Mapping Viral Functional Domains for Genetic Diversity in Plants

Abstract

Cucumber mosaic virus (CMV) comprises numerous isolates with various levels of in-host diversity. Subgroup-distinctive features of the Fny and LS strains provided us with a platform to genetically map the viral control elements for genetic variation in planta. We found that both RNAs 1 and 2 controlled levels of genetic diversity, and further fine mapping revealed that the control elements of mutation frequency reside within the first 596 amino acids (aa) of RNA 1. The 2a/2b overlapping region of the 2a protein also contributed to control of viral genetic variation. Furthermore, the 3′ nontranslated region (NTR) of RNA 3 constituted a hot spot of polymorphism, where the majority of fixed mutations found in the population were clustered. The 2b gene of CMV, a viral suppressor of gene silencing, controls the abundance of the fixed mutants in the viral population via a host-dependent mechanism.

INTRODUCTION

RNA viruses are often characterized by their large genetic variation, inherent in error-prone replication (1, 2), and their short generation times (2). As a consequence, RNA viruses are difficult to control. They can adapt easily to new environments to expand their host ranges. A prime example is the single-stranded positive-sense RNA virus Cucumber mosaic virus (CMV) (genus Cucumovirus; family Bromoviridae), which has been used extensively as a model system for virus evolution studies (3).

The CMV genome is divided into three single-stranded positive-sense RNAs designated RNAs 1, 2, and 3 (4, 5). RNA 1 encodes protein 1a, which is involved in virus replication (6, 7) and contains the putative methyltransferase domains (8) in the amino-terminal region and the helicase domains (9) in the carboxy-terminal portion. RNA 2 encodes two proteins, the 2a protein, also involved in virus replication and encoding of the RNA-dependent RNA polymerase (RdRp) (6, 7), and the 2b protein, which is expressed from RNA 4a, a subgenomic RNA derived from RNA 2, and is involved in virus movement, pathogenicity, and suppression of RNA silencing (10–13). RNA 3 also encodes two proteins, the 3a movement protein and the coat protein (CP), both required for virus movement (14, 15).

CMV is one of the most evolutionarily successful viruses known, with the largest host range of any known virus, infecting more than 1,200 plant species (16). In previous studies, CMV and two other alpha-like plant viruses, Tobacco mosaic virus (TMV) (genus Tobamovirus; family Virgiviridae) and Cowpea chlorotic mottle virus (CCMV) (genus Bromovirus; family Bromoviridae), were used for a direct comparison of viral genetic variation in Nicotiana benthamiana (17), and the viruses had differences in the levels of population variation that correlated with their relative host range sizes. In a study that followed, CMV was used to further assess these differential viral genetic variations in various plant hosts (18). These two studies clearly demonstrated that the level of viral variation found in a particular host was controlled by both the virus and the plant host, but they did not address the mechanism(s) that would result in a specific level of diversity for a particular host-virus interaction. However, the variety of available CMV strains and the divided genome of the virus provide a system to genetically map the viral element(s) that controls the differential variations in plants.

Based on serological, biological, and sequence properties, strains of CMV are divided into two major subgroups, I and II, and subgroup I is further divided into subgroups IA and IB (19). In this study, we used two strains of CMV: Fny, belonging to subgroup IA, and LS, belonging to subgroup II. We showed that the two strains have significantly different mutation frequencies, and we used reassortant and intermolecular recombinant viruses between Fny- and LS-CMV to conduct genetic mapping of the viral domains controlling genetic variations.

MATERIALS AND METHODS

Viruses.

Fny- and LS-CMV and the procedure to make infectious RNA transcripts from these CMV strains were previously described (20, 21).

Hosts and plant inoculations.

The plant hosts used were tobacco (Nicotiana tabacum cv. Xanthi nc), N. benthamiana, pepper (Capsicum annuum cv. Morengo), squash (Cucurbita pepo cv. Elite), and tomato (Solanum lycopersicum cv. Rutgers). The plants were maintained in a greenhouse with daytime temperatures of 28°C, nighttime temperatures of 22°C, and a 16-h day. Using transcripts from cDNA clones of Fny- and LS-CMV RNAs 1, 2, and 3, infection was established in 10 3-week-old plants with all possible combinations of genomic RNAs, i.e., Fny RNA 1 (F₁), Fny RNA 2 (F₂), and Fny RNA 3 (F₃), and LS RNA 1 (L₁), LS RNA 2 (L₂), and LS RNA3 (L₃), representing the parental viruses. The reassortants included F₁F₂L₃, F₁L₂L₃, F₁L₂F₃, L₁L₂F₃, L₁F₂L₃, and L₁F₂F₃. Control plants were inoculated with buffer only (50 mM Na₂HPO₄, pH 9).

Total-RNA extraction and HiFi RT-PCR.

At 15 days postinoculation (p.i.), total RNA was isolated from systemically infected leaves using Tri-reagent, according to the manufacturer's protocol (Molecular Research Center Inc., Cincinnati, OH). One-fifth of the total RNA extraction was then used in a high-fidelity reverse transcription (HiFi RT). SuperScript III was used for RT reactions, as recommended by the manufacturer (Invitrogen), with the first-strand primer-1 (GGCTGCAGTGGTCTCCTT), which is specific for the 3′ ends of both Fny- and LS-CMV RNA 3. After a 30-min incubation at 37°C, the cDNAs obtained were used as templates for thermal-cycling reactions with the same first-strand primer-1 and forward primer-2 (CCCCCCGAATTCTCATGGATGCTTCTCCGCGAG), specific for both Fny- and LS-CMV. The generated fragments of 1,060 nucleotides (nt) for Fny RNA 3 or 1,062 nt for LS RNA 3, encompassing part of the intergenic region, the coat protein, and the 3′ nontranslated region (NTR), were used as the reporters (Fig. 1C). The thermal-cycling reactions were carried out as described previously (22).

Fig 1.

Genomic organization of the three RNAs of Fny- and LS-CMV and schematic diagram of intermolecular recombinants and RNA 3 reporters. The lines represent noncoding regions, while the rectangles represent open reading frames. (A) L_{1 M}F_1H was constructed by exchanging sequences between RNAs 1 of Fny- and LS-CMV. The stippled and white rectangles indicate that the sequences are from LS- and Fny-CMV, respectively. The black rectangle represents a stretch of 22 nucleotides that are conserved between Fny-, LS-, and L_{1 M}F_1H. (B) F_2aLS_2b was constructed by replacing the 2b gene of Fny-CMV with that of LS-CMV. To create F_2aΔ2b with a deleted 2b ORF, all three AUGs that are found in the N terminus of the 2b protein were mutated. The dashed rectangles indicate that the 2b ORF is no longer encoding. (C) Portions of RNA 3 encoding the CP and flanking regions that were amplified and cloned for sequence analysis. The solid arrows indicate primer positions, and the numbers correspond to nucleotide positions.

Cloning and sequence analyses of the progeny viral RNA.

The HiFi RT-PCR products were ligated into the pGEM-T Easy vector (Promega) and used to transform DH5α competent cells that were later plated onto LB agar containing iso-propyl-β-d-thiogalactopyranoside (IPTG) and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal). Twenty-four white colonies from each plate were transferred to a broth culture, grown overnight, and used for a small-scale plasmid preparation. The sequences of the inserts of the resulting plasmids were determined using the DNA analyzer ABI 3730 and analyzed using Clustal W. The mutation frequencies were calculated as the number of mutations per kilobase of RNA. Similarly, the background mutation frequency of the HiFi RT-PCR was determined in a control reaction done simultaneously, using in vitro-generated transcripts.

Construction of intermolecular recombinant and 2b mutant clones.

The majority of the helicase domain of LS-CMV was replaced with that of Fny-CMV by subcloning (Fig. 1A). Plasmids pFny86.1 and pLS1 (21, 23) were digested with MfeI and BsiWI, and L_{1 M}F_1H was constructed by ligating the 1,876-nt and 5,842-nt fragments obtained, respectively, from pLS1 and pFny 86.1, into the appropriate sites created by digestion with the same enzymes. The first 596 amino acids (aa) of L_{1 M}F_1H (containing the entire methyltransferase domain and the 43 N-terminal amino acids of LS RNA 1) are 100% identical to those of LS RNA 1, whereas the remaining 397 aa containing the majority of the Fny helicase domain are 100% identical to those of Fny RNA 1.

To replace the 2b open reading frame (ORF) of Fny-CMV with that of LS-CMV (F_2aLS_2b), we used cDNA clone pFny 209 (20) as the template DNA in a three-step PCR amplification with Taq DNA polymerase (Roche) and Pfu (Stratagene). For the first step, primer pairs 1 (AACAGCGAAAGAATTATGGATGTGTTGACAG) and 2 (ATGCGGAAGGGGAGGTTTCAAAACGACCCTTCG) were used to amplify the LS 2b ORF, 3 (TGGAACCTGCCGTACCAT) and 4 (CTGTCAACACATCCATAATTCTTTCGCTGTTT) were used to amplify a fragment of Fny RNA 2 located upstream of Fny 2b (Fny 5′), and 5 (CGAAGGGTCGTTTTGAAACCTCCCCTTCCGCAT) and 6 (GGCTGCAGTGGTCTCCTT) were used for a fragment located at the 3′ end of Fny RNA 2 (Fny 3′). For the second step, a mixture of the amplified products of the LS 2b ORF and Fny 5′ was used as the template for amplification with primers 3 and 2. For the final step, a mixture of the amplified Fny 3′ product and the product of the second PCR step was used as the template for the final amplification with primers 1 and 6, which contain the restriction sites BSpM I and PstI, respectively. The product of the final PCR was reinserted into plasmid pFny 209 to generate construct F_2aLS_2b (Fig. 1B).

To create mutant F_2aΔ2b, corresponding to Fny RNA 2 with a deleted 2b ORF (Fig. 1B), all three AUGs that are found in the N terminus of the 2b protein and in frame with each other were mutated. Since the 2b gene of Fny-CMV overlaps the 2a gene, the mutations introduced were designed to keep the amino acid sequence of the 2a protein intact. Forward primers GCGAAAGAATTACGGAATTGAACG, GGTGCAACGACAAACGT, and ACTGGCTCGTACGGTGGAGGCGA were used to mutate the first, second, and third AUG start codons, respectively. Reverse primers CGTTCAATTCCGTAATTCTTTCGT and CCTCCACCGTCAGAGCC were used in the first PCR amplification step. For each primer, the mutated nucleotide is represented in boldface. The 5′ and 3′ external primers for the first and second PCR amplification steps were GCTCACTTCATGAGC and GGCGCAGTGGTCTCCTT, respectively. The final PCR products were reinserted into plasmid pFny 209 at the restriction sites NarI (position 2273) and AvrII (position 2691).

Statistical analysis.

Differences in mutation frequencies among viral populations were tested for statistical significance using a one-way analysis of variance (ANOVA) test from StatPlus:mac LE.

RESULTS AND DISCUSSION

Interviral recombinants in progeny L₁L₂F₃.

The parental viruses (Fny- and LS-CMV) and all possible reassortant viruses were simultaneously inoculated on the five hosts studied. Their respective mutation frequencies were used as markers to identify fluctuations in levels of population variation that are specific to particular genomic RNAs. RNA 3 was used as a reporter for the progeny viral populations by analyzing individual cDNA clones. The L₁L₂F₃ progeny were mostly recombinants between RNAs 2 and 3. This recombination event was not considered in calculating mutation frequency, and the recombinants are described in a separate publication (24).

Mutational hot spots and fixed mutations.

Mutational hot spots were observed in the F₃ reporter sequence around base 2000, with the most common mutations in this region being an insertion or a deletion within the stretch of six uracil residues (nucleotides U-2001 to U-2006) and C-to-U substitutions at bases 1951 and 1957. These mutations often occurred with additional mutations in clones from the same plant sample. Therefore, each of them was considered an independent mutation in calculating the mutation frequency. There were no such obvious mutational hot spots observed in the L₃ reporter sequence. However, we observed numerous individual mutants that appeared to be partially fixed in L₃ populations. These mutations do not occur with additional mutations in the same clone, so we think fixation is more likely than a mutational hot spot, but we cannot fully distinguish fixation from a mutational hot spot. Hence, these mutations either were recorded as individual mutants for each clone where they occurred (mutational hot spot) or were counted as one mutation (fixation) irrespective of the number of times they appeared in a population. Data are presented for both alternatives.

Differential genetic variations between LS- and Fny-CMV.

As shown by Schneider and Roossinck (17, 18), different related Sindbis-like plant viruses, including CMV, maintain different levels of population diversity that are controlled by both the viruses and the hosts. To identify the viral determinant(s) for these underlying genetic variations, we first measured the mutation frequencies of LS- and Fny-CMV in five different hosts, with three plant replicates per host. We found that, irrespective of the host tested, the mutation frequencies calculated for Fny-CMV (F₁F₂F₃) were higher than those for LS-CMV (L₁L₂L₃) (Table 1). However, the small size of the treatments (three plant replicates per host) did not allow a full statistical comparison within a host. Nonetheless, by pooling all the numbers for each virus, irrespective of the host, we generated treatments of adequate size for ANOVA analyses. Subsequently, we used this method to compare the viral populations. We found that the calculated mutation frequencies for Fny-CMV (F₁F₂F₃) were significantly higher than those for LS-CMV (L₁L₂L₃), with a P value of <0.001 (ANOVA test). Since only 1 out of 23 control clones (4%) derived from in vitro transcripts contained a single mutation, the background level of experimental mutation frequency, at 4.00 × 10⁻⁵ mutations per nucleotide, was well below the experimental values.

Table 1.

Genetic variation in Fny- and LS-CMV populations in the five hosts tested

Virus	Host	Total no. of mutations/bases sequenced		Mutation frequency (10−4)a
		With mutational hot spotb	With fixed mutationsc	With mutational hot spot	With fixed mutations
Fny-CMV (F1F2F3)	N. benthamiana	21/56,425	20/56,425	3.72	3.54
	Tobacco	28/51,952	28/51,952	5.39	5.39
	Pepper	43/82,735	43/82,735	5.19	5.19
	Squash	42/91,658	42/91,658	4.58	4.58
	Tomato	22/61,814	22/61,814	3.56	3.56
LS-CMV (L1L2L3)	N. benthamiana	5/69,378	5/69,378	0.72	0.72
	Tobacco	3/43,307	3/43,307	0.62	0.69
	Pepper	21/81,934	21/81,934	2.56	2.56
	Squash	10/60,356	10/60,356	1.65	1.65
	Tomato	12/61,124	12/61,124	1.96	1.96

^aSubstitutions, insertions, and deletions were all counted equally in determining mutation frequency. We could not statistically determine the within-host differences in diversity due to the small size of the treatments.

^bThe mutation frequency was calculated assuming redundant mutants were due to a mutational hot spot (see the text for details).

^cThe mutation frequency was calculated assuming redundant mutants were partially fixed (see the text for details).

RNAs 1 and 2 determine the levels of genetic variation.

We measured the mutation frequencies of all the reassorted and recombinant viruses and analyzed the levels of population variation by pooling all the numbers for each virus, irrespective of the host.

The progeny derived from F₃ reassortants/viruses indicated that the mutation frequencies calculated for L₁L₂F₃ were significantly lower than those calculated for F₁F₂F₃, irrespective of the hosts tested (P < 0.001; ANOVA test), indicating that the levels of population variation depend on the nature of the replicase. Furthermore, we found that the calculated mutation frequencies for both F₁L₂F₃ and L₁F₂F₃ were significantly higher than those for L₁L₂F₃ (P < 0.001 and P = 0.002, respectively; ANOVA test). This indicates that the level of population variation depends on the nature of the replicating RNA 1 or RNA 2; the presence of F₁ or F₂ is sufficient to switch L₁L₂F₃ to a high-diversity virus. We obtained similar conclusions regardless of the method used to calculate the mutation frequencies.

Similarly, analysis of the progeny when L₃ was the reporter indicated that, irrespective of the hosts tested, the calculated mutation frequencies of L₁L₂L₃ were significantly lower (P < 0.001 by ANOVA test) than those of F₁F₂L₃ independent of the calculation method. This confirmed the conclusion that the nature of the replicase controls the levels of population variation in plants, irrespective of the replicating RNA 3. However, the role of the replicating RNA 1 or RNA 2 in controlling the genetic diversity when L₃ was the reporter was not as obvious as in the F₃ progeny, because of an apparently strong selection leading to the partial fixation of particular mutants in L₃ reassortant progeny. If these mutants arose due to a mutational hot spot, the role of F₁ or F₂ in controlling the genetic variation is clearer, because calculated mutation frequencies for F₁L₂L₃ or L₁F₂L₃ are significantly higher than those for L₁L₂L₃ (P < 0.001). However, this is not the case when all the redundant mutants present in a given plant population are considered a single partially fixed mutation, particularly for pepper plants.

Figure 2 depicts the specific differences in mutation frequencies among the reassortants. By pooling all the numbers for each virus irrespective of the host (as we did for the ANOVA tests), we found that irrespective of the reporter (F₃ or L₃) and irrespective of the calculation method (hot spots or fixed mutants), the calculated mutation frequencies for L₁L₂ were significantly lower than those for F₁F₂ (Fig. 2), confirming that replicases from different strains of the same virus could generate different genetic variations in plants. Also, when we considered all the fixed mutants in the calculation hot spots, the data showed that the control elements for mutation frequency reside within both RNAs 1 and 2; replacement of RNA 1 of LS-CMV (L₁) with that of Fny-CMV (F₁) led to a dramatic increase in the mutation frequency in F₁L₂L₃ (F₁L₂L₃-HS) compared to the parental LS-CMV (Fig. 2). Similarly, LS-CMV gained a higher mutation frequency when its genomic RNA 2 (L₂) was replaced by F₂ (L₁F₂L₃-HS). This is less perceptible when the redundant mutations are counted as fixed, particularly for reassortant F₁L₂L₃ (F₁L₂L₃-F).

Fig 2.

Genetic variation in populations of parental and reassortant CMVs. The error bars correspond to the standard errors. HS, redundant mutations calculated as a mutational hot spot; F, redundant mutations calculated as fixed mutants.

To consider the host-specific factors affecting population variation, we also pooled the mutation frequency data for each host irrespective of the viruses tested (Fig. 3). This highlighted once more the important role of the host in developing or maintaining viral population diversity, as the mutation frequencies of viruses in pepper (particularly when calculated with redundant mutations as hot spots) are significantly higher than those obtained in N. benthamiana, tobacco, tomato, or squash (Fig. 2).

Fig 3.

Host factors determine the levels of genetic variation in CMV populations. The standard errors of the mean (error bars) of mutation frequencies, calculated irrespective of the viruses tested, emphasized the role of host factors in generating and maintaining viral genetic variations. HS, redundant mutations calculated as a mutational hot spot; F, redundant mutations calculated as fixed mutants; Nb, N. benthamiana; Tob, tobacco; Tom, tomato; Pep, pepper; Squ, squash.

Endornavirus and high mutation frequencies in pepper.

The ability of pepper to support more diverse populations described here was also seen previously (18), and CMV was also shown to have a higher indel mutation rate in pepper (22). The mechanisms responsible are unknown; however, bell peppers, including the Morengo cultivar used in these studies, are naturally infected by an endornavirus, bell pepper endornavirus (BPEV) (25, 26). We tested for a possible link between the presence of BPEV and the level of CMV genetic variation in pepper by comparing CMV mutation frequencies in BPEV-infected pepper plants (+pepper) and a BPEV-free pepper line (−pepper). A single −pepper plant was discovered after planting 137 Morengo pepper seeds. With seeds collected from this −pepper plant, we obtained several −pepper plantlets that we used to test the possible role of BPEV in CMV mutation frequency. We compared mutation frequencies in −pepper and +pepper plants using the parental viruses LS- and Fny-CMV (Table 2); however, the calculated mutation frequencies were not significantly different between −pepper and +pepper plants, indicating that the endornavirus is not involved in the high mutation frequencies observed in pepper.

Table 2.

Genetic variation in various CMVs in pepper and tobacco

Host	Virus	Mutation frequency (10−4)a
+Pepperb	F1F2F3	4.18A
	L1L2L3	2.00B
−Pepperc	F1F2F3	4.46A
	L1L2L3	1.50B
Tobacco	F1F2F3	2.93A
	F1F2aLS2bF3	0.76B
	L1L2L3	0.94B
	L1 MF1HL2L3	0.66B
	L1F2aL2bL3	1.29B

^aSubstitutions, insertions, and deletions were all counted equally in determining mutation frequency. Differences in genetic variations were determined using the ANOVA test (P < 0.05). Statistically significant differences in diversity levels are marked by capital letters next to the mutation frequencies. Mutation frequencies with the same letter are not statistically different.

^bBell pepper infected with bell pepper endornavirus.

^cBell pepper lacking bell pepper endornavirus.

Fine mapping of mutation frequency control in RNAs 1 and 2.

For further fine mapping of the viral control element(s) of genetic variation, we constructed recombinant viruses between portions of LS- and Fny-CMV RNAs 1 and 2 (Fig. 1). We replaced most of the helicase domains of LS-CMV with that of Fny-CMV to form L_1MF_1H, and we replaced the 2b gene of Fny with that of LS-CMV to form F_2aL_2b. These RNAs were mixed with either L₂ and L₃ or L₁ and L₃, respectively, and inoculated onto tobacco for more detailed mapping. The parental viruses F₁F₂F₃ and L₁L₂L₃ were retained as internal controls, since changing conditions can have some effects on ultimate mutation frequencies, although not on comparative mutation frequencies.

If the control elements for high mutation frequencies reside within the helicase-like domain of Fny-CMV, L_1MF_1HL₂L₃ would have triggered an increase in the mutation frequency of LS-CMV. This is not the case, as the mutation frequency of the recombinant virus L_1MF_1HL₂L₃ is not significantly different from that of L₁L₂L₃ in tobacco (Table 2). Therefore, the helicase-like domain does not contain the mutation frequency control elements, suggesting that the first 596 aa of the 1a protein, containing the entire methyltransferase-like domain and 43 N-terminal amino acids of the helicase-like domain (Fig. 1A), may harbor those elements. Unfortunately, we could not directly test this hypothesis, because we could not establish infection with the recombinant virus L_1MF_1HF₂F₃ and we were not able to construct the recombinant F_1ML_1H due to sequence differences between LS and Fny RNAs 1 (the two virus strains share only about 70% nucleotide identity).

The mutation frequency of L₁F_2aL_2bL₃ was not significantly different from that of L₁L₂L₃ (Table 2), indicating that the 2a region was not responsible for the increased mutation frequency of Fny-CMV. However, the replacement of Fny 2b with LS 2b in the context of F₁ and F₃ (F₁F_2aL_2bF₃) resulted in a significant reduction of the mutation frequency of Fny-CMV (Table 2), indicating that the 2b or the 2a/2b overlapping region may control viral genetic variations. Indeed, as the 2b ORF overlaps the N-terminal region of 2a, we cannot exclude the possibility that the control elements of mutation frequency map in that region and not in 2b. To clarify this, we constructed an Fny 2b deletion mutant (F_2aF_Δ2b) in which all 3 AUGs that are found in the N-terminal region of the 2b protein (in frame with each other) were mutated (Fig. 1B). The mutations were designed to keep the amino acid sequence of the 2a protein intact. We compared the mutation frequencies of the Fny-CMV mutated virus, F₁F_2aF_Δ2bF₃, with those of Fny- and LS-CMV in tobacco plants. A few changes were incorporated in our experimental design, because F₁F_2aF_Δ2bF₃ induced milder symptoms than the wild-type Fny-CMV. Infected plants completely recovered from the infection around 8 days p.i., and we could barely detect the virus at 15 days p.i. Therefore, we collected infected leaves and measured the mutation frequencies at 3 and 7 days p.i. instead of 15 days p.i.

Since the calculated mutation frequencies between the two time points chosen were not significantly different for each virus, we averaged them to obtain the mutation frequencies of the virus tested (Table 3). This did not change the significance, as the trend observed before persisted: the mutation frequency of F₁F₂F₃ (1.72 × 10⁻⁴) remained more than 1 order of magnitude higher than that of L₁L₂L₃ (3.10 × 10⁻⁵). We obtained a calculated mutation frequency of 6.26 × 10⁻⁴ for F₁F_2aF_Δ2bF₃, which is more than three times higher than that of F₁F₂F₃ (Table 3). This may indicate that the RNA 2 control elements for genetic variation reside in 2b and not within the 2a C-terminal region. However, when the redundant mutations were considered fixed rather than a mutational hot spot, we obtained a mutation frequency of 2.78 × 10⁻⁴ for F₁F_2aF_Δ2bF₃, which is not statistically different from that of F₁F₂F₃ (Table 3). This was due to the presence of a particular mutant, G₁₉₆₄ to A, which is partially fixed in the population (Table 4). The same mutation was observed in 25 clones out of the 75 clones analyzed for F₁F_2aF_Δ2bF₃ versus 1 clone out of 72 for F₁F₂F₃. Consequently, we concluded that the control elements for mutation frequency reside within the overlapping 2a/2b region, since the 79 aa of the C terminus of the 2a protein are the only differences between the 2a protein sequences of F_2aL_2b and F_2aF_2b. The 2b gene interaction with the host seems to be the driving force controlling the fixation of the mutants in the viral populations.

Table 3.

Genetic variation in Fny-, LS-, and Fny-CMV with 2b deleted in tobacco plants

Virus	Mutation frequency (10−4)a
L1L2L3	0.31A
F1F2F3	1.72B
F1F2aFΔ2bF3-HSb	6.26C
F1F2aFΔ2bF3-Fc	2.78B

^aSubstitutions, insertions, and deletions were all counted equally in determining mutation frequency. Differences in genetic variations were determined using the ANOVA test (P < 0.05). Statistically significant differences in diversity levels are marked by capital letters next to the mutation frequencies. Mutation frequencies with the same letter are not statistically different.

^bHS, redundant mutations calculated as a mutational hot spot.

^cF, redundant mutations calculated as fixed mutants.

Table 4.

Distribution of partially fixed mutations observed in the 3′ NTR of RNA 3 in clones derived from Fny-CMV with 2b deleted, Fny-CMV, and LS-CMV

Virus	Plant	Mutation(s)a		% Mutated clonesb
		3 days p.i.	7 days p.i.
F1F2aFΔ2bF3	1	G1964 to A (7/7)	G1964 to A (9/14)	50 (38/75)
			C1974 to U (3/14)
			Ins G1984 (1/14)
	2	U1958 to C (1/10)	A1997 to G (1/16)
		Ins U2006 (1/10)
	3	G1964 to A (8/13)	G1964 to A (1/15)
			Ins U2006 (3/15)
		Ins U2006 (1/13)	C2064 to U (1/15)
			G2067 to U (1/15)
F1F2F3	1	NA	Ins U2006 (1/12)	18 (13/72)
			Ins G1900 (1/12)
	2	Ins U2006 (3/18)	G1964 to A (1/15)
		U1969 to C (1/18)	Ins U2006 (1/15)
		C1994 to U (1/18)	U2112 to C (2/15)
	3	Ins U2006 (1/12)	Ins G1948 (1/15)
L1L2L3	1	NA	A1992 to G (1/18)	3.61 (3/83)

^aIns, insertion; NA, not applicable. The numbers in parentheses are number identified/number of clones sequenced.

^bThe numbers in parentheses are number of mutated clones/number of clones sequenced.

Closer examination of F₁F_2aF_Δ2bF₃-derived mutants revealed that they entirely clustered within the 3′ NTR of RNA 3. The position of the G₁₉₆₄-to-A mutation on RNA 3 is important, because the crossover that produced recombinant 4 (Table 5) (24) also occurred at nucleotide 1964. This prompted us to analyze further the 3′ NTRs of the viruses tested. We found that the deletion of the 2b gene of Fny-CMV triggered an increase in the percentage of mutated clones in that region from 18% to 50% (Table 4), indicating a possible interaction between the 2b protein and the 3′ NTR.

Table 5.

Distribution of partially fixed mutations in clones derived from parental and reassortant populations in the five hosts tested

Virus	Plant	Mutation(s)a
		Solanaceae				Cucurbitaceae
		N. benthamiana	Tobacco	Pepper	Tomato	Squash
F1F2L3	1	G1958 to A (5/18)	0/20	A1964 to U (15/22)	0/20	0/21
	2	A1915 to G (2/19)	0/20	C1932 to U (5/19)	0/23	0/19
	3	A1953 to G (4/19)	NA	G1958 to A (13/20)	C1932 to U (3/18)	0/13
F1L2L3	1	Ins A1973 (18/24)	Ins A1973 (5/21)	Ins A1973 (14/15)	Ins A1973 (15/17)	0/15
	2	Ins A1973 (9/21)	Ins A1973 (2/18)	Ins A1973 (17/19)	Ins A1973 (9/17)	0/21
	3	Ins A1973 (8/23) G2044 to A (4/23)	NA	Ins A1973 (15/20)	Ins A1973 (4/18)	0/17
L1F2L3	1	Ins A1955 (11/22)	0/24	Ins A1955 (12/22)	0/23	0/23
	2	Ins A1955 (10/22)	0/19	Ins A1955 (8/22)	0/23	0/24
	3	Ins A1955 (14/23)	NA	NA	0/23	0/23
L1L2L3	1	0/23	0/24	0/24	0/24	0/24
	2	0/20	0/16	0/20	0/24	0/24
	3	0/24	NA	0/24	NA	NA
F1F2F3	1	InsU2006 (3/22)	InsU2006 (3/23) C1957 to U (9/23)	InsU2006 (4/24) C1957 to U (8/24)	InsU2006 (1/24)	InsU2006 (5/24)
	2	InsU2006 (2/18)	InsU2006 (4/24) C1957 to U (1/24)	InsU2006 (2/24) C1957 to U (1/24)	InsU2006 (2/24)	InsU2006 (12/24)
	3	InsU2006 (2/17)	NA	InsU2006 (4/23)	NA	InsU2006 (4/24)
F1L2F3	1	InsU2006 (2/23)	InsU2006 (1/24)	InsU2006 (1/24) C1951 to U (1/24)	InsU2006 (2/23)	InsU2006 (2/23)
	2	C1951 to U (5/19)	0/19	InsU2006 (4/23)	A1759 to G (4/24)	InsU2006 (4/23)
	3	InsU2006 (9/24)	NA	InsU2006 (1/23)	DelU2006 (3/22) InsU2006 (3/22)	NA
L1L2F3	1	Rec-1 (6/10)	Rec-1 (6/24) Rec-2 (17/24) Rec-3 (1/24)	Rec-1 (21/21)	Rec-1 (23/24) Rec-4 (1/24)	Rec-1 (18/22)
	2	Rec-1 (13/13)	NA	Rec-1 (22/22)	Rec-1 (19/19)	Rec-1 (18/22)
	3	Rec-1 (21/24)	NA	Rec-1 (4/20) Rec-5 (16/20)	Rec-1 (6/23) Rec-4 (17/23)	Rec-1 (23/23)
L1F2F3	1	InsU2006 (1/15)	0/14	DelU2006 (3/15) InsU2006 (1/15)	InsU2006 (4/18)	InsU2006 (3/20)
	2	InsU2006 (1/14)	InsU2006 (3/19) A1890 to C (4/19)	InsU2006 (1/15)	InsU2006 (1/22)	InsU2006 (1/15)
	3	0/17	NA	DelU2006 (2/15) InsU2006 (2/15)	InsU2006 (3/17)	InsU2006 (1/17)

^aThe numbers in parentheses or alone are number identified/number of clones sequenced. NA, not applicable; Ins, insertion; Del, deletion; Rec, recombinant.

We extended this analysis to all the reassortant viruses used in this study and focused on the distribution of the fixed mutations along the entire sequence of the reporters, encompassing part of the intergenic region, the coat protein, and the 3′ NTR. We found that, irrespective of the host tested, fixed mutations were not observed in L₁L₂L₃ progeny (Table 5), and in other populations, they were mostly constrained to the 3′ NTR, except for two instances, A₁₇₅₉ to G and A₁₈₉₀ to C, which are located within the CP ORF.

With the exception of a few mutants, like Ins-U₂₀₀₆ (mutational hot spot), which was observed in the progeny of several viruses, of the fixed mutants identified were specific to a particular reassortant. For example, mutant Ins A₁₉₇₃ was specific to F₁L₂L₃, and fixed recombinants were observed only in L₁L₂F₃ progeny (Table 5) (24). Since several of these mutants (like mutant Ins A₁₉₇₃) were detected in all the hosts tested (in multiple plants), except in squash, it is unlikely that they are artifacts of the RT-PCR. Hence, these data show that the 3′ NTR of CMV RNA 3 may be under constant evolutionary pressure, and the selection exerted by 2b-host interaction is exacerbated (or more perceptible) in that region. This is not surprising, since the region contains the promoter for minus-strand synthesis of viral RNA during replication, and its primary function involves an interaction with the replicase (i.e., the 1a and 2a proteins, along with host factors) (27). Also, there are functional demands on the 3′ NTR to serve roles typical of cellular mRNA, including the regulation of RNA stability and translation (28–30). Therefore, finding the 3′ NTR to be a hot spot of polymorphism fits very well in a model in which the domains controlling viral genetic variation reside within the 1a and 2a proteins and involve the 2b gene as a modulator of variability. The existence and the partial fixation of a whole range of CMV mutants in Solanaceae (N. benthamiana, tobacco, pepper, and tomato) that were not found in Cucurbitaceae (squash) (Table 5) are important findings with significant implications for viral-disease evolution. Since host shifting is thought to play a major role in disease emergence (31), identifying the factors that drive the specific adaptation of these mutants to these different hosts (Solanaceae and Cucurbitaceae) may help us understand the spread of new diseases.

Previous studies demonstrated that the diversity levels of viral populations are controlled by both the virus and the host (18), and this was attributed to either differences in polymerase fidelity or different bottlenecks that limited the accumulation of diversity in different hosts. In fact both phenomena were observed for CMV (22, 32–34). The capacity for host-specific adaptation was also evaluated for other virus-host systems (35). This study demonstrated significant differences in the capacities of West Nile virus and St. Louis encephalitis virus to adapt to mosquitos in vivo. This was also attributed to bottlenecks. Here, we presented several new findings to fill the gaps in our understanding. (i) We found that, irrespective of the reporter, L₃ or F₃, the calculated mutation frequencies were significantly lower when the replicase was derived from LS-CMV than when it was from Fny-CMV. This suggests differences in the abilities of the two parental viruses either to generate diversity through replicase error or to maintain a large quasispecies population. Hence, differences in genetic diversity in plants between strains of the same virus could be due to a difference in replicase fidelity. (ii) We determined that RNA 1 and RNA 2 could act independently of each other in modulating virus diversity. Fine mapping with recombinant viruses indicated that the variation control domains map to the first 596 aa of the 1a protein and to the last 79 aa of the 2a protein. There is a 79-aa difference between Fny- and LS-CMV in the first 596 1a amino acid sequences, randomly distributed along the domain. Therefore, further genetic mapping by site-directed mutagenesis may be difficult. (iii) Recently, it was shown that one of the domains identified as a genetic diversity determinant here, the 2a/2b overlapping region, also controls symptom induction and viral-RNA accumulation in CMV (36). Studies with poliovirus, a human Enterovirus (family Picornaviridae), clearly established a direct link between genetic diversity and viral pathogenesis (37, 38). Similarly, our data indicate a positive correlation between virulence and levels of genetic variation, as LS-CMV, which is less pathogenic, has low genetic variation, whereas the highly pathogenic Fny-CMV displayed higher genetic diversity. Subgroup I strains are in general more virulent than subgroup II strains, and studies have suggested that this differential virulence is mediated by the 2b gene (39). The 2b gene also modulates the phenomenon of fixation described here. For example, the G₁₉₆₄-to-A mutant was observed only once in Fny progeny out of 72 clones sequenced, but it appeared in 25 clones out of the 75 clones sequenced when Fny 2b was not expressed (Table 4). This indicates that this particular mutant can be generated by the Fny replicase, but it is probably under a purifying selection exerted by the 2b-host interaction, limiting its accumulation in the viral population.

(iv) Our attempt to associate BPEV with the high mutation frequencies calculated in pepper failed, as we could not establish a positive correlation between the elevated mutation frequency and the presence of the persistent virus. Although significant differences in bottlenecks were found between several hosts (33), they also did not correlate with host-specific differences in mutation frequencies. We previously found that the rate of indel mutations was much higher in pepper than in tobacco (22), but as we do not have data on the substitution rates (usually the predominant mutations) in these hosts, the mechanism(s) by which CMV develops high genetic variations in pepper remained an open question. However, in light of the host-dependent 2b-driven selection described here, high genetic variation in pepper could be due in part to differential selection pressures imposed by host-2b interactions. Indeed, if we considered the redundant mutations hot spots rather than fixed mutants in calculating mutation frequencies, CMV variation in pepper was in the range of the other hosts, squash and tomato (Fig. 3). It was shown that the 2b gene promotes the selection of interviral recombinants in planta (40). We have also observed similar interviral recombinants in CMV progeny. However, we demonstrated that the 2b gene was partially involved in their selection only in the context of a homologous RNA 1 and RNA 2 (24). These examples further showed that CMV 2b is undoubtedly a driving force in the strong selective pressure exerted on the mutant cloud generated in plants and emphasize that this selection depends on environmental factors (virus strains and hosts).