Effects of genotype and host environment on the cuticular hydrocarbon profiles of Lysiphlebus parasitoids and aggression by aphid-tending ants

Abstract

Parasitoids in the genus Lysiphlebus specialize on ant-tended aphids and have previously been reported to mimic the cuticular hydrocarbon (CHC) profiles of their aphid hosts to avoid detection by ants. However, the precise mechanisms that mediate reduced ant aggression toward Lysiphlebus spp. are not known, nor is it clear whether such mechanisms are broadly effective or specialized on particular aphid hosts. Here we explore the effects of wasp genotype and host environment on Lysiphlebus CHC profiles and ant aggression. Rearing asexual Lysiphlebus lines in different host aphid environments revealed effects of both wasp line and aphid host on wasp CHCs. However, variation in genotype and host affected different features of the CHC profile, with wasp genotype explaining most variation in linear and long-chain methyl alkanes, while aphid host environment primarily influenced short-chain methyl alkanes. Subsequent behavioural experiments revealed no effects of host environment on ant aggression, but strong evidence for genotypic effects. The influence of genotypic variation on experienced ant aggression and relevant chemical traits is particularly relevant in light of recent evidence for genetic divergence among Lysiphlebus parasitoids collected from different aphid hosts.

1. Introduction

Mimicry is widespread in nature and has been the focus of considerable empirical and theoretical research [1–5]. A key factor influencing the evolutionary dynamics of mimicry systems is the extent to which relevant traits of the mimic track those of the model via genetic evolution alone or are mediated by phenotypic plasticity [6–8]. In avian brood parasites, for example, host specialization is more common when the traits by which parasites mimic host brood are genetically specified rather than being acquired during development (e.g. via learning of host begging calls) [7,9,10]. Similar patterns have been described in insect systems, including brood and social parasites [6,11–13], where mimicry frequently entails the imitation of chemical cues and signals either via direct biosynthesis of relevant compounds or via acquisition from the host or its environment [8,14,15]. However, because the specific mechanisms underlying chemical mimicry systems frequently remain unknown, we often have limited information about the genetic and environmental origins of relevant chemical traits.

Distinguishing between genetic and environmental influences on mimicry traits is particularly relevant in the case of mimics that are obligate parasites of their model organisms—and therefore typically experience strong selection pressure to successfully track relevant host traits [16]. Parasitoid wasp species in the genus Lysiphlebus (see [17] for latest phylogeny) are specialists on ant-tended aphid hosts, and previous studies have reported that Lysiphlebus spp. elicit relatively low levels of aggression from ants in multiple genera (Lasius niger, Myrmica ruginodis [18]) compared with other parasitoids. Past work has also shown that this reduced aggression is mediated by cuticular hydrocarbons (CHCs) [19–21], which are used by ants to recognize aphid mutualists, which they defend against attack by parasitoids and predators in exchange for carbohydrate-rich honeydew [22–24], and possibly also to discriminate between aphid lines on the basis of reward quality [25,26]. Moreover, similarity in the CHC profiles of Lysiphlebus parasitoids and aphid hosts suggests that this interaction entails chemical mimicry (C. Hertaeg, C. Vorburger, C. M. De Moraes, M. C. Mescher 2023, in preparation), although the detailed mechanisms underlying these interactions remain unknown, and other forms of biochemical crypsis cannot be ruled out [5,27]. It also remains unclear whether chemical mimicry (or crypsis) by Lysiphlebus spp. is broadly effective in avoiding detection by ants or specialized to particular aphid hosts, and previous studies have not explored whether these mechanisms have a proximate genetic basis or are mediated by phenotypic plasticity (e.g. via the acquisition of CHC cues from the aphid host during development).

These questions are particularly relevant in light of recent work documenting genetic divergence among Lysiphlebus individuals collected from different aphid hosts [28], which suggests a possible role for host specialization in driving the formation of host races or even speciation [29]. Biochemical adaptation (e.g. to mimic host CHC profiles) could plausibly play a role in such specialization, and similar processes have been documented in other chemical mimicry systems [30,31]. Such adaptation could also create barriers to host switching, which is thought to be an important factor in host-associated differentiation in parasitic arthropods [32], although the potential importance of such processes again hinges on whether there is genetic variability in the CHC traits that mediate reduced ant aggression toward Lysiphlebus and on whether the underlying mechanisms are specific to particular aphid hosts. Even if these mechanisms are broadly effective in reducing ant aggression across multiple hosts, however, relevant genetic variation might still enable selection for better specialization to particular aphid species.

The communicative functions of CHCs can rely not only on the presence of particular compounds but also on ratios of specific compounds or compound classes [22,33–35], and it is known that both genetic and environmental factors can influence relevant features of the CHC profile [36–39]. In social insects for example, nestmate recognition can involve recognition of CHC cues that are inherited from parents [40] or acquired from the colony environment [41]. In parasitoid systems, there is evidence (solely from ectoparasitoids) that association with different host species can influence parasitoid CHC profiles [42–45], although implications for chemical mimicry or camouflage have not been explored. It is known that arthropod CHC profiles can be modified by diet [46–48], which may have relevance for parasitoid wasps that feed exclusively on host tissues during their larval stages. In addition, insect CHCs can be influenced by other environmental factors, including exposure to plant lipids, nest and substrate materials, and close contact with other insects [11,37,49]. Hence, features of the CHC profiles of Lysiphlebus (and other) parasitoids can plausibly be acquired directly from the host, or influenced by the host environment during development, but may also be influenced by the genotypes of the wasps themselves.

The current study explores the importance of wasp genotype and host environment for the CHC profiles of Lysiphlebus fabarum, which parasitizes multiple host aphid species from the genus Aphis, and for ant aggression toward L. fabarum. To assess genetic and host influence on parasitoid CHCs, we reared four different asexual lines of L. fabarum (collected on four different host aphids) on each of three different host aphid species (Aphis fabae cirsiiacanthoidis, Aphis gossypii and Aphis urticata) in a full factorial design. We then extracted and analysed aphid and wasp CHC profiles and calculated the effect size of wasp genotype and host aphid species on each compound in the wasp CHC profile. In subsequent behavioural experiments, we first assessed the influence of host-acquired compounds on ant aggression by exposing wasps from the same genetic line that had developed in three different aphid host species to aphid-tending L. niger ants associated with each of the three aphids. In a second set of behavioural experiments, we investigated the influence of genetic differences by confronting ants with wasps of all four genotypes that had developed in the same aphid species as the one that ants tended during the experiment.

2. Results and discussion

(a) . Both genotype and environment influence wasp cuticular hydrocarbon profiles

In experiments where four asexual L. fabarum lines were reared on each of three different aphid host environments, chemical analyses revealed clear separation of L. fabarum CHC profiles when grouped either by host aphid environment (figure 1a) or by wasp genotype (figure 1b). These results demonstrate both genetic and environmental influences on wasp CHC profiles (figure 1a–c and electronic supplementary material, figure S1). Furthermore, separation between aphid species (figure 1c) suggests that the similar separation of wasps by aphid host might be based on the acquisition of CHCs or their precursors from host aphids, as this separation remains distinctive when we account for wasp genotype (figure 1c). Furthermore, CHC profiles of all wasp genotypes react similarly to the rearing environments. Compound classes showing marked variation between host aphid profiles included the most abundant linear alkanes and many short-chained monomethyl alkanes (electronic supplementary material, table S1). Overall amounts of CHCs varied markedly among aphid species and among wasps across all lines that developed in these host environments, while differences between wasp lines were smaller (electronic supplementary material, table S1). Furthermore, wasp CHC amounts were 2–6 times larger than CHC amounts on aphids (electronic supplementary material, table S1). Chromatograms of host aphid species and wasp lines reared on different host aphids are presented in electronic supplementary material, figure S2.

Figure 1.

Discriminant analyses of principal components (DAPCs) on cuticular hydrocarbon (CHC) profiles of (a) wasps reared on different host aphids (8 PCs), (b) different wasp lines (3 PCs), and (c) aphids grouped by species and wasps grouped by wasp line and aphid host species (8 PCs). Species names in brackets indicate the aphid species on which wasp lines were collected in the field. Wasp line, aphid host and their interaction had significant effects on wasp CHC profiles (permutational multivariate analysis of variance (PERMANOVA): wasp line: pseudo F = 16.79, R² = 0.25, p < 0.001; aphid host: pseudo F = 30.73, R² = 0.31, p < 0.001; interaction: pseudo F = 2.23, R² = 0.07, p < 0.001). CHC profiles of the three host aphids were also significantly different (PERMANOVA: F = 135.18, R² = 0.76, p < 0.001). Electronic supplementary material, table S1 shows percentages of all 38 compounds detected in the samples, and all pairwise comparisons between wasp and aphid CHC profiles are presented in electronic supplementary material, table S2.

(b) . Wasp genotype and aphid host influence different features of the wasp cuticular hydrocarbon profile

Calculating effect sizes of wasp line and aphid host environment on individual CHCs showed wasp genotype to be the primary driver of variation in linear alkanes (the most abundant compound class), as well as long-chain dimethyl and trimethyl alkanes (figure 2). In contrast, the composition of short-chain monomethyl alkanes was mainly influenced by the aphid host environment (figure 2). Moreover, although both linear and short-chain methyl alkanes varied considerably among aphid species (electronic supplementary material, table S1), only host differences in short-chain methyl alkane compositions were reflected in wasp CHC profiles. Host influence on methyl alkanes might reflect variable amino acid availabilities in the hosts, as the biosynthesis of these compounds uses valine, isoleucine and methionine as precursors [50,51]. Aphid amino acid content could, in turn, reflect variation in the nutritional composition of their phloem sap diet, different biosynthetic abilities, or the presence of endosymbiotic bacteria that synthesize amino acids [52,53] and influence aphid CHC profiles [25].

Figure 2.

Effect of wasp line, aphid host and their interaction on individual compounds in the cuticular hydrocarbon (CHC) profile of Lysiphlebus fabarum wasps. Effect sizes (η²) were calculated using the EtaSq function in R, after performing a type III ANOVA for every compound detected in the wasp profiles. Linear alkanes (shortest to longest) are listed first and are followed by all methyl-branched alkanes (shortest to longest). Unknown methyl-branch positions are marked with an X.

(c) . Cuticular hydrocarbon features varying with wasp genotype are likely more important in eliciting ant aggression

We next conducted a series of behavioural experiments with ants to explore how wasp genotype and aphid host influence ant aggression. In an initial experiment, we examined the responses of aphid-tending L. niger ants to L. fabarum wasps (from line 1) reared on three different host aphids. These trials were carried out in ant-tended colonies of all three aphid species on which the wasps had been reared (aphid backgrounds). We observed no significant variation in ant aggression toward wasps from the three different aphid hosts when tested in any of the three aphid backgrounds (figure 3a, table 1); however, as expected, aggression toward L. fabarum wasps reared on all three aphid hosts was much lower than that observed for non-mimic wasps (Aphidius colemani), which served as positive controls for the ant response. A similar pattern was observed for time ants spent antennating wasps from each treatment, with higher rates of aggression being associated with short antennation times (figure 3b and table 1b). These experiments also revealed variation in overall rates of ant aggression when ants tended different aphid species (figure 3b and table 1b), which might reflect differences in the quality of aphid honeydew [54,55].

Figure 3.

Effect of aphid host on ant aggressiveness towards Lysiphlebus fabarum wasps (line 1). Non-mimic Aphidius colemani wasps served as negative controls. (a) Attack rate of ants towards A. colemani wasps and L. fabarum wasps that developed on three different host species (Aphis fabae cirsiiacanthoidis, Aphis gossypii, Aphis urticata) presented in three different aphid backgrounds (A. fabae cirs., A. gossypii, A. urticata). Error bars depict confidence intervals for binomial probabilities (confidence coefficient = 0.95) and numbers under the bars show sample size. Aphid background and wasp species (A. colemani or L. fabarum line 1) had a significant influence on ant attack rate (generalized linear mixed-effects model (GLMER): type of wasp: χ² = 27.2, d.f = 3, p ≤ 0.001; aphid background: χ² = 6.9, d.f = 2, p = 0.03). Differences in attack rate between L. fabarum wasps that developed on different hosts were not significant (pairwise comparisons between backgrounds and wasps can be found in table 1a). (b) Time ants spent antennating A. colemani and L. fabarum wasps that developed on three different host species during the maximum 10 min long observations. Box-plot elements: centre line, median; box limits, first and third quartile; whiskers, largest or smallest observation ≤ or ≥ the upper or lower box limits plus or minus 1.5 times the interquartile range. Outliers are represented by large points. Numbers under the bars show sample size. Aphid background and wasp species had a significant influence on antennation time (linear mixed-effects model (LMER): type of wasp: χ² = 52.8, d.f = 3, p ≤ 0.001; aphid background: χ² = 16.7, d.f = 2, p < 0.001). Differences in antennation time between L. fabarum wasps that developed on different hosts were not significant (pairwise comparisons between backgrounds and wasps can be found in table 1b).

Table 1.

Pairwise differences (emmeans) between wasps and aphid backgrounds. (a) Differences in attack rates across Lysiphlebus fabarum wasps reared on different host aphids and non-mimetic Aphidius colemani wasps, as well as differences between aphid backgrounds. (b) Differences in antennation time between different wasps and aphid backgrounds.

(a) attack rate	estimate	s.e.	z-ratio	p
contrast wasp species/wasp host
A. colemani/L. fabarum (A. fabae cirs.)	6.80	1.36	5.00	<0.001
A. colemani/L. fabarum (A. gossypii)	6.42	1.30	4.93	<0.001
A. colemani/L. fabarum (A. urticata)	7.34	1.47	5.00	<0.001
L. fabarum (A. fabae cirs.)/L. fabarum (A. gossypii)	−0.38	0.67	−0.06	0.942
L. fabarum (A. fabae cirs.)/L. fabarum (A. urticata)	0.54	0.84	0.46	0.917
L. fabarum (A. gossypii)/L. fabarum (A. urticata)	0.92	0.84	1.10	0.688
contrast aphid background
A. fabae cirs./A. gossypii	1.03	0.68	1.51	0.288
A. fabae cirs./A. urticata	2.28	0.91	2.51	0.032
A. gossypii/A. urticata	1.25	0.94	1.32	0.382
(b) antennation time	estimate	s.e.	t-ratio	p
contrast wasp species/wasp host
A. colemani/L. fabarum (A. fabae cirs.)	−292.80	42.20	−6.93	<0.001
A. colemani/L. fabarum (A. gossypii)	−244.30	41.60	−5.88	<0.001
A. colemani/L. fabarum (A.urticata)	−262.30	44.20	−5.93	<0.001
L. fabarum (A. fabae cirs.)/L. fabarum (A. gossypii)	48.50	30.60	1.58	0.390
L. fabarum (A. fabae cirs.)/L. fabarum (A. urticata)	30.50	33.20	0.92	0.795
L. fabarum (A. gossypii)/L. fabarum (A. urticata)	−18.10	33.40	−0.54	0.949
contrast aphid background
A. fabae cirs./A. gossypii	−112.60	30.60	−3.68	0.001
A. fabae cirs./A. urticata	−99.90	31.60	−3.16	0.005
A. gossypii./A. urticata	12.70	31.40	0.40	0.914

Although we only tested ant reactions towards one wasp line, these findings suggest that variation in CHCs associated with development in different aphid host environments does not have strong effects on ant aggression. Lysiphlebus fabarum wasps from line 1 elicited low levels of ant aggression regardless of which aphid host they were reared on or the aphid background in which trials were conducted. This strongly suggests that host-derived compounds are not essential to mimicry and furthermore that the mechanisms underlying reduced ant aggression are not narrowly restricted to the context of particular aphid hosts, as wasps experienced low ant aggression in different aphid hosts, albeit from the same genus. However, a previous study [21] reported that Lysiphlebus wasps experience higher levels of ant aggression on an aphid host from a different genus than the one from which they had been collected in the field. Moreover, the current findings do not exclude the possibility of host specialization, particularly for L. fabarum lines that form long-term associations with particular aphid hosts under field conditions.

In a second series of behavioural experiments, we held the aphid background constant and assessed variation in ant aggression toward L. fabarum parasitoids from four different genetic lines. As in the first experiment, wasps from all L. fabarum lines elicited lower levels of aggression than the positive control (A. colemani); however, here we also saw significant variation among L. fabarum lines (figure 4a and table 2). As in the previous experiments, antennation time displayed a similar pattern, being negatively correlated with aggression (figure 4a and table 2). Furthermore, the lowest level of aggression was observed for L. fabarum line 3; while this difference was only statistically significant in comparison with line 2, it is consistent with the prediction that line 3 should be best adapted to A. fabae cirsiiacanthoides (the aphid being tended in these experiments), as it was initially collected from this species.

Figure 4.

Effect of Lysiphlebus fabarum wasp line on ant aggressiveness. Non-mimic Aphidius colemani wasps serve as negative controls. Species names in brackets indicate aphid species on which wasp lines were collected in the field. (a) Ant attack rates in Aphis fabae cirsiiacanthoides background towards A. colemani and L. fabarum lines 1–4 reared in A. fabae cirsiiacanthoides. Error bars depict confidence intervals for binomial probabilities (confidence coefficient = 0.95) and numbers under the bars show sample size. Pairwise comparisons of attack rates between wasp lines 1–4 and A. colemani can be found in table 2. (b) Time ants spent antennating A. colemani and L. fabarum lines 1–4 during the maximum 10 min long observations in an A. fabae cirsiiacanthoides background. Box-plot elements: centre line, median; box limits, first and third quartile; whiskers, largest or smallest observation ≤ or ≥ the upper or lower box limits plus or minus 1.5 times the interquartile range. Outliers are represented by large points. Numbers under the bars show sample size. Pairwise comparisons of antennation time between L. fabarum lines 1–4 and A. colemani can be found in table 2.

Table 2.

Pairwise differences (emmeans) between attack rates and antennation time that different Lysiphlebus fabarum wasp lines and Aphidius colemani wasps experienced in an Aphis fabae cirsiiacanthoides background.

	attack rate				antennation time
contrast wasp species/line	estimate	s.e.	z-ratio	p	estimate	s.e.	d.f	t-ratio	p
A. colemani/line 1	4.09	0.82	5.00	<0.001	−6.32	0.87	231	−7.23	<0.001
A. colemani/line 2	2.64	0.81	3.26	0.010	−3.11	0.87	231	−3.55	0.004
A. colemani/line 3	5.08	0.85	5.95	<0.001	−7.45	0.88	231	−8.48	<0.001
A. colemani/line 4	3.84	0.81	4.72	<0.001	−3.06	0.88	231	−3.48	0.005
line 1/line 2	−1.45	0.46	−3.13	0.015	3.21	0.87	231	3.68	0.003
line 1/line 3	0.99	0.50	1.99	0.271	−1.13	0.88	231	−1.29	0.701
line 1/line 4	−0.25	0.45	−0.55	0.982	3.26	0.88	231	3.71	0.002
line 2/line 3	2.44	0.52	4.72	<0.001	−4.34	0.88	231	−4.94	<0.001
line 2/line 4	1.20	0.46	2.63	0.065	0.05	0.88	231	0.05	1.000
line 3/line 4	−1.24	0.50	−2.48	0.095	4.39	0.88	231	4.97	<0.001

When we repeated this experiment in a different aphid background (A. gossypii) we observed much higher rates of overall aggression, so that there were no consistent differences between L. fabarum and non-mimic (control) wasps (electronic supplementary material, figure S3 and table S3). We did still observe some evidence for variation in ant responses to different L. fabarum lines (electronic supplementary material, figure S3c), and L. fabarum line 3 again elicited the lowest levels of aggression (electronic supplementary material, figure S3 and table S3). However, the failure to observe consistent differences in ant aggression toward L. fabarum lines versus the positive control (A. colemani) warrants significant caution in drawing conclusions about these results—because differential ant responses to these species are confirmed not only by other trials in the current study, but also by previous findings [56,57]. It seems likely that the high overall levels of ant aggression observed in this experiment are an artefact of our experimental conditions, and this interpretation is further supported by the divergent results observed for one wasp–aphid combination (L. fabarum line 1 in A. gossypii) that was tested in both this experiment and experiment 1: while this line elicited high levels of ant aggression (and no significant differences versus A. colemani) in the current experiment, it elicited low levels of aggression (and strong differences from A. colemani) across 52 replicates in experiment 1, as well as in preliminary experiments (not presented).

Taken together, the results of our cross-fostering experiment and chemical analyses clearly distinguish between the effects of wasp genotype and host environment on Lysiphlebus CHC composition, providing the first characterization of intraspecific genetic and environmental contributions to the composition of parasitoid CHC profiles. Our chemical analyses also indicate that genotype and host environment largely affect different features of the CHC profile, with genotype being the primary driver of variation in linear alkanes, as well as long-chain dimethyl and trimethyl alkanes, while short-chain monomethyl alkanes were mainly influenced by the aphid host (figure 2). Meanwhile, our behavioural studies revealed no effects of host environment on ant aggression (figure 3), but stronger evidence for genotypic effects (figure 4 and table 1; electronic supplementary material, figure S3 and table S3), which further suggests that genetically influenced features of the CHC profile may play an important role in mediating these interactions.

The first study to examine chemical mimicry by Lysiphlebus reported that wasp (Lysiphlebus cardui) and aphid CHC profiles were most similar within the mass range of C₂₃ to C₃₀ [19]. The current data confirm these findings, but also suggest that the striking similarity in short-chain methyl alkanes is at least partially based on compounds acquired from the host, while the composition of linear alkanes and long-chain methyl alkanes is influenced by wasp genotype, which appears to have stronger effects on ant behaviour in the present study. Linear alkanes are ubiquitous in insects and were previously suggested to be of minor importance in communication [58]; however, subsequent work has shown that L. niger ants use these compounds, for example in distinguishing between suitable and unsuitable aphid mutualists [23], and that ants themselves have many long-chain methyl-branched alkanes on their cuticles [59], which they use as nestmate recognition cues [59].

An important goal for future studies is to link the observed variation in wasp CHC profiles with underlying functional genetic variation. Shared biosynthetic pathways within wasps, and resulting covariation of CHC compounds, could limit the extent to which wasps can adapt their CHCs to host profiles. Although the pathways underlying CHC biosynthesis are known [60] and efforts to elucidate the genetic bases of CHC biosynthesis are ongoing [36,61,62], we do not currently understand the genetic bases of variation in relative CHC amounts in Lysiphlebus wasps. However, recent comparative analyses of the annotated genomes of L. fabarum and Aphidius ervi (a non-mimetic parasitoid with an overlapping host range) [63] have yielded interesting results, including missing desaturase genes in L. fabarum which may explain the absence of alkenes in their CHC profiles, and thus increased chemical similarity between L. fabarum and its host aphids, which also lack alkenes.

As discussed, the relative contributions of genotype and host environment to relevant biochemical traits can have important implications for understanding the potential role of chemical mimicry in driving host specialization. While a few previous studies have explored chemical mimicry in parasitoids [64–67], questions relating to host specialization have primarily been addressed in social parasites that use CHCs to exploit their hosts. These include systems based on environmentally acquired compounds that do not lead to host specialization [6,68], as well as other cases in which chemical mimicry has been shown to drive specialization [12,13,69,70]. There is also evidence that specialization can be driven by selection for other forms of crypsis, as observed in stick insects which experience strong selection to match the colour patterns of different host plants [71]. The current findings reveal effects of both genotype and host environment on Lysiphlebus CHC profiles, but suggest that CHC variability between genotypes has a greater influence on ant aggression. This, in turn, suggests that selection favouring enhanced chemical mimicry could play a role in host specialization [29] and may therefore be relevant to the genetic divergence observed among Lysiphlebus wasps collected on different aphid hosts [28]. However, elucidating the significance of this role will require further investigation, preferably involving potentially co-evolved wasp and aphid lines sampled from the field.

3. Material and methods

(a) . Cross-fostering experiment

To determine the genetic and host influence on wasp CHCs, we reared four asexual lines of L. fabarum wasps on three different host aphid species. All wasp lines were collected in Switzerland in different years and host aphid species: line 1 from Aphis hederae (2017), line 2 from Aphis fabae fabae (2007), line 3 from Aphis fabae cirsiiacanthoides and line 4 from Aphis ruborum (both 2009). Each line was initiated with a single asexual female (monoclonal) and subsequently reared on susceptible A. fabae fabae aphids on Vicia faba bean plants until we used them for the experiment. Additionally, we collected and reared three common European aphid species that are suitable hosts for L. fabarum: (1) Aphis fabae cirsiiacanthoides on Helianthus annuus plants, (2) Aphis urticata on Urtica dioica plants and (3) Aphis gossypii on Cucumis sativus plants. We established monoclonal lines from field samples collected in June and July 2017 in northeastern Switzerland. To ensure that aphids were suitable hosts for parasitoid wasps, we tested them for the presence of the protective endosymbiont Hamiltonella defensa using diagnostic PCR (described in Vorburger et al. [72]). We reared all aphids and wasps under summer-like conditions (22°C, 16 h light, 8 h dark).

In our cross-fostering experiment (overview in electronic supplementary material, figure S4), we reared L. fabarum lines 1–4 on all three host aphid species (A. fabae cirsiiacanthoidis, A. gossypii, A. urticata) in a complete factorial design. We replicated each of the 12 combinations ten times. Each replicate consisted of one aphid-infested plant (H. annuus, U. dioica or C. sativus in a 9 cm pot with standard potting soil) covered with a cellophane bag. We started the aphid colonies with ten adult A. urticata on four-week-old U. dioica plants, and with five adult A. fabae cirsiiacanthoides and A. gossypii on two-week-old H. annuus and C. sativus plants, respectively. We allocated the bagged plants randomly to 30 trays in the same climate chamber (22°C, 16 h light, 8 h dark). Six days later, we collected, pooled and froze eight adult aphids from every replicate for CHC analysis. The following day, we added between six and 14 L. fabarum wasps (depending on the availability of different wasp lines) to the aphid colonies. Trays 21–30 were set up two weeks after trays 1–20 owing to the limited number of wasps available at the start. However, we kept all trays in the same climate chamber and used the same aphid source colonies. Two weeks after adding the first wasps, the next wasp generation emerged. We collected and pooled five wasps from every replicate for CHC analysis and all surplus wasps for usage in the behavioural experiments. We stored all insect samples at −80°C until we extracted the CHCs or used them for behavioural experiments.

By rearing asexual lines of parasitoids on asexual hosts, we capture only a restricted amount of the natural variation present in the study species, and we exclude more specialized wasp lines that may exhibit stronger chemical differentiation. However, this approach minimizes within-group variation and is therefore ideal for quantifying the relative contributions of wasp genotype and host environment on specific features of the wasp CHC profile.

(b) . Chemical analysis of insect cuticular hydrocarbons

CHC analysis follows the protocol in Hertaeg et al. [25]. In short, we thawed and air-dried the frozen insects for approximately 10 min before we did three consecutive 5 min immersions in 200 µl of hexane. We collected the resulting 600 µl of crude extract in a clean vial and applied it onto a 0.1 g SiOH column (silica gel 60, 230–400 mesh ASTM, particle size 0.04–0.063 mm, Fluka) to obtain only the nonpolar fractions. Then we eluted the CHCs from the column with 1 ml of hexane before we dried the samples under a gentle flow of nitrogen to remove remaining volatile compounds. Next, we re-suspended the samples, transferred them into a low-volume glass insert, dried them again and re-suspended them in 24 µl of hexane with 4 ng µl⁻¹ of nonyl acetate as an internal standard. We analysed 2 µl of each sample on an Agilent GC-MS (Agilent 7890B/5977A GC-MSD (EI), Agilent Technologies AG) equipped with a DB-1 silica capillary column (30 m × 0.25 mm internal diameter × 0.5 µm film thickness, Agilent Technologies AG). We used helium as carrier gas at a constant flow rate of 2 ml min⁻¹. The inlet temperature was set to 250°C and the split/splitless injector to pulsed splitless mode. We heated the column with the following programme: 60°C for 2 min, 60–200°C at a rate of 60°C min⁻¹, 200–250°C at a rate of 8°C min⁻¹ and 250–320°C at a rate of 4°C min⁻¹, followed by 320°C for 10 min. Compounds were detected with an electron impact single quadrupole mass spectrometer (70 eV), which allowed us to collect data for both mass spectra and flame ionization detection (FID) simultaneously. We later analysed the data using Mass Hunter Software (Agilent technologies) and performed FID-based quantification. We identified the linear n-alkanes by comparing their mass spectra and retention times to a C8–C40 alkane calibration standard (Supelco, USA). To identify the methyl-branched alkanes, we used Kovats retention indices and characteristic ions [73,74]. Across all samples, we detected 38 different compounds, including n-alkanes, mono-, di- and trimethyl alkanes.

(c) . Behavioural experiment with ants

We conducted two behavioural experiments to investigate how Lasius niger ants react to host-acquired or genetically influenced components of wasp CHC profiles. Lasius niger, one of the most common ant species in central Europe, frequently tends a multitude of different aphid species. For the first experiment, we used four and for the second experiments five ant colonies (queenright with 50–100 workers). We purchased them from Antshop Switzerland and maintained them under a 16 h light, 8 h dark cycle at 22°C in open plastic boxes (32.5 × 26.5 × 15 cm) with a plaster floor, a test tube nest (15 × 1.8 cm, half filled with water and plugged with cotton wool), and a strip of Teflon PTFE DISP30 Fluoropolymer Dispersion (Chemours) to prevent ants from escaping. We provided the ants with water and 10% honey water ad libitum, and mealworms (Tenebrio molitor) twice weekly. Before each experiment, we starved the ants for 4 days to raise their motivation to forage.

In our first behavioural experiment, we tested whether host-acquired compounds affect the chemical mimicry of L. fabarum wasps. We used only wasp line 1, which we selected because it was a good mimic and caused little ant aggression in preliminary trials. We presented freeze-killed wasps (reared on either A. fabae cirsiiacanthoidis, A. urticata or A. gossypii) to ants tending the same or a different aphid species from the one that wasps developed in. We started aphid colonies on small, 2–3-week-old Cynara scolymus (instead of H. annuus), C. sativus and U. dioica plants potted in 50 ml Falcon tubes with holes for water drainage. Three to four days later, we introduced these small aphid-infested plants into the ant boxes. Before starting the observations, we waited 1 h for the ants to start tending the aphids and thawed wasp samples only a few minutes before placing them among the ant-tended aphids. During the 10 min long observations, we used the Observer XT 12 Software (Noldus Information Technology, Wageningen, The Netherlands) to record the time ants spent antennating the wasp and noted whether they attacked the wasp and carried it away to their nest. We did 20 replicates of every wasp–background combination and additionally tested eight A. colemani females in every aphid background as negative (non-mimicry) controls. Aphidius colemani are sexual parasitoids frequently used as biocontrol agents that do not mimic aphid CHCs. We purchased them from Andermatt Biocontrol (Grossdietwil, Switzerland) and froze them right after they emerged. Wasp samples that were not contacted by any ant during these 10 min were excluded from the analysis. All ant colonies were placed in the same climatized chamber (22°C, 16 h light, 8 h dark), where they always tended the same aphid species for two consecutive days. Each ant colony was used equally often and given the same wasp treatments. However, to avoid systematic carryover or learning effects, we varied the order of the aphid species that the ants were tending and the sample order for every ant colony.

In the second behavioural experiment, we tested how ants react to the genetic differences in wasp CHCs. We used freeze-killed wasps from lines 1–4 that developed in A. fabae cirsiiacanthoidis or A. gossypii aphids and presented them to ants that were tending the same aphid species as the one the wasps developed in (either A. fabae cirsiiacanthoidis on Cynara scolymus or A. gossypii on C. sativus). We did 25 replicates of every wasp line–aphid background combination, including 25 female A. colemani wasps in each background as negative control. Each ant colony was used equally often and was presented with equal numbers of samples from the same treatments. We alternated the aphid background that ant colonies were tending every day. The plant and aphid colony preparation, as well as the procedure of the behavioural observations, were identical to those described for the first behavioural experiment. Owing to unforeseen results in the A. gossypii background, we did this experiment twice.

All field collections and experiments with insects and plants complied with the ETH Zürich Guidelines for research integrity RSETHZ 414.

(d) . Statistical analyses

(i) . Analysis of aphid and wasp cuticular hydrocarbons

We performed all statistical analyses in R v. 4.0.3 [75]. First, we excluded all aphid samples from replicates from which no wasps emerged. Second, we converted the area of each of the 38 detected peaks in the chromatograms to their proportional contribution to the total peak area of every sample before we visualized them using discriminant analysis of principal components (DAPC) [76]. We used the optim.a.score function (adegenet 2.0.0 package [77]) to determine the number of principal components to use. In a first DAPC, we visualized differences between wasps grouped by host aphid. In a second one, we grouped the aphid data by aphid species, in a third one the wasp data by wasp line, and in a fourth DAPC the wasp data by wasp line and aphid host. Next, we centre-log-ratio (clr) transformed [78] the proportions of the 38 peaks and used all of them as response variables for permutational multivariate analyses of variance (PERMANOVA) [79] based on Euclidean distances. We performed a first PERMANOVA using the wasp data with wasp line, aphid host and their interaction as explanatory variables. In a second PERMANOVA we used the aphid data with aphid species as explanatory variable. To show significant pairwise differences between aphid species, wasps grouped by aphid species and wasps grouped by wasp line, we performed three pairwise PERMANOVAs using the pairwise.adonis2 function [80]. To investigate which factors influenced individual compounds in the wasp CHC profiles, we fitted a linear model for every compound with the proportion of the compound as response variable and wasp line, aphid host and their interaction as explanatory variables. Then, we ran an ANOVA type III followed by effect size calculations (η², EtaSq function from DescTools package [81]) of wasp line, aphid host and their interaction. We visualized the results in a heatmap.

(e) . Analysis of the ant behavioural experiments

(i) . Ant reaction towards host influenced differences in wasp cuticular hydrocarbons

To compare differences in the duration of antennation that wasps from line 1 reared on different host aphids received, we ran a linear mixed-effects model (lmer function from the lme4 package [82]) with total antennation time as response variable, host aphid and aphid background species as fixed effects, and ant colony and the day of experiment as random effects. To analyse differences in attack rate, we used a generalized linear mixed effects model (glmer function from the lme4 package) with attack as binary response variable, host aphid and aphid background species as fixed effects, and ant colony and day of the experiment as random effects. Both models were followed by an ANOVA Type II and we calculated pairwise differences (emmeans [83]) between wasps that developed in different host aphids and control wasps, and between aphid backgrounds.

(ii) . Ant reaction towards genetic differences in wasp cuticular hydrocarbons

Since the ant reactions varied considerably across aphid backgrounds but not between the two experiments, we analysed the antennation times and attack rates towards the different wasp lines separately for each background but pooled the data from both experiments. For the A. fabae cirsiiacanthoidis background data, we ran an lmer with square-root-transformed total antennation time as response variable, aphid line as fixed effect, and ant colony, day and experiment as random effects. To compare attack rates in the A. fabae cirsiiacanthoidis background we ran a Bayesian linear mixed-effects model (bglmer function from blme package [84]), since using a glmer was impossible owing to singularity issues. Instead, we used a bglmer with binomial distribution and a logit link function with attack rate as response variable, wasp line as fixed effect, and day, ant colony and experiment as random effects. We used the Wishart distribution (default) to impose a prior over the covariance of the random effects. The distribution of the total antennation times in the A. gossypii background was strongly skewed. Therefore, we ran a bglmer with a γ-distribution and an identity link function with total antennation times plus 1 as response variable, wasp line as fixed effect, and day of experiment and ant colony as random effects. To analyse the attack rates in the A. gossypii background we ran a bglmer model with binary distribution and a logit link function with attack rate as response variable, wasp line as fixed effect, and ant colony as random effect. All models were followed by an ANOVA Type II and calculations of pairwise differences (emmeans).

Ethics

This work did not require ethical approval from a human subject or animal welfare committee. All field collections and experiments with insects and plants complied with the ETH Zürich Guidelines for research integrity RSETHZ 414.

Data accessibility

The data and code that support the findings of this study are available on Dryad: https://doi.org/10.5061/dryad.1g1jwsv2t [85].

Supplementary material is available online [86].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

C.H.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing—original draft; C.V.: conceptualization, funding acquisition, methodology, supervision, writing—review and editing; C.M.D.M.: conceptualization, funding acquisition, methodology, project administration, resources, supervision, writing—review and editing; M.C.M.: conceptualization, funding acquisition, methodology, project administration, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This study was supported by a Research Grant from ETH Zürich (ETH-44-16-1).