Evaluation of semiochemical based push-pull strategy for population suppression of ambrosia beetle vectors of laurel wilt disease in avocado

Ambrosia beetles (Coleoptera: Curculionidae: Scolytinae and Platypodinae) bore into tree xylem to complete their life cycle, feeding on symbiotic fungi. Ambrosia beetles are a threat to avocado where they have been found to vector a symbiotic fungus, Raffaelea lauricola, the causal agent of the laurel wilt disease. We assessed the repellency of methyl salicylate and verbenone to two putative laurel wilt vectors in avocado, Xyleborus volvulus (Fabricius) and Xyleborus bispinatus (Eichhoff), under laboratory conditions. Then, we tested the same two chemicals released from SPLAT flowable matrix with and without low-dose ethanol dispensers for manipulation of ambrosia beetle populations occurring in commercial avocado. The potential active space of repellents was assessed by quantifying beetle catch on traps placed ‘close’ (~5–10 cm) and ‘far’ (~1–1.5 m) away from repellent dispensers. Ambrosia beetles collected on traps associated with all in-field treatments were identified to species to assess beetle diversity and community variation. Xyleborus volvulus was not repelled by methyl salicylate (MeSA) or verbenone in laboratory assays, while X. bispinatus was repelled by MeSA but not verbenone. Ambrosia beetle trap catches were reduced in the field more when plots were treated with verbenone dispensers (SPLAT) co-deployed with low-dose ethanol dispensers than when treated with verbenone alone. Beetle diversity was highest on traps deployed with low-dose ethanol lures. The repellent treatments and ethanol lures significantly altered the species composition of beetles captured in experiment plots. Our results indicate that verbenone co-deployed with ethanol lures holds potential for manipulating ambrosia beetle vectors via push-pull management in avocado. This tactic could discourage immigration and/or population establishment of ambrosia beetles in commercial avocado and function as an additional tool for management programs of laurel wilt.

of an olfactometer described below creating two treatment arms and two control arms. Before placing the wicks in the arms, the cotton wicks were left for 15 min in a fume hood for solvent evaporation.
A four-choice olfactometer was used to test the repellency of the two beetle species to verbenone and MeSA in the laboratory. Dichloromethane was used as the carrier solvent for the treatments and used as the control (100 μL). The four-choice olfactometer used in this study was previously described in Martini et al. 35 . Briefly, the four choice olfactometer has four arms that depart from a 30 × 30 cm polytetrafluoroethylene (PTFE) square arena. Four independent odor fields were created in the chamber by a constant airflow of 0.35 l/min pushed through each arm of the olfactometer and by pulling air (0.50 l/min) (as shown in Martini et al. 2015) through the floor's central air evacuation hole. The olfactometer floor and arms were covered with filter paper (25 cm diameter laboratory filter paper, Curtin Matheson Scientific, Houston, TX) to assist beetle movement. The filter paper was changed between each bioassay and the olfactometer was washed with acetone and Sparkleen detergent (Fisherbrand, Pittsburgh, PA) to remove traces of the treatment chemicals. Each arm of the olfactometer was connected to a 350 ml glass vial the collected the beetles as they moved toward the treatments and then to a custom-made air delivery system (ARS, Gainesville, FL). The ambient air supply was purified through a charcoal filtration system. A flowmeter was used to measure airflow (Varian, Walnut Creek, CA) to ensure equivalent airflow within each arm. The olfactometer was centered under a 150W high-pressure sodium grow light (Hydrofarm, Petaluma, CA). For each species tested, 25 adult female beetles were released between 16:00 and 17:00 hr in the center of the olfactometer. After 16 hours, the number of beetles in each of the four chambers was counted and recorded. The experiment was run six times for each treatment and species combination. field trials. A 10-acre avocado orchard was selected in Homestead, FL (25.509735, −80.521506) based on known incidence of laurel wilt disease symptoms. Avocado trees (West Indian) were 11 years old and planted on a 3 × 6 m spacing with average canopy height of 4 m. This orchard was adjacent to a second neighboring orchard approximately 10 m away and separated by a dirt roadway. The experimental orchard was managed through fungicide treatments and symptomatic tree removal and had low incidence of disease symptoms during the study while the neighboring orchard was characterized by visibly symptomatic trees and unknown management strategies. The experiment was deployed on the exposed edge of the managed orchard across from the orchard with disease symptoms and unknown management. We chose this area for investigation of chemical repellants and/ or push-pull management based on the assumption that an orchard border facing a large area of marginally protected avocado should encourage high pressure from target beetles. The SPLAT (ISCA Technologies, Riverside, CA, USA) flowable matrix was used in the field as the release device for putative repellant chemicals. SPLAT containing either 10% w/v of verbenone or MeSA was dispensed as eight individual 1 g dollops per tree using a caulk gun (HDX, Heavy Duty Caulk Gun, Home Depot, Cobb County, GA, USA) to all trees per 3 × 3 tree plot. The application dosage was based on field efficacy observed previously with release of both verbenone and MeSA in previous field studies with X. glabratus 27 .
The initial field trial was established 4 May 2018 (Supplementary Figure 1). The objectives of the first experiment were: (1) to evaluate verbenone and MeSA as possible repellents for ambrosia beetle species occurring under field conditions in avocado, and (2) to determine whether active space of repellent dispensers was beyond 1 m in distance. Treatments were applied in to the 3 × 3 plots (0.05 ha) in a randomized complete block design with three treatments per block: verbenone, MeSA, untreated control and five replicates per treatment. Replicate blocks were separated by at least 20 m and treatment plots by at least 15 m. Four white, unbaited sticky traps were placed in each replicate block (Wing Trap Liners IPM-103, Great Lakes IPM, Vestaburg, MI, USA) with two traps on each of two trees, 1.5M above the ground. In each tree, two traps were stapled (top and bottom of the trap) onto a tree 1 m apart and at the same height. One trap was placed ~5-10 cm away from SPLAT dispensers (termed 'close'), while the other was placed ~1-1.5 m away from dispensers (termed 'far'). Traps were collected and replaced in the field every ten days for 30 days; traps were collected 10 days after treatment (DAT), 20 DAT, and 30 DAT which resulted in three sets of traps for the duration of the experiment. All ambrosia beetles caught on the sticky traps were counted, regardless of species.
The second field trial was established 12 June 2018 (Supplementary Figure 2). The objective of this experiment was to test the hypothesis that combining the forces of push (repellent, SPLAT with verbenone or methyl salycilate) and pull (attractant, an ethanol lure) would have a greater effect of inhibiting beetle populations as indirectly measured by capture on unbaited sentinel traps than deploying repellent alone. The same distribution of traps per plot was used; on each of two trees, two traps were placed with one "close" (~5-10 cm) and one "far" (~1.5 m) away from repellent. The attractant ('pull' factor) was ethanol dispensers (Product Number 3344, Synergy Semiochemicals, Burnaby, BC, Canada). The ethanol dispensers mimic stressed and declining trees and are attractive to a broad range of species of ambrosia beetles in avocado (see Results). The trial was deployed in the avocado orchard described above. The experiment was arranged as a split plot design replicated three times with the 'pull' as whole plot factor at two levels: (1) control, (2) ethanol dispensers and 'push' as the split plot factor at three levels: (1) control, (2) MeSA, and (3) verbenone. Like the first trial, SPLAT was applied to each tree in a 9 tree square plot (3 trees × 3 trees) with a buffer row in-between each plot. In the block with the 'pull' ethanol dispensers, the dispensers were put in the buffer row in-between SPLAT treatment plots. The 'pull' whole plot factor was established by deploying two ethanol dispensers per west facing plot edge along with a white sticky trap as described above to monitor beetle populations surrounding the 'pull' treatment. Control plots received unbaited sticky traps deployed without lures at west facing edge. This edge of plot borders faced the unmanaged avocado orchard described above. Beetle densities in each plot were measured with white, unbaited sticky traps placed by 'close' and 'far' away from nearest SPLAT dispenser, as described above. All replicate whole plots were separated by at least 20 m and split plots by at least 15 m. All traps were collected and replaced in the field every ten days. All Scolitydae beetles were identified to species level with the use of the most recent taxonomic key 33  Statistical analyses. Beetle preference measured using the four-choice olfactometer was analyzed using a two-tailed, paired t-test in GraphPad Prism 7.05 (GraphPad Software, La Jolla, CA, USA). Field results were analyzed using a Generalized Linear Model (Proc GLM) to perform a repeated measures analysis of variance in SAS 9.4 (SAS, Cary, NC, USA). For the first field trial, the fixed variables in the model were: days after treatment (DAT), block, treatment, and trap distance. For the push-pull experiment, the fixed variables were: DAT, block, push (repellent), pull (attractant), and trap distance. A priori contrasts were applied to determine which treatment differed when significant difference were found. Analyses of community composition were performed with the statistical software R v 3.3.1 (R core team, Vienna, Austria). Diversity of Scolytineae beetles was examined with a principal component analysis (PCA) conducted with the package FactoMineR 36 . Linear correlation between principal components and beetle density were subsequently conducted.
Field trial: repellents only. In the first field study (repellent only), days after treatment and sentinel trap distance from dispensers impacted number of beetles caught. There was no overall effect of repellent plot treatment on numbers of beetles caught on the associated monitoring traps. However, a trend for treatment effect was observed for reduced captures on sentinel traps deployed close to dispensers (F 1 = 3.37, P = 0.0682). DAT influenced variation in trap captures; the number of beetles captured 20 DAT was significantly higher than captures at 10 or 30 DAT (F 2 = 3.45, P = 0.0340; Fig. 3). DAT also had an effect on number of beetles captured in treatment plots by trap distance. At ten days (Fig. 3A) and thirty days (Fig. 3C), there was no statistical difference between captures on close or far traps in any of the plot treatments. However, when analyzed within the treatment times, at 20 DAT, fewer beetles were captured on traps placed "close" to verbenone dispensers than on control traps (F 1 = 4.75, P = 0.0429; Fig. 3B). There was a similar trend observed at twenty days with reduced catches on traps located "close" to MeSA dispensers than on control traps or traps placed "far" from these dispensers (Fig. 3B). Overall, beetle catch on traps placed far away from either MeSA or verbenone did not differ from the controls.
Field trial: complete push-pull system. In the second field experiment, an ethanol dispenser (attractive pull) sub-factor was included in the design. We observed a significant effect of the interaction between the push (repellents in SPLAT) and the pull (ethanol attractant) treatments on beetles captured per trap deployed near treated trees ( Table 1). The numbers of beetles captured in plots treated with the push and pull were significantly lower than in control plots at 10 (Fig. 4A) and 30 DAT (Fig. 4C). This was true for both push-pull using verbenone of MeSA as the repellent. Interestingly, not a single beetle was captured in close traps placed in plots treated with verbenone in the push-pull blocks throughout the entire experiment. Trap captures on the "close" traps did not www.nature.com/scientificreports www.nature.com/scientificreports/ vary significantly with sampling date while those on the "far" traps varied significantly by sampling date (Table 1). There was also an interaction between repellent and DAT, as the repellents were less effective at 20 than at 10 and 30 DAT (Table 1, Fig. 4). Captures of beetles per trap placed far from the treated trees were not significantly different from captures in control plots for either push (repellent alone) or combined push-pull treatments ( Table 1, Fig. 4).
Ambrosia beetle diversity and community composition. Sixteen species of scolytine beetles were captured during the course of the experiment ( Table 2). The vector of R. lauricola in redbay X. galbratus was not found during either field study. However, five species known to carry and for some of them transmit R. lauricola were found: Xyleborus ferrugineus, Xyleborus gracilis, X. bispinatus, X. volvulus, and Xyleborus crassisculus. Trees with the ethanol pull and untreated trees were both dominated by a single species [X. saxesenii 75% (Fig. 5A) and E. fornicatus 85% (Fig. 5B), respectively]. The community composition of beetles was different on traps placed in the repellent blocks from the composition on the traps placed with ethanol lures; verbenone associated traps only captured three species of beetles while the methyl salicylate and the untreated control captured seven or more species (Fig. 5D). The most species of xyleborine ambrosia beetles were captured on traps associated with the ethanol lure (Fig. 5A). One species was only found on traps in plots treated with 10% w/v methyl salicylate released from SPLAT: Cyclorhipidion bodoanum; but in very low numbers (Table 1, Fig. 5C). Capture of no species increased in the presence of verbenone; E. fornicatus were captured in higher numbers on traps in untreated control traps than on traps deployed in any other treatment plots.
A principal component analysis was performed to determine the main contributions to the differences in beetle diversity associated with various semiochemical treatments. The first (PC1) and second (PC2) principal components accounted for 18.91 and 15.24% of the variance, respectively and the other principal components explained less than 10% of the variance. PC1 indicated positive correlation between multiple species and the ethanol lures [X. saxesenii (t = 10.941, P < 0.001), X. affinis (t = 6.063, P < 0.001), and X. ferrugineus (t = 11.360, P < 0.001)] (Fig. 6A). PC2 was negatively correlated with E. fornicatus (t = 2.767, P = 0.007), and positively correlated with X. crassisculus (t = 7.094, P = <0.001), X. compactus (t = 7.921, P = <0.001), and X. graciliis (t = 5.792, P = <0.001) (Fig. 6A). Overall, there was a clear separation in the species associated with the ethanol lures and the repellents with larger diversity associated with the repellents (Fig. 6B). Among the beetle species captured, X. affinis, and X. saxensenii were most attracted by ethanol lures.

Discussion
Redbay ambrosia beetles, X. glabratus, and laurel wilt have decimated redbay trees in the southeastern US 36 . Laurel wilt has also severely impacted the avocado industry in southern Florida. Redbay ambrosia beetles are not the main vector(s) of R. lauricola, the causal fungus of laurel wilt in commercially grown avocado. Previous investigations indicate that avocado is among the least attractive of the Lauraecae to X. glabratus 28 . Therefore, X. glabratus rarely occurs in avocado orchards 22 , which was further confirmed by the current results; no X. glabratus were captured during both field experiments. The transmission of R. lauricola to avocado is likely the result of attack by a diversity of ambrosia beetles that acquire the pathogenic fungi by horizontal transfer in infected Lauracea (redbay or swamp bay) 7 . In Florida, avocado groves are located nearby the Everglades National Park that has been severely impacted by laurel wilt 37 . It is possible that the Everglades, as well as, small urban forest patches serve as a source of infected ambrosia beetles that migrate subsequently in avocado. Herein, we found 16 species of ambrosia beetles, with some (X. crassisculus, X. bispinosa and X. gracilis) confirmed vectors of R. lauricola 7 . www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, Ploetz et al. 38 reported that most ambrosia beetles harbor viable R. lauricola spores. This suggests that control of laurel wilt in avocado should not focus on a single species, but on the community of ambrosia beetles occurring in the crop when R. lauricola is present or nearby.
Traps placed with ethanol lures captured the highest diversity of xyleborine ambrosia beetles (Fig. 5A). In the field, X. volvulus was only found on traps in plots treated with the low-dose ethanol lures (pull) (Fig. 5A). E. fornicatus and X. saxesnii were captured on traps placed in all treatment plots. Trap captures of X. ferrugineus, X. saxesnii, and X. affinis were higher in plots treated with ethanol lures than in all plots without the pull treatment. Furthermore, the MeSA treatment slightly increased capture of three species compared with captures in control plots indicating that MeSA may also act as attractant for certain ambrosia beetle species 5 . Our results suggest that management of a complex of Scolytineae beetles in avocado with semiochemical and in particular the xyleborine beetles, would likely require use of a pull treatment (ethanol lures) in addition to some general repellent treatment. Variation among host plant preference between ambrosia beetle species may explain the variation observed in the communities of beetle species found in avocado and their contribution to disease incidence.
With the high risk of laurel wilt to avocado orchards in southern Florida, it is essential to develop tools for prevention and management of immigrating ambrosia beetles. It is unclear which specific species contribute most to disease spread in avocado or their proportional contribution as vectors 7,9 . With the threat of lateral Figure 3. Effect of field-applied repellants by days after treatment and trap distance from dispensers on scolytine beetle trap captures without ethanol lures. Mean (±SEM) number of beetles without an asterisk did not differ significantly from its paired mean (α = 0.05). Traps were collected at ten day intervals with panel A representing captures from 0-10 days after application, panel B representing captures from 10-20 days after application, and panel C representing captures from 20-30 days after application. 'Close' indicates monitoring traps were ~5-10 cm from repellent dispenser, while 'far' indicates traps were ~1-1.5 m from repellent. (2020) 10  www.nature.com/scientificreports www.nature.com/scientificreports/ pathogen spread between various species, it is a conservative approach to assume all species are potentially capable of promoting pathogen spread. The push-pull strategy of applying combined repellent and attractant stimuli in order to manipulate the distribution of a pest is an integrated pest management tool that has been established in several systems 39 . The use of repellents has been successful in natural stands of Laureacae against laurel wilt 27 , but in a monoculture, agricultural setting, we hypothesized that the additional pull treatment may be necessary to affect beetle populations based on our first trial where repellents alone in avocado were ineffective.
Under laboratory conditions, we assessed the repellency of verbenone and methyl salicylate on two putative vectors of R. lauricola: X. bispinatus and X. volvulus. Verbenone was selected because it has been used in multiple systems as a general bark beetle repellent and has also been proven highly effective in natural forests against the redbay ambrosia beetle, X. glabratus 27,[40][41][42][43][44] . Methyl salicylate, the other putative repellent used in our studies, is a volatile chemical commonly associated with plant stressors 45 . It may or may not be attractive or repellent to beetle species depending on their life history strategy 5 . MeSA was found to be repellent to X. glabratus in laboratory studies 34 , and confirmed later under field conditions 27 . This is congruent with the life history strategy of X. glabratus, which is known to attack healthy trees 20 . However, its behavioral effect on other Ambrosia beetle species is unclear. It could potentially be an attractant for secondary and tertiary colonizers and a repellent for others.
The laboratory studies did not provide a clear connection to the field studies. In our laboratory studies, methyl salicylate did not appear to be a broad-spectrum repellent of Ambrosia beetles species with low-level repellency to X. bispinatus, but not to X. volvulus. In our field studies, X. bispinatus and X. volvulus were not found in the verbenone treatment but X. volvulus was found in the methyl salicylate treatment which validates the lab data showing it is not a repellent for this species. X. bispinatus was only found in the low-dose ethanol baited traps but in very low numbers. While laboratory olfactometer assays can be useful for preliminary screening, they may not effectively predict what is to occur in the field because of the potential effects of uncontrolled variables such as presence of host and non-odors from surrounding plants, varying levels of disease, and a heterospecific community of beetles. This is particularly the case with verbenone, that did not statistically repel either X. bispinatus or X. volvulus in the laboratory, but was effective in repelling most of the beetles present in commercial avocado, despite the limited radius of activity from dispensers tested here.
Information gathered from the first field study guided the design of second experiment. Beetle captures in the first study without the ethanol lures varied over time following treatment application. This variation seemed to be due to natural population fluctuations of beetles. However, at 20 DAT when the majority of beetles were captured, there was also an effect of treatment. Thus, it was prudent to hypothesize that effectiveness of a repellent treatment may vary in proportion to fluctuations in beetle population density when planning the second experiment. The results of the second experiment incorporating the pull treatment (ethanol lure) appeared to confirm this hypothesis. The synergy between the push and pull treatments was observed as significant reductions in beetle captures relative to control plots at 10 and 30 DAT, but not at 20 DAT, when the density of beetles was lower than during the other sampling dates. Catches of beetles were reduced most effectively relative to control plots when combining the forces of push and pull as a treatment. However, this was only evident on traps placed "close" (~5-10 cm) to verbenone dispensers indicating a small effective repellent radius [sensu effective attraction radius (EAR) measure 46 ] of this treatment. The effective repellent radius of the verbenone-releasing dispensers tested here was less than 1 meter. Thus, there would need to be a high density of these dispensers for practical application. Because of the small effective radius, SPLAT loaded with the 10% w/v rate of verbenone would likely need to be applied to every tree within a monoculture of avocado and in combination with low-dose ethanol lures on field edges. In the present study, the treatments were effective up to 30 DAT, which is consistent with the upper limit of longevity observed previously (Martini, unpublished data).
The field site was selected because it was highly managed with fungicides and had no active disease symptoms, but was in close proximity to a low-management site with active laurel wilt. The flight capability of X. glabratus and another ambrosia beetle species, Monarthrum mali Wood & Bright, have been assessed and indicate that ambrosia beetle may potentially immigrate from distances up to 100 m 47 . An edge is a boundary between two distinct habitats or vegetation types or in this case, two management styles. The effect of edges has been long studied in other contexts such as wildlife conservation 48 . Thus, a follow-up study is warranted to investigate the edge effect to determine the distance immigrating beetles penetrate into avocado orchards in order to best determine where and how semiochemical dispensers and sentinel traps should be placed.
Management is crucial to prevention of the spread of laurel wilt within avocado orchards. Our results suggest that the push-pull strategy may hold promise as a management tool for this pathosystem and deserves further investigation. It is unlikely that it could adequately manage beetle populations as a stand-alone treatment and would likely function best in association with additional preventive measures such as removal of infected trees. Several follow up investigations are warranted. It is possible that long-term effects of beetle population suppression with push-pull could reduce disease incidence over time, if semiochemicals were deployed on an area-wide scale (see Conlong et al.) 49 . This could be evaluated by tracking infections in high-risk Lauraceous hosts occurring nearby commercial avocado 19 in addition to infection sentinels