Is Vector Control Sufficient to Limit Pathogen Spread in Vineyards? Author(s): M. P. Daugherty, S. O'Neill, F. Byrne, and A. Zeilinger Source: Environmental Entomology, 44(3):789-797. Published By: Entomological Society of America URL: http://www.bioone.org/doi/full/10.1093/ee/nvv046
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/ terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.
PEST MANAGEMENT
Is Vector Control Sufficient to Limit Pathogen Spread in Vineyards? M. P. DAUGHERTY,1,2 S. O’NEILL,1 F. BYRNE,1 AND A. ZEILINGER3
Environ. Entomol. 44(3): 789–797 (2015); DOI: 10.1093/ee/nvv046
ABSTRACT Vector control is widely viewed as an integral part of disease management. Yet epidemiological theory suggests that the effectiveness of control programs at limiting pathogen spread depends on a variety of intrinsic and extrinsic aspects of a pathosystem. Moreover, control programs rarely evaluate whether reductions in vector density or activity translate into reduced disease prevalence. In areas of California invaded by the glassy-winged sharpshooter (Homalodisca vitripennis Germar), Pierce’s disease management relies heavily on chemical control of this vector, primarily via systemic conventional insecticides (i.e., imidacloprid). But, data are lacking that attribute reduced vector pressure and pathogen spread to sharpshooter control. We surveyed 34 vineyards over successive years to assess the epidemiological value of within-vineyard chemical control. The results showed that imidacloprid reduced vector pressure without clear nontarget effects or secondary pest outbreaks. Effects on disease prevalence were more nuanced. Treatment history over the preceding 5 yr affected disease prevalence, with significantly more diseased vines in untreated compared with regularly or intermittently treated vineyards. Yet, the change in disease prevalence between years was low, with no significant effects of insecticide treatment or vector abundance. Collectively, the results suggest that within-vineyard applications of imidacloprid can reduce pathogen spread, but with benefits that may take multiple seasons to become apparent. The relatively modest effect of vector control on disease prevalence in this system may be attributable in part to the currently low regional sharpshooter population densities stemming from areawide control, without which the need for within-vineyard vector control would be more pronounced. KEY WORDS ease incidence
Xylella fastidiosa, vector-borne pathogen, transmission efficiency, disease spread, dis-
The control of arthropod vectors is often a central component of managing vector-borne diseases that pose risks to human health, wildlife management, and agricultural production (Walker and Lynch 2007, World Health Organization [WHO] 2011). Control measures, including insecticide use, often are predicated on the expectation that reducing vector densities will slow or halt disease spread. However, the relationship between vector density and disease spread—typically formulated with respect to the epidemiological parameter R0 (i.e., pathogen net reproduction rate)—is often nonlinear and depends on the particular biology of a pathosystem (Perring et al. 1999, Madden et al. 2000, Wonham et al. 2006). For example, models suggest that vector densities may need to be reduced far lower to curb outbreaks of a persistent-propagative pathogen than for a pathogen that is transmitted in a semipersistent manner (Madden et al. 2000). Similarly, vector control is less effective for systems in which primary spread—defined as transmission occurring from primarily outside the 1 Department of Entomology, University of California, Riverside, CA 92521. 2 Corresponding author, e-mail:
[email protected]. 3 Initiative for Global Change Biology, University of California, Berkeley, CA 94720.
focal host population—is more important in the epidemic process than secondary spread—transmission among individuals within a host population (Perring et al. 1999). As a result, chemical control of vectors may offer highly variable reductions in disease risk (e.g., Perring et al. 1999, Erlanger et al. 2008). Here we evaluate whether attempts to control an invasive vector translate into demonstrable reductions in disease prevalence in vineyards. Xylella fastidiosa is a xylem-limited plant-pathogenic bacterium, with likely origins of its most dominant subspecies in Central America and North America (Nunney et al. 2010). It is a generalist pathogen that infects numerous native, weedy, crop, and ornamental plant species (Purcell 1997, Hopkins and Purcell 2002). For susceptible hosts, X. fastidiosa infection leads to progressively more severe leaf scorch, defoliation, or stunting symptoms, and eventually plant death. Among the most susceptible hosts to X. fastidiosa are grapevines (Vitis vinifera L.), in which infection causes Pierces disease (Purcell 1997). Pierce’s disease first was documented in the late 1800s, during a significant outbreak in southern California (Pierce 1892). More than 100 yr later, in the late 1990s, severe Pierce’s disease outbreaks occurred again in the southern part of California, this time associated with the invasion of the
C The Authors 2015. Published by Oxford University Press on behalf of Entomological Society of America. V
All rights reserved. For Permissions, please email:
[email protected]
790
ENVIRONMENTAL ENTOMOLOGY
glassy-winged sharpshooter [Homalodisca vitripennis Germar (Hemiptera: Cicadellidae)] several years prior (Blua et al. 1999). X. fastidiosa is primarily insect borne, most often associated with the activity of the sharpshooter leafhoppers [(Hemiptera: Cicadellidae); Severin 1949]. For these vectors there is no evidence of transovarial transmission of X. fastidiosa (Freitag 1951) or a protracted latent period in the insect (Purcell and Finlay 1979), but once acquired by adults the infection is maintained persistently (Purcell and Finlay 1979). Compared with native vectors, such as the blue-green sharpshooter [Graphocephala atropunctata Signoret (Hemiptera: Cicadellidae)], H. vitripennis is less efficient at transmitting X. fastidiosa to grapevines (Almeida and Purcell 2003, Daugherty and Almeida 2009). Thus, it is likely that H. vitripennis’s ability to drive disease outbreaks is tied largely to its high population densities in select nonvineyard habitats (Blua et al. 1999). These areas of high H. vitripennis population density have been the focus of Pierce’s disease management efforts in California (Sisterson et al. 2008). It is important to note that control of H. vitripennis in California is occurring at multiple scales. At the larger scale, starting in the late 1990s, area-wide control programs were established that employ chemical control of H. vitripennis populations, primarily targeting favored reproductive hosts such as citrus, to constrain spread into nearby vineyards (Toscano et al. 2004, Sisterson et al. 2008). A biological control program also was established, with releases of parasitoids occurring throughout the area where H. vitripennis established (Pilkington et al. 2005). The area-wide programs are credited with successfully reducing sharpshooter population densities in most of the counties where they were implemented, with anecdotal observations that Pierce’s disease declined substantially in these areas (Toscano et al. 2004). At a finer scale, individual grape growers in southern California have adopted aggressive measures to limit H. vitripennis density or activity in their vineyards. Based on research showing that moderate concentrations of systemic insecticides, particularly imidacloprid, in grapevine xylem sap can cause high mortality of sharpshooters for prolonged periods (Byrne and Toscano 2006), vineyard managers frequently apply imidacloprid to vineyards to complement area-wide management of H. vitripennis. The degree to which insecticide applications in vineyards reduce further vector pressure and Pierce’s disease prevalence remains unknown. In other words, it is unclear how important within-vineyard vector control is for disease management, especially considering the existence of the area-wide control programs. In addition to an incomplete knowledge of the effect of within-vineyard sharpshooter control on pathogen spread, the risk posed to nontarget organisms has not been evaluated. Little published data exist on the potential of generalist predators to suppress H. vitripennis populations. The few studies that exist suggest that between 3% and 18% of spiders surveyed in California habitats—including vineyards—tested positive for H. vitripennis eggs, and 8 to 40% of hemipteran
Vol. 44, no. 3
predators tested positive (Fournier et al. 2008, Hagler et al. 2013). Although it is doubtful that these generalist predators strongly regulate H. vitripennis populations, the same taxa can suppress populations of other vineyard pests (Walton et al. 2012). Based on work in other agroecosystems, imidacloprid does not appear to strongly affect many spiders, with the exception of Tetragnatha spp. (Kunkel et al. 1999, Tanaka et al. 2000, Marquini et al. 2002). Other generalist predators, however, are affected; Nabis and Geocoris spp. (Hemiptera) suffer 15 to 30% mortality from exposure to imidacloprid (Boyd and Boethel 1998). Thus, it is plausible yet unknown whether systemic insecticide applications targeting H. vitripennis may inadvertently promote secondary outbreaks of other vineyard pests. Despite extensive attempts to manage Pierce’s disease via control of vector populations within and outside vineyards, to date, evaluations of the importance of control measures are lacking. In the present study, we investigated the epidemiological value of within-vineyard sharpshooter control, especially given recent area-wide control efforts, relative to potential nontarget effects of systemic insecticide use. Specifically, we documented imidacloprid use in vineyards over 2 yr and then compared densities of Homalodisca spp., generalist predatory arthropods, and other pest taxa among treatment categories. In addition, we compared Pierce’s disease prevalence among treatment categories. Materials and Methods Field Site Selection and Characteristics. Prior to initiating the field surveys, we gathered information from grapegrowers and vineyard managers in the Temecula Valley area (Riverside County, CA) to identify vineyard sites of known treatment history over the prior 4–5 yr. Of the 88 distinct vineyard blocks for which we acquired information, 66 (i.e., 75%) were reported to have been treated yearly with imidacloprid, 14 were treated intermittently (i.e., “mixed” treatment history—two or three applications in the last 5 yr), and 8 were not treated with imidacloprid for at least the last 4 yr. This last category included three vineyards that were essentially unmanaged in recent years and five organic blocks, three of which had no insecticide applications for the last 3 yr. Although it was not possible to reconstruct all pesticide applications made at all of the sites, especially those few organic sites, information garnered from grape growers indicated that imidacloprid was far more frequently used in the region than other conventional insecticides—often the only one applied. Throughout, we use the term “treatment history” to describe the one of three treatment patterns (i.e., untreated, mixed, or regularly treated) over the preceding years, and “treatment category” to describe whether a given vineyard had been treated or not in a given year. From our list of 88 properties, we selected 34 in which to work that included the range of imidacloprid treatment histories over the preceding 5 yr; this included selecting at random 12 of the 66 sites believed to be treated yearly, to have a sample size more similar
June 2015
DAUGHERTY ET AL: SHARPSHOOTER CONTROL AND DISEASE PREVALENCE
to those of the intermittent and untreated sites (14 and 8, respectively). Sites were spread throughout the region, ranging in distance apart from !0.25 to >6 km, ranging in size from !0.5 to >10 ha, and included several different red and white wine grape varieties— though each block sampled included a single variety. At each of these sites, we collected detailed measures of imidacloprid concentration, vector abundance, natural enemy abundance, and disease prevalence over two successive field seasons. One site was lost in the second year because the vineyard was replanted. Imidacloprid Assays. To verify recent imidacloprid treatments, we collected leaf samples from vines at each site in the late summer 2011 and again in midsummer 2012. Preliminary analyses indicated there was little variability in imidacloprid concentration among vines in the same row (M.P.D., unpublished data). Therefore, we collected samples from multiple vines in each of 10 different rows spread throughout the vineyard. The collected samples were stored on ice and returned to the lab for processing. Samples from different vines in the same row were pooled for processing. In the laboratory, we quantified imidacloprid residues in pooled leaf samples using methods described by Byrne and Toscano (2007). Briefly, leaf tissue was homogenized in 100% methanol to extract the insecticide. After extraction, the samples were centrifuged to pellet the plant particulate matter. Ten-microliter aliquots from each extract were added to 1.5-ml microcentrifuge tubes, dried completely in a TurboVap LV evaporator (Caliper Life Sciences, Hopkinton, MA) and then resuspended in water containing 0.05% Triton X-100. Imidacloprid concentrations then were quantified without further treatment using a commercially available (QuantiPlate Kit for Imidacloprid; cat. # EP 006, EnviroLogix Inc., Portland, ME) competitive enzyme-linked-immunosorbent-assay (ELISA) specific to imidacloprid. As a reference point for imidacloprid activity against glassy-winged sharpshooter, prior research showed a significant decline in sharpshooter nymph populations at concentrations of imidacloprid between 5 and 10 ppb in citrus xylem sap (Castle et al. 2005), which other research has shown may equate to !1/10th the concentration in corresponding leaf tissue samples (Byrne and Toscano 2007). Thus, 100 ppb of imidacloprid in leaf tissue is expected to approximate conservatively the lower threshold for insecticidal effects in these assays. The replicate imidacloprid estimates per vineyard were used to verify imidacloprid treatments using discriminant function analysis (Everitt 2010). Specifically, mean concentration, median concentration, number of samples with concentrations above 100 ppb, minimum concentration, and maximum concentration of the 10 samples per site were used to determine the likelihood that a site fit into either a “treated” or “untreated” category in a given year. Separate analyses were conducted for 2011 and 2012. Next, permutation multivariate analysis of variance [adonis() in the vegan package (Oksanen et al. 2013)] was used to test for significant differences in the number of samples greater than 100 ppb and mean imidacloprid concentration between
791
the putative treatment categories. All analyses were conducted in R v. 2.15.2 (R Core Team 2014). Arthropod Monitoring. Arthropods were monitored in each of the vineyard sites using two separate methods. First, in 2011 and 2012, we employed yellow panel sticky traps to regularly monitor for sharpshooters and common generalist natural enemies. Second, in the summer of 2012 we conducted tap sampling to monitor the abundance of all arthropods. For the sticky trap monitoring, we deployed between four and eight 14- by 22-cm traps (Seabright Laboratories, Emeryville, CA) per site depending on the size of the vineyard. This was intended to standardize sampling effort among different sized vineyards. Traps were spaced throughout the vineyard and attached on sticks !1 m above the top of the vineyard trellis. Each trap was collected monthly and a new trap was deployed. Collected traps were returned to the lab to identify and count insects under the microscope. H. vitripennis and a native congener, Homalodisca liturata Ball (smoke-tree sharpshooter), were identified to species and recorded separately. We pooled counts of the most common generalist predators, for which different groups were identified to different taxonomic levels, from order (i.e., Aranae spp.) to genus (i.e., Geocoris sp.). The most common predatory taxa were lady beetles (Coccinellidae), spiders, minute pirate bugs (Anthocoridae), and green lacewings (Chrysopidae). Big-eyed bugs (Geocoris), predatory stink bugs (Pentatomidae), assassin bugs (Reduviidae), and snakeflies (Raphidioptera) were collected but were much rarer. In the summer of 2012, we conducted tap sampling at all of the vineyards. For each site, we collected 40 tap samples using 38-cm-diameter sweep nets—from each of five vines spaced approximately equidistant throughout each of six rows. Samples among vines within rows were pooled. The collections were returned to the laboratory to identify and count insects under a dissecting microscope. As with the sticky traps, identification was carried out to different taxonomic scales from order (i.e., spiders, thrips) down to individual species (i.e., H. vitripennis vs. H. liturata). For analysis, arthropod counts were pooled into three different functional groups: herbivores, predators, and “other.” Common herbivore taxa included western grape leafhopper (Erythroneura elegantula) and thrips (Thysanoptera spp.), common predators included spiders and Anthocorids, and common “other” taxa included Psocoptera and fungivorous lady beetles (Psyllabora sp.). For the sticky traps, H. vitripennis, H. liturata, and generalist predator densities were analyzed in separate analyses. The dataset was truncated to the primary growing season, months IV through XI. For each group, we calculated the mean cumulative season-long density in each year (i.e., total count divided by the number of traps deployed over the season). The cumulative density was analyzed with a linear mixed-effects model with fixed effects of year (2011 or 2012) and imidacloprid treatment category in the current year (“treated” or “untreated” based on the discriminant function analysis), and a random effect of vineyard
792
ENVIRONMENTAL ENTOMOLOGY
block to account for the repeated measurements made at each of the sites (Crawley 2007). Cumulative densities were square-root transformed to meet the test assumptions, and model simplification via stepwise deletion was employed to arrive at the minimum adequate model (Crawley 2007). For the tap sampling, the six separate collections per vineyard were pooled to calculate cumulative totals. We analyzed the effect of imidacloprid treatment category on the totals of the three functional groups (i.e., herbivores, predators, others) collectively using a multivariate permutation test. A significant effect was followed up with univariate generalized linear models (GLM) for the total of each functional group separately, with a fixed effect of imidacloprid treatment, a quasipoisson error distribution to cope with overdispersion in the data (Crawley 2007), and with Bonferroni correction for multiple tests (ac ¼ 0.0167). Finally, we conducted a correlation analysis between total predators caught in a vineyard and total herbivores or total other taxa (Crawley 2007). Disease Monitoring. In the fall of 2011 and again in 2012, we conducted successive surveys of Pierce’s disease prevalence. This time of year was chosen because symptoms are most apparent then, facilitating more precise estimates of prevalence. Individual vines were inspected noting, categorically, whether they exhibited any of the common leaf scorch, matchstick petioles, or green islands on canes that are associated with X. fastidiosa infection in grapevines. We did not score vines for severity of disease symptoms. At the eight largest vineyard blocks, we inspected at least one third of the vines (i.e., every third row) for symptoms, and at the remaining 26 sites all vines were inspected. Vines that were dead at the time of inspection, for which the cause was not known, were excluded from the survey counts. Finally, we collected leaf samples from up to 50 apparently diseased vines per vineyard for plate culturing to determine definitively infection by X. fastidiosa (Hill and Purcell 1995). This culturing was used as a correction factor for disease prevalence estimates, to minimize biases from any false positives in visual assessment, by multiplying the proportion of culture-positive samples by the proportion of vines showing visual symptoms. Thus, this corrected estimate should be viewed as a relatively conservative estimate of disease prevalence. Disease surveys were analyzed in two ways. First, we analyzed initial disease prevalence via the effects of prior treatment history the previous 4 to 5 yr (i.e., “treated,” “mixed,” or “untreated” based on interviews) on estimated prevalence in 2011. This consisted of a multivariate permutation test on 2011 visual symptoms and 2011 corrected prevalence, collectively, with a fixed effect of treatment history. We followed-up this main test with univariate GLMs for visual prevalence and corrected prevalence, separately, both of which assumed quasibinomial error to cope with overdispersion and employed correction for multiple comparisons (ac ¼ 0.025). Second, we analyzed effects on the change in disease prevalence between years by subtracting for each vineyard the corrected disease prevalence in 2011
Vol. 44, no. 3
from the corrected prevalence in 2012 (cdiff ¼ c12 # c11). This difference is used as a proxy for disease incidence, with positive values indicating disease spread. Effects on corrected prevalence were tested with a multiway analysis of variance that included a fixed effect of imidacloprid treatment in 2012, covariates of cumulative H. vitripennis density in 2012 and corrected prevalence in 2011 (i.e., prior disease), and interactions between imidacloprid treatment and H. vitripennis density or prior disease. We then conducted model selection using information theory criteria to arrive at the minimum adequate model (Bolker 2008). Results For the 2011 imidacloprid analyses, 32 of 34 sites fit into treated or untreated categories with greater than a 95% posterior probability. The remaining two sites group in the untreated category at more than 74 and 87% posterior probability. Based on this, 19 sites were viewed as treated in 2011 and 16 untreated. Similar results occurred in 2012, with 32 of 33 sites grouping with greater than a 95% posterior probability. The one remaining site fit in the untreated category with a 73% posterior probability, for a total of 15 treated in 2012 and 18 untreated. Based on these groupings, the permutation MANOVAs found significant effects of imidacloprid treatment in both 2011 (F1,32 ¼ 15.693, P < 0.001) and 2012 (F1,31 ¼ 9.340, P < 0.001). Mean ELISA readings were !10-fold higher, and the mean number of samples with concentrations >100 ppb were nearly ninefold higher in treated compared with untreated sites (Fig. 1). For the sticky trap monitoring, a total of >450 H. vitripennis and nearly 200 H. liturata were caught over the 2 yr. The final model included significant effects of year (v2 ¼ 26.380, df ¼ 1, P < 0.0001) and df ¼ 1, imidacloprid treatment (v2 ¼ 15.536, P < 0.0001) on H. vitripennis catch, but the final model for H. liturata included nonsignificant effects of both year (v2 ¼ 0.414, df ¼ 1, P ¼ 0.5199) and treatment (v2 ¼ 0.161, df ¼ 1, P ¼ 0.6885). Glassy-winged sharpshooters were approximately twice as abundant in 2012 compared with 2011 and were up to twice as abundant in untreated compared with treated vineyards (Fig. 2A). Collectively, >30,000 generalist arthropod natural enemies were caught, !66% of which were Coccinellids, 25% were spiders, 6% were Anthocorids, 1% were lacewings, and collectively big-eyed bugs, assassin bugs, and snakeflies made up the remaining. The final model included a significant effect of year (v2 ¼ 59.523, df ¼ 1, P < 0.0001) and nonsignificant effect of imidacloprid treatment on natural enemy abundance (v2 ¼ 0.001, df ¼ 1, P ¼ 0.9743). The cumulative abundance of all predatory taxa was !40% higher in 2011 than in 2012 (Fig. 2B). Individual analyses for some of the common taxa showed the same patterns (M.P.D., unpublished data). Sharpshooters were rare in the tap samples (21 of 925 total herbivores), whereas grape leafhoppers were super abundant in some vineyards (783 of 925 total herbivores). Spiders constituted >80% (481 of 592) of
June 2015
DAUGHERTY ET AL: SHARPSHOOTER CONTROL AND DISEASE PREVALENCE
793
Fig. 1. Differences in estimated imidacloprid concentration ( 6 SE) and number of samples (out of 10; 6 SE) with concentrations above 100 ppb between treated (closed symbols) and untreated (open symbols) vineyards in (A) 2011 and (B) 2012.
Fig. 2. Cumulative density (number per trap per season) of (A) H. vitripennis and (B) all generalist predatory arthropods collected on sticky traps in imidacloprid-treated versus untreated sites and between years.
the predatory taxa collected—an order of magnitude more than other predators. Psocoptera were the most dominant “other” taxa, constituting approximately half (216 of 428) of all other taxa collected. The permutation MANOVA showed a significant difference in the collective total number of predators, herbivores, and other athropod taxa caught between imidacloprid treatments (F1,31 ¼ 9.796, P ¼ 0.005). The three univariate tests on predators (v2 ¼ 6.446, df ¼ 1, P ¼ 0.0017), herbivores (v2 ¼ 8.013, df ¼ 1, P ¼ 0.0046), and other taxa (v2 ¼ 6.546, df ¼ 1, P ¼ 0.0105) were also significant. For all three response variables abundance was greater for the untreated compared with treated vineyards (Fig. 3). For example, !70% (336 of 481) of spiders collected came from untreated sites, and >99% (777 of 783) of grape leafhoppers collected were in untreated vineyards. Notably, there were no obvious secondary
outbreaks in treated sites. Overall, predator count was positively correlated with the abundance of herbivores (r ¼ 0.381, t31 ¼ 2.294, P ¼ 0.0288) and other taxa (r ¼ 0.342, t31 ¼ 2.024, P ¼ 0.0517). Disease prevalence over the 2 yr was low across the vineyard sites, with estimated mean (6SD) prevalance based on visual symptoms of 1.68% (61.75) and corrected prevalence of 0.52% (61.18). Prior imidacloprid treatment history had a marginally significant effect on initial prevalence in the permutation MANOVA (F2,31 ¼ 2.348, P ¼ 0.076). Treatment history had a marginally significant effect on 2011 prevalence of visual symptoms (v2 ¼ 6.446, df ¼ 2, P ¼ 0.0398), and a significant effect on the 2011 corrected prevalence (v2 ¼ 9.092, df ¼ 2, P ¼ 0.0106). Vineyards that were untreated for the prior few years had higher disease prevalence than those that were regularly treated or,
794
ENVIRONMENTAL ENTOMOLOGY
Fig. 3. Cumulative number of arthropods (number per 40 taps per site) in three functional groups (predators, herbivores, and other taxa) collected during tap sampling in imidacloprid-treated and untreated vineyards.
especially, those that were intermittently treated (Fig. 4). For the change in disease prevalence between 2011 and 2012, a full model and 10 reduced models were evaluated using Akaike information criteria (AIC), for which lower values indicate a more ideal combination of parsimony or model fit. Based on AIC rankings, the preferred model was one with only an effect of H. vitripennis density (Table 1; Model 1). Yet three other models can be viewed as approximately equivalent (i.e., DAIC
