Wolbachia manipulate fitness benefits of olfactory associative learning in a parasitoid wasp

The fitness benefits of decision-making processes play important roles in uncertain environments, such as foraging for food or learning (Katz and Naug, 2015; Addicott et al., 2017). Through decision making, animals maximize rewards whilst minimizing costs by learning (Korn and Bach, 2015; Budaev et al., 2019). Trade-offs exist between decision costs and benefits in many ecologically relevant tasks such as: spatial exploration strategies; choosing between plant species by pollinators; and prey and predator choice (Chittka et al., 2009; Chittka and Raine, 2006). This trade-off characterizes the balance between reaping the benefits of spending energy to learn new cues or acting innately.

Several factors such as stress (Nishio et al., 2001; Shors, 2004), aging (De Bartolo et al., 2010; Weiler et al., 2008), repetition (Toppino and Bloom, 2002) and reward value (Adcock et al., 2006; Kruidhof et al., 2012; Hoedjes et al., 2011) may affect the learning ability and memory retention of animals. Reward value can affect different behaviors such as flower odor preference in honeybees (Russell et al., 2016), win–shift tendency in birds (shift away from/avoid locations that have recently yielded food) (Sulikowski and Burke, 2007), as well as changes in the duration of memory, e.g. for odors in fruit flies (Burke and Waddell, 2011). Kruidhof et al. (2012) hypothesized that consolidation of different memory forms may be induced by exposure to the different value of the same type of rewards. The importance of reward value in behavior modification of an organism is variable in relation to association ability between learned information and the reward value (Savage and Ramos, 2009; Kaiser et al., 2017). To study reward value effects on memory duration in an ecologically relevant context, parasitic wasps are ideal model organisms as learning to associate odor cues with host presence may lead to optimization of their host searching efficiency (Hoedjes et al., 2011).

Wolbachia (Rickettsiales: Anaplasmataceae) (α-proteobacteria) is the most widespread endosymbiotic bacterium and infects arthropod species, including a high proportion of insects (Kageyama et al., 2017; Kajtoch et al., 2019). Several studies have focused on the effects of Wolbachia on bionomics (Hohmann et al., 2001; Grenier and De Clercq, 2003; Miura and Tagami, 2004) and behavior (Almeida et al., 2010; Ramirez-Romero et al., 2012; Furihata et al., 2015) of parasitic wasps, and have resulted in contradictory conclusions. Wasps of the genus Trichogramma (Hymenoptera: Trichogrammatidae), like many other parasitic wasps, harbor this symbiotic bacterium (Hurst et al., 1999; Werren, 1997). The parasitoid wasp Trichogramma brassicae Bezdenko (Hymenoptera: Trichogrammatidae) has been used as a biological control agent against various Lepidoptera pest species in agricultural systems to attack pest egg stages (Hoffmann et al., 1995; Pinto, 1998; Mansfield and Mills, 2002; Babendreier et al., 2003; Kuske et al., 2003). It has been shown that two modes of reproduction in Trichogramma wasps frequently occur, as is also the case in T. brassicae individuals: arrhenotokous parthenogenesis, in which unfertilized eggs develop into haploid male offspring, which are viable and fertile (Hohmann et al., 2001; Pannebakker et al., 2004; Ma and Schwander, 2017); and thelytokous parthenogenesis, in which female offspring are produced from unfertilized eggs (Hohmann et al., 2001; Poorjavad et al., 2012; Ma and Schwander, 2017). Researches have shown that infection with the endosymbiotic bacterium Wolbachia leads to thelytoky and consequently the offspring sex ratio will be female biased (Farrokhi et al., 2010; Kishani Farahani et al., 2015; Poorjavad et al., 2012, 2018).

Kishani Farahani et al. (2017) showed differences in learning ability and memory retention between Wolbachia-infected and -uninfected T. brassicae and suggested that it may be due to negative side effects of Wolbachia infection. However, the possibility that the learning impairment between infected and uninfected strains may be due to them obtaining different benefits from learning tasks has not been explored. The main question of this paper is: do Wolbachia-infected wasps make decisions, learn and memorize new stimuli, without considering the future fitness benefits of their decisions? We hypothesized that Wolbachia-infected wasps face the same fitness benefits from learning low- and high-quality rewards because they exhibit the same resultant life history traits irrespective of reward quality. We expected that uninfected wasps would show differences in behavior, learning ability and memory retention according to host quality. In contrast, we predicted that Wolbachia-infected wasps would display the same learning ability and the same memory duration irrespective of reward quality.

All methods were carried out in accordance with Iranian and European regulations. Experimental protocols were approved by the University of Tehran. The study was carried out in compliance with the ARRIVE guidelines.

The current study did not include any potentially harmful manipulations and invasive samples. At the end of the experiments, all wasps were kept under the same conditions and were fed with undiluted honey, and during this period they were exposed to hosts to enable them to go through routine life stages including feeding and oviposition.

We compared two strains of T. brassicae: one infected with Wolbachia (Wolbachia wBaT.bra registered as FJ441291 in GenBank) and one not (uninfected). Both wasp strains were provided by the Biological Control Research Department (BCRD) of the Iranian Research Institute of Plant Protection (IRIPP). The original strains of these parasitoids were collected from the same corn fields located in Baboulsar Region (southern coastline of Caspian Sea, in 2016). It has been shown by previous studies, carried out based on ITS2 and COI sequencing, that the infected and uninfected lineages have the same genetic background (Kishani Farahani et al., 2015; Poorjavad et al., 2018). The wasps' genetical background characteristics allowed us to avoid the potential confounding effects, both physiological and behavioral, of antibiotic treatment against the infected linage (Timmermans and Ellers, 2009; Dedeine et al., 2001; Hoffman et al., 1990).

To determine Wolbachia prevalence in female wasps which produced only female offspring (Huigens and Stouthamer, 2003), a PCR method based on the Wolbachia surface protein (wsp) was implemented (81F/691R primers; Braig et al., 1998). Wolbachia presence was verified by PCR performed as described for Trichogramma species identification (Braig et al., 1998) with the following two modifications: (1) the primers used to amplify the wsp region were 5′-TGGTCCAATAAGTGATGAAGAAAC-3′ (forward) and 5′-AAAAATTAAACGCTACTCCA-3′ (reverse), and (2) the cycling program was: 3 min at 94°C, 40 cycles of 1 min at 94°C, 1 min at 50°C and 1 min at 72°C, followed by 5 min at 72°C after the last cycle. Throughout the entire experimental duration, Wolbachia presence in the infected line was screened by PCR to ensure that all different behaviors were due to the Wolbachia (Kishani Farahani et al., 2015).

For all experiments described below, we used the flour moth Ephestia kuehniella (Lepidoptera: Pyralidae) to rear parasitoids. Lepidoptera eggs were obtained from a culture maintained at the Insectary and Quarantine Facility, University of Tehran. The host eggs were maintained at 25±1°C on host larvae fed by wheat flour and yeast (5%). Glass containers (500 ml) were used to keep newly emerged and mated moths to provide moth eggs. Ephestia kuehniella eggs were collected daily, and loaded onto egg cards (2×5 cm, 200 eggs) for rearing parasitoids. The egg cards were placed into emergence canisters – 500 ml cardboard cylinders (63×161 mm) inside a 50 ml glass vial (26×93 mm) inserted through the lid – and held in incubators at 25±1°C, on a 16 h:8 h light:dark cycle at 50±5% relative humidity for more than five generations.

Females are more strongly phototactic than males (Morrison et al., 1978; Kishani Farahani et al., 2015), which facilitated the collection of newly emerged females into the glass vials. Once approximately 20 parasitoids had emerged into a glass vial, the vial was removed, provisioned with undiluted honey as a food source, and closed with a ventilated plastic cap to serve as a holding container until females were 24 h old.

Throughout experiments, lab-reared wasps were fed daily on a droplet of 10% honey solution, spread and rubbed on the internal surface of each tube to avoid damaging the wings, and consequently affecting the behavior of the wasps.

We tested the effects of Wolbachia presence and reward value – using young E. kuehniella eggs, less than 12 h old, as high-value hosts, and old eggs, 30 days old, as low-value hosts – on adult wasp learning ability and memory retention. The factorial design comprised four treatments: T1, uninfected wasps conditioned with young eggs; T2, uninfected wasps conditioned with old eggs; T3, infected wasps conditioned with young eggs; and T4, infected wasps conditioned with old eggs.

To study trade-offs in adult decision making when learning a new odor, we recorded life history traits of wasps reared on high- and low-quality rewards for both uninfected and infected wasps.

To correlate body size with fitness parameters, the length of the left hind tibia and wing surface of each individual was measured using a binocular microscope (0.5×6.3, Olympus SZ-CTV) connected to a video camera (JVC KY-F). Tibia length is commonly used as an indicator of body size in parasitoid wasps and correlates strongly with other measures, e.g. dry mass (Godfray, 1994). From photographed images, the wing surface area and tibia length of 40 wasps per strain (from each host quality, 160 wasps in total) were determined using ImageJ software. As body mass is thought to correlate with size, a minimum of 40 females per strain reared on high- or low-quality hosts (i.e. a total of 160 females) were selected randomly and frozen in liquid nitrogen on emergence, and weighed on a microbalance to an accuracy of ±0.1 µg (Mettler Toledo XP2U) (Ismail et al., 2012).

Following wasp emergence, adult longevity without food (but with access to water) was measured to estimate longevity with only available capital resources, i.e. the energy reserves within the body after development (n=40 females per strain reared on high- or low-quality hosts, i.e. a total of 160 females). Individual adults were placed in small tubes (1.5 cm diameter and 10 cm long) and were monitored until death (Kishani Farahani et al., 2016).

Innate preference of 50 wasps towards one odor (peppermint and lemon essential oil, at least 98% purity; Adonis Gol Darou Group) was tested as detailed by Kishani Farahani et al. (2017) in a wind tunnel as previously described by Yong et al. (2007), and their responses were recorded. We performed the cognitive tests in a flight tunnel (200×50×50 cm l×w×h) made of transparent Plexiglas. A smaller chamber (50×20×20 cm, l×w×h), centered within the main chamber and open at both the upwind and downwind ends, served as the experimental arena. The experimental room was illuminated at 2000 lx with LEDs (Pars Shahab Lamp Co.). Air was driven through the flight tunnel by a fan located at the upwind end, and extracted outside by a fume hood at the downwind end. The end opposite to the start zone of the tunnel was divided by a glass separator wall into two decision chambers. Each decision chamber contained an odorant stimulus presented on a filter paper attached to a glass pipette placed vertically on a stand. Single naive female wasps were introduced into the flight chamber and exposed to odor versus clean air: 25 of the 50 wasps underwent this procedure using the peppermint odor and the other 25 underwent the procedure using the lemon odor. The responses of the wasps to the odors were observed in the flight tunnel during a flight time of 15 min. Any individual that landed or hovered on an odor site for more than 2 min was recorded as a responder wasp. Females that did not complete a flight or did not fly over the start area in the flight chamber were scored as displaying no response (Kishani Farahani et al., 2017).

The ability of wasps to learn was determined using a Pavlovian conditioning procedure whereby an odor stimulus was associated with the reward of ovipositing (van Baaren et al., 2005; Bleeker et al., 2006). Both strains of wasps were naive without any former oviposition experience to avoid the variability in sequence and the retention of behavioral events associated with learning from the first host encountered (Mills and Kuhlmann, 2004). Sixty-five 1 day old naive females were exposed individually to host eggs for 15 min to gain oviposition experience. This conditioning time allowed wasps to oviposit on hosts and associate oviposition with the odor in the conditioning tank. As some wasps died, were lost or did not oviposit during the manipulation, approximately 60 wasps per treatment were kept for the experiment. Half of the tested individuals (n=30) were conditioned using peppermint odor and the remaining half with lemon odor. For conditioning, one adult wasp was introduced into a vial (2×10 cm) containing 100 eggs from high- or low-quality hosts, glued on cardboard, and transferred into the conditioning tank (25×25×25 cm). During experiments, the conditioning odor (either peppermint or lemon) was pumped into the tank at a speed of 1 m s⁻¹ speed. The conditioning process lasted a total of 2 h and was repeated for both uninfected and infected females (60 females of each of the treatments T1–T4). The conditioning time of 2 h was set based on the average time of patch leaving for 100 adult wasps exposed to 100 eggs.

Fifteen minutes after conditioning, wasps were placed individually into the flight chamber. The response of 50 female wasps (randomly selected from the surviving wasps of the 60 conditioned), 25 conditioned on peppermint and 25 conditioned on lemon, was tested for the four treatment groups (totaling 4×50=200 females). The responses of the wasps to the conditioned odors were observed in the flight tunnel during a flight time of 15 min. If females displayed a preference towards the conditioned odor (i.e. the individual landed or hovered on the conditioned odor site for more than 2 min), it was assumed that associative learning between the odor and the reward of oviposition had occurred. The number of wasps that landed on an unconditioned odor and the number of non-responding wasps were recorded to determine behavioral response variation by both strains. Females that did not complete a flight or did not fly after 5 min were scored as displaying no response. All flight responses were tested at 25°C, 50% relative humidity and a light intensity of 2000 lx (Kishani Farahani et al., 2017).

Memory (retention) was defined as present when wasps showed a significant preference for the conditioned odor (peppermint/lemon) when conditioned on peppermint or lemon, respectively, as compared with the unconditioned odor, after learning. To determine the duration of memory, experienced wasps of both strains were kept 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24 and 30 h after training until used in a bioassay in the same conditions as above. For each time interval, 50 wasps of both strains were observed in the wind tunnel as described above, representing a total of 1600 tested wasps (Kishani Farahani et al., 2017).

Before running statistical analyses, all data were tested with Kolmogorov–Smirnov test (UNIVARIATE procedure, SAS software v.9.1) for normality.

Numerical data (mass, longevity, wing surface and tibia length) were analyzed by generalized linear models (GLMs) based on a Poisson distribution and log-link function. In all cases, strain and host quality were considered factors. When a significant effect of treatment was found, a Bonferroni's post hoc multiple comparison test was performed, and the two-by-two comparisons evaluated at the Bonferroni-corrected significance level of P=0.05/k, where k is the number of comparisons. The statistical analyses were performed using SAS software (SAS Institute Inc., 2003).

The innate responses of the two strains were compared by chi-square analysis using SAS software (SAS Institute, 2003). To compare the responses of the two strains before and after conditioning, we used a GLM implemented in the GENMOD procedure (SAS software v.9.1), with the binomial family error and logit link. After this global test, the least square estimates of the proportions in each level were compared by the chi-square approximation (an option offered by GENMOD).

Host quality and strain effects and interaction of these parameters on life history traits of uninfected and infected wasps are shown in Table 1. Mass (χ²=61.83, P<0.0001), longevity (χ²=27.2, P<0.0001), tibia length (χ²=7.75, P=0.005) and wing surface area (χ²=33.61, P<0.0001) were significantly higher for uninfected wasps reared on high-quality hosts than for uninfected wasps reared on low-quality hosts (Fig. 1).

For infected wasps, we observed no significant difference in mass (χ²=1.29, P=0.259), longevity (χ²=0.49, P=0.489), tibia length (χ²=0.08, P=0.779) or wing surface area (χ²=8.2, P=0.004) between females reared on high- and low-quality hosts (Fig. 1).

The effects of reward quality and conditioning, and the interaction between these factors, on the responses of uninfected and infected wasp are shown in Table 2.

Associative learning towards conditioned stimuli was observed by uninfected wasps conditioned with a high-quality reward, with the wasps displaying attraction responses towards the conditioned stimuli. Uninfected wasps associated the conditioned odor with oviposition (lemon odor: χ²=6.30, P=0.012, N=25; peppermint odor: χ²=21.67, P<0.0001, N=25) (Fig. 2A). In addition, the neutral response by uninfected wasps was significantly lower after conditioning (lemon odor: χ²=14.62, P=0.001, N=25; peppermint odor: χ²=30.39, P<0.0001, N=25) (Fig. 2B).

Conditioned uninfected wasps conditioned with a low-quality reward displayed aversive learning by showing significantly fewer positive responses towards the conditioning odor (lemon odor: χ²=8.47, P=0.003, N=25; peppermint odor: χ²=4.86, P=0.0275, N=25) (Fig. 2A). Neutral responses of uninfected wasps were significantly lower after conditioning (lemon odor: χ²=4.51, P=0.033, N=25; peppermint odor: χ²=32.97, P<0.0001, N=25) (Fig. 2B).

Infected females significantly preferred the odor in the presence of a high-quality reward to which they were conditioned (i.e. olfactory learning) to the unconditioned odor (lemon odor: χ²=7.59, P=0.005, N=25; peppermint odor: χ²=13.9, P=0.0002, N=25) (Fig. 3A). Infected wasps showed no significant differences in neutral responses before and after conditioning (lemon odor: χ²=1.58, P=0.20, N=25; peppermint odor: χ²=2.36, P=0.124, N=25) (Fig. 3B).

Positive responses of infected wasps conditioned with a low-quality reward (lemon odor: χ²=11.43, P=0.0007, N=25; peppermint odor: χ²=8.7, P=0.003, N=25) showed a significant difference after conditioning (Fig. 3A). Furthermore, no significant differences were observed in neutral responses of infected wasps conditioned on a low-quality reward (lemon odor: χ²=3.3, P=0.067, N=25; peppermint odor: χ²=2.76, P=0.096, N=25) (Fig. 3B).

Differences in memory retention between the strains were highly significant (F=7.36, P=0.02), as were those between the reward values (F=12.03, P=0.008), while odor type (F=1.43, P=0.26) and their interaction (strain×reward value; F=0.54, P=0.71) were not significantly different. For the uninfected strain, memory retention was longer for wasps conditioned with a high-quality reward than on a low-quality reward (P=0.02 for the lemon odor; P=0.03 for the peppermint odor) (Fig. 4). For the infected strain, memory duration did not vary significantly with reward value (P=0.49 for the lemon odor; P=0. 8 for the peppermint odor) (Fig. 4). Memory retention of infected wasps was lower than that of uninfected wasps when they were conditioned with high-quality hosts (P=0.03 for the lemon odor; P=0.01 for the peppermint odor). However, memory retention did not significantly differ between infected and uninfected wasps when they were conditioned on low-quality hosts (P=0.73 for the lemon odor; P=0.35 for the peppermint odor).

In accordance with our hypotheses, uninfected wasps changed their behavior in response to the reward value. Conditioned uninfected wasps with a low-value reward showed aversive learning and avoided the conditioned odor. In contrast, conditioned uninfected wasps with high-quality hosts were attracted to the conditioned stimuli and showed olfactory learning. Uninfected wasps, conditioned on high-quality rewards, showed significantly longer memory retention than those conditioned on low-quality hosts. Wolbachia-infected wasps showed associative learning in response to high- and low-value rewards and were attracted to conditioned stimuli. Memory retention of infected wasps did not significantly differ between those conditioned on high- and low-quality rewards and their memory duration was lower than that of uninfected wasps conditioned on a high-quality reward. According to the obtained life history traits, different host qualities result in different fitness benefits for uninfected wasps by producing bigger and thus likely more fecund individuals. However, infected wasps obtained the same fitness benefits from the low- and high-quality hosts, displaying the same value as the reward for infected wasps.

The host represents the sole physiological and nutritional environment during immature development for insect parasitoids (Jervis and Kidd, 1986). Consequently, host quality is important for overall parasitoid growth and development and may affect the immature developmental time, longevity, mortality rate and fecundity (Harvey and Strand, 2002; Sampaio et al., 2008). For example, Kant et al. (2012) showed that Diaeretiella rapae females emerging from high-quality hosts produced 62% more offspring than females emerging from low-quality hosts. Body size has an impressive role in ecology and evolution theories (Angilletta et al., 2004), as it affects all aspects of physiology, life history and fitness of animals (Thorne et al., 2006; Chown and Gaston, 2010; Wilder et al., 2016). A larger body size is linked with greater fitness in several animal species (Kingsolver and Huey, 2008; Amarillo-Suárez et al., 2011; Goldbogen, 2018; Chown and Gaston, 2010). According to Lacoume et al. (2007), small males of the parasitoid wasp Dinarmus basalis (Hymenoptera: Pteromalidae) possessed a reduced size and sperm stock when compared with larger males. As a result, the smaller males fertilized fewer females in competitive situations. According to our results, uninfected wasps reared on low-quality hosts showed a smaller size compared with those reared on high-quality hosts (reward). As a consequence of parasitizing low-quality reward hosts, smaller offspring would be produced. Therefore, this could be a reason for aversive learning towards low-quality hosts observed by uninfected wasps.

In the current study, the low-quality reward led to aversive behavior in uninfected wasps. The importance of reward value in the modification of behavior varies depending on the ability to associate the learned information with the reward value. This ability, in turn, has consequences for the fitness of the organism. It allows the organism to focus exclusively on learned cues that profit the organism with more efficient feedback. Different kinds of aversive stimuli have been studied, such as electric shock and inedible food (Mery and Kawecki, 2005; Zhao et al., 2004; Salloum et al., 2011; Kandori and Yamaki, 2012; Unoki et al., 2005). In the case of aversive learning, studies on insects have been thus far limited to a few species such as the fruit fly, Drosophila melanogaster (Diptera: Drosophilidae), in which flies learn to associate an odor and an electric shock (Malik and Hodge, 2014; Le Bourg, 2012). Zhang et al. (2005) showed that Caenorhabditis elegans (Rhabditida: Rhabditidae) modifies its olfactory preferences after exposure to pathogenic bacteria, avoiding odors from the pathogen and increasing its attraction to odors from familiar non-pathogenic bacteria. This behavior may lead to increased dispersion of uninfected wasps to avoid low-quality hosts while infected wasps may spend more time in patches containing low-quality hosts.

It has been shown by Kruidhof et al. (2012) that different values of reward, even of the same reward type, induce consolidation of a different memory form in a parasitoid wasp. The importance of a reward may vary in relation to its benefit to an animal's fitness and the reliability of the association between the learned cues and the linked reward (Kruidhof et al., 2012). Soldati et al. (2017) showed that red-footed tortoises, Chelonoidis carbonaria (Turtle: Testudinidae), are able to recall the learned information about the stimuli as indicators of relative reward values, which demonstrates a memory for the relative quantity and quality of food. As longer-lasting memory is more energetically costly than shorter-lasting memory, longer memory consolidation by a low reward value may not be optimal (Mery and Kawecki, 2002). However, consolidation of longer memory may not be beneficial if making irrelevant associations is a high-risk action (Smid et al., 2007). The observed similar memory retention and host size of infected wasps reared on high- and low-quality hosts suggests that there are similar fitness benefits to be gained from the high- and low-quality hosts.

We observed two different memory forms, aversion and attraction, and different durations of memory retention by uninfected wasps when conditioned with different reward qualities. Based on our results, aversive memory was significantly shorter than attraction learning of uninfected wasps. Avoiding and remembering associatively learned predictors of unpleasant stimuli are crucial for survival (Krypotos et al., 2015). Memories can, however, become counter-adaptive when they are overly generalized to harmless cues and contexts (König et al., 2017). Associatively learning the predictors of noxious events is useful for survival as it enables pre-emptive avoidance. Depending on the nature of the unpleasant experience, its memory can last for different durations (Davis and Squire, 1984; Dudai, 2012; Kandel et al., 2014). Amano and Maruyama (2011) showed that the nematode C.elegans displayed shorter memory retention when conditioned by negative reinforcements than when conditioned with pleasant stimuli. Memory formation is a costly process which consumes much of the daily energy budget (Burns et al., 2011; Placais et al., 2017; Liefting et al., 2019). Energy efficiency is indeed put forward as a major factor of the selective pressure driving the evolution of nervous system function and efficiency (Placais et al., 2017). As we have shown, uninfected wasps display a smaller size on low-quality hosts, which may lead to lower energy resources. However, it seems that shorter aversive memory retention may play an evolutionary role for uninfected wasps as these animals are time-limited foragers and therefore may be able to learn new stimuli related to the presence of a higher reward.

Our results showed that Wolbachia-infected wasps associated the conditioned odor with both types of reward, high and low quality, which means they behave similarly towards hosts that represent different qualities for uninfected wasps. This behavior can aid Wolbachia transmission towards uninfected individuals. Indeed, in addition to vertical transmission of Wolbachia, which ensures the durability of intracellular symbioses, horizontal transmission must also occur (Tolley et al., 2019). Horizontal transfer was observed when infected and uninfected parasitoid larvae shared the same host (Huigens et al., 2004). According to Huigens et al. (2004), horizontal transfer can occur in nature only when a transferor and a recipient host are in close contact. Huigens et al. (2000) reported that high rates of natural horizontal transfer occur between conspecifics in Trichogrammakaykai, when Wolbachia-infected and uninfected T. kaykai larvae shared the same host. According to Kishani Farahani et al. (2015), Wolbachia-infected T. brassicae lack host evaluation/host discrimination ability, and superparasitize hosts more frequently than uninfected strains. It seems that the higher superparasitism rate and the behavior of parasitizing hosts of all qualities may allow Wolbachia to be transmitted between Trichogramma strains horizontally. As the two strains share the same niches, associative learning may allow Wolbachia-infected wasps to find hosts of all qualities, oviposit on them and thereby increase the chance of infection of new individuals, which will lead to an increase in infection rate within the species.

In conclusion, learning represents trade-offs between the cost of learning and the associated benefits. If infected wasps gain similar benefits from ovipositing in high- and low-quality hosts, there is no additional benefit of associated learning and enhanced memory retention, and thus fixed memory retention would be favored. It seems that uninfected wasps make decisions based on the future fitness benefits; the wasps will learn a cue if it is beneficial to future behavior and fitness. In contrast, decision making by infected wasps is detrimentally affected by Wolbachia. The behavior of infected wasps such as decision making may be manipulated by Wolbachia to enhance dispersal to new environments and through new hosts. However, other possibilities such as potential negative side effects of Wolbachia infection on mobility behavior, perception capacity or learning ability should be considered.

We thank Dr Lucy Alford for her help and comments in improving the English in the manuscript.