Tetranychus evansi was obtained in 2002 from a population infesting tomato plants (Solanum lycopersicum, variety Santa Clara I-5300) in a greenhouse at the Federal University of Viçosa, Brazil. The population of T. urticae was obtained from infested tomato plants of the same variety as described above in a greenhouse at the Federal University of Viçosa. Both T. evansi and T. urticae were cultured on tomato plants (variety Santa Clara I-5300). Because the rate of oviposition varies with age in spider mites [28], young adult females of T. urticae and T. evansi of similar age were used in all the experiments. To obtain such cohorts, several adult females were allowed to lay eggs on detached uninfested tomato leaves on wet cotton wool, placed inside a tray. The adults were removed after 24 h and the mites hatching from the eggs were reared until adulthood.

Tomato seeds were sown in trays with a commercial substrate (Bioplant®, Bioplant Misturadora Agrícola LTDA, Nova Ponte, MG), composed of vermiculite and organic fertilizer and were kept in mite-proof screen cages in a greenhouse. Plants (21 days old) were transplanted to plastic pots (2L) that contained a mixture of soil, bovine manure (3∶1) and fertilizer (4-14-8 N-P-K). The plants were subsequently grown in mite-proof screen cages in a greenhouse until they were 45 days old and had at least four completely developed leaves. Subsequently, the plants were used either for the experiments or for spider-mite rearing.

To evaluate the effect of damage and web of T. evansi on the oviposition of T. urticae, discs (diam. = 1 cm) were made from leaves of clean tomato plants and subsequently infested with T. evansi females. The leaf discs were placed in groups of 10 in trays (20×10 cm) containing a piece of foam rubber covered with wet cotton wool and soaked with water to prevent the mites from escaping from the discs. Each disc received 10 mated T. evansi females, which were allowed to produce web and feed for a period of two days. Thereafter, all the females were carefully removed from the leaf discs, while taking care not to damage the web [30]. This resulted in discs that were damaged by feeding activities of T. evansi and entirely covered with a fine layer of web. The eggs and web were removed from about half of these discs, resulting in discs with damage by T. evansi. The eggs and web were left intact on the other half of the discs, and the eggs were counted. As a control, leaf discs from clean plants, without damage and web of T. evansi, were used. Subsequently, one 2-day-old mated female T. urticae was released on each disc, and the discs were incubated inside a climate box (28±2°C; relative humidity 70±10% and 12 hours of light) for a period of four days. Subsequently, the eggs produced by the T. urticae females were counted. Each replicate consisted of a group of 10 leaf discs and each treatment was replicated 11 times through time.

Likewise, we performed the reverse experiment, evaluating the effect of damage and web of T. urticae on oviposition by T. evansi. Leaf discs were treated as above, but with T. urticae instead of T. evansi, and the production of eggs by 2-day-old mated female T. evansi was measured on leaf discs that were previously damaged by T. urticae and contained eggs and web of this species, leaf discs that were previously damaged by T. urticae, but without eggs and web, and clean leaf discs. The rate of oviposition of T. urticae and T. evansi in each experiment was analysed using a linear mixed-effects models (lme) with treatment as fixed effect and replicate as a random factor (see Statistical Tables S1). Contrasts among treatments were calculated by aggregating non-significant treatment levels in an a posteriori stepwise procedure [31]. All analyses were done using R statistical software [32].

Tomato plants were infested with three 2-day-old mated females of T. evansi, T. urticae, or both, using a fine brush. There were three treatments, each replicated four times (i.e. four plants): plants infested with either species separately and plants infested with both species. When plants were simultaneously infested with both species, the two species were placed on different leaves. Hence, if T. evansi would down-regulate plant defence throughout the plant, we would expect T. urticae to profit from it and perform better on plants with T. evansi than if alone. Tetranychus urticae systemically induces direct plant defences, such as the increased activity of proteinase inhibitors [14], [16], and these defences may negatively affect T. evansi. Plants were maintained inside a climate room (28±2°C; relative humidity 70±10% and 12 hours of light) during the entire experimental period. On the plants, adult female mites of each of the two mite species were counted every 3 days until the plants died of overexploitation by the mites (after 33–35 days). The two species could easily be distinguished because adult females of T. evansi are red and the strain of T. urticae used here is white with two clearly visible dark spots. Log-transformed densities of spider mites were analyzed using linear mixed-effects models (lme), to correct for repeated measures with the presence/absence of the other species and time as factors and replicate as random factor.

In this experiment, we tested whether T. urticae was hindered by the web of T. evansi. We assessed whether the mites were capable of reaching the leaf surface when it was covered with web of T. evansi. Tetranychus urticae usually resides on the surface at the underside of leaves, where they spend 50% of their time feeding [33]. Leaflets of clean tomato plants were put upside down on moist filter paper, which in turn was put on a sponge inside a Petri dish (diam. 13.5 cm) filled with water. The edges and the petiole of the leaflet were covered with moist cotton wool to prevent mites from escaping and to maintain leaf turgor. One hundred adult female T. evansi were put on each leaflet and were incubated for 4 days (25±5C; 70±10%, 12∶12 L∶D photoperiod) to produce web. Subsequently, the web was gently removed from half of the leaflets with a fine brush. This treatment did not result in complete removal of the web, and occasionally one or a few mites were removed together with the web (0–15 per leaflet). A small strip of Parafilm® (3×6 mm) was put on top of the central vein, and 10 adult female T. urticae were released on this strip, from where they could descent onto the web or the leaf. Their position (on the strip, in the web, on the leaf surface or dead/escaped) was evaluated 1.5 h and 17.5 h after release. Moreover, we assessed which mites were feeding, judged by an inclined posture and the mouthparts (pedipalps) touching the leaf, after 17.5 h. In case of doubt, we checked for movements of fluid through the mite gut, which is a genuine sign of feeding. Ten replicates (leaflets) were done per treatment (with web removed or left intact). The data were analysed with a generalized linear model (glm) with a binomial error distribution and the presence or absence of web as fixed effect; mites that had died or escaped were omitted from the analysis.

In this experiment, we assessed the effect of cues associated with T. urticae on oviposition and web production by T. evansi. Prior to measuring the oviposition and webbing of T. evansi, plants were infested with either T. evansi or T. urticae by putting four small, infested tomato leaflets on each plant, inside a mite-proof screen cage in a greenhouse. As a control, a group of plants of the same age was left clean. The plants were incubated in this way for one week, resulting in 300–400 adult mites per plant with mites. Subsequently, the web, mites and eggs were removed from leaflets using a fine brush, and five leaf discs (Ø = 1 cm) were punched from the leaflets of each plant. Seven replicates were done in blocks through time, with one plant for each treatment per block. The leaf discs were kept in groups of 15 in plastic trays (30×20 cm) containing a piece of foam rubber covered with wet cotton wool and filled with water to prevent the mites from escaping from the discs. One 2-day-old mated female T. evansi from a cohort was individually placed on each leaf disc. Four days later, the eggs produced by the mites were counted and the web produced was measured with a technique adapted from [34] [see 35]. Soil particles (0.177 mm) were sprinkled over a leaf disc with web, using a fine brush. Particles were counted and their position (in the web or on the leaf) scored using a stereo microscope (32× magnification). Web density was calculated as the percentage of particles that was found suspended in the web [35].

Rates of oviposition rates and web densities were averaged per plant and these averages were analysed with linear mixed effects models (lme) with treatment (damage by T. evansi, by T. urticae or no damage) as fixed effect and replicate as random factor. Contrasts among treatments were calculated as above.

Here, we evaluated the effect of cues emanating from leaves infested with T. evansi or T. urticae, as well as from clean leaves (control) on the production of eggs and web by T. evansi females. A tomato leaflet infested with 50 T. evansi or 50 T. urticae females or a clean leaflet was placed in the centre of a Petri dish (diam. = 10 cm) containing wet cotton wool and this leaflet served as a source for cues. Subsequently, a group of 6 clean leaf discs (Ø = 1 cm) were placed in a circle around the leaflet, and a mated T. evansi female, 2-day-old since her last moult, was placed on each leaf disc. Thus, these leaf discs received volatile cues as well as water-soluble cues from the leaflet in the centre. Each replicate consisted of a group of 6 leaf discs in a Petri dish. Each Petri dish (Ø = 10 cm) was placed inside a plastic tray that was sealed with PVC film and kept inside a climate room (28±2°C; relative humidity 70±10% and 12 hours of light). Each treatment was replicated 23 times (number of Petri dishes).

Because not all experiments were carried out simultaneously, and plants may differ with season and season-specific greenhouse conditions, we carried out the experiment in blocks through time, each block consisting of an equal number of each treatment. The results of the different treatments should therefore be compared to their respective controls, carried out at the same time. The oviposition rate and relative web density were analysed using linear mixed-effects models (lme) with replicate as random factor. Contrasts among treatments were calculated as above.

There was a significant effect of treatments on the rate of oviposition of T. urticae (lme, Log likelihood ratio (L-ratio) = 185.8, d.f. = 2, p<0.0001, Fig. 1a); the oviposition of T. urticae was highest on leaf discs with damage but without web of T. evansi, intermediate on leaf discs with damage and webbing of T. evansi, and lowest on clean leaf discs (Fig. 1a, Log likelihood ratio (L-ratio), all p's<0.0001). The rate of oviposition of T. evansi was also affected by treatment (lme, L-ratio = 16.7, d.f. = 2, p = 0.0002). It was significantly higher on clean leaf discs than on discs previously damaged by T. urticae, either with or without web (Fig. 1b, L-ratio = 14.2 and 11.4 respectively, p's<0.001). There was no effect of the presence of web produced by T. urticae on oviposition of T. evansi (L-ratio = 0.043, p = 0.84). These results show that T. urticae can profit from the reduced direct defence of plants attacked by T. evansi, but was negatively affected by T. evansi web. Furthermore, T. evansi was not hindered by the web of T. urticae, but was negatively affected by the defence induced by T. urticae.

Numbers of T. urticae were significantly lower on plants where they were together with T. evansi than on plants free of T. evansi (lme, F_1,6 = 160.7, p<0.0001, Fig. 2a). The population of T. urticae showed a slight increase in the first two weeks, but then declined and went extinct during the last 3 weeks of the experiments. In contrast, there was no significant difference between the population size of T. evansi on plants where it was together with T. urticae and the plants where it was alone (lme, F_1,6 = 3.53, p = 0.108, Fig. 2b). The population of T. evansi increased over the 5 weeks of the experiments (Fig. 2b), and reached much higher levels than the population of T. urticae, even when the latter was alone on a plant (Fig. 2).

Although we tried to carefully remove the web from half of the leaflets, there was invariably some web left, and the mites furthermore continued producing web during the experiment. Therefore, we found mites in the web in both treatments, but fewer in the treatment without web (55% with web vs 17% with web removed after 1.5 h and 57.4% vs 12.5% after 17.5 h). After 1.5 and 17.5 h, a significantly lower proportion of adult females were found on the leaf surface when web was not removed (Fig. 3: glm: 1.5 h: Chi² = 4.42, d.f. = 1, p = 0.035; 17.5 h: Chi² = 10.08, d.f. = 1, p = 0.0015). When most of the web was removed, the proportion of mites on the leaf surface increased over time (Fig. 3). This did not occur when the web was not removed. Significantly more mites were feeding when web was removed than when the web was left intact (Fig. 3: glm: Chi² = 10.10, d.f. = 1, p = 0.0015).

The oviposition of T. evansi was significantly affected by previous damage by T. urticae or by conspecifics (Fig. 4a, lme: L-ratio = 43.7, d.f. = 2, p<0.0001). As reported before [16], the oviposition rate was significantly higher on tomato leaf discs that previously received damage by conspecifics than on leaves that were previously damaged by T. urticae or on clean discs (Fig. 4a. L-ratio = 47.2 and 28.1 respectively, p's<0.0001). As found above, the oviposition rate of T. evansi was significantly lower on leaf discs that were previously damaged by T. urticae than on clean leaf discs (L-ratio = 20.2, p<0.0001) (cf. Fig. 2b).

There was a significant effect of treatment on the web production by T. evansi (lme: Fig. 4b, L-ratio = 23.5, d.f. = 2, p<0.0001). The relative web density was a factor 3 higher on leaf discs from plants that were previously damaged by T. urticae than on leaf discs from plants that had received damage by T. evansi (L-ratio = 23.4, p<0.0001), and almost a factor 2 higher on clean discs than on discs previously damaged by T. evansi (L-ratio = 6.38, p = 0.012).

The rate of oviposition of T. evansi was not affected by cues from leaflets infested with T. urticae (lme, L-ratio = 0.05, d.f. = 1, p = 0.82, Fig. 5a), or with conspecific mites (lme, L-ratio = 0.006, d.f. = 1, p = 0.94, Fig. 5b). However, there was a significant effect of cues from leaflets infested by T. urticae on web production by T. evansi (lme, L-ratio = 5.0, d.f. = 1, p = 0.025, Fig. 5c), but not of cues from leaflets infested with conspecifics (lme, L-ratio = 0.15, d.f. = 1, p = 0.7, Fig. 5d).