends on genomes and lies upstream of the junction, and the other specifically extends on antigenomes and lies downstream of the junction. The four genome-specific and four antigenome-specific primers are used in individual reactions to synthesize cDNA with RT. Samples of the RT reaction mixture are then used in real-time PCR amplification along with the reverse primer located across the junction. This approach eliminates the contribution of mRNAs to the amount of antigenomes determined (only discistronic mRNAs that are very rare in SeV infections will be scored). At least three sets of primers (and fluorescent probes) were used for each determination, and the average genome/antigenome ratio from at least two infections of each SeV is shown. The ratios determined by RT-PCR agree well with those determined by primer extension analysis (in parentheses in Fig. 2), and require considerably less viral RNA as starting material. Relative to SeV-wt infections, where ca. 10 times as many genomes as antigenomes are found intracellularly, SeVAGP55 infections (in which the entire 55 nt of tr is replaced with the le sequence) contain equal levels of the two (17) (Fig. 2). (AGP refers to a virus or DI particle. When “AGP” is 9149 followed by a single number, as in AGP55, this indicates that positions 1 to 55 of AG/Pr have been exchanged with le sequences.) SeV-AGP65 infections, which contain gs-1 in addition to the le sequence at AG/Pr, yielded similar results (Fig. 2). The same genome-to-antigenome ratios were found in virions (data not shown), since, similar to rabies virus infections (7), SeV genomes are not being selectively exported during virion assembly (14). Genome/antigenome ratios intracellularly should then simply reflect the frequencies with which vRdRP productively initiates at the template 3⬘ ends and begins to encapsidate nascent le/tr RNAs, to produce genome or antigenome nucleocapsids. The efficiency of this process is one measure of promoter strength. By this criterion, the replacement of the tr region of AG/Pr with the le sequence alone (SeV-AGP55) strongly decreases AG/Pr strength. Since SeV-AGP55 replicates to levels similar to those of SeV-wt, it appears that G/Pr is inherently weak only in the presence of the stronger AG/Pr. Thus, as described for rabies virus infections (7), competition between G/Pr and AG/Pr for vRdRP appears to be the primary determinant of the relative amounts of SeV genomes and antigenomes formed (Fig. 1B) The genome/antigenome ratio of 1 in AGP55/65 infections suggests that promoter strength is primarily determined by tr versus le sequences themselves, in agreement with previous studies of DI genome replication (3). However, it is also possible to construct SeV in which the first 42 or 48 nt of the le sequence of G/Pr is replaced with the equivalent tr sequences (SeV-GP42/48) without seriously affecting viral mRNA synthesis (8, 9). SeV-GP42 and SeV-GP48 are notable in that their infections fail to induce apoptosis. Nevertheless, SeV-GP42/48 infections still accumulate 10 times as many genomes as antigenomes intracellularly, like SeV-wt (only SeV-GP42 is shown in Fig. 2). This result was unexpected because, if promoter strength is determined primarily by the nature of first 30 nt of le/tr sequence, the modified G/Pr of SeV-GP42/48 should have been equivalent in strength to AG/Pr. Possible explanations for why G/Pr of SeV-GP42/48 remains weaker than AG/Pr include (i) the remainder of the le region (positions 49 to 55) is specifically important for G/Pr strength in the context of nondefective infections and (ii) in some situations, the presence of gs-1 within G/Pr does indeed reduce G/Pr strength, despite the presence of 48 nt of tr sequence. To directly examine this question, a matched series of rSeV isolates were prepared that bridged the transition from SeV-wt to SeV-AGP65 (Fig. 2). Except for the ambisense SeV (i.e., SeV that expresses ambisense mRNA) that grow less well than SeV-wt, all the other SeV grow similarly to SeV-wt; i.e., they all accumulate similar levels of viral macromolecules intracellularly. SeV-wt⫹MCS contains a cassette at the L gene end/tr junction with three elements, a poly(A)-stop site that terminates mRNAs transcribed from the antigenome 3⬘ end, a multicloning site (MCS), and a duplicated L gene end site. SeVwt⫹MCS infections produce genome/antigenome ratios of 10 like SeV-wt; thus, the inclusion of this cassette appears not to have affected relative promoter strengths. SeV-AGP56-65 is identical to SeV-wt⫹MCS except that the 10 nt adjacent to tr (i.e., the L gene end site) are converted to the complement of gs-1. Remarkably. the simple inclusion of these 10 nt (gs-1) in AG/Pr of AGP56-65 leads to transcription of a short (80 nt) ambisense mRNA from this antigenome that terminates just 9150 LE MERCIER ET AL. J. VIROL. FIG. 2. 3⬘-end promoter mutations of nondefective SeV and the ratio of genomes to antigenomes formed during infection. (A) The SeV genomes are shown as a series of boxes (not drawn to scale). (B) The 96-nt genome 3⬘-end promoter (on the left) and antigenome 3⬘-end promoter (right) are shown as narrow boxes. The remainder of the genome (from the N gene AUG codon to the UAA that terminates the L open reading frame) is shown as wider boxes. The genome and antigenome 3⬘-end promoters, G/Pr and AG/Pr, are composed, respectively, of 55 nt of le (white box) or tr sequence (black box), gs-1 (white circle with forward arrow) or the complement of the L gene end signal (black circle with stopped arrow), a short spacer sequence, and the [C1N2N3N4N5N6]3 repeat or its complement (BB box). The cassettes added at the L gene end/tr junction to control ambisense mRNA expression are marked with a dotted line. White and grey indicate the various G/Pr and AG/Pr sequences present at each 3⬘-end promoter. In naming the viruses, GP and its associated numbers refer to those nucleotides of G/Pr that have been replaced with the equivalent sequences of AG/Pr, or vice versa for AGP viruses. Total RNA from cytoplasmic extracts of A549 cells infected for 24 h with 20 PFU/cell of the various ND-SeV was used to estimate the relative amounts of genomes and antigenomes by real-time RT-PCR. Four sets of primers and probes that spanned different gene junctions (A) were used (to minimize the contribution of mRNA to the antigenome determination [see Materials and Methods]). At least three sets of primers and probes were used for each determination, and the average G/AG ratio from at least two infections of each SeV is shown. The numbers in parentheses show the ratios determined by 32P-labeled primer extension. before the opposing L gene. SeV-AGP56-65 infections, however, now accumulate almost as many antigenomes as genomes. Thus, gs-1 placed within AG/Pr not only functions well but also apparently acts to equalize the two promoters, presumably by weakening AG/Pr. SeV-AGP52-65 and AGP48-65 differ from AGP56-65 by the further conversion of the 4 and 8 nt adjacent to the ambisense gene start site. The exchange of these nucleotides further increases the relative amounts of antigenomes, but only slightly. Thus, AG/Pr can be weakened relative to G/Pr in one of two ways: by exchanging the resident tr sequences for le sequences or by introducing an ectopic gs-1 within this 3⬘-end promoter. The introduction of gs-1 within AG/Pr automatically destroys the L gene end site. It thus remains possible that the destruction of this sequence is responsible (at least in part) for weakening AG/Pr. We therefore attempted to prepare derivatives of wt⫹MCS in which the L gene end site is mutated to something other than the complement of gs-1. The complement of the L gene end site (5⬘ UAAGAAAAA) was mutated to 5⬘ UGCCGCAUG and UUUCUUUUU. However, we were VOL. 77, 2003 PROMOTER COMPETITION FOR SENDAI VIRUS RNA POLYMERASE unable to prepare these rSeV in three separate attempts, and we sought another manner to examine this question. The mutated L gene end sites were built into mini-genomes expressing green fluorescent protein (GFP) (Fig. 3), and these were recovered from DNA by coinfection with SeV-AGP55 (which contains two weak replication promoters [see below]). We were again unable to recover the 5⬘ UGCCGCAUG minigenome (DI-Ge-1; replicable genomes may not tolerate certain sequences here). However, DI-Ge-3 (5⬘ UUUCUUUUU) could be recovered almost as well as the wt. When RNA from these mixed infections were examined for their ratio of DI genomes to antigenomes, DI-Ge-3 and DI-wt were found to contain the same excess of genomes over antigenomes (approximately threefold in this case; the ratio of DI genomes and antigenomes are not necessarily the same as that of ND genomes and antigenomes). Thus, the loss of the complement of the L gene end site per se does not alter AG/Pr strength. We conclude that it is the presence gs-1 within AG/Pr that is responsible for weakening this 3⬘-end promoter. Competition between DI genomes and ND helper genomes for vRdRP. The relative amounts of genomes and antigenomes formed during infection is one measure of the relative strengths of G/Pr and AG/Pr of nondefective SeV. This manner of characterizing promoters stems from early studies of copy back DI genomes, which have deleted most of the genome and simultaneously exchanged ⬎96 nt of G/Pr with the equivalent sequences of AG/Pr (13). Copy back DI genomes do not transcribe mRNAs and have a clear replicative advantage over ND virus. Moreover, in contrast to their ND helper virus, these coinfections accumulate equal amounts of copy back DI genomes and antigenomes. Further studies suggested that their replicative advantage did not stem from their shorter length, and that relative promoter strength is determined by the terminal 96 nt of G/Pr and AG/Pr. Internal-deletion (int-⌬) DIs that maintain the wt promoter constellation can also be generated on high-MOI passage of SeV stocks. However, these DIs are considerably more difficult to generate, and the reasons for their replicative advantage over their ND helper virus are unknown. In further contrast to copy back DIs, int-⌬ DI genomes cannot be recovered from DNA using SeV-wt as helper, even though they are efficiently amplified when the replication substrates (N, P, and L) are provided from plasmids. More importantly, the same int-⌬ DIs are efficiently recovered from DNA using SeV-AGP55 or AGP65 as helper virus (Fig. 3 and 4). Thus, although int-⌬ DIs presumably have no advantage in competing with wt helper virus for limiting RdRP during recovery from DNA, int-⌬ DIs can clearly compete with ND helpers whose AG/Pr is weakened (as defined by ND genome/antigenome ratios). We have adapted this system to study promoter strength by including a GFP transcription unit within int-⌬ DIs. BSR T7 cells infected with various helper viruses are transfected with plasmids expressing various mini-antigenomes (see time line in Fig. 3). The particles released 36 to 48 h postinfection are then used to infect fresh cultures that are examined by fluorescent microscopy for GFP expression. This simple test reveals whether the DI particles have been generated from plasmid DNA during the initial infection/transfection, and amplified by the helper virus during subsequent coinfection (P1) (Fig. 3). Using this DI reporter system, we have reexamined whether 9151 gs-1 within G/Pr affects the ability of G/Pr to compete for vRdRP during coinfection. A series of mini-genome constructs that varied only in their G/Pr was used to determine what changes were required so that the int-⌬ DI could be recovered and amplified by SeV-wt helper (5 top constructs, Fig. 4). We first exchanged the first 42 or 55 nt of le of G/Pr with the equivalent tr sequences (GP42/wt and GP55/wt), but only background fluorescence was found at passage 1. We then added a further 12 nt of the trailer sequence but maintained the remainder of DI GP55/wt, so that the GFP expression was retained and now started at nt 68 rather than nt 56 (DI GP68⫹start/wt). However, this DI as well could not be recovered (in six attempts) with SeV-wt. Only when the mRNA start site was eliminated from G/Pr of DI GP55/wt, by exchanging it for the equivalent AG/Pr sequence (DI GP96/wt), was this DI recovered with SeV-wt. (The elimination of gs-1 of course also eliminates GFP expression, and DI RNA levels were determined by real-time RT-PCR [Fig. 4, bottom].) Thus, int-⌬ DI genomes successfully compete for vRdRP with wt helper genomes only when the mRNA start site within G/Pr has been eliminated; simply moving gs-1 downstream by 12 nt (DI 68⫹start/wt) has no effect. Similar experiments were carried out with DI wt/AGP55, which contains le in place of tr sequences at AG/Pr (bottom three constructs in Fig. 4). Just as DI-wt/wt cannot be rescued by coinfection with SeV-wt, DI wt/AGP55 cannot be rescued by coinfection with SeV-AGP55 (presumably because both G/Pr and AG/Pr of the DI and ND-helper virus are again equivalent). When the first 42 nt of G/Pr are exchanged for the equivalent tr sequences, this DI GP42/AGP55 cannot be rescued by SeV-AGP55. However, DI GP96/AGP55, in which the first 96 nt are trailer sequence and gs-1 is eliminated, can be rescued with SeV-AGP55. Taken together, these results indicate that gs-1 within G/Pr clearly decreases the ability of this 3⬘-end promoter to compete for vRdRP during recovery from DNA with helper SeV and subsequent coinfection. DISCUSSION The 96-nt G/Pr contains 55 nt of le sequence (light color, weak), gs-1 (oval with arrow) and a simple hexamer repeat (BB box), whereas the symmetrical AG/Pr contains tr sequence (dark color, strong), the L gene end site (empty oval), and a BB box (Fig. 5). AG/Pr is a stronger replication promoter than G/Pr, as defined by the preponderance of [⫺] genomes formed during virus replication. In these and other experiments, the two BB boxes appear to be interchangeable without effect (data not shown). The major finding of this work is that gs-1 within 3⬘-end promoters negatively influences promoter strength. Our results are summarized schematically in Fig. 5. G/Pr contains two elements that weaken this 3⬘-end promoter relative to AG/Pr; the presence of the le as opposed to the tr sequences, and gs-1. Simply replacing the tr sequence of AG/Pr with le sequence reduces AG/Pr strength such that SeV-AGP55 infections now accumulate equal amounts of genomes and antigenomes (Fig. 2). Similarly, introducing gs-1 into AG/Pr of SeV-wt⫹MCS (SeV-AGP56-65) reduces the genome/antigenome ratio from 12 to 2.5 (Fig. 2B). Thus, the stronger AG/Pr apparently can be weakened in two different ways. G/Pr has been examined via DI mini-genomes as well as 9152 LE MERCIER ET AL. J. VIROL. FIG. 3. Defective-interfering mini-genomes expressing GFP. (A) DI-wt, in which the entire viral coding sequences (from the N gene AUG to the UAA of the L open reading frame) were replaced with the GFP open reading frame, is shown schematically as for Fig. 2. The sequences present at L gene end site of DI-wt, DI-Ge-1 and DI-Ge-3 (see text) are indicated. (B) A time line of DI particle recovery from DNA is shown (see text for details). Green fluorescent photos of the first passage cultures (P1) that contained equal densities of cells are shown. To measure viral RNA levels, two additional virus passages were carried out to minimize spill-over from biased genome/antigenome ratios during DI recovery from DNA. Genome/antigenome levels were then determined by real-time RT-PCR with the two primers pairs and probe as indicated in panel A. Transfection of DI-wt without helper virus (first photo in panel A), and DI-Ge-1 transfection with helper virus (third photo in panel A) serve as negative controls that show the amplification specificity for the DI RNAs. Error bars, standard deviations. VOL. 77, 2003 PROMOTER COMPETITION FOR SENDAI VIRUS RNA POLYMERASE 9153 FIG. 4. DI particle recovery from DNA with various nondefective helper SeV. The various DI constructs are shown schematically and are named as for Fig. 2 and 3. For those constructs with a functional gs-1 driving GFP expression, recovery was scored (⫹ or ⫺) by fluorescent microscopy at P1 (Fig. 3). For those constructs without functional gs-1 (marked with asterisks), recovery was scored by determining DI genome levels by real-time RT-PCR and is plotted in the bar graph below. Error bars, standard deviations. ND-SeV, because gs-1 is not essential for DI-genome replication. Exchanging the first 42 or 48 nt of the le region of G/Pr of ND viruses with the equivalent tr sequences (SeV-GP42/48) does not diminish the genome/antigenome ratio of 10, nor does the analogous exchange of the first 55 nt of G/Pr improve the ability of int-⌬ DI genomes to be rescued by SeV-wt (Fig. 4). Thus, simply exchanging le for tr sequences apparently cannot overcome the negative effects of gs-1 within G/Pr. The simultaneous elimination of the mRNA start site within G/Pr of int-⌬ DI genomes (DI GP98/wt and DI GP98/AGP55) was 9154 LE MERCIER ET AL. J. VIROL. FIG. 5. A model for promoter competition in cis for vRdRP. The various 96-nt G/Pr and AG/Pr are shown schematically as for the other figures, with the weak le sequences shown by light shading and the strong tr sequences shown by dark shading. The first 12 nt of each promoter are conserved and are shown as a separate shading. The larger oval near the end of the le/tr regions represents vRdRP that has initiated at the 3⬘ end, has cleared this promoter, and has just released the le/tr RNAs. If vRdRP remains attached to the N:RNA without a nascent chain, it is free to scan the template (in either direction) for another RNA start site, which can be either the 3⬘-end promoter or gs-1 (horizontal arrows below oval). The relative strength of the 3⬘-end promoter, i.e., its ability to compete for limiting vRdRP during infection relative to other 3⬘-end promoters, is indicated by the size of bent arrow at the 3⬘ ends. Relative 3⬘-end promoter strength is conditioned both by the presence of the le/tr sequences and that of gs-1 (circle with bent arrow). found to be essential for their rescue by SeV-wt and SeVAGP55, respectively (Fig. 4). Thus, the ability of DI genomes to compete with helper virus genomes for vRdRP is dominated by the presence of gs-1 within the DI genome. We note that there are also examples at AG/Pr where the converse situations applies, namely, where the presence of le as opposed to the tr sequences predominates, and effects due to the presence of gs-1 are not detected (e.g., SeV-AGP55 and SeV-AGP65 rescue int-⌬ DI genomes with equal efficiency). Almost all our DI constructs were efficiently amplified when their replication substrates were provided via plasmids (in the absence of other competing SeV templates) (Fig. 4). The inability of some constructs to be rescued with SeV-wt, coupled with their efficient rescue with SeV-AGP55/65, suggests that DI rescue depends on the ability of the DI 3⬘-end promoters to compete with the helper genomes for vRdRP. It is from these experiments that the conclusion that gs-1 negatively affects 3⬘-end promoter strength comes through most clearly. On what basis, though, does gs-1 within G/Pr diminish the ability of DI genomes to compete with ND genomes for replication substrates? Simple competition of resting ND and DI N:RNAs for a common pool of vRdRP (off the template) may be part of the explanation. In this case, the competition implies that the eventual transcriptases and replicases cannot be distinguished off the template. However, it is also possible that this competition occurs in cis, once vRdRP has engaged G/Pr, as this might offer an explanation for why the presence of gs-1 within G/Pr is so deleterious for 3⬘-end promoter strength. A model for this in-cis competition is shown in Fig. 5. vRdRP in this cartoon (elongated oval) has already initiated at the genome 3⬘ end and cleared the promoter. In those cases where RNA synthesis and assembly with N become coupled, vRdRP is committed to replication and this P4-L does not reenter the pool until it has completed N:RNA synthesis. G/Pr strength, or its relative abil- ity to initiate RNAs that are subsequently coassembled with N, is presumably conditioned by the presence of the le versus tr sequences within G/Pr. These sequences as [⫺] genome would promote vRdRP initiation, and as nascent [⫹] le RNA would promote N assembly. When RNA synthesis and its assembly with N do not become coupled, vRdRP releases the nascent le chain at or near gs-1. If vRdRP is simultaneously released from the template along with the le RNA, this P4-L rejoins the vRdRP pool off the template and is free to interact with all the available 3⬘-end promoters in the coinfected cell (Fig. 1B). However, if vRdRP is not released from the template along with le RNA, it may then be free to scan the template (in both directions) for the nearby gs-1 as well as the genome 3⬘ end, similar to vRdRP that has released its mRNA at gene junctions (6, 23). Scanning confers an advantage to this 3⬘-end promoter, precisely because vRdRP is restricted to reinitiating on this template. The presence of gs-1 within G/Pr weakens this replication promoter by diverting these scanning vRdRP from reinitiating at the genome 3⬘ end, thus decreasing its relative replication promoter strength. A corollary of this competition-in-cis model is that relative initiation from gs-1 will similarly depend on competition with the 3⬘-end promoter; i.e., gs-1 will be initiated more frequently when present within or near G/Pr (near weaker le sequence) than within or near AG/Pr (near stronger tr sequence). This aspect has recently been tested with DI mini-genomes containing tandem 96-nt promoters in which gs-1 of the external promoter was corrupted and the replicative ability of the internal promoter was disabled, such that gs-1 of the internal promoter was now found at position 146, i.e., 50 nt downstream of the external 96-nt replication promoter. In this location, gs-1 was used 15 times more frequently when downstream of a weak G/Pr than downstream of a strong AG/Pr (D. Vulliemoz, P. Le Mercier, and L. Roux, personal communication). Again, the obvious conclusion is that gs-1 and nearby replication promot- VOL. 77, 2003 PROMOTER COMPETITION FOR SENDAI VIRUS RNA POLYMERASE ers compete with each other, presumably in cis, for a common pool of vRdRP in coinfected cells. 13. ACKNOWLEDGMENTS We thank Laurent Roux for numerous discussions. This work was supported by a grant from the Swiss National Science Fund. REFERENCES 1. Buchholz, U. J., S. Finke, and K. K. Conzelmann. 1999. Generation of bovine respiratory syncytial virus (BRSV) from cDNA: BRSV NS2 is not essential for virus replication in tissue culture, and the human RSV leader region acts as a functional BRSV genome promoter. J. Virol. 73:251–259. 2. Calain, P., and L. Roux. 1993. The rule of six, a basic feature for efficient replication of Sendai virus defective interfering RNA. J. Virol. 67:4822–4830. 3. Calain, P., and L. Roux. 1995. Functional characterisation of the genomic and antigenomic promoter of Sendai virus. Virology 211:163–173. 4. Curran, J., R. Boeck, N. Lin-Marq, A. Lupas, and D. Kolakofsky. 1995. Paramyxovirus phosphoproteins form homotrimers as determined by an epitope dilution assay, via predicted coiled coils. Virology 214:139–149. 5. Egelman, E. H., S.-S. Wu, M. Amrein, A. Portner, and G. Murti. 1989. The Sendai virus nucleocapsid exists in at least four different helical states. J. Virol. 63:2233–2243. 6. Fearns, R., and P. L. Collins. 1999. Model for polymerase access to the overlapped L gene of respiratory syncytial virus. J. Virol. 73:388–397. 7. Finke, S., and K.-K. Conzelmann. 1997. Ambisense gene expression from recombinant rabies virus: random packaging of positive- and negative-strand ribonucleoprotein complexes into rabies virions. J. Virol. 71:7281–7288. 8. Garcin, D., T. Pelet, P. Calain, L. Roux, J. Curran, and D. Kolakofsky. 1995. A highly recombinogenic system for the recovery of infectious Sendai paramyxovirus from cDNA: generation of a novel copy-back nondefective interfering virus. EMBO J. 14:6087–6094. 9. Garcin, D., G. Taylor, K. Tanebayashi, R. Compans, and D. Kolakofsky. 1998. The short Sendai virus leader region controls induction of programmed cell death. Virology 243:340–353. 10. Gubbay, O., J. Curran, and D. Kolakofsky. 2001. Sendai virus genome synthesis and assembly are coupled: a possible mechanism to promote viral RNA polymerase processivity. J. Gen. Virol. 82:2895–2903. 11. Hoffman, M. A., and A. K. Banerjee. 2000. Precise mapping of the replication and transcription promoters of human parainfluenza virus type 3. Virology 269:201–211. 12. Iseni, F., F. Baudin, D. Garcin, J. B. Marq, R. W. Ruigrok, and D. Kolakof- 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 9155 sky. 2002. Chemical modification of nucleotide bases and mRNA editing depend on hexamer or nucleoprotein phase in Sendai virus nucleocapsids. RNA 8:1056–1067. Kolakofsky, D. 1976. Isolation and characterization of Sendai virus DIRNAs. Cell 8:547–555. Kolakofsky, D., and A. Bruschi. 1975. Antigenome in Sendai virions and Sendai virus-infected cells. Virology 66:185–191. Kolakofsky, D., T. Pelet, D. Garcin, S. Hausmann, J. Curran, and L. Roux. 1998. Paramyxovirus RNA synthesis and the requirement for hexamer genome length: the rule of six revisited. J. Virol. 72:891–899. Lamb, R. A., and D. Kolakofsky. 2001. Paramyxoviridae: the viruses and their replication, p. 1305–1340. In D. M. Knipe and P. M. Howley (ed.), Fields virology. Lippincott, Williams & Wilkins, Philadelphia, Pa. Le Mercier, P., D. Garcin, S. Hausmann, and D. Kolakofsky. 2002. Ambisense Sendai viruses are inherently unstable but are useful to study viral RNA synthesis. J. Virol. 76:5492–5502. Murphy, S. K., Y. Ito, and G. D. Parks. 1998. A functional antigenomic promoter for the paramyxovirus simian virus 5 requires proper spacing between an essential internal segment and the 3⬘ terminus. J. Virol. 72:10– 19. Murphy, S. K., and G. D. Parks. 1999. RNA replication for the paramyxovirus simian virus 5 requires an internal repeated (CGNNNN) sequence motif. J. Virol. 73:805–809. Pelet, T., C. Delenda, O. Gubbay, D. Garcin, and D. Kolakofsky. 1996. Partial characterization of a Sendai virus replication promoter and the rule of six. Virology 224:405–414. Rager, M., S. Vongpunsawad, W. P. Duprex, and R. Cattaneo. 2002. Polyploid measles virus with hexameric genome length. EMBO J. 21:2364–2372. Skiadopoulos, M. H., L. Vogel, J. M. Riggs, S. R. Surman, P. L. Collins, and B. R. Murphy. 2003. The genome length of human parainfluenza virus type 2 follows the rule of six, and recombinant viruses recovered from nonpolyhexameric-length antigenomic cDNAs contain a biased distribution of correcting mutations. J. Virol. 77:270–279. Stillman, E. A., and M. A. Whitt. 1999. Transcript initiation and 5⬘-end modifications are separable events during vesicular stomatitis virus transcription. J. Virol. 73:7199–7209. Tapparel, C., D. Maurice, and L. Roux. 1998. The activity of Sendai virus genomic and antigenomic promoters requires a second element past the leader template regions: a motif (GNNNNN)3 is essential for replication. J. Virol. 72:3117–3128. Tarbouriech, N., J. Curran, R. W. Ruigrok, and W. P. Burmeister. 2000. Tetrameric coiled coil domain of Sendai virus phosphoprotein. Nat. Struct. Biol. 7:777–781. Vulliemoz, D., and L. Roux. 2001. “Rule of six”: how does the Sendai virus RNA polymerase keep count? J. Virol. 75:4506–4518.