Tomato yellow leaf curl virus manipulates Bemisia tabaci, MEAM1 both directly and indirectly through changes in visual and volatile cues

Insect colonies

Wildtype specimens of Bemisia tabaci MEAM1 were collected initially from infested straightneck squash, Cucurbita pepo L. cv. recticollis, in Quincy, Florida, in fall 2018. Whitefly biotype testing was conducted by genetic analysis following the published protocol by McKenzie & Osborne (2017). We then divided the collected whiteflies into two separate colonies: one with TYLCV-infected tomato plants and insects and one without TYLCV. We reared whiteflies on collard, Brassica oleracea, of the Flash variety (Seedway^® 674580NT) for the uninfected colony in a small greenhouse. We placed several sentinel tomato plants in the greenhouse and periodically tested them for the presence of the virus using traditional PCR to ensure that TYLCV was absent in the uninfected colonies. The whiteflies used for the 2-choice assay involving visual targets were reared in laboratory conditions (22.1 ± 0.23 °C, 42.9 ± 1.3% RH) in screen cages on 5–10 collard plants under a 16 h L:D photoperiod.

Hybrid tomatoes, Solanum lycopersicum, of Florida 47 R variety (Seedway^® 818212NT), were utilized in both whitefly colonies. Methods for insect maintenance were similar to those described in Johnston & Martini (2020). Tomato seeds were planted on plastic trays (51 × 25 cm) containing agricultural substrate (Jolly Gardener Proline C/B Growing Mix). Seedlings were then transplanted into 11.5 cm² black nursery pots (one seedling per pot) three weeks after sowing and maintained under greenhouse conditions. After one month, plants to be used for the experiments were moved to white mesh screen cages (40 × 40 × 60 cm) and inoculated them with TYLCV, by introducing to each cage 100 adult B. tabaci from the TYLCV whitefly rearing. After 14 days, PCR tests ran on a subset of plants confirmed that tomato plants were positive for TYLCV, and we isolated thirty of them in a small rearing chamber maintained at a constant temperature of 22 °C (12 h L: D period). After one month, a new batch of uninfected tomato plants was added to the infected batch and labeled them with the initial infection date. This process monthly was repeated to obtain a range of diseased plants, with the older plants showing the most severe TYLCV symptoms. It was assumed that all whiteflies collected on TYLCV-infected plants were also TYLCV-positive, having acquired the virus through feeding on infective tissue since 100% viral acquisition from tomatoes after 12 h of feeding has been demonstrated (Jiu, Zhou & Liu, 2006).

Olfactometer choice assays

Olfactometer assays were conducted with the same methods than in Johnston & Martini (2020). Whitefly choice tests between healthy tomato plant cues and TYLCV-infected tomato plant cues were conducted on three levels: (1) visual cues only, (2) olfactory cues only, and (3) visual + olfactory cues. Forty individual whiteflies were evaluated at each level for both TYLCV-infected and uninfected populations, resulting in a total of 80 recorded choices per level. In addition, whitefly preference to TYLCV-infected tomato plants was tested at three different stages of infection: (1) presymptomatic (pre), or plants inoculated with TYLCV two weeks prior that have yet to show symptoms (Fig. 1A), (2) mid-stage symptomatic (mid), or plants that show visible TYLCV symptoms approximately one month after inoculation (Fig. 1B), and (3) late-stage symptomatic (late), or plants that show severe TYLCV symptoms approximately two or more months after inoculation (Fig. 1C). For the experiments that only used visual cues, a glass Y-tube horizontally was positioned under a fluorescent light source mounted within a white fiberboard box for uniform light diffusion. Then, visual targets were placed (tomato leaves representing different levels of TYLCV disease) underneath each arm of the glass Y-tube and a single whitefly was released at the bottom of the tube (Johnston & Martini, 2020). Each whitefly had 15 min to choose between the two arms, where crawling at least one cm into either arm constituted a choice. The visual targets were switched to opposite arms after every two whitefly choices to ensure no bias between the right and left arm of the olfactometer.

For experiments that only used odor cues, the odor source consisted of a single tomato plant, either TYLCV-infected or uninfected. Tomato plants were positioned within a cylindrical glass jar (16 × 50 cm) sealed with a Teflon lid (Sigma Scientific, Gainesville, FL, USA). Two outlets were placed on the Teflon lid with Teflon tubing connecting one outlet to the airflow source and the other outlet to a single arm of the Y-tube glassware. Glass jars received charcoal-purified and humidified air at 0.1 L/min from a custom-made air delivery system (Sigma Scientific, Gainesville, FL, USA). After allowing air to circulate for at least 2 min through the apparatus, a single whitefly was introduced to the bottom of the Y-tube and allowed to choose the same manner as the visual cues-only experiments. The third set of experiments that evaluated host choice under visual and olfactory cues combined the visual tomato leaf targets underneath the Y-tube glassware with the setup of the olfactory cues only experiments.

Since we reared uninfected whiteflies and TYLCV-infected whiteflies on different hosts (collard and tomato, respectively), we conducted additional control replicates to ensure that host rearing did not bias the final host selection. These control replicates examined the whitefly choice between collard and uninfected tomato volatiles using the Y-tube olfactometer setup previously described. In the control trials, 40 individual choices were made by whiteflies reared on collards and compared to 40 individual choices by whiteflies reared on uninfected tomato plants. Once we completed the control trials and confirmed the null hypothesis (see results), we used collard as the host-rearing plant for all non-viruliferous whiteflies for logistical reasons.

Y-Tube Methyl salicylate choice assays

Methyl salicylate (MeSA) volatiles increased significantly in TYLCV infected tomatoes (see results); therefore we conducted an experiment to test the response of whiteflies to different MeSA concentrations. MeSA (≥ 99% (GC), #M6752; Sigma-Aldrich) concentrations tested were created from a stock solution and subsequently serial diluted with a mineral oil base. We conducted whitefly choice tests between MeSA volatile cues at three concentrations: (1) 1.0 µg/µl MeSA vs. mineral oil, (2) 0.1 µg/µl MeSA vs. mineral oil, and (3) 0.01 µg/µl MeSA vs. mineral oil. We evaluated forty individual whitefly choices at each level using TYLCV-infected and uninfected populations, recording 80 whitefly choices at each level.The setup consisted of a glass Y-tube positioned vertically with a burette stand and clamp under a fluorescent light source mounted within a white fiberboard box for uniform light diffusion. A single cotton roll (Dynarex Corporation, #3250) was infused with 100 µL of one of the MeSA concentrations or the mineral oil control and added to an oven bag (Reynolds Kitchens^® Oven Bags, 482 mm × 596 mm) baked at 60 °C overnight. We positioned two outlets on the oven bag, connecting one outlet to the airflow source and the other to a single arm of the Y-tube glassware with Teflon tubing. Teflon bags received charcoal-purified and humidified air at 0.1 L/min from a custom-made air delivery system. After allowing air to circulate for at least 2 min through the apparatus, a single whitefly was introduced to the bottom of the Y-tube and allowed to choose in the same manner as the visual cues only experiments.

Volatile collection

Data for the profile of TYLCV-infected tomato plant volatile odors were collected as previously described in Strzyzewski, Funderburk & Martini (2023). Specifically, the volatile collection system consisted of the top half of the tomato main stem enclosed within an oven bag (Reynolds, Lake Forest, IL, USA) and tied at the top and bottom with zip ties. The oven bag was baked overnight at 60 °C to reduce contaminants. Incoming air was purified via a charcoal filter and pushed in at the top of the oven bag at a rate of 1.0 L/min. The volatiles were forced to the bottom of the bag by pulling air at 0.5 L/min with a controlled vacuum from the automated volatile collection system through a volatile collection trap (7.5 cm long) with 30 mg of HayeSep Q adsorbent (Volatile Assay Systems, Rensselaer, NY, USA) connected to the bottom of the bag with a PTFE fitting. Volatiles were collected for 24 h, andvolatile samples were extracted in vials with 150 µl of dichloromethane. Beforehand, 1 µg of nonyl acetate as an internal standard was added to the samples. 1 µL of each sample was injected into GC-MS (Thermo Scientific ISQ) using an autosampler. Helium was used for the carrier gas at a two mL/min linear flow velocity. All samples were analyzed on a fused silica TG-5MS column (5% phenyl methyl polysiloxane) (Thermo Fisher Scientific, Waltham, MA, USA) 30 min × 0.25 mm ID. The column oven temperature was maintained at 40 °C for 1 min and increased at a rate of 7 °C/min to a final temperature of 300 °C and maintained at 300 °C for 6 min. The injector temperature was set at 270 °C, with the detector set at 200 °C. Quantitation was assigned by comparing peak areas of known amounts of internal standard, nonyl acetate, with the area under the peak of compounds extracted from the infected plants. Compounds were first tentatively identified by comparison of mass spectra with available mass spectra libraries (NIST 11, Wiley 9, mainlib), then confirmed for those available at a reasonable price by injection of analytical standards. Every compound in Table 1 was confirmed by GC-MS analyses with their corresponding synthetic compounds purchased from Sigma-Aldrich. There were three to five replicates per TYLCV-infection stage.

Two choice visual assay

The study assessed viruliferous and nonviruliferous whiteflies’ visual preferences when exposed to filtered light visual targets. Based on the results from the previous experiments, we tested whiteflies’ visual preference against yellow and green visual cues using a two-choice visual target test. The setup included an enclosed space with a base chamber and adhesive translucent panels at the top for capturing the whiteflies, after Paris et al. (2017). This enclosed space was a black cylindrical tube, 12 cm in diameter and 30 cm high, with a small chamber (4 cm by 6.5 cm) attached to its lower center. We coated the inside of the tube and chamber in matte black paint. We attached a shutter mechanism beneath the tube and inserted the chamber into this shutter. The chamber’s base was covered with a black plastic lid, measuring 4.5 cm in diameter and 3 cm in height.

A 400-watt metal halide lamp, positioned 55 cm above the enclosure, provided illumination. This lamp delivered around 2000 lux and emitted both UV and visible light. We covered the chamber’s top with an inverted, transparent Petri dish (150 mm in diameter and 15 mm high) and lined its interior base with a thin adhesive coating (Tangletrap1, Grand Rapid, MI, USA). Visual stimuli consisted of dyed polyester filters attached to the top of the inverted Petri dish, each covering half of the surface (either right or left side). To prevent ambiguity in determining the whiteflies’ color choice at the meeting point of the two colored filters in the middle of the Petri dish, we placed a strip of black electric tape (1.905 cm wide) across the dish’s diameter. We counted whiteflies that landed on the black electric tape portion as non-responders. We randomly alternated the placement of these filters (left or right side of the Petri dish) between experiments.

Plants infected with TYLCV showed altered reflectance spectra, primarily in the visible spectrum. Therefore, we chose the selected filters for their capacity to block UV while transmitting visible light. These included Roscolux1 yellow (#4530), Moss green (#89), UV block (#3114) with 94% transmission, and a neutral density filter (#3415) with 0.15 optical density (Rosco, Stamford, CT), cut to fit half of the Petri dish. To further impede UV light transmission through the colored filters, we layered UV-blocking filters atop the colored ones. Although these UV-blocking filters were primarily transparent to visible light, they significantly reduced UV transmission to less than 10% of its initial intensity. Figure 2 depicts the percent reflectance of each filter.

For the experiments, we collected whiteflies into plastic vials (3 cm wide ×6 cm tall) with lids and fitted gaskets around the vials to ensure a snug fit in the assay arena’s holding chamber. Before starting an assay, we closed the manual shutter, tapped the vial to move the whiteflies to the bottom, removed the cap, and then quickly placed the vial into the holding chamber. A black coverslip was positioned on the chamber’s bottom, keeping the whiteflies in total darkness.

We dark-adapted the whiteflies for 10 min inside the holding chamber before each test. Then, we placed the filters on top of the test arena, allowed the whiteflies to adapt to the dark, opened the shutter, and allowed the whiteflies an hour to respond to the visual target. After the test, we considered the whiteflies stuck to the sticky layer as responders and determined their sex. We designated whiteflies that could not be identified to sex as unknown sex.

Whiteflies in the chamber or landing on the black line of the Petri dish’s sticky layer were labeled non-responders. We used CO2 to immobilize these non-responders for sex determination. We carried out the experiments between 08:00 and 18:00 h, with each test involving 20 to 30 whiteflies. We calculated the total response of whiteflies to either visual target by counting the whiteflies that went to either side of the visual target. In contrast, we counted non-responders as those that did not land on either visual target and were anywhere else in the arena. To ensure statistically representative proportions, we only included tests in the analysis where 40 percent or more of the whiteflies responded to the color choice of visual targets. We calculated the response as a proportion of the total number of responders. We repeated this process 25 times to replicate the assays.

Color spectrometry

We used a spectrometer (UV–VIS Black-Comet; StellarNet Inc, Tampa, FL, USA) to measure the spectral reflectance of tomato leaves at varying stages of TYLCV infection. The spectrometer, equipped with a focus lens, had seven exterior fibers illuminated by the light source and an interior fiber that enabled us to pinpoint the area of spectral reflectance. A Tungsten Halogen (350–2300 nm) and Deuterium (200–400 nm) light source were merged using a fiberoptic cable and equilibrated with filters. We used a small insect pin to create a hole in the leaf to record the locations where we made visual spectral reflectance readings (Fig. 1). The leaves were then photographed on a black background by a Nikon HDSLR (D5300) using an AF-P DX NIKKOR 18–55 mm f/3.5−5.6G VR lens. In Adobe Photoshop, we removed the background surrounding the leaves and replaced the holes made by the insect pin with colored digital circles. If the leaves were healthy control leaves, we took only three measurements at similar colored regions of the leaf. If the leaves were moderately expressing phenotypic characteristics of TYLCV, we took a single measurement at one of the three color regions if green, yellow, or dark yellow were present. The spectrometer produced electromagnetic spectrum values every 0.5 nm (Fig. 1). Researchers categorized the leaves by infection status, including presymptomatic (green throughout), mid-stage symptomatic (green, yellow, and dark yellow spots), and late-stage symptomatic (green, yellow, and dark yellow spots).

Statistical analysis

We conducted all statistical analyses using the statistical software R (RStudio Team, 2020; R Core Team, 2022). We used the following packages to run the analyses: readxl, readr, dplyr, lsr, car, citation, nlme, reshape2, multcomp, and emmeans (Pinheiro & Bates, 2000; Wickham, 2007; Hothorn, Bretz & Westfall, 2008; Navarro, 2015; Fox & Weisberg, 2019; Lenth, 2022; Wickham & Bryan, 2022; Wickham et al., 2022a; Wickham, Hester & Bryan, 2022b; Dietrich & Leoncio, 2023). We conducted a general linearized model (GLM) with binomial distribution for all replicates with plant stage of infection (early, mid, late) and whitefly viral status (positive, negative) treated as categorical factors to determine differences in whitefly choice. We also conducted a chi-square test as a post-hoc analysis for olfactometer choice experiments (α = 0.05).

For the spectrometer measurements, we divided the color region of the electromagnetic spectrum into two categories of interest: between 400 and 750 nm, 496 and 570.5 nm for green, and 571 and 590.5 nm for yellow. Using the Akaike information criterion (AIC), we modeled the percent reflectance of the leaves using the following variables: percent reflectance (p), wavelength in nanometers (λ), regression coefficient (a), leaf with level of exposure to TYLCV (lf), and color region of the electromagnetic spectrum (cr). The percent reflectance of leaves fits a quartic polynomial model: (1)${\displaystyle {p=a}_0{+a}_1{\lambda +a}_2{\lambda }^2{+a}_3{\lambda }^3{+a}_4{\lambda }^4+cr+lf.}$ We compared the percent reflectance of each leaf with different exposure times using ANOVA. We carried out subsequent post-hoc tests using the Holm-Sidak multiple comparison test with a significance level of P = 0.05.

For all GC-MS spectra, the internal standard ratio was calculated by taking the total peak area for each compound (counts ×min) and dividing by the peak area of the IS, nonyl acetate (RT = 19.36). We calculated all peak area percentages and IS ratios by taking the mean value of the total number of replicates for each treatment. For each compound, differences between treatments were tested with a Krikal-Wallis test.

Volatile collection

GC-MS analysis yielded several interesting compounds unique to TYLCV-infected tomatoes compared to uninfected control plants (Table 1). Of particular interest is that MeSA, an ester upregulated by many plants in response to biotic stress, increased significantly in late-stage TYLCV infections, faintly in mid-stage and presymptomatic infections, and was absent in controls (Table 1).

In addition, 3,7,7-trimethyl-1,3,5-cycloheptatriene, a homoterpene that was already found to increase in tomatoes following herbivory damage (Ayelo et al., 2021), increased significantly in all TYLCV-infected tomato stages. The production of terpene volatiles, including α-pinene, β-phellandrene, 4-carene, caryophyllene, α-terpinene, and humulene also increased in TYLCV-infected tomatoes compared to controls, but the difference was not statistically significant.

Color spectrometry

We obtained the spectral reflectance data of 64 tomato leaves and a yellow card, including presymptomatic, mid-stage symptomatic, and late-stage symptomatic leaves from tomato plants at one of three stages of TYLCV infection (Fig. 3, Table 2). We measured the spectral reflectance on each of the 32 leaves in the presymptomatic category. The percent reflectance relative to standard on green areas in control leaves was similar to midstage and presymptomatic tomato leaves for long wavelengths, and only with pre-symptomatic for medium wavelengths (Fig. 3A, Table 2). However the main differences in refelectance among mid and late infection stage were found in yellow and dark yellow areas that differed significantly depending on the infection status (up to 33% more reflectant in late stage infection as compared to mid stage) (Table 2). Late-stage leaves had a higher percent reflectance across all wavelength ranges than the others (Fig. 3C, Table 2). The yellow card used as a control had a much higher reflectance than the mid-stage or late-stage symptomatic at all wavelengths (Fig. 3B).

Y-tube choice assays

In control choice assays where all whiteflies and plants were TYLCV-uninfected, B. tabaci showed a preference for volatiles produced by tomatoes over those produced by collard, regardless of host rearing (Fig. 4). There was no difference in host preference overall between the B. tabaci reared on collard compared to B. tabaci reared on tomato (χ² = 0.065, df = 1, P = 0.799).

Uninfected whiteflies were overall more attracted to TYLCV-infected tomatoes than viruliferous whiteflies independently of the cues (olfactory or visual) used (GLM with binomial distribution, χ² = 6.516, df = 1, P = 0.011). For experiments using only visual cues, we found that the viral status of the whitefly (TYLCV-infected or uninfected) significantly influenced host choice across all stages of infection (χ² = 6.03, df = 1, P = 0.025). Viruliferous whiteflies showed no preference for a healthy versus an infected host plant despite a slightly higher choice count towards healthy tomatoes at any stage of the disease (Fig. 5A). Additional analysis revealed that uninfected whiteflies were visually more attracted to TYLCV-infected tomatoes at the mid-stage of infection as compared to uninfected tomato (χ² = 4.05, df = 1, P = 0.004) (Fig. 5B). TYLCV-infected tomatoes visual cues were more attractive to uninfected whiteflies than to viruliferous whiteflies (Fig. 5C). For experiments using only olfactory cues, whitefly viral status also had a significant influence on all stages of plant infection in the same manner as bioassays with vision only (χ² = 3.75, df = 1, P = 0.021) where viruliferous whiteflies preferred healthy tomato and uninfected whiteflies preferred TYLCV-infected tomato (Fig. 6C). GLM analysis also found late-stage symptomatic tomato plant volatiles to be more attractive to uninfected whiteflies than healthy tomato volatiles (χ² = 3.2, df = 1, P = 0.017) (Fig. 6B). Finally, combined visual and olfactory experiments also yielded similar results where uninfected whitefly prefered TYLCV-infected tomato rather than uninfected conterparts (χ² = 4.15, df = 1, P = 0.027) (Fig. 7B). As seen in olfactory cues only experiments, choice assays that combined both cues saw significant preference of uninfected whiteflies selecting TYLCV-infected tomato at late-stage symptoms (χ² = 4.05, df = 1, P = 0.004) (Fig. 7C).

We exposed uninfected whiteflies to different concentrations of MeSa. The concentrations of methyl salicylate that were not more attractive than the control were 0.1 µg/ µL (χ² = 1.23, df = 1, P = 0.26, Fig. 8B) and 1 µg/ µL (χ² = 0.03, df = 1, P = 0.87, Fig. 8B), while 0.01 µg/ µL was more attractive compared to the control (χ² = 7.81, df = 1, P < 0.01, Fig. 8B). Viruliferous whiteflies did not show preference to different concentrations of methyl salicylate compared to the control for 1 µg/ µL (χ² = 3.46, df = 1, P = 0.06, Fig. 8A), 0.1 µg/ µL (χ² = 3.27, df = 1, P = 0.07, Fig. 8A), or 0.01 µg/ µL (χ² = 1.32, df = 1, P = 0.25, Fig. 8A).

Two choice visual assay

The intensity of the green visual target was 94.39 ± 0.22 µwatts/cm²/sond relative was adjusted to be nearly equal to the yellow visual target, which was 92.53 ± 0.07 µwatts/cm²/sond using neutral density filters. The total response of whiteflies to either filter was not significant (χ² = 16.91, df = 11, P = 0.11). Whitefly response to the visual target was not affected by filter placement randomly on the left or right side (χ² = 0.22, df = 1, P = 0.64) or sex of the whitefly (χ² = 2.39, df = 2, P = 0.30). However, the whiteflies’ response to visual targets significantly differed based on whether researchers reared them on a TYLCV positive or negative host plant (χ² = 6.72, df = 1, P < 0.01). Whiteflies reared on TYLCV-negative collards preferred green visual targets over yellow ones (χ² = 204, df = 144, P < 0.001, Fig. 9A). In contrast, whiteflies reared on TYLCV-positive tomatoes show no preference between yellow or green visual targets (χ² = 288, df = 256, P = 0.08, Fig. 9B).