Introduction

Hyles is a successful hawk moth genus with a world-wide distribution. Molecular phylogenetic analyses have indicated that the genus evolved in the Neotropics (Hundsdoerfer et al. 2009) and experienced a Pleistocene burst of diversification in the Northern Hemisphere (Hundsdoerfer et al. 2017). Herbaceous plants of the Asphodelaceae, Elaeagnaceae, Ericacae, Euphorbiaceae, Nyctaginaceae, Onagraceae, Polygonaceae, Rubiaceae, and Zygophyllaceae are used as larval source of food (HOSTS database: http://www.nhm.ac.uk/our-science/data/hostplants/search/index.dsml; Hundsdoerfer et al. 2017; Pittaway 2018). Larvae of the three species which branch off first in the radiation (Hundsdoerfer et al. 2017; Fig. 1) are polyphagous (Hyles lineata, H. annei, H. euphorbiarum; Ureta and Donoso 1956). This also holds true for the two species of the Hawaiian radiation (H. perkinsi and H. wilsoni), as well as the Palearctic H. gallii and H. livornica. Larvae of most other species are specialised on specific host plant families. Some rely on plant genera that are known to contain no acutely toxic secondary compounds, such as H. centralasiae on Eremurus (de Freina and Geck 2014; Vanselow et al. 2016), for which Ali and Qaiser (2009) reported use of leaves as medicinal herbal remedy or H. hippophaes on Hippophae rhamnoides (Pittaway 2018), for which Ali et al. (2012) reported non-toxicity of a herbal antioxidant supplement prepared from the leaves (containing flavonoids, Zu et al. 2006). The larvae of Hyles nicaea, Hyles dahlii and the Hyles euphorbiae complex (HEC s. str.; Hundsdoerfer et al. 2011, 2017; Mende et al. 2016), however, are specialized on toxic Euphorbia (Euphorbiaceae; Pittaway 2018). Comparison of Euphorbia monophagy with phylogenetic relationships (Hundsdoerfer et al. 2009) suggests that this ecological trait evolved at least twice independently, since H. nicaea is not part of the HEC. The clade containing H. nicaea and the HEC includes many non-Euphorbia specialists (Hundsdoerfer et al. 2009). In the Western Palearctic, North Africa and Macaronesia, the most common food plants of H. nicaea and the HEC are E. characias, E. cyparissias, E. dendroides, E. myrsinites, E. nicaea, E. obtusifolia, E. paralias and E. segetalis, but host plants in the Onagracae and Vitacae are accepted by H. nicaea larvae as food plants in captivity. Hundsdoerfer et al. (2017) have postulated that using Euphorbia as host plant may have led to an increased speciation rate (and the emergence of at least five species) within the crown group, the HEC.

Fig. 1
figure 1

Phylogenetic tree of the genus Hyles taken from Hundsdoerfer et al. (2017) based on combined mitochondrial (1531 bp COI & 681 bp COII) and nuclear (772 bp EF1alpha) sequences of a single individual per species analyzed using BEAST; matrix lists species’ larval host food plant, dietary breadth and TPA sensitivity

Full size image

Leafy spurges in the Euphorbia subgenus Esula have a clear Eurasian centre of diversity (480 temperate Eurasian species; Peirson et al. 2014) and include all known natural Hyles food plants in the family Euphorbiaceae (numerous subgenera; Riina et al. 2013). The majority of New World Euphorbia diversity (32 species) occurs in a single section of the subgenus Esula, Tithymalus (Peirson et al. 2014). The sister lineage to the New World clade of Tithymalus, European weedy E. peplus, also occurs in the New World, but is not favoured as food plant of the HEC species H. euphorbiae (AKH, pers. obs.) in Europe.

Euphorbia plants contain diterpene esters with numerous harmful biological activities (for example tumour-promotion, cytotoxicity and skin irritation; overview in Shi et al. 2008). TPA (12-tetradecanoyl-phorbol-13-acetate; also called phorbol 12-myristate 13-acetate, PMA) is the most abundant biologically active phorbol diester of croton oil (Croton tiglium, Euphorbiaceae) and also the most irritant and co-carcinogenic compound (Hecker 1968a, b) in this oil. It is known to induce genes with a specifically conserved 9-bp motif in the promoter region (e.g. collagenase, stromelysin; Angel et al. 1987), and it is also present (as an isomeric form) in E. cyparissias (Ott and Hecker 1981). Hyles euphorbiae has been shown to be insensitive to high doses of this ‘standard’ phorbol ester (Hundsdoerfer et al. 2005). Although these larvae are conspicuous, indicating that they could be harmful for predators, the study indicated that the larvae metabolise TPA, rather than sequestering it as protection from predators. Several of the above-mentioned polyphagous species (H. gallii, H. livornica, H. lineata, H. perkinsi) feeding mainly on Rubiaceae and Onagraceae (as well as plants of many other families) are also able to use food plants of the genus Euphorbia (pers. obs. AKH, MAO; Danner et al. 1998; Pittaway 2018; HOSTS database, link above).

The primary aim of this paper is to examine the evolution of phorbol ester insensitivity in Hyles species, to better evaluate the importance of this trait as a possible key innovation causing accelerated speciation in the genus. The interpretation that Euphorbia feeding evolved at least twice independently could be incorrect as this metabolic ability could be plesiomorphic for the entire genus Hyles. The secondary loss of Euphorbia feeding has been identified previously (e.g. Hundsdoerfer et al. 2009), but an interesting question remains: whether non-Euphorbia feeders are nevertheless able to detoxify phorbol esters. Larvae of numerous Hyles species were fed with artificial diet containing TPA (H. annei, H. centralasiae, H. dahlii, H. euphorbiae, H. euphorbiarum, H. gallii, H. lineata, H. livornica, H. tithymali, H. vespertilio and H. zygophyllii). As a comparison, we also tested TPA sensitivity of a species in the closely related genus Hippotion (Hippotion celerio), in which no species feed on Euphorbia.

Additional questions concern the mechanism of tolerance of Hyles larvae to phorbol esters, i.e. how do larvae avoid intoxication? Hundsdoerfer et al. (2005) showed that no TPA is found in the integument upon oral consumption. Only 10–30% of the TPA consumed was found in the faeces; thus, 70–90% was metabolised. If injected, TPA was neither found in the integument nor in the haemolymph, but many (uncharacterised) metabolites were observed in the integument (Hundsdoerfer et al. 2005). However, questions remained about the mode of metabolism and the existence of other preventive measures by the organism. Recently, Barth et al. (2018) reported nearly 400 differentially expressed RNA transcripts involved in detoxification of TPA by H. euphorbiae. Several enzymes of the four classes involved in ‘drug metabolism’ cytochrome P450 (CYP), carboxylesterases (CES), glutathione S-transferases (GST) and lipases were found to be upregulated upon oral consumption of TPA by the larvae, corroborating metabolic detoxification as a mechanism to deal with this toxin.

Three possible mechanisms of preventing intoxication by TPA are under focus in this study: target site modification, physical barrier or metabolism (in this case CYPs).

In other insects which store toxic secondary metabolites, tolerance is achieved through (secondary compound) target site modification. For example, the binding site for cardiac glycosides has been changed in the monarch butterfly (Danaus plexippus) by two amino acid exchanges, so that binding of these compounds to the Na,K-ATPase is strongly reduced (Aardema et al. 2012; Dobler et al. 2012; Holzinger and Wink 1996). TPA is known to strongly activate the alpha form of the protein kinase C (PKC; cPKC and nPKC isozymes; Blumberg 1988; Nishizuka 1984) by redistribution of an inactive cytosolic form of the enzyme to the membrane (Ohno et al. 1990 and references therein). PKC is a serine/threonine kinase taking a central role in essential biochemical pathways in vertebrate as well as insect tissues (e.g. Ojani et al. 2016; Pakpour et al. 2013) by modulating several signal transduction systems. For an example, PKC phosphorylates epidermal growth factors, blocks the elevation of cAMP in mouse skin, influences the phosphatidylinositol turnover pathway and is involved in the expression of some oncogenes (Blumberg 1988 and references therein). Although there are other cellular targets of phorbol esters (Brose and Rosenmund 2002), phorbol ester insensitivity could at least partly be caused by lack of TPA binding to a modified PKC in tolerant Hyles species. We therefore compare the PKC alpha DNA and translated protein sequence of insensitive H. euphorbiae to that of the sensitive Hippotion celerio and Manduca sexta (Lepidoptera, Sphingidae) to find possible mutations. The second aim of this paper is thus to gain first insights into the possible importance of target site modification in Hyles TPA insensitivity.

An alternative possible way of avoiding intoxication by phorbol esters is via a physical intestinal barrier which prevents uptake in Hyles larvae. The peritrophic matrix (PM) lining the midgut epithelium in invertebrates plays an important role in preventing potentially damaging substances from being up taken into the body (Hegedus et al. 2009; Wang and Granados 2001). Calcofluor is a substance that inhibits PM formation (Wang and Granados 2000), thus enabling substances from the food bolus to freely cross the intestinal barrier. If larvae are TPA insensitive only due to the PM preventing uptake, they should become intoxicated without a PM, i.e. the addition of calcofluor to their diet would lead to decreased larval vigour or death. However, if calcofluor has no effect on TPA insensitivity, this suggests that the larvae are able to metabolise the toxin, either in the gut or within their tissues. The third aim of this paper is therefore to report on TPA tolerance in H. euphorbiae larvae lacking a PM, in order to assess the importance of this physical barrier.

The fourth and final aim of this work is an initial evaluation in-vivo of the possible role of cytochrome P450 monooxygenases (CYPs) as potential Hyles TPA metabolism detoxification enzymes. CYPs are well-studied enzymes of insect xenobiotic metabolism (Liu et al. 2015) and strong candidates for this function. Piperonyl butoxide (PBO) is an insecticide (Dove 1947) which is known to inhibit the activity of cytochrome P450 enzymes for several hours (Kinsler et al. 1990). Treatment of H. euphorbiae larvae with PBO prior to TPA consumption will therefore lead to intoxication (decreased larval vigour or death) if CYPs were sole key enzymatic pathways of detoxification.

Materials and methods

Evolution of TPA insensitivity in Hyles

L1-L4 larvae of H. annei, H. euphorbiarum, H. gallii, H. lineata, H. livornica and H. vespertilio were reared on cut shoots of non-toxic Epilobium species (E. angustifolium and unidentifiable natural hybrids). Hyles centralasiae larvae were reared on the flowers of potted plants of an Eremurus hybrid (probably the hybrid robustus × olgae), or cut inflorescence. Hyles zygophyllii was reared on Zygophyllum fabago. Hyles dahlii, H. euphorbiae and H. tithymali larvae were reared on cut shoots of Euphorbia segetalis and E. myrsinites in the laboratory. In L5 the larvae were fed an artificial diet (Hundsdoerfer et al. 2005). This diet was specially developed for hawk moths (Harbich 1994) and consists of an agar–agar mass, enriched with nutritional additives plus a small amount of dried larval food plant of the respective species to improve the taste (Epilobium angustifolium leaves, the above-mentioned Eremurus hybrid flowers or Euphorbia segetalis leaves). In the case of H. zygophyllii, a different diet without agar–agar or food plant powder was used. (The ingredients are: 30 ml water, 12 g wheat/spelt breadcrumbs, 3.3 g wheat germ, 2 g corn starch, 0.3 g brown sugar, 0.3 g vitamin C powder). When larvae had reached a minimal body weight of 2.6 g (2.1 g in the case of H. euphorbiae, since they tolerate more TPA and 1.5 g in the case of H. zygophyllii, since the species generally has smaller body sizes), they were used for the feeding experiment. Depending on the weight and feeding behaviour of the experimental larvae, they were offered 1–3 cubes of 0.3, 0.5 g or 1 g diet containing 0.3–2.0 mg TPA (Fluka, Sigma) each during a time interval of 4–5 h. The TPA was introduced into the diet by injecting a TPA solution (concentration 100 µg TPA / µl solvent DMSO) into the slightly melted, softened cube. The control larvae were fed cubes of diet mixed with DMSO only. The amount of TPA applied to the first larva of a non-Euphorbia feeding species was between 0.14 and 0.31 mg TPA/g body weight. The total amount of TPA applied per larva was adapted according to the tolerance of the first dose, but no larva was given more than a total of 2 mg TPA. A dose was seen as applied if the larva ingested the cube within max. 4 h. The TPA dose of the subsequent larvae depended on the reaction of the first larva. If the first larva of a species tolerated the TPA dose applied without losing vitality (no paralysis, normal feeding and excretion), the dose was increased in subsequent experiments, if not, it was decreased. Mortalities of larvae after TPA consumption were scored and sensitivity mapped on a recent, well-supported phylogenetic tree (Fig. 1; Hundsdoerfer et al. 2017; their Fig. 2) to discuss evolutionary aspects.

Fig. 2
figure 2

Deduced amino acid sequences of the C1 domain of the PKC alpha of H. euphorbiae and H. celerio, compared to those of B. mori (out-group; GB Acc. No. BAE17022.1) and M. sexta (GB Acc. No. Msex2.04738-RB). The conserved cysteine residues are marked by asterisks, residues in white on black are conserved in all species, in white on grey amongst most of them (e.g. due to missing data in one species)

Full size image

Target site modification

To investigate the possibility that target site insensitivity towards TPA might have evolved in Hyles euphorbiae, we sequenced the cysteine-rich C1 domain (CRD) of the protein kinase C alpha gene and compared it to that of the non-adapted out-group species H. celerio and of the closely related species M. sexta (Lepidoptera, Sphingidae; Msex2.04738-RB). Degenerated primers were developed based on the deduced amino acid sequences of Bombyx mori (Lepidoptera, Sphingidae; BAE17022.1), Apis mellifera (Hymenoptera, Apidae; XP_391874), Drosophila melanogaster (Diptera, Drosophilidae; NP_001097350), Aedes aegypti (Diptera, Culicidae; XP_001652409), and Tribolium castaneum (Coleoptera, Tenebrionidae; XP_969570.2). Two primer pairs that included the CRD were designed and predicted to yield PCR products of 500 to 650 bp (Table 1). Tissue samples of about 30 mg of H. euphorbiae and H. celerio preserved in RNA later (Ambion) were pulverised under liquid nitrogen and RNA extracted using the RNA tissue kit (Qiagen) according to manufacturer instructions. RNA was reverse transcribed to cDNA with Superscript II (Invitrogen) using random hexamers and a T17 primer. PCR proved to be most successful in touch down programs starting with an annealing temperature of 57 °C for 1 min and lowering the temperature by 0.3 °C in each of a total of 32 cycles followed by 10 more cycles with an annealing temperature of 40 °C. In cases where PCR products were too weak for direct sequencing they were reamplified with the same primer combination. PCR products were cycle sequenced with the Big dye kit (Amersham Bioscience) and analyzed on a 96 capillary automated sequencer (ABI; sequencing facility University Hospital Eppendorf, Hamburg). Visualisation and alignment of sequence fragments were performed with Sequencher 5.1 (GeneCodes Corporation), for translation to amino acid sequences and alignment with other species Mega 7 (Kumar et al. 2016), and Vector NTI was used (Thermo Fisher Scientific). All sequences have been submitted to Genbank (Acc.no. LR135166-LR135167).

Table 1 Primer sequences for the amplification of the PKC alpha C1 domain
Full size table

Physical barrier

Ten larvae of the Euphorbia-monophagous species H. euphorbiae and H. tithymali (each) were fed with artificial diet containing 1% calcofluor white. Upon dissection, the guts were indeed observed to appear ‘thinner’ than expected. After 24 h, five of these were subsequently fed with 2 mg TPA in artificial diet and the vitality of every individual was scored in three parameters (from 0 to 5 points) within a time interval of 4 h: (1) overall impression (0 points dead, 5 points for a strong and mobile larva showing the healthy reaction of clinging onto the finger upon handling), (2) feeding after treatment (0 points if no food and 5 points if the entire food was consumed), (3) defecation after treatment (0 points for no, 1 for very liquid, 4 for paste like and 5 points for normal faeces and 2 or 3 points for two conditions in between). The points were summed for every experimental set-up and converted to percentages (100% representing the maximum of 15 points per individual) and an average (from the five individuals) used to plot a bar chart for every experimental set-up of this and the next series of experiments.

The role of cytochrome P450 monooxygenases in metabolism

Ten larvae of H. euphorbiae and H. tithymali (each) were injected with 1.5 mM PBO (end concentration in larva 1 mM). After 24 h, five of these were subsequently fed with 2 mg TPA in artificial diet and the vitality scored by means of the same three parameters described above.

In addition, ten larvae of both species (each) that had consumed artificial diet containing 1% calcofluor white, and thus presumably had no more PM, were injected with PBO as above. Again, five of these were treated with 2 mg TPA after 24 h. The same vitality values were scored.

Results

Evolution of TPA insensitivity in Hyles

The experimental data of larval survivorship to TPA ingestion in the Hyles species tested are reported in Table 2. The total number of larvae available for experiments varied strongly between species. The majority of the breeding stocks were strong, and experiments could be replicated multiple times, with the exception of H. centralasiae which only provided two larvae. The control larva survived on the artificial diet but the TPA-fed one died of intoxication at 0.29 mg TPA/g body weight (Table 2). We also attempted to rear further species, such as non-Euphorbia feeding H. hippophaes, but unfortunately these rearings did not yield any testable larvae.

Table 2 Overview of TPA feeding experiments with mortalities (in numbers of larvae) and the maximal TPA doses applied and tolerated
Full size table

Neither a difference in acceptance of cubes with or without solvent (DMSO) nor a physiological reaction to cubes with solvent could be detected in the control larvae of any species. Nevertheless, in five species control larvae also died (Table 2): Hyles annei, H. dahlii, H. euphorbiarum, H. vespertilio and H. zygophyllii. Remarkably, in two of these, H. dahlii and H. euphorbiarum, the mortality rate of the TPA-fed larvae was 0%.

Experimental larvae consumed the cubes with a TPA concentration of up to 2 mg/g diet at different speeds. After a few bites into the TPA treated cubes of the artificial diet, larvae (of several species, H. euphorbiae, H. gallii and H. zygophyllii) stopped eating for several minutes. Thereafter, they appeared to consume the TPA cubes more avidly than the control larvae and the untreated ones. Some larvae consumed the cube very rapidly (most H. annei, H. euphorbiarum, H. dahlii, H. lineata and H. livornica) and the maximal dose of 2 mg per larva was reached (the fastest larva, a H. dahlii, ingested 2 mg TPA in one cube within about 2 h), whereas others did not consume the entire cube (H. centralasiae, H. annei, H. vespertilio, most H. zygophyllii, Hippotion celerio). The larvae of some species died within a day (H. centralasiae, Hippotion celerio, and the single H. lineata which died), but most deaths occurred after several days of physiological breakdown involving liquid defecation, cessation of feeding and eventual haemolymph loss.

Larvae of seven Hyles species survived the TPA treatment without loss (Hyles dahlii, H. euphorbiae, H. euphorbiarum, H. gallii, H. livornica and H. tithymali). Hyles centralasiae had a mortality rate of 100%, but this result is based on only one larva and cannot be considered statistically significant. Hyles lineata and H. zygophyllii had low mortality rates of 17 and 25%, respectively, based on six and three individuals, respectively (Fig. 1). Hyles annei and H. vespertilio and the out-group species Hippotion celerio had high mortality rates of between 70–85% (based on 5 individuals or more). The only H. vespertilio larva (of seven tested) that survived the treatment, did not consume the entire TPA containing cube of artificial diet, but the material consumed caused significant physiological breakdown. (It did not resume feeding and therefore lost weight continuously prior to pupation.)

The South American species H. annei showed one of the most complex reactions to TPA treatment. During their consumption of TPA, most TPA-fed H. annei larvae showed higher vigour and appetite than those within the control group. However, about 10 h after TPA consumption, these larvae were producing more moist faeces than those in the control group and were losing weight. Nevertheless, the larvae exposed to TPA continued to feed well and (re-)gained weight until they showed signs of wanting to pupate. However, nearly half of the 10 treated larvae failed to pupate. Four TPA-fed (0.2 mg/g body mass) pupated successfully. The two larvae that had consumed the highest TPA doses (0.5 mg/g body mass) managed to form pupae, but they were faulty with an open cuticle, leading to rapid black fungal infection and death. Since larvae showed no signs of acute intoxication, TPA appears to have been metabolised by the larvae. Nevertheless, the treatment led to lethal developmental changes in most individuals (70%). High doses of TPA appear to interfere with the process of metamorphosis. It should be noted though that the H. annei stock was not healthy: 60% of control larvae reared under laboratory conditions died in our laboratory and a further cohort reared at another location also failed to yield any healthy pupae (Heimo Harbich, pers. comm.).

Mapped onto a phylogenetic tree (see Fig. 1 with out-group omitted, taken from Hundsdoerfer et al. 2017), it becomes clear that TPA sensitivity is not correlated with relatedness. With a mortality rate of over 70%, the two species H. centralasiae (n = 1) and H. vespertilio (n = 7) are very sensitive to TPA, but they do not form a monophyletic group. Conversely, H. dahlii, H. euphorbiae, H. euphorbiarum, H. gallii, H. livornica and H. tithymali tolerate TPA without loss (0% mortality), but these species also do not form a clade. Of the three most basal species, H. euphorbiarum (0% mortality) and H. lineata (20%) show low mortalities. The three polyphagous species which include Euphorbia in their natural host plant spectrum were able to metabolise TPA efficiently: Hyles gallii and H. livornica are insensitive, whereas H. lineata has a low sensitivity (20% mortality).

Target site modification

According to the PCR results with degenerated primers and sequencing of PCR products, a single paralog of the classic PKC is present in Hyles euphorbiae just as in the case of Bombyx mori or Manduca sexta. The deduced amino acid sequences of the C1 domain of the PKC of Hyles euphorbiae and Hippotion celerio did not show any conspicuous differences among each other or compared to M. sexta and B. mori, the closest relatives for which a PKC gene sequence is available (Fig. 2). The conserved cysteine-rich motif that acts as a binding domain for TPA and other phorbol esters, CX2CX13CX2CX7CX7C, was present and repeated in tandem in both of these species. The corresponding sequence for H. euphorbiae only differed by one amino acid substitution, threonine instead of alanine, at position 117 in the sequence for M. sexta. Unfortunately, a corresponding clean sequence was not available for the TPA sensitive species Hippotion celerio. The only other substitution within the cysteine-rich domains was at position 111 which differentiated B. mori from the three sphingid species. It should be noted that these sphingids differ in TPA sensitivity, H. euphorbiae is insensitive, whereas Hippotion celerio and M. sexta are not (see Wink and Theile 2002). As a consequence, position 111 is thus unlikely to be of relevance for this difference in TPA sensitivity (Fig. 2).

Physical barrier

Control larvae of H. euphorbiae coped well with calcofluor in their diet (Table 3; Fig. 3), implying that artificial removal of the PM does not influence vigour in this species. Treatment with calcofluor and subsequently with TPA also did not lead to any losses and larvae remained nearly as vigorous as the control larvae (calcofluor only) or larvae treated with TPA alone.

Table 3 Experimental data of TPA, calcofluor and PBO treatments (a) of H. euphorbiae larvae (n = 5 for every group), summarised to show the reaction parameters (ranges are given for the 5 values of body weight and relative TPA doses) and (b) of H. euphorbiae (same 5 individuals) and H. tithymali larvae (also n = 5) as observations in percent vitality (plotted as a bar chart in Fig. 3)
Full size table
Fig. 3
figure 3

Summary of mean vitality scores of experimental larvae of H. euphorbiae and H. tithymali (n = 5 per column), the scale on the x-axis is in percent and lies between 100% representing agile, healthy larvae feeding and defecating normally and 0% for dead larvae (not having fed or defecated). The y-axis lists the experimental set-ups (different treatments). C control, T TPA treatment, TF treatment with calcofluor and subsequent TPA, TFP treatment with calcofluor, PBO and subsequent TPA, eH. euphorbiae, tH. tithymali

Full size image

The role of cytochrome P450 monooxygenases in metabolism

Hyles euphorbiae larvae fell into an initial body paralysis after PBO injections (Table 3), but recovered fully after about 0.5 h. After recovery, under PBO plus TPA, calcofluor plus PBO treatments or calcofluor plus PBO and subsequent TPA treatments, the larvae remained almost (Fig. 3) as vital as the control larvae. The three substance treatments lead to a slight reduction in vitality of experimental larvae (Table 3; Fig. 3).

Discussion

Evolution of TPA insensitivity in Hyles

Due to the variation in the availability of breeding stock for the species studied, the number of larvae which could be tested varied to such an extent between species, that a detailed comparison is difficult. Although an asymmetric sample size is well accommodated by nonparametric tests, any quantitative test with a sample size of n = 1 or even n < 5 is likely to contain artefacts (due to low n), preventing quantification or statistical comparisons of our results. Working with live stock from so many species was a major challenge in this work; the results were hard to obtain and are thus exclusive. For example, because of the poor physiological condition of the stock (leading to mortality of control larvae, possibly due to bacterial and/or viral infections), the results of testing TPA sensitivity in H. annei and H. zygophyllii (obligate on Z. fabago, Zygophyllaceae in its natural habitat) must be interpreted with caution. Nevertheless, H. annei appears to show a level of sensitivity to Euphorbia, which lies between that of species like H. zygophyllii (starts to feed but dies before pupation) and H. lineata (where significant numbers of individuals are able to complete their development).

Specialist non-Euphorbia feeders

Two non-Euphorbia feeders within the HEC, H. costata and H. zygophylii are specialists on other food plant families (see Fig. 1; also Fig. 3 in Hundsdoerfer et al. 2017). They have presumably lost Euphorbia feeding secondarily (Hundsdoerfer et al. 2009). They could, however, have retained the ability to detoxify phorbol esters as this metabolic ability represents a plesiomorphic character within the HEC. Tests with H. zygophyllii larvae for TPA sensitivity showed good survival after TPA treatment (3 out of 4) and additional feeding experiments (MAO) corroborate their ability to use Euphorbia species as food. Nevertheless, these latter larvae were not able to complete development arguing for a limited capacity of H. zygophyllii to deal with Euphorbia toxins. It is likely that testing H. costata (specialised on Polygonaceae) and also H. livornicoides would further enlighten the question, but unfortunately, we could not obtain livestock of these species.

H. centralasiae is sensitive to TPA, and H. vespertilio is minimally able to tolerate TPA in their diet (85% mortality; Table 2). As it was based on two individuals (one control, one experimental), the result for H. centralasiae warrants further investigation. Since H. centralasiae is obligate on a not acutely toxic plant genus, it is likely to have lost the relevant metabolic pathways for detoxifying and thus utilising Euphorbia food plants. The result of the single larva not surviving TPA treatment is thus not unexpected and is most likely representative for the species.

Polyphagous non-Euphorbia feeders

Hippotion celerio was tested as an out-group that is sensitive to TPA in its diet. The impressive list of 99 host plant species in the HOSTS database (see above) shows that its polyphagy includes utilisation of toxic plants genera such as Araceae and Solanaceae; however, no members of the genus Euphorbia are listed. This implies that although H. celerio is able to detoxify secondary compounds produced by some genera (e.g. alkaloids from Nicotiana and Arum), it is not likely to have evolved the metabolic pathways to detoxify phorbol esters.

Although Hyles euphorbiarum was successfully reared on E. characias, E. amygdaloides, E. peplus, E. pulcherrima and even E. cyparissias (pers. obs. by MAO), the species is not specialised on Euphorbia (Ureta and Donoso 1956) and probably does not feed on this genus in its natural habitat, where the abundance of leafy spurges (subgenus Esula) is not high. Nevertheless, H. euphorbiarum is able to tolerate high doses of TPA (Table 2), so it is likely that the species is able to utilise Euphorbia species, such as Poinsetta, E. pulcherrima (subgenus Chamaesyce; possibly less toxic), when other food plants are unavailable. Unfortunately, the results of the feeding experiment with its sister species H. annei have to be interpreted with caution (see above). Three of ten larvae survived TPA treatment in spite of health problems in the breeding stock, but two of the pupae in the control group did not survive. As with H. zygophyllii (see above), a minimal conclusion can be drawn that H. annei retains a limited capacity to metabolise spurge toxins.

A possible explanation for the survival of TPA-fed larvae in low-quality breeding stock is that the toxicity of TPA provides a level of protection from bacterial infection (reduced infection hypothesis). Informal feeding experiments have shown that the mortality of low quality inbred H. gallii and H. livornica captive broods can be reduced by 60–70% by using Euphorbia sp. as their food plant (pers. obs. by MAO). This could also be a major advantage for larvae feeding on Euphorbia in natural environments. Euphorbia toxins could also provide protection from parasitoids (“nasty host” hypothesis). These two observations suggest that Euphorbia-feeding Hyles species have claimed an empty niche, utilising food plants which cannot be metabolised by competing herbivores, enabling them their success (Hundsdoerfer et al. 2009, 2017). There are few other herbivores that feed on Euphorbia species in Central Europe (Manojlovic and Keresi 1997; pers. obs. by AKH), which underlines the effectiveness of phorbol esters as a chemical defence against herbivores.

Polyphagous species which include Euphorbia in their host plant spectrum

H. lineata has been observed to utilise Euphorbia as natural food plant (Leatherman 2014; Moss 1912) and has also been successfully reared using species in this genus (pers. obs. MAO). In our experiments, 80% of the larvae were insensitive to the toxicity of TPA, confirming the ability of this species to metabolise Euphorbia toxins.

Both Hyles gallii and H. livornica proved insensitive to TPA in our experiments. These two species thus appear to possess effective detoxification mechanisms, although they are not Euphorbia specialists.

Specialist Euphorbia feeders

The insensitivity to TPA demonstrated by the three Euphorbia specialists H. dahlii, H. euphorbiae and H. tithymali is not unexpected. Only H. euporbiae has been tested before (Hundsdoerfer et al. 2005), and the species tolerates 2 mg TPA in one dose without showing any loss of vigour. All three species showed 0% mortality when exposed to TPA. We expect that all three species will be able to tolerate much higher doses of TPA, but this was not investigated in this study due to the high cost of standard TPA.

Hyles species generally occupy open and dry habitats. Within the Old World, these are also the areas in which Euphorbia species have their diversity and distribution centres. Spurge feeding was found to be an ancestral character state for the HEC, which in turn revealed an ancestral range in the Palearctic (Hundsdoerfer et al. 2017). This implies that obligate spurge feeding would have originated in the Old World (probably in Central Asia). In these arid regions, being able to feed on succulents (when basal food plants have all been consumed) is highly adaptive. In the Old World, those succulents are likely to be Euphorbia sp. and H. tithymali larvae have been found to be able to use Euphorbia trigona (a cactus like spurge) in captivity.

In summary, whether the natural food plant spectra include Euphorbia (amongst others) is a good predictor of TPA sensitivity (Fig. 1). Although H. euphorbiarum does not normally use Euphorbia, it is able to utilise host plants in this genus in captivity and proved to be TPA-insensitive. The species is expected to be able to use natural Euphorbia, but has simply not been observed on it.

Phylogenetic deductions

From a phylogenetic standpoint, the species branching off earliest in the Hyles tree (Fig. 1), H. lineata, includes Euphorbia in its natural food plant spectrum. Next to branch off are the polyphagous species H. annei and H. euphorbiarum. These species do not include Euphorbia in their natural food plant spectra, but like H. lineata they are able to tolerate TPA in artificial diet (at least a few H. annei; see results), which implies they possess the necessary detoxification mechanisms to metabolise phorbol esters. We thus postulate phorbol ester insensitivity to be the plesiomorphic state for the genus. Moving further towards the crown of the tree, H. biguttata (Madagascar) is specialised on host plants of the genus Ericacae (Attié et al. 2010) and H. livornicoides (Australia) is polyphagous utilising several plant families (Moulds 1998), but does not include Euphorbia in its (observed) host plant spectrum. Unfortunately, the question of whether these species have lost the phorbol ester detoxification pathways, as H. vespertilio and H. centralasiae have, remains open. Continuing along the Hyles evolutionary succession, two of the polyphagous species of the Hawaiian radiation, H. perkinsi and H. wilsoni include Euphorbia in their host plant spectrum, whereas the third species H. calida does not (HOSTS database, see above; Attié et al. 2010). Unfortunately, no live material could be obtained in order to test the extent of the specific TPA sensitivity in these three species. Branching off early within the Palearctic radiation, the two polyphagous species H. livornica and H. gallii both include Euphorbia in their host spectrum. The latter is the sister species to the Euphorbia specialist H. nicaea, and it actually also utilises an Euphorbia species as its sole food plant in part of its range (Nepal; Daniel 1966). Hundsdoerfer et al. (2017) have postulated that H. nicaea has specialised on Euphorbia from a polyphagous ancestor which included Euphorbia in its host plant spectrum. The nine crown group species (see Fig. 1) include obligate host plant specialists which utilise both Euphorbia and other non-related plant genera. The ancestral state reconstruction (see Hundsdoerfer et al. 2017) supports that host plant specialisation in the genus progresses from an ancestral predominance of polyphagy to strict specialisation in this crown group (Hundsdoerfer et al. 2017).

Tolerance to phorbol esters

A lethal effect on M. sexta was produced at phorbol ester concentrations between 1 and 10 µg/100 mg diet (Wink et al. 1997). However, H. euphorbiae caterpillars tolerated TPA doses in excess of 2 mg (in one case in 2 g diet, i.e. an average concentration of 100 µg TPA/100 mg diet) within 4 h and were able to survive being injected with 1.5 mg TPA (Hundsdoerfer et al. 2005). They thus tolerated at least a 10–100 times higher dose in the diet than M. sexta and an even higher magnitude of the comparable dose when directly injected. M. sexta includes an Euphorbiaceae species in its host plant spectrum (Cassava, Manihot esculenta), but none within the genus Euphorbia itself (HOSTS; link above).

Wink and Theile (2002) demonstrated that H. euphorbiae is not generally more tolerant to plant toxins than M. sexta. When injected with the alkaloid nicotine (from Nicotiana), the caterpillars suffered strongly and although they were still alive after 72 h, they were in a very poor physiological state. M. sexta tolerates the highest amount of nicotine of all Lepidoptera studied, which is not surprising because it naturally feeds on Nicotiana. As expected, sphingid species appear to have evolved efficient detoxification mechanisms for the specific plant toxins which they encounter in their diet. As a further study it would also be interesting to test TPA sensitivity in other non-Hyles genera closely related to Hyles, such as Theretra, Deilephila and Xylophanes (and others).

Target site modification

One possible way to avoid intoxication by phorbol esters could theoretically be achieved by modifying the stereochemistry of their main target, the protein kinase C. Such an avoidance of toxic effects has been detected in a multitude of insect species exposed to cardiac glycosides in their food plants which would normally block the ubiquitous Na, K-ATPase (Dobler et al. 2012; Holzinger et al. 1992; Labeyrie and Dobler 2004; Petschenka et al. 2017; Zhen et al. 2012). The interaction between phorbol esters and protein kinase C likewise involves a precise molecular target, a highly conserved cysteine-rich motif in the enzyme’s C1 domain (Parker et al. 1986; Rosenthal et al. 1987). This interaction is so prominent that protein kinase C was first described as the phorbol ester receptor before diacylglycerol, its physiological ligand, was recognised and the vast number of signalling cascades initiated by this receptor–ligand complex described (Blumberg 1988, Nishizuka 1984). Although protein kinase C is the most prominent ‘receptor’ for TPA and other phorbol esters, it is now recognised that the C1 domain is also present in other enzymes which interact with TPA and other phorbol esters (Brose and Rosenmund 2002). Possibly this larger number of binding sites and the huge importance of signalling cascades initiated by diacylglycerol render target site insensitivity impossible. At least our sequences of H. euphorbiae PKC yielded no evidence for a disruption of the decisive cysteine-rich motif that is essential for phorbol ester binding and revealed an almost identical sequence to the one of the TPA sensitive M. sexta (Fig. 2). However, the identity of the amino acid at position 117 of Hippotion celerio needs to be verified with fresh material since a potential threonine-alanine substitution could have a three-dimensional effect on the enzyme, as threonine is polar and also bulkier than alanine. If Hippotion celerio features the conserved threonine like B. mori and M. sexta, the effect of this substitution would need to be further investigated as it could thus conceivably block or inhibit the binding of TPA to the cysteine-rich region. A point mutation at position 368 (arginine instead of lysine; Ohno et al. 1990) changed the putative ATP binding site of PKC in rat fibroblasts and caused resistance to TPA-mediated PKC activation.

Physical barrier and the role of cytochrome P450 monooxygenases in metabolism

There are large differences in the larval body weight among the groups, yet due to availability it was not possible to level body weights among the experimental groups. It is not impossible that the results are influenced by this, but it appears not to be the case, since all H. euphorbiae and H. tithymali larvae were equally insensitive, regardless of body weight. Both species showed no loss of vigour after chemical destruction of their PM (dietary calcofluor; Fig. 2). Subsequent treatment with the maximal individual 2 mg TPA dose did not reduce vigour in H. euphorbiae much either. When CYPs were reduced in metabolic activity by PBO forming a metabolite–inhibitory complex with the enzymes, larvae of both species went into paralysis for around 30 min, indicating the strength of the effect. After recovery, larvae were as vital as before and fed and defecated normally. Larvae of H. euphorbiae lacking a PM and having the metabolic activity of CYPs reduced (dietary calcofluor plus PBO injections; Table 3), showed only a slight (15%) loss of vitality (Fig. 2) after TPA uptake (24 h after calcofluor plus PBO treatments), indicating other detoxification pathways also to be involved. The effects observed were stronger in closely related H. tithymali, vitality drops slightly (to 85%) upon chemical destruction of their PM, and strongly upon additional reduction of CYP metabolic activity (down to 22%, one of five individuals died). The P450 pathway is thus postulated to be one of the enzymatic cascades involved in the detoxification machinery making these species insensitive to phorbol esters. This was also postulated by Barth et al. (2018), who detected an upregulation of transcription of CYP enzymes upon TPA treatment. However, additional enzyme classes in ‘drug metabolism’, carboxylesterases (CES), glutathione S-transferases (GST) and lipases, had also been triggered. Detoxification is a multi-step process (Bury et al. 2014) often involving an initial functionalisation by cytochrome P450 (CYP) enzymes, oxidases etc., followed by subsequent modifications such as, e.g. conjugation by glutathione-S-transferases (GST).

In conclusion, neither the PM acting as a physical barrier, nor a point mutation in the phorbol ester binding site of the PKC appears to cause H. euphorbiae insensitivity to phorbol esters. Further clarification of the potential of a possible threonine-alanine substitution just outside the binding site will be needed. The P450 metabolism appears to be an important, but probably not the only detoxification machinery preventing intoxication. The largest radiation of the genus Hyles occurred in the Old World, correlating with the origin and main distribution area of leafy spurges (480 species, Peirson et al. 2014). The Palaearctic, polyphagous species H. gallii and H. livornica include Euphorbia in their natural food plant spectrum and are, accordingly, also insensitive, whereas two specialist non-Euphorbia feeders have completely lost this ability. Comparatively fewer Euphorbia species occur in South America, the origin of the genus Hyles (Hundsdoerfer et al. 2017). Since the three South American species (H. lineata also occurs in North America) are able to detoxify the standard Euphorbiaceae phorbol ester TPA (at least in part for H. annei), we deduce the hypothesis that phorbol ester insensitivity is a plesiomorphic character state within the genus that enabled its adaptive radiation in the Palearctic.