Introduction

Herbivores are suppressed by both bottom-up and top-down processes (Price et al. 1980; Stiling and Rossi 1997; Denno et al. 2002). Bottom-up processes are associated with direct, chemical and physical plant defences that negatively affect herbivore performance (Price et al. 2011; Kant et al. 2015). Top-down processes result from the actions of natural enemies of the herbivores. The enemies often interact with the plants, which possess traits that enhance their attraction or arrestment, often referred to as indirect plant defences (Sabelis et al. 1999, 2001; Schmitz et al. 2000). These two plant defences are not necessarily compatible; besides their negative effects on herbivores, direct plant defences can also negatively affect natural enemies (Eisner et al. 1998; Rutledge et al. 2003; Simmons and Gurr 2005; Riddick and Simmons 2014; Legarrea et al. 2022). For example, glandular trichomes are outgrowths of the plant epidermis in which specialized metabolites are synthesized and stored (Schuurink and Tissier 2020). They predominantly serve as broad-spectrum direct defence against herbivores (Glas et al. 2012), but can also increase the mortality of natural enemies (Rabb and Bradley 1968; van Haren et al. 1987; Verheggen et al. 2009; Legarrea et al. 2022).

Tomato (Solanum lycopersicum L.) is a classic example of a plant species with glandular and non-glandular trichomes (Kennedy 2003; Glas et al. 2012). The glandular trichomes are multicellular and consist of basal, stalk, and apical cells. The apical cells contain metabolites that can entrap small insects and predatory mites and can be highly toxic to mites (van Haren et al. 1987; Sato et al. 2011; van Houten et al. 2013; Legarrea et al. 2022). Nevertheless, tomato plants are still attacked by herbivores that are adapted to these direct defences (Sarmento et al. 2011; Glas et al. 2014). For example, the tomato russet mite Aculops lycopersici (Tryon) (Acari: Eriophyidae) is one of the most destructive pests of tomato worldwide (Lindquist et al. 1996; Van Leeuwen et al. 2010; Vervaet et al. 2021). It is so small (adults are 150–200 μm in length, Vervaet et al. 2021) that it can easily move among the trichome stalks without touching the apical cells. Thus, they are protected from larger competitors and predators by the glandular apical cells above them (van Houten et al. 2013). The mobile stages of this pest feed on the epidermis of foliage, stems, inflorescences and (young) fruits, causing browning, shrivelling and necrosis of stems and leaves, flower abortion and russeting of fruits (Royalty and Perring 1988; Duso et al. 2010; Vervaet et al. 2021). At high infestations, tomato plants may shrivel, leading to complete production loss (Vervaet et al. 2021). Currently, the russet mites are only controlled through pesticide applications (e.g., abamectin and sulphur, Duso et al. 2010; Vervaet et al. 2021), but this is often ineffective (Lindquist et al. 1996; van Houten et al. 2013), or require frequent applications throughout a cropping period. Furthermore, a meta-analysis of published data from field experiments shows that chemical pest control often fails in the presence of natural enemies (Janssen and van Rijn 2021). It is therefore preferable to search for efficient natural enemies to avoid pesticide use.

Predatory mites of the Phytoseiidae family are one of the main groups of natural enemies studied for tomato russet mite control (Brodeur et al. 1997; Park et al. 2010; Al-Azzazy et al. 2018). Although some phytoseiid mites may be or become adapted to tomato plants and their trichomes (Drukker et al. 1997; da Silva et al. 2010; Sato et al. 2011), the effectiveness of most phytoseiid mites on tomato plants is limited by the presence of glandular trichomes, and to date, no known phytoseiid can control the russet mite on tomato plants with trichomes (Legarrea et al. 2022). We hypothesised that predatory mites found naturally associated with tomato plants and their pests would be more adapted to these trichomes and could therefore control pests on this crop.

Recently, it was demonstrated that the predatory mite Amblyseius herbicolus (Acari: Phytoseiidae), originally collected from feral tomato plants (see Materials and Methods), was able to establish on tomato plants when supplemented with pollen as alternative food and could significantly reduce whitefly densities on these plants (Cardoso 2019). This last study also demonstrated that this predator can develop and reproduce when feeding on A. lycopersici, but its effectiveness in controlling russet mites on tomato plants has not yet been evaluated.

We have also encountered a new iolinid species, cf. Homeopronematus anconai sp. nov., on tomato plants. Several iolinid genera are currently subject of a taxonomic revision (Ueckermann & De Vis et al. 2024) and the new species is currently being described by R. De Vis and colleagues. The closely related iolinid predatory mite Homeopronematus anconai (Baker) and the more distantly related Pronematus ubiquitus (McGregor) (Acari: Iolinidae) were reported to be adapted to tomato plants (Kawai and Haque 2004; Pijnakker et al. 2022b) and to feed and reproduce on A. lycopersici (Hessein and Perring 1986; Kawai and Haque 2004; van Houten et al. 2020; Pijnakker et al. 2022a, 2022b). Like A. lycopersici, both species are so small that they can move in between the stalks of the trichomes without touching the apical cells. These mites are omnivorous, feed on plant tissue (Pijnakker et al. 2022b; Vervaet et al. 2022), and can only be reared in the presence of plant tissue (except for Pronematus sp., De Vis et al. 2006; Ueckermann & De Vis et al. 2024). However, they cannot complete their life cycle feeding on tomato leaves alone but require additional prey, pollen or fungi (Hessein and Perring 1986; Kawai and Haque 2004; van Houten et al. 2020; Pijnakker et al. 2022a, 2022b). In a recent study, Pijnakker et al. (2022b) showed that establishing P. ubiquitus on tomato plants supplemented with pollen prevented A. lycopersici infestations in a greenhouse. They also demonstrated that it effectively controlled powdery mildew infections (Oidium neolycopersici L.), a devastating tomato disease (Jones et al. 2001; Kiss et al. 2001; Pijnakker et al. 2022b).

Current management practices of powdery mildew include fungicide applications, the use of sulphur, the development of resistant crop varieties and technology-based climate control in greenhouses crops (Castañé et al. 2020; Vielba-Fernández et al. 2020). Successful biological control of powdery mildew in vegetables is achieved almost exclusively by antagonistic microorganisms such as the fungus Ampelomyces quisqualis (Kiss et al. 2004; Caffi et al. 2013). Although several studies demonstrated the ability of fungivorous arthropods (mainly mites) to reduce powdery mildew infestations (Hessein and Perring 1986; English-Loeb et al. 1999; Sutherland and Parrella 2009; Pijnakker et al. 2022b), fungivores are not often used for biological control of fungal plant diseases. We therefore tested the capacity of two predatory mites originating from tomato plants to control the tomato russet mite and powdery mildew on individual plants. We assessed the reproductive performance of A. herbicolus and the cf. H. anconai sp. nov. when feeding on A. lycopersici and pollen. Subsequently, we evaluated the ability of both predators to control of russet mites on tomato plants, as well as their impact on fruit production. Additionally, we examined the effectiveness of the iolinid mite in suppressing natural powdery mildew infections on tomato plants.

Materials and methods

Plant materials

Tomato seeds (Solanum lycopersicum var. Aguamiel (EX V305), Limagrain®, imported by Vilmorin Seed Generation, Campinas—Brazil) were sown in a commercial plant substrate (MECPLANT®—Mec Prec, Telêmeco Borba), in polystyrene trays (8 × 16 cells). This variety is susceptible to powdery mildew but is resistant to various plant-pathogenic viruses. Fifteen days after germination, the seedlings were transplanted to plastic 3 L pots with the same substrate. The plants were fertilized every week with a solution of 50 g of N-P-K (20-05-20) and 100 g of simple superphosphate in 20 L water and were tied with a string to bamboo sticks for support. The plants were kept inside cages in a room with natural light at 23 ± 3 °C and 70 ± 10% relative humidity. Pollen was collected from cattail plants (Typha sp.) from the campus of the Federal University of Viçosa. The pollen was dried in an oven at 60 °C for 24 h and stored in a freezer at − 20 °C (van Rijn and Tanigoshi 1999; Montserrat et al. 2006).

Mite rearing

Tomato russet mites were collected from tomatoes grown in a rural area of Coimbra in 2021 (MG, Brazil, 20° 51′ 42.4″ S, 42° 50′ 14.4″ W) and were maintained on tomato plants inside cages (0.5 × 0.5 × 1.0 m) in a room at 22 ± 2 °C, 70 ± 10% RH, 12:12 L:D. A new tomato plant with at least four completely developed leaves was added to each cage every month and the oldest plants were removed.

The laboratory rearing of cf. H. anconai sp. nov. was initiated with mites collected on russet-mite-infested tomato plants in the house of the first author in Viçosa in 2021 (20°45′35.3″S 42°52′58.3″W). The mites were reared on tomato plants, to which cattail pollen was added twice a week as extra food. The tomato plants were kept inside cages (0.5 × 0.5 × 1.0 m), which were kept in climate rooms as above. To obtain cohorts of adult females of similar age, larvae were carefully placed on clean plants using a fine brush. Three weeks later, the plants harboured young adult females, which were used for experiments.

The rearing of A. herbicolus was started with individuals collected from tomato plants in gardens in the urban and rural areas of Prados (MG, Brazil) (Cardoso 2019). The species was identified by Manoel Guedes Correa Gondim Junior from UFRPE (PE, Brazil). This predator reproduces through thelytokous parthenogenesis; hence, populations consist of females only (De Moraes and Mesa 1988). They were reared on arenas made of black PVC sheets (15 × 10 cm) on top of foam pads (h = 3 cm), which were kept in plastic trays (29 × 14 × 4 cm) filled with water (van Rijn and Tanigoshi 1999). Wet tissue paper, in contact with the water in the tray, covered the edges of the arenas, and prevented the mites from escaping while serving as a water source. Small pieces of cotton wool were placed on the arenas as oviposition sites and were covered with pieces of black, tent-shaped PVC sheet (1.5 cm2) for shelter. Cattail pollen was offered as food two times per week. The rearing was maintained in climate rooms as above. To obtain cohorts of young adult females, adult females were placed on a new arena with food and a small piece of cotton wool. The next day, the piece of cotton wool with eggs was put on a new arena with pollen and a piece of cotton wool. Ten—12 days later, the eggs had developed into young adult females, which were used for experiments (Kalile et al. 2021).

Predation and oviposition

First, we investigated the ability of both predatory mites to feed on adult russet mites. This experiment was performed on tomato leaf discs (Ø = 2 cm), arranged abaxial side up on wet tissue paper (Ø = 3 cm) and placed in Petri dishes (Ø = 5 cm, 1.5 cm high), which were closed with lids to prevent mites from escaping. Preliminary experiments were done to check how many individuals should be offered to avoid prey depletion. Adult female predators were placed singly on the leaf discs and were starved for 24 h prior to the experiments to prevent effects from the previous diet. We then placed ten adult russet mites on each leaf disc. Ten replicates were carried out per predator species. Predation was scored 24 h later. The predation rates of russet mite adults of the two predators were compared using a generalized linear model (GLM) with a Poisson error distribution.

To assess the ability of both predatory mites to oviposit when feeding on russet mites, we introduced a young adult female of one of the two predatory mite species from a cohort as explained in the previous section on a tomato leaf disc with c. 100 russet mites (mixed stages, including eggs) in experimental units as above. Typha sp. pollen on clean leaf discs was used as food in a control treatment. Oviposition rates were assessed every 24 h for four days. The oviposition rates of predatory and phytophagous mites peak soon after the pre-oviposition period and are a good proxy for the total fecundity and the intrinsic rate of increase (Janssen and Sabelis 1992; Sabelis and Janssen 1994). After each evaluation, predators were transferred to new experimental units containing fresh prey or pollen. Ten replicates were carried out per predator species. To prevent effects from the previous diet or the lack thereof, oviposition of the first day was excluded from the analysis (Sabelis 1990). The tests were performed in a room at 22 ± 2 °C, 70 ± 10% RH, 12:12 L:D. Because of considerable non-normality of the errors, also after transformation, oviposition rates were summed per individual over the last three days of the experiment and compared with a Wilcoxon-Mann–Whitney test (function wilcox.test of the R package coin, Hothorn et al. 2008). Nevertheless, we present oviposition data per day below to show trends through time.

Tomato russet mite control on tomato plants, powdery mildew levels and fruit production

To test for the capacity of A. herbicolus and cf. H. anconai sp. nov. to control russet mites, we carried out experiments in three screen cages (2 × 2 × 2 m—Howitec Netting®, Joure, the Netherlands), under natural conditions (mean temperature 16.1 °C, range 6.7–27.4 °C; mean relative humidity 81%, range 25–99%, Universidade Federal de Viçosa—UFV 2023), but protected from occasional rain. Each cage contained 12 tomato plants, each plant being a replicate. The plants were spaced out in the cage, and each plant pot was put in a dish with water, which served as barrier to prevent cross-contamination by dispersing mites. On June 9th, 2022, a tomato leaf disc (3 cm2) with 5 adult females A. herbicolus was put on all leaves of the tomato plants (4-leaf stage), with a total of 20 predators per plant. The cf. H. anconai nov. sp. were introduced in a similar way, but with 10 mobile stages per leaf disc at a total of 40 individuals per plant. Using a fine brush, about 1 mg of Typha sp. pollen was applied to the top leaf of each plant twice a week, including in the control treatment (without predators). Plants were not sampled during the first 4 weeks to avoid disturbances. A few days after releasing the predatory mites, natural infestations of powdery mildew occurred in all treatments. The presence of the mildew was quantified during four weeks by counting the leaflets with visual mildew lesions on the intact plants.

Four weeks after the release of predatory mites, we assessed the numbers of predatory mites by collecting six leaflets per plant (two leaflets from the bottom, middle and top section of each plant, respectively). The leaflets were stored in separate boxes for evaluation in the laboratory with a stereo microscope. On the same day, plants of all treatments were infested with russet mites as follows. Discs were cut from leaflets picked from a well-infested tomato plant, and each disc harboured on average c. 50 mobile russet mite stages. Each tomato plant received five such leaf discs with russet mites, one disc each on the second to the sixth leaf from the bottom, hence, about 250 mobile stages in total. The discs were positioned on the middle of each leaf. Plants were subsequently sampled for four more weeks (week 5–8) as above and the numbers of predatory mites and russet mites on the sampled leaflets were assessed. At the last sampling, tomato fruits were collected, counted and weighed using an analytical balance (Balança-AD200, Marte Científíca, São Paulo).

The densities of A. lycopersici were log (x + 1) transformed before the analysis and compared with a linear mixed effects model (LME) with time and treatment and their interaction as fixed factors (lme of the package nlme, Pinheiro et al. 2020). Replicate (plant) was used as a random factor to correct for repeated measures. The numbers of tomato fruits and weights of fruits per plant were compared among treatments using a GLM with Gaussian error distribution (identity link). The numbers of leaflets with powdery mildew lesions in the first four weeks of the population dynamics experiment were compared using a generalized linear mixed effects model (GLMER, glmer of the package lme4, Bates et al. 2015) with a Poisson error distribution, treatment and time as fixed factors and replicate (plant) as random factor to correct for repeated measures. All models here and below were checked with normal error plots and plots of residuals against fitted values. Non-significant factors and interactions were removed until a minimal adequate model was obtained (Crawley 2013). Significance of factors and interactions were determined with likelihood ratio (L.R.) tests. Contrasts among treatments were assessed with the Tukey method using the package emmeans (Lenth 2019). All statistical analyses were performed with software R, version 4.2.1 (R Core Team 2023).

Powdery mildew control by cf. Homeopronematus anconai sp. nov.

Based on the results with powdery mildew obtained in the above experiment, we conducted an experiment to evaluate whether the iolinid could affect mildew progress on different strata (bottom, middle and top) of tomato plants. As powdery mildew initially starts on lower leaves of plants, fifty iolinid motiles were introduced on the bottom (= oldest) leaf from each 6-leaf-stage tomato plant (5 plants per treatment). Typha sp. pollen (1 mg per plant) was applied two times per week, divided over all leaves, also on plants of a control treatment without predators. The plants were kept in two separate cages (1 × 1 × 1 m). Plant apices were subsequently removed to keep plants at a constant size and prevent the leaves from touching the top of the cages. For four weeks (July 11—August 8, 2022, mean temperature 17.2 °C, range 9.1–28.6 °C; mean relative humidity 80%, range 34–98%), we counted the leaflets with visible powdery mildew lesions on the entire plants. Powdery mildew occurrence on tomato plants was compared in two steps using a GLMER with a binomial error distribution: first we compared the occurrence of mildew per leaf in the three strata, and then the number of leaflets with mildew for those leaves that had symptoms of mildew.

Results

Predation and oviposition

The average consumption of russet mite adults by cf. H. anconai sp. nov. (0.8 ± 0.2 day−1) and A. herbicolus (5.6 ± 0.8) (mean ± SE) differed significantly (GLM; Deviance = 42.7, d.f. = 1, p < 0.001). The phytoseiid mite thus consumed about 7 times more russet mite adults than the iolinid.

The two predators were able to oviposit by feeding on A. lycopersici. Oviposition rates were significantly affected by diet in both the iolinid (Fig. 1A; Wilcoxon test: Z = 3.52, p = 0.0004) and A. herbicolus (Fig. 1B; Z = 3.75, p = 0.0002). Feeding on cattail pollen resulted in the highest oviposition rates for both predators.

Fig. 1
figure 1

Average (± SE) oviposition rates of cf. H. anconai sp. nov. A and A. herbicolus B on the 2nd—4th day of feeding on all stages of the tomato russet mite A. lycopersici or cattail pollen. Different letters next to final points indicate significant differences in oviposition rates between diets (p < 0.001)

Full size image

Tomato russet mite control on tomato plants, powdery mildew levels and fruit production

There was a significant effect of the presence of predators on densities of A. lycopersici through time (Fig. 2A; LME: interaction of treatment with time: L.R. = 14.4, d.f. = 2, p = 0.0007). The iolinid mites successfully suppressed the increase of A. lycopersici, but A. herbicolus did not (Fig. 2A). Although the iolinid did not eradicate the russet mites, the densities were near zero, whereas they increased in the treatments with A. herbicolus and without predators (Fig. 2A). At the same time, the densities of cf. H. anconai sp. nov. increased 12-fold during the experiment, whereas A. herbicolus disappeared from all plants from week 7 onwards (Fig. 2B).

Fig. 2
figure 2

Population dynamics of Aculops lycopersici, the iolinid cf. H. anconai sp. nov. and A. herbicolus on tomato plants. A Average densities (± SE) of tomato russet mites in the presence of the iolinid (triangles), A. herbicolus (squares) and no predators (circles). Different letters in the legend indicate significant differences between treatments and asterisks after the week number show significant differences within each week between the treatment with cf. H. anconai sp. nov. and the two other treatments; these latter two did not differ significantly (contrasts through model simplification after LME). B Log densities (± SE) of all stages of both cf. H. anconai sp. nov. (squares) and A. herbicolus (circles). The plants were infested with 250 mobile A. lycopersici in week 4. Predators were released in week 0 and were fed Typha sp. pollen

Full size image

The different treatments significantly affected the number of leaflets with powdery mildew lesions through time during the first four weeks (Fig. 3; interaction of treatment with time, GLMER; Chi2 = 26.6, d.f. = 2, p < 0.0001). In the seventh week, the plants showed a high incidence of powdery mildew, except in the treatment with cf. H. anconai sp. nov. The severity of the disease resulted in the drying and senescence of some leaflets, directly affecting the predators still present on the plants. The iolinid mites reduced the spread of the disease with 86.5% and 87.6% compared to the treatments with A. herbicolus and without predators, respectively.

Fig. 3
figure 3

Average (± SE) number of leaflets with powdery mildew lesions in the first four weeks of the population dynamics experiment in presence of the predatory mites cf. H. anconai sp. nov. (triangles), A. herbicolus (squares) and no predators (circles). Different letters in the legend indicate significant differences among treatments through time and asterisks after the week number show significant differences within each week between the treatment with cf. H. anconai sp. nov. and the two other treatments; these latter two did not differ significantly (contrasts through model simplification after GLMER)

Full size image

There was no significant effect of treatment on the numbers of fruits produced (Fig. 4A; GLM; Deviance = 5.79, d.f. = 2, p = 0.055), but plants with the iolinid produced somewhat fewer fruits than plants in the other treatments. These fruits were somewhat, not significantly, heavier than those of plants with A. herbicolus and without predators (Fig. 4B; GLM; F2,33 = 2.0, p = 0.15), and the total fruit weight per plant also did not differ significantly among treatments (Fig. 4C; GLM; F2,33 = 0.55, p = 0.58).

Fig. 4
figure 4

A Average numbers of fruits per tomato plant (± SE) B average weight of individual tomato fruits per plant and C total fruit weight per plant infested with A. lycopersici in the presence of cf. H. anconai sp. nov., A. herbicolus and no predators, after an 8-week experiment. Differences among treatments were not significant

Full size image

Powdery mildew control by cf. H. anconai sp. nov.

The occurrence of powdery mildew on leaves in the three strata of tomato plants was significantly affected by the presence of cf. H. anconai sp. nov. (Fig. 5A–C; interaction of treatment with strata, GLMER; Chi2 = 7.3, d.f. = 2, p = 0.026). Occurrence in the lowest stratum did not differ between treatments (Fig. 5C) but was significantly lower with the iolinids in the two higher strata (Fig. 5A, B). The proportion of leaflets of the leaves that had mildew was also affected by the presence of the iolinid (interaction of treatment with time, GLMER; Chi2 = 31.5, d.f. = 1, p < 0.0001, treatment with stratum; Chi2 = 22.1, d.f. = 2, p < 0.0001). In the upper parts of the plant, almost no leaflets with powdery mildew lesions were found in the presence of the iolinid. In conclusion, the control of powdery mildew by cf. H. anconai sp. nov. was evident, limiting the spread of the disease to higher parts of the tomato plant.

Fig. 5
figure 5

Average (± SE) number of leaflets with powdery mildew lesions in three different plant strata: top A, middle B and bottom C of tomato plants in the presence and absence (no predator) of cf. H. anconai sp. nov. Differences between treatments were significant in the top and middle strata, as indicated with letters to the right of the last data points (contrasts through model simplification after GLMER)

Full size image

Discussion

We show that both the phytoseiid A. herbicolus and cf. H. anconai sp. nov. were capable of feeding and reproducing on tomato russet mites. However, A. herbicolus consumed approximately 7 times more adult russet mites than the iolinid in a leaf disc experiment. These findings align with other studies that reported low consumption rates of adult russet mites by H. anconai (Brodeur et al. 1997; Vervaet et al. 2022), but consumption of other stages occurred (Brodeur et al. 1997). In our study, both predatory mites showed higher oviposition rates when feeding on cattail pollen than on a diet of tomato russet mites (Fig. 1A, B). Providing pollen before pest introduction allowed the establishment of both mites on tomato plants. However, only the densities of the iolinid increased and effectively controlled russet mites (Fig. 2A), thereby compensating its lower individual predation rate compared to A. herbicolus. Moreover, our results indicate that cf. H. anconai sp. nov. was also effective in controlling the spread of powdery mildew (Figs. 3 and 5). The plants with iolinids were in the same area as those without these mites and as the spores of powdery mildew can easily travel through the air, the plants were all equally exposed to infections with this fungus. In conclusion, early releases of the iolinid mite with the addition of Typha pollen can prevent infestations by both russet mites and powdery mildew on tomato plants. This is the third iolinid known to exert dual control, parallel to the findings of Pijnakker et al. (2022b, 2023) concerning P. ubiquitus and H. anconai. Possibly, other iolinids might also have this capacity. We suggest that further research could investigate potential differences among iolinid species and genera and determine if feeding on plants, fungi and prey is a common characteristic of these mites. Feeding on fungal spores has previously been demonstrated in several tydeoid mites (Hessein and Perring 1986; English-Loeb et al. 1999; Norton et al. 2000; Duso et al. 2005). Hessein and Perring (1988) reported that adults of H. anconai survived when fed Cladosporium cladosporioides spores, but this diet did not allow for reproduction. Future research should investigate whether cf. H. anconai sp. nov. can develop on a diet of powdery mildew alone.

The glandular tomato trichomes provide a refuge for russet mites by protecting them against larger natural enemies, which are negatively affected by these trichomes (van Haren et al. 1987; Sato et al. 2011; van Houten et al. 2013; Riddick and Simmons 2014; Legarrea et al. 2022; Pijnakker et al. 2022a). This does not hold for iolinid mites, which are small enough to move unhindered under the glands of the trichomes where they can find and prey on russet mites (Pijnakker et al. 2022b). Pronematus ubiquitus, H. anconai and our new species cf. H. anconai sp. nov. appear to be well-adapted to tomato plants because they are small enough to walk in between the trichomes without touching the glandular cells, and can thus contribute to biological control of russet mites. We expected that other predators that naturally occur on tomato plants, such as A. herbicolus, would also be more adapted to these defences. Whereas Amblyseius herbicolus did initially establish on the plants (Fig. 2B), it subsequently vanished. Cardoso (2019) showed that A. herbicolus supplemented with large amounts of pollen could significantly reduce whitefly densities on tomato plants, with predator densities increasing on average up to eight times their release densities. Because A. herbicolus can feed and oviposit on tomato russet mites (Fig. 1A), we suggest that A. herbicolus disappeared in our experiment not because of lack of food, but due to the infection by powdery mildew and subsequent plant deterioration. Severe powdery mildew infestations lead to leaf chlorosis, which results in reduced plant quality (Jones et al. 2001). Perhaps the limitation of physical resources such as oviposition sites directly affected A. herbicolus, and better control of russet mites by A. herbicolus might have been achieved if the plants had been less affected by powdery mildew.

In our experiment on controlling tomato russet mites, we provided the predatory mites with pollen as alternative food, which is an increasingly common practice in biological control (Nomikou et al. 2002, 2010; van Rijn et al. 2002; Duso et al. 2004; Gonzalez-Fernandez et al. 2009; Delisle et al. 2015; Duarte et al. 2015; Leman and Messelink 2015). Here, it was done to allow the predators to establish before pest infestation, and this indeed resulted in a 20-fold increase of the densities of the iolinid during the four weeks after their introduction. Although the addition of alternative food may initially result in a lower predation rate on the pest (Nomikou et al. 2004; Vangansbeke et al. 2016), it often leads to better control after some time (van Rijn et al. 2002). The presence of alternative food such as pollen in a crop can enhance the efficacy of pest control by increasing the densities of natural enemies, which subsequently leads to a reduction in pest densities, a phenomenon similar to apparent competition (Holt 1977; van Rijn et al. 2002; Park et al. 2010; Delisle et al. 2015; Duarte et al. 2015; Leman and Messelink 2015; Kasap 2019). We show here that establishing cf. H. anconai sp. nov. with pollen on tomato plants before pest infestation indeed resulted in effective control: the density of russet mites per leaflet sampled in the presence of the iolinid was nearly zero (Fig. 2A). Hence, predator satiation and a possible preference of the iolinid for pollen over prey were eventually offset by an overall increase in predation at the population level. Despite A. herbicolus exhibiting an oviposition rate comparable to that of the iolinid and a higher predation rate, this did not result in better control on tomato plants.

It is also known that plant feeding by various omnivorous natural enemies can trigger the induction of plant defences (Pappas et al. 2015; Pérez-Hedo et al. 2015; Zhang et al. 2018). The enhanced control by cf. H. anconai sp. nov. is perhaps partly caused by the induction of plant defences by the iolinid, which deserves further testing.

Tomato plants with cf. H. anconai sp. nov. produced fewer, but heavier fruits than plants of the other treatments (Fig. 4A), although the total fruit weight did not differ among treatments. Some plant hormones involved in defences, such as jasmonic acid, also play a role in flowering (Wasternack et al. 2013). If the iolinid indeed induces plant defences, this may lead to a higher allocation of resources to vegetative growth or defences and a lower investment in reproduction (Herms and Mattson 1992). For example, a previous study showed that sweet pepper plants exposed to the omnivorous plant-feeding predator Macrolophus pygmaeus had fewer leaves and flowers than unexposed plants (Zhang et al. 2019). It is important to note that the tomato plants with the iolinid remained of good quality until the end of our experiment, unlike the plants of the other treatments. It is therefore highly likely that the plants with iolinids would continue to produce fruits for a longer period than the plants of the other treatments, and further studies are required to evaluate the life-time flower production and fruit setting of plants with and without the predator.

In conclusion, we propose that the early establishment of the omnivorous iolinid mite on tomato plants can effectively control two important tomato pests: the tomato russet mite and powdery mildew. Further studies are required to assess whether augmentative releases of the iolinid can maintain these pests below economic threshold densities in the field. Additionally, investigating the potential of the iolinid predatory mite as a biological control agent in other crop systems warrants further investigation. Lastly, the combined release of A. herbicolus and the iolinid could be a new biological control strategy for tomato: while the iolinid mite could target russet mites and powdery mildew, A. herbicolus could contribute to the reduction of A. lycopersici and could also control the whitefly B. tabaci (Cardoso 2019). However, the interaction between these two predators requires further investigation.

Author contributions

ÍM, RDV and AJ conceived the ideas and designed the methodology; IM and LSF collected the data; ÍM and AJ analysed the data; ÍM, AP, MMF and AJ led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.