Beet mosaic virus : epidemiology and damage

<strong>Overview:</strong>The aim of the studies described in this thesis was to obtain a thorough understanding of the main factors determining the spread of a potyvirus in a high plant density crop. The factors studied included the relationships between virus, host and vector, the spread of the virus around an initial virus source consisting of one or more infected plants, the spread of the virus by the prevailing aphid population, and the effect of plant density on the spread of the virus. A time-save sampling technique was developed and the damage caused was estimated. This study was made with the system beet - beet mosaic virus (BtMV), a potyvirus infecting sugar beet, as a model pathosystem. Sugar beet is a herbaceous plant widely cultivated in The Netherlands. The crop, which has a cycle of approximately 8 months, is cultivated in fields at a density of 7 to 10 plants/m <sup>2</SUP>. The disease in this crop is polycyclic, as several infection cycles occur during the growing season.The spread of a potyvirus in a crop starts with a primary infection, either introduced by migrating aphids from sources outside the field or by the use of infected seed or propagative plant material. These plants form the sources from which the virus is spread secondarily in the field. The primary infections are in most cases scattered over the field, whereas secondary infections are aggregated around early-infected plants. Studies on the spread of a virus from a known source are few, as primary introductions are difficult to prevent in many crop virus system. Primary infections are frequently introduced at erratic moments and increase in virus incidence, due to plants infected from outside sources is superimposed on the secondary spread ongoing within the field. As BtMV is only rarely encountered in The Netherlands and not seed transmitted, this pathosystem is a good model to analyze secondary spread using a known virus source in the experimental plots. Spread was expected only to occur from these sources and not from outside sources.<strong>Development of a time saving transect sampling method:</strong>Spread of BtMV occurred around the virus source in a clustered isotropic pattern with a negative exponential gradient. Such a spread is common for polycyclic epidemics of potyviruses in annual crops (Dahal, 1992; Eckel and Lampert, 1993; Nelson and Campbell, 1993; Perring et al., 1992). The isotropic spatial pattern of spread found in all plots showed that a simple sampling method, called transect sampling method, could be developed and used to monitor the development of the infection. This method consisted of monitoring the plants on two orthogonal transects extending diagonally across the rows from the source plants in the plot. In the analysis of transect data, the uneven representation of the sampled plants at each distance class must be taken into consideration. The temporal and spatial spread of the BtMV disease could be described as reliably using the transect method, as by monitoring the whole plot, provided that a lower precision per repetition is compensated by raising the number of repetitions. This result suggests that by using this less labor intensive and less time consuming sampling method, more sites or more treatments can be studied.This sampling method can also be applied to study the spread in other pathosystems such as <em>Papaya</em><em>ringspot</em><em>virus</em> (PRSV) in cucurbits, <em>Potato</em> vi <em>r</em> us <em>Y</em> (PVY) in crops of various solanaceous species, and <em>Soybean mosaic virus</em> (SMV) in soybeans, when the virus source is known or can be found. This sampling method can potentially also be applied to semi-persistently and persistently transmitted viruses such as <em>Beet yellows virus (</em> BYV and <em>Beet mild yellowing virus</em> (BMYV) (van der Werf, personal communication), which form usually clusters with an isotropic spatial pattern around the primarily infected virus source, in a similar fashion as BtMV.<strong>Modeling spread as a function of migrating aphid flights:</strong>Under natural conditions, aphids transmit potyviruses in a non-persistent manner. Although the interaction between the virus and the vector is specific (Shukla et al., 1994), apparent specificities in the epidemiological relationships between potyviruses and aphid species have not been elucidated. The role of the individual aphid species in the spread of potyviruses has been analyzed by different analytical methods. The simplest approach is to plot virus incidence and number of aphids counted on plants or collected with traps on a common time axis and to subjectively compare the curves obtained for the spread and the number of aphids obtained for each individual species or the total aphid population. Eckel and Lampert (1993), van Hoof (1977) and Karl et al. (1983) used this approach but could not find any relation between the species and the spread of the potyviruses studied. A pitfall of this approach is that population trends of different aphid species over time may be collinear (Chapter 4). Thus, the role of one species might not be isolated from the other. Correlation and regression analysis also usually fail to relate spread to aphid species or total counts (Madden et al., 1987; Mora-Aguilera et al., 1992; Watson and Healy, 1953).Garrett (1988) demonstrated that, in lupine, <em>Clover yellow vein virus</em> was mostly spread by two aphid species ( <em>Aphis craccivora</em> and <em>Myzus persicae</em> ) using multiple regression analysis to relate the rate of spread of this potyvirus to the species that compose the aphid population. In the studies described here, no single species could be associated with the spread applying correlation or regression analyses. A good correlation could be detected between the total daily number of alatae caught and the spread of BtMV.Based on the collected data, a deterministic simulation model was developed to study the spread as a function of the migrating aphid population (Chapter 4). This model was based on a logistic population growth applied to plant diseases (van der Plank, 1963). The rate of the disease was, in this model, proportional to the virus sources, healthy plants, latent and incubation periods of the virus in the plant, the total number of aphids caught in a suction trap (not discriminating species) and a parameter ( <em>r</em> ) that represented all aspects of vector activity relevant to virus spread (Jeger et al. 1998). This parameter <em>r</em> , describing the relationship between the daily catches of aphids and the number of newly infected plants, was quite robust among experiments. Remarkably, <em>r</em> appeared to be independent of the moments at which the primary inoculum sources were introduced, confirming that the chosen model and common parameter value give a seasonable mechanistic description of epidemics started at different dates. These results confirm and extend the conclusions of Di Fonzo et al. (1997), Madden et al. (1987), Mora-Aguilera et al. (1992) and Nemecek (1993), that migrating aphids, regardless of the species, play a major role in the spread of non-persistently transmitted viruses.The simulation model used in this study only accounted for secondary spread as introductions from external virus sources rarely occur in the Netherlands. The absence of any spread from outside sources allowed inoculating the field at different dates. This simulation model could simulate the final number of plants showing symptoms. A rough approximation, using the averaged obtained value for <em>r</em> and aphid catches showed that the number of infected plants to occur could be predicted two weeks in advance. The use of this model as a predictive tool for the whole crop cycle is premature because it does not model the development of the aphid population. Although <em>r</em> was conserved between the inoculation dates in each experiment, it varied between experiments. The species composition of the aphid population varies each year. Although the total number of aphids caught could be related to virus spread, the rate by which the virus will be spread will differ among years and among locations. In more elaborate models, <em>r</em> must be decomposed into different components representing the behavior of the aphid population such as the acquisition and inoculation rates, the infectious period of the virus in the vector, the vector turn over, the feeding time per vector per day, and the distance hopped by aphids (Jeger et al., 1998).The simulation models used by Nemecek (1993) and Sigvald (1992), to predict <em>Potato virus Y</em> spread in potatoes, included some behavioral characteristics and a more detailed description of the aphid species composition. These models could be used to simulate the final disease incidence in crops, which were initially infected with different numbers of virus sources. The studies in <em>Soybean mosaic virus</em> presented by Ruesink and Irwin (1986) also included some behavioral aspects of the vector and could be used to predict yield and level of seed transmission. The complementary information added by the present study was an experimentally demonstration that spread of BtMV is related to the major migrating aphid flight. The calibration studies using a simulation model confirmed that this spread could be described by one absolute rate parameter. It can be concluded that management strategies to control virus spread have to be focused on a delay of virus introductions in the field, or alternatively, to restrain aphid dispersal early in the season.The deterministic model developed in Chapter 4 was adapted to include a factor that describes the effect of plant density in the rate equation. This factor could be included on the assumption that plant density would affect the spread by affecting the number of aphids per plant, and the number of available plants, while the other parameters related to spread would remain constant. The spread was indeed inversely proportional to the plant density in the first weeks after the virus started to spread. However, analyzing the incidence for the whole growing season, the model failed to explain the observed spread. Values of <em>r</em> estimated by calibration, to experimental field data, showed that the rate of spread in low-density plots was lower than the rate expected by the hypothesis that spread is proportional to the number of aphids per plant. The factors that lead to the strong aggregation of the infected plants around the primarily infected plant might have affected the rate of spread in these low-density plots.The contrast between bare soil and plants will be larger in low-density plots than in plots with standard density. By the middle of July, when most of the aphid migration occurred in both years, the canopy was closed in the standard density plots while bare soil was still visible in the low-density plots. An attraction exerted on the aphids by the contrast between plants and bare soil might have affected the mobility of the aphids within the plot, reducing the distance hopped between plants and, consequently, the spread of the disease as the infected plants might be re-inoculated frequently. A factor considering this distance and/or the spatial pattern of the spread must be included to improve the model. Improvements of this model must, therefore, concentrate on the inclusion of a set of equations representing the vector behavioral components and the spatial pattern of spread.<strong>BtMV and damage in sugar beet:</strong>Experiments to determine damage due to virus infections are laborious. It is assumed that yield will depend on date of inoculation, initial inoculum levels, rate at which the virus spreads, disease incidence at the moment of harvest, and others. Since many factors will affect the yield, it will be difficult to estimate the crop losses caused by a pathogen. The use of a crop growth model can overcome these difficulties, assuming that the parameters related to damage can be determined and incorporated in the model. Information on damage caused by BtMV is rare in the literature, but it is generally accepted that this disease has little impact on yield of sugar beet (Watson and Watson, 1953).The effect of BtMV infections on the yield of the sugar beet crop was evaluated by simulation using a crop growth model (SUCROS). This analysis was experimentally grounded by determination of the light response curve, light absorption and transmission, and other parameters on healthy and BtMV infected leaves showing mosaic symptoms. The model dynamically simulates the carbon budget and growth of the crop by integrating leaf photosynthesis over time and leaf area, taking into account incident light, leaf area index, proportion of mosaic-affected leaf area, and optical characteristics of the leaves and the canopy. The damage simulated for early-infected crops was estimated to be, under the most extreme situation, approximately 20%. However, as the infection usually starts to spread in the second half of the growing season, the estimated damage in a fully infected crop after July was less than 3%. This value can be neglected considering the damage due to other diseases, harvesting and processing of the roots. Injury component analyses indicated that the direct effect due to both reduction in maximum rate of photosynthesis ( <em>P</em>_m ) and increase in dark respiration ( <em>R</em>_d ) were the major causes of the simulated damage (Chapter 6).The simulation studies demonstrated that the usually observed negligible damaging effect of BtMV is due to the late occurrence of the spread of the disease under field conditions (Chapters 3 and 4). When infection takes place, the crop has already a large enough area of healthy leaves to sustain the yield, even if all plants in the field were infected after the middle of the growing season (Chapter 6). This study is probably the first which simulates crop damage caused by a potyvirus, and it is certainly the first simulation study of the damage caused by BtMV in sugar beet.<strong>Concluding remarks:</strong>As a model is a simplification of reality, perfection is not expected (Ruesink and Irvin, 1986). Several improvements can be made to the simulation model presented in this study to describe the disease incidence. The present version allowed to test the raised hypothesis that spread was a function of the migrating aphid population, for every date at which the inoculum source was introduced. The results of the analysis of the rate of spread of the field experiments, together with the modeling studies, could indicate the model has to be improved by including an aphid population sub-model that describes vector behavior. It suggested also that the spatial characteristics of spread must be taken in account.