Volatile organic compounds from entomopathogenic and nematophagous fungi repel banana weevil (Cosmopolites sordidus) under banana field conditions

Volatile organic compounds and pheromones

We have used volatiles from biocontrol fungi (C1, C2, C5 and C7). C1 (styrene) and C2 (benzothiazole) were produced by the isolates Bb203 and Bb1TS11 form the entomopathogenic fungi Beauveria bassiana (Lozano-Soria et al., 2020, 2024). C5 (1-octen-3-ol) was produce by the isolate Mr 4TS04 of the entomopathogenic fungus Metarhizium robertsii and the isolate Pc123 of the nematophagous fungus P. chlamydosporia (Lozano-Soria et al., 2020, 2024). C7 (cycloheptene-1-one) was identified from the entomopathogenic and nematophagous fungi (Lozano-Soria et al., 2020, 2024). All the strains are preserved in the Spanish Type Culture Collection (CECT) (Table S1). These VOCs were formulated by mixing them individually (500 µL) with silica gel (2 g) (60A, 70–200 µ; CARLO ERBA Reagents S.A.S, Cornaredo, Italy). Formulations were placed in miracloth article (Calbiochem Millipore Corp., Burlington, MA, USA) envelopes (3.5 × 2.5 cm) (Fig. S1B). These were sealed and stored at 4 °C in the dark in closed zip bags enclosed in vacuumed bags before use. Cosmolure™ (P160-Lure90; Chemtica International S.A., Costa Rica) or Ecosordidina30 (ECOBERTURA S.A, Canary Islands, Spain), both commercial BW aggregation pheromone formulations with an expected field efficacy of 90 and 30 days, respectively, were also used in experiments (Figs. S1A, S1E).

Banana weevil traps

Chemicals (VOCs and pheromones) were placed in open dispensers in pitfall traps (Scyll Agro^®, Hastingues, France) (Piedra-Buena Díaz et al., 2021). They consist of an empty vessel buried in the ground that contains the active compound (pheromones and/or VOCs) and a lid enclosing the trap, leaving a space through which the weevils fall into the trap (Figs. S1C, S1D).

Field experiments

Seven field experiments were carried out in the Canary Islands between April 2019 and December 2021. Four treatments were used in the traps: (i) pheromone only, (ii) pheromone + VOC, (iii) VOC, and (iv) empty traps as a control to estimate natural captures (no chemicals; Table 1). Traps were placed in commercial banana fields in North Tenerife (Canary Islands, Spain), mostly with high natural BW infestations identified previously (Fig. 1). Traps with treatments were placed randomly within each field. Traps were set ten meters apart in rows. Distance between adjacent rows was 18 m, generating a trap network covering the area of the field. The total number of traps (N) ranged from 18 to 36 traps/field depending on the size of the field (Table 1). BW adults captured in the traps were periodically counted (7–14 days) and removed to avoid reintroduction of insects in the field. VOCs were replaced every 3 weeks to maintain their efficacy (Table 1). Besides, cumulative (summation of BW captured in all time points) captures per trap were calculated by adding all BW scored during the whole experimental period.

Volatile organic compound analysis

Tenax TA fritted glass TD tubes were used to detect VOCs in the field (Cai et al., 2015). Seven tenax TA fritted glass TD tubes (17.8 cm length × 6 mm OD × 4 mm ID; 60–80 mesh; 60 mm/180 mg; Supelco, Bellafonte, PA, USA) were used for capturing VOCs in the field. Prior to use, they were conditioned (GERSTEL TC) and sent to the Canary Islands. They were then placed at the “Los Llanos” (LL) site (Fig. 1) one tenax next to one trap baited with C5 another close to a C5+pheromone baited trap and a further one close to an empty trap. They were left for 24 h at 0, 10 and 25 cm from traps, except for the empty trap where only one tube was placed at the closest distance (Fig. S2). After 24 h capturing VOCs, Tenax fibers were sent to Alicante and placed into the Gases Mases Spectrophotometry Service in the Common Research Facilities at University of Alicante. Them, tenax fibers were desorbed in a Thermal Desorption System (Gerstel TDS-2), from 40 °C (0.5 min) to 250 °C (60 °C/min, 6 min). VOCs were then injected in a Gas Chromatograph coupled to a Mass Spectrometer (Agilent model 6890N gas chromatograph-Agilent 5973 Network Mass Spectrometer; column: DB-624 30 m, 0.25 mm ID 1.4 m, J&W Scientific, Folsom, CA, USA). The chromatography program used had an initial temperature of 35 °C (5 min) and a 3 °C/min increasing curve until 150 °C. Afterwards, a 5 °C/min increasing curve was used until 250 °C. Total analysis time was 64.33 min. The electron impact ionization was 70eV at 230 °C. A single quadrupole was used as an analyzer at 150 °C in the Mass Spectrometer working in SCAN mode (range of 15–350 m/z) as a detector. Wiley275 library was used for identifying VOCs.

Meteorology data

Meteorological data were collected from El Drago, Buenavista del Norte and La Orotava meteorological stations (Agrocabildo, 2015). Daily measurements of mean temperature (°C), mean relative humidity (%) and precipitation (mm) were used for the analysis. Mean meteorological data were plotted weekly.

Banana weevil trap capture interpolation maps

Surface maps were plotted from BW captures over time using Inverse Distance Weighted (IDW) interpolation with QGIS Software (www.qgis.org). QGIS generates an IDW interpolation of a point vector layer (insect trap data). IDW estimates unknown values by averaging the values of sample data points in the vicinity. The closer the sampled point (known value) to the estimated point (unknown value), the more influence it has in the process. It assumes that the variable being estimated decreases in influence with distance from its sampled location. IDW uses spatial autocorrelation and is a quick deterministic interpolator (Li & Heap, 2011). IDW is an exact interpolator, where the maximum and minimum values in the interpolated map can only occur at sample points. IDW interpolations can generate predictions even when pest presence is low (Cohen et al., 2022).

Statistical analysis

One-way ANOVA followed by Tukey’s HSD was used to analyze differences between treatments (fixed factor) in cumulative values of BW captures per trap for the entire experimental period. Cumulative values of BW captures per trap/treatment were calculated by successively adding (summing) the BW captures/trap found in all samples taken in each field experiment. Prior to ANOVA, homogeneity of variances was checked using the Levene test. The normality of the distribution of the residuals in the models was checked using the Shapiro test. If the homogeneity of variance of the data or the normality of the residuals of the ANOVA model were not met, the data were transformed using the square root or logarithms. When transformed data did not meet the parametric assumptions, the Kruskal–Wallis non-parametric test was used.

A negative binomial generalized linear mixed model fit by maximum likelihood (Laplace approximation; GLMM: ‘glmer.nb’ function from the ‘lme4’ package; Bates et al., 2015) was used to estimate fixed and random effects in the response of BW captured in traps. As traps were always the same, only the volatiles (repellents and sordidin) were replaced, a repeated measures model was used on the same trap, with trap considered as a random effect and treatment and time as fixed factors in the model. Post-hoc analyses were also performed to identify differences in response variables between treatments at each monitoring time. Pairwise comparisons with estimated marginal means (emmeans function and package; Lenth, 2022) were performed using Sidak’s HSD test for the GLMM data. Diagnostics of the models were examined using the DhARMa package (Hartig, 2022). To analyze the relationship between BW captures and each meteorological variable, Pearson’s correlation coefficients were calculated for each experiment. The normality distribution of the data was checked with Shapiro test. In case that normality distribution of the data was not met, we performed Spearman’s test. All statistical analyses and graphs were performed in R version 3.6.1 (R Core Team, 2021) and using GraphPad Prism version 6.01 (GraphPad, 2021). The data analysis workflow is shown in Fig. S3.

Fungal volatile organic compounds performance vary in their repellence to banana weevil under field conditions

The most repellent effect was observed for 1,3-dimethoxybenzene (C5) in the “Las Toscas” experiment (Fig. 2). In this experiment (Experiment 4, Table 1), the presence of VOC C5 in traps together with the pheromone Cosmolure™ significantly reduced the cumulative value of BW adults captured per trap (26.8 ± 16.752) compared to traps in which the pheromone was applied alone (14.6 ± 7.106) (one-way ANOVA, F = 4.495, DF = 1.18, p = 0.048; Fig. 2A). In addition, cumulative values of BW adults captured per trap with C5 only in the “Los Llanos” experiment (Experiment 7) were significantly (one-way ANOVA, F-value = 66.02, p-value < 0.0001) lower (7.22 ± 5.23) than those in traps with pheromone only (100.77 ± 37.48) and, also with pheromone and C5 together (97.55 ± 22.75) (Fig. 2B). We detected, under field conditions, C5 VOC in the air close to traps baited with the volatile (Table S2). As expected, C5 amount decreased with distance from traps.

Styrene (C1) and benzothiazole (C2) baited traps (Fig. S4) caused a reduction tendency in BW cumulative captures in Experiments 2 (33.00 ± 22.80 and 36.66 ± 23.53) and 3 (41 ± 32.59 and 34.77 ± 26.38) compared to those found in pheromone baited traps (39.3 ± 25.47 and 44.4 ± 31.75). However, these cumulative captures were not significantly different from those in pheromone only containing traps (Figs. S4B and S4C). 2-cyclohepten-1-one (C7) showed no effect on BW captures with either low or high BW population levels in the field (Experiment 5 and 6) (Figs. S4D, S4E).

Site and season affect volatile organic compounds repellence to banana weevil

Number of BW adults captured per trap weekly or biweekly was highly variable. Experiment 1 where VOCs (C1 and C2) were not replaced indicated a 3-week period of VOCs activity (Figs. S5 and S6). This was reflected in a tendency of increasing capture in traps with VOCs (C1 or C2) only form third weeks onwards. Therefore, VOCs were replaced after 3–4 weeks for the rest of the experiments (Table 1).

VOC C5 in “Las Toscas” Experiment 4 reduces BW captures in traps containing both VOC and BW pheromone (2.43 ± 2.061) compared to traps with BW pheromone only (4.47 ± 4.534) (Fig. 3). No significant differences were found in treatment:time interaction (Fig. 3). However, significant differences were found in factor treatment and factor time. Pheromone traps only significantly attracted more BWs than traps containing both active compounds (GLMM Negative Binomial, Estimate = 1.0046, z-value = 3.005, p-value = 0.00225; Table S3, Diagram S1 and S2) than traps containing both active compounds. The winter period coincided with a reduction in temperature and precipitation (Fig. S7). The variance of the trap was 0.1412.

In the second assay performed with VOC C5 (“Los Llanos”, Experiment 7), empty traps and traps with C5 only were also included. The number of BWs was significantly lower in traps with fungal VOC C5 only than in traps with pheromone only (GLMM Negative Binomial, Estimate = 1.917, z-value = 5.036, p-value < 0.001; Fig. 4, Table S3, Diagram S1 and S2). In this experiment, meteorological events (Fig. S8) conditioned field results. For instance, a mild precipitation in the third week resulted in a significant increase in BW captures (Fig. 4). There was no significant difference in the number of BWs caught across the treatments (GLMM Negative Binomial Estimate 0.71488; z-value = 1.691, p-value = 0.0908). A high precipitation occurred at the end of the experiment. This may have affected trap captures since some of them were filled with rainwater. Capture in the last weeks were consequently lower than in the previous one (Fig. 4). The variance of the trap capture rate is 0.05187.

No significant differences were found in BW trap captures for C1 and C2 in Experiments 2 and 3 for treatment factor nor for treatment:time interaction (Figs. S9–S12). C7 was first evaluated in “Las Toscas”, Experiment 5 (Table 1). BW captures were extremely low (2–3 individual per trap) for the whole experiment and no significant differences were found for treatment, time, and the interaction (Fig. S13). C7 was also tested at “La Orotava”, Experiment 6, during summer months (Figs. S14, S15). In this case only time show significant reduction in BW during summer. No differences were found for treatment factor nor for the treatment:time interaction.

VOC 1,3 dimetoxybenzene modifies banana weevil spatial ecology under field conditions

BW shows an aggregate distribution in the field. Surface model maps are different at the start of the experiments from those at the end (Figs. 5, S16 and S17). Videos showing time lapses of BW captures are included as GIF files (Figs. S18–S20). They reveal areas with high (<45 BW/trap/week) and low (≲5 BW/trap/week) BW density. IDW interpolations indicated the areas of occurrence and movement of the BWs in the field.

In “Las Toscas” (Experiment 4) (Table 1) with fungal VOC C5, there were five initial BW foci (outbreaks) across the field (Fig. 5A). They were associated with pheromone baited traps (3) and C5+Pheromone traps (2). BW field populations decreased with time (Fig. S18, Top). Foci associated with C5+pheromone completely disappear (Fig. S18, Top). At the end of the experiment, two new emerging foci, characterized by middle values of BW captures per trap, appear in pheromone only containing traps (Fig. 5B). “Los Llanos” C5 (Experiment 7) started with three main foci (associated to pheromone baited traps) in the south part of the field plus three further foci in the northern part in this case associated with pheromone+C5 baited traps (Fig. 5C). BW then spread throughout the field (Fig. S18, Bottom). At the end the experiment, BW population decreased, but the initial three foci, associated with pheromone baited trap, now with lower captures, remained (Fig. 5D). The other three initial foci with C5+pheromone completely disappear (Fig. S18, Bottom).

In the C1–C2 “Alcaravanes” (Experiment 1), there were two early BW outbreaks in the northern area of the field (associated with pheromone+C1 baited traps) and a main one in the south-east part close to pheromone baited trap (Fig. S16A). In the subsequent weeks, the south focus increased and moved to the east, to a Pheromone+C1 baited trap (Fig. S19, Top). At the end only two emerging foci in pheromone traps were observed (Fig. S16B). Regarding the C1–C2 (Experiment 2), performed in early summer, BW was less present compared to prior experiments (Fig. S16C). However, foci shifted with time (Figs. S16D and S19, Middle). In the last experiment in “Alcaravanes” (Experiment 3) two big BW outbreaks associated with pheromone+C2 and pheromone baited traps in the south-west area of the field (Fig. S16E) decreased and moved with time (Fig. S19, Bottom). At the end of the experiment, in the middle of the winter, BW was almost absent from traps (Fig. S16F).

In “Las Toscas” C7 (Experiment 5) (mesh-covered field) BW was scarcely present (Figs. S17A, S17B). The geostatistical model did not perform well with such low BW captures (Fig. S20, Top). In C7 “Los Llanos” (Experiment 6) there was a large initial BW outbreak in the middle of the field close to a fungal VOC C7+pheromone and pheromone baited traps (Fig. S17C). In the next weeks, the big outbreak split into two little foci areas at the south-east part of the field (Fig. S20, Bottom) finally associated with pheromone traps (Fig. S17D).