Effect of host switching simulation on the fitness of the gregarious parasitoid Anaphes flavipes from a novel two-generation approach

Herbivorous insects can escape the strong pressure of parasitoids by switching to feeding on new host plants. Parasitoids can adapt to this change but at the cost of changing their preferences and performance. For gregarious parasitoids, fitness changes are not always observable in the F1 generation but only in the F2 generation. Here, with the model species and gregarious parasitoid Anaphes flavipes, we examined fitness changes in the F1 generation under pressure from the simulation of host switching, and by a new two-generation approach, we determined the impact of these changes on fitness in the F2 generation. We showed that the parasitoid preference for host plants depends on hatched or oviposited learning in relation to the possibility of parasitoid decisions between different host plants. Interestingly, we showed that after simulation of parasitoids following host switching, in the new environment of a fictitious host plant, parasitoids reduced the fictitious host. At the same time, parasitoids also reduced fertility because in fictitious hosts, they are not able to complete larval development. However, from a two-generation approach, the distribution of parasitoid offspring into both native and fictitious hosts caused lower parasitoid clutch size in native hosts and higher individual offspring fertility in the F2 generation.

Natural enemies keep herbivores at densities lower than those that would deplete host plants 1 . Parasitoids are important natural enemies that maintain the natural balance of ecosystems 2,3 , with all insects in a given habitat potentially being hosts for parasitoids 4 . However, herbivorous insects, as the most vulnerable hosts 4 , are not passive participants; they evolve mechanical, physiological and immune defences against parasitoids 2 and are in a constant evolutionary arms race. Regardless of the defensive mechanisms the host invents, parasitoids evolve to overcome them 5,6 .
In addition to direct defence, herbivorous insects can escape parasitoids by switching to a novel host plant and obtaining enemy free space 7 . Herbivores usually experience lower performance when colonizing a novel host plant 8 , but the associated fitness reduction may be compensated by lower predation and parasitism on this plant 9,10 . Although locating hosts in a new microhabitat is not easy for parasitoids, there are known cases of this switching 11,12 . These new habitats pose several challenges for parasitoids 13 , that might reduce their fitness, either directly (e.g. encounter of a plant toxin or metabolite by the juvenile stage) or indirectly (e.g. effect of decreased host size and quality on the parasitoid) 14 . In addition to possible changes in parasitoid fitness related to the new host plant of their host 8,15 , the new environment may present not only new predators of the parasitized hosts 16 but also new potential host species for parasitoids 17 .
In this changing environment, it is critical whether the parasitoid's reproductive strategy is solitary or gregarious: solitary parasitoids respond to these changes primarily by choosing the sex of the offspring developing in the host 18 , whereas gregarious parasitoids can choose both sex ratio and clutch size 19,20 . Female solitary parasitoids usually lay one egg or sometimes multiple eggs in a single host, but only one offspring completes development.

Discussion
In general, parasitoids show remarkable differences in their reproductive strategies, with their success depending on many factors such as ovigeny, fecundity, life span and many other 51 . In this study, we focused on one of the factors that affects the successful reproduction of parasitoids 52,53 , namely the localization of a suitable host in relation to a distinct host plant. A considerable number of studies show that parasitoids are influenced by plant defenses, either by changes in plant volatiles that hosts can use to defend themselves against parasitation or, conversely, these toxins degrade host quality for parasitoids [54][55][56] . However, there is a need to examine the effect of plants, herbivores and parasitoids in broader ("tritrophic") context, and to uncover many direct and indirect interactions that will vary in strength and significance 57 . Thus, here, we presented the effect of host plants and the effect of a fictitious host that the parasitoid may encounter in the new environment of the fictitious plant on the fitness of parasitoids from a novel two-generation approach 45 . Using the model species A. flavipes, in Fig. 3, we show the hypothetical importance of this approach for studying parasitoid fitness. For example, in a situation  www.nature.com/scientificreports/ where founders would refuse the host on the fictitious host plant, the changes in fertility in relation to a different type of host plant may not be observed in the F1 generation but will appear in the F2 generation.
Recognition of the host plant by the parasitoid is fixed before adult emergence 58,59 , immediately following emergence 60 , or during oviposition 61 . In our no-choice tests, females of A. flavipes parasitized more host eggs on the substrate on which they hatched (control substrate; Fig. 4c) than on the usual and fictitious host plant (Fig. 4a,b), which would indicate that females learn about the host plant before or shortly after hatching. This finding may also be related to the fact that A. flavipes females hatch with a final number of own eggs (pro-ovigeny type) 62 that can be laid immediately after hatching 46 . Based on the fact that synovigenic species are longer lived than proovigenic species 63 , it can be assumed that synovigenic species have a longer time to recognize the host plant. However, the influence of the ovaries in relation to host plant recognition is debatable because the imprinting of the host plant after the emergence of the parasitoids and, in relation to this, their higher fitness have been confirmed in quite a number of studies 64,65 and these studies included both synovigenic (Trichogramma brassicae Bezdenko, 1968) 63 and proovigenic (Cotesia Cameron 1891) 63 parasitoid species.
Interestingly, if females had a choice between the usual host plant, fictitious host plant and the control substrate, they chose to parasitize more host eggs on the usual host plant than on the learned substrate during emergence (control substrate) (Fig. 4d). This would indicate that the females can recognize the usual host plant if they have a choice between substrates, i.e., learning during oviposition 61 .
When comparing the results of the choice test and the no-choice test, females with a choice between substrates on which host eggs are offered for parasitization always have significantly lower values of fitness characteristicsthe number of parasitized hosts, the number of offspring and the clutch size (Fig. 4). Although many factors are involved in fitness in the broader sense, one of the main indicators is still undoubtedly the number of offspring, especially when fitness is measured by maternal fertility according to the traditional approach 43,44 . However, the two-generation approach involves a change in fertility between the F1 and F2 generations caused by different clutch sizes 45 . In this study, in the choice test (Fig. 4d), the females had fewer offspring than females in the nochoice test (Fig. 4a-c) but also a smaller clutch size, thus ensuring a larger offspring body size and higher fertility, which increased by approximately 13 percent in the F2 generation (Fig. 4d). Other changes in the fertility of the F2 generation were not obvious (Fig. 4); however, for interest, the seemingly lower fertility in the F1 generation in relation to the fictitious host plant may ensure higher fertility in the F2 generation (Fig. 4b compared to 4a).
Interpretation of our results in the natural environment will be difficult because in nature, the offspring of switched hosts can use toxins from new host plants for their own protection 66,67 . In our experiments, the native host consumed only their native host plants; however, interestingly, with the simulation of switched hosts, the www.nature.com/scientificreports/ wasps with choice between the host plants had a lower fitness. It can be assumed that lower parasitoid fitness can be compensated by possible lower predation of fictitious hosts 68 . The switched host usually escapes the strong pressure of predators and parasitoids on their native plant 9,10 . Although under certain conditions, the predator avoids parasitized prey 69 , generally the parasitoid fitness is reduced more than predator fitness by intraguild predation 70,71 . Thus, for parasitoids, specialization on switched hosts should be stronger and more advantageous than for predators 12 .
Following host switching in a new environment, parasitoids can adapt to new potential host species 4 . In the second part of this study, we simulate the impact of a fictitious host on the fitness of parasitoids when they follow a switched host into a new environment. In the new environment of the fictitious host plant, the A. flavipes females parasitize the native host as well as the fictitious host. By parasitism, they reduce the fertility of the native and fictitious hosts and, concurrently, partly their own fertility because they were not able to complete larval development in the fictitious host (same as in our previous study Samková et al. 45 ). Here, we show that the fertility of switched parasitoids is lower (females have 6 native and 6 fictitious hosts on the fictitious host plant; Fig. 5c) than the fertility of parasitoids on the usual host plant, which has high (12 native hosts; Fig. 5a) or low (6 native hosts; Fig. 5b) host population density. However, from a two-generation approach, if switched parasitoid females distribute their offspring between both the native and fictitious hosts, the clutch size decreases, and the offspring in the native host obtain a larger body size and higher fertility in the F2 generation ( Fig. 5; similarly as in Samková et al. 45 ). It can be assumed that in the new environment of a new host plant, among other factors 16,55 , parasitoid fitness will be reduced by parasitism of fictitious hosts in which the parasitoid is unable to complete larval development, but using this ovipositing behaviour, gregarious parasitoids increase the fertility of offspring developed in native hosts ( Fig. 5; Samková et al. 45 ). In addition to higher fertility, larger female offspring also obtain more females in the F3 generation offspring compared to smaller females 72 , and it can also be assumed that there is higher flight efficiency 40,73 , longevity 74 or longer egg-laying time 75 . Additionally, by parasitizing the fictitious host, parasitoids gain the opportunity for specialization on new potential host species.

Material and methods
Parasitic wasps. Anaphes flavipes was reared from host eggs (Oulema spp.) collected in cereal fields in Prague (50.136° N, 14.363° E) from the end of April until the end of June in 2019. The parasitized host eggs were stored in Petri dishes with only moistened filter papers without host plants until adult wasps emerged. These "wild" wasps were used as an initial population from which the next generations of parasitoids were reared in an environmental chamber at 22 ± 2 °C with 40-60% relative humidity and continuous illumination. Subsequent generations of females and males were used for experiments. All wasps used in the experiment were naive to the host plant, that is, they were kept in Petri dishes without a host plant only on filter paper. Mated females (not older than 24 h postemergence) were placed in Petri dishes with host eggs. The females were not fed before the start of the experiment or during the experiment, and they had free access to water (modified from Samkova et al. 20,41,45,72 ). and the habitual/common hosts of the Oulema species complex (including two very ecologically close species, O. duftschmidi (Redtenbacher, 1874) and O. melanopus (Linnaeus, 1758) (Coleoptera: Chrysomelidae)) were used; these species were used identically as in our previous studies e.g. 20,41 and in other studies 46,76 because they were determined only on the basis of genital preparation 77 . In the current study, the host culture was established from adults collected in Prague (50.136° N, 14.363° E) and in Police n/Met (50.527° N, 16.245° E). The adults were kept in plastic boxes with moistened filter papers, were fed of their native host plant (G. viridula were fed by R. obtusifolius and Oulema spp. were fed by T. aestivum), and had unlimited access to water. They were allowed to lay their eggs on leaves at 22 ± 2 °C, a relative humidity of 40-60% and a 16:8-h L:D cycle. We used host eggs no older than 24 h (modified from Samkova et al. 20,41,45,72 ).
Host plant. The substrates on which host eggs were offered for parasitization were as follows: the usual host plant was Triticum aestivum L. (Poaceae), and Rumex obtusifolius L. (Polygonaceae) was used as the fictitious host plant. The usual host plant is a typical plant on which the host Oulema spp. occurs naturally in Europe and North America 77 , and the parasitoid Anaphes flavipes also occurs in this environment 46 . In the Czech Republic, R. obtusifolius, a plant host of G. viridula, which is ecologically and phylogenetically close to Oulema spp. is found near the land of this native plant (pers. obs.) , so this plant was chosen as a fictitious plant because there is a possibility that the parasitoid of A. flavipes will encounter this plant. The host plants were obtained at localities in Prague (50.1367° N, 14.3638° E), and they were used within three hours of collection; prior to their use, they were stored in water. The host eggs were offered to wasps for parasitization on one (no-choice experiment) or two pieces (choice experiment) of the same leaf size. As a control, the host eggs were laid loosely on filter paper. Laboratory experiments. All laboratory experiments were performed in Petri dishes (8.5 cm) in a thermal cabinet at 22 ± 2 °C and 40-60% relative humidity under a 16:8 h (L:D) photoperiod. Individual parasitized host eggs were moved to 1.5-ml plastic tubes on the 9th or 10th day after parasitization and stored at the same temperature in a thermal cabinet. The number and sex ratio of the wasps that emerged from each parasitized host egg were measured (modified from Samkova et al. 20,41,45,72 ).
Experimental design.

No-choice test of host plants
Twelve

Choice test of host plants
Each female (n = 20) had 12 host eggs available for parasitization for 8 h on the three different substrates in the same Petri dish (4 host eggs on the usual host plant-crop (T. aestivum), 4 host eggs on the fictitious host plant (R. obtusifolius) and 4 host eggs on filter paper).

Effect of fictitious host
Each female had host eggs offered for parasitization in three groups: (1) simulation of switched host to the fictitious host plant (6 native and 6 fictitious host eggs on the fictitious host plant; n = 15); (2) simulation of native environment with a low population density of host (6 native host eggs on the usual host plant; n = 11); and (3) simulation of native environment with a high population density of host (12 native host eggs on the usual host plant; n = 19). For fictitious hosts, we measured the rate of emergence in (1) experiments with native and fictitious hosts (n = 114), and for controls, we measured the rate of emergence for fictitious hosts without parasitoids (n = 100).
Simulation of novel two-generation reproductive approach. In our previous study, we have shown that with higher clutch size the body size of the A. flavipes offspring decreases and at the same time, that the offspring body size determines their fertility 41 . Therefore, in the study by Samková et al. 45 we proposed a two-generation model to see intergenerational changes in fertility when clutch size changes in the F1 generation. In this study, we applied this model to the groups of experiments where there was a change in clutch size in the F1 generation, respectively a change in offspring body size-Figs. 3, 4 and 5. In these figures, the distribution of offspring in host eggs (number and sex ratio) was determined using the frequencies of each clutch size in real data (Suppl. Mat. 4). Fertility of hypothetical group in Fig. 3 was extrapolared by formula "fertility = 20-sqrt(clutch size)", which fitted means of all measured fertility levels (rounded to the nearest integer).
The sex ratio of offspring did not differ significantly between groups (see Results), hence the 3:1 ratio used by Anderson  www.nature.com/scientificreports/ Statistical data processing. All statistical analyses were performed in R 4.0.3 (R foundation for statistical computing 2020 78 ). The models used were specific for each dependent variable. For the number of offspring per dish, we used linear models (LM), and for the rate of parasitized eggs, we used generalized linear models for binomial distribution (GLM-b). The offspring sex ratio was analysed by means of mixed-effect generalized linear models for binomial distribution (GLMM-b), and clutch size was analysed either by mixed effect linear models (LMM) or by mixed-effect generalized linear models for Poisson distribution (GLMM-p, data transformation: x-1) depending on the distribution of a particular dataset. An ID specific for each dish was used as a random factor in all mixed-effect models. The mixed-effect models were built in the R package lme4 79 and analysed by their comparison with simpler models (see below) by means of ANOVA. Basic models always included a single fixed factor. The factors were egg substrate (Triticum/Rumex/control filter paper), substrates per dish (single substrate/mixed), or presence of leaf bug eggs (yes/no). Basic mixedeffect models were compared with null models that included the random factor only. Models assessing interactive effects between factors (e.g., different effects of substates in single-substrate and mixed dishes) always included two factors and their interaction. Mixed-effect models with interactions were compared with a model including the same two factors but not the interaction.
We calculated 95% confidence intervals for normally distributed data in R 4.0.3 (R foundation for statistical computing 2020 78 ). For binomial and Poisson-distributed data, 95% confidence intervals were calculated using the R packages Hmisc 80