Introduction

Insect selection of host plants to settle, feed or oviposit is affected by many factors; among others, previous insects experience (including habituation or sensitization (Heard 1999), plant quality for self or the progeny (Awmack and Leather 2002), the presence of plant viruses (Chen et al. 2017; Mann et al. 2009) or beneficial plant-associated microorganisms (Grunseich et al. 2019). Insects also choose their host plants based on defenses that plants possess or may produce (Rodriguez-Lopez et al. 2020) due to previous herbivore infestation levels, either by conspecifics (Zhang et al. 2014) or by heterospecific herbivores (Saad et al. 2015). Numerous studies have focused on how herbivorous insects use chemical cues to select their host plants (preference). Such cues include volatile organic compounds (VOCs) detected remotely through the olfactory system, as well as non-volatiles detected after direct contact with the plant (Schoonhoven et al. 2005). In the case of the tomato plants (Solanum lycopersicum, Solanaceae) the presence of trichomes and other secondary metabolites (e.g., acyl sugars, alkaloids, methyl ketones) also play a role in host selection by the tomato herbivores (Oliveira et al. 2012; Labory et al. 1999; Sohrabi et al. 2016; Tian et al. 2012).

Volatile organic compounds are secondary metabolites released by plants that mediate important ecological processes, including interactions between plants and their herbivores (Furstenberg-Hagg et al. 2013), their pollinators (Raguso 2004), beneficial natural enemies such as predators and parasitoids (Dicke 2015), and even other plants in the surroundings (Baldwin et al. 2006). VOCs are differentially produced according both to environmental factors that affect plant development (Beck et al. 2014) and to biotic stress affecting the plant (Lucas-Barbosa 2016; Venkatesan 2015). Herbivore damage induces changes in the plant’s emitted volatiles, socalled herbivore-induced plant volatiles (HIPVs), which may modulate plant-herbivore interactions and mediate attraction of herbivore natural enemies (Ayelo et al. 2021c). These HIPVs function as indirect defenses and are modulated mainly by the activation of the jasmonic acid (JA) and salicylic acid (SA) pathways (Glas et al. 2014). Cross-talking between pathways regulates and balances the final outcome of the plant’s induced response (Glas et al. 2014). These pathways are differentially activated depending on the insect feeding habits: while chewer insects activate the JA pathway; phloem feeders activate mainly the SA pathway (Glas et al. 2014; Lin et al. 2019; Pieterse et al. 2012; Thaler et al. 2012). Therefore, plants modulate their induced defensive response to different herbivores because the JA and SA defense pathways usually exhibit negative cross-talk; that is, the upregulation response produced by one of the hormones lowers the response regulated by the other one (Pieterse et al. 2012). Herbivores, in turn, are able to manipulate these plant defenses (Pieterse et al. 2012).

In the case of tomato plants, HIPVs play a role in their interactions with different herbivore guilds. These include the phloem-feeder whiteflies, such as Bemisia tabaci and Trialeurodes vaporariorum (Hemiptera: Aleyrodidae) (Lorenzo et al. 2016; Rodriguez-Lopez et al. 2020), as well as leaf chewers of the order Lepidoptera (Tian et al. 2014), including Tuta absoluta (Lepidoptera: Gelechiidae) (Anastasaki et al. 2018), and Coleoptera (Tian et al. 2012). Tuta absoluta specializes in Solanaceae and particularly in tomato plants as its main host (Silva et al. 2021b). Their larvae are chewing miners that feed in the leaf mesophyll, producing galleries through which they move to other parts of the plant such as the fruits (Gontijo et al. 2013), causing their decay (Bentancourt and Scatoni 1995; Da Silva Galdino et al. 2015). While some evidence on leaf mesophyll feeding has been reported for adults (Baetan et al. 2015), they feed mostly on nectar, without causing significant damage to the plant. When choosing their host plant, T. absoluta females respond preferentially to tomato volatiles over potato volatiles (Caparros Megido et al. 2014) and oviposit at higher rates in tomato than in potato (Caparros Megido et al. 2014; Sridhar et al. 2015) or eggplants (Sridhar et al. 2015). Female ability to discriminate among plant volatiles was also observed when given the option between non-damaged and damaged (by conspecific larvae) tomato plants, preferring to oviposit on undamaged plants (Anastasaki et al. 2018; Bawin et al. 2014; Maneesha et al. 2021). Trialeurodes vaporariorum and B. tabaci are polyphagous and cosmopolitan leaf sucker pests that cause direct and indirect damage to plants both as nymphs and as adults (Rodríguez et al. 2003). Feeding and oviposition preferences differ between both whitefly species: while T. vaporariorum prefers to settle and oviposit on tomato over pepper plants, B. tabaci prefers to settle on pepper but lays more eggs on tomato plants (Lorenzo et al. 2016). Bemisia tabaci is attracted to, prefers to settle and oviposit, and performs better, on conspecific-damaged tomato plants rather than on undamaged plants (Su et al. 2018). Trialeurodes vaporariorum males are more attracted to conspecific-damaged plant volatiles, while females are attracted to volatiles from undamaged plants. These attraction responses and preferences seem to be guided by plant volatiles, since volatiles produced by the insects themselves are not attractive to conspecific females or males (Darshanee et al. 2017).

While several studies have focused on insect-plant interactions involving whiteflies and T. absoluta, only a few have studied these interactions when other herbivores are involved. It is known that previous herbivore attack changes other herbivores’ preferences for settling, feeding, ovipositing, or even their performance (Karban 1989). For instance, B. tabaci prefers to settle and lay eggs on cucumber plants previously infested by Tetranychus cinnabarinus (Acari: Tetranychidae), but not on plants pre-infested by Phenacoccus solenopsis (Hemiptera: Pseudococcidae) (Lin et al. 2019) or by Myzus persicae (Hemiptera: Aphididae) (Tan et al. 2014). In the latter case, B. tabaci preference for undamaged vs. damaged plants was correlated with the emission of HIPVs by tomato (Tan and Liu 2014). The differential preference of B. tabaci towards damaged plants may arise from the different feeding habits of the herbivores, which activate different defensive pathways in the plants (Lin et al. 2019). In the case of Trichoplusia ni (Lepidoptera: Noctuidae) oviposition preferences changed depending on which herbivore has previously damaged the plants. It prefers ovipositing on undamaged soy plants rather than on Spodoptera frugiperda (Lepidotera: Noctuidae) or B. tabaci-damaged plants. However, T. ni prefers laying eggs on soy plants where there are heterospecific eggs (S. frugiperda) rather than on undamaged ones (Coapio et al. 2016). In the case of T. absoluta, mated females were more attracted to volatiles from, and preferred laying eggs on undamaged plants rather than on plants damaged by Liriomyza trifolii (Diptera: Agromyzidae) (Maneesha et al. 2021). Finally, T. absoluta larva feeding on the same leaf as B. tabaci nymphs showed a decreased performance; but the performance of B. tabaci nymphs was not affected by previous feeding of T. absoluta larvae (Mouttet et al. 2013).

We have observed both insects usually co-occurring on the same plants in greenhouse crops (unpublished), so we set up to study this intraguild herbivore-plant system in more detail. Specifically, this study focuses first on the production of volatiles by tomato plants infested by two of the main insect pests in South America: the tomato moth T. absoluta and the whitefly T. vaporariorum. Second, we aimed to evaluate how pre-infestation of T. absoluta or T. vaporariorum affects their oviposition and settlement preferences. Understanding these intraguild interactions between herbivore insects that are relevant as tomato pests may provide valuable inputs for tomato production.

Methods and Materials

Plants and Insects

All plants and insects were reared under the same controlled onditions (25 °C, 16:8 L:D, 17,640 lx, r.H. = 70%). Tomato plants, S. lycopersicum cv. San Pedro (seeds were from Beltrame & Co, https://beltrame.com.uy/), were grown in individual pots (12 cm h x 12 cm diam.) and watered either three times a week or as needed according to the plant water demand. Tuta absoluta were continuously reared in laboratory cages (15 × 15 × 30 cm, 6.7 L) covered by voile and fed with potted fully-grown tomato plants with at least 7 leaves with 7 leaflets each. For the initial settlement of the laboratory colony, T. absoluta adults were collected on tomato plants at organic farms nearby Montevideo, Uruguay, and new field-collected individuals have been added every year. The plants were about 1 month old and were replaced twice a week. The adults were separated from the larvae every 3 days. Trialeurodes vaporariorum were weekly collected as adults from pesticide-free tomato crops grown in greenhouses from organic farms nearby Montevideo (34°28’18.2"S 55°57’40.8"W), then kept in the laboratory on potted tomato plants like those described for T. absoluta.

Damaged Plants

Damage procedures were done under the same controlled laboratory conditions mentioned above. Damage was performed for each insect on fully grown plants confined in cages (55 cm, 166 L) by exposing the plants to the insects for 24 h. The plants were then used either for bioassays or for headspace collection of plant volatiles. We induced plant damage by using T. absoluta larvae, the stage responsible for damaging the plants, and T. vaporariorum adults, which unlike nymphs (Rodríguez et al. 2003), can be handled during flight without injury. In the case of T. absoluta-damaged plants, preliminary tests were performed with different ratio of larvae/leaflet to obtain plants with at least 25% of the leaflet surface damaged after 24 h, but without dying during the assay time (72 h). The ratio of 1 larva (L3) for every two leaflets (that is between 20 and 25 larvae per plant) was chosen. Leafminer larvae were not removed from the plants after the 24-h initial damage to cause continuous damage throughout the experiments, and to avoid causing mechanical damage to the plant, which might modify plant volatile emissions (Raghava et al. 2010). At the end of the experiments leaf damage was visually evaluated resulting in ca. 33% leaf surface damage. In the case of T. vaporariorum-damaged plants, 75 field-collected adults were placed on fully grown- tomato plants for 24 h. As these insects were not sexed in advance, damage to the plants could have been caused by both feeding and oviposition. As with leafminer larvae, adults were not removed from the plant after the initial 24-h damage period, to have continuous damage on the plant. Therefore, when performing settling preference assays with T. vaporariorum, the number of settled adults was corrected as explained below.

Preference Bioassays

All bioassays were run under the same environmental conditions used to rear plants and insects as choice experiments in which one intact and one damaged plant of the same age and foliar development were offered in opposite sides of a cage (55 × 55 × 120 cm, 363 L). To assess oviposition preference by T. absoluta, 50 adults of both sexes were released. After 72 h, the number of eggs laid on each plant was registered (n = 12 for T. absoluta-damaged vs. undamaged plants; n = 7 for T. vaporariorum-damaged vs. undamaged plants). For assessing T. vaporariorum settlement preferences, 75 adults were released, and the number of whiteflies settled on each plant was registered at the end of the assay (72 h) for undamaged vs. T. absoluta-damaged plants (n = 11), or after 24, 48 and 72 h for undamaged vs. T. vaporariorum-damaged plants (n = 11).

To account for the previous presence of T. vaporariorum in the damaged plants, we ran preliminary tests that showed that no more than 2.5% of whiteflies migrate from the damaged to the healthy plant in the 72-h assay time window. Besides, the percentage of whitefly death during 72 h was 30% in the first day, 14.4% in the second and 13.7% in the third (n = 15). According to these preliminary results, when counting settled whiteflies in conspecific-damaged plants, corrections were made by subtracting the whiteflies initially used for damaging the plants; that is, 52 (30% of 75) individuals in day one, 45 (14.4% of 52) in day two, and 39 (13.7% of 45) in day three. Both preference bioassays and volatile collection in the case of T. absoluta-damage were conducted in January-March 2019; and in January-March 2020 for T. vaporariorum’s.

Plant Volatile Chemistry

To obtain volatile extracts from the same plants before and after damage, volatile collections were first performed from undamaged plants for 72 h, then the same plants were damaged as described previously, and another volatile collection (72 h) was performed from the damaged plants. This collection-damage-collection procedure was performed five times in blocks of 3 plants, reaching 15 plant volatile extracts for both T. absoluta- and T. vaporariorum-damaged plants. Two plants and their volatile collections had to be discarded from each group, therefore reaching the final sample sizes as 13 plant volatile extracts sampled before and after damage by each herbivore.

Volatile collections were done by enclosing potted tomato plants, with their pots wrapped in aluminum foil, in Teflon-sealed glass cylinders (37 cm h x 28 cm d; 17 L) at room temperature. A background volatile control was done in parallel for each group of three sampled plants, using a pot with soil wrapped with aluminum foil. A stream of charcoal-filtered air was pushed through each of the cylinders at 2 L/min using an electric air compressor (Toshiba TOSCON) and simultaneously pulled from the plant chamber at a flow of 1 L/min for 72 h, using a CASELLA Apex 2 pump. Volatiles were adsorbed on 50 mg HayeSep Q (Hayes Separations, Inc.), then eluted with 1 mL of double-distilled hexane and added with tridecane (24 µg) as internal standard (IS). After elution, the solution was concentrated to 100 µL under a N2 stream.

A Shimadzu QP2010 plus gas chromatograph coupled to mass spectrometer (GC-MS) was used for volatiles characterization. The analyses were performed with an OPTIMA-5-MS column (30 m x 0.25 mm id x 0.25-µm film thickness; Macherey-Nagel). The analytical conditions were as follows: gas carrier: He (1 mL/min); oven temperature: from 40 °C (1 min) to 160 °C (3 °C/min), 235 °C (5 °C/min) and finally to 280 °C (20 °C/min, 2 min); injector and detector temperatures: 250 °C; injection 1 µL in the splitless mode; ionization potential: 70 eV; scan range: 40–350 m/z. The identification of volatiles was performed by comparing (± 5 units of difference was considered a match) calculated Retention Indices (RI) with those reported by Adams (2007) and by comparison of fragmentation patterns with those contained in NIST 05 (Linstrom and Mallard 2005) and Adams (2007) mass spectrum libraries. The amount of each volatile was quantified by manually integration and expressed as µg of internal standard.

Statistical Analyses

In bioassays, the number of eggs laid by T. absoluta and the corrected number of T. vaporariorum adults settled in undamaged vs. damaged plants were compared with Wilcoxon signed-rank tests, using the VassarStats website (Lowry 2023). Volatile profiles of damaged and undamaged plants were analyzed using the online Metaboanalyst platform (Chong et al. 2018). The GC-MS data were not filtered before multivariate analyses. Normalization, scaling, and centering of the data was done (operations used are indicated in each result) (Chong et al. 2018). Outliers were not detected. GC-MS profiles of volatiles were first explored with unbiased Principal Component Analyses (PCA). Further analyses for identifying peaks that contribute to the differentiation of samples were done using Partial Least Square-Discriminant Analyses (PLS-DA). In these supervised models, the health status (2 levels: damaged and undamaged plants) was included as the classification variable in the model. The PLS-DA models were cross validated with permutation tests (number of permutations as indicated in each result). Then, the PLS loading and variable influence on the projection (VIP) scores were used to make a selection of peaks of interest (VIP > 1) (Xia and Wishart 2016) that contributed to the differentiation of undamaged vs. T. absoluta-damaged plants and between undamaged and T. vaporariorum-damaged plants. Random Forest Analyses were also run. In this case, during tree construction for the classification process, about one-third of the instances were left out of the bootstrap sample. The out-of-bag (OOB) error was calculated and used for variable importance estimation; and precision (percent of correct predictions), recall (percent of correct classification); and Prior Probability (percent expected by chance) were calculated. Besides, data on individual compounds were also subjected to conventional univariate analyses (t-tests on paired samples).

Results

Preference Bioassays with T. absoluta-Damaged Plants

Tuta absoluta females preferred to lay eggs on conspecific-damaged plants (41 ± 5 total number of eggs/plant) rather than on undamaged plants (22 ± 3 total number of eggs/plant; p = 0.005, Wilcoxon test, Fig. 1A). Trialeurodes vaporariorum also preferred to settle on T. absoluta-damaged plants (20 ± 5 whiteflies/plant) rather than on undamaged plants (9 ± 1 whiteflies/plant; p = 0.03, Wilcoxon test, Fig. 1B).

Fig. 1
figure 1

Oviposition preference of T. absoluta (A, N = 12) and settling preference of T. vaporariorum adults (B, N = 11) after 72 h between undamaged and T. absoluta-damaged plants in choice experiments (numbers of eggs and individuals respectively). * denotes significant differences (p < 0.05, Wilcoxon signed-rank tests; results shown as mean ± standard error)

Full size image

Preference Bioassays on T. vaporariorum-Damaged Plants

Tuta absoluta preferred laying eggs on T. vaporariorum-damaged (42 ± 5 total number of eggs/plant) rather than on undamaged plants (20 ± 4) (p = 0.03, Wilcoxon, Fig. 2A). For conspecific-damaged vs. undamaged plants, T. vaporariorum preferred settling on damaged plants in all of the three days assessed: day one (44 ± 5) vs. (17 ± 2), day two (46 ± 6) vs. (16 ± 3) and day three (38 ± 6) vs. (13 ± 2) (p = 0.006, p = 0.001, p = 0.021, respectively, Wilcoxon tests, Fig. 2B).

Fig. 2
figure 2

Oviposition preference of T. absoluta (A, N = 7) and settling preference of T. vaporariorum adults (N = 10) after 24 h (B), 48 h (C) and 72 h (D) between undamaged and T. vaporariorum-damaged plants in choice experiments (* denotes significant differences at p < 0.05,Wilcoxon signed-rank tests; results shown as mean ± standard error)

Full size image

Plant Volatile Chemistry

The volatile profiles emitted by tomato plants were complex as expected (Fig. S1). Combining the volatiles from the four plant treatments studied, 147 chromatographic peaks were detected (Tables 1 and 2). Among these peaks, 103 were from plants damaged by T. absoluta and their corresponding undamaged plants, and 92 were from plants damaged by T. vaporariorum and their corresponding undamaged plants (Table 3). While the tomato cultivar was the same for all experiments (San Pedro), volatiles from undamaged plants used for T. absoluta damage experiments differed from the volatiles from undamaged plants used for T. vaporariorum damage (Tables 1 and 2 and Fig. S1). These differences may arise from within-cultivar genotypic variations between the two experimental years, or from other experimental conditions beyond our control. Therefore, volatiles from T. absoluta damaged plants and from T. vaporariorum damaged plants will be analyzed separately. While only 49 out of the 147 peaks detected were present in all plants in the different treatments, these common compounds accounted for most of the volatiles (average range from 73 to 78% among the four treatments), and most of them have been reported in previous studies (Anastasaki et al. 2018; Ángeles López et al. 2012; Ayelo et al. 2021a, c; Caparros Megido et al. 2014; Milonas et al. 2019; Proffit et al. 2011; Silva et al. 2017, 2018). Identified compounds belong to the common classes usually found in plant volatiles (Table 3).

Table 1 Volatiles (µg eq IS ± SE) emitted by undamaged and T. absoluta-damaged tomato plants (N = 13 pairs of undamaged/damaged plants)
Full size table
Table 2 Volatiles (µg eq IS ± SE) emitted by undamaged and T. vaporariorum-damaged tomato plants (N = 13 pairs of undamaged/damaged plants)
Full size table
Table 3 Volatile analysis overview of undamaged and damaged tomato plants
Full size table

In the case of T. absoluta-damaged plants, despite the similarity in the qualitative analysis, there was a tendency to increase volatile emissions (56 ± 5 in damaged plants vs. 22 ± 5 in undamaged plants -µg eq IS ± SE; p = 0.05, Wilcoxon test, Table 3). Of the 103 quantified peaks, 91 of them were successfully identified (Tables 1 and 3). Ten out of the 103 compounds were emitted only by damaged plants. On the other hand, only one unidentified compound, Unk3, was emitted in undamaged plants but not in damaged ones (Table 1). Univariate analyses on the 103 compounds showed that 12 of them varied when comparing the volatiles from undamaged and damaged plants (Table 1, paired t-tests, p < 0.05). These 12 compounds included 1 monoterpene (carvacrol), 1 aldehyde (2E-decenal), 2 sesquiterpenes (β-caryophyllene, α-humulene), 1 diterpene (E,E-geranyl linalool), 2 hydrocarbons (undecane and tetradecane), 1 alcohol (2-hexyl-1-decanol), 1 esters (3Z-hexenyl butanoate), 2 aromatics (indole, benzophenone), and the carotenoid (E)-β-ionone.

Multivariate analyses were performed on the matrix generated (103 compounds x 2 plant treatments), with previous data normalization (Log10) and scaling (Pareto scaling) (Alaerts et al. 2010). First, the PCA (singular value decomposition, Fig. S2A) showed that the data was well explained (80% of variance) by 5 components (Component 1: 49%; Component 2: 16%; Component 3: 6%; Component 4: 5%; Component 5: 4%). When modeling these data by the PLS-DA, the model has a p value > 0.1 in the permutation tests, so no conclusion could be drawn from this analysis (Fig. S3). Worth noticing, a Random Forest analysis correctly classified samples from the two plant treatments with an OOB error = 0.11 (precision 92% and 85% for damaged and undamaged plants respectively). All these data together show that volatiles emitted by undamaged plants present some compounds in significantly different amounts than volatiles from plants damaged by T. absoluta.

In the case of T. vaporariorum-damaged plants, there was a significant reduction in the total amount of volatiles emitted after damage (26 ± 7 vs. 13 ± 3 µg eq IS ± SE, p = 0.002, Wilcoxon test). Of the 92 quantified peaks, 64 of them were successfully identified (Tables 2 and 3). In this case, all 92 peaks were detected in both kinds of plants, in different amounts. Univariate analyses on the 92 compounds showed that 2 of them varied when comparing the volatiles from undamaged and damaged plants: δ2-carene and β-phellandrene (Table 2, paired t-tests, p < 0.05). Multivariate analyses were then performed on the matrix generated (92 Compounds x 2 kind of plants). Previous to statistical analysis, these data were normalized (square root), and scaled (Pareto scaling) (Alaerts et al. 2010). First, the PCA (singular value decomposition, Fig. S2B) showed that the data was well explained (67% of variance) by 5 components (Component 1: 28%; Component 2: 14%; Component 3: 12%; Component 4: 7%; Component 5: 6%). Then a partial least squares-discriminant analysis (PLS-DA, Fig. 3) was used to model the differences between undamaged and damaged plants. Permutation tests based on separation distance were applied to evaluate the reliability of the model (2000 permutations, p = 0.04). Overall, the PLS-DA model was found to be an acceptable model for discrimination between the plant status. The validated model had 5 components, with R2 = 0.95, Q2 = 0.85 and accuracy of 1 (Fig. 3). From the model built, 26 compounds with a Variable Importance in Projection (VIP) greater than 1 (Chong et al. 2018) were found (Table 2). Nineteen out of these 26 were identified (Table 2): 4 were monoterpenes (δ2-carene, p-cymene, β-phellandrene, (E)-β-ocimene), 1 aldehyde (Undecanal), 4 sesquiterpenes (δ-elemene, farnesane, β-caryophyllene, 2,6,10-trimethyltridecane -modified ST-), 7 hydrocarbons (tetradecane, octyl-cyclohexane, 5-methyl-tetradecane, 2-methyl-tetradecane, 3-methyl-tetradecane, pentadecane, hexadecane), 1 alcohol (2-hexyl-1-decanol) and 2 esters (isopropyl myristate, methyl hexadecanoate). Finally, Random Forest analyses correctly classified samples from the two plant statuses with an OOB error = 0.08 (precision 85% and 100% for damaged and undamaged plants respectively). All these data together also clearly show that the volatiles from undamaged plants can be differentiated from the volatiles from species damaged by T. vaporariorum.

Fig. 3
figure 3

PLS-DA of undamaged vs. T. vaporariorum-damaged plants volatiles: Score plot of PC1 vs. PC 2 (the explained variances are shown in brackets, A) and VIP score plot showing compound with VIP > 1 (B). *Unk: Unknown (in the Case on Unk HC 3, it was classified as hydrocarbon (HC).

Full size image

Discussion

Tuta absoluta females laid more eggs on conspecific-damaged plants in two-choice bioassays with undamaged plants as an option. These results may be due to different plant chemistry, either volatile or non-volatile, resulting from the previous damage by conspecifics. Our volatile analysis in tomato plants (discussed later) did show significant changes due to T. absoluta damage, but we cannot rule out other possible effects, chemical or otherwise, related to the presence of T. absoluta larvae feeding on the plants. Beyond the mechanistic explanation, the oviposition preference we found may be also discussed in adaptive terms. It has been shown that lepidopteran larvae may benefit from developing together, concerning both biotic and abiotic stressors (Tsubaki 1981). Moreover, aggregation of lepidopteran larvae (Jin et al. 2016; Jumean et al. 2004; Tsubaki 1981) and attraction of females to oviposit (Sun et al. 2014) on plant areas with presence of conspecifics have been previously reported for different families. While T. absoluta females lay eggs singly, several larvae usually coexist in the same plant, so potential adaptive explanations for T. absoluta females to prefer ovipositing on conspecific-damaged plants are not unforeseen. This preference may be counterintuitive if factors such as intra-specific resource competition, cannibalism or induced plant defenses are considered. On the other hand, infested plants may be more suitable for future larvae, or may result in a decreased likelihood of predation or parasitism. A more specific discussion would require a thorough knowledge of the natural history of T. absoluta in its native range and trophic web. While the literature on T. absoluta is quite extensive, the focus has been on applied aspects and simplified systems (our study is no exception).

Our oviposition preference results differ from previous studies that reported T. absoluta females ovipositing more eggs on undamaged plants in comparison with conspecific-damaged ones (Anastasaki et al. 2018; Bawin et al. 2014; Maneesha et al. 2021). These contradictory results may stem from methodological differences or even from plant cultivar characteristics. Bawin et al. (2014) found oviposition preferences for undamaged plants but no volatile-mediated preferences in wind tunnel assays, pointing to other plants cues such as non-volatile induced defenses. Maneesha et al. (2021) did not report experimental conditions such as the size of the bioassay cages, larval infestation levels or whether damaging larvae were retrieved before the preferences assays, all factors that may have influenced the preference results. Anastasaki et al. (2018) worked with preference bioassays using shorter distances (60 cm vs. 120 cm) and smaller containers than ours (2.1 L vs. 3.6 L), which may have influenced the results due to the closeness of the contrasting plants. Infestation levels were also different both in the number and instar of damaging larvae: twenty larvae (L1) were used by Anastasaki et al. (2018), while we used 25 older larvae (L3), likely increasing the level of damage and induced response by the plant. Indeed, it has been reported that the age of T. absoluta larvae feeding on tomato plants affects adult attraction to the plants (Abdelhady et al. 2020). All in all, different results may arise from different experimental conditions or cultivars; San Pedro in our study, Moneymaker in Bawin et al. (2014), Semiramis in Anastasaki et al. (2018) and Sahoo TO-3251 in Maneesha et al.’s work (2021). Different cultivars are not only reported to impact in T. absoluta fitness (Mathieu W. Sawadogo et al. 2021) and oviposition preferences (Cherif 2013; Proffit et al. 2011), but also differ in the volatiles emitted (see below) as well as in the defensive secondary metabolites and trichome density, which are known to play a role in tomato resistance to T. absoluta (Sohrabi et al. 2016) and differentially affect T. absoluta oviposition behavior (Oliveira et al. 2012; Labory et al. 1999).

Our results also showed that T. vaporariorum preferred to settle on plants with previous damage by conspecifics. This is in line with results with B. tabaci females, which prefer to lay on plants previously occupied by conspecifics (Silva et al. 2021a; Su et al. 2018). Previous olfactometer studies with T. vaporariorum adults, however, showed conflicting results. Trialeurodes vaporariorum adults were more attracted to volatiles emitted by undamaged plants than to those from conspecific-damaged plants. This was the case for two tomato cultivars, but did not hold for two other cultivars (Deletre et al. 2022) (none of these cultivars are the same as in this study). In a different study, T. vaporariorum males were more attracted to conspecific-infested tomato plants than to undamaged ones, but females showed the opposite results (Darshanee et al. 2017). Therefore, attraction to undamaged over damaged plants for T. vaporariorum depends on plant cultivar and whitefly sex. In our work, experimental T. vaporariorum adults were not separated for sex, so we cannot rule out that the settling preference we found was not affected by sex ratio of the tested insects.

Our findings of T. vaporariorum settling preferences may also have adaptive implications. It has been shown that whiteflies may benefit from conspecific feeding aggregations (facilitation) via sink modification, a mechanism by which whiteflies can control the sap flow within the plant to their advantage. This in turn may reduce the nutritional quality of the plants to other competing herbivores, especially non-sap-feeders (Inbar and Gerling 2008). In addition, whitefly feeding has been reported to suppress the effects of the JA pathway by activation of the SA pathway (Zhang et al. 2013), which may also benefit the formation of aggregations.

In our crossed herbivore preference bioassays, T. absoluta preferred ovipositing on T. vaporariorum damaged plants, and T. vaporariorum preferred settling on T. absoluta damaged plants. Partly in line with our results, T. absoluta larval development was positively affected by the previous presence of B. tabaci, but only on locally-damaged leaves and not when the damage occurred in different leaves (Mouttet et al. 2013). This effect was not symmetrical, since previous infestation by T. absoluta did not affect the development of B. tabaci nymphs (Mouttet et al. 2013. Our results are also in line with a study with the leaf-chewer Pieris brassicae (Lepidotera: Pieridae) and the phloem-sucking aphid Brevicoryne brassicae (Hemiptera, Aphididae), which showed better performance of the larvae on heterospecific damaged cabbage plants, in comparison with undamaged plants (Soler et al. 2012). Additional preference experiments for crossed herbivory effects may be the oviposition preference of T. vaporariorum and the adult settlement preference of T. absoluta. These were not attainable in our working conditions but would likely provide interesting additional information.

The difficulty for the plant to defend itself when it has been already attacked by a pest that activates a different defensive route has already been studied in various systems in relation to the crosstalk effect between hormone-controlled pathways (Thaler 2012). The previous results by Mouttet et al. (2013) rise the hypothesis that the tomato plants are able to defend itself at the same time locally to herbivore damages but not systemically for both herbivores. A similar scenario may allow T. absoluta and T. vaporariorum coexistence. No oviposition preference of T. absoluta has been previously documented, as far as we know, related to the previous presence of T. vaporariorum.

To sum up, our laboratory findings did indeed confirm the co-occurrence that was previously observed in local greenhouses. In the case of T. vaporariorum-damaged plants, preference would be caused by a downregulation of plant indirect defenses as happens when B. tabaci damages the tomato plant (Zhang et al. 2019). In the case of T. absoluta-damaged plants, more studies are needed to elucidate the underlining mechanisms that favor the whitefly settling. The differences found compared to previous reports, and among them, may be explained due to the cultivars of tomato plants used, the distance between the stimuli, the number of larvae of the tomato leafminer or nymphs or adult whiteflies causing damage, among other effects. Besides, since damaging insects were not removed, we cannot rule out that attraction to damaged plants may be also influenced by some insect cue. Although, in the case of T. vaporariorum attraction to conspecifics seems not to take place (Darshanee et al. 2017).

Regarding volatiles emitted by the tomato plants here studied, as mentioned, T. absoluta damaged plants emitted more volatiles after the damage (Table 3), as it was previously reported (Anastasaki et al. 2018; Ayelo et al. 2021a; Chen et al. 2021; Silva et al. 2017, 2018). Among the many reports on T. absoluta-damaged plants volatiles, although the main identified compounds are similar among works, there is an enormous variation on the reported compounds (Anastasaki et al. 2015, 2018; Ayelo et al. 2021a; Caparros Megido et al. 2014; Chen et al. 2021; De Backer et al. 2015; Gontijo et al. 2019; Milonas et al. 2019; Proffit et al. 2011; Silva et al. 2017, 2018). In these works, together, more than 200 compounds are reported, of which none of them are common to all works, and for instance, only 32 (16%) are reported in at least four of them (considering the works where all compounds are listed). The different tomato cultivars, soil, location and season of the year in which the volatiles collections are done are some of the factors that may explain these differences in volatile profiles (Holopainen and Gershenzon 2010). Except for the work by Chen et al. (2021), none of the other publications reviewed here quantify more than 60 compounds. In this work, 103 peaks (91 identified) are reported, allowing to show change in minor peaks after the damage. Among the 12 compounds found here to vary significantly in T. absoluta-damaged plants, β-caryophyllene was previously reported to cause a physiological response in T. absoluta’s antennae (Anastasaki et al. 2018; Miano et al. 2022; Pagadala Damodaram et al. 2021), and to increase significantly its amount after the damage (Anastasaki et al. 2018; Ayelo et al. 2021a, b; Maneesha et al. 2021; Silva et al. 2018), as in our results (Table 2). However, in another report, even if β-caryophyllene was detected in HIPVs from T. absoluta-damaged plants, its amount did not vary after being damaged just as the amount of the rest of the sesquiterpenes did not vary either (De Backer et al. 2015). Tetradecane (Maneesha et al. 2021), (Z)-3-hexenyl butanoate (Ayelo et al. 2021a), Indole (Silva et al. 2017), (E)-β-ionone and α-humulene (Ayelo et al. 2021a; Silva et al. 2018) were also previously reported to increase significantly when the tomato plants are damaged by T. absoluta larvae. Besides, α-Humulene and Undecane were also previously reported to cause a physiological response in T. absoluta’s antennae (Pagadala Damodaram et al. 2021) although Undecane was not reported to increase after damage, as it did in our work. However, in the case of Tetradecane and (Z)-3-hexenyl butanoate, no antennal response to them was detected (Anastasaki et al. 2018). The other five compound (2-E-Decenal, Carvacrol, 2-hexyl-1-Decanol, Benzophenone and E, E-Geranyl linalool, Table 1) that vary significantly their emision after T. absoluta damage in tomato plants were not previously reported in any of the above-mentioned works.

Damaged plants by T. vaporariorum emitted less volatiles than undamaged tomato plants (Table 2), similarly to the previously reported not only for T. vaporariorum (Deletre et al. 2022) but also for B. tabaci (Silva et al. 2017). However, this pattern is not general as it has been also reported that HIPVs after T. vaporariorum damage increased when tomato plants were infested by 100 whitefly adults (Ayelo et al. 2021c). Interesting, this last work also showed that the emission of HIPVs decreased when infestation was higher (with 200 adults) (Ayelo et al. 2021c). In the case of B. tabaci-damaged tomato plants, an increase in HIPVs has also been reported (Silva et al. 2021a). As mentioned, tomato volatile characterization has been previously reported not only for T. vaporariorum damage (Ángeles López et al. 2012; Ayelo et al. 2021c; Darshanee et al. 2017; Deletre et al. 2022) but also for B. tabaci’s (Chen et al. 2020; Silva et al. 2017, 2018, 2021a; Su et al. 2018). Among these reports and ours, more than 150 compounds were quantified. Of the 26 compounds that were here found to vary significantly after damage by T. vaporariorum, only three compounds, (E)-β-ocimene, terpinolene and β-caryophyllene are quantified in all reports. δ2-Carene and β-phellandrene, were previously reported to decrease significantly after damage by T. vaporariorum (Ángeles López et al. 2012). β-Phellandrene also decreased in four tomato cultivars after T. vaporariorum’s damage (Deletre et al. 2022; Silva et al. 2018) but increased after B. tabaci’s (Chen et al. 2020; Silva et al. 2018). On the other hand, β-caryophyllene and (E)-β-ocimene increased significantly in the different works (Ángeles López et al. 2012; Ayelo et al. 2021c; Silva et al. 2018). Although β-caryophyllene increased after damage of 100 whiteflies, decreased after damage of 200 whiteflies. p-Cymene increased in the only work that was quantified and was also reported to be repellent for T. vaporariorum (Ayelo et al. 2021c). Our results showed that p-cymene and (E)-β-ocimene decreased significantly, and adults were attracted to damaged plants (see above). One wonders whether the decrease in repellent compounds (Deletre et al. 2022) may favor the settling of the whiteflies. δ-Elemene was quantified in two publications and in both it did not vary significantly (Ayelo et al. 2021b; Silva et al. 2018). The four compounds just mentioned that vary plus δ2-Carene and hexadecane are reported to cause a physiological response in T. absoluta’s antennae (Anastasaki et al. 2018; Chen et al. 2021; Miano et al. 2022; Pagadala Damodaram et al. 2021). The other 18 compounds that vary significantly in our work were not quantified in any of the above-reviewed reports. This comparative analysis shows the huge variation among works in volatiles emitted by whiteflies-damaged tomato plants. This variation is such that in some works volatiles are up-regulated and in others are down-regulated, either as whole or individually. Given that, at least in the case of T. vaporariorum, the gregariousness does not respond to own cues (Darshanee et al. 2017), it is highly probable that the stimuli of the plants contribute to the gathering of the insects. Clearly then, further studies like olfactometry assays and electroantennography analyses are needed to understand the role of these compounds.

Comparing the variation of volatiles emitted after been damaged by T. absoluta and T. vaporariorum, as mentioned, 12 and 26 compounds vary significantly respectively, but only two of them, β-caryophyllene and tetradecane, vary in both cases. Both compounds increased after T. absoluta damage and decreased after T. vaporariorum’s. Having both insects different feeding habits (Pieterse et al. 2012), they would activate different defensive routes in tomato plants (Walling 2000), leading probably to different variations in volatiles. Since both insects were attracted to plants damaged by conspecific and heterospecific individuals, which exhibited different volatile variations, the attractiveness to damaged plants could be guided by the mixture of compounds rather than by any of them individually, and by the balance with other non-volatile stimuli present.

Conclusions

In this study, T. absoluta and T. vaporariorum preferred laying eggs or settling respectively on damaged plants by either a conspecific or the other herbivore. Our results on the insects´ preferences agree with both, our own field observation and with previous reports in the case of T. vaporariorum. However, the oviposition preference observed for T. absoluta differs from some previous studies. This difference probably arises from different experimental conditions and tomato cultivars used in the studies. The preference for conspecific-damaged plants may be driven by gregariousness or a general decrease in plant defenses following an attack. On the other hand, the preference for heterospecific-damaged plants might be due to the challenge plants face in defending themselves against simultaneous attacks from both insects, as reported in other systems. The compounds emitted by the tomato plants varied in a different way when attacked by one or the other insect; and differed also from the variations of individual compounds previously reported in tomato HIPVs. Interestingly, some individual compounds showed opposing variations under the damage caused by both herbivores. Nonetheless, plants emitting these compounds remained consistently attractive to the herbivores. This suggests a hypothesis that preferences are determined by the mixture of volatiles rather than individual compounds, as well as possibly modulated by other stimuli. To establish a definitive conclusion on whether this chemical variation in volatiles governs insects’ choice, further studies utilizing olfactometer preference bioassays and wind tunnel experiments are warranted, as understanding the behavior of these herbivores (and the underlying cues) in the presence of other one is essential for designing a strategy to control them.