The pesticides used in this work are listed in Table 1, reporting their trade names, formulations, suppliers, active ingredients, chemical families, field rates and modes of action. All tested insecticides were stocked, prepared and applied according to manufacturer guidelines. The concentration tested was the maximal recommended by the manufacturer for citrus. (Table 1).

Medflies and parasitoids were obtained from laboratory colonies maintained in the facilities of the High Agronomic Institute of Chott-Mariem (ISA-CM), Sousse (Tunisia). Both insects, D. longicaudata and C. capitata , were reared under controlled conditions (25±2ºC, 65±10% RH and 16:8 h L:D) in climatic chambers (constant climate chamber in-house built, Memmert® GmbH, D-91126 Schwabach, Germany).

Ceratitis capitata colony was established in 2012 from infested citrus fruits (collected at Chott-Mariem, Tunisia). Adults are maintained in plastic cages (40 × 30 × 30 cm) with fine mesh framed lateral sides to allow oviposition and egg collection according to the rearing procedure described by Sabater-Muñoz et al. (2009) and Martins et al. (2010) . Adults were provided with a d libitum water, sugar and a mixture 1:4 of yeast hydrolysate: household sugar. Larvae of the medfly were fed on artificial diet based on wheat bran yeast and sugar ( Sabater-Muñoz et al., 2009 ; Martins et al., 2010 ).

Diachasmimorpha longicaudata colony was also establishedin 2012 with imported parasitized pupae from the Instituto Valenciano de Investigaciones Agrarias (IVIA) research station (Valencia, Spain). Adults were maintained in plastic cages (40 × 30 × 30 cm), similar to those of medfly, with approximately 4,000 to 5,000 females per cage, and provisioned with ad libitum water, honey and household sugar. Medfly third instar larvae (within its artificial diet) were exposed daily to the parasitism of D. longicaudata females through a fine mesh framed window located on the upper side of the rearing cages. Exposed larvae were allowed to develop in separate cages until the emergence of the new parasitoid generation approximately two weeks later. A cohort of 6-8-d-old parasitoids (≈1:1 sex ratio female:male) were established from this rearing colony for each treatment.

To evaluate the residual toxicity of pesticides in D. longicaudata , a laboratory method was used ( Contreras et al., 2005 ). Briefly described, sour orange ( Citrus × aurantium ) fully expanded young leaves (approx. 10 days after flush fully expanded) were collected from 10-yr-old ornamental untreated trees located at ISA-CM (Chott-Mariem, Tunisia) and delicately brushed then rinsed in distilled water to remove dust, bird feces or other accidently present arthropods, prior treatment. Clean leaves were treated with selected pesticides (Table 1) by the leaf-dip method ( Immaraju et al., 1990 ). Distilled water was used as control throughout the bioassay. Treated leaves were placed individually in experimental units (isolators) in the laboratory to allow aging of pesticides for 1 hour (fresh residue), 3, 7, 14, 21 or 28-d under ambient conditions, including light exposure (natural photoperiod) to simulate open field conditions. Ambient conditions including sunlight exposure (what we call field conditions) have been stated to decrease the activity and cause degradation (by sun fotobleaching) of several insecticide residues, which are not taken up by tree leaves, as abamectin and spinosad ( Demchak & Dybas, 1997 ; Urbaneja et al., 2009 ). Isolator consisted on two superposed plastic glasses (600 mL and 100 mL). The top glass had a central hole on its bottom to allow citrus leaf petiole to reach the water present in the bottom glass allowing to keep leaf turgidity and metabolic actions during the pesticide aging process. A fine mesh cloth was fixed on the upper opening of the top glass to allow ventilation. This cup system was previously validated in similar side-effects studies by Zappalà et al. (2012) and Biondi et al. (2013) .

The bioassays were conducted under controlled environmental conditions (25±2°C, 60±10% RH, 16:8 h L:D) in a climatic chamber. Bioassay arena consisted in a polypropylene transparent box (14 × 10 × 9 cm) with a mesh covered aeration window (10 cm ² ) in the lid. Treated leaves were transferred individually to bioassay arena, replacing the 100 mL water container by a 1.5 mL micro-centrifuge vial attached with modeling clay to the bottom of bioassay arena. A cohort of 40 6-8 days-old D. longicaudata adults (≈ 1:1 females: males ratio) was used in each replica, being provided with a honey solution (in a cylindrical 0.5 × 1.5 cm (diameter × length) dental cotton roll) as food source. Residues of the selected pesticides were assayed at 1h (fresh), 3, 7, 14, 21 or 28 days-old post application in a sequential way ( i.e. aged residues, which do not show significant differences with control, were not further continued). Each treatment was replicated three times, assessing mortality after three days of exposure to treated leaves. Parasitoid specimens were considered as dead if no response was observed under stereomicroscope after being stimulated with the tip of a soft paintbrush.

Data were first tested for normality and homogeneity of variance using Shapiro-Wilk test ( SPSS, 2011 ). The mortality was then compared using one-way analysis of variance (ANOVA) at p <0.05. Means were then separated using the LSD post hoc test. When significant differences were detected between the control and the pesticides, the mortality values were corrected using Abbot’s formula ( Abbott, 1925 ). Then, corrected mortality percentages were used to rank insecticides and their residues according to the IOBC Working group "Pesticides and Beneficial Organisms" standards toxicity classes ( Sterk et al., 1999 ) as follows: (1) harmless, mortality <30%; (2) slightly harmful, 30-79%; (3) moderately harmful, 80-99%; and (4) harmful, mortality >99%.

Malathion and methidathion belong to the same organophosphate chemical family but with differentiated mode of action (Table 1). These insecticides have been recently banned from the European Union ( OJEU, 2015 ) by their human health concerns ( Flessel et al., 1993 ; Marty et al., 1994 ) and their harmful effects on non-target arthropods ( Ehler & Endicott, 1984 ; Urbaneja et al., 2009 ) which lead in some cases to outbreaks of secondary agricultural pests that were under control of natural enemies. Despite this banning, Malathion is still used in the Mediterranean area [including some EU members by some considerations in the OJEU (2015) ], including Tunisia, as efficient pesticide against C. capitata . However, in some of these countries, it has been reported the presence of malathion resistant medfly populations after an increase of the number of treatments by the citrus growers ( Magaña et al., 2007 ; Vontas et al., 2011 ). These results for C. capitata along with the ones presented here, which classifies Malathion as IOBC class 4 even at 7-d-post-treatment to D. longicaudata , allowed us to ask for a complete removal from the Tunisian IPM programme against the medfly. Concerning methidathion, in Tunisia it is used against scales and whiteflies, which to date are not showing any resistance. But, due to the human health and environmental concerns ( Flessel et al., 1993 ; Marty et al., 1994 ), along with the residual toxicity for D. longicaudata determined in this work, our proposal is similar to those of malathion, just to be replaced by other less harmful substances.

Acetamiprid belongs to the neonicotinoids pesticide family and have been considered as rational alternative to the organophosphates due to their high specificity, elevated efficacy and relatively low toxicity to the environment ( Tomizawa & Casida, 2005 ). Neonicotinoids are widely used in Tunisian citrus orchards to control hemipteran pests considering their ovicidal and larvicidal activities and systemic action. However, in the past 3-5 years, eco-toxicological studies revealed a wide range of adverse side effects on non-target arthropods, including the worldwide-threatened honey bee Apis mellifera L. (Hymenoptera: Apidae), which remains the model organism against the use of neonicotinoids ( Laurino et al., 2011 ). Regarding tephritid fruit fly parasitoids, neonicotinoids have been previously tested, being the active ingredient imidacloprid, not the acetamiprid used in this work. Liburd et al. (2004) determined lethal effects of imidacloprid on D. longicaudata when in use in treated spheres for the management of key fruit fly pests. Whereas Adán et al. (2011) assessed lethal and sublethal toxicity of imidacloprid on Psyttalia concolor Szépligeti (Hymenoptera: Braconidae), depending on the application mode (cover sprays become lethal whereas bait sprays remained sublethal). So taking into consideration our results, we can confirm that neonicotinoids should be used with caution when beneficial hymenopterans are present in the agro-ecosystem.

Whilst the first belongs to the pyrethroids class, the second belongs to the neonicotinoids class, and many of the effects detected can be attributed to the lethal effects of the neonicotinoid as the observed in past works. But neonicotinoids can have either a nitro group or a cyano. Those that have the cyano group, as thiacloprid, show relative lower toxicity attributed to different receptors, metabolism and secondary metabolites production ( Suchail et al., 2004 ; Jones et al., 2006 ). Despite this lower toxicity, some beneficial arthropods exhibit sublethal side effects, like flight-navigation problems, reduction in attack rate, increase of handling time or even reduced emergence success or sex ratio distortion, factors that decrease the chance for parasitoid establishment in new territories ( Krepsi et al., 1991 ; Garcia et al., 2009 ; Carmo et al., 2010 ; Wang et al., 2012a , b ; Fischer et al., 2014 ). In our study, although determined an initial acute toxicity, its effect on D. longicaudata significantly decreased with time (Tables 2 and 3), rendering it as compatible within the IPM for citrus pests. Despite this, more research is needed especially regarding long-term toxicity and sublethal side effects for D. longicaudata not done within this work.

Abamectin, a mixture of avermectins, is a natural insecticidal, acaricidal and nematicidal compound derived from the bacteria Streptomyces avermitilis . This agricultural compound was approved as a plant protection agent and as a veterinary drug for control of endo- and ecto-parasites ( FAO, 1996 ). Spinosad, a bacterial insecticide derived from the actinomycete Saccharopolyspora spinosa , shares some relations with abamectin.

Even if we have classified both as slightly harmful for D. longicaudata (IOBC class 2) as in other risk assessment studies ( Stark et al., 2004 ), opinions are still divergent about their compatibility with many biocontrol agents of other plant pests, especially parasitoids with emphasis on its possible trans-generational sublethal side effects ( Bueno & Freitas, 2004 ; Biondi et al., 2012b , 2013 ; Costa et al., 2014 ; Abbes et al., 2015 ). As for the precedent substance, more research is required to determine the sublethal side effects on D. longicaudata .

Azadirachtin, a limonoid tetranor-triterpenoid chemical derived from neem tree ( Azadiracta indica ), has been extensively used in other countries for the control of phytophagous ( Stark et al., 1992 ; Stara et al., 2011 ; Alvarenga et al., 2012 ) and livestock arthropod ( Ruiu et al., 2008 ) pests by its environmental compatibility and extremely low acute mammalian toxicity ( Schmutterer, 1990 and references herein). Our results indicate that this was the less harmful pesticide for D. longicaudata , even as fresh residue, as similarly determined by Stark et al. (1992) . However, its application form and dose could be the responsible of the differential results observed by other authors. When D. longicaudata hosts were fed with neem seed cake, with a high content in azadirachtin, the parasitoid emergence rate was significantly lower than the control not feed with this neem cake ( Alvarenga et al., 2012 ). Whereas D. longicaudata survival and emergence rate did not differ from control when hosts (either Dacus dorsalis Hendel (Diptera: Tephritidae) or C. capitata ) were subjected to azadirachtin sprays at doses that inhibits host emergence ( Stark et al., 1992 ).