Mallophora ruficauda eggs were collected from herbaceous vegetation in grasslands located in Moreno (34°46′S, 58°93′W), Pilar (34°28′S, 58°55′W) and Mercedes (34°65′S, 59°43′W), three localities associated with apiaries in Buenos Aires province, Argentina, during the summers (January to March) from 2006 to 2012. We collected the egg-clutches by cutting the branch where they were attached to and stored them individually in Falcon type tubes of 30 mL until eggs started hatching. Immediately after egg hatching, neonate larvae were separated individually in Eppendorf type tubes of 1.5 mL with a moistened piece of filter paper as substrate. Drops of clean water were added when necessary to keep the humidity inside the tubes at 100%, in order to avoid dehydration of the larvae. Tubes were stored in complete darkness at room temperature between 24°C–27 °C until larvae were used in the experiments.

Hosts were collected as third instar larvae from soil samples in the same localities from where parasitoid egg-clutches were collected. The white grubs were collected from the late summer and through the whole autumn season (February to June) from 2006 to 2012 in grasslands near bee hives where robber fly activity had been observed during the previous summers. We performed a random sampling method and collected all scarab beetle larvae from each site digging the soil with a shovel at 0.30 m depth. Each individual was identified to the species level in the laboratory using the taxonomic key designed specifically to determine rhizophagous white grubs from Buenos Aires province [47]. We also recorded with a magnifier lens (16×) the number of larvae of M. ruficauda per beetle larva, which were attached externally to the host cuticle. Hosts were then classified according to the degree of parasitism in unparasitized and parasitized hosts. The unparasitized hosts were used for obtaining artificial parasitism while already parasitized hosts were used only to study the field parasitism pattern (Section 2.2.1). Hosts were kept individually at room temperature in black tubes filled with soil, and fed weekly with fresh pieces of carrots.

We studied if the different attachment percentage of parasitoids to the hosts was due to behavioral non-immunological defenses through aggressive reaction when the larvae touched the host. For this, we measured the reaction to a simulated parasitism and constructed a Simulated Attack Reaction Index (SARI). To simulate the parasitoid attack, we placed the white grub on a flat surface caring that the individual had not been disturbed previously, and gently punctured its dorsal abdomen just behind the third pair of legs with a blunt end stick. This procedure avoids the permanent injury of the host and the subsequent death. Then, the type and intensity of the response was registered. The type of the response was ranked according to the movement the white grub made during the procedure. If there was no apparent response, a 1 was assigned. Then, increasing values were assigned when: (2) movement of the tip of the abdomen to the head, (3) the head was moved forwards and backwards, (4) an inversion of the body loosing the characteristic “C” shape, (5) torsion of the body accompanied by an upward movement of the legs and (6) direct attack to the stick with the mandibles and legs. Also, the intensity of the response was gradually ranked between 1 when no reaction was observed, 2 when there was little movement, 3 when the insect moved very perceptibly and 4 when the individual moved violently. With this, the SARI index was constructed combining the type and intensity reaction with the following formula:

SARI = ((T/6) + (I/4))/2

where T stands for type of reaction and I for intensity of reaction. The index varies from 0 when there is no reaction and 1 when there is a violent reaction to the simulated attack. All the assays were performed under gloom light conditions and laboratory temperature (T° = 20.5–26.8 °C). Although white grubs inside the soil matrix could move more less than in a flat surface, this experimental design is able to measure a behavioral reaction reflecting an ecological relevant defense against parasitoid larvae. It is possible that host reaction could be increased by this method, because white grubs can move freely if not surrounded by soil, but since all individuals of each species were treated identically the index is affected in the same way.

After obtaining the pattern constructed with the SARI, we analyzed the influence of host weight on host acceptance to study if there is a relation between the defensive behavior and body size. If any size dependent defense of the hosts exists, it should be evinced after weighting. For this, we weighted every host with a commercial precision scale and we compared that weight pattern and the SARI indices to the field parasitism patterns.

The influence of non-immunological defenses, morphological and life cycle characteristics on parasitism by M. ruficauda were studied by generalized additive models with a Gaussian distribution and identity link function [54]. The smooth function was represented using thin plates regression splines with the best k for each variable [54]. Response variable was constructed as:

Magnitude of non-immunological response = 100-LP

where LP stands for the percentage of laboratory artificial parasitism for each species studied. Performance in SARI, individual weight and voltinism were included as explanatory variables in the models. Criterion as to if an explanatory variable was statistically significant, was evaluated with the deviance change of the model when excluded, and a posterior ANOVA between both models [55]. Also, the generalized cross-validation (GCV) score was used to further support the decision as to which model was better [55,56]. The GCV is an information criterion that is lower for better models [56]. All the analyses were done using the mgcv package available in R language.

We compared the field and laboratory parasitism and found that the percentage of field parasitism (FP) is different from that of laboratory parasitism (LP), the latter being always higher for all species (Table 1). This difference indicates that the first defenses, or pre-parasitism defenses, occurring when the parasitoid contacts the host are not the only ones involved in the interaction between M. ruficauda and hosts. Also, the results denote that white grubs might have both immunological and non‑immunological defenses but they seem to have different relevance in each species, as is indicated by the magnitude of the difference in 100-LP percentages. The FP pattern shows the picture after host location and acceptance occurs, hence, after parasitism is established, whereas LP shows acceptance and parasitism prior to the expression of immunological defenses. It has been previously observed in the laboratory that parasitized hosts can escape from parasitism some weeks after the attack of M. ruficauda larvae [57], indicating that some physiological defenses could be expressed. However, to determine if multiple defenses are involved in these processes, further experimental investigation is required. Also, due to the magnitude of the differences between FP and LP, the possibility that differences in parasitism rates could arise from manipulation during experimental infection could not be discarded, as observed in other systems [58]. The most important fact is that acceptance frequencies showed the same pattern in both situations, and so could be taken as reliable indicators of parasitoid host use, which is determined by the physiological capacity of larval parasitoid to exploit the host, the behavioral ability of females to find resources, and the effectiveness of behavioral defenses of the host [58].

To assess the importance of non-immunological defenses on parasitism avoidance, we studied the influence of host mean weight and SARI on host escape from parasitoid contact constructing three models. In the full model, we included all the variables as explanatory variables in the model (mean host weight and SARI; model 1). Then, we deleted SARI and left host weight (model 2), and finally we made a final model which included SARI alone (model 3). The results of each model performance are shown in Table 2.

We found no initial support to keep mean host weight in the final model (p = 0.292, Table 2). Moreover when we removed mean host weight from the model, the deletion of that variable did not cause a significant change in deviance (ANOVA (model 1 vs. model 3), deviance = 0.013, residuals df = 3.077, p = 0.289, Table 2). When analyzing the performance of model 2, we found that the deletion of the variable SARI increased the deviance significantly (ANOVA (model 2 vs. model 3), deviance = 0.535, residuals df = 3.077, p > 0.050, Table 2). Moreover, the GCV score showed that model 1 did better than model 2 (Table 2). Then, we analyzed model 3 and found that the deviance explained is almost the same (96% vs. 98%, Table 2) as that of model 1. However, GCV score showed a slightly better performance of model 3 over model 1 (Table 2). So, the best model for our data included only SARI (model 3) as the variable that explained most of the variance in the model. The results of model 3 are shown in Figure 1. It shows that when the SARI increases, the percentage of escape from parasitism also increases at almost a linear rate. However, there seems to be a SARI value (~0.55) beyond which there is no substantial increase in the percentage of escape from parasitism and the relation is almost constant. This result shows that violent body movements and torsion, as a behavioral reaction to an attack, are very effective non-immunological defenses preventing a host from being parasitized.

These kinds of defense have also been found in some hymenopterans where active defensive behaviors reduce the probability of being parasitized because the parasitoid chooses other hosts or because they directly kill the infective stage [4,59,60]. Also, it has been found that the scarab Geotrupes spiniger Marsham (Coleoptera: Geotrupidae) larvae were sensitive to surface contact, making a coordinated attack on a source of stimulation showing defensive behavior [61]. Some studies have demonstrated that certain defenses are more efficient against hymenopteran parasitoids than against dipteran parasitoids and vice versa [62]. For example, Gentry and Dyer (2002) [63] found that avoidance behaviors such as thrashing, dropping, or biting were effective against female hymenopteran parasitoids, but not against dipteran parasitoids. The fact that many dipteran parasitoids use indirect methods, such as microtype eggs or planidial larvae to infect their host, buffers the female parasitoid from coming into direct contact with the insect, thus avoiding these defensive behaviors. However, after entering the host body for the melanization process, the parasitoid attack behavior has almost no influence because once the larvae/eggs are inside the host’s hemocoel, the response is universal, both for Hymenopteran and Dipteran parasitoids [6]. For example, Cyclocephala grubs parasitized by larvae of Tiphia spp. Fabricius are likely to be found deeper in the soil than nonparasitized grubs and that this phenomenon results from a behavioral change in the host, rather than physical actions by the wasp or its larva [64]. Also, it could be that the grub’s downward movement represents a defensive response to parasitism, because movement of parasitized grubs through the soil sometimes dislodges the egg or the developing ectoparasitic larva [64,65]. Nevertheless, in this work we have found that aggressive behavior against larvae of M. ruficauda is a very effective strategy in white grubs that contributes to avoid attachment of the ectoparasitic larvae to its cuticle. This behavioral strategy allows scarab larvae to partially escape parasitism.

As a result of the first barrier white grubs form against this parasitoid, the parasitism preference of M. ruficauda by Cyclocephala species could be determined by the level of aggression of its host. This kind of behavior has been observed in intraspecific host instar selection. For instance, the parasitoid Lipolexis oregmae Gahan (Hymenoptera: Aphidiidae) can oviposit and develop on every larval instar of the host Toxoptera citricida Kirkaldy (Hymenoptera: Aphidiidae) but higher mortality risks exist when it develops on the fourth instar compared to the second [66]. Although there are not so many studies where interspecific host preference related to host aggressive behavioral defenses was analyzed, some authors found that sometimes more aggressive hosts might interrupt the parasitoid oviposition, thus increasing its handling time [59]. In M. ruficauda, the host C. signaticollis showed the least aggressive behavior and the highest percentage of parasitism. Moreover, the three species of the genus Cyclocephala show lower indices regarding aggressive behavior. Another aspect of the host's body morphology that can influence their behavioral defenses is the amount of sensorial mechanoreceptors distributed along the body of every species. It is interesting that species of the genus Cyclocephala have fewer setae than DA which has in turn less setae than the species AT and PB. These last two species have a high density of setae on every segment and this could provide them with a high sensibility to detect an organism that contacts it.

There are also some physical factors that can increase the host's probability of parasitism escape from parasitoids. Some studies suggest that high environmental temperatures influence the host immune system and resistance, increasing the probability of a host killing its parasitoid [67,68]. For example in some Hymenopteran parasitoids it was shown that egg hatchability, postembryonic development and successful attack are lower at high rather than at low temperatures [69,70] because of either more effective host defenses or a lower effectiveness of the substances injected by the female parasitoid at oviposition. Also, it has been observed that effective encapsulation of eggs is correlated with ambient temperature [71]. These examples suggest that an increase in the environmental temperature can increase the host defensive capabilities against immature parasitoids, resulting in a decrease in parasitization success, and can thus strongly affect the host-parasitoid population dynamics by altering the capacity of parasitoids to attack hosts [67]. According to parasitism success in C. signaticollis, there is some evidence that M. ruficauda larvae attach to preferential places in the host body and this could contribute to increase the survival probabilities of the parasitoid larva [72]. It has been shown that some immature parasitoids exploit preferentially host parts that provide privileged access to energetic resources or oxygen sources [73].

Although the experiments performed in this work deal with the decision of the parasitoid larva to accept, i.e., decide to parasitize or not a host once it is on its body, the larva of M. ruficauda orientates actively to the host. In this sense, the chemical cues involved in the orientation to the host in this system are also very important. Previous studies showed that M. ruficauda’s larvae orientate to chemical cues from CS, CM, CP, AT, DA but not from PB. This information is relevant because it shows that non-immunological defenses are very important in most of the host species in preventing parasitism. The fact that M. ruficauda does not orientate to PB could be interpreted as a species with very good defenses or it could explain the phenology of this species. Phenology, i.e., the timing when the vulnerable stage appears in the field, of species in a host-parasitoid system is a very important factor determining the persistence of both systems [74]. Also, spatial and temporal asynchrony could promote persistence or disruption in these systems [74,75,76]. For instance, when only a small fraction of the host population is overlapped with the parasitoid population, a partial refuge for hosts is created and persistence is increased [76]. From an evolutionary perspective, voltinism (i.e., number of generations per year) can also play an important role in the host-parasitoid interactions. In this context, potential host species that reduce overlapping with the parasitoid would escape more effectively from parasitism. Philochloenia bonariensis is a species with a life cycle at least longer than one year of duration [47]. This characteristic could explain why this species is not parasitized. Something very similar could happen with the species A. vervex that has a life cycle of three years [47]. However, in the latter case, very low levels of abundance in the area of M. ruficauda also explains why this species is not parasitized.