Experimental evolution of the host range dynamics in two isolates of potato virus Y
Ivair J Morais, João M F Silva, Alice K Inoue-Nagata, Santiago F Elena
TL;DR
This study explores how potato virus Y evolves in different host plants, showing how host environments influence viral adaptation and survival.
Contribution
The study experimentally tracks viral evolution across multiple host species, revealing how host switching affects genetic diversity and lineage survival.
Findings
Bentham’s tobacco supports high viral RNA accumulation and genetic stability, while potato causes strong bottlenecks and lineage extinction.
Tomato acts as an intermediate host with variable outcomes, and host switching helps prevent viral lineage extinction.
PVYNb shows higher diversity and nonsynonymous changes, while PVYSt has genomic stability and purifying selection.
Abstract
The emergence of plant viruses is a complex process influenced by viral genetic variation, host species, and environmental factors. To better predict and manage new plant diseases, it is important to understand how viruses adapt to novel hosts. In this study, we examined how two isolates of potato virus Y (PVY), PVYNb, and PVYSt, evolve when repeatedly passed through three solanaceous plants: Bentham’s tobacco (Nicotiana benthamiana), potato (Solanum tuberosum), and tomato (Solanum lycopersicum). We also tested whether switching between hosts could reduce the impact of strong population bottlenecks, which often occur in poorly suited hosts. Our findings show that benthamiana supports high viral RNA accumulation and genetically stable diversity, consistent with large effective population sizes. In contrast, potato creates strong bottlenecks, often leading to viral lineage extinction and…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10| Sample | Passage | %SNPs | Fixed | Synonymous | Nonsynonymous |
|---|---|---|---|---|---|
| PVYNb | 0 | – | – | – | – |
| PVYNb/Sl.L2 | 1 | 20 | 0 | 0 | 0 |
| PVYNb/St.L1 | 4 | 0 | 23 | 14 | 9 |
| PVYNb/St.L2 | 4 | 0 | 353 | 251 | 92 |
| PVYNb/Nb.L1 | 4 | 40 | 0 | 0 | 0 |
| PVYNb/Nb.L1 | 10 | 0 | 0 | 0 | 0 |
| PVYNb/Nb.L2 | 4 | 30 | 0 | 0 | 0 |
| PVYNb/Nb.L2 | 10 | 20 | 0 | 0 | 0 |
| PVYNb/CTF.L1 | 4 | 0 | 333 | 239 | 94 |
| PVYNb/CTF.L2 | 4 | 0 | 225 | 138 | 79 |
| PVYNb/MIX.L1 | 3 | 0 | 0 | 0 | 0 |
| PVYNb/MIX.L2 | 4 | 0 | 94 | 52 | 42 |
| PVYNb/MIX.L2 | 9 | 0 | 24 | 9 | 15 |
| PVYSt | 0 | – | – | – | – |
| PVYSt/St.L1 | 4 | 12.5 | 0 | 0 | 0 |
| PVYSt/St.L1 | 6 | 12.5 | 185 | 143 | 36 |
| PVYSt/St.L2 | 4 | 18.75 | 0 | 0 | 0 |
| PVYSt/St.L2 | 10 | 12.5 | 1 | 0 | 0 |
| PVYSt/Nb.L1 | 4 | 12.5 | 0 | 0 | 0 |
| PVYSt/Nb.L1 | 10 | 6.25 | 0 | 0 | 0 |
| PVYSt/Nb.L2 | 4 | 18.75 | 0 | 0 | 0 |
| PVYSt/Nb.L2 | 10 | 12.5 | 0 | 0 | 0 |
| PVYSt/CTF.L2 | 1 | 0 | 0 | 0 | 0 |
| PVYSt/MIX.L1 | 2 | 6.25 | 1 | 0 | 0 |
- —Generalitat Valenciana10.13039/501100003359
- —CNPq10.13039/501100003593
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsPlant Virus Research Studies · Plant and Fungal Interactions Research · Evolution and Genetic Dynamics
Introduction
Emerging plant viruses pose a serious threat to global agriculture, biodiversity, and ecosystem stability. Their expansion is driven by a complex mix of ecological factors, virus-host interactions, and the ability of the viruses to evolve and adapt to new environments (Morse 1995, Jones 2009, Lefeuvre et al. 2019). In recent years, viruses like tomato brown rugose fruit virus, cucumber green mottle mosaic virus, cassava brown streak virus, tomato torrado virus, and rice stripe virus have caused major economic damage. These cases highlight the urgent need to understand the ecological and evolutionary forces behind virus emergence (Jones 2009, Lefeuvre et al. 2019).
A key step in virus emergence is the ability to infect and persist in new hosts. This usually begins with low viral fitness in the new host, followed by either extinction or successful adaptation that allows the virus to replicate and spread (Morse 1995). Several factors influence this process, including viral genetic diversity, compatibility with the host, dispersal conditions, and demographic events like population bottlenecks.
One useful way to understand these dynamics is the source-sink model. In this model, viruses move between hosts that differ in how well they support viral replication. ‘Source’ hosts allow high replication and large viral populations, while ‘sink’ hosts are more restrictive, leading to low viral RNA accumulation and a higher risk of extinction unless the virus adapts (Holt and Gaines 1992, Gomulkiewicz and Holt 1995, Morse 1995, Antia et al. 2003). Whether a virus can adapt in a sink host depends on how severe the bottlenecks are and how often it returns to source hosts. If the cost of infecting the sink is too high, the virus may go extinct before it can adapt (Holt and Gomulkiewicz 1997). However, occasional returns to source hosts can act as a ‘rescue,’ keeping the virus population alive long enough for adaptation to occur (Morse 1995). While this theory is well established, there is still limited experimental evidence in plant virus systems.
Potato virus Y (PVY), one of the most economically damaging plant viruses worldwide, is a great model for studying these processes. PVY infects solanaceous crops like Bentham’s tobacco (Nicotiana benthamiana Domin), potato (Solanum tuberosum L.) and tomato (Solanum lycopersicum L.). It has a single-stranded, positive-sense RNA genome ~9.7 kb long, with a VPg at the 5′ end and a poly(A) tail at the 3′ end (Shukla et al. 1994). The genome encodes a large polyprotein (~3 062 amino acids) that is processed into 10 mature proteins. An additional peptide is produced via a − 1 ribosomal frameshift in the P3 gene. PVY replicates using its own RNA-dependent RNA polymerase (NIb) (Li et al. 1997), which has low fidelity, leading to high mutation rates of around 10^−6^ substitutions per site per replication for related potyviruses (Sanjuán et al. 2009, Tromas and Elena 2010, De la Iglesia et al. 2012).
Experimental evolution has helped us to understand how viruses adapt to new hosts. These studies show that viruses can evolve quickly, changing how they infect, replicate, and mutate (Remold et al. 2008, Hillung et al. 2014). Evolution experiments, where viruses are repeatedly passed through different hosts, often show that viruses become specialized. Lineages evolved in one host tend to perform better in that host but may lose fitness in others (Wallis et al. 2007, Agudelo-Romero et al. 2008, Bedhomme et al. 2011, Hillung et al. 2014, Navarro et al. 2022). While this has been well studied in potyviruses like tobacco etch virus and turnip mosaic virus, research on PVY is more limited and often focuses on single-host systems or specific resistance genes (Cuevas et al. 2012, Fabre et al. 2012, Kutnjak et al. 2017, Rousseau et al. 2017, Da Silva et al. 2020).
To fill this gap, we conducted an evolution experiment using two PVY strains from different hosts. We infected three solanaceous species under five different conditions and tracked viral evolution over 10 serial passages. We hypothesized that host permissiveness and bottlenecks shape PVY specialization, with benthamiana acting as a source and tomato as a sink. We measured viral titers using reverse transcription quantitative polymerase chain reaction (RT-qPCR) and sequenced viral genomes at different stages using high-throughput sequencing (HTS). We also recorded symptoms and plant height to link genetic changes with viral evolution and virulence over time.
Materials and methods
Plants and growth environment
We used three plant species as experimental hosts for the evolution experiments: N. benthamiana, S. lycopersicum cv. Marmande, and S. tuberosum cv. Kennebec. Plants were maintained in a growth chamber, with a light period of 16 h at 24°C (LED tubes at PAR 90–100 μmol m^−2^ s^−1^), a dark period of 8 h at 20°C, and 40% relative humidity. Individual plants were transplanted into pots, with each pot containing two plants, except for the potato plants, which were cultivated in separate pots. The soil substrate comprised a mixture of DSM WNR1 R73454 substrate (Kekkilä Professional, Vantaa, Finland), grade 3 vermiculite, and 3–6 mm perlite in a ratio of 2:1:1.
Prior to the experiment, the batch of seed potatoes was tested by RT-PCR to ensure the absence of PVY infection. Infection of the tuber by other viruses was not expected, since all used tubers were certified as free from viral infections. The potato tubers were cut into two or three sections, each containing at least one eye, and submerged in a 2 ppm gibberellic acid solution for ~1 hour before planting. To standardize the experimental conditions, only one potato stem was retained for each plant before the inoculation process, with additional stems removed. This approach aimed to minimize variability and ensure the uniformity of the experimental setup.
Isolates, inoculation, and collection
Throughout the experiments, we used two isolates of PVY (species Potyvirus yituberosi, genus Potyvirus, family Potyviridae). PVYSt (N-Wi strain) was originally collected by the authors in 2020 from a commercial potato field (cv. Atlantic) in Brasília, DF (Brazil). After collection, the isolate was maintained only once in its original host (potato cv. Atlantic), from which the plant material used in this study was obtained. In contrast, PVYNb (O strain), has been maintained in benthamiana through continuous mechanical inoculations over multiple generations, although the exact number of passages is unknown. To recover the inoculum for this experiment, PVYNb was passed once more in benthamiana. The consensus genomes analysed here represent the contemporary laboratory variants of each isolate at the time this study was initiated. The two isolates were assigned to strains by their inclusion in the corresponding clades in a maximum-likelihood (ML) phylogenetic tree constructed using IQ-TREE version 2.0 (Minh et al. 2020), with 10 000 bootstrap pseudosamples used to evaluate the statistical confidence of clades. This analysis utilized a dataset of 447 representative PVY haplotypes downloaded from GenBank (downloaded in 25/12/2023), along with the consensus sequences of PVYNb and PVYSt (Supplementary Fig. S1), determined by HTS as described below.
Mechanical inoculation was performed using phosphate inoculation buffer, pH 7, with 3% polyethylene glycol, and Carborundum. For inoculation, 20 μl inoculum solution were deposited per leaf on two leaves per plant, and inoculation was done manually. To account for morphological differences in leaf structure among the three host species, inoculations were standardized as follows: for tomato and potato, which have compound leaves, the entire second or third fully expanded compound leaf below the apex was inoculated. For benthamiana, which has simple leaves, inoculation was performed also in the second or third fully expanded single leaf below the apex. This approach aimed to standardize the amount of tissue exposed to the viral inoculum across hosts and minimize differences in the effective population size at the initiation of infection.
Plant preparation varied among host species to ensure uniform developmental stages at the time of inoculation. Tomato seeds were sown and transplanted two weeks later; one-week after transplantation (i.e. three weeks post-sowing), plants were inoculated. Potato plants were grown from tubers planted three weeks before inoculation. Benthamiana plants were sown, transplanted at three weeks, and inoculated one-week post-transplantation.
Ten days post-inoculation (dpi), the three uppermost non-inoculated leaves were collected. At this stage, plants were well developed, and collected tissues exhibited distinct symptomatology depending on the host. Subsequently, these plant tissues were flash-frozen in liquid N_2_, powdered, and homogenized. The homogenized tissue was resuspended directly in the same phosphate inoculation buffer initially used, using ~150 μl of tissue based on visual estimation. This homogenate was then used to prepare the inoculum for subsequent passages. All collected samples were preserved at −80°C to maintain their molecular integrity and ensure the preservation of viral particles for subsequent analyses.
Primer design and RNA amplification
In our study, we employed two approaches for viral genome amplification depending on the objectives. The first aimed to quantify viral RNA using RT-qPCR, while the second focused on detecting the virus using standard RT-PCR. We designed two sets of primers for these purposes.
The RT-PCR primer set targeted a highly conserved region within the NIb gene of PVY: forward 5’-ACTATGATTTTTCGTCGAGAACAA-3′ (Universal PVY forward primer, UYF) and reverse 5’-CGCGAGGTTCCATTTTCAATGC-3′ (Universal PVY reverse primer, UYR). Total RNA extractions were performed using the NZY Plant/Fungi RNA Isolation Kit (NZYtech, Lisbon, Portugal). RT-PCR was carried out using NZYSupreme One-step RT-qPCR Probe Master Mix 2× (NZYtech) under the following conditions: 50°C for 20 min, 95°C for 5 min, followed by 40 cycles of 95°C for 5 s, 60°C for 40 s.
The RT-qPCR primer set targeted a conserved region of the viral CP gene (qYF, 5’-CAATCACAGTTTGATACGTGG-3′ and qYR 5’-GGCGAGGTTCCATTTTCAATGC-3′) and a common housekeeping gene, the glyceraldehyde-3-phosphate dehydrogenase (GAPDH), highly conserved among plant taxa (Martin and Cerff 1986) (qGAPDHF 5’-CTGTAACCCCAYTCGTTGTC-3′ and qGAPDHR 5’-GTKGTKTCMAMWGAYTTTGTKGG-3′).
To generate the standard curves, a series of cDNA dilutions of PVY ranging from 50 to 0.005 ng/μl was prepared. Each dilution was tested in triplicate. The standard curves were used to calculate the qPCR amplification efficiency and the accuracy of the quantification, utilizing the linear regression equations derived from the C_T_ values versus the logarithm of the initial RNA concentration. The amplification efficiencies (%) were calculated based on the slope (s) of the standard curves using the expression efficiency = 100 × (10^–1/s^ − 1). The efficiency of PVY primers for the amplification of the portion of PVYNb and PVYSt genome was 94% (R^2^ = 99.9%) and 104% (R^2^ = 99.1%), respectively. For the GAPDH primers, the amplification efficiency was 89% for benthamiana (R^2^ = 99.3%) and tomato (R^2^ = 98.4%), and 91% (R^2^ = 99.9%) for potato (Supplementary Fig. S2). Thus, the RT-qPCR method was validated for adequate quantification of PVY RNA in the plant samples.
RNA extraction for this set was conducted using Sigma STRN250 Spectrum Plant Total RNA Kit with DNase treatment (Invitrogen TURBO DNA-free Kit AM1907, California, USA). The samples were checked for concentration and quality using NanoDrop One (Thermo Fisher Scientific, Wilmington NC, USA) and normalized to 50 ng/μl. RT-qPCR was performed using qPCRBIO SyGreen 1-Step Go Hi-Rox (PCR BIOSYSTEMS, London, UK) with at least three replicates for each sample. The conditions included 45°C for 10 min, 95°C for 2 min, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. All RT-qPCR results were filtered based on quality: only reactions with C_T_ between 5 and 35 and with replication standard deviation \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \le\end{document} 0.3 were retained. For each sample, mean C_T_ per amplicon was calculated and ΔC_T_ = C_T_^PVY^—C_T_^GAPDH^ was computed. The ΔΔC_T_ method (Schmittgen and Livak 2008) was used with the initial sample of each virus isolate as calibrator and relative quantities were calculated as RQ = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} {2}^{-\Delta \Delta{C}T}\end{document} . Viral RNA accumulation was analysed using a linear model applied to log_10RQ, which stabilizes variances and yields group comparisons. Before analysis, we also removed a small number of biologically inconsistent values to ensure that viral accumulation trajectories reflected true evolutionary dynamics rather than technical artefacts.
Transmission rate evaluation experiments
In the passage experiment, we recognized the potential for a loss of quantification accuracy due to pooling all 16 plants during collection and further inoculation. To address this issue, we devised a transmission rate quantification experiment. In this setup, we inoculated 50 benthamiana, 50 potatoes, and 50 tomatoes using the founder PVYNb or PVYSt. Each plant was individually collected to quantify the number of positively infected plants.
For this analysis, viral detection was performed by endpoint RT-PCR and the samples were subjected to 1% agarose gel electrophoresis with SYBR green. Among the positive plants, we randomly selected three, except for tomatoes infected with PVYSt, for which only two positive plants were obtained. Subsequently, each selected positive sample was used to inoculate another set of 50 plants for each species.
This process allowed us to assess the likelihood of a virus passing through the same host. Utilizing the same approach of RT-PCR and gel electrophoresis, we systematically analysed the infection rates and dynamics within each host species. This individualized sampling strategy aimed to provide a more accurate and detailed understanding of virus transmission patterns among the different plant species.
Evolution experiment
Twenty evolution experiments were simultaneously initiated, each with a total of 10 serial passages; half were started with PVYNb and the other half with PVYSt. Five treatments were tested differing in their host plant composition as follows: viruses only inoculated to tomato plants (Sl); only to potato plants (St); or only to benthamiana plants (Nb); viruses alternating inoculations in the tomato-potato-benthamiana sequence (correlated temporal fluctuations, CTF); and viruses inoculated into a mixture of the three plant species at equal proportions at each passage (MIX). Two independent evolution lineages per treatment were generated (L1 and L2). At each passage, the host population size was 16 plants. A full experiment design can be seen in the Fig. 1.
Schematic illustration of the evolution experiment of PVY into different host setups. Two isolates were used: PVYNb, originally from benthamiana (represented in orange), and PVYSt, originally from potato (in green). Each isolate was used to initiate five independent evolutionary lineages with two replicates each, totaling 10 lineages: Sl.L1 and Sl.L2 (S. lycopersicum); St.L1 and St.L2 (S. tuberosum); Nb.L1 and Nb.L2 (N. benthamiana); CTF.L1 and CTF.L2 (correlated temporal fluctuations); and MIX.L1 and MIX.L2 (mixed host species). Each plant symbol in the schematic corresponds to 16 inoculated plants (blue background), or, in the case of MIX, six plants per species (yellow background). Using the last positive sample, 10 plants of each species were inoculated (purple background). Arrows indicate mechanical inoculation between passages. Passages selected for HTS are marked with a star beneath the plant icon. A list of the last positive passages per lineage is available at Supplementary Table S1.
All evolving lineages underwent simultaneous inoculation on the same day, and the plant symptoms were daily monitored until 10 dpi. Then, leaves of the 16 plants were pooled and representative samples were collected. The tissue was powdered in liquid N_2_, a portion used for inoculation of the next passage, and another portion was used for RNA extraction. Following quantification via relative RT-qPCR, only the positive lineages were continued. For negative lineages, the inoculation process was repeated using tissue from the previous positive passage to minimize potential inoculation error. Importantly, no previously negative lineage yielded a positive result after the second trial, ensuring the reliability of the experimental outcomes.
In each passage, six plants from each species inoculated only with phosphate buffer served as mock-inoculated controls. In addition, six plants of each species were employed as negative controls.
For the evaluation of disease phenotypic effects, the plant height was measured from the base of the plant to the apical tip. Measurements were taken 1 day before inoculation and one day before collection (9 dpi). As a measure of the impact of infection on plant growth, we evaluated the effect of viral infection in plant growth relative to the mean growth of mock inoculated plants, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} V=1-\frac{\Delta{L}{infected}}{\left\langle \Delta{L}{mock}\right\rangle },\end{document} in which L is the plant height measured at the time of inoculation and 9 dpi and the 〈·〉 stands for the average value. V < 0 indicates stunting relative to the average mock-treated plant, while V > 0 indicates infection induces overgrowth compared to noninfected plants. It is important to note that this index captures the magnitude of deviation in growth due to infection, regardless of direction, and thus is a measure of symptoms rather than of virulence (i.e. alteration in plant reproductive success).
At the end of the passages, an infectivity assay was performed using viral extracts obtained from the last positive passage of each lineage, rather than the final attempted passage in cases where the lineage was discontinued due to extinction (i.e. two consecutive failures to establish infection). Each viral extract was mechanically inoculated onto 10 plants of each of the three host species. Individual RT-PCR were performed on each plant to detect the number of positive samples. Negative and mock controls were used during the inoculation and detection steps.
HTS and sequence analyses
To determine the genome changes of the virus during the passage experiment, we employed Illumina HTS on three time points: (i) the initial PVYNb and PVYSt inoculum source, (ii) the 4^th^ passage (n = 12), and (iii) the latest available passage of each lineage (n = 17; Table 1). Total RNA was extracted from fresh or dried leaf tissue using STRN250 Spectrum Plant Total RNA Kit (Sigma-Aldrich, Burlington MA, USA), then they were treated with Invitrogen TURBO DNA-free Kit (Thermo Fisher Scientific). The RNA concentration and the ratio of absorbance at 260/230 and 260/280 nm for quality were checked using Nanodrop before being sent for sequencing at Macrogen Inc. (Seoul, South Korea). RNA GTR5001-S screwcap microtubes (GenTegra, Pleasanton CA, USA) ensured secure sample storage and transportation during sequencing. Libraries were prepared using TruSeq Stranded Total RNA Library with Ribo-zero Plant (Illumina Inc., San Diego CA, USA) and sequenced in a NovaSeq 6 000 platform (Illumina Inc.) at a 100 × 2 bp configuration.
Consensus generation and variant calling
To generate a consensus sequence for each sample (n = 31), reads were trimmed with BBDuk (https://sourceforge.net/projects/bbmap/), assembled with MEGAHIT (Martin and Cerff 1986, Li et al. 2015) and subjected to DIAMOND (Buchfink et al. 2015) BLASTX searches against the non-redundant NCBI database (download on the 24/06/2024). The longest contigs identified as PVY for each sample were then subjected to BLASTN searches against the nucleotide database to identify their closest isolate (HM367076 for PVYNb and MW685829 for PVYSt). Then, reads were aligned against their respective reference sequence with BBMap v39.01 (https://sourceforge.net/projects/bbmap/) and a consensus sequence for each isolate was generated with Geneious Prime build 2022-03-15 (Dotmatics, Boston MA, USA). Reads were aligned to the corresponding consensus sequence of PVY with BBmap with the vslow option. Base recalibration was then performed with GATK v4.0.5.1 (Van der Auwera and O’Connor 2020). Then, single nucleotide polymorphisms (SNPs) were called with LOFREQ2 (Wilm et al. 2012). To further mitigate false positives due to sequencing error, only SNPs with frequency > 1% and with coverage depth > 50 were retained for downstream analyses. Additionally, only sequences with > 100 average coverage depth were used for diversity analyses between and within samples.
Diversity estimation
The diversity of each sequenced line was calculated from the Shannon entropy of each polymorphic site and by the alignment pairwise distance (APD) method (Shao et al. 2014). For each population, the mean APD across all sites was used as a summary measure of within-population genetic diversity. To assess genetic differentiation between populations, we estimated the fixation index (F_ST_). F_ST_ was computed from pairwise comparison between populations derived from the same viral isolate with Popoolation2 (Kofler et al. 2011), accounting only for SNPs with a minimum count of two and at positions with > 50 coverage depth and using the default population size of 500.
To visualize the diversity along the genome of PVY, site-wise Shannon entropy values were plotted along the PVY genome for each population, and positions of fixed SNPs (as defined in Section 2.7) were overlaid and classified as synonymous, nonsynonymous, or located in UTR according to the PVY genome annotation. If two or three SNPs in the same codon were fixed, they were accounted for simultaneously. We also constructed the ML-tree using IQ-TREE v2 with 10 000 bootstrap pseudoreplicates with all consensus generated sequences.
Estimation of synonymous and nonsynonymous diversity
To quantify selective pressures acting on each coding region, we used SNPGenie (Nelson et al. 2015), which computes per-site synonymous (πS) and nonsynonymous (πN) nucleotide diversity as well as the ratio πN/πS. SNPGenie was run in ‘pooled’ mode using allele-frequency files generated from LOFREQ2 variant calling (see Section 2.7) and default parameters. For each population, πN, πS, and πN/πS were estimated for all annotated cistrons of the PVY polyprotein. Values of πN/πS > 1 were interpreted as potential evidence of positive selection, whereas values < 1 indicate purifying selection, acknowledging that large variance in low-diversity samples may inflate individual estimates.
Results
Test of transmission efficiency across host species
We began by assessing how potential host-imposed constraints might influence the adaptation of the two PVY isolates. To explore whether host species differentially modulate the early infection success of PVY isolates, we compared the transmissibility of PVYNb and PVYSt across benthamiana, potato and tomato plants by mechanical inoculation. Clear differences in infection success were observed both between the viral isolates and among host species (Fig. 2), suggesting that the likelihood of establishing successful infections depends on specific host-virus combination.
Evaluation of PVYSt and PVYNb infection rates on each of the three alternative hosts. Fifty plants of each species were inoculated, and their infection status was individually determined by RT-PCR. Three randomly chosen positive plants were used to inoculate a second batch of 50 plants of the same species. The numbers below or at the side of each plant represent the infected out of the inoculated plants.
Data shown in Fig. 2 were fitted to a logistic-regression using a generalized linear model (GLM) with a Binomial distribution probability and logit-link function. Firstly, this analysis confirmed the highly significant differences between the two viral isolates (χ^2^ = 154.204, 1 d.f., P < .001), with PVYNb transmission efficiency across hosts being 0.84 ± 0.04 (±1 SE), while it was 88.1% lower (0.10 ± 0.03) for PVYSt. Host species also significantly affected transmission (χ^2^ = 9.821, 2 d.f., P = .007), with benthamiana (0.6 ± 0.1) and potato (0.53 ± 0.05) showing similar susceptibility, while tomato was remarkably less susceptible (0.26 ± 0.08). More interestingly, a significant interaction between viral isolate and host species was found (χ^2^ = 30.608, 2 d.f., P < .001), indicating that transmissibility depends on the specific combination of host and virus.
In the first round of inoculation, PVYNb transmission peaked in benthamiana (0.96 ± 0.03) and was lowest in potato (0.70 ± 0.07). Conversely, PVYSt transmitted best in potato (0.36 ± 0.07) but showed poor transmission in tomato (0.04 ± 0.03).
In a second round using new inoculum from three infected plants (or two in the case of PVYSt in tomato), PVYNb maintained high transmissibility in benthamiana (0.92 ± 0.02) but suffered drastic declines in tomato (0.07 ± 0.02) and potato (0.07 ± 0.02). PVYSt showed a similar pattern: high transmission in benthamiana (0.94 ± 0.02), reduced in potato (0.12 ± 0.03), and null in tomato. Differences among the two viral isolates remained significant after this second infection (χ^2^ = 4.001, 1 d.f., P = .045), although in this case the average transmission efficiency for the PVYNb-derived samples was 0.29 ± 0.04 but null for the PVYSt-derived ones.
This pattern was supported by a strong interaction between inoculum origin and recipient host (χ^2^ = 12.770, 2 d.f., P = .002). Notably, PVYNb and PVYSt propagated from benthamiana maintained high infectivity in this host but failed almost completely when inoculated in tomato or potato.
Transmission changes between the first and second inoculations were host-dependent for both isolates (Fig. 3). For PVYNb, transmission remained stable in benthamiana (0.96 ± 0.03 versus 0.92 ± 0.02; sequential Bonferroni post hoc test, P = .779) but dropped sharply in tomato (0.74 ± 0.06 versus 0.07 ± 0.02; P < .001) and potato (0.70 ± 0.07 versus 0.07 ± 0.02; P < .001) (χ^2^ = 6.946, 2 d.f., P = .031). For PVYSt, a contrasting pattern was observed (χ^2^ = 123.735, 2 d.f., P < .001). Transmission efficiency increased dramatically in benthamiana from 0.06 ± 0.03 to 0.94 ± 0.02 (P < .001), suggesting recovery of infectivity. In sharp contrast, no change in efficiency was observed for tomato (0.04 ± 0.03 versus 0.000 ± 0.000; P = .447) and a significant reduction in potato (0.36 ± 0.07 versus 0.12 ± 0.03; P = .006) was found.
Transmission efficiencies of PVYNb and PVYSt in benthamiana, tomato, and potato plants based on the results of the infection rate experiment during two sequential passages using the same host. The data were fitted to a logistic-regression using a generalized linear model (GLM) with a binomial error distribution and logit-link function. Error bars represent ±1 SE.
In conclusion, these experiments show that the transmission efficiency of PVY isolates is highly dependent on the combination of viral genotype and host species. PVYNb exhibited higher transmissibility across all hosts compared to PVYSt, with benthamiana being the most susceptible. Sequential inoculations revealed that PVYNb transmission efficiency decreased significantly in potato and tomato, while remaining stable in benthamiana. In contrast, PVYSt improved its transmission in benthamiana but struggled in tomato and potato. These results suggest that different plant species can act as either facilitators (sources) or barriers (sinks) to viral transmission and adaptation.
Variation of viral RNA accumulation along the evolutionary passages
Viral RNA accumulation, estimated by log_10_-transformed relative RT-qPCR values, was taken as a proxy to within-host replicative fitness. Figure 4 shows the evolution of viral RNA accumulation, for both PVY isolates under each of the five experimental host treatments. Only samples meeting the qPCR quality criteria were retained for analysis (see Section 2.3). Data were fitted to a GLM with a Gamma probability function and a log-link function; viral isolate, experimental treatment and passage were included in the model as orthogonal factors and lineage was nested within the interaction of viral isolate and experimental treatment.
Viral accumulation (log10-relative quantifications, RQ) of PVY across passages during the evolution experiment. Each panel represents one virus isolate (columns) and one host treatment (rows). Points show log10RQ for each independent evolutionary lineage (line 1 in blue, line 2 in red). Coloured solid lines represent lineage-specific mean trajectories, the dashed black line represents the overall mean for each virus-treatment combination, and the shaded area correspond to this mean ± 1 SD.
Firstly, a net effect of passage was observed (χ^2^ = 3984.677, 9 d.f., P < .001) due to temporal fluctuations in viral accumulation across passages (Fig. 4). Secondly, net differences exist between both PVY isolates (χ^2^ = 1691.114, 1 d.f., P < .001), with PVYNb, on average, accumulating more than PVYSt (marginal mean values per isolate: 292 ± 16 versus 0.023 ± 0.002, respectively). Thirdly, differences exist between the five host treatments (χ^2^ = 5682.249, 4 d.f., P < .001). Overall, the most permissive host for virus replication was benthamiana [marginal mean accumulation (2.33 ± 0.03)·10^6^] distantly followed by tomatoes (161 ± 18) and potatoes (13.0 ± 0.8). Viral RNA accumulation estimated for the two mixed-hosts treatments were low [(1.0 ± 0.2)·10^−6^ for CTF and (2.8 ± 0.3)·10^−2^ for MIX]. These values refer to the estimated marginal means from the GLM model, not the raw log_10_RQ values displayed in Fig. 4.
Interestingly, a significant interaction between PVY isolate and experimental conditions was observed (χ^2^ = 2532.462, 4 d.f., P < .001), indicating that the accumulation of each isolate depended on the host in which it was measured. Furthermore, this effect also depends on passage number (χ^2^ = 1498.356, 9 d.f., P < .001), suggesting that the difference between isolates changed along evolutionary time. For both isolates, benthamiana showed the highest accumulations, confirming its role as source host [(2.75 ± 0.05)·10^6^ for PVYNb and (1.98 ± 0.04)·10^6^ for PVYSt]. However, in the case of PVYNb, CTF showed the second largest accumulation, (2.1 ± 0.1)·10^3^, whilst PVYSt in potato ranked second (1.028 ± 0.003)·10^3^. The lowest accumulation of PVYNb was observed in potato (0.16 ± 0.02), while for PVYSt it was so in the two treatments involving the three hosts [(5 ± 2)·10^−16^ in CTF and (2.6 ± 0.5)·10^−6^ in MIX]. In tomatoes, PVYNb was unable to infect plants in the second passage after an initial infection, being restricted to the first passage (Fig. 4). PVYSt also showed difficulty in infecting tomatoes, but PVYSt/Sl.L1 was maintained in tomatoes for five serial passages (Fig. 4). Interestingly, PVYSt/Sl.L1 showed an increase in viral RNA accumulation during the passages, but after the peak of virus abundance, it failed to infect the plants in the next passage. This result is in line with the evidence that tomato is a sink host for the propagation of the two used PVY isolates. In potatoes, both PVY isolates could infect systemically and be passed through at least five passages. The variation in viral RNA accumulation during passages did not follow a consistent pattern. PVYNb lines infected potatoes for five passages, showing a similar infection pattern except for the fourth passage, which decreased dramatically and resulted in absence of infection in the next passage. PVYSt/St.L1 could infect potato plants for six passages and PVYSt/St.L2 for 10 passages. PVYSt/St.L2 exhibited a significant increase in virus RNA accumulation after the 8^th^ passage, suggesting adaptation to the host. In benthamiana, the four experimental lineages reached 10 passages. PVYNb/Nb.L1, PVYNb/Nb.L2, PVYSt/Nb.L1, and PVYSt/Nb.L2 exhibited high viral RNA contents compared to other lines (Fig. 4). The viral RNA accumulation was high from the 1^st^ passage and remained high throughout all passages. Both PVYNb lineages had a close detection pattern and constant virus amount with minimal variation. In contrast, PVYSt lineages displayed more variation but seemed to have adapted to the host in the last two passages. For the CTF treatment, PVYSt/CTF.L1 could not infect tomato plants, while in PVYSt/CTF.L2 only the 1^st^ passage contained infected plants, but failed to infect potato plants in the next passage. PVYNb had an initial low infection ability in tomatoes but it increased in subsequent passages with potatoes and benthamiana, decreasing again when returning to tomatoes. PVYNb/CTF.L1 reached the 4^th^ passage but could not infect potatoes after being inoculated in tomatoes, while PVYNb/CTF.L2 stopped at the 8^th^ passage, unable to pass from potatoes to benthamiana. In MIX treatment, PVYSt/MIX.L1 was terminated at the 2^nd^ passage with a low viral RNA accumulation, and PVYSt/MIX.L2 caused no infection. PVYNb was detected until the 3^rd^ passage, with similar viral RNA accumulation in both lineages. Lineage PVYNb/MIX.L2 reached the 9^th^ passage with significant variation between passages. It was unclear which host contributed to virus replication as plant tissues were pooled during RNA extraction and virus detection.
The observed effects of passages on viral RNA accumulation did not necessarily represent an overall trend to increase or decrease in virus accumulation, but simply uncorrelated significant differences among passages. Indeed, for lineages evolved in tomato (partial correlation coefficient controlling for viral strain and lineage: r_p_ = 0.289, 34 d.f., P = .088), benthamiana (r_p_ = −0.121, 130 d.f., P = .167) and CTF (r_p_ = 0.156, 53 d.f., P = .254) treatments, there was no significant correlation between the number of the passage and the accumulation of viral RNA. A weak positive yet significant correlation (r_p_ = 0.208, 108 d.f., P = .029) was found for viral lineages evolved in potato, while a significant negative correlation was found for the MIX lineages (r = −0.360, 59 d.f., P = .004).
Based on these experiments, we concluded that a host species significantly influences the pace of PVY phenotypic evolution. The permissiveness of different hosts varied widely, with benthamiana plants demonstrating their role as source host supporting high viral replication and efficient transmission across passages, while tomato plants acting as sink hosts, often failing to support the virus beyond the initial passages. Host switching revealed that initial low infection rates in less permissive hosts could improve later in more permissive hosts, and mixed host lines showed varied infection outcomes.
Infectivity of evolved lineages depends on both the local environment and the test host
Using the last positive sample of PVYNb as inoculum, 10 plants of each host were inoculated and their infection status determined by RT-PCR detection (Fig. 5 and Supplementary Table S1). Fitting these infectivity data to a logistic regression, the analysis confirmed a significant effect of the host compositions during serial passages (χ^2^ = 35.722, 4 d.f., P < .001), with viruses passaged in benthamiana being, on average, the most infectious, followed by those evolved in potato. No net significant effect was observed for the host in which the infectivity of the evolved lineages was tested (χ^2^ = 0.000, 2 d.f., P = 1.000), although a significant interaction existed between the evolved host and the test host (χ^2^ = 33.915, 8 d.f., P < .001). PVYNb/Sl.L1 could not infect any hosts, while PVYNb/Sl.L2 infectivity of benthamiana plants was 0.2 ± 0.2 (LaPlace estimator of the binomial parameter with 95% adjusted Wald CI). After five passages in potatoes, PVYNb/St.L1 could not infect any hosts, but PVYNb/St.L2 infectivity in benthamiana was 0.8 ± 0.2, 0.8 ± 0.2 in tomato and 0.3 ± 0.2 in potato plants. After 10 passages in benthamiana, PVYNb/Nb.L1 infectivity in benthamiana was 0.8 ± 0.2 and 0.9 ± 0.1 in tomato plants but no potatoes could be infected. PVYNb/Nb.L2 infectivity in benthamiana and tomatoes was 0.9 ± 0.1 but dropped down to 0.2 ± 0.2 in potatoes. After four passages, switching hosts and ending in tomatoes, PVYNb/CTF.L1 infectivity in benthamiana was zero, 0.5 ± 0.3 in tomato and 0.3 ± 0.3 in potato plants. PVYNb/CTF.L2, after eight passages, ending in potato plants, could not infect any hosts. After three passages, PVYNb/MIX.L1 infectivity in benthamiana was 0.8 ± 0.2 and 0.8 ± 0.2 in tomato plants but zero in potatoes, and PVYNb/MIX.L2, after nine passages, could not infect any hosts.
Endpoint infectivity profiles of evolved PVYNb and PVYSt lineages across three hosts. Each tile represents the fraction of 10 plants infected by the endpoint inoculum of a lineage when mechanically inoculated onto benthamiana, potato, or tomato. Raw infection counts are available in Supplementary Table S1.
Following the same approach with the PVYSt isolate, we also found significant effects of the host composition during serial passages in the infectivity of the evolved viruses (χ^2^ = 58.871, 4 d.f., P < .001). Again, with lineages evolved in benthamiana being the most infectious. However, for this viral isolate, significant differences among the three test hosts were observed (χ^2^ = 17.708, 2 d.f., P < .001), with tomato showing more infected plants than the other two hosts. A significant interaction between evolution conditions and test host was also observed (χ^2^ = 35.799, 8 d.f., P < .001). PVYSt/Sl.L1 could not establish infection in any host, while PVYSt/Sl.L2, after one passage, infectivity in benthamiana was 0.3 ± 0.2 and 0.8 ± 0.2 in tomato plants but null in potatoes. PVYSt/St.L1 reached six passages and could not infect benthamiana but had an infectivity of 0.4 ± 0.3 in tomatoes and of 0.3 ± 0.2 in potatoes. PVYSt/St.L2 reached 10 passages but had an infectivity of 0.2 ± 0.2 in potato plants and zero in the other two hosts. In benthamiana, PVYSt/Nb.L1 and PVYSt/Nb.L2 reached 10 passages. Lineage PVYSt/Nb.L1 infectivity in benthamiana was 0.8 ± 0.2, 0.9 ± 0.1 in tomato and 0.2 ± 0.2 in potato, while lineage PVYSt/Nb.L2 infectivity in benthamiana and in tomato was 0.9 ± 0.1 and 0.3 ± 0.2 in potato. In switching hosts, only PVYSt/CTF.L2 achieved one passage and had infectivity 0.3 ± 0.2 in tomato and in potato. Lineage PVYSt/MIX.L1 achieved the second passage and showed infectivity 0.2 ± 0.2 in benthamiana and in potato.
All together, these observations further support that benthamiana is the most permissive host for PVY isolates, consistently supporting sustained infection regardless of the viral lineage. In contrast, potato acted as a highly restrictive environment, limiting the ability of most lineages to maintain infectivity, especially those evolved in other hosts. Tomato exhibited an intermediate behaviour, acting as a sink for some lineages but supporting relatively high infectivity in others, particularly for PVYSt, which performed better in tomato than PVYNb. These results also highlight clear evolutionary trade-offs, where adaptation to one host often reduced the virus’ ability to infect alternative hosts. The outcomes from the MIX and CTF treatments further illustrate how complex host-virus interactions become when the selective environment is dynamic, suggesting that both demographic bottlenecks and adaptive responses jointly shape the infectivity patterns of PVY lineages.
Changes in virulence and symptomatology
Next, we sought to evaluate the virulence of the PVY evolving lineages. Virulence was measured after each of the serial passages. Data in Fig. 6 were fitted to a GLM with a Normal probability distribution and identity-link function. Viral isolate, experimental passage, host treatment, and test host were used as orthogonal main factors, while experimental lineage was nested within the interaction between viral isolate and host treatment. Overall significant differences were found between the two viral isolates (χ^2^ = 4.680, 1 d.f., P = .031), being the marginal mean virulence for PVYNb (0.04 ± 0.03) 4-fold larger than for PVYSt (0.01 ± 0.08). Recall that positive virulence values correspond to inter-branch elongations as a symptom of infection. No main effects were associated for the host environment in which lineages evolved (χ^2^ = 7.488, 4 d.f., P = .112) nor for the host in which virulence was tested (χ^2^ = 3.542, 2 d.f., P = .170). However, significant differences among viral lineages evolved in a particular host environment and the host species in which virulence was evaluated have been found (χ^2^ = 8.441, 2 d.f., P = .015), confirming that virulence is indeed dependent on the interaction between viral strain, host environment and test host. For example, the largest reduction in growth induced by PVYNb infection was observed for lineages evolved in potatoes and tested in the same host (−0.13 ± 0.08), while the smallest virulence was observed for lineages evolved in potato but tested in tomato (0.00 ± 0.10).
Evolution of virulence (relative effect of infection in plant growth). Data are organized by PVY isolate (columns) and passage treatment (rows). Plant species in which virulence was evaluated are indicated by colours: red: tomato, green: potato, orange: benthamiana. Viral lineages evolved in CTF were tested on the plant species corresponding to each passage. Viral lineages evolved in MIX were tested in all three plant species.
Analysis of the infection impact index (V) revealed complex patterns across lineages. While in classical terms virulence corresponds to reduced host fitness, here positive V indicated stunting. Interestingly, some lineages evolved in benthamiana exhibited negative V values, reflecting abnormal stem elongation rather than enhanced fitness. This phenomenon illustrates a key limitation of using V as a linear measure of virulence: it captures changes in growth but does not differentiate between detrimental stunting and pathological overgrowth. A more complete assessment of virulence would require complementary metrics such as biomass accumulation, reproductive output, or physiological stress markers.
Next, we decided to evaluate the possible effect of the source host on the evolution of symptoms in the most permissive host, i.e. benthamiana. To do so, we monitored the presence or absence of symptoms in inoculated plants over 9 dpi. Mean time to the appearance of first visible symptoms was estimated using the Kaplan–Meier regression of the number of infected plants to days after inoculation. Figure 7 shows the evolution of this mean time along the passage experiment. Remarkably, only the lineages evolved in benthamiana plants and lineage MIX.L2 were able to generate visible symptoms along all the passages (Supplementary Fig. S3). Indeed, for these lineages, a significant negative correlation exists between mean time to symptoms and passage number (partial correlation coefficient controlling for viral isolate and lineage: r_p_ = −0.601, 34 d.f., P < .001), indicating that symptoms tend to appear faster in benthamiana plants as the virus was evolving in this plant host. In other instances, symptoms appeared only sporadically (e.g. lineage PVYNb/CTF.L2 recovered from benthamiana plants or early passages of PVYSt/MIX.L1), making additional statistical analyses unreliable.
Evolution of the mean time for the appearance of the first symptoms in benthamiana plants inoculated with the different evolving lineages (indicated by colours), divided by treatment (Nb, CTF and MIX).
We next asked whether variation in virulence at the end of the evolution experiment could be explained by differences in viral RNA accumulation of the inoculum solutions. Qualitatively, lineages that maintained high virulence on benthamiana also tended to show high endpoint accumulations in this host (Fig. 4), whereas lineages with low or null infectivity on tomato or potato often had reduced viral accumulation. However, across all host-treatment combinations we did not detect a simple one-to-one correspondence between final viral RNA accumulation and virulence class: some treatments with intermediate viral accumulations showed either mild or severe symptoms, and vice versa. These observations suggest that virulence differences cannot be attributed solely to final inoculum titre but instead emerge from the combined effects of host constraints and the demographic history of each lineage.
Taken together, these findings highlight the multifaceted nature of virus infection on plant growth, emphasizing the importance of considering various factors and their interactions in understanding the impact of viral infections on plant phenotypes. Importantly, it is well established that virulence and viral genome accumulation can be uncoupled in plant viruses, and our results are consistent with this view: virulence classes were not simply predicted by endpoint accumulations, indicating factors beyond viral accumulation also contribute to the observed phenotypic differences.
Genome polymorphisms rising during experimental evolution
Now we focus on characterizing the genomic modifications observed in selected time points. Sequencing was based on HTS from the original isolates, those at 4^th^ passage and the latest viable passage for all the 20 lineages. Firstly, the coverage along the genome differed for each sample (Supplementary Fig. S4). Seven samples did not meet our threshold of 100× average coverage across the genome and were excluded from further analysis due to potential biases [PVYNb/Sl.L1 and PVYSt/Sl.L2 (both 1^st^ passage), PVYNb/St.L1 and PVYNb/St.L2 (both 5^th^ passage), PVYNb/CTF.L2 (8^th^ passage), PVYSt/Sl.L1 (4^th^ passage), and PVYSt/Sl.L1 (5^th^ passage)]. Interestingly, the tomato plant was the immediate previous host of PVYNb/CTF.L2 at 8^th^ passage and PVYSt/Sl.L1 at the 4^th^ and 5^th^ passage, consistent with observations described above that tomato is a sink host for PVY replication. Their low coverage could lead to reduced accuracy in variant calling and lower confidence in quantitative analyses, thereby introducing uncertainty.
Using the reliable dataset, we calculated the genetic differentiation between every population pair using pairwise F_ST_ (Fig. 8a). Low F_ST_ values indicate genetically similar populations with extensive sharing of variants, whereas high values signal strong differentiation and limited overlap in allele frequencies.
(a) Pairwise FST between populations derived from each viral isolate. The left heatmap displays populations evolved from PVYNb and the right panel those from PVYSt. Colours represent the degree of differentiation, from dark (FST ≈ 0, genetically similar) to light (FST > 0.5, strongly differentiated). Rows and columns show host treatment, lineage, and passage number. (b) Average pairwise distance (APD) of each lineage showing changes in population diversity.
Overall, genetic differentiation between populations was higher for PVYNb (mean F_ST_ = 0.59) compared to PVYSt (mean F_ST_ = 0.42). PVYNb populations can be divided into two clusters: (i) the first cluster, which includes the initial population, also comprises the tomato population Sl.L2 at the 1^st^ passage and the benthamiana populations. Within this cluster, lineages L1 and L2 remained highly similar, and F_ST_ values were consistently low across passages, indicating limited divergence and extensive sharing of variants over time in tomato and especially in the permissive host benthamiana; (ii) the second cluster comprises potato, CTF at the 4^th^ passage (tomato) and MIX lineages. These showed markedly higher F_ST_ values relative to the first cluster, suggesting reduced gene flow and more pronounced population differentiation, likely driven by repeated bottlenecks or host-specific constraints. An exception was MIX.L1 at the 3^rd^ passage, which was closer to the first cluster.
PVYSt populations were overall more homogeneous but still resolved into two clusters: (i) the first cluster includes the initial PVYSt, benthamiana populations, CTF.L2 (1^st^ passage), and mixed population MIX.L1 (2^nd^ passage). These populations showed uniformly low F_ST_ values and thus remained genetically close to the initial inoculum, again consistent with the permissive nature of benthamiana and the limited divergence observed in early passages; (ii) the second cluster includes potato populations with elevated F_ST_. This pattern suggests that the passage through potato plants filtered certain haplotypes and that these hosts may exert a selective pressure that reduces genetic diversity. In contrast, this filtering effect was not observed after passage through benthamiana and tomato plants, for which the virus populations tended to maintain greater genetic diversity and remained more closely related to the initial population.
Genetic diversity, measured by APD (Fig. 8b), fluctuated depending on the host and passage. In most cases, APD was lower at the initial inoculum, which can indicate genetic stability. For PVYNb, APD decreased in tomato Sl.L2. Other populations showed higher APD compared to the initial population, with St.L1 (4^th^ passage) and MIX.L1 (3^rd^ passage) having the highest APD.
Notable increases in APD observed in PVYNb/St.L1, PVYSt/St.L1, and PVYSt/St.L2 at the 4^th^ passage might reflect the isolate adaptative response to potato plants, driving the generation of new variants that can better exploit the host environment. Curiously, the APD decreased in later passages in potato plants. The PVYSt/St.L1 at the 6^th^ and PVYSt/St.L2 at the 10^th^ passage showed a reduction in the genetic diversity, potentially indicating fixation of specific mutations, population homogenization or stabilization of the virus within the host environment. In contrast, the APD slightly increased in benthamiana populations from 4^th^ to 10^th^ passages to both isolates. This suggested a slight increase of genetic variability, indicating that over the time, viral populations might be experiencing less bottleneck effects or more balanced selective pressures, allowing a broader range of genetic variants to coexist.
The number of SNPs differed between populations, reflecting a combination of demographic processes, such as population bottlenecks and replication dynamics, and potential selective pressures (Table 1). Comparing to the initial population (PVYNb time 0), PVYNb/Sl.L2 retained 20% of segregating SNPs in the 1^st^ passage, while potato, CTF and MIX lineages presented none. SNP percentages in PVYNb/Nb.L1 decreased from 40% at the 4^th^ passage to 0% at the 10^th^, while PVYNb/Nb.L2 decreased from 30 to 20%. For PVYSt, PVYSt/St.L1 retained 12.5% from the 4^th^ to 6^th^ passage, and PVYSt/St.L2 decreased from 18.75 to 12.5% from the 4^th^ to the 10^th^ passage. SNP percentages in PVYSt/Nb.L1 decreased from 12.5 to 6.25%, and in PVYSt/Nb.L2 from 18.75 to 12.5% from the 4^th^ to 10^th^ passage. PVYSt/MIX.L1 had 6.25% at the 2^nd^ passage, and PVYSt/CTF.L2 had zero at the 1^st^ passage. A mutation introducing a premature stop codon (C737U; polyprotein position 565; reference codon CGA, encoding arginine) becomes fixed in PVYNb/St.L1 passage 4^th^, PVYNb/MIX.L2 passage 9^th^, and PVYNb/CTF.L2 passage 4^th^. It is also highly prevalent in PVYNb/CTF.L1 passage 4^th^ (> 98%) and MIX.L2 passage 4^th^ (> 99%). However, a second SNP at another position within the same codon (A739C; polyprotein position 567) converts this stop codon (UGA) into cysteine (UGU). This compensatory mutation occurs at > 99% frequency in all samples where C737U is present, except in PVYNb/MIX.L2 passage 9^th^, where its frequency is 70.9%, indicating the coexistence of truncated and full-length genomes in this population.
These fluctuations suggest that the dynamics of SNP retention and fixation are influenced by lineage-specific patterns of population size and replication success across hosts. Indeed, hosts that imposed stronger bottlenecks, such as potato and CTF, were associated either with complete loss of segregating SNPs or with accumulation of many fixed substitutions (e.g. PVYNb/St.L2 and PVYNb/CTF.L1), whereas permissive hosts like benthamiana maintained polymorphism with little or no fixation. As shown in Fig. 9, lineages with numerous fixed changes tend to cluster substitution in specific genomic regions and include a mixture of synonymous and nonsynonymous variants, while most PVYSt lineages either lack fixed SNPs or harbour only a small number, predominantly synonymous (e.g. PVYSt/St.L1 at passage 6^th^). The combined patterns of entropy and fixation are consistent with host-dependent demographic bottlenecks and drift as major drivers of genome changes during the passage experiment.
Genome-wide distribution of Shannon entropy at each site (black dots) along PVY genome, with fixed SNPs superimposed as vertical bars. Synonymous fixed mutations are shown in dark grey, nonsynonymous in red and those present at untranslated regions in light grey.
Overall, while patterns of SNP retention and fixation suggest differences in the evolutionary dynamics between PVYNb and PVYSt, a clearer understanding requires evaluating whether these mutations are subjected to differential selective pressures. Therefore, we next analysed the distribution of nonsynonymous and synonymous mutation across lineages to assess whether the observed genomic changes are better explained by adaptive evolution or neutral processes driven by population dynamics.
Estimation of selective pressure on evolved PVY populations
To assess whether the genomic changes observed across lineages were primarily driven by selection or by neutral processes, we computed synonymous (πS) and nonsynonymous (πN) nucleotide diversities and their ratio (πN/πS) on a per-cistron basis (Fig. 10). Full cistron-level diversity summaries, including πN and πS distribution, are provided in Supplementary Fig. S5.
Cistron-specific πN/πS estimates. Points represent individual populations (colour-coded by treatment). The dashed line indicates πN/πS = 1.
Overall, πN/πS < 1 for the majority of cistrons and populations indicated pervasive purifying selection acting on the PVY polyprotein. Notably, a small number of isolated cistron-lineage combinations showed πN/πS > 1, and all of these occurred in benthamiana at the 10^th^ passage: P1 (PVYNb/Nb.L2), NIa-Pro (PVYNb/Nb.L1), and CP (PVYNb/Nb.L2), the latter attaining the largest πN/πS estimate, ~ 8). For PVYSt, πN/πS > 1 was observed only in NIa-Pro for PVYSt/Nb.L2 passage 10^th^. These late, benthamiana-specific peaks, suggest possible lineage-restricted diversification in genes involved in replication (P1 and NIa-Pro) and encapsidation (CP).
However, several important caveats temper a straightforward adaptive interpretation. Firstly, elevated πN/πS estimates frequently coincided with very low πS, a condition that can inflate the ratio in small or bottlenecked populations. Secondly, the peaks were not replicated across independent lineages or isolates and were restricted to single late passage samples, a pattern more consistent with stochastic events (drift, demographic idiosyncrasies or occasional hitchhiking) than with recurrent positive selection. Thirdly, populations that accumulated many fixed nonsynonymous substitutions in restrictive hosts (e.g. PVYNb/St.L2 and PVYNb/CTF.L1) did not show elevated πN/πS across cistrons, further supporting a role for drift under severe host-imposed bottlenecks, rather than directional selection.
In summary, the cistron-level πN/πS analyses indicate that purifying selection is the dominant force shaping PVY diversity, with rare lineage-specific deviations compatible with weak or transient adaptive events observed primarily in benthamiana at late passages.
As a final step, a ML phylogenetic tree was constructed using all consensus genomes, as depicted in Supplementary Fig. S6. Note that sequences that were previously excluded due to not meeting the 100× average coverage threshold were retained here. The tree shows well-structured group formations with highly intriguing clades. PVYNb and PVYSt isolates tended to cluster with isolates derived from the same initial virus. Additionally, the host played a significant role in shaping the phylogenetic structure, as seen in benthamiana isolates from both viruses, which showed to be closely related to the initial population and to each other. Indeed, the initial PVYNb clustered with all the benthamiana treatments (Nb.L1 and Nb.L2 at 4^th^ and 10^th^ passage) and tomato isolates in the first generation, a pattern also observed for PVYSt. This indicated that source hosts such as benthamiana preserve ancestral genomic signatures by enabling large effective population sizes, whereas tomato restricts diversification by acting as a sink host with low viral accumulation. It may explain why tomato plants in general behaved as sink host for PVYNb and PVYSt, since these isolates were not adapted to tomato plants, replication was severely limited and, thus, production of new variants was unlikely.
Another notable aspect is the tendency of potato isolates to remain grouped together, often forming a distinct clade separate from other isolates. This observation highlights the potato as a host that can rapidly favour the fixation of certain genomic changes within a short time frame. These alterations result in isolates that cluster differently from the initial population. For example, while PVYSt/St.L2 in the 4^th^ passage appears significantly distinct from the original population, by the 10^th^ generation, this difference diminishes. This shift underscores the strong bottleneck effects during passage in the potato host, where isolates with new genomic characteristics can emerge. This clustering pattern reflects the influence of the host and passage number on the phylogenetic relationships among virus populations, a pattern consistent with the strong bottlenecks imposed by potato. Many of the fixed SNPs observed in these lineages did not correspond to elevated πN/πS, indicating that fixation likely resulted from drift in small effective populations rather than from widespread positive selection.
Discussion
In this study, we have explored how host permissiveness and host alternation influence the evolution of PVY. By experimentally evolving two PVY isolates across three solanaceous hosts, benthamiana, potato and tomato, alternating hosts, and in mixed-host regimes, we examined how demographic factors like bottlenecks and selection shape viral persistence, genetic diversity, and adaptation.
Our results show that PVY evolution is shaped by a balance between adaptation and genetic drift, with outcomes varying by host. In benthamiana, the most permissive source host, we observed consistent increases in viral titers and earlier symptom onset, suggesting adaptation. In contrast, viral accumulation and virulence in tomato and potato fluctuated without clear directional trends, indicating limited or inconsistent adaptation. Repeated lineage extinctions and reduced infectivity, particularly in tomato, point to strong genetic bottlenecks in a sink host. These bottlenecks likely magnified genetic drift, limiting the role of selection in restrictive hosts. The patterns of SNP fixation and declining diversity in several lineages further support the dominance of drift in sink environments.
Population differentiation analysis further clarifies how host environments shaped PVY evolution. Populations evolving in benthamiana showed uniformly low F_ST_ and modest increases in APD, consistent with a highly permissive host that maintains large effective population sizes and extensive sharing of variants (Elena and Sanjuán 2007, Goodin et al. 2008). In contrast, potato passages produced markedly higher F_ST_ and transient increases in diversity, matching empirical evidence that restrictive hosts impose strong bottlenecks and accentuate drift-driven divergence in plant RNA viruses (Li and Roossinck 2004). Tomato showed an intermediate pattern, consistent with multi-host evolution studies where host-specific constraints differ in strength (Bedhomme et al. 2011). Host alternation (CTF) occasionally generated highly differentiated lineages, as expected when permissive and restrictive phases interact to modify bottleneck severity (Fabre et al. 2012). Overall, these patterns support a source-sink dynamic in which benthamiana maintains genetic connectivity, while potato acts as a restrictive sink that fragments populations through repeated demographic constraints.
Taken together, our results suggest that PVY evolution during serial passage is shaped by a dynamic interplay between adaptation and genetic drift. The relative contribution of each process depends on host permissiveness, bottleneck severity, and the standing genetic variation in the viral population.
These findings are consistent with general models of viral emergence, where novel host infection typically begins with low fitness in a novel host, followed by either extinction or eventual adaptation (Morse 1995). In our study, potato served as the native host, while tomato and benthamiana were novel. Our use of mechanical inoculation likely reduced the transmission bottlenecks that occur during vector-mediated infection and may differ from natural aphid-mediated transmission (Da Silva et al. 2020). Our mixed-host treatments simulated coarse-grained temporal heterogeneity, where the virus alternated between permissive and restrictive hosts. This setup mirrors agricultural practices like crop rotation (Bargués-Ribera and Gokhale 2020). We found that exposure to permissive hosts helped prevent lineage extinction in restrictive environments, supporting the idea that intermittent access to favourable hosts can act as a rescue mechanism and promote adaptation (Morse 1995).
PVYNb, originally maintained in benthamiana, showed greater evolutionary flexibility and higher standing genetic variation, which likely enhanced its adaptability. However, this also made it more prone to bottlenecks when transferred to restrictive hosts. PVYSt, isolated from potato, exhibited greater genomic stability and stronger purifying selection, suggesting a more conservative evolutionary strategy (Cuevas et al. 2012). In permissive hosts, large effective population sizes limit the impact of drift and allow purifying selection to efficiently remove deleterious variants, stabilizing viral genomes (Hillung et al. 2014, Martson et al. 2017, González et al. 2019, Navarro et al. 2022, Van Insberghe et al. 2024).
Signals of positive selection were scarce and only a few combinations of lineage and cistron showed πN/πS > 1, all restricted to late passages in benthamiana. None of these signals were replicated across lineages or across isolates, and several occurred in cistrons with very low πS. Taken together, these results indicate that purifying selection predominated across the PVY polyprotein, and that most genomic changes resulted from drift rather than sustained adaptive evolution.
Similar idiosyncratic and drift-dominated trajectories have been reported for plant DNA viruses, such as cassava begomoviruses, where stochastic changes and demographic bottlenecks outweighed directional selection (Aimone et al. 2021). Comparable patterns are also well documented in RNA viruses and phages, where repeated host-imposed bottlenecks often drive rapid fixation of neutral or mildly deleterious mutations, even under conditions where selection could in principle operate efficiently (Barrick and Lenski 2013, Cuevas et al. 2015). These studies collectively demonstrate that strong genetic drift, restricted population sizes and host-specific replication constraints consistently shape short-term viral evolution. The predominance of purifying selection and the lineage-specific, non-replicated deviations in πN/πS observed in our experiments therefore align with general principles emerging from virus evolution work across multiple biological systems.
Our findings highlight how host variability and switching frequency influence virus persistence and adaptation, factors often overlooked in models of viral emergence (Coffey and Vignuzzi 2011, Novella et al. 2012, Hillung et al. 2014). Although tomato functioned as a sink host under our experimental conditions, recent PVY outbreaks in Brazilian tomato crops (Lucena et al. 2025) demonstrate that this species can support infection under field conditions. This discrepancy may be due to differences in isolate adaptation, environmental factors, or cultivar susceptibility, factors not fully captured under our controlled conditions.
In conclusion, this study underscores the complex interplay between host permissiveness, genetic drift, and selection in shaping PVY evolution. The greater adaptability of PVYNb and the genomic stability of PVYSt illustrate how evolutionary strategies can diverge depending on the host environment and transmission dynamics. A better understanding of these processes is essential for anticipating and managing the emergence of plant viruses. Future research should investigate the impact of host cultivar choice on viral evolution, the effects of prolonged incubation periods, environmental influences on adaptation, and how these interactions play out under real-world field conditions.
Supplementary Material
Supplementary_file_for_review_veag010
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Agudelo-Romero P, de la Iglesia F, Elena SF. The pleiotropic cost of host-specializati Eon in tobacco etch potyvirus. Infect Genet Evol 2008;8:806–14. 10.1016/j.meegid.2008.07.01018765302 · doi ↗ · pubmed ↗
- 2Aimone CD, Lavington E, Hoyer JS et al. Population diversity of cassava mosaic begomoviruses increases over the course of serial vegetative propagation. J Gen Virol 2021;102:001622. 10.1099/jgv.0.00162234310272 PMC 8491896 · doi ↗ · pubmed ↗
- 3Antia R, Regoes RR, Koella JC et al. The role of evolution in the emergence of infectious diseases. Nature 2003;426:658–61. 10.1038/nature 0210414668863 PMC 7095141 · doi ↗ · pubmed ↗
- 4Bargués-Ribera M, Gokhale CS. Eco-evolutionary agriculture: host-pathogen dynamics in crop rotations. P Lo S Comput Biol 2020;16:e 1007546. 10.1371/journal.pcbi.100754631945057 PMC 6964815 · doi ↗ · pubmed ↗
- 5Barrick JE, Lenski RE. Genome dynamics during experimental evolution. Nat Rev Genet 2013;14:827–39. 10.1038/nrg 356424166031 PMC 4239992 · doi ↗ · pubmed ↗
- 6Bedhomme S, Lafforgue G, Elena SF. Multihost experimental evolution of a plant rna virus reveals local adaptation and host-specific mutations. Mol Biol Evol 2011;29:1481–92. 10.1093/molbev/msr 31422319146 · doi ↗ · pubmed ↗
- 7Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIAMOND. Nat Methods 2015;12:59–60. 10.1038/nmeth.317625402007 · doi ↗ · pubmed ↗
- 8Coffey LL, Vignuzzi M. Host interaction of chikungunya virus increases fitness while restricting population diversity and adaptability to novel selective pressures. J Virol 2011;85:1025–35. 10.1128/JVI.01918-1021047966 PMC 3020036 · doi ↗ · pubmed ↗