Adult mites of O. ilicis were collected from leaves o pesticide-free coffee plants (C. arabica) in the municipality of Lavras, state of Minas Gerais, Brazil, taken to the laboratory and reared under controlled conditions at 25˚C ± 2˚C of temperature, 70% ± 10% of RH, and 14 h of photophase. The mites were reared on detached coffee leaf placed on top of a foam pad of 1 cm thickness, which occupied the entire bottom of a petri dish of 15 cm diameter, without a lid. The foam was constantly kept moist with distilled water, and a strip of hydrophilic cotton wool was placed around the leaf to maintain turgidity and act as a physical barrier to prevent the mites from escaping of the arena [20] [21] .

The predatory mites, A. herbicolus and E. alatus, were collected in the same pesticide-free coffee plants as O. ilicis and, in the lab were kept in 9.5 cm diameter arenas, made of flexible black polyvinyl chloride (PVC), floating on distilled water in 15 cm diameter Petri dishes, without lid [22] . These arenas were maintained under the same rearing laboratory conditions as described previously for O. ilicis breeding. Predatory mites were fed on pollen of castor bean plant (Ricinus communis L., Euphorbiaceae) and O. ilicis from the lab rearing.

Thirty coffee plants (C. arabica) of approximately 100 cm high, were kept in pots with 20 liters’ capacity, in a greenhouse covered with transparent plastic and shading material of 50% sun protection factor, with fine mesh in all the sides, promoting an ambient temperature of 25˚C ± 5˚C.

For the infestation of coffee plants with pest mite, 40 O. ilicis adult females from laboratory rearing were transferred, with aid of a fine-tipped paint brush, to the leaves located in the middle part of each plant, totaling 1,200 adult females.

Twenty days after the infestation with O. ilicis, the predatory mites A. herbicolus and E. alatus were released in a separate experiment for each predatory mite species. There were a total of six treatments for each predatory mite species, each treatment representing an amount of the predatory mite releases set as follow: 0 (control), 2, 4, 6, 8, and 10 predatory mites per coffee plant.

The experimental design was completely randomized, and each treatments being replicated five times.

At 7, 15, 21 and 28 days after predatory mites’ release, a leaf of the central part of each plant was detached, totalizing 30 leaves/coffee plant, and taken to the laboratory. The number of all O. ilicis mobile phases―larvae, protonymphs, deutonymphs and adults―as well as the number of laid eggs of the pest mite on the leaf surface was recorded under a 40× binocular stereomicroscope.

Due to the high mobility of the phytoseiid predatory mites on the plants, it was impossible to determine the total number of them, for this reason they were not count.

Given the specifications of the experiment mentioned above, the number of the O. ilicis mobile phases and eggs laid by them on the surface of the leaf were analyzed using the methodology of generalized linear models (GLM) [23] .

It was proposed the Poisson regression model and was considered a logarithmic linkage function, with a systematic component given by η i = β 0 + β 1 x i + β 2 z i , with x i the quantity of predators released and z i a categorical variable related to A. herbicolus ( z i = 0 ) and E. alatus ( z i = 1 ) species. Parameters β j , when j = 0, 1, 2, represented the effects associated with each variable and the log-linear linkage function log μ i = η i , being that μ i represented the mean of the response variable y_i in the quantity of predators released x i of species z_i; i.e. the model that linked the mean number of preys with variables x’s and z’s is given by Equation (1), considering ε_i the experimental error associated with the ith portion.

μ i = e β 0 + β 1 x i + β 2 z i + ε i (1)

The regression model with a negative binominal response was used, upon occurrence of over dispersed or overlapping responses. This model is the same as Equation (1) and the parameters estimation is similar to what was explained previously. In this case, an additional parameter to the Poisson model was considered and estimated, called dispersion parameter and denoted by ϕ [24] .

Percentage rates denoted as ψ , were calculated with the objective of measuring the effect of a treatment in relation to the control group; considering the reference model obtained when z = 0 and the model for a given z. The percentage rates, or the odds ratio, between the models is given by Equation (2) [25] .

ψ = e β j (2)

Therefore, when the species E. alatus was considered, Equation (2) represented a percentage change in the expected number of the prey mite O. ilicis that were present on the plant. If β j > 0 , an increase in the expected number was entailed and if β j < 0 a decrease in the expected number was entailed; and in this case, it is common to demonstrate Equation (2) as 1 − ψ [23] .

In view of the above, the parameters of the models were estimated with the use of the maximum likelihood method [24] , and the validation was carried out through the envelope chart, in order to verify the negative binominal regression model suitability. In addition, under certain conditions of regularity and if the current model describes the data well, the deviance has an asymptotically and approximately chi-square distribution with n − p degrees of freedom, being n the number of observations and p the quantity of parameters that were considered in the model [26] .

All the results of the statistical analyses were obtained using the R statistical software [27] .

Considering the total number of O. ilicis mobile phases―larvae, protonymphs, deutonymphs and adults―that were alive, it was observed that the species E. alatus provided an increase of approximately 13.7% in the expected number of alive O. ilicis in relation to the predatory mite A. herbicolus; however, this is not a significant difference, considering a 5% significance level. Therefore, E. alatus showed a similar predation rate to A. herbicolus (Table 1).

When analyzing the number of predatory mites that were released throughout the experiment, it was observed that as the number of predatory mites increased there was a 16.8% decrease in the expected number of alive O. ilicis mobile phases in relation to the previous number (Table 1).

The adjusted model for both predatory mite species, A. herbicolus and E. alatus, considering the number of adults that were released, is shown in Figure 3. A similar behavior between the two predatory mite species was observed, and as the number of predatory mites increases the similarity between the numbers of alive mobile phases of O. ilicis is more evident (Figure 3).

For both predatory mites, the release of 10 mites per plant provided a more efficient control of O. ilicis in coffee plants (Figure 3).

Findings of this study are similar to those already reported in the literature which performed a study with different ratios between the number of Tetranychus urticae Koch (Tetranychidae) and Neoseiulus californicus (McGregor) (Phytoseiidae) on strawberry (Fragaria spp., Rosaceae) and concluded that at ratios of 1:10 (predatory mite/pest) it prevented the pest mite from reaching the economic damage level, i.e. 50 T. urticae/leaflet [28] .

It has also been observed that N. californicus, when released at ratios of up to 1:10 (predatory mite/pest) was effective in controlling T. urticae too on strawberry too, both in the field and greenhouse, maintaining the pest mite population at low levels for long period [29] .

Considering the total number of O. ilicis laid eggs, it was observed that the species E. alatus caused a reduction of approximately 0.25% in the expected number of O. ilicis eggs in relation to the predatory mite A. herbicolus, however this difference is not significant, considering a 5% significance level. So, E. alatus presented a similar predation rate to A. herbicolus when the number of O. ilicis laid eggs was also considered (Table 2).

The adjusted model for both predatory mite species in relation to the number of O. ilicis laid eggs, considering the number of predatory mites released, is shown in Figure 4, where is observed an almost identical behavior between both studied species, A. herbicolus and E. alatus.

Therefore, the results showed that both predatory mites species, A. herbicolus and E. alatus, are efficient in reducing the O. ilicis mobile phases―larvae, nymph (protonymphs and deutonymphs), and adults―of the southern red mite, in the greenhouse coffee growing (Figure 3). These results are also confirmed by the reduction of the number of eggs that were found in the leaves, that is, reduction of the number of eggs placed as a consequence of the reduction of the pest mite number by predation, in addition to the direct predation of the eggs (Figure 4).

This results are similar to those obtained in laboratory studies on the predation potential of the phytoseiid A. herbicolus, and other predatory mite species, over several developmental phases of O. ilicis, were those showed also that the adult female phase of the predatory mites is the most efficient in predation, followed by the phases of nymph (protonymphs and deutonymphs), adult male, and larvae, respectively [16] [30] .

Considering the phases of O. ilicis altogether, webbing reduced the predation potential of I. zuluagai and E. citrifolius predatory mites, but not of A. herbicolus [17] studied in this work.

It has been demonstrated also under laboratory conditions that E. alatus adult male is an efficient predatory mite of O. ilicis in all its developmental phases [18] , result also found in this study in the greenhouse coffee growing.

Initial infestations of 15 - 120 females of O. ilicis per coffee leaf can result in a significant decrease in photosynthesis, causing substantial damage to coffee crops [10] . As the present result indicates, E. alatus and A. herbicolus would be capable of controlling the population of O. ilicis before it would reach such damaging level, what show the importance of conservation and augmentation of this biological control agent in the coffee growing field.

Releases of the predatory mites A. herbicolus and E. alatus efficiently reduced O. ilicis populations on greenhouse-grown and the strategy of the conservative biological control in the field coffee growing is also essential for implementing the southern red mite integrated management tactic, since this species is of natural occurrence in coffee plants, as already reported in the literature [13] [14] [15] . In addition, considering the main control method adopted by the coffee producers, i.e., the use of plant protection products, those products that present some type of selectivity might be prioritized, which would maintain predatory mites that are natural enemies of the southern red mite.

Among the most successful biocontrol programs in the world are those of mite control, implemented in plants grown under greenhouse or field. Brazil is starting in this practice of control, but still have limited and demand for information on the interactions of predatory mite/prey in different production systems are necessary.

Because they are of naturally occurrence in coffee plants, the strategy of conservation and increase of predatory mites is essential to the tactics of integrated management of the southern red mite. If necessary the use of plant protection chemical products should be prioritized those that show some type of selectivity, thus preserving the predatory mites, natural enemies of pest mites.