Plant community diversity influences vector behaviour and Barley yellow dwarf virus population structure




The species composition of a plant community can affect the distribution and abundance of other organisms including plant pathogens. The goal of this study was to understand the role of host diversity in the transmission of two Barley yellow dwarf virus (BYDV) species that share insect vectors and hosts. Greenhouse experiments measured the transmission rate of BYDV species PAV and PAS from infected oat plants to healthy agricultural and wild grasses and from these species back to healthy oat seedlings. In the field component of the study, the rate of spread of PAV and PAS was measured in monoculture plots planted with agricultural grasses. In greenhouse experiments, the aphid vector more readily transmitted PAV from agricultural grasses and more readily inoculated PAS to the wild grass species assayed. In the field experiment, disease prevalence was greater in wheat, but there was no difference in the rate of spread of PAV and PAS. These results indicate an interaction between vector and host genotype that selects for greater PAV transmission in grain crops, contributes to differences in disease prevalence between grass types, and maintains pathogen diversity within the larger plant community (i.e. agricultural and non-agricultural hosts).


Hosts are the context for pathogen reproduction and selection. The quantity of available hosts and the relative ability of a host type to support a productive pathogen infection (termed competence) determine the degree of pathogen multiplication and transmission. If there are different host types in a community, each may elicit disparate responses from a given pathogen genotype. The present study explores how host diversity influences disease prevalence and pathogen population structure. This is accomplished by experimental assessment of phenotypic traits that affect the dispersal of two closely related species of Barley yellow dwarf virus (BYDV; Luteoviridae).

BYDV infects a wide array of graminaceous species (including wild grasses, cereals and grasses used in biomass production) and is obligately aphid transmitted. A feeding period of several hours may be required for an aphid to transmit BYDV (Gray et al., 1991). Consequently, BYDV species are distributed in a plant community according to the host preference of the aphid vector (Power & Remold, 1996; Leclercq-LeQuillec et al., 2000). Within the limits set by aphid feeding behaviour, host genotype may influence pathogen population structure by host susceptibility to infection.

PAV is the BYDV species most often identified in grain crops (Clement et al., 1986; Fabre et al., 2005) and yield reductions >60% have been reported (Riedell et al., 2003). Another species of BYDV, PAS, produces more severe symptoms on susceptible and resistant varieties of oat (Chay et al., 1996) and barley (Bencharki et al., 1999). In addition to having sympatric distributions (Chay et al., 1996; Mastari et al., 1998; Bencharki et al., 1999), PAV and PAS share the same vector species –Rhopalosiphum padi and Sitobion avenae. PAS and PAV are distinguished by approximately 10 and 22% nucleotide divergence in their coat protein and replication related genes, respectively (Bencharki et al., 1999; Hall, 2006). However, the majority of published field surveys conflate the prevalence of PAV and PAS because they are indistinguishable by the most commonly employed antibody-based survey methods (Chay et al., 1996).

Virus concentration in leaf tissue influences aphid transmission of cucumoviruses (Bromoviridae; Pirone & Megahed, 1966; Banik & Zitter, 1990), Potato virus Y (Potyviridae; Marte et al., 1991) and BYDV (Gray et al., 1991, 1993). There may then be a link between host identity and BYDV population structure via the effect of hosts on the probability of aphid acquisition of a given genotype. There is only a handful of studies that examine how the host-pathogen interaction affects virus disease epidemics (Padgett et al., 1990; Gray et al., 1994; Steinlage et al., 2002). In each of these studies, within-host virus dynamics were manipulated through the use of susceptible and resistant host genotypes, and in every case plots with resistant plants (in which virus replication would be restricted) had significantly less spatial or temporal disease spread. No study has yet examined the epidemic spread of disease while varying both the host and virus genotype.

The present study compares PAV and PAS replication and aphid transmission in the greenhouse, using agricultural grasses Hordeum vulgare (barley cv. Romulus), Avena sativa (oat cv. Astro) and Triticum aestivum (wheat cv. Barry), as well as a weedy wild species, Lolium multiflorum (Italian rye grass). Virus epidemics were also initiated and monitored in monoculture field plots planted with the same agricultural grasses. The context for the experiments is provided by a survey over one growing season of PAV and PAS populations in agricultural fields in central New York state. The combination of approaches undertaken in this study allows evaluation of the effects of within-host virus population dynamics and host diversity on the structure of a virus population in a managed plant community.

Materials and methods

Field survey of PAV and PAS populations

From May to July 2002 wheat plants were collected from agricultural fields in central New York State. Only field edges were sampled but within this area plants were selected randomly. Double antibody sandwich-enzyme linked immunosorbent assay (DAS-ELISA) with PAV polyclonal antibodies (Agdia IN) was performed (as described by Gray et al., 1991) to detect virus infected plants. PAV polyclonal antibodies also react with PAS isolates (Chay et al., 1996). Total nucleic acids were extracted from plants that showed an antibody reaction according to the protocol provided by Hall (2006). Reverse transcription-PCR (RT-PCR) was used to amplify the major coat protein gene (Hall, 2006), followed by digestion of the products with the restriction enzyme BstNI (Bencharki et al., 1999) to differentiate PAV and PAS infected plants.

Virus isolates

All virus isolates used in the greenhouse experiments were collected from agricultural fields in central New York state. PAS-129 was isolated in 1992 from migrating alate R. padi alighting on winter wheat, and infected leaf tissue was provided by S. Gray (USDA, Cornell University). PAV-FA2k298 was isolated in 1998 from apterous R. padi collected from oat fields. PAV-WS32 and PAS-WS179 were isolated in the winter of 2004 from wheat plants. To establish isolates in greenhouse culture plants, adult R. padi (from laboratory maintained, disease-free colonies) were fed for 48 h on field collected leaves then transferred to healthy oat seedlings. Aphids were allowed a 5 day inoculation access period, after which plants were fumigated and placed in the greenhouse. The same plant inoculation protocol was used for all subsequent experiments. Hereafter PAV-FA2k298, PAV-WS32, PAS-129 and PAS-WS179 will be referred to as PAV 1, PAV 2, PAS 1 and PAS 2, respectively.

Virus population size and transmission efficiency

Three experiments were conducted to evaluate within-host virus population growth and virus transmission efficiency. Experiment 1 examined the susceptibility to virus infection of barley, oat, wheat and Italian rye grass. Experiment 2 monitored the concentration of PAV and PAS antigen in the three agricultural hosts over time. Experiment 3 compared the transmission success of PAV and PAS by single aphids.

In Expt. 1 plants were individually grown from seed in four-inch pots. At the two-leaf stage (10 days after planting) plants were challenged with two or eight R. padi previously fed on detached oat leaves infected with one of the four virus isolates. The oat plants used as virus source had been infected for approximately 60 days. Six to twelve plants were inoculated per treatment. At 21 days post-inoculation (dpi) leaf tissue from all plants was collected and virus content was analyzed by DAS-ELISA. A logistic regression was used to model the probability that a plant became infected given its species, the number of aphids used to carry out the inoculation, and the virus isolate inoculated. The coefficient of the predictor variable can be interpreted as the change in odds of disease for plants that are in the membership category versus those in the reference category. One can determine the odds of disease (probability of disease/1 – probability of disease) given the predictors in the model by exponentiation of the regression coefficients. Logistic regression was carried out in stata version 9·1 (StataCorp).

In Expt. 2, barley, oat and wheat were grown from seed in four-inch pots. Upon reaching the two-leaf stage, 10 plants of each species were inoculated with PAV 1 or PAS 1. At 8, 20 and 33 dpi the most recently emerged leaf from all plants was collected, weighed, homogenized 1:10 (w:v) with phosphate-buffered saline, and antigen concentration was measured by a semi-quantitative ELISA. At each harvest date the homogenized leaf tissue sample was loaded on two microtitre plates. The same PAV and PAS positive control sap was loaded in duplicate on each plate. The absorbance values for both plates were recorded when the positive controls had the same value on each plate. This procedure allowed samples to be compared between plates on the same harvest date but not between harvests, as new controls were used on each date. A two-factor anova implemented with the r software package (R Foundation for Statistical Computing) was used to compare antigen concentration between treatments. The main effects of the analysis were virus species and plant species. Tukey’s HSD test was used to compare antigen concentration between virus species infecting a single plant species and between plant species for a given virus species.

In Expt. 3, two plants from each treatment of Exp. 2 were randomly selected at 8 dpi. The whole plant was used as inoculum source for R. padi. After a 2 day acquisition period, single aphids were transferred to healthy oat seedlings. At 14 dpi, ELISA was used to determine the presence or absence of infection in the indicator plants. A two-factor anova was used to analyze the rank of the proportion of plants that became infected with PAV versus PAS isolates. The main effects of the analysis were virus species and plant species.

Field experiment

In the summer of 2004 the epidemic spread of PAV and PAS isolates in barley, oat and wheat was evaluated in a field experiment. Plant species and virus isolates were randomly assigned to 4 × 1 m plots arranged in four blocks (156 plots in total). There was 1 m of bare earth between plots in the same block and 2 m of bare earth between blocks. On 7 and 8 June seeds were hand sown in five rows per plot at a density of 2·5 (oat and wheat) or 2 (barley) bushels per acre. Hand weeding of the experimental field was done as needed over the course of the growing season. Plots were uncaged; therefore, one plot of each species per block was uninoculated (control plots) so that one could assess whether plots became contaminated with BYDV from outside the plot. Plants that would serve as the source of virus infection within each plot were inoculated on one of three dates post plant emergence (dpe), 18, 26 and 36 dpe. On the first inoculation date, four plots of each plant species were infected with each virus isolate (eight plants inoculated per treatment). Plots were inoculated by placing 20 viruliferous adult R. padi on two adjacent plants in the centre (2 m from one end) in the centre row. At 26 dpe, 3–6 previously uninoculated plots per block of each plant species were infected with a given virus isolate (6–12 plants inoculated per treatment). Inoculations were carried out as on the first date. On the third date, plots inoculated at 26 dpe were inoculated a second time with 10 viruliferous aphids placed on one plant located 1 and 3 m from the end in the centre row (6–12 plants inoculated per treatment). After all species had reached anthesis, ELISA was used to identify infected plants. Sampling began with the inoculated source plants. Then two contiguous plants on both sides of the source plant were collected. If at least one of the two plants was infected, two additional plants along the centre row were analyzed. If neither of the plants was infected, infection was no longer monitored along that length of row. Sampling was repeated along the row until neither plant in a two plant sample was found to be infected.

Two types of statistical analyses were performed. First, logistic regression was used to model the probability that the inoculated source plant became infected. Predictor variables in the model were plant species inoculated, virus species and plant age at the time of inoculation. To evaluate if the two-way interaction terms significantly improved the ability to predict disease, a likelihood ratio test was applied to models with and without each set of terms (plant species * isolate, inoculation date * plant species, and isolate * inoculation date). Likelihood ratio tests were performed by comparing the change in −2 ln(likelihood) between the full and reduced models against a chi-squared with degrees of freedom equal to the number of parameters dropped from the model (df = 1, α = 0·05). The second analysis performed was a two-factor anova of the distance disease spread per plot. Pre-planned contrasts were used to compare levels of the main effects – PAV versus PAS and barley versus oat versus wheat.


Survey of PAV and PAS in agricultural fields

Of the 262 total plants sampled, 40 were found to be infected with PAV or PAS. Restriction fragment analysis of the virus content in these plants revealed that 24 were infected with PAV, seven were infected with PAS, two were infected with both PAV and PAS and seven had no recoverable RNA.

Virus population size and transmission efficiency

In Expt. 1, barley, oat and wheat plants were inoculated with PAV or PAS and harvested 21 dpi. The odds of disease (OR) were significantly greater for plants inoculated with eight aphids versus two aphids (OR = 3·5, = 0·004; Table 1). There was no difference in the odds of disease among agricultural grasses, but the odds of disease decreased when inoculating the former versus Italian rye grass (OR ≤ 0·008, < 0·001). Compared to PAV 1 there was a significant decrease in the odds of disease when plants were inoculated with PAS 1 (OR = 0·104, < 0·001) or PAS 2 (OR = 0·272, = 0·045). Compared to PAV 2 the odds of disease decreased when inoculations were carried out PAS 1 (OR = 0·19, = 0·005) but not PAS 2 (OR = 0·51, = 0·28). There were also differences among PAS isolates. The odds of disease significantly increased when PAS 2 was inoculated to plants versus PAS 1 (OR = 2·6, = 0·044).

Table 1.   Parameter estimates for logistic regression modelling the odds of Barley yellow dwarf virus infection for greenhouse grown plants inoculated using two or eight Rhopalosiphum padi aphids
Reference categoryMembership categoryCoefficientaOdds of disease (P value)b
  1. aCoefficient of predictor variable in the membership category.

  2. bChange in odds of disease for plants in the membership category versus those in the reference category.

2 Aphids per plant8 Aphids per plant1·253·5 (0·004)
PAV 1PAV 2−0·630·53 (0·36)
PAV 1PAS 1−2·260·104 (<0·001)
PAV 1PAS 2−1·290·27 (0·045)
PAV 2PAS 1−1·620·19 (0·005)
PAV 2PAS 2−0·660·51 (0·28)
PAS 1PAS 20·962·6 (0·044)
BarleyOat0·862·36 (0·103)
BarleyWheat0·431·53 (0·374)
BarleyRye grass−4·780·008 (<0·001)
OatWheat−0·430·65 (0·43)
OatRye grass−5·640·003 (<0·001)
WheatRye grass−5·20·005 (<0·001)

In Expt. 2, virus antigen concentration in barley, oat and wheat at 8, 20 and 33 dpi was estimated by semi-quantitative ELISA (Fig. 1). At 8 dpi virus concentration was dependent on the plant species infected (= 32·5, < 0·0001) but not the virus species (= 0·5, = 0·48). At 20 dpi there was a significant effect of plant species (= 16·5, < 0·0001) and virus species (= 82·3, < 0·0001) on virus concentration. At 33 dpi there was a significant effect of plant species (= 28·5, < 0·0001) and a significant plant species by virus species interaction (= 3·71, = 0·03). The latter was due to the higher concentration of PAS in barley and oat and the higher concentration of PAV in wheat.

Figure 1.

Barley yellow dwarf virus (BYDV)-PAV 1 (white bars) and BYDV-PAS 1 (grey bars) antigen concentration in greenhouse grown barley, oat and wheat harvested 8, 20 and 33 days post inoculation (dpi). At each harvest date, treatments with different letters where found to be significantly different using Tukey’s HSD multiple comparison method.

In Expt. 3, infection status of oat inoculated by single aphids fed on PAV and PAS infected barley, oat and wheat was determined by ELISA at 14 dpi. The number of indicator plants that became infected was dependent on the virus species (= 13·5, = 0·007) in the source leaf but not the plant species (= 0·23, = 0·8). Across plant species 94% (±6) and 63% (±13) of plants became infected with PAV and PAS, respectively (Table 2). When virus antigen concentration in the source leaf was used as a covariate in the analysis it had no impact on effect of virus (= 0·02, = 0·9). This indicates that the greater transmission success of PAV was independent of virus concentration in the source leaf.

Table 2.   Number of healthy oat indicator plants that became infected when inoculated with single Rhopalosiphum padi carrying Barley yellow dwarf virus (BYDV)-PAV or BYDV-PAS isolates from infected barley, oat or wheat
Source plantInoculumReplicate source plantPlants inoculatedPercent infected plants
barleyPAV 1110100
barleyPAV 12989
barleyPAS 111020
barleyPAS 12888
oatPAV 11683
oatPAV 121090
oatPAS 111060
oatPAS 12863
wheatPAV 117100
wheatPAV 1210100
wheatPAS 11880
wheatPAS 121070

Field experiment: infection success and virus spread in field plots

A logistic regression was used to model the probability that inoculated plants became infected (Table 3). Plant age at the time of inoculation (x2 = 41·1, < 0·01) and plant species (x2 = 22·5, < 0·05), but not virus species (x2 = 1·47, = 0·23) were significant predictors of the odds of disease. The odds of disease significantly decreased when plants were inoculated 26 (< 0·01) or 36 dpe (< 0·01) instead of 18 dpe. There was no significant change in the odds of disease when plants were inoculated 26 vs.36 dpe (= 0·16). There was a significant decrease in the odds of disease when inoculating oat versus wheat (< 0·01) and barley versus wheat (= 0·045). The odds of disease significantly increased if barley was inoculated as opposed to oat (< 0·01). The two-way interaction terms were not significant predictors of the probability of disease (data not shown).

Table 3.   Parameter estimates for logistic regression of plants inoculated with Barley yellow dwarf virus in the field
Reference categoryMembership categoryCoefficientaOdds of disease (P value)b
  1. aCoefficient of variable in the membership category.

  2. bChange in the odds of disease for plants in the membership category versus those in the reference category.

  3. cPlants were inoculated 18, 26, 36 days post emergence (dpe).

Inoc. 18 dpecInoc. 26 dpe−1·60·2 (0·0001)
Inoc. 18 dpeInoc. 36 dpe−2·060·13 (0·0001)
Inoc. 26 dpeInoc. 36 dpe−0·460·63 (0·16)
wheatbarley−0·670·51 (0·045)
wheatoat−1·60·2 (0·0001)
oatbarley0·932·5 (0·006)

There was too little disease spread in plots inoculated 26 or 36 dpe to perform a meaningful statistical analysis; therefore, only data for plots inoculated 18 dpe are presented here (Fig. 2). In these plots the identity of the host species (df = 2, = 4·4, = 0·02) but not virus isolate (df = 3, = 0·89, = 0·46) had a significant effect on the distance of disease spread. The distance of disease spread in wheat plots was significantly greater than barley (= 0·03) or oat plots (= 0·009), but there was no difference between barley and oat (= 0·47) plots.

Figure 2.

 Mean distance of spread of Barley yellow dwarf virus (BYDV)-PAV (white bars) and BYDV-PAS (grey bars) isolates spread from inoculated source plants in 4 × 1 m plots of barley, oat or wheat.

To evaluate whether experimental plots may have experienced contamination with outside inoculum, five plants randomly collected between 1 and 3 m of row length in the centre row of the (uninoculated) control plots were assayed for virus. There were three wheat plants and one barley plant infected in block 1, two wheat plants infected in block 2, one wheat plant infected in block 3 and no plants infected in block 4. Based upon these data it seems that there might have been preferential colonization of wheat plots by aphid immigrants. Block was found to have no effect on the measure of disease spread (df = 3, = 0·34, = 0·8), indicating that if there was contamination of plots with virus from outside of the plot it occurred randomly across blocks.


The survey of BYDV population density in local agricultural fields found that the prevalence of PAV was at least three times greater than that of PAS. It is hypothesized that the asymmetrical distribution of virus species in the host population was the result of differences in their growth rate within the host. One might also expect that if virus infection rates are not uniform across plant species, communities with different host compositions might also have a different virus population structure. The greenhouse experiments conducted validated the hypotheses, in part, and demonstrated more complex interactions between vector, virus and plant than had initially been proposed.

In Expt. 1 the transmission success of PAV to barley, oat and wheat was greater than PAS. This experiment alone cannot explain why PAV was transmitted more often than PAS. First, virus concentration in source leaves was not measured. Secondly, PAS infected source plants (especially PAS-129 plants) were shorter and had a greater number of chlorotic tillers and leaves than PAV infected plants. Thus, the higher infection rates of PAV infected plants may reflect virus titre in the source leaf or the quality of the source leaf used for aphid feeding. Italian rye grass differed from the other plant species in that there was a complete failure of PAV isolates to cause infection. This finding coupled with the low, but existent, probability of successful PAS infection of Italian rye grass provides evidence that host diversity facilitates the persistence of diversity in the virus population. It should be noted that at the time of fumigation, aphids inoculated to Italian rye grass, and all other plant species, had produced nymphs. Feeding is a prerequisite for nymph production. Thus, the lack of virus transmission was not due solely to aphid mortality or an inability to feed.

Host selection of virus genotypes has been experimentally demonstrated for an alstroemeria-infecting Cucumber mosaic virus (ALS-CMV) isolate that contains a mutation in the 3′ untranslated region of RNA 3 and RNA 4 (Chen et al., 2002). The variant ALS-CMV isolate had greater fitness on alstromeria than tobacco. Furthermore, when exclusively passaged on tobacco for 5 years ALS-CMV lost the 3′ mutation and had lower replication rates in alstromeria than non-passaged isolates. Another example can be found in Prunus necrotic ringspot virus (PNRSV; Moury et al., 2001). Assays of 27 isolates from rose and Prunus species found low infectivity of Prunus isolates on roses but not the reciprocal, pointing to adaptation of Prunus isolates to their host plants.

In Expt. 3 transmission assays with single aphids were conducted prior to the development of symptoms in source plants (8 dpi) and before there were significant differences in PAV and PAS concentration (Expt. 2, Fig. 1). Nonetheless, PAV transmission was significantly greater than PAS. This suggests that greater transmission success of PAV is independent of virus titre, and due to the interaction between viral proteins and receptors that mediate virus movement across membranes in the aphid. If further testing corroborates the findings of vector selection, the ecological consequence would be a decrease in virus population diversity. Adding to the complexity of BYDV epidemiology is the variability in virus transmission efficiency among aphid vector species and clones of a single species (Zavaleta et al., 2001). In other geographic regions the relative abundance of vector species and the transmission efficiency of each species and existent clones may produce a different disease pattern than has been documented in the current study.

The greater transmission success of PAV isolates in the greenhouse did not translate to a higher incidence of PAV in the field experiment. It is likely that the onset of mature plant resistance arrested disease spread and prevented the development of asymmetry in the distribution of PAV and PAS (Lindblad & Sigvald, 2004). The predominant agricultural practice in New York state is to sow grain crops in the fall and let them overwinter as seedlings. As a result, aphid migrants leaving summer crops have an opportunity to initiate and enlarge disease foci while plants are still very young. In greenhouse Expt. 1, PAS transmission success significantly increased when eight aphids instead of two were used to carry out inoculations. This implies that the effects of vector selection and within-host dynamics on virus population structure will be dependent on the number of vectors present. It is possible that a high aphid density in field plots homogenized any potential differences in the spread of PAV and PAS.

At each harvest date of Expt. 2, virus antigen concentration was significantly greater in oat plants; yet in the field, disease prevalence was greatest in wheat plots. A study by Power (1991) found that the genotype of the host had a significant impact on aphid movement rates and BYDV spread. Taken as a whole the results of the present study indicate that the density of infective vectors, vector activity relative to the availability of susceptible hosts, and vector behaviour in response to the host genotype are important determinants of disease outcomes in the host community. The selection of PAV by R. padi from agricultural hosts can explain its greater prevalence in wheat fields. Selection of PAS by the wild Italian rye grass suggests a mechanism by which the virus species is not competitively excluded from the population by PAV. A major unanswered question of the present study is how does diversity in the virus population influence host population structure? In order to address this question, future research should conduct a detailed survey of PAV and PAS population densities in hosts at agricultural and non-agricultural sites coupled with measures of host mortality and fecundity.


The authors thank Kai Blaisdell, John Chau and Craig Kahlke for field and laboratory assistance. PCR and restriction fragment analysis were performed in the Evolutionary Genetics Core Facility in the Department of Ecology and Evolutionary Biology at Cornell University. Research was supported by the National Institutes of Health Ruth L. Kirschstein individual research fellowship and the Cornell University Andrew W. Mellon student research grant.