Phytophagous insects tend to be either cryptic and solitary, or brightly colored and gregarious, as a defense against vertebrate predators. Here, we tested whether potent defensive chemicals produced de novo by larvae of Argidae and Pergidae sawflies have influenced the evolutionary relationship between larval appearance and levels of gregariousness. Phylogeny-based correlation analyses indicated only a weak trend for solitary species to be cryptic, and for gregarious ones to be conspicuous. Numerous Argidae were cryptic–solitary or conspicuous–gregarious, whereas most Pergidae were conspicuous–gregarious. Both families also included not truly gregarious but aggregated species, i.e. with individuals more evenly distributed on the host plant. By considering two specific morphological traits, predominant body coloration and contrasting spots on body, each one was (weakly) associated with appearance but none with gregariousness, which reflects the functional relevance of appearance as a whole. Furthermore, Argidae can display alternate appearances during successive larval instars. Finally, an independent contrasts test showed no obvious correlation between two major toxic peptides. Our results point towards diversely combined patterns of linked ecological traits in these insects. By assuming that warning coloration is more warranted against vertebrate than invertebrate predators, we suggest that the occurrence itself of toxins allowed this diversity via differing predator guilds and environmental factors, to which these insects were confronted during evolution.
Most insects feed on plants and are specialists1. They are also frequently defended against attacking predators by containing defensive chemicals, i.e. allomones2,3. Their capability to metabolize, store, and emit such compounds is directly linked to ecological, behavioral, morphological, and physiological adaptations4,5,6, the whole constituting a chemical defense strategy. Additionally, plants also contain deterrents and toxins, against the attack of phytophagous insects7,8. Thus, plant chemistry is potentially shaping the diet breadth and modulating the allomonal profile of phytophagous insects.
The biosynthetic origin of allomones should influence how ecological and defensive traits are articulated among each other. Insects rely for anti-predator defense on exogenous (i.e. sequestered) and/or endogenous (de novo-produced) compounds; exogenous ones are typically plant-derived and endogenous ones can be produced by endosymbiotic organisms9,10,11. At the level of plant–insect relationships, host shifts are expected to occur more frequently in insects with endogenous allomones, although a specialist species defended by plant-derived allomones can also shift from one host to another12,13,14,15. At the level of predator–prey interactions, the autonomy procured by endogenous allomones may impact life history traits related to anti-predator defensive strategies, but this assumption remains to be tested at a large phylogenetic scale.
For insects defended by endogenous chemicals, we can expect an increased likelihood not only of host shifts but also ‘defense shifts’, that is, switches between crypsis and conspicuousness, and across various levels of gregariousness. In insects, the chemical adaptation of unpalatability preceded the morphological adaptation of warning coloration as well as the behavioral adaptations of gregariousness, maternal care, and subsociality16,17,18,19. Most studies on the evolution of predator–prey interactions tend however to focus on the signaling of defenses towards birds20, although unpalatability itself is most probably driven by both vertebrate and invertebrate predators, as shown in leaf beetles21,22, tiger moths23, and sawflies24. In tritrophic approaches of ecosystems, birds preying on herbivores have been shown to top-down regulate the plant level25, and predatory insects, especially ants, act on plants in a same way26,27,28. This emphasizes indirectly that guilds of predators exert a high selective pressure, as plants do, on phytophagous insects.
The larvae of sawflies (Hymenoptera, Symphyta) offer an ideal opportunity to estimate the evolutionary mode of inherently intricate insect–plant and predator–prey interactions. Sawfly larvae generally feed on plant foliage whereas adults scarcely feed (e.g. on nectar). The three largest sawfly families are the Tenthredinidae with c. 5,700 species worldwide, the Argidae comprising 920 species, and the Pergidae 440 species29,30. In the Tenthredinidae, major defensive strategies are easy bleeding of deterrent hemolymph31,32 and emission of volatiles by eversible ventral glands33, but the group shows an overall high variety in life history traits related to defense24.
Here, we tested whether there is an evolutionary relationship between chemical protection, larval appearance and levels of gregariousness. To test our hypothesis, we used larvae in the sawfly families Argidae and Pergidae. The larvae of most analyzed species of these closely related groups34 contain hepta- and octapeptides35,36. These chemicals are unique among insects and, seemingly even in nature, because of a high proportion of D to L amino acids and due to the presence of phosphoserine35. They are lethal to vertebrates by hepatotoxicity37,38,39 and probably have deleterious effects also on invertebrates40. Their biosynthetic origin remains uncertain; some publications evoke the possible role of endosymbionts and at least exclude the host plant as a source of the peptides35. The exceptionality of the chemical structures also supports this conclusion. Thus, there are good evidences that the larvae basically rely on these endogenous toxins for defense.
Larvae of nearly all Argidae and Pergidae species feed on angiosperms, but some species rely on fungi, dead leaves, or aquatic ferns41,42,43,44. Both families are predominantly distributed in Asia, Australia, Africa, and the Neotropics29, and the hosts and habits remain unknown for many species. The general occurrence of endogenous allomones across these two related and relatively large lineages allowed studying the evolutionary consequences of the particular and potent defensive compounds on life history traits of these sawflies. The larvae are either cryptic or brightly colored, and they are distributed on their host plant by being solitary, aggregated (i.e., several individuals settled on one plant but not forming a dense group), or truly gregarious (i.e., forming a dense group)44,45.
Here we first infer the molecular phylogeny of Argidae and Pergidae based on DNA sequence data from two mitochondrial genes and a nuclear one, and then use the phylogeny to map host-plant and chemical data as well as the traits gregariousness and appearance. We assessed the effect of coding larval appearance as either cryptic versus conspicuous or based on two specific morphological traits, on the output of phylogenetic correlations. We also explore how peptide quantities are distributed across the reconstructed phylogenetic trees. The hypothesis is that, by generally occurring in these insects, the toxic peptides impacted the association between levels of gregariousness and appearance. We therefore discuss whether the commonness of the peptides influenced the evolution of predator–prey interactions.
Results and Discussion
Our Bayesian phylogenetic analysis confirms the reciprocal monophyly of the families Argidae and Pergidae (Figs 1 and 2) and strongly supports their sister-group relationship34,46. A previous study by Schmidt and Walter46 focused on the phylogeny of the Pergidae, and all of their studied taxa were used for the present analysis (Fig. 1). Major pergid subfamilies like the Pterygophorinae and Perginae came out as monophyletic, whereas other subfamilies appear to be paraphyletic, including the Perreyiinae with respect to the group that was formerly recognized as the Australian Euryinae, and, albeit with low branch support, the Acordulecerinae to the Perginae and the Australian Phylacteophaga (formerly Phylacteophaginae) that is nested within the Neotropical Acordulecerinae46. Within the Argidae, the Erigleninae genus Sericoceros comes out as the sister group of the remaining Argidae consisting of the Sterictiphorinae + Athermantinae + Arginae (Fig. 1). Within the latter subfamily, the tree indicates the monophyly of each of the genera Antargidium and Scobina, both being supported by high posterior probabilities. The tree also shows that Arge is paraphyletic with respect to the monophyletic genus Spinarge. By contrast, the position of Cibdela (Athermantinae) remains unclear because its placement within Arge is supported by low posterior probabilities.
Chemical defensive strategies
Several defense mechanisms against natural enemies are known in the Argidae and Pergidae. Some of these defenses are restricted to certain species groups, such as the oral discharge of an oily fluid from a pharyngeal diverticulum in Perginae47,48,49, and volatiles emitted by non-eversible ventro-abdominal glands in species of Arge and Acordulecera40,47. Other mechanisms that are putatively involved in defense include the sticky secretion oozing from flask-shaped latero-abdominal glands in Atomacera and Sterictiphora45,47, the fluid secreted from the coxal bases in Sericoceros50, the eversible dorsal glands in Austrosteenisia blackii51, and the caudal filaments in Lagideus podocarpus52. Still other traits related to defense are known in larvae of Schizocerella, Corynophilus and Phylacteophaga that are leaf-miners (although S. pilicornis shows also an externally feeding biotype), in those of Enjijus that are shoot borers44,53, and in many Argidae larvae that raise their abdomen permanently or upon disturbance.
The most striking and regularly encountered defensive strategy among both Argidae and Pergidae larvae is to contain toxic peptides. However, these compounds were neither detected in the sterictiphorines Sterictiphora geminata and Aproceros leucopoda, nor in the pergines Perga affinis and Pergagrapta polita36. Pergines use mandibular structures to mechanically sequester toxic leaf compounds as a defense against natural enemies49.
Our independent contrasts test revealed a significant cross-species correlation between the amounts of pergidin and lophyrotomin (r = 0.415, r2 = 0.173, P = 0.018; N = 32 ingroup species; Fig. 3). Among the species used in this test, however, C. janthina contains particularly high amounts of each peptide36, and statistical significance was lost when excluding this species from the test (r = 0.174, r2 = 0.030, P = 0.349; N = 31). Moreover, our stochastic mapping analyses indicated that peptide quantity was not associated with the other traits: diet breadth, gregariousness, predominant body coloration, contrasting spots on body, appearance, and host-plant diversity (Table 1). Thus, it seems that intraspecific quantities between peptides are phylogenetically unrelated, as are also peptide quantities and other defensive traits. It is surprising that peptide quantities were related to none of the three visual traits, since there are evidences that the level of toxicity is a predictor of warning coloration in ladybirds (Coccinellidae)54 as well as poison frogs (Dendrobatidae)55. Possibly this is related to the lower mobility of insect larvae compared to organisms like ladybirds and frogs. Vertebrate predators such as birds are more mobile than invertebrate predators like ants, and they may impose a higher predation pressure on mobile prey, thus linking levels of toxicity to strength of the visual signaling. However, the association between warning signal intensity and chemical defenses remains difficult to understand because it can be positive, negative, or absent depending on the studied system56.
Gregariousness and appearance
Among the 17 pairwise comparisons performed using stochastic mapping, only three yielded an overall statistical significance (uncorrected P = 0.008–0.046), but none of them remained significant after Holm’s sequential Bonferroni correction (Table 1; Supplementary Table S1). This general non-significance of the corrected results of D values would lead to the conclusion that none of the studied characters is associated with another. Cautiousness is even more basically requested when using methods of stochastic mapping of categorical characters, because in these methods, cases of within-clade pseudo-replication can be confounded with associations that are evolutionary independent57. The evolutionary association between appearance (cryptic or conspicuous) and levels of gregariousness (solitary or gregarious, respectively) is well documented in insects such as leaf beetles, lepidopterans17,18, and sawflies24,58,59. In a previous study focusing on tenthredinid sawflies, none of the phylogenetic correlations between predominant body coloration, contrasting spots on body, and other larval traits was found to be significant24. In our results, raw P values were only significant in the pairwise comparisons of appearance with gregariousness, predominant body coloration, and contrasting spots on body (Table 1); the two latter are morphological traits not always unambiguously linked to appearance18,60. Gregariousness was associated with none of the morphological traits (Table 1). Our results suggest that compared to appearance, the two specific morphological traits constitute a more accurate and neutral description of the visual pattern displayed by the insect, but they may be less adequate in reflecting the effective visual impact of the patterns on foraging vertebrate predators, and how this impact is amplified when the sawflies are gregarious. The statistical weakness of the gathered associations would advocate, however, for a general, alleviated explanation, that is, gregariousness and appearance evolved rather independently from each other in the Argidae and Pergidae. The mapping of traits on the phylogenetic tree (Fig. 2) revealed a discrepancy between the two families in that Argidae species are frequently cryptic–solitary or conspicuous–gregarious (i.e., aggregated or truly gregarious), whereas Pergidae are predominantly conspicuous–gregarious. The question why cryptic–solitary species are particularly rare in Pergidae remains puzzling. But one can conclude at least that the Argidae contributed more than the Pergidae in gathering the aforementioned (weakly) significant results.
Since a large majority of Argidae and Pergidae contain toxic peptides, any conspicuous species may be considered “aposematic” in the strict sense since a (chemical) defense is underlying the signaling. The next step is to understand why all the toxic species are not aposematic. Generally, insects are often preyed upon by insectivorous birds, which undoubtedly respond to visual signals and are eventually able of avoidance learning18. Insects are also the prey of numerous insectivorous insects for which detailed body patterns of the prey are less significant. Moreover, repellent volatiles can function as a primary defense, i.e. a signal to warn predators that rely on chemical cues during foraging and which is the case for many insects but also birds61. We can assume that the typical evolutionary route of a chemical defense strategy against predation by vertebrates led a chemical adaptation to be followed by warning coloration, itself preceding gregariousness. In contrast, such a route against invertebrate predators could have been shorter by ‘skipping’ the shift of appearance in that a chemical adaptation leads directly to gregariousness, if we assume, all other things being equal, that grouped individuals are better chemically defended than single ones. A predator should then face a higher total dose of distasteful compounds. In experiments not using predators but where sawfly larvae were repeatedly harassed to produce a chemical defense, however, group size decreased under high ‘attack’ intensity62. This supports the idea that in the evolution of insects, a conspicuous body coloration and the chemical defense itself are differently influenced by avian predators, invertebrate ones, and even more generally, by other natural enemies, host-plant quality as well as abiotic conditions60,62.
The assumption of a similarly important impact of vertebrate versus invertebrate predators implies the possible existence of prey insect species that are chemically defended but solitary and cryptic, or even gregarious without being necessarily conspicuous. Such species actually occur among sawflies. On the one hand, at least five Argidae species are known to contain toxic peptides while being solitary and cryptic36 (Fig. 2). On the other hand, gregarious and cryptic species do not occur at last larval instar, but some sawflies combine these trait states sequentially during their development, as known in a nematine and some diprionid sawflies33,63. However, intraspecific variation in levels of appearance is common in the Argidae, having been described in Arge captiva64, Arge enkianthus65, Arge pagana40, and Spinarge flavicostalis66. Thus, sawfly species containing toxic peptides seem to have a propensity for varying appearances during ontogeny. Although numerous Argidae and Pergidae larvae still remain to be described, the known life histories of the Argidae already include diverse combinations across and/or within species between the traits gregariousness and appearance (Fig. 2).
The patchy distribution of an herbaceous host plant can force an insect to be aggregated rather than solitary, an explanation not holding for insects living on larger plants. Our results highlight the fact that Argidae and Pergidae are quite often aggregated although most studied species feed on shrubs or trees (Fig. 2), which suggests that aggregation is really an adapted intermediate stage of gregariousness in these sawflies. Indeed, gregariousness was uncorrelated to host-plant size (Table 1). Since endogenous chemicals may allow rather frequent host shifts14, these would be further enhanced if the hosts belong to a diversified group containing many taxonomically and chemically related species. However, diet breadth and host-plant diversity were uncorrelated traits (Table 1), indicating that host shifts are governed by other factors than the biosynthetic origin of allomones in Argidae and Pergidae.
Predator–prey interactions at large paleo-geographical scale
The paradox of commonly occurring and undoubtedly toxic peptides apparently unconnected to complementary defensive traits can find some further explanations from the perspective of trends in habitat conditions and predation risks encountered by the sawflies, in geological times as well as in their recent history.
Approximately one third of the Argidae species known worldwide live in the Palearctic, whereas no single Pergidae species is recorded from this geographical region29. All studied Argidae species are from this region and they tend to be cryptic and solitary, or conspicuous and gregarious (i.e. aggregated or truly gregarious); by contrast, the Pergidae are rarely cryptic and solitary (Fig. 2) and we collected them mainly in Australia, although the family is also well represented in the Neotropics. Sericoceros is the sister-group of all other Argidae and Aulacomerus the one of all other Pergidae (Fig. 1). These two genera are only known from the Neotropics29, which may suggest that the Argidae–Pergidae split occurred in Gondwana, the Argidae subsequently extending their geographical distribution towards Laurasia (which includes the extant Palearctic). The split between the two families happened ≈ 153 million years ago, with further main splits at least in the Pergidae during the Cretaceous46. This is well before the adaptive radiation during the Paleocene–Eocene of ants67 and modern, terrestrial birds68,69, both groups including major taxa of predators of phytophagous insects4 including sawflies24,70. The modern birds diversified and spread in the Holarctic zone during the Paleocene or early Eocene71, although birds as terrestrial and arboreal insectivores are today the most diversified in tropical zones72. Ants are also at present especially diversified and abundant in the (sub)-tropics where they forage at all vegetation strata73,74, and, more generally, predation inflicted by invertebrates is pronounced at low latitudes75. Thus, both Argidae and Pergidae were confronted to vertebrate as well as invertebrate predators during longstanding geological times, but the Argidae fauna from the Palearctic may be comparatively more subjected to avian predation since the Paleocene–Eocene.
Open and heterogeneously structured habitats are more common in the temperate and subarctic zones of the Palearctic than in the (sub)-tropical zones of the Southern Hemisphere that are partly dominated by dense and multi-strata rain forests76. In larvae of the lepidopteran genus Papilio, the signal environment is an important determinant in the evolution of aposematic coloration, and such advertising signals are likely more efficient in open than closed environments77. Thus, our results may reflect a rather frequent need for Argidae, at least from the Palearctic, to be defended against avian predators and which is allowed by the chemical defense eventually leading to conspicuousness and a gregarious behavior. Comparatively, the defense strategy of the Pergidae would be less impacted by birds, the toxic peptides nevertheless allowing the sawflies to become gregarious in response to predation by invertebrates.
Phylogenetic correlations potentially reveal the existence of relevant ecological and/or functional associations across traits57. However, a trait in a prey can variably evolve under the selective pressures of several differing guilds of predators, conversely undermining the significance of associations between traits. In the Argidae and Pergidae, none of the tests of phylogenetic correlation remained significant after statistical correction, with only uncorrected results indicating gregariousness and two specific morphological traits to be associated with appearance. The emergence of each of the traits appearance and gregariousness requires a (chemical) defense78, a prerequisite fulfilled in most representatives of the two families. The widespread occurrence of original and potent peptides may have allowed these sawflies to adjust the ecological traits of conspicuousness and gregariousness in response to specific ecological conditions, and more in terms of predator–prey than insect–plant relationships. Although gregariousness rarely evolves without warning coloration in a prey lineage79 under the impact of vertebrate predators, we suspect the likelihood of a direct causal relationship linking chemical defense to gregariousness to be higher when considering invertebrate predators. To the extent that the evolution of Argidae and Pergidae is driven by both types of predators, the combined selective pressures can explain the weak association between conspicuousness and gregariousness. Moreover, peptide quantities were related to none of the three visual traits.
Plant-derived and endogenous chemical defenses of phytophagous insects do not differ in the magnitude of their effects on predators21. The availability of the endogenous chemicals used by the sawflies studied here may have contributed to the geographical expansion and radiation of these insects. Yet, such conclusions are subject to caution due to incompleteness of our knowledge about the biology and ecology of the sawflies, especially for species from the Southern Hemisphere and that remain rarely collected. Nevertheless, we identified multi-guild predation as a determinant in the evolution of these sawflies, probably in combination with environmental factors that differ between habitats at a large paleo-geographical scale.
Our phylogenetic analysis was based on 38 Argidae and 60 Pergidae species collected mainly in Europe, Japan, Australia, and South America (Supplementary Table S2). The exemplar species represent four out of the seven subfamilies of Argidae, and ten out of the twelve subfamilies of Pergidae. Worldwide, 39% of Argidae species belong to the genus Arge29, from which 22 species were included in our analysis. As outgroups, we sampled ten species from three sawfly families (Supplementary Table S2). Vouchers are kept at the Royal Belgian Institute of Natural Sciences, and SNSB - Zoologische Staatssammlung München.
DNA was extracted from a leg taken off specimens stored in ethanol or from dry mounted museum specimens using Qiagen DNeasy Tissue Kits according to the manufacturer’s protocol, except that lysis took several hours or overnight. Three gene regions were amplified: (1) the 5′ end of the mitochondrial cytochrome c oxidase 1 gene (COI; 658 bp), (2) a fragment of the mitochondrial 16 S rDNA gene (397 bp), and (3) the D2 loop of the nuclear 28 S rDNA gene (523 bp). For primers and sequencing protocols, see Schmidt and Walter46. Sequences were edited using Sequencher version 4.9 (Gene Codes Corp., Ann Arbor, MI, USA) and are available in GenBank (accession numbers KF318463–KF318519, KC567820–KC567882, KF445006–KF445070).
We aligned the ribosomal sequences (28S-D2 and 16 S rDNA) using the T-Coffee software package80, which allows incorporating information on secondary structure predicted using RNAplfold, which is part of the ViennaRNA package81. COI sequences were aligned with MUSCLE82. Bayesian inference was conducted with MrBayes version 3.2.283. The combined analyses employed separate, unlinked substitution models for each gene region, and distinct models for codon positions of the COI region. In addition, gaps were coded as binary characters using FastGap version 1.284. For nucleotides, a mixed substitution model with gamma-categorized rate variation was employed, whereas gaps were included as a separate partition of variable, gamma-corrected binary characters. Two independent runs with one cold and three heated chains were run for five million generations, after which the average standard deviation of split frequencies had reached values lower than 0.01. Trees were sampled every 1,000 generations and the suitability of a relative burn-in fraction of 0.25 was examined using Tracer 1.6, prior to calculating a Bayesian consensus tree including all compatible groups.
Character coding and chemical measurements
To infer the evolutionary history of ecological and defensive traits within Argidae and Pergidae, we gathered data on larval traits and host features for the species included in the phylogenetic tree. The data were categorized and coded into two to four states for each character (see Fig. 2).
To determine diet breadths, species-specific host plants were compiled from reliable literature sources (e.g.30,85,86). Data on larval gregariousness and appearance were extracted from standard works on sawflies30,45, specific works on Argidae (e.g.85, and literature therein), and unpublished observations and sources. Since gregariousness as well as appearance can vary from one larval instar to another (e.g.40,85), only the last instar preceding the eonymph stage was considered. Gregariousness was coded into three states: solitary, aggregated (i.e., larvae distributed on a plant, generally less than three per leaf), or truly gregarious (larvae on one leaf or several adjacent leaves).
In a previous phylogenetic analysis of tenthredinid sawflies24, larval appearance was indirectly studied via two morphological traits, predominant body coloration and the presence of contrasting spots on body. The predominant body coloration corresponded to the major color of the body, here being either green, white, yellow, or brown. Since the larvae feed on the foliage, green was set as one state, the three other body colors as another state because all three appear in contrast to the close environment. Spots on the body were considered in relation to their contrasting effect with the predominant body coloration; such spots were generally dark but could be bright and whitish. These two morphological traits can be coded more objectively than appearance per se, because determining whether a species is cryptic or conspicuous partly depends on factors such as the immediate visual environment (background) of larvae, larval body size, visual capabilities of the observer, etc.54,77,87,88,89,90,91. To assess the impact of the coding method on the results, we included here the two aforementioned morphological traits plus the more subjective one, appearance, as either cryptic or conspicuous. We judged the appearance on the basis of the general visual effect rendered by the larva settled on the foliage. Note that aposematism in the strict sense is linked to an underlying defense4, which was an attribute not considered when determining the appearance. Furthermore, the link between the two morphological traits and appearance is not always obvious. For instance, instead of rendering the prey more conspicuous, dark body marks can actually enhance crypsis, and their presence or absence can be related to other functions than defense18,60.
Besides the traits directly linked to the sawfly larvae, we included in our analyses two specific plant traits that may however influence larval defensive strategies. Data about host-plant size and diversity were obtained from standard floras (e.g.92). Host-plant diversity refers here to the maximum number of known (extant) species of a plant genus. If a sawfly species feeds on more than one plant genus, this number was then the sum of all the species in the corresponding plant genera.
Data on peptide quantities (in µg per larva) were available for 28 Argidae and eight Pergidae species36, whereas earlier literature sources35,93 unambiguously identified and quantified the peptides in only one argid and three pergid species. Cumulative quantities of the three major peptides, pergidin, 4-valinepergidin and lophyrotomin, were coded. No complete information exists about possibly differing toxicities of these peptides on a predator, except about the lethal dose (LD) for mice by intra-peritoneal injection: LD50 ≈ 10 mg pergidin/kg, and 2 mg lophyrotomin/kg35,94. Such toxicity tests clearly do not reflect natural predatory-prey interactions95. We cannot exclude that the two compounds strongly differ in their feeding deterrence and/or toxicity, knowing however that these hepta- and octapeptides are closely related35,36. We assumed therefore that the chemical defense efficiency of a species, all other things being equal, is proportional to the total amount of peptides it contains.
Ancestral-state reconstructions and phylogenetic correlation analyses
In our ancestral-state reconstructions, we took phylogenetic uncertainty into account by maximum-likelihood (ML) optimizing states of the focal ecological, morphological, and chemical characters across 500 trees sampled from the Bayesian posterior distribution. These trees were obtained by randomly subsampling 250 post-burnin trees from each of the two MrBayes runs using the sample.trees function of the package ape96 in R97. The ape package was also used for pruning the trees to include a single representative of each species for which states were known for at least three characters. ML optimization was done using Mesquite v. 3.4098, and the results were summarized on the nodes of the similarly pruned Bayesian maximum clade credibility (MCC) tree. Because some of the most polyphagous exemplar species use several plant orders and multistate taxa cannot be ML optimized in Mesquite, ancestral host use was inferred by simple maximum-parsimony optimization of plant orders across nodes of the MCC tree.
Correlations between characters and character states were inferred by Bayesian stochastic mapping99,100 in SIMMAP v. 1.5.2101. Only ingroup species were included in these analyses and, in order to accommodate phylogenetic uncertainty, all correlation analyses were performed across 500 pruned post-burnin trees (rescaled to a length of 1). Seventeen character-pair comparisons were selected out of the 28 possible ones among the eight focal traits (Supplementary Table S1; Table 1). This selection was guided by supposing an evolutionary: (i) independence between the occurrence of toxic peptides and specific visual traits (gregariousness and appearance), (ii) non-independence between appearance and morphological traits (predominant body coloration and contrasting spots on body), and (iii) independence between plant- and sawfly-related features. State-by-state associations between characters were evaluated based on the dij statistic, which measures co-occurrence of states i and j across branches in relation to the expectation under independent evolution100, and overall character correlations were measured using statistic D, which is the sum of the absolute values of individual dij’s between characters100. All characters were treated as unordered, and pairwise correlation analyses were configured using bias and rate priors determined using the MCMC-based approach of Schultz and Churchill102 in SIMMAP. For binary characters, best-fitting priors for the bias parameter (=beta distribution parameter α) and the overall evolutionary rate parameters (=gamma distribution parameters α and β), were determined based on the pruned MCC tree, which had been rescaled to a length of 1. MCMC runs were performed using the default number of distribution discretization categories (31 for the bias parameter and 60 for the rate parameters), cycles (100,000), sampling frequency (200), burnin (10,000), and upper rate bound (1,000). Results were extracted in R v. 3.3.297 using the sumprmcmc.r script provided in the SIMMAP installation package. In the case of multistate characters, we used an empirical prior for the bias parameter, while rate-parameter priors were determined as described for binary characters. The number of samples, prior draws, and predictive samples were set to 1, meaning that both observed and predictive sample sizes equaled 500 for each character pair.
Independent contrasts test
The major toxic peptides are pergidin and lophyrotomin in Argidae, whereas pergidin and 4-valinepergidin in Pergidae36. Such data about peptide contents were gathered using the last larval instar, and are expressed in absolute amounts of µg per individual. In the context of predator–prey interactions, an absolute amount of peptides well reflects the actual (maximum) amount to which a predator is confronted while attacking a prey item. The amount is potentially determined by factors of the sawfly, such as species, body size, population, etc.35,36. To analyze the distribution of defensive chemicals in relation to the phylogenetic position of species, Felsenstein’s103 independent contrasts method was applied on the published dataset36 of peptide quantities, using specifically the quantities of pergidin versus lophyrotomin. The peptide 4-valinepergidin was discarded due to the higher number of both chemical and genetic data available for Argidae (24 species) than for Pergidae (8). The tree used in these analyses was obtained by pruning the Bayesian MCC tree to include only species for which chemical data were available, and contrasts were calculated using the PDAP:PDTREE package v. 1.15104 in Mesquite.
The datasets generated during and/or analyzed during the current study are available from the corresponding author on request.
Jaenike, J. Host specialization in phytophagous insects. Annu. Rev. Ecol. Syst. 21, 243–273 (1990).
Whitman, D. W., Blum, M. R. & Alsop, D. W. In Insect Defenses: Adaptive Mechanisms and Strategies of Prey and Predators. (eds Evans, D. L. & Schmidt, J. O.) 289–351 (State University of New York Press, 1990).
Meinwald, J. & Eisner, T. The chemistry of phyletic dominance. Proc. Natl. Acad. Sci. USA 92, 14–18 (1995).
Pasteels, J. M., Grégoire, J.-C. & Rowell-Rahier, M. The chemical ecology of defense in arthropods. Annu. Rev. Entomol. 28, 263–289 (1983).
Hoy, C. W., Head, G. P. & Hall, F. R. Spatial heterogeneity and insect adaptation to toxins. Annu. Rev. Entomol. 43, 571–594 (1998).
Dettner, K. In Insect Molecular Biology and Ecology. (ed. Hoffmann, K. H.) 39–93 (Taylor & Francis, 2015).
Cseke, L. J. et al. Natural Products from Plants. (Taylor & Francis, 2006).
Agrawal, A. A. Macroevolution of plant defense strategies. Trends Ecol. Evol. 22, 103–109 (2007).
Kellner, R. L. L. Stadium-specific transmission of endosymbionts needed for pederin biosynthesis in three species of Paederus rove beetles. Entomol. Exp. Appl. 107, 115–124 (2003).
Laurent, P., Braekman, J.-C., Daloze, D. & Pasteels, J. Biosynthesis of defensive compounds from beetles and ants. European J. Org. Chem. 2003, 2733–2743 (2003).
Opitz, S. E. W. & Müller, C. Plant chemistry and insect sequestration. Chemoecology 19, 117–154 (2009).
Termonia, A., Hsiao, T. H., Pasteels, J. M. & Milinkovitch, M. C. Feeding specialization and host-derived chemical defense in Chrysomeline leaf beetles did not lead to an evolutionary dead end. Proc. Natl. Acad. Sci. USA 98, 3909–3914 (2001).
Kergoat, G. J., Delobel, A., Fédière, G., Le Rü, B. & Silvain, J.-F. Both host-plant phylogeny and chemistry have shaped the African seed-beetle radiation. Mol. Phylogenet. Evol. 35, 602–611 (2005).
Engler-Chaouat, H. S. & Gilbert, L. E. De novo synthesis vs. sequestration: negatively correlated metabolic traits and the evolution of host plant specialization in cyanogenic butterflies. J. Chem. Ecol. 33, 25–42 (2007).
Futuyma, D. J. & Agrawal, A. A. Macroevolution and the biological diversity of plants and herbivores. Proc. Natl. Acad. Sci. USA 106, 18054–18061 (2009).
Sillén-Tullberg, B. Evolution of gregariousness in aposematic butterfly larvae: a phylogenetic analysis. Evolution 42, 293–305 (1988).
Guilford, T. In Insect Defenses: Adaptive Mechanisms and Strategies of Prey and Predators. (eds Evans, D. L. & Schmidt, J. O.) 23–61 (State University of New York Press, 1990).
Ruxton, G. D., Sherratt, T. N. & Speed, M. P. Avoiding Attack. The Evolutionary Ecology of Crypsis, Warning Signals, and Mimicry. (Oxford University Press, 2004).
Chaboo, C. S., Frieiro-Costa, F. A., Gómez-Zurita, J. & Westerduijn, R. Origins and diversification of subsociality in leaf beetles (Coleoptera: Chrysomelidae: Cassidinae: Chrysomelinae). J. Nat. Hist. 48, 2325–2367 (2014).
Skelhorn, J. & Ruxton, G. D. Ecological factors influencing the evolution of insects’ chemical defenses. Behav. Ecol. 19, 146–153 (2007).
Zvereva, E. L. & Kozlov, M. V. The costs and effectiveness of chemical defenses in herbivorous insects: a meta-analysis. Ecol. Monogr. 86, 107–124 (2016).
Zvereva, E. L. et al. Defence strategies of Chrysomela lapponica (Coleoptera: Chrysomelidae) larvae: relative efficacy of secreted and stored defences against insect and avian predators. Biol. J. Linn. Soc. 124, 533–546 (2018).
Rojas, B. et al. How to fight multiple enemies: target-specific chemical defences in an aposematic moth. Proc. R. Soc. B Biol. Sci. 284:20171424 (2017).
Boevé, J.-L., Blank, S. M., Meijer, G. & Nyman, T. Invertebrate and avian predators as drivers of chemical defensive strategies in tenthredinid sawflies. BMC Evol. Biol. 13:198 (2013).
Mäntylä, E., Klemola, T. & Laaksonen, T. Birds help plants: a meta-analysis of top-down trophic cascades caused by avian predators. Oecologia 165, 143–151 (2010).
Way, M. J. & Khoo, K. C. Role of ants in pest managment. Annu. Rev. Entomol. 37, 479–503 (1992).
Dyer, L. A. A quantification of predation rates, indirect positive effects on plants, and foraging variation of the giant tropical ant. Paraponera clavata. J. Insect Sci. 2, 18 (2002).
Rosumek, F. B. et al. Ants on plants: a meta-analysis of the role of ants as plant biotic defenses. Oecologia 160, 537–549 (2009).
Taeger, A., Blank, S. M. & Liston, A. D. World Catalog of Symphyta (Hymenoptera). Zootaxa 2580, 1–1064 (2010).
Schmidt, S. & Smith, D. R. Pergidae of the World – An online catalogue of the sawfly family Pergidae (Insecta, Hymenoptera, Symphyta). World Wide Web electronic publication. Available at: http://pergidae.snsb-zsm.de. (Accessed: 25th July 2016) (2018).
Boevé, J.-L. & Schaffner, U. Why does the larval integument of some sawfly species disrupt so easily? The harmful hemolymph hypothesis. Oecologia 134, 104–111 (2003).
Boevé, J.-L. In Functional Surfaces in Biology. Little Structures with Big Effects , Volume 1 (ed. Gorb, S. N.) 31–43 (Springer Netherlands, 2009).
Boevé, J.-L. & Heilporn, S. Secretion of the ventral glands in Craesus sawfly larvae. Biochem. Syst. Ecol. 36, 836–841 (2009).
Malm, T. & Nyman, T. Phylogeny of the symphytan grade of Hymenoptera: new pieces into the old jigsaw(fly) puzzle. Cladistics 1–17, https://doi.org/10.1111/cla.12069 (2014).
Oelrichs, P. B. et al. Unique toxic peptides isolated from sawfly larvae in three continents. Toxicon 37, 537–544 (1999).
Boevé, J.-L., Rozenberg, R., Shinohara, A. & Schmidt, S. Toxic peptides occur frequently in pergid and argid sawfly larvae. PLoS One 9(8), e105301 (2014).
Dadswell, L. P., Abbott, W. D. & McKenzie, R. A. The occurrence, cost and control of sawfly larval (Lophyrotoma interrupta) poisoning of cattle in Queensland 1972-81. Aust. Vet. J. 62, 94–97 (1985).
Dutra, F., Riet-Correa, F., Mendez, M. C. & Paiva, N. Poisoning of cattle and sheep in Uruguay by sawfly (Perreyia flavipes) larvae. Vet. Hum. Toxicol. 39, 281–286 (1997).
Soares, M. P. et al. Experimental intoxication by larvae of Perreyia flavipes Konow, 1899 (Hymenoptera: Pergidae) in pigs and some aspects on its biology. Toxicon 39, 669–678 (2001).
Petre, C.-A., Detrain, C. & Boevé, J.-L. Anti-predator defence mechanisms in sawfly larvae of Arge (Hymenoptera, Argidae). J. Insect Physiol. 53, 668–675 (2007).
Smith, D. R. A synopsis of the sawflies (Hymenoptera: Symphyta) of America south of the United States: Argidae. Mem. Am. Entomol. Soc. 39, 1–201 (1992).
Naumann, L. D. & Balciunas, J. K. A sawfly larva feeding on an aquatic fern (Hymenoptera: Symphyta: Pergidae). Aust. Entomol. 24, 39–47 (1997).
Flores, C., Ugalde, J., Hanson, P. & Gauld, I. In Hymenoptera: Evolution, Biodiversity and Biological Control. (eds Austin, A. D. & Dowton, M.) 258–266 (CSIRO Publishing, 2000).
Schmidt, S. & Smith, D. R. An annotated systematic world catalogue of the Pergidae (Hymenoptera). Contributions of the American Entomological Institute 34(3), 1–207 (2006).
Lorenz, H. & Kraus, M. Die Larvalsystematik der Blattwespen (Tenthredinoidea und Megalodontoidea). (Akademie-Verlag, 1957).
Schmidt, S. & Walter, G. H. Young clades in an old family: Major evolutionary transitions and diversification of the eucalypt-feeding pergid sawflies in Australia (Insecta, Hymenoptera, Pergidae). Mol. Phylogenet. Evol. 74, 111–121 (2014).
Maxwell, D. E. The comparative internal larval anatomy of sawflies (Hymenoptera: Symphyta). Can. Entomol. 87, 1–132 (1955).
Morrow, P. A., Bellas, T. E. & Eisner, T. Eucalyptus oils in the defensive oral discharge of Australian sawfly larvae (Hymenoptera: Pergidae). Oecologia 24, 193–206 (1976).
Schmidt, S., McKinnon, A. E., Moore, C. J. & Walter, G. H. Chemical detoxification vs mechanical removal of host plant toxins in Eucalyptus feeding sawfly larvae (Hymenoptera: Pergidae). J. Insect Physiol. 56, 1770–1776 (2010).
Smith, D. R. In The Hymenoptera of Costa Rica. (eds Hanson, P. E. & Gauld, I. D.) 157–177 (Oxford University Press, 1995).
Schmidt, S., Walter, G. H., Grigg, J. & Moore, C. J. In Recent Sawfly Research: Synthesis and Prospects. (eds Blank, S. M., Schmidt, S. & Taeger, A.) 173–193 (Goecke & Evers, 2006).
Boevé, J.-L., Marín-Armijos, D. S., Domínguez, D. F. & Smith, D. R. Sawflies (Hymenoptera: Argidae, Pergidae, Tenthredinidae) from southern Ecuador, with a new record for the country and some ecological data. J. Hymenopt. Res. 51, 55–89 (2016).
Gorske, S. F. & Sell, D. K. Genetic differences among purslane sawfly biotypes. J. Hered. 67, 271–274 (1976).
Arenas, L. M., Walter, D. & Stevens, M. Signal honesty and predation risk among a closely related group of aposematic species. Sci. Rep. 5, 1–12 (2015).
Summers, K. & Clough, M. E. The evolution of coloration and toxicity in the poison frog family (Dendrobatidae). Proc. Natl. Acad. Sci. USA 98, 6227–6232 (2001).
Summers, K., Speed, M. P., Blount, J. D. & Stuckert, A. M. M. Are aposematic signals honest? A review. J. Evol. Biol. 28, 1583–1599 (2015).
Maddison, W. P. & FitzJohn, R. G. The unsolved challenge to phylogenetic correlation tests for categorical characters. Syst. Biol. 64, 127–136 (2014).
Ross, J. A. Social saw-fly larvae. Vic. Nat. 46, 21 (1929).
Carne, P. B. Primitive forms of social behaviour, and their significance in the ecology of gregarious insects. Proc. Ecol. Soc. Aust. 1, 113–141 (1966).
Kingsolver, J. G. Butterfly thermoregulation: organismic mechanisms and population consequences. J. Res. Lepid. 24, 1–20 (1985).
Guilford, T., Nicol, C., Rothschild, M. & Moore, B. P. The biological roles of pyrazines: evidence for a warning odour function. Biol. J. Linn. Soc. 31, 113–128 (1987).
Lindstedt, C. et al. Ecological conditions alter cooperative behaviour and its costs in a chemically defended sawfly. Proc. R. Soc. B 285 (2018).
Lindstedt, C., Huttunen, H., Kakko, M. & Mappes, J. Disengtangling the evolution of weak warning signals: high detection risk and low production costs of chemical defences in gregarious pine sawfly larvae. Evol. Ecol. 25, 1029–1046 (2011).
Shinohara, A., Hara, H. & Kim, J. The species-group of Arge captiva (Insecta, Hymenoptera, Argidae). Bull. Natl. Museum Nat. Sci. Ser. A (Zoology), Tokyo 35, 249–278 (2009).
Hara, H. & Shinohara, A. Arge enkianthus n. sp. (Hymenoptera, Argidae) feeding on Enkianthus campanulatus in Japan. Bull. Natl. Museum Nat. Sci. Ser. A (Zoology), Tokyo 38, 21–32 (2012).
Shinohara, A., Kojima, H. & Hara, H. New host plant records and life history notes on Spinarge flavicostalis (Hymenoptera: Argidae) in Japan. Bull. Natl. Museum Nat. Sci. Ser. A (Zoology), Tokyo 39, 185–191 (2013).
Wilson, E. O. & Hölldobler, B. The rise of the ants: A phylogenetic and ecological explanation. Proc. Natl. Acad. Sci. USA 102, 7411–7414 (2005).
Brodkorb, P. In Avian biology, Volume I. (eds Farner, D. S., King, J. R. & Markes, K. C.) 19–53 (Academic Press, 1971).
Pacheco, M. A. et al. Evolution of modern birds revealed by mitogenomics: timing the radiation and origin of major orders. Mol. Biol. Evol. 28, 1927–1942 (2011).
Woodman, R. L. & Price, P. W. Differential larval predation by ants can influence willow sawfly community structure. Ecology 73, 1028–1037 (1992).
Claramunt, S. & Cracraft, J. A new time tree reveals Earth history’s imprint on the evolution of modern birds. Sci. Adv. 1, e1501005 (2015).
Kissling, W. D., Sekercioglu, C. H. & Jetz, W. Bird dietary guild richness across latitudes, environments and biogeographic regions. Glob. Ecol. Biogeogr. 1–13, https://doi.org/10.1111/j.1466-8238.2011.00679.x (2011).
Fittkau, E. J. & Klinge, H. On biomass and trophic structure of the Central Amazonian rain forest ecosystem. Biotropica 5, 2–14 (1973).
Andersen, A. N. Ant megadiversity and its origins in arid Australia. Austral Entomol. 55, 132–137 (2016).
Roslin, T. et al. Higher predation risk for insect prey at low latitudes and elevations. Science 356, 742–744 (2017).
Tuanmu, M.-N. & Jetz, W. A global, remote sensing-based characterization of terrestrial habitat heterogeneity for biodiversity and ecosystem modelling. Glob. Ecol. Biogeogr. 24, 1329–1339 (2015).
Prudic, K. L., Oliver, J. C. & Sperling, F. A. H. The signal environment is more important than diet or chemical specialization in the evolution of warning coloration. Proc. Natl. Acad. Sci. USA 104, 19381–19386 (2007).
Curley, E. A. M., Rowley, H. E. & Speed, M. P. A field demonstration of the costs and benefits of group living to edible and defended prey. Biol. Lett. 20150152, https://doi.org/10.1098/rsbl.2015.0152 (2015).
Tullberg, B. S. & Hunter, A. F. Evolution of larval gregariousness in relation to repellent defences and warning coloration in tree-feeding Macrolepidoptera: a phylogenetic analysis based on independent contrasts. Biol. J. Linn. Soc. 57, 253–276 (1996).
Wilm, A., Higgins, D. G. & Notredame, C. R-Coffee: a method for multiple alignment of non-coding RNA. Nucleic Acids Res. 36, e52 (2008).
Lorenz, R. et al. ViennaRNA Package 2.0. Algorithms for Molecular Biology. 6, 26, https://doi.org/10.1186/1748-7188-6-26 (2011).
Edgar, R. C. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32, 1792–1797 (2004).
Ronquist, F. & Huelsenbeck, J. P. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572–1574 (2003).
Simmons, M. P. & Ochoterena, H. Gaps as characters in sequence-based phylogenetic analyses. Syst. Biol. 49, 369–381 (2000).
Shinohara, A., Ibuki, S. & Hara, H. Arge meliosmae n. sp. (Hymenoptera, Argidae) feeding on Meliosma myriantha in Japan. Bull. Natl. Museum Nat. Sci. Ser. A (Zoology), Tokyo 37, 99–112 (2011).
Blank, S. M., Groll, E. K., Liston, A. D., Prous, M. & Taeger, A. ECatSym - Electronic World Catalog of Symphyta (Insecta, Hymenoptera). Digital Entomological Information, Müncheberg (2016).
Nilsson, M. & Forsman, A. Evolution of conspicuous colouration, body size and gregariousness: a comparative analysis of lepidopteran larvae. Evol. Ecol. 17, 51–66 (2003).
Merilaita, S. & Tullberg, B. S. Constrained camouflage facilitates the evolution of conspicuous warning coloration. Evolution 59, 38–45 (2005).
Tullberg, B. S., Merilaita, S. & Wiklund, C. Aposematism and crypsis combined as a result of distance dependence: functional versatility of the colour pattern in the swallowtail butterfly larva. Proc. R. Soc. B 272, 1315–1321 (2005).
Remmel, T., Davison, J. & Tammaru, T. Quantifying predation on folivorous insect larvae: the perspective of life-history evolution. Biol. J. Linn. Soc. 104, 1–18 (2011).
Stevens, M. & Ruxton, G. D. Linking the evolution and form of warning coloration in nature. Proc. R. Soc. B 279, 417–426 (2012).
Tutin, T. G. et al. Flora Europaea, Volumes 1–5. (Cambridge University Press, 2001).
Oelrichs, P. B., MacLeod, J. K., Seawright, A. A. & Grace, P. B. Isolation and identification of the toxic peptides from Lophyrotoma zonalis (Pergidae) sawfly larvae. Toxicon 39, 1933–1936 (2001).
Oelrichs, P. B. et al. Lophyrotomin, a new toxic octapeptide from the larvae of sawfly, Lophyrotoma interrupta. Lloyidia 40, 209–214 (1977).
Weldon, P. J. Poison frogs, defensive alkaloids, and sleepless mice: critique of a toxicity bioassay. Chemoecology 27, 123–126 (2017).
Paradis, E., Claude, J. & Strimmer, K. APE: Analyses of phylogenetics and evolution in R language. Bioinformatics 20, 289–290 (2004).
R Development Core Team. R: A language and environment for statistical computing. Available at: http://www.r-project.org/ (2016).
Maddison, W. P. & Maddison, D. R. Mesquite, a modular system for evolutionary analysis. Version 2.75, Available at: http://mesquiteproject.org (2011).
Nielsen, R. Mapping mutations on phylogenies. Syst. Biol. 51, 729–739 (2002).
Huelsenbeck, J. P., Nielsen, R. & Bollback, J. P. Stochastic mapping of morphological characters. Syst. Biol. 52, 131–158 (2003).
Bollback, J. P. SIMMAP: Stochastic character mapping of discrete traits on phylogenies. BMC Bioinformatics 7, 88 (2006).
Schultz, T. R. & Churchill, G. A. The role of subjectivity in reconstructing ancestral character states: A Bayesian approach to unknown rates, states, and transformation asymmetries. Syst. Biol. 48, 651–664 (1999).
Felsenstein, J. Phylogenies and the comparative method. Am. Nat. 125, 1–15 (1985).
Midford, P. E., Garland, T. & Maddison, W. P. PDAP:PDTREE: A translation of the PDTREE application of Garlands et al.’s Phenotypic Diversity Analysis Programs. (2010).
Chase, M. W. et al. An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants: APG IV. Bot. J. Linn. Soc. 181, 1–20 (2016).
We warmly thank Ewald Altenhofer, Stephan M. Blank, Eric Deulufeut, Erich Diller, Jan Grigg, Herbert R. Jacobson, Manfred Kraus, Christian Kutzscher, Mamoru Matsuki, Chris Lambkin, Andrew D. Loch, Anne-Marie McKinnon, José H. Pedrosa-Macedo, Serge Quilici†, Alexander Riedel, Henri Savina, Urs Schaffner, Olga Schmidt, Klaus Schönitzer, Jeff Skevington, M. Snižek, Noel Starick, Martin Steinbauer, H. Vargas, and Alexey G. Zinovjev for providing sawfly specimens. T.N. is supported by the Academy of Finland (grant #294466).
The authors declare no competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.