The olive tree ( Olea europaea L.) is one of the main crops in the Mediterranean basin, where the region alone produces 91.2% and consumes 72% of the world's olive oils (International Olive Oil Council, http://www.internationaloliveoil.org , data updated to 2015/2016 crop year). Losses up to 80% of the oil value and 100% of some table cultivars are caused by the obligate olive key pest, the olive fruit fly Bactrocera oleae (Rossi, 1790) (Diptera: Tephritidae) ( Daane & Johnson, 2010 ; Malheiro et al. , 2015 ). The pest indirectly damages the crop by ovipositing inside the fruit and feeding upon the pulp, until reaching the pupal stage. Historically confined to the Mediterranean Basin, B. oleae has spread in almost every country where olive is cultivated for commercial purposes ( Augustinos et al. , 2002 ). In addition, a predicted 2°C global warming in the Mediterranean Basin between 2030 and 2060 ( Giannakopoulos et al. , 2009 ) is alerting bioeconomics and producers, as interactions between olive and B. oleae are expected to be enhanced ( Ponti et al. , 2014 ). For the conservation of olive orchard agroecosystem, economy and cultural heritage, holistic strategies are urgently needed ( Ponti et al. , 2016 ; Marchini et al. , 2017 ).

Conventional control strategies against B. oleae have proved to have a negative impact both on consumers' health and non-target species ( Vickerman & Sunderland, 1977 ; Amvrazi & Albanis, 2009 ), not to mention the development of resistance by the pest to frequently used insecticides ( Daane & Johnson, 2010 ; Pavlidi et al. , 2018 ). Initial inroads have been made in understanding the role of resident natural enemies of B. oleae . Indeed, the assessment of pest control ecosystem service would be of utmost importance in the design of environmental-friendly management plans. Literature on conservation biological control of B. oleae has been addressing at the interactions between potential soil predators and pre-imaginal stages of the pest ( Orsini et al. , 2007 ; Santos et al. , 2007 ; Dinis et al. , 2016 ). The third instar larvae of the overwintering generation leave the drupe to pupate in the first centimeters of the soil ( Dimou et al. , 2003 ). Here, they spend several months, being thus exposed to a guild of soil predators and abiotic factors, until they emerge as adults ( Bateman, 1972 ; Cavalloro & Delrio, 1975 ).

DNA-based techniques have been providing valua­ble results in unraveling these interactions between the pest and its natural enemies, mostly when direct obser­va­tion of predation in the field are impractical ( González-Chang et al. , 2016 ). Polymerase chain reaction (PCR)-based analysis of unique prey DNA sequences has proven to be a highly sensitive, specific and cost-effective tool for detecting prey's remains in the predator's gut ( Symondson, 2002 ). Several factors affect the probability of detecting prey DNA sequences, such as the identity and physiology of the predator, the collection method, the preservation of the specimens and the choice of the molecular markers, and some methodological issues still need to be optimized ( Juen & Traugott, 2006 ; Weber & Lundgren, 2009 ). Since the likelihood of detecting the prey target sequences decreases with post feeding time, one of the main ecological issues is inspecting which is the maximum time that prey consumption can be detected, in order not to underestimate predation ( Greenstone et al. , 2007 ). While in laboratory trials the post feeding time can be controlled, this issue is particularly relevant in field analyses, where predators are collected after an unknown post feeding time. In addition, predator identity ( e.g. taxonomy identity and sex) may affect post feeding prey DNA detection intervals ( Zaidi et al. , 1999 ; Sheppard et al. , 2005 ). For a given specific prey, calibratory feeding trials are therefore required to assess the robustness and versatility of the PCR-based diagnostic assay for a broad range of predators ( King et al. , 2008 ; Aebi et al. , 2011 ).

Despite the accumulating evidence that carabid beetles (Coleoptera: Carabidae) are a dominant compo­nent in the predatory fauna of olive orchard ( e.g. , Lasinio & Zapparoli, 1993 ; Gonçalves & Pereira, 2012 ), a better understanding of the chance to molecularly detect the olive pest in adult carabid's gut is required. Recently, Rejili et al. (2016) designed B. oleae specific primers based on mtDNA sequences of cytochrome oxidase subunit I ( COI ) gene to successfully detect the prey consumption by the carabid Pterostichus globosus (Fabricius, 1792) up to 16 h after prey's ingestion, in laboratory conditions. To validate the versatility of the above PCR-based diagnostic assay, we tested Pseudoophonus rufipes (= Harpalus rufipes ) (De Geer, 1774), a medium-size (11-16 mm length) carabid species, generalist feeder, with a wide (Palearctic) distribution, preferring open habitats ( Thiele, 1977 ; Honek et al. , 2013 ). It is among the most active species in arable lands ( Lövei & Sárospataki, 1990 ; Thomas et al. , 1997 ; Avgin & Luff, 2009 ) being mentioned as a biocontrol agent of several pests ( Kromp, 1999 ; Sunderland, 2002 ; Boreau de Roincé et al. , 2012 ). Its life cycle has been detailed described ( Matalin, 1997 ) and the species has showed peaks of activity during the period of highest abundance of B. oleae pupae ( Albertini et al. , 2017 ). In this study, we tested the post feeding detection of B. oleae within P. rufipes gut up to 20 h in both female and male predators, to assess if B. oleae COI fragments were still detectable in the carabids' gut. Finally, we extracted prey's DNA both from dissected and not dissected carabid specimens, aiming at improving Rejili et al. (2016) protocol in terms of efficiency and user-friendly features.

Specifically, this study was designed to compare: i) B. oleae mtDNA detectability at different post feeding time intervals; ii) carabid female and male differences in prey detection time; iii) the sensitivity of two different pair primers designed to amplify two fragments of mtDNA COI gene of B. oleae and iv) the efficiency of the use of two different sample types for extracting DNA ("gut" vs "whole body" samples).

The PCR-based assay optimization here presented may allow accurate analyses by means of molecular tools, prompting the development of quantitative measurements of predation in-field by the natural enemies of B. oleae pre-imaginal stages.

A specific amplification with the predicted products of B. oleae COI gene fragments (108 and 214 bp) was generated by the use of both pairs of primers up to 20 h after prey ingestion for both female and male predators (Fig. 1). No amplification was revealed when DNA extracted from carabids fed with C. capitata or starved carabids were used (Fig. 2).

For all the post feeding treatments, 160 PCRs were carried out in total, 146 of which showed specific amplification. For all sample types, 90.36% showed positive amplifications with the first pair primers (SBo1-F/SBo1-R, 83 PCRs in total) and 92.21% with the second pair (SBo2-F/SBo1-R, 77 PCRs in total). No significant differences were found when compared the overall success of the two pair primers and when compared short, medium and large time intervals (Table 1).

Female carabids obtained 89.29% of positive amplifications, male carabids 93.42% and no statis­tically significant differences were found between sexes (Table 1). Considering "gut" and "whole body" samples, 89.84% and 96.86%, respectively, responded positively to B. oleae detection, differences being again not statistically significant (Table 1).

For the "gut trial" samples (128 PCRs), B. oleae DNA fragments were detected in all the post feeding time treatments, ranging from 79.41% to 100% of positive amplifications. Higher rates were found for 2, 4, 6 and 8h post feeding (from 83.33% to 100%). Nine-15 and 16-20 h treatments had, respectively, 79.41% and 89.29% of successful amplifications. In comparing the total percentage of successful amplifications using the two pair primers, the first pair had 87.5% of positive results and the second 92.19% (Fig. 3). Females had 88.06% of successful amplifications and males 91.80%. Comparisons among the efficiency of the two pair primers and any differences among sexes revealed no statistically significant differences (Table 1).

For the "whole body trial" samples (32 PCRs), the per­centage of positive amplifications was 100% for all post feeding time treatments, except for 16-20 h treatment where there was 80% of success (4 positive samples out of 5 assays). The first pair of primers gave 100% of positive results, the second pair 92.31% (12 positive samples out of 13 assays) (Fig. 3). Females had 94.12% of positive results and males 100%. Comparisons among the efficiency of the two pair primers and among sexes revealed no statistically significant differences (Table 1).

When comparing the quantity and quality of the extracted DNA, the average value was 457.90 (SE=31.80, n=58) μg/μL and the ratio 260/280 was 2.11 (SE=0.02, n=58) for the "gut trial" while for the "whole body trial" values were 586.74 (SE=84.81, n=12) μg/μL and 2.10 (SE=0.03, n=12) respectively. Differences were statistically not significant for DNA quantity ( p =0.10) nor quality ( p =0.37).

Our laboratory study has demonstrated that it is possible to detect B. oleae consumption in carabids' gut at least up to 20 h since pupa ingestion, for both female and male predators that consumed one B. oleae pupa only. There are new elements compared to Rejili et al. (2016) , which confirm the validity of their protocol and prove its versatility and sensitiveness. There is a general increasing interest in the tested species P. rufipes , which has already been proposed both as pest control agent and useful seed consumer ( Wallinger et al. , 2015 ). Since it is also active in olive orchards when pest pupae are more abundant in the soil ( Albertini et al. , 2017 ), P. rufipes may be used as a good model species for testing its efficiency as B. oleae natural enemy.

DNA-based diet analyses may allow verifying prey con­sumption on field, once the calibratory fee­ding trials have been carried out in laboratory. The chance to detect the DNA of a prey for a given predator can differ significantly and refers specifically to that prey-predator system ( King et al., 2010 ). The laboratory feeding trial are a needed step for the meaningful interpretation of the field results and the extrapolation of the prey-detection period from one prey-predator system to another. In field experiments, traps are checked the day after their activation, as most carabids have nocturnal predatory habits. Zaidi et al. (1999) suggested that covering an interval of at least 12 h between supposed prey consumption during the night and predator collection the next morning is an adequate amount of time for subsequent molecular analysis. Since B. oleae consumption can be detected up to 20 h post feeding time, this may allow researchers to activate the traps for a longer time still obtaining reliable results. Moreover, recent studies have proved for P. rufipes a DNA detectability over 30 h for other preys (seeds included) ( Monzó et al. , 2011 ; Wallinger et al. , 2015 ). In addition, since Cornic (1973) reported for P. rufipes seasonal variations in the type of ingested food, this element should be taken into consideration both for field and laboratory experiments.

Recently Lantero et al. (2017) detected B. oleae consumption in two Mediterranean carabid species up to 72 h, however sex bias was not tested. Predator sex is frequently considered a factor affecting food intake and thus prey detection ( Sunderland, 1975 ) and for carabids results are contradictory. We did not find any statistically significant difference among sexes in prey detection, confirming the results obtained by Zaidi et al. (1999) . However, there could be behavioural differences in the field, being males generally more active and females showing diet seasonality ( Sheppard et al. , 2005 ; Šeric Jelaska et al. , 2014 ; Šeric Jelaska & Symondson, 2016 ). Therefore, we still suggest considering sex identity for future analysis.

Meal size and amount have been often shown to affect the DNA detectability in the predator's gut ( King et al. , 2008 ), however responses are not homogeneous. For example, Zaidi et al. (1999) demonstrated that there is no correlation between the number of preys consumed by the predators and the likelihood of successful prey DNA amplification, while King et al. (2010) found a strong effect of the meal size. In addition, it has been hypothesized that feeding the starved individuals with a large amount of prey, if not ad libitum , may bias towards longer post ingestion detection ( Aebi et al. , 2011 ). In our test only one pupa was provided to each carabid, its consumption being detected with high percentage of success. Unless feeding trials with specific purposes should be carried out, we suggest providing the tested predators with the minimum amount of food, necessary and sufficient for a reliable molecular analysis. In such way there are three main advantages: it is a cost-effective solution, it reflects better the prey availability and accessibility in the field for a given predator, and it can be helpful when prey population size is controlled in experimental settings.

We compared the efficiency of two pair primers, amplifying for a shorter (108 bp) and a longer (214 bp) mtDNA region. Comparisons were made considering the overall reactions and reactions for specific post feeding time, and no statistically significant differences were found. Fragment size can affect DNA detectability, and short fragments (< 300 bp) are usually suggested ( King et al. , 2008 ), even if for P. rufipes also medium-sized fragments allowed long prey DNA detection intervals ( Waldner et al. , 2013 ). Besides both pair primers we tested proved to be highly efficient and specific for B. oleae , it is still preferable to use both of them, especially for field-collected carabids, where the probabilities to obtain positive results are lower than in laboratory-controlled experiments ( Sheppard et al. , 2005 ; Aebi et al. , 2011 ). In fact, for field-collected predators there is no a priori knowledge whether predation has occurred, and a number of additional factors, including PCR-inhibitory substances that edaphic arthropods may enter in contact with, may affect successful amplification of prey DNA ( Juen & Traugott, 2006 ). A less time-consuming approach may be the use of a multiplex PCR ( Harper et al. , 2005 ), instead of two separate singleplex PCRs, one for each pair primer.

No significant differences were found in the success of prey detection between carabids used in the "gut trial" and those used in the "whole body trial". Nor DNA quantity nor DNA purity were affected by the two sample types. The choice in dissecting or not dissecting a specimen has been discussed in King et al. (2008) . The carabid species we used was relatively large in size, thus dissection was not impeded by its body dimension. It has to be stated that prey DNA extraction from the predator's gut relies on a certain expertise of the researcher in opening the gut and selecting the target carabid's material. In this case, the preparation of each dissected sample for the DNA extraction usually requires, as a rule of thumb, 20 min on average, before the 24 h of the sample incubation. On the opposite, DNA extraction from the whole body is a faster and more user-friendly procedure, as the beetle is not manipulated (except for bigger and chitinous appendices removal). However, its efficiency strongly relies on the body grinding procedure with liquid nitrogen, as a fine powder is needed. Because of this procedure, we estimate that the preparation of each non-dissected sample requires again 20 min on average. Being the two methods almost comparable for a predator of medium to large body size, the choice of the most convenient method may be driven by laboratory's equipment and expertise level.

In conclusion, we used the protocol of Rejili et al. (2016) introducing relevant modifications prior to the step of DNA extraction. The new protocol still ensures the extraction of quality DNA and offers many practical advantages such as cost-effective feeding trials and easy handling of specimens. The protocol is particularly suitable for field experiments, in which we suggest the collection of predators by means of "semi-dry" pitfall traps that allow carabids collection without the use of PCR inhibitors. This trapping system, together with the protocol, has been recently used for field-collected staphylinids in olive orchards, successfully demonstrating B. oleae consumption in Ocypus olens (Müller, 1764) (Coleoptera: Staphylinidae) ( Albertini et al. , 2018 ). The assessment of B. oleae consumption by field predators may be particularly advantageous in developing biological control strategies against the most harmful pest of olive orchards.

We are particularly grateful to Gaia Monteforti for collecting carabids and Prof. Angelo Canale for providing C. capitata pupae. We thank Teresa Gomes, Gisela Fernandes and Diogo Mina for their laboratory assistance. We thank the anonymous reviewers and the editor for their comments to improve our manuscript.