Pleiotropic Effects of Resistance-Breaking Mutations on Particle Stability Provide Insight into Life History Evolution of a Plant RNA Virus

ABSTRACT

In gene-for-gene host-virus interactions, virus evolution to infect and multiply in previously resistant host genotypes, i.e., resistance breaking, is a case of host range expansion, which is predicted to be associated with fitness penalties. Negative effects of resistance-breaking mutations on within-host virus multiplication have been documented for several plant viruses. However, understanding virus evolution requires analyses of potential trade-offs between different fitness components. Here we analyzed whether coat protein (CP) mutations in Pepper mild mottle virus that break L-gene resistance in pepper affect particle stability and, thus, survival in the environment. For this purpose, CP mutations determining the overcoming of L³ and L⁴ resistance alleles were introduced in biologically active cDNA clones. The kinetics of the in vitro disassembly of parental and mutant particles were compared under different conditions. Resistance-breaking mutations variously affected particle stability. Structural analyses identified the number and type of axial and side interactions of adjacent CP subunits in virions, which explained differences in particle stability and contribute to understanding of tobamovirus disassembly. Resistance-breaking mutations also affected virus multiplication and virulence in the susceptible host, as well as infectivity. The sense and magnitude of the effects of resistance-breaking mutations on particle stability, multiplication, virulence, or infectivity depended on the specific mutation rather than on the ability to overcome the different resistance alleles, and effects on different traits were not correlated. Thus, the results do not provide evidence of links or trade-offs between particle stability, i.e., survival, and other components of virus fitness or virulence.

IMPORTANCE The effect of survival on virus evolution remains underexplored, despite the fact that life history trade-offs may constrain virus evolution. We approached this topic by analyzing whether breaking of L-gene resistance in pepper by Pepper mild mottle virus, determined by coat protein (CP) mutations, is associated with reduced particle stability and survival. Resistance-breaking mutations affected particle stability by altering the interactions between CP subunits. However, the sense and magnitude of these effects were unrelated to the capacity to overcome different resistance alleles. Thus, resistance breaking was not traded with survival. Resistance-breaking mutations also affected virus fitness within the infected host, virulence, and infectivity in a mutation-specific manner. Comparison of the effects of CP mutations on these various traits indicates that there are neither trade-offs nor positive links between survival and other life history traits. These results demonstrate that trade-offs between life history traits may not be a general constraint in virus evolution.

INTRODUCTION

A major goal in the study of plant virus evolution is understanding resistance breaking, i.e., the appearance and increase in frequency within virus populations of genotypes able to infect, multiply in, and cause disease in otherwise resistant host genotypes. Resistance breaking has important consequences for the sustainable control of viral diseases in agricultural systems, as growing resistant crop cultivars is a highly effective, target-specific, and environmentally friendly strategy for disease control. However, the protection conferred by resistance is often not durable, as selection for resistance-breaking genotypes in virus populations eventually renders resistance inefficient (1–4). Many plant-virus interactions conform to the gene-for-gene (GFG) model, in which the recognition of viral proteins by host proteins encoded in resistance genes (R) triggers a defense reaction limiting virus multiplication to the infection site (5–8). Mutations in these viral proteins that impair their recognition by R proteins result in resistance breaking. Under the GFG model, consecutive events of resistance breaking eventually result in the virus's capacity to infect all host genotypes (9). Resistance breaking in GFG interactions is thus a case of host range expansion. Pleiotropic effects of resistance-breaking mutations may result in across-host fitness trade-offs, known as resistance-breaking costs, so that the resistance-breaking genotypes are less fit than the non-resistance-breaking ones in susceptible hosts. The relevance of resistance-breaking costs for plant-virus coevolution and the durability of crop resistance has led to important efforts aimed at their detection and at understanding the mechanisms generating them, and the study of resistance breaking by plant viruses has contributed significantly to understanding plant virus evolution (1, 10). Evidence of resistance-breaking costs in plant viruses is abundant, but it derives almost exclusively from the analysis of the within-host multiplication component of virus fitness (11–17). In order to fully understand the evolution of resistance breaking, it is necessary to consider potential costs on other fitness components, as the evolution of parasites, including viruses, may be constrained by conflicting trade-offs between different fitness components (18–20). Life history evolution theory predicts that one such trade-off would be between increased reproduction and extended survival. For viruses, the reproduction-survival trade-off would be between the intracellular (within-host multiplication) and extracellular (survival outside the host) stages of the virus life cycle (19). Why this trade-off should arise in viruses is unclear, as virus survival outside the host is linked to particle stability and there is no obvious mechanism linking particle stability and within-host multiplication (19). A link between survival and virulence has also been proposed. In directly transmitted parasites, virulence is assumed to be traded against transmission (18). Parasites transmitted through the environment need not be constrained by this trade-off, and a sit-and-wait strategy, in which particle survival in the environment exceeds the infectious period of the host, may result in a positive correlation between virulence and survival (21). This hypothesis has been called “curse of the pharaoh,” and the conditions for an increase in virulence being linked to an increase in survival have been explored through model analyses (22, 23). Hypotheses linking survival and other viral life history traits have seldom been tested for viruses infecting eukaryotes (19) and, to our knowledge, have never been tested for plant-infecting viruses. We do so in this work, focusing on Pepper mild mottle virus (PMMoV), an important pathogen of pepper crops.

The present work builds on previous results from our group on resistance-breaking costs, and their generating mechanisms, in the interaction between virus species in the genus Tobamovirus and pepper (Capsicum annuum L.) (12, 24, 25). Tobamoviruses have a single-stranded, monopartite genome encapsidated into rod-shaped particles, whose structure has been resolved by X-ray crystallography for several species (26–28). Pepper-infecting tobamoviruses are transmitted horizontally through plant-to-plant contact or vertically through the seed. After the pepper crop is harvested, infected plant tissue degradation liberates virus particles to the soil, where infectious particles can survive for several months, being a major source of primary inoculum for epidemics (29–31). Tobamovirus interaction with pepper conforms to the GFG model, being determined by the interaction between proteins encoded by alleles at the pepper L resistance locus (32, 33) and the virus coat protein (CP) (34–37). Recognition of the CP by L alleles elicits resistance, expressed as the development of necrotic local lesions (NLL) that limit virus multiplication to infection sites. According to the capacity to elicit resistance or to infect pepper genotypes carrying the different L resistance alleles, tobamovirus species and genotypes are classified into pathotypes. Plants homozygous for allele L⁺ are susceptible to all described pathotypes (P₀, P₁, P_1,2, P_1,2,3, and P_1,2,3,4), L¹/− plants are resistant only to pathotype P₀, L²/− plants are resistant to pathotypes P₀ and P₁, L³/− plants are resistant to pathotypes P₀, P₁, and P_1,2, and so on. Pathotype P_1,2,3,4 is able to infect all known L-gene host genotypes. While P₀ and P₁ pathotype isolates belong to different tobamoviral species, such as Tomato mosaic virus, Tobacco mild green mosaic virus (TMGMV), or Paprika mild mottle virus, P_1,2, P_1,2,3, and P_1,2,3,4 isolates are all genotypes of PMMoV (38–40).

Conclusive evidence of resistance-breaking costs, expressed as reduced virus multiplication in susceptible host genotypes, was obtained from the analysis of Spanish field isolates of TMGMV (P₀ pathotype) and PMMoV (P_1,2 and P_1,2,3 pathotypes) and of PMMoV CP mutants engineered into infectious cDNA clones (12, 25). Resistance-breaking costs were due to pleiotropic effects of resistance-breaking mutations and depended on the specific genotype of the susceptible host and the type of infection, single or mixed (25). Also, the analysis of P_0, P_1,2, and P_1,2,3 pathotype field isolates showed that the stability of their particles differed and that the capacity to survive in the soil was positively correlated with particle stability (24). Particle stability ranked P_1,2 > P_1,2,3 > P₀, which indicated that resistance breaking may involve costs that are expressed in fitness components other than within-host multiplication. However, the various P_0, P_1,2, and P_1,2,3 pathotype field isolates analyzed differed in their CP amino acid sequences in various positions, in addition to those reported as determinants of resistance breaking. Thus, previous work did not establish whether differences in particle stability were due to resistance-breaking mutations. Here we analyzed whether resistance-breaking CP mutations have pleiotropic effects on particle stability, which would affect survival in the environment. Analyses focused on the mutations in the PMMoV CP that have been reported as determinants of the conversion of pathotype P_1,2 to pathotypes P_1,2,3 and P_1,2,3,4 (34, 38, 41–43). We also quantified the within-host multiplication and the virulence of the various virus genotypes and analyzed the relationship of these traits to particle stability as a proxy for survival. Last, for a subset of genotypes we analyzed whether differences in particle stability translated into differences in infectivity, another component of the virus fitness. The results show that resistance-breaking mutations may result in altered survival due to altered virion stability and that mutations also translated into differences in within-host multiplication, virulence, and infectivity. However, the sense and magnitude of these effects depended on the specific resistance-breaking mutation, and there was no evidence of trade-offs between survival and reproduction or of a link between survival and virulence.

RESULTS

Kinetics of virus particle disassembly in vitro.

The kinetics of particle disassembly of wild-type genotype MG (MG-WT) (pathotype P_1,2) and its derived CP mutants of the P_1,2,3 pathotype [MG-(M138N), MG-(T43K+D50G), and MG-(L13F+G66V)] and of the P_1,2,3,4 pathotype [MG-(A86G)] were analyzed both in 0.1 M Tris-HCl (pH 8.75) and in 6 M urea. The single mutants from which the two double mutants of the P_1,2,3 pathotype originated were also analyzed. Representative electrophoreses of disassembly in 6 M urea (pH 7.4) are shown in Fig. 1. The electrophoreses shown illustrate how RNA mobility increases from that of RNA encapsidated in the assembled virus particle (C, time zero controls) to that of free RNA after full disassembly (as for the purified genomic RNA, indicated as RNA) as particles disassemble during treatment (e.g., genotype A86G) or how mobility does not change because particle disassembly does not occur (e.g., genotype L13F+G66V). Note that the ethidium bromide fails to stain the RNA in particles of genotype T43K and only weakly stains the RNA in particles of genotype T43K+D50G. Under the three assayed conditions, the kinetics of disassembly of genotype TS were the same as for MG, and those of TS-(M138N), TS-(L13F+G66V), and TS-(T43K+D50G) were the same as for MG-(M138N), MG-(L13F+G66V), or MG-(T43K+D50G), respectively (not shown), indicating that the kinetics depended only on the specific CP mutations. Thus, in the rest of this section we refer to CP mutations regardless of the genetic background.

FIG 1.

Disassembly of virus particles of PMMoV MG-WT (WT) and coat protein mutants in 6 M urea, pH 7.4. Each panel indicates the incubation time in minutes. RNA, electrophoretic mobility of genomic RNA, C, electrophoretic mobility of virus particles at incubation time zero (negative control).

The amount of RNA in virus particles was quantified by densitometry, and exponential curves describing the disassembly kinetics were adjusted on the data of at least three replicate experiments (Fig. 2). Note that no densitometry, and hence no disassembly curves, could be obtained for T43K, as encapsidated RNA did not stain. Curves for T43K+D50G are partly artifactual, showing an apparent increase in encapsidated RNA with incubation time, again due to poor staining of encapsidated RNA at early times followed by a more efficient staining as the particle structure relaxes upon treatment, although particles of this genotype do not disassemble under any assayed conditions (compare Fig. 1 and 2; data not shown). Comparison of the slopes of linearized disassembly kinetics curves showed that particle stability in Tris-HCl (pH 8.75) was ranked as T43K+D50G > D50G = G66V = L13F+G66V > L13F = A86G = WT > M138N (P ≤ 0.018) (Fig. 2A), particle stability in 6 M urea (pH 7.4) was ranked as T43K+D50G = G66V > D50G = L13F = L13F+G66V = WT > M138N > A86G (P ≤ 0.023) (Fig. 2B), and particle stability in 6 M urea (pH 10) was ranked as T43K+D50G = D50G > G66V = L13F+G66V = WT > L13F = M138N = A86G (P ≤ 0.030) (Fig. 2C).

FIG 2.

Kinetics of disassembly of PMMoV MG-WT (WT) and coat protein mutants in 0.1 M Tris-HCl (pH 8.75) (A), 6 M urea (pH 7.4) (B), and 6 M urea (pH 10.0) (C). Disassembly curves are presented as the percentage of encapsidated RNA over a period of 90 to 120 min. Curves were adjusted using data from at least three replicated assays and according to exponential functions of the form y = ab^x.

The results from the in vitro disassembly experiments indicate that resistance-breaking mutations in the CP responsible for the conversion of pathotype P_1,2 to pathotype P_1,2,3 or P_1,2,3,4 have pleiotropic effects on virus particle stability, but the magnitude and sense of these effects depended on the specific mutations rather than on the pathotype.

Analysis of molecular interactions among coat protein subunits in virus particles.

The structure of the CP of isolate U2 of TMGMV (U2-TMGMV) was determined at a 3.5-Å resolution (PDB entry 1VTM) in 1992 (27), whereas that of isolate U1 of Tobacco mosaic virus (TMV) (U1-TMV) was refined at a 2.4-Å resolution (PDB entry 1EI7) a few years later (44). The structures of the two coat proteins are virtually indistinguishable. Since the 1VTM PDB file includes the 49 transformation matrices needed to construct models of biological helical assemblies, this U2-TMGMV structure was used as reference for our analyses of interactions between subunits in the virus particles of the various PMMoV CP sequence variants. In spite of sequence differences between CPs of TMGMV and PMMoV, the modeled geometries for PMMoV variants superimposed indistinguishably on that of TMGMV (Fig. 3A). Intermolecular interactions between a reference CP subunit (A in Fig. 3B) and its eight neighbor subunits (L, R, U, D, T, V, C, and E in Fig. 3B) in a core nonamer (Fig. 3B and C) were calculated for each of the CP WT and mutants. The results found upon classifying interactions as contacts, H bonds, and ionic pairs are summarized in Table 1.

FIG 3.

(A) Ribbon diagram of the structure of the coat protein subunit of PMMoV MG-WT and the indicated coat protein mutants modeled upon the crystal structure of U2-TMGMV and indicating the position of the genomic RNA. Root mean square deviations computed for backbone atoms are indicated. (B and C) Molecular surface of a three-turn section of the nine virus particle coat protein subunits that integrate a core nonamer used to analyze protein-protein interactions. Panel B shows a left, side view, and panel C shows a right, top view.

TABLE 1.

Classification of intersubunit contacts, H bonds, and ionic pairs between central subunit A and neighbor subunits^a

Virus genotype	Total no./specific type
	Contactsb	H bondsc	Ionic pairsd
WT	48/13R (6s, 7w), 19L (10s, 9w), 8D (4s, 4w), 6U (2s, 4w), 2T (2s)	18/2R (2w), 10L (3s, 3m, 4w), 4D (1s, 3w), 2U (1m, 1w)	6/3R (3.30, 3.46, 3.82), 3L (3.30, 3.46, 3.82)
M138N	32/12R (4s, 8w), 18L (7s, 11w), 1U (1s), 1E (1w)	10/2R (1s, 1w), 6L (1s, 2m, 3w), 1T (1m)	6/3R (3.36, 3.42, 3.91), 3L (3.36, 3.42, 3.91)
A86G	50/22R (14s, 8w), 18L (10s, 8w), 5U (2s, 3w), 1C (1s), 4E (1s, 3w)	17/6R (4m, 2w), 7L (1s, 4m, 2w), 2U (1m, 1w), 2C (1m, 1w)	8/2R (3.40, 3.75), 2L (3.40, 3.75), 2U (3.37, 3.57), 2C (3.37, 3.57)
L13F+G66V	53/19R (13s, 6w), 15L (12s, 3w), 2D (1s, 1w), 6U (3s, 3w), 3T (2s, w), 8E (4s, 4w)	16/5R (4s, 1m), 2L (1s, 1m), 4D (1s, 3w), 4U (1s, 1m, 2w), 1E (1m)	0
T43K+D50G	54/15R (8s, 7w), 28L (13s, 15w), 2D (2w), 5U (2s, 3w), 3T (3s), 1E (1w)	24/8R (1s, 5m, 2w), 10L (1s, 6m, 3w), 2D (1s, 1m), 3U (2s, 1m), 1T (1m)	10/5R (2.55, 3.00, 3.28 3.60, 3.66), 5L (2.55, 3.00, 3.28 3.60, 3.66)
T43K	54/15R (8s, 7w), 28L (13s, 15w), 2D (2w), 5U (2s, 3w), 3T (3s), 1E (1w)	24/8R (1s, 5m, 2w), 10L (1s, 6m, 3w), 2D (1s, 1m), 3U (2s, 1m), 1T (1m)	11/5R (2.55, 3.00, 3.28, 3.60, 3.66), 5L (2.55, 3.00, 3.28, 3.60, 3.66), 1T (3.11)
D50G	45/14R (9s, 5w), 23L (12s, 11w), 2D (2w), 3U (1s, 2w), 3T (3s)	24/8R (1s, 5m, 2w), 10L (1s, 6m, 3w), 2D (2m), 4U (3m, 1w)	4/2R (3.00, 3.28), 2L (3.00, 3.28)

^aThe neighbor subunits are T, U, V, L, R, C, D, and E, as labeled in Fig. 3B.

^bContacts are considered strong (s) if the interatomic distance d is ≤3.2 Å and weak (w) if d is >3.2 Å.

^cThe strength of A-H—B hydrogen bonds is classified in terms of the d(A—B) distance as strong (s) if d = 2.2 to 2.5 Å, moderate (m) if d = 2.5 to 3.2 Å, and weak (w) if d = 3.2 to 3.7 Å.

^dAttraction energy in ionic pairs is inversely proportional to the interatomic distance; values in parentheses are in angstroms.

In all cases, the great majority of contacts involve intradisk side interactions (A and either R or L), while there are fewer contacts involving interdisk axial interactions (A and either D or U) and rather scarce diagonal interactions (A and T, V, C, or E). For H bonds, side interactions dominate again, but the number of axial links is quite significant. The much stronger attractive ionic pairs refer almost exclusively to side interactions (details of the atoms involved in these ionic attractions are available from the authors upon request). Full intersubunit interactions (Table 1) indicate that the M138N mutant would have viral particles less stable than those of the WT, while the A86G mutant and, to a lesser degree, the L13F+G66V double mutant, show an overall balance similar to that of the WT. All the interactions considered favor the stability of the viral particle of the T43K+D50G double mutant with respect to the WT structure. Not only the number of strong H bonds but, above all, the number of ionic pairs at distances shorter than 3.3 Å are greater in the T43K+D50G mutant than in the WT. Given the markedly different features of interactions in this double mutant, we analyzed separately the contributions from mutations in positions 43 and 50 with the same methodology. The results (Table 1) reveal that the dominant effect on intersubunit interactions comes from the change T43K, whereas the D50G mutation implies fewer side contacts and especially fewer ionic side attractions.

Analysis of electrostatic features of coat protein subunits in virus particles.

Nearly all the charged amino acids are conserved between TMV and PMMoV (not shown). Thirteen out of the 15 acidic residues in the CP of TMV are present in PMMoV, even with similar side chain conformations. Only 2 aspartates of TMV (positions 64 and 66) are absent in PMMoV. Furthermore, the essential acidic residues (Caspar carboxylates) that might play a major role in the disassembly of the virus particle in TMV (45) are conserved, with the only replacement that of E50 (TMV) by D50 (PMMoV). For basic residues, 11 out of the 13 present in the CP of TMV are also present in PMMoV, although with some differences in side chain conformations. Again, only 2 arginines in TMV (positions 46 and 141) are absent in PMMoV. Due to this conservation of charged amino acids, similar electrostatic features are expected in the CP structures of TMV and PMMoV, although subtle differences may affect binding to the RNA, which is deeply embedded and spirally bound in the inner side of the viral particle. In fact, the electrostatic potential mapped onto the surfaces of RNA-interacting regions in TMV shows distinctive features. Within disks, one side of the structure has a regular distribution of guanine-binding pockets with positive potential surrounded by regions of negative potential, while the other side features a continuous RNA-binding large groove showing a much more positive potential (not shown). Since the CP interacts with negatively charged phosphates in RNA, these positive-potential regions show the electrostatic complementarity underlying the interactions with RNA.

Comparing electrostatic potentials in WT PMMoV and mutants reveals some differences (Fig. 4). The region of guanine-binding pockets (upper row in Fig. 4) is rather similar in all cases, with the single exception of A86G, which has a nearly neutral potential (white in Fig. 4). In contrast, the intensely positive RNA-binding groove region (lower row in Fig. 4) displays a different electrostatic pattern in the WT and the T43K+D50G mutant on one side and the M138N and A86G mutants on the other side, with the L13F+G66V mutant somewhat between. Summarizing, compared with the WT CP, the electrostatic potentials on surfaces suggest a slightly stronger binding to RNA in the T43K+D50G double mutant, a slightly weaker binding in the L13F+G66V double mutant, and the weakest binding in the M138N and A86G single mutants.

FIG 4.

Poisson-Boltzmann electrostatic potential mapped onto the molecular surface of disk coat protein monomers in the viral particle of PMMoV MG-WT or the indicated coat protein mutants, showing regions directly involved in binding to RNA. The structures in the upper row should locate just above those in the lower row in the viral particle upon 180° rotation around a horizontal axis.

Analysis of within-host fitness and virulence of WT PMMoV and CP mutants.

To analyze whether CP mutations in the eight PMMoV single and double mutants derived from MG-WT, previously analyzed for particle stability, were associated with altered fitness in susceptible hosts, assays were done in C. annuum cv. Dulce Italiano (L⁺/L⁺). The assays followed a random-block design with 9 virus genotype treatments plus one mock-inoculated control, with 8 replicated plants per treatment/control. Plants were harvested at 21 days postinoculation (dpi), virus RNA was quantified in systemically infected leaves (data not shown), and fitness was estimated as described in Materials and Methods. Generalized linear model (GzLM) analysis showed that genotype was a factor determining fitness (Wald χ²_8,70 = 45.6, P < 0.001). Fitness data relative to MG-WT are shown in Table 2. The A86G, G66V, and T43K mutants were significantly less fit than the parental MG-WT, while the rest of the mutants were as fit as MG-WT (Table 2). In addition, pairwise comparisons of mutant fitness relative to that of MG-WT showed that the fitness of the A86G, T43K, and G66V mutants was similar to (P > 0.109) and lower than (P < 0.045) the fitness of the M138N, T43K+D50G, D50G, L13F+G66V, and L13F mutants, with no significant differences among the fitness of this second group of mutants (P > 0.080).

TABLE 2.

Magnitude and sense of fitness differences between mutant and parental genotypes

Virus genotype	Wma	Wpb	Wm − Wp	Pc
MG-M138N	1.197	1.207	−0.010	0.650
MG-A86G	1.142	1.207	−0.066	0.017
MG-(L13F+G66V)	1.013	1.029	−0.015	0.477
MG-L13F	1.042	1.029	0.013	0.658
MG-G66V	0.925	1.029	−0.104	0.001
MG-(T43K+D50G)	0.988	1.029	−0.041	0.312
MG-T43K	0.926	1.029	−0.102	0.001
MG-D50G	1.060	1.029	0.031	0.259

^aW_m, average fitness of the mutant in each host at 21 dpi.

^bW_p, average fitness of the parental genotype MG-WT.

^cFitness differences significant at a P value of ≤0.05 level according to an LSD analysis are underlined.

To analyze the possible effect of the above set of mutations on virulence, virulence was quantified as the effect of infection on plant biomass (data not shown), and the derived virulence values are presented in Table 3. General linear model (GLM) analyses showed that genotype was a significant factor determining virulence (F_8,71 = 7.248, P < 0.001). Pairwise comparisons showed that genotype A86G was more virulent than the other eight (P < 0.001), L13F was more virulent than T43K, D50G, and T43K+D50G (P < 0.033), and virulence did not differ for MG-WT, M138N, L13F+G66V, G66V, T43K+D50G, T43K, and D50G (P > 0.115).

TABLE 3.

Virulence of nine PMMoV genotypes in C. annuum cv. Dulce Italiano (L⁺/L⁺)

Virus genotype	Virulencea
MG-WT	0.157 ± 0.026
MG-M138N	0.105 ± 0.034
MG-A86G	0.505 ± 0.032
MG-(L13F+G66V)	0.130 ± 0.054
MG-L13F	0.224 ± 0.052
MG-G66V	0.107 ± 0.037
MG-(T43K+D50G)	0.061 ± 0.084
MG-T43K	0.057 ± 0.072
MG-D50G	0.055 ± 0.054

^aVirulence was computed as V = 1 − (P_i/P_m), where P_i is the dry weight of the total biomass of each infected plant and P_m the mean dry weight of mock-inoculated plants. Data are means ± standard errors from eight replicated plants.

Thus, the nine PMMoV genotypes ranked differently for within-host fitness and virulence and for each of these two traits and particle stability.

Analysis of the infectivity of PMMoV genotypes which differ in particle stability.

It has been reported that TMV mutants with particles that are hyperstable to disassembling conditions had a reduced infectivity (46). To test whether the observed differences in particle stability among PMMoV CP mutants translated into quantitative changes in infectivity, the infectivities of the two mutants with lower and higher resistance to disassembly, MG-(M138N) and MG-(T43K+D50G), respectively, were compared with that of MG-WT in a local-lesion assay in the resistant host Nicotiana tabacum cv. Xanthi-nc. The number of NLL per half leaf was used to estimate infectivity. For all three genotypes, the number of NLL per half leaf varied linearly with virus concentration in the inoculum (r > 0.954, P < 1 × 10⁻⁴) (Fig. 5). The regressions of number of NLL per half leaf to inoculum concentration significantly differed among genotypes both in the intercept (F_2,5 = 42.74, P < 1 × 10⁻⁴) and in the slope (F_2,5 = 54.49, P < 1 × 10⁻⁴). Comparisons between any two genotypes were also significant (F_1,3 ≥ 8.50, P ≤ 0.004), and slopes indicated that infectivity was ranked as MG-(M138N) > MG-(T43K+D50G) > MG-WT. Differences in the number of local lesions increased with increased concentration of the inoculum. When the data for the maximum inoculum concentration (18 μg/ml) were eliminated from the analysis, regressions still differed significantly, indicating that infectivity was ranked as MG-(M138N) = MG-(T43K+D50G) > MG-WT. Thus, for these three genotypes, infectivity and particle stability ranked differently.

FIG 5.

Infectivities of PMMoV MG-WT, MG-(M138N), and MG-(T43K+D50G). Infectivity was estimated by the number of necrotic local lesions (y axis) caused in N. tabacum Xanthi-nc by each viral genotype at different inoculum concentrations (x axis). The represented linear regressions of infectivity on inoculum concentration are y = −6.51 + 17.45x for MG-WT (blue symbols), y = −3.29 + 25.78x for MG-(M138N) (red symbols), and y = 4.57 + 21.41x for MG-(T43K+D50G) (green symbols).

DISCUSSION

The analysis of resistance breaking by plant viruses offers an excellent opportunity to address different issues of virus evolution. There is abundant evidence of resistance-breaking fitness costs in plant viruses, which are expressed as reduced multiplication in susceptible host genotypes compared with the wild-type virus genotypes (11, 12, 14, 16, 25, 47–49). Evidence also indicates that a major cause of across-host fitness trade-offs in RNA viruses (1, 50–52) is antagonistic pleiotropy (53), i.e., opposite phenotypic effects of mutations in different environments. When resistance-breaking mutations occur in structural proteins, pleiotropic effects on virion stability could be expected in addition to, or instead of, effects on virus multiplication, thus potentially affecting survival, a key component of fitness (54). The effect of survival on virus evolution or, more generally, on parasite evolution is an underexplored subject (but see references 21, 46, 55–58), despite trade-offs being expected to occur between different fitness components of parasites (18–20) and despite a positive correlation between survival and virulence for parasites transmitted through the environment being predicted (21–23).

The mutations that determine the breaking of L-gene resistance in Capsicum by tobamoviruses occur in the CP (34–37), so it could be hypothesized that they have effects on particle stability. It has been shown that tobamovirus field isolates of pathotypes P₀, P_1,2, and P_1,2,3 differ in particle stability, which was ranked as P_1,2 > P_1,2,3 > P₀. Particle stability was positively correlated with survival of infectious particles in the soil for extended time periods of up to 6 months, which is usually more than the duration of the crop cycle and hence more than the infectious period of infected hosts (24). These results demonstrated that increased resistance-breaking capacity from P₀ to P_1,2 was not associated with a cost on survival but that there was a cost from P_1,2 to P_1,2,3. The field isolates of each pathotype assayed by Fraile et al. (24) differed in various amino acid positions in the CP, so that different haplotypes were described. About 30% of the amino acid positions of the CPs of P₀ isolates (TMGMV) differ from those of P_1,2 and P_1,2,3 isolates (PMMoV). Hence, that study could not determine whether changes in particle stability were due specifically to resistance-breaking mutations. This question was analyzed in the present work by comparing the effect of reported CP mutations responsible for the conversion of PMMoV pathotype P_1,2 to pathotypes P_1,2,3 and P_1,2,3,4 after engineering these mutations in the CP gene of an infectious clone of pathotype P_1,2 derived from a field isolate, P84/8 of CP haplotype 8, in the study by Fraile et al. (24). The results demonstrate that resistance-breaking mutations have effects on particle stability, as estimated by the in vitro kinetics of disassembly under conditions of high pH and high urea concentration. These assay conditions were chosen because they have been used extensively in the past to analyze the disassembly kinetics of TMV, providing useful information on its kinetics and mechanisms (46), not because they mimic natural soil conditions. Still, as noted above, stability under these assay conditions positively correlates with tobamovirus particle survival in the soil (24). Considering together the data from the three assayed conditions for disassembly, the stability of the different mutants analyzed depended on the specific mutation and was ranked as MG-(T43K+D50G) (P_1,2,3) > MG-(L13F+G66V) (P_1,2,3) ≥ MG-WT (P_1,2) ≥ MG-(A86G) (P_1,2,3,4) ≥ MG-(M138N) (P_1,2,3).

The analysis of inter-CP subunit interactions in the modeled virus particle agrees with and explains the results of the in vitro disassembly. Each CP subunit of the low-stability mutant MG-(M138N) establishes fewer H bonds and contacts with the neighbor subunits than MG-WT, and, importantly, this decrease involves mostly axial interactions. Mutant MG-(L13F+G66V) is slightly more stable than MG-WT in Tris-HCl (pH 8.75), in spite of forming fewer ionic pairs and fewer H bonds with neighboring units, although it forms more and stronger axial H bonds than MG-WT. These data highlight the relevance of axial interactions in disassembly, as proposed in classical studies of TMV particle disassembly. Disassembly proceeds in an orderly manner, mainly from the end of the rod corresponding to the 5′ end of the virus RNA (59–61), suggesting that it is due primarily to the destabilization of axial interactions among CP subunits. The most stable mutant, MG-(T43K+D50G), is able to establish more ionic pairs, more H bonds, and more contacts with its neighbor subunits than MG-WT. Interestingly, the excess of ion pair interactions in MG-(T43K+D50G) with respect to MG-WT (10 versus 6), all in side interactions, may explain the poor access of ethidium bromide to the encapsidated RNA. This hypothesis is supported by the fact that the encapsidated RNA is also nonaccessible to ethidium bromide in the single mutant MG-(T43K), which forms the same number of side ionic pair interactions, while the RNA is accessible in MG-(D50G), which forms only 4 ionic pairs. The differences in CP-RNA interactions in the virus particle shown by Poisson-Boltzmann (PB) analyses do not seem to be large enough to translate into noticeable differences in the in vitro disassembly assays.

Virus particles are metastable structures that must protect the viral genome while in the environment and deliver it for infection upon entry in the infected cell. For TMV, virion disassembly has been much analyzed both in vitro and in vivo (reviewed in references 45 and 46). Caspar (62) proposed that stability switching for infection depends on the protonation state of carboxyl-carboxylate groups from adjacent CP subunits, identifying the pairs E50-D77 (axial interaction) and E95-E106 (side interaction) in the virion structure (63). These carboxylate groups have since been called “Caspar carboxylates,” and later mutational analyses confirmed their role in disassembly, with the axial interaction E50-D77 having the highest effect (64, 65). In the environment, interactions between these Caspar carboxylates will be stabilized by Ca²⁺ ions or protons, and the higher pH and lower Ca concentration in the cell will result in repulsion between carboxylates and destabilization of the particle structure, which will make the RNA accessible for cotranslational disassembly (66). Mutations of amino acids involved in the stability switch showed that increased stability resulted in decreased infectivity in an NLL assay and less efficient translation of virion particles in rabbit reticulocyte lysate (64, 65), thus suggesting a trade-off between particle stability and infectivity, two components of the virus fitness. Our previous analysis of the infectivities of field isolates of TMGMV P₀ and PMMoV P_1,2 isolates, which differed in particle stability, did not show pathotype-associated differences in infectivity (24). However, factors other than particle stability can determine the infectivities of isolates from two different virus species. In this study, we detected differences in infectivity between MG-WT and the low- and high-stability mutants MG-(M138N) and MG-(K43T+D50G), respectively. While MG-(M138N) was more infectious than MG-WT, as expected from the lower particle stability and lower axial CP interactions, MG-(K43T+D50G) was also more infectious than MG-WT, although less infectious than MG-(M138N), contrary to expectations from published results for TMV (46). The increased stability to in vitro disassembly resulting from the mutation D50G in MG-(D50G) and MG-(K43T+D50G) (Fig. 2) strongly suggests a role of D50 in the PMMoV stability switch, similar to the role of E50 in TMV. The stronger binding of the RNA to the CP units in MG-(K43T+D50G) relative to MG-WT is also contrary to higher infectivity being due to easier disassembly and access of RNA for translation. Thus, other, unknown factors in addition to ease of disassembly or carboxylate repulsion may be involved in linking particle structure and infectivity. In any case, our results show that a trade-off between particle stability and infectivity is not a universal feature among tobamovirus genotypes, as previous work (65) suggested.

The present results show pleiotropic effects of resistance-breaking mutations on PMMoV particle stability, a proxy for survival in the soil. These results confirm our previous finding that deployment of resistance in the host population results in selection for resistance breaking (12) and for differential survival, a trait unrelated to the plant-virus interaction (24). As was the case for virus multiplication in susceptible hosts (25), the sense and magnitude of the pleiotropy depend on the specific mutations that determine resistance breaking and not on the pathotype. Hence, there is no trade-off in this system between increased host range (i.e., resistance breaking) and survival, as has been reported for bacteriophages (55, 58). Also, the analysis of the within-host fitness of these coat protein mutants (Table 2) shows that fitness and particle stability rank differently. Moreover, no correlation was found between the values of fitness and those of the parameters of exponential curves describing disassembly kinetics (R² ≤ 0.074 and P ≥ 0.514 in a Spearman test for any assayed disassembly conditions), so there is no evidence of a trade-off between virus multiplication (i.e., reproduction) and particle stability (i.e., survival). This result is consistent with previous ones, as data on the within-host fitness of some of the CP mutants analyzed here, reported by Moreno-Pérez et al. (25), showed no correlation with our present stability data in different susceptible hosts (not shown). To our knowledge, the relationship between survival and multiplication has not been studied for other plant viruses, but studies with bacterial or animal-infecting viruses have demonstrated that such a trade-off is not general across systems (55–58, 67–72). Trade-offs between survival and reproduction, predicted by the life history theory, need not apply for viruses, because there is no obvious mechanistic reason to expect them due to the differences between the extracellular and intracellular environments where survival and reproduction, respectively, occur, as pointed out by Goldhill and Turner (19).

Survival may be particularly relevant for pathogens transmitted through the environment or through both the environment and direct contact (21, 23, 73), as is the case for tobamoviruses. Transmission through the environment will break the trade-off between virulence and transmission in directly transmitted pathogens (18), establishing a positive correlation between survival and virulence and allowing the evolution of highly virulent strains with high survival, the “curse of the pharaoh” hypothesis (22). Again, our data show that the PMMoV CP mutants do not rank similarly for virulence and particle stability, and there is no correlation between the values of virulence and the parameters of exponential curves describing disassembly kinetics (R² ≤ 0.159 and P ≥ 0.207 in a Spearman test for any assayed disassembly conditions). Again, this result is consistent with previous ones, as comparison of the virulence on different host genotypes of some of the PMMoV mutants analyzed here (25) and their particle stability do not provide evidence of a relationship between survival and virulence. Therefore, no survival-virulence correlation appears to occur in the analyzed system, even though it fulfills the conditions predicted by theoretical models for such a positive correlation to occur, i.e., that infection prevalence is not at equilibrium and that the death rate of the infected host is not much higher than that of virions in the environment (22). In epidemics of PMMoV on susceptible or resistant pepper cultivars, prevalence will not be in equilibrium, and PMMoV infection has little effect on plant survival (our unpublished observations). Similarly, our present and previous (24) data show no trade-off between survival and infectivity. Structural analysis of the wild-type and mutant CP interactions in particles significantly contributes to understanding the mechanistic bases of the occurrence or nonoccurrence of trade-offs between survival and other fitness components. This is to be underscored, as an understanding of the mechanistic bases of evolutionary trade-offs is rarely attained. A notable exception is the pioneering work of Dessau et al. (56) on the structural bases of thermal stability of bacteriophage Φ6.

Our results show that joint consideration of different life history traits is necessary for understanding virus evolution, as shown with experimental populations of phages or wild populations of avian influenza virus (56, 71, 72). The consequence of the reported pleiotropic effects of resistance-breaking mutations for the evolution of resistance-breaking is a complex one. Selection on traits that determine different components of the virus fitness may be in the same or opposite directions, which may favor or limit the durability of the resistance, depending on the specific resistance-breaking mutations. An additional conclusion from the results presented here, relevant to the understanding of virus evolution, is that trade-offs between different life history traits, predicted by theory for parasite evolution, may not apply to the evolution of viruses.

MATERIALS AND METHODS

PMMoV genotypes.

Viral genotypes were multiplied in Nicotiana clevelandii Gray after inoculation with transcripts from full-length infectious cDNA clones. Virus particles were purified as described by Bruening et al. (74). Plasmid pMG, generating transcripts named genotype MG, was obtained from field isolate P84/8 (12) as described by Moreno-Pérez et al. (25). PMMoV genotype MG is of a P_1,2 pathotype (25). PMMoV genotype MG is considered the wild type (WT) in this study, as site-directed mutagenesis of pMG was used to introduce mutations in the CP gene as described by Moreno-Pérez et al. (25), resulting in the following set of mutant genotypes: MG-(M138N), MG-(T43K+D50G), and MG-(L13F+G66V), of the P_1,2,3 pathotype, and MG-(A86G), of the P_1,2,3,4 pathotype. The single mutants MG-(T43K), MG-(D50G), MG-(L13F), and MG-(G66V) were also obtained. The list of mutants and their pathotype is presented in Table 4. Amino acids are numbered according to their position in the CP. Note that genotype MG-(A86G) was designated MG-(A87G) by Moreno-Pérez et al. (25), following the denomination used by Antignus et al. (38). However, codon 87 in the CP gene encodes amino acid 86 in the mature CP, and hence we modified the numbering for consistency with all other mutants. PMMoV genotypes TS (pathotype P_1,2) and TS-(M138N), TS-(T43K+D50G), and TS-(L13F+G66V) (pathotype P_1,2,3), described by Moreno-Pérez et al. (25) were also used in this work.

TABLE 4.

Pathotypes of PMMoV genotypes used in this work

Virus genotype	Phenotype forb:				Pathotype
	L1/L1	L2/L2	L3/L3	L4/L4
MG	+	+	NLL	NLL	P1,2
MG-(M138N)a	+	+	+	NLL	P1,2,3
MG-(T43K+D50G)a	+	+	+	NLL	P1,2,3
MG-(T43K)	+	+	NLL	NLL	P1,2
MG-(D50G)	+	+	NLL	NLL	P1,2
MG-(L13F+G66V)a	+	+	+	NLL	P1,2,3
MG-(L13F)	+	+	NLL	NLL	P1,2
MG-(G66V)	+	+	NLL	NLL	P1,2
MG-(A86G)	+	+	+	+	P1,2,3,4

^aThis mutation was also introduced in P_1,2 clone TS, as reported by Moreno-Pérez et al. (25).

^b+, systemic infection; NLL, resistance expressed as necrotic local lesions.

Analysis of the in vitro stability of virus particles.

Virus particles were purified simultaneously for all genotypes from N. clevelandii-infected plants, and particle stability was analyzed by incubating 200 μg of particles under different conditions. For stability at basic pH, particles were incubated in 0.1 M Tris-HCl (pH 8.75) on ice for different times (5, 10, 15, 30, 45, 60, 90, and 120 min), and the process of disassembly was stopped by addition of MgCl₂ to a final concentration of 50 mM (59). For analysis of stability at high ionic strength, particles were incubated in 6 M urea (pH 7.4) in phosphate buffer on ice for 5, 10, 15, 30, 45, 60, and 90 min, with disassembly being stopped by adding two volumes of 20 mM phosphate buffer, pH 6.9 (60). A third treatment was included, involving incubation in 6 M urea at pH 10.0 under the conditions specified for 6 M urea at pH 7.4. Disassembly was assessed by electrophoresis in 1.2% agarose gels in Tris-borate-EDTA (pH 8.1) (75) and ethidium bromide staining. For each genotype, in vitro stability analyses were repeated at least three times. For each virus genotype and treatment, the kinetics of disassembly was monitored by the amount of RNA encapsidated in full-length particles, which was quantified by densitometry using Image J. Values were adjusted to exponential functions of the form y = ab^x, and linearized curves were compared using analysis of variance (ANOVA) to test the equality of slopes (with the intercept being always 100%) on the basis of at least three replicates. Statistical analyses were performed using the statistical software package SPSS 21.0 (SPSS Inc., Chicago, IL).

Model analysis of interactions among CP subunits in virus particles.

The structure of the CP subunit assembled into virus particles of PMMoV was modeled separately for all the CP sequence variants by using the hybrid method I-TASSER (76, 77). In all cases, the best model was selected and compared with the crystal structure of U2-TMGMV (Fig. 3A), PDB code 1VTM (27). The biological assembly with helical symmetry made of 49 subunits in 3 turns of the helix was generated for all the CP haplotypes using the 49 transformation matrices of 1VTM. A core nonamer of CP subunits (Fig. 3B and C) was then selected to compute protein-protein interactions as follows. One central subunit (labeled A in Fig. 3B) was taken as a reference to compute the interactions with all its neighbor subunits arranged in four adjacent subunits (up, down, left, and right [labeled U, D, L, and R, respectively, in Fig. 3B]) and four diagonally adjacent subunits (labeled T, V, C, and E in Fig. 3B).

Protein-protein contacts, hydrogen bonds (H bonds), and ionic pairs were then calculated on the core nonamers with the PDB2PQR 2.1 program (78) and arranged in sets of interactions between the reference central subunit A and its eight neighbor subunits. PDB2PQR considers contacts between atoms “strong” if their distance is ≤3.2 Å and “weak” if it is between 3.2 and 4.0 Å. H bond cutoffs are 30° for acceptor hydrogen-donor angles and 3.7 Å for donor-acceptor distance, with the H bond considered “weak,” “moderate,” or “strong” if that distance is >3.2, between 3.2 and 2.5, and ≤2.5 Å, respectively. For ionic pairs, a cation-anion cutoff distance of 4.0 Å is assumed, with the strength of coulomb attraction inversely proportional to the interatomic distance. Poisson-Boltzmann (PB) electrostatic potentials for the WT and mutants were computed with APBS 1.4 (79) using AMBER99 charges (80) and atomic radii assigned with PDB2PQR 2.1 (78). Potential values in units of kT (where k is Boltzmann's constant and T is absolute temperature) were then mapped onto molecular surfaces and rendered with PyMOL 1.8 (PyMOL molecular graphics system, version 1.8.4; Schrödinger LLC).

Quantification of within-host virus multiplication and virulence.

Assays of virus multiplication and virulence were conducted in the susceptible host C. annuum cv. Dulce Italiano (L⁺/L⁺) as described by Moreno-Pérez et al. (25). Briefly, plants were grown at 23 to 25°C with a 16-h light photoperiod in a P2 level biological containment greenhouse. The first two true leaves were inoculated with 400 ng of freshly purified virus particles suspended in 0.1 M phosphate buffer, pH 7.2. Virus multiplication was estimated in infected plants at 21 dpi as viral RNA accumulation in all systemically infected leaves. Leaf RNA was extracted using TRIzol reagent (Life Technologies), and viral RNA accumulation was determined by quantitative real-time reverse transcription-PCR (RT-PCR), as described by Fraile et al. (24), using primers 5′ AACTGCCGAGACGCTT 3′, identical to nucleotides 5996 to 6011 of PMMoV, and 5′ GAGTTGTAGCCCAGGTG 3′, complementary to nucleotides 6137 to 6153 of PMMoV (accession no. KX063611). Levels of viral RNA were deduced from comparison with standard curves generated using a set of serial dilutions of purified virion RNA. Virus fitness was estimated from virus RNA accumulation as described by Lalic et al. (50) using the Malthusian parameter m, which represents the population exponential growth rate at a time t after inoculation, with m = (1/t) ln Q, where Q is the virus accumulation. Fitness (W) is then computed as W = e^m (59).

Virulence was quantified as the effect of infection on the above-ground host plant biomass, with V = 1 − (P_i/P_m), where P_i is the dry weight of the total biomass of each infected plant and P_m is the mean dry weight of mock-inoculated plants (81).

The distribution of fitness and virulence data was tested against the null hypothesis of normality using Kolmogorov-Smirnov test, and homogeneity of variances was analyzed with the Levene test. The distribution of fitness data was neither normal nor homoscedastic, while that of virulence data was. Thus, differences in relative fitness were analyzed using generalized linear models (GzLM), as this method is robust with respect to data distribution, and differences in virulence were analyzed using general linear models (GLM). The significance of the differences among classes within a factor was tested using the least significant difference (LSD) test. All statistical analyses were implemented in SPSS 21.0 (SPSS Inc.).

Infectivity quantification assay.

Genotypes MG, MG-(M138N), and MG-(T43K+D50G) were multiplied in N. clevelandii plants, and particles were purified simultaneously to avoid differences attributable to different storage times. Virus particles were resuspended in 1 mM EDTA, the concentration was adjusted to 18 μg/ml, and then the mixture was diluted at 1/3, 1/6, 1/9, 1/27, and 1/81. Twenty microliters of each dilution was inoculated in half leaves of two consecutive fully expanded leaves per plant of Nicotiana tabacum cv. Xanthi-nc. Twenty replicated half leaves were inoculated per treatment in a fully randomized design. At 4 days postinoculation, when NLL were apparent, leaves were harvested and NLL were counted. The initial virus concentration and dilutions were chosen on the bases of previous assays, so that the mean number of NLL per half leaf was above 0 and below 500.

Data on number of necrotic local lesions were normally distributed (Kolmogorov-Smirnov test) and showed homogeneity of variances among genotypes (Levene test). The relationship between infectivity and virus concentration was tested by bivariate analysis considering linear and nonlinear models. Infectivity curves were compared using ANOVA to test the equality of slopes and intercepts. Statistical analyses were performed using SPSS 21.0 (SPSS Inc.).