Experimental design and field management.

A study using a randomized complete block design with three blocks, each representing a replication, was conducted at the Wiregrass Research and Extension Center, Auburn University, Headland, AL, USA (31°21′26″N, 85°19′24″W), to evaluate the effects of different wild tomato accessions on B. tabaci populations. The treatments included seven wild tomato accessions (detailed in Supplemental Table 1) and five tomato cultivars as controls. Five wild accessions—G29258, PI127826, PI134418, PI209978, and PI126449—from S. habrochaites were sourced from the US Department of Agriculture-Agricultural Research Service Germplasm Resources Information Network-Global. The wild accessions LA1401 (S. galapagense), LA1932 (S. chilense), and the cultivar LA3475 (S. lycopersicum ‘M-82’) were obtained from the C.M. Rick Tomato Genetics Resource Center, University of California, Davis, CA, USA. Four S. lycopersicum cultivars were also included: the beefsteak tomatoes Patsy (Bejo Seeds, Oceano, CA, USA) and Mountain Man (Syngenta US, Greensboro, NC, USA), and the cherry type of tomatoes Cherry Bomb and Apple Yellow (Johnny Seeds, Waterville, ME, USA). Each experimental unit was a single row with five plants, with data collected from the central three plants to reduce edge effects.

Tomato seeds for wild accessions and cultivars were sanitized in a 10% sodium hypochlorite solution for 5 minutes before sowing. On 21 Jul 2023, seeds were planted into 36 mm peat pellets (Jiffy Group, Lorain, OH, USA) and placed in 28 °C growth chambers for germination. Seedlings were transplanted into the field on 23 Aug 2023. The field, with sandy soil, featured 15-cm-tall raised beds spaced 1.8 m apart, with 30 cm between plants. Uniform cultural practices, including irrigation, fertilization, pest, disease, and weed management, were applied according to the Southeastern US 2022 Vegetable Crop Handbook (Kemble et al. 2022).

Data collection.

Whitefly populations were monitored weekly, starting 30 d after transplanting (DAT), when the first adult whiteflies were observed. Counts were conducted at 30, 37, 44, 51, and 58 DAT by visually inspecting the underside of two leaves from the lower third of three different plants per plot, sampled before 0900 HR to minimize the influence of temperature, sunlight, or insect movement. This sampling strategy resulted in a total of 90 leaves per genotype (n = 2 leaves × 3 plants × 3 plots × 5 DATs). In addition, counts of whitefly nymphs and eggs were conducted in the laboratory at 44, 51, and 58 DAT to evaluate the effect of the treatment on immature stages. On each date, one leaf per plant was collected from three plants per plot across three replicate plots, resulting in 27 leaves per genotype (n = 1 leaf × 3 plants × 3 plots × 3 DATs). Collected leaves were placed in a zip sealed bag and examined under a Leica^® M165 C Stereo Microscope (Leica Microsystems, Wetzlar, Germany) with a ×5 to ×20 magnification. A 1-cm² area from each leaflet’s terminal part was randomly selected, and the numbers of whitefly nymphs and eggs on the abaxial surface of that area were recorded for statistical analysis.

Environmental data (rainfall and daily air temperature) were recorded using an on-site Vantage^® Pro2 Plus (Davis Instruments, Hayward, CA, USA) weather station to contextualize fluctuations in whitefly populations and assess the impact of environmental factors on plant performance and whitefly infestations across treatments (Supplemental Table 2).

Tomato leaf trichome characterization.

At 62 DAT, one fully developed young leaf from the upper third of each plant at the pre-flowering stage was collected from three plants per plot across three replicate plots, resulting in nine leaves per genotype (n = 1 leaf × 3 plants × 3 plots × 1 DAT) for trichome identification and quantification. The leaflets were placed in zip seal bags and transported to the laboratory, where they were stored at room temperature (25 °C) until scanning electron microscope (SEM) analysis was conducted the following day. For SEM preparation, three small paradermal fragments (10 mm² each) from the upper, middle, and lower abaxial surfaces of the leaflets were mounted on aluminum stubs with carbon tape and coated with gold using a sputter-coating machine (EMS Q150R SCD, Quorum Technologies, Calgary, AB, Canada). Photomicrographs were captured with an SEM (EVO 50; Carl Zeiss Vision Inc., Hebron, KY, USA) at ×500 magnification and 20 kV voltage. Trichomes on the abaxial leaflet surface were counted and classified as glandular or nonglandular according to established methodology (Channarayappa et al. 1992; Toscano et al. 2001).

Reagent, solvent, and terpene extraction.

Certified terpene mixture (2500 μg·mL⁻¹ in isopropanol of eucalyptol, (−)-α-bisabolol, camphene, δ-3-carene, β-caryophyllene, geraniol, (−)-guaiol, α-humulene, p-cymene, isopulegol, D-limonene, linalool, β-myrcene, nerolidol, β-ocimene, α-pinene, (−)-β-pinene, α-terpinene, γ-terpinene, terpinolene, and (−)-caryophyllene oxide) were purchased from Restek Corp (Bellefonte, PA, USA). α-zingiberene (≥95%) was obtained from Alfa-chemistry (Ronkonkoma, NY, USA).

A stock standard solution (125 μg·mL⁻¹) of terpenes was prepared by appropriately diluting the original standard in hexane and stored at −18 °C. Working standard solutions at different concentration levels were prepared by appropriately diluting the stock solutions in n-hexane (VWR International LLC, Radnor, PA, USA) and stored at −18 °C.

Terpenes were extracted from intact tomato leaflet trichomes using a standardized leaf dip method (Pizzo et al. 2024). For each genotype, leaves were collected from plants grown in separate replicate plots. From each plant, 0.1 to 0.5 g of tomato leaflets were placed in centrifuge tubes with 5.0 mL of n-hexane. The mixture was shaken for 2 min at 25 ± 2 °C. After extraction, the solution was centrifuged at 9000 g_n for 2 min, and the supernatant was transferred into gas chromatography–mass spectrometry (GC-MS) glass vials.

GC-MS analysis.

GC-MS analysis was performed using an Agilent 5977B GC/MSD (Agilent Technologies, Santa Clara, CA, USA) following established methodology (Pizzo et al. 2024). Terpenes were separated on an HP-5MS column (30-m × 250-μm-diameter capillary, 0.25-μm film thickness). A 1.0 μL injection was made in splitless mode at 300 °C, with a solvent delay of 4 min. The temperature program for the GC oven was as follows: 50 °C for 1 min, then increased to 300 °C at a rate of 7 °C/min, followed by an increase to 320 °C at 20 °C/min, where it was held for 2 min. Helium was used as a carrier gas at a constant flow rate of 1 mL·min⁻¹, and terpenes were ionized using electron ionization at 70 eV. The transfer line and ion source temperatures were maintained at 250 and 230 °C, respectively. MS data were acquired in scan mode over 50 to 550 m/z. All samples and standards were injected into triplicate. To prevent contamination, a hexane blank was included after every three sample runs.

Terpenes were identified through retention time (RT) matching and mass spectral library searches, including the National Institute of Standards and Technology (NIST) MS spectral database (version 2.4, 2020). Authentic standards were used to confirm the identity of specific compounds when available. In cases in which standards were unavailable, identification was based on achieving an NIST library matching score greater than 80%. Compounds were categorized as “similar to” the matched compound scores between 80% and 90% and definitively identified for scores between 91% and 100%. Compounds with matching scores below 79% were classified as unidentified and excluded from the reported results. Identification was performed using MassHunter Qualitative Analysis software (version 10.0).

Quantitative analysis was performed using the external calibration curves created with a terpene standard mix. Calibration curves were constructed at seven concentrations ranging from 0.2 to 10 μg·mL⁻¹. The method's linearity, limits of detection (LOD), and limits of quantification (LOQ) were evaluated. LOD values were calculated as 3.3 times the standard deviation (SD) divided by the slope (b) of the calibration curve and LOQ was determined as 10 × SD/b. Quantification was conducted with Agilent MassHunter Quantitative Analysis software (version 10.2). A qualifier ion was employed to identify each analyte, and a quantifier ion was used to determine the analyte response via peak calculation from the extracted ion chromatogram. Specific standards were used to semiquantify the leaf extract components for components not included in the terpene standard mix. D-limonene, 3-carene, and α-humulene from the terpene standard mix were used as external standards for monocyclic monoterpenes, bicyclic monoterpenes, and monocyclic sesquiterpenes, respectively. The terpene content was expressed as μg·g⁻¹ fresh weight.

Statistical analysis.

Repeated measures analysis was used to evaluate temporal changes in whitefly adults, nymphs, and eggs, modeling the variance-covariance structure with PROC-GLIMMIX in SAS (version 9.4, 2024) using restricted maximum likelihood. This approach accounted for correlations from repeated sampling, employing a heterogeneous compound symmetry structure for the variance-covariance matrix. The significance of correlations was assessed at P ≤ 0.05 with Pearson’s coefficient. Accessions were considered fixed effects and blocks were included as random effects. Orthogonal contrasts facilitated comparisons between wild accessions and cultivars. Nymph, egg, and trichome counts were analyzed under a Negative Binomial distribution with a logit link function using the Laplace approximation. Tukey’s test was applied for post hoc mean comparisons with a significance threshold of P ≤ 0.05.

Multivariate analysis included whitefly nymph counts, glandular and nonglandular trichome densities, and terpene content detected via GC-MS to identify associations between wild tomato accessions and tomato cultivars and test for correlation among measurements. Pearson’s correlation matrix (Supplemental Table 3) significant correlations (P ≤ 0.05 or P ≤ 0.01) were submitted to correlation-based network analysis (CNA) to investigate and graphically represent the correlation among variables analyzed (Epskamp et al. 2012; Wei and Simko 2021). A heatmap displayed trait patterns, while Euclidean distance and unweighted pair group method with arithmetic hierarchical clustering produced dendrograms grouping accessions by variable similarity (Gu et al. 2016). Principal component analysis (PCA) was further used to evaluate relationships among whitefly nymphs, trichome densities, and terpene profiles based on their contributions to principal components (Kassambara 2017; Lê et al. 2008).

Whitefly population and trichome density (nonglandular vs. glandular).

There was no statistically significant interaction between repeated assessments of whitefly populations over time and tomato plant treatments (P > 0.05). However, a significant main effect of wild tomato accessions and tomato cultivars on whitefly populations was observed in the density of nymphs per cm² (P ≤ 0.05) (Table 1). Although no significant differences were found for adults and egg densities, there were significantly fewer nymphs on the wild tomato accessions, including S. habrochaites (G29258, PI126449, PI127826, PI134418, PI209978), S. galapagense (LA1401), and S. chilense (LA1932) as compared with the cultivars of S. lycopersicum (LA3475, Cherry Bomb, Apple Yellow, Patsy, and Mountain Man) (Table 1). On average, the cultivars harbored 1.8 adults per leaf, 5.6 nymphs/cm², and 1.8 eggs/cm², whereas wild accessions harbored 1.4 adults per leaf, 0.3 nymphs/cm², and 0.7 eggs/cm² (Table 1 and Fig. 1A). Among cultivars, Apple Yellow and Cherry Bomb were the most susceptible, with infestation rates of eight and seven nymphs per cm², respectively.

Table 1. Mean numbers of Bemisia tabaci adults, nymphs, and eggs, and the distribution of nonglandular and glandular trichome densities from wild tomato accessions of Solanum habrochaites (G29258, PI126449, PI127826, PI134418, and PI209978), Solanum galapagense (LA1401), and Solanum chilense (LA1932), and cultivars of Solanum lycopersicum (LA3475, Cherry Bomb, Apple Yellow, Patsy, and Mountain Man).

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Significant differences were also observed in trichome numbers and types across accessions and cultivars (Table 1). S. lycopersicum cultivars exhibited a higher density of nonglandular trichomes, averaging 55 nonglandular trichomes per μm², compared with an average of 20 nonglandular trichomes per μm² in wild accessions. Among the cultivars, Apple Yellow had the highest density, with 73 nonglandular trichomes per μm². Most wild accessions had lower densities of nonglandular trichomes, with LA1401, PI127826, and PI209978 completely lacking them. In contrast, glandular trichomes were present in wild accessions, averaging 15 glandular trichomes per μm², but were generally absent in S. lycopersicum cultivars. The wild accessions LA1401 (27 glandular trichomes per μm²) and PI209978 (26 glandular trichomes per μm²) had the highest densities of glandular trichomes. All S. lycopersicum cultivars (LA3475, Cherry Bomb, Apple Yellow, Patsy, and Mountain Man) had no glandular trichomes. Scanning electron micrographs of trichomes across different genotypes are shown in Fig. 1B.

Terpene profile and content in tomato plants.

The external calibration curves for the 22 terpene standards demonstrated high linearity within the concentration range of 0.2 to 10 μg·mL⁻¹, with r² values ranging from 0.9936 to 0.9994 (Table 2). This strong correlation between analyte concentration and detector response indicates that the method can accurately quantify terpenes across a broad range of concentrations. The LOD for the various terpenes ranged from 0.02 to 0.30 μg·mL⁻¹ and the LOQ varied between 0.2 and 0.9 μg·mL⁻¹ (Table 2). These values demonstrate the method's sensitivity for detecting and quantifying terpenes in extracts from tomato leaflet trichomes. A hexane blank was randomly included in each run to assess potential carryover. No detectable analyte peaks were observed in the blank runs, indicating minimal carryover.

Table 2. Linear range, analytical curve equation, coefficient of determination (r²), limit of detection (LOD), and limit of quantification (LOQ) of terpenes using gas chromatography–mass spectrometry analysis.

View Fullscreen

The field-grown wild tomato accessions and tomato cultivars were subjected to leaf dip extractions and were analyzed by GC-MS. The analysis identified 24 terpenes based on their RT and mass spectral patterns. Table 3 presents the identified compounds, including their RT, qualitative and quantitative ions (m/z), molecular formula, terpene content (μg·g⁻¹), and classification. The classifications include acyclic monoterpenes, bicyclic monoterpenes, monocyclic monoterpenes, acyclic sesquiterpenes, monocyclic sesquiterpenes, and bicyclic sesquiterpenes. The variability of terpenes among wild accessions and cultivars highlighted substantial differences. Among the terpenes identified, α-pinene (<LOD – 15.99 μg·g⁻¹), 4-carene (0.29–23.04 μg·g⁻¹), α-phellandrene (<LOD – 3.46 μg·g⁻¹), β-phellandrene (<LOD – 10.11 μg·g⁻¹), β-caryophyllene (<LOQ – 23.51 μg·g⁻¹), δ-elemene (<LOQ – 2.41), humulene (<LOQ – 7.81 μg·g⁻¹), and α-zingiberene (1.27–27.77 μg·g⁻¹) are common. However, their concentrations varied significantly across tomato accessions and cultivars.

Table 3. Terpene content in wild tomato accessions (Solanum habrochaites: G29258, PI126449, PI127826, PI134418, and PI209978; Solanum galapagense: LA1401; Solanum chilense: LA1932) and tomato cultivars (Solanum lycopersicum: LA3475, Cherry Bomb, Apple Yellow, Patsy, and Mountain Man) extracted from intact tomato leaflet trichomes.

View Fullscreen

PI209978 and PI127826 exhibited notably high concentrations of zingiberene-related sesquiterpenes, including 9-hydroxy-10,11-epoxy-zingiberene (28.93 μg·g⁻¹ and 46.82 μg·g⁻¹, respectively) and 9-hydroxy-zingiberene (30.19 μg·g⁻¹ and 24.06 μg·g⁻¹, respectively), which were absent in other tomato plants. Similarly, α-zingiberene was abundant in PI209978 and PI127826, with 27.77 μg·g⁻¹ and 27.07 μg·g⁻¹, respectively. In contrast, it was present at lower levels in other wild tomato accessions, including G29258 (2.96 μg·g⁻¹), LA1401 (2.89 μg·g⁻¹), LA1932 (1.44 μg·g⁻¹), and in the cultivar Mountain Man (1.27 μg·g⁻¹).

Multivariate analysis.

Pearson’s correlation matrix (Supplemental Table 3) highlighted key relationships among tomato plant traits, particularly the interplay between terpene content, trichome types, and pest resistance indicators, such as nymph density per cm². The CNA established significant correlations among terpenes associated with resistance traits (Fig. 2A). In this network, positively correlated compounds exhibited strong associations, particularly among 9-hydroxy-10,11-epoxy-zingiberene, 9-hydroxy-zingiberene, and α-zingiberene, as well as terpinolene, 9-hydroxy-zingiberene, and α-zingiberene. In addition, a significant correlation was observed between the number of nymphs per cm² and the number of nonglandular trichomes per μm². Conversely, negative correlations linked specific terpenes to nonglandular trichomes, highlighting an inverse relationship: tomato plants with higher concentrations of certain terpenes generally had fewer nonglandular trichomes and more glandular trichomes. Also, the number of glandular trichomes negatively correlates with the number of whitefly nymphs and nonglandular trichomes per μm².

View Figure

• Download as Powerpoint
• Download Figure

The cluster analysis and heatmap (Fig. 2B) categorized wild tomato accessions and tomato cultivars based on terpene content, revealing distinct clustering patterns that highlight biochemical diversity. Wild accessions PI209978 and PI127826 formed a unique cluster characterized by higher levels of specific terpenes and a greater density of glandular trichomes, as well as a reduced number of nonglandular trichomes and lower nymph density—traits associated with enhanced pest resistance.

The PCA biplot (Fig. 2C) revealed two primary clusters, with principal components 1 (PC1, 71.2%) and 2 (PC2, 16.9%) explaining 88.1% of the total variance. Cluster 1, which includes PI209978 and PI127826, showed a high concentration of specific terpenes (9-hydroxy-10,11-epoxy-zingiberene, 9-hydroxy-zingiberene, and terpinolene) and a strong association with glandular over nonglandular trichomes. In contrast, cluster 2, which consists of the remaining wild accessions and all cultivars, had lower terpene concentrations, indicating that the wild accessions in cluster 1 have distinct biochemical traits, potentially contributing to their enhanced pest resistance.