INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The alphaviruses have similar replication and transcription strategies in that all synthesize a minus-strand RNA that is complementary to the genomic plus-strand RNA (25, 27, 37, 41, 42). The minus strand is then used as template for genomic replication and subgenomic mRNA transcription (14, 19, 20, 28, 29, 37). With respect to transcription, Ou et al. identified a highly conserved region at the junction of the alphavirus nonstructural protein (nsP) and structural (STR) coding regions, 19 nucleotides (nt) upstream and 2 nt downstream of the subgenomic mRNA start site, that was hypothesized to serve as the promoter for alphavirus subgenomic mRNA transcription (26). Indeed, a minimal promoter region 19 nt upstream and 5 nt downstream of the subgenomic mRNA start site (19/+5, encompassing the conserved sequence element identified by Ou et al. (26) was sufficient for detectable levels of subgenomic mRNA transcription (19) and a 3-nt insertion after position 6 dramatically reduced transcription levels and virus growth (8).

Heterologous alphavirus minimal promoter sequences can be recognized quite efficiently by Sindbis virus (SIN) for transcription, suggesting that transcriptional requirements might be a major contributor to the conservation of the junction region (11). While the junction region of SIN also encodes the C terminus of nsP4 (whose termination codon spans positions +2 to +4), its contribution to the conservation of the junction region may be smaller, at least in the minimal promoter: The nsP4 termination codon of the equine encephalitis viruses and the aquatic alphaviruses is at positions 8 to 6. Promoter sequences between 5 and +4 of these viruses could potentially change in the absence of nsP4 coding constraints, but they remain conserved. In Semliki Forest virus, the nsP4 termination codon is at positions +11 to +13, and so the length of the carboxyl terminus of alphavirus nsP4 does not appear to be strongly constrained. Additionally, the wobble positions for nsP4 amino acid codons should be able to change in the absence of other constraints, but they remain conserved. The contribution of other functions of the region to sequence conservation remains to be determined. Positions +1 onward encode the 5' end of the subgenomic mRNA, which might affect its capping, stability or translation efficiency. The viral RNA II (believed to be a stalled replication product) terminates at position 4, and the amount of RNA II that is made appears to depend on the junction region (46). If RNA II does play a functional role during the virus life cycle, perhaps the regulation of the relative levels of viral replication versus transcription (46) might also account for some of the sequence conservation.

In vivo evolution experiments suggested that the conserved nucleotides of the alphavirus junction region may in fact be optimized for promoter function (9, 10). The 13 to 9 region was randomized, in the absence of nsP4 coding constraints, to create a library of promoter variants. Viruses in the library were competed against each other to identify those that grew best. Even though some non-wild-type promoter sequences were found that could support transcription and virus growth, the wild-type promoter sequence was strongly preferred (9, 10).

By itself, the 19/+5 promoter region is three- to sixfold less active than the 98/+14 promoter region (32), suggesting that regions upstream and/or downstream of this element may be required for full promoter activity. In this study, we used deletion analysis of the 98/+14 promoter region to identify two regions about the SIN minimal promoter, at positions 40 to 20 and +6 to +14, that were sufficient to achieve promoter activity comparable to the 98/+14 promoter region. In addition, we used the in vivo evolution method (9, 10) to identify the sequence preference and the location of positions within the 40 to +14 promoter region required for promoter function and virus growth.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell lines. Baby hamster kidney (BHK-21) (ATCC CCL10; between passages 9 and 20) cells were grown at 30 or 37°C in minimal essential medium with Earle's salts, supplemented with 10% heat-inactivated fetal calf serum. C7-10 cells (between passages 30 and 50) from Aedes albopictus larvae (34) were grown at 30°C in the same medium supplemented with 10% tryptone phosphate buffer.

Deletion analysis. The TCS clone (~12.6 kb) (9, 10) contains the nonstructural (nsP) and structural (STR) coding regions of SIN, separated by the chloramphenicol acetyltransferase (CAT) gene. It contains two 98/+14 subgenomic promoters (9). The first, designated the CAT promoter, drives the expression of the CAT gene. The second, designated the STR promoter, drives the expression of the STR genes and is downstream of the CAT promoter. The 19/+5 (9) 19/+14, 40/+14, 40/+5, and 98/+5 (46) clones are identical to TCS except that the STR promoter of TCS was replaced by the respective 19/+5 (9-11), 19/+14, 40/+14, 40/+5, or 98/+5 promoters.

Promoters with single point mutations. Two 48-nt oligonucleotides were synthesized that contained the SIN 19/+5 minimal promoter. During the synthesis of one oligonucleotide, each purine position of the minimal promoter was replaced by a mixture of ~90% wild-type nucleotide and ~3% each of the remaining nucleotides. The second oligonucleotide was made similarly, except that each pyrimidine position of the minimal promoter was replaced by the mixture. In each, the minimal promoter sequence was flanked by 4-nt spacers and KpnI and XbaI restriction sites for cloning. Each oligonucleotide library was made double stranded via PCR, using primers complementary to the ends of each oligonucleotide library. The PCR products were digested with KpnI and XbaI and directionally cloned into the JUNCAT plasmid (11). Clones were screened using low-stringency colony hybridization with an end-labeled negative-sense oligonucleotide spanning the wild-type 19/+5 minimal promoter sequence, and 162 clones that contain mutant promoters were identified. Of these, 22 had point mutations. They were digested with ApaI and PvuI and directionally cloned into the DIC20e plasmid (32), replacing the 3' promoter sequence with the mutant promoters. The DIC20e plasmid contains the cDNA sequence of a double subgenomic mRNA promoter construct of a defective interfering (DI) genome with all the cis-acting sequences required for replication and packaging (20). The 5' promoter of the DIC20e plasmid consists of the wild-type SIN minimal promoter, which serves as an internal control. It is required for the synthesis of a 1.8-kb mRNA (0.35 kb of spacer sequence and the 3' promoter, followed by the CAT sequence). The 3' promoter is the test promoter used to transcribe a 1.45-kb CAT mRNA. The DIC20a clone is identical to the DIC20e derivatives, except that its 3' promoter is the SIN wild-type minimal promoter sequence (19/+5), used as a control to compare test promoters of interest.

In vitro transcription, transfection, isolation of total RNA, and RT-PCR. Virus clones or libraries were linearized with SstI and transcribed in vitro using the Epicentre (Madison, Wis.) SP6 RNA-dependent RNA polymerase (46). Transfections were performed as described previously (22, 23) unless indicated otherwise. Approximately 1 µg of total RNA isolated from Trizol (Life Technologies, Inc.)-solubilized cell samples was subjected to reverse transcription (RT) with SuperScript II (Life Technologies, Inc.) in a 10-µl reaction volume as specified by the manufacturer. PCR used the MasterAmp Tth DNA Polymerase (Epicentre) as specified by the manufacturer.

Viral RNA labeling and analysis. Approximately 1 × 10⁶ to 3 × 10⁶ C7-10 cells or 7 × 10⁵ BHK-21 cells were infected at a multiplicity of infection (MOI) of 1 to 5, and viral RNAs were labeled with [³²P]orthophosphate, as described previously (46). Approximately one-fifth of the total RNA from each sample was denatured (3), electrophoresed through 1% agarose gels in 1× TBE (33), and exposed to preflashed film. Exposures were performed at 20°C for 2 to 5 days with (C7-10) or without (BHK-21) an intensifying screen. Radiolabeled RNA was quantitated on a phosphorimager (Bio-Rad) by integrating the area under each radioactive peak. Baselines were set at the valley between genomic and CAT subgenomic RNA. The relative molar ratios of STR subgenomic mRNA versus CAT subgenomic mRNA was calculated as 1.2 × (STR mRNA counts/CAT mRNA counts), since the length of CAT mRNA (4,991 nt) is 1.2 times greater than that of the STR mRNA (4,156 nt).

Competition assays. Defined molar ratios of in vitro-transcribed RNA (~5 µg total) from TCS, STR promoter deletion clones, or individual promoter clones were mixed and transfected into C7-10 cells. The media were harvested at 26 h (individual promoter clones), 30 h (STR promoter deletion clones), or 48 h (for 19/+5 and 19/+14 competition) postelectroporation (p.e.). Virus stocks obtained after transfection were designated passage 1 (P1) virus, while viral RNA contained within transfected cells was designated P0 RNA. P1 virus mixtures were subjected to titer determination on C7-10 cells and used to infect fresh C7-10 cells at an MOI of 0.1. At 24 h (30 h for the 19/+5 and 19/+14 competition) postinfection (p.i.), the medium (P2 virus) and infected cells (P1 RNA) were harvested. Virus populations were passaged on fresh C7-10 cells until P2 or P3 RNA was obtained. Total RNA from each Trizol-solubilized sample was reverse transcribed using the 1926 primer (Table 1). PCR was performed on 10 to 15% of each cDNA sample using the 1941 and 1926 primers (Table 1) for 10 cycles and then held at 94°C for the addition of 2 pmol of [5'-³²P]labeled 1941 primer (Table 1), and PCR was performed for 10 more cycles. The PCR products were resolved on 8% denaturing polyacrylamide gels (33), visualized by autoradiography, and quantitated on a Bio-Rad phosphorimager.

cDNA libraries. The previous cloning strategy (9) was used to make nine different full-length cDNA libraries named according to the positions of the randomized nucleotides within the 40/+14 promoter region. Oligonucleotides containing a block of five random nucleotides at various positions within the 40/+14 promoter region were synthesized by Integrated DNA Technologies (Coralville, Iowa) for the 35, 25, 20, 15, 5, and +1 oligonucleotides or by Life Technologies, Inc., for the 30, 10, and +5 oligonucleotides (Table 1). The quality of the randomized oligonucleotides was assessed by sequencing them after PCR amplification using the appropriate 5' or 3' primers (Table 1). Based on previous sequence analyses, detection of a given base at any position in a mixture required that it be present at an abundance of 0.2 (9). PCR products which appeared random by this criterion were used to construct the libraries.

Passaging of virus libraries in cultured cells. Approximately 3 µg (C7-10) or 5 µg (BHK-21) of in vitro transcripts of each library was transfected in duplicate into C7-10 cells or singly into BHK-21 cells. An aliquot from each transfected C7-10 or BHK-21 cell sample was serially diluted for an infectious-center assay to determine the number of successful transfection events, from which the minimal size of the library transfected into host cells may be estimated (9) (Table 2). The first set of C7-10 plates, used to estimate the diversity of each library shortly after transfection (see below), was incubated at 30°C until 1 to 2 h p.e. (35, 30, 25, 20, 15, 10, 5, and +5 libraries) or 8 h p.e. (+1 library). The media were removed, and the transfected cells were solubilized with 3 ml of Trizol reagent for isolation of the P0 RNA. The second set of transfected C7-10 cells was incubated at 30°C until 30 h p.e. The transfected BHK-21 cells were incubated at 37°C until 21 h p.e. At these time, the media containing the P1 virus populations were harvested. For the next passage, approximately 9 × 10⁵ C7-10 cells or 7 × 10⁵ BHK-21 cells were seeded onto 35-mm-diameter wells and incubated for 1 day at 30°C for C7-10 or at both 30° and 37°C for BHK-21, at which time they were ~85% confluent. Cells were infected with the P1 virus population at a MOI of 0.1. At 1 h p.i., the inoculum was removed and replaced with 1 ml of the appropriate media. At 24 h p.i. for C7-10 cells, 5 h p.i. for BHK-21 cells (37°C), or 8 h p.i. for BHK-21 cells (30°C), the media containing the P2 virus were collected. Cells remaining in the plate were solubilized with 1 ml of Trizol reagent for isolation of the P1 RNA. Passaging continued in this manner until P3 RNA for C7-10 cells or P4 RNA for BHK-21 cells was obtained.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Sequence elements required for full promoter activity. The alphavirus junction region contains two regions of conservation (Fig. 1): one identified by Ou et. al. (26), between positions 19 and +2 (with respect to the start site of alphavirus subgenomic mRNA) and a smaller one at positions 35 to 30. Previous studies demonstrated that the 19/+5 region of SIN (11, 19) was sufficient for directing mRNA transcription. However, it was approximately three- to sixfold less active in subgenomic mRNA synthesis than the 98/+14 region was (32, 46) and supported slower virus growth (10). Sequences required for the higher levels of transcription appear to be confined to the 40/20 and +6/+14 regions immediately adjacent to the minimal promoter in mammalian cells (46) and also in mosquito cells (Fig. 2A [values under each lane designate promoter activity relative to the wild-type CAT promoter serving as an internal control in each clone and normalized to the TCS wild-type clone] [32]). While the activity of the 40/+14 region is comparable to that of the 98/+14 region of TCS (Fig. 2A), absence of the +6/+14 region (clones 98/+5, 40/+5, and 19/+5) or the 40/20 region (clone 19/+14) decreased STR mRNA transcription by about two- to fourfold, respectively.

View larger version (47K): [in this window] [in a new window]   FIG. 1.   Subgenomic mRNA promoter region of alphaviruses. The numbering denotes positions relative to the subgenomic mRNA initiation site. Identity to the SIN sequence is denoted by dashes. The boxes enclose the two conserved regions. The nsP4 sequence of SIN is shown above the sequence alignment. Abbreviations: A86, GIR, YN8, OCK, XJ1, Sindbis-like S.A.AR86, Girdwood, YN87448, Ocklebo and XJ-160 viruses, respectively; BFV, Barma Forest virus; AUR, Aura virus; SFV, Semliki Forest virus; MBV, Middelburg virus; RRV, Ross River virus; SAG, Sagiyama virus; ONN, O'nyong nyong virus; IGB and SG6, O'nyong nyong virus-like Igbo Ora and SG650 viruses; WEE, western equine encephalitis virus; EEE, eastern equine encephalitis virus; VEE, Venezuelan equine encephalitis virus; SPD, salmon pancreas disease virus; SDV, rainbow trout sleeping disease virus (4, 13, 16, 17, 21, 24, 26, 35, 38, 40, 43, 45).

View larger version (71K): [in this window] [in a new window]   FIG. 2.   Regions flanking the minimal SIN promoter which enhance transcription and virus fitness. (A) The TCS, 98/+5, 40/+14, 40/+5, 19/+14, and 19/+5 viruses were used to infect C7-10 cells at a MOI of 3. Viral RNA was labeled in vivo with [32P]orthophosphate between 21.5 and 24 h p.i. Total RNA was resolved on a 1% agarose gel for analysis via autoradiography. The relative STR promoter activities shown below each lane were calculated by dividing the abundance of CAT subgenomic mRNA by that of STR subgenomic mRNA for each virus, and normalizing these values to that obtained for TCS. *, RNA II terminating at the STR promoter; #, RNA II terminating at the CAT promoter. (B) Effect of the 40/20 and +6/+14 regions on viral fitness. Mixtures of the in vitro transcripts listed in each panel (I = 98/+5, 40/+5, and 19/+5 mixture; II = 98/+14 [TCS], 40/+14, and 19/+14 mixture; III = 19/+14 and 19/+5 mixture; IV = 40/+14 and 40/+5 mixture) were transfected into C7-10 cells and passaged two or three times using a MOI of 0.1 (P0 = transfected cells). Total-cell RNA was isolated at each passage, and the viral STR promoter regions were 32P radiolabeled during RT-PCR. The PCR products were resolved on an 8% sequencing gel for analysis by autoradiography. The relative abundance of each virus in the mixes is listed below each lane, ranked by the size of their PCR products, in bases: TCS (233 bases), 98/+5 (224 bases), 40/+14 (175 bases), 40/+5 (166 bases), 19/+14 (154 bases), and 19/+5 (147 bases).

Sequence preference for promoter function. The importance of only a few positions in the minimal promoter has been identified to date (6, 8-10). Since the 40/+14 region appears to be functionally similar to the 98/+14 region, we focused on it to identify the sequences required for promoter function. We used an in vivo evolution method (9) that efficiently samples the functions of thousands of different promoter sequences in parallel in the context of the normal infection cycle. Nine virus libraries were generated. Like the STR promoter deletion clones described above, viruses in the libraries contain two subgenomic mRNA promoters, the CAT and the STR promoters. Each library consists of a randomized sequence of five contiguous nucleotides located at various positions within the 40/+14 STR promoter. The libraries are named according to the location of the randomized sequences (e.g., positions 35 to 31 are randomized in the 35 library [Table 1]). In vitro-transcribed RNA from each library was transfected into C7-10 and BHK-21 cells and serially passaged three (C7-10) or four (BHK-21) times. At each passage, the sequence of the promoter region of the virus population as a whole was determined (Fig. 3). The promoter sequences shown are those of the minus strand.

View larger version (77K): [in this window] [in a new window]   FIG. 3.   Evolution of the 35/+9 promoter region during passaging in cultured cells. In vitro-transcribed RNA from each library was transfected into C7-10 cells (A) or BHK-21 cells (B). The virus produced by the transfected cells were then passaged three or four times at a MOI  0.1 (see Materials and Methods). Intracellular viral RNA during each passage was isolated, and the STR promoter region was RT-PCR amplified. The RT-PCR product was purified and sequenced directly to obtain the consensus sequence of each population after each passage (P0 = transfected cells). The sequences shown are those of the minus strand.

View larger version (12K): [in this window] [in a new window]   FIG. 4.   SIN promoter. (A) Consensus sequence of the virus populations after three passages on C7-10 cells at 30°C (Fig. 3A). Consensus bases are those that show obvious enrichment over the other bases (Fig. 3A). Bases that were found to be wild type are underlined. Positions that appeared to prefer more than one base are designated by the standard single-letter code (R = G or A; Y = U or C; S = G or C; W = A or U; K = G or U; H = A, U, or C; D = A, G, or U; N = A, C, G, or U). (B) Consensus sequence of individual clones sampled from the C7-10 P3 virus population (Table 3). Unambiguous bases are those found at a frequency of 0.6, except for positions 25 and 16, where the dominant base had a frequency of 0.5 but the other bases each had frequency of 0.2. (C). Consensus sequence of the virus populations after four passages on BHK-21 cells at 37°C (Fig. 3B). (D) Consensus sequence of the 35 and +5 virus populations after four passages on BHK-21 cells at 30°C (Fig. 3B). Blanks indicate that populations were not tested.

Promoter preference in mammalian cells. Previous studies indicated that growth in mosquito cells poses more stringent demands on promoter activity than does growth in hamster cells (10). Consequently, it is possible that some positions in the 40/+14 promoter region might exhibit different selection intensities or preferences for particular bases in alternate hosts. To test this, transcripts of the libraries were transfected into BHK-21 cells and passaged at 37°C (Fig. 3B and 4C). As observed in C7-10 cells, wild-type bases were selected for in BHK-21 cells in most of the 15 to +5 region. Thus, at most positions in the minimal promoter, the wild-type base appears to be optimal for subgenomic mRNA transcription in both insect and mammalian cells. Selection for promoter function does appear less stringent in hamster cells, since higher levels of non-wild-type bases were found, e.g., at positions 15, 12, 6, and +4, compared to that observed in C7-10 cells (Fig 3A and B and 4A and C). Furthermore, there was no obvious selection at many positions in the 35 to 16 and +6 to +9 regions in BHK-21 cells compared to what is seen in C7-10 cells (Fig. 3A and 4A). Where selection can be perceived, e.g., at positions 30 to 28, the bases selected for are generally the same as those selected for in C7-10 cells. The only exception is at position 34, where U is preferred in BHK-21 cells and A is preferred in C7-10 cells.

Promoter activity and the fitness of P3 viruses. The in vivo evolution experiments showed that the wild-type base was generally preferred at positions 15 to +5. In contrast, positions 35 to 16 and +6 to +9 generally preferred non-wild-type bases or exhibited relaxed preferences. This suggests that some non-wild-type sequences function as well as the wild type does. If this is true, the activity of these promoters should be comparable to that of the wild type. Additionally, they should support virus growth as well as the wild-type promoter does. To test this, we cloned the most frequently sampled promoter (Table 3) from each P3 population into the wild-type background to ensure that the non-wild-type promoter sequences isolated were not accompanied by compensatory mutations outside of the 40/+14 promoter region which allowed the virus to grow well. The resulting clones are identical to the 40/+14 clone except at the indicated positions (Table 3). The recloned promoters supported high-titer virus growth after transfection (data not shown), comparable to that of TCS and 40/+14 and better than that of the 19/+5 clone, especially in C7-10 cells (data not shown). This suggests that new mutations were not required for good growth. The activity of the promoters in these clones was then measured in C7-10 and BHK-21 cells (Fig. 5A). All had activity comparable to that of the 40/+14 and 98/+14 promoters and clearly higher than that of the 19/+5 promoter in both host cells. We also tested a randomly chosen promoter from the 20 P0 population. Its sequence is entirely non-wild type, but it had promoter activity comparable to that of the 98/+14 and 40/+14 promoters. This provides further evidence that the wild-type sequence at positions 20 to 16 is not required for promoter function.

View larger version (80K): [in this window] [in a new window]   FIG. 5.   Promoters of P3 clones. (A) Promoter activity. C7-10 or BHK-21 cells were infected with the most frequently isolated virus from each P3 population (Table 3) and an isolate from the 20 library at P0 with promoter sequence of 3'AAGCU. Viral RNAs were radiolabeled in vivo with [32P]orthophosphate at 3 h (BHK-21) or 21.5 h (C7-10) p.i. and isolated at 5 or 24 h p.i., respectively. Total RNA was denatured and resolved on a 1% agarose gel. [32P]orthophosphate-radiolabeled RNA was examined via autoradiography and phosphorimager analyses. G denotes the genomic RNA of each clone; CAT denotes CAT mRNA transcribed by the 98/+14 CAT subgenomic mRNA promoter; and STR denotes STR mRNA transcribed by the STR promoter region of each virus. Relative promoter activity, listed below each lane, was determined by finding the ratio of STR to CAT mRNA for each sample and normalizing the values against that calculated for TCS. (B) Fitness of the viruses. The viruses were competed against TCS by transfecting approximately equal amounts of the respective in vitro transcripts into C7-10 cells. The resulting virus populations were passaged three more times. The promoter region of viruses in each mix is amplified by RT-PCR, labeled, and gel resolved as described in the legend to Fig. 2B. The relative abundance of each P3 virus relative to that of TCS is listed below each lane. The size of the PCR product from TCS is 233 bp, that from the P3 clones is 175 bp, and that from the 19/+5 virus is 147 bp.

The effect of point mutations on promoter activity. The in vivo evolution studies show that the wild-type base is strongly preferred at most positions in the minimal promoter (Fig. 3 and 4). The apparent superiority of the wild-type base was verified by measuring the activity of 22 minimal promoters with point mutations (Fig. 6) using double promoter constructs (32) derived from a SIN DI genome (18, 20) that include the wild-type SIN 19/+5 promoter as internal reference (32) (see Materials and Methods). None of the 14 point mutations resulting in promoter activities of 60% were present in the promoters sampled after three passages in C7-10 cells (Tables 3), and most are clearly depleted in the P3 population as a whole (Fig. 3A). For example, mutations 6C and 11U result in promoter activities approximately 60% that of the wild type (Fig. 6). They were at best faintly present in the population after three passages (Fig. 3A). In comparison, five of the eight point mutations resulting in promoter activities 80% of the wild-type activity (Fig. 6) were found among the clones isolated after three passages in mosquito cells (Table 3). This suggests that the promoter must be 80% as active as the wild type for viral fitness to be high enough to survive three cycles of selection in mosquito cells. Comparable results were found in BHK-21 cells; for example, 11U persisted at relatively low frequencies to at least four passages (9, 10).

View larger version (5K): [in this window] [in a new window]   FIG. 6.   Effect of point mutations in the 19/+5 region on promoter activity. DI constructs with double promoters were used to measure the activity of each mutant promoter relative to the wild-type internal control (see Materials and Methods). The values obtained were then normalized against that obtained for the DIC20a construct, both of whose promoters were the wild-type 19/+5 promoter, and expressed as a percentage of wild-type activity. The top line shows the wild-type SIN sequence. Each value in subsequent lines corresponds to the promoter activity due to the base change at the indicated nucleotide position. For example, an A-to-G mutation at position 18 has 90% of wild-type activity.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The complete SIN promoter was mapped to within the 40/+14 region encompassing the subgenomic mRNA start site. In vivo evolution was used to identify positions in the 35/+9 region, as well as the specific bases at each position that are preferred for promoter function (Fig. 4). We reasoned that positions where specific bases were most rapidly selected for are likely to be essential for promoter function while those that evolve more slowly probably contribute to achieving full promoter activity. By these criteria, the promoter consists of an essential region from 15 to +5 required for promoter function and a number of sites flanking it that contribute to full promoter activity in mosquito cells (Fig. 3). The essential region is the same in mammalian cells, but few of the flanking sites are required, some of which also depend on the culture temperature. These conclusions are consistent with previous studies showing that the 19/+5 region is the smallest region still giving detectable promoter activity and that full promoter activity could be obtained with the 98/+14 region (19, 32). For example, deletion of downstream sequences, between +6 and +14, should diminish promoter activity, but deletions into the essential region, beyond position +5, should abolish promoter function. Similarly, deletion of upstream sequences, in the 40 to 16 interval, should progressively reduce promoter activity, eventually rendering promoter activity undetectable even if the deletions did not extend into the essential region. This model of the promoter is verified by examining the properties of mutant promoters. Point mutagenesis experiments showed that base changes at many sites in the essential region result in defective promoter function (Fig. 6). The available data suggest that even a small decrease in promoter activity, ca. 40%, significantly decreases viral fitness. Complementary information is provided by characterizing promoters from the P3 populations, which are essentially mutants with one to five base changes. The results identify the mutations at specific positions that are consistent with normal promoter function (Table 3; Fig. 5). Notably, all but three of these positions are outside of the 15 to +5 essential region. Examination of the consensus sequence (Fig. 4A) and the sequence of individual isolates (Table 3) did not reveal any obvious secondary structure in either the essential or the flanking regions. In this respect, the SIN promoter resembles the brome mosaic virus subgenomic promoter (5) and is unlike the cucumber mosaic virus (5) or rubella virus (7) promoters, which are members of the alphavirus-like supergroup (39).

In theory, positions unrelated to promoter function should exhibit no base preference. However, most positions outside of the essential region do exhibit some base preference, but they typically prefer the non-wild-type base. In addition to contributing to promoter function, they may have other roles. The 5' end of the subgenomic mRNA is encoded by the region studied, and selection for expression of optimal amounts of the structural proteins may also operate on the quality of the mRNA. Selection might also operate on the viral RNA II (46), which terminates at position 4, and the amount of RNA II that is made appears to depend on the 40 to +20 region (Fig. 2A). Further studies are needed to disentangle the possibly overlapping sequence requirements of these functions from those required just for promoter function. The junction region also encodes the carboxyl terminus of nsP4. The genome of the viruses used in this study was designed to remove this constraint; therefore, the observed base preference is not due to selection for nsP4 function. The fact that positions in the 35 to 16 region generally prefer non-wild-type bases shows that nsP4 coding constraints in the normal genome context play a major role in limiting the spectrum of allowable changes, even when the changes were preferred for promoter function in its absence. Thus, it is likely that the wild-type sequence outside the essential region reflects a compromise that best satisfies the potentially conflicting demands of the several functions of the region.

When provided with a choice of all four bases, the wild-type base is strongly preferred in both mosquito and mammalian cells, especially within the 15/+5 essential region (Fig. 4), suggesting that the wild-type sequence is optimal for promoter function and that the optimum is the same in both types of hosts. Similarly, point mutagenesis studies of the 3' conserved sequence of alphaviruses, presumably required for initiation of minus-strand synthesis, showed that all mutations but one examined were deleterious (14), suggesting that this sequence too had been optimized. For the positions that are conserved among the alphaviruses (Fig. 1), sequence optimization probably preceded the divergence of the viruses. For the positions that are different among the alphaviruses, optimization probably occurred in parallel with sequence divergence. For example, the point mutagenesis results (Fig. 6) show that 10A of salmon pancreas disease virus, 7U of salmon pancreas disease virus, rainbow trout sleeping disease virus, and Venezuelan equine encephalitis virus and +3A of Semliki Forest virus are individually deleterious in the SIN context. All these viruses probably have high mutation rates, and their promoters are likely to be as important for their life cycle as it is for SIN. We therefore predict that promoter sequence changes during the divergence of the alphaviruses were accompanied by compensatory changes elsewhere in the promoter or the cognate viral transcription factor, since any host factors involved would evolve much too slowly. While the compensatory changes remain to be identified, it is likely that promoter recognition was optimized rapidly. Variants generated by mutation during replication have an initial abundance of ca. 10⁴ to 10⁵, about 10 to 100 times lower than the abundance of the wild-type base in our libraries. Since only a few passages were sufficient for the wild-type base to be selected for in the in vivo selection experiments, optimization in nature probably requires only a modest number of replication cycles.

The cis-acting sequences of many other RNA virus families are also very well conserved (see reference 12 for a review). The available evidence suggests that they too have been optimized. For example, point mutagenesis studies of the influenza A virus vRNA promoter (30), the transcription signals of vesicular stomatitis virus (1, 2) and human respiratory syncytial virus (15), the 3' transcription signal of Rift Valley virus (31), and the replication signal of a group A rotavirus (44) show that most if not all point mutations disrupt normal functioning.

Sequence optimization might not be confined to viral cis-acting sequences. Steinhauer and Holland (36) proposed that the sequence of RNA virus populations can remain stable during extended periods of environmental stability despite their high mutation rates, because none of the variants is more fit than the wild type; i.e., the wild-type sequence is optimal. Indeed, the high mutation rates of RNA viruses might actually expedite genome optimization whenever the environment changes, by generating an abundant diversity of variants to be tested by selection. It will be interesting to determine if many other parts of the virus genome are also optimized.