Diet influences resource allocation in chemical defence but not melanin synthesis in an aposematic moth Free

Organisms need to invest simultaneously in various life-history traits such as growth, energy maintenance, reproduction, predator avoidance and protection from pathogens (Stearns, 1992; Roff and Fairbairn, 2007). According to life-history theory, limited resources need to be distributed among various competing traits, inevitably leading to trade-offs in resource allocation (i.e. resource-allocation hypothesis; Bazzaz et al., 1987; Stearns, 1989; Glazier, 2002). Examples include trade-offs between larval and adult traits (Stevens et al., 1999), or between predator avoidance and immune function (Rigby and Jokela, 2000). In general, allocation costs lead to trade-offs between traits that shape the demography of populations (McCauley et al., 1990; Boggs, 1992) and can significantly impact eco-evolutionary dynamics in nature (Schroderus et al., 2010; Farahpour et al., 2018). In organisms with complex life cycles, such as insects, considering how resources are allocated across and within life stages is essential to understanding how trade-offs shape fitness and different traits (Burdfield-Steel et al., 2019; Lindstedt et al., 2016).

Melanin plays an important role in the warning signals of many aposematic organisms (Fabricant et al., 2013; Lindstedt et al., 2020), those in which a primary defence (e.g. a warning signal such as colour, odour, sound) is coupled with secondary defences (chemical, morphological, behavioural) (Poulton, 1890). Melanin is a phenolic biopolymer widely present across the animal kingdom, whose function in organisms as distant as mammals and insects is similar despite differences in the enzymes and substrates that regulate its production across species (Sugumaran, 2002). In vertebrates, melanin has a pivotal role in protection against UV radiation (Jablonski and Chaplin, 2017; Nicolaï et al., 2020), clarity of vision (Ostrovsky et al., 2018), as well as in sexual (Roulin, 2016), and social (Møller, 1987) signalling and protective colouration (Cuthill et al., 2017). In invertebrates, melanin, specifically eumelanin, is also crucial in physiological and ecological processes such as thermoregulation (True, 2003; Trullas et al., 2007), UV protection (Hessen, 1996), desiccation resistance (Ramniwas et al., 2013), immune defence (Wilson et al., 2001; Sugumaran, 2002), wound healing (Bilandžija et al., 2017) and protection from predators (Majerus, 1998; Hegna et al., 2013). Whether the expression of melanin is prioritised or canalised (Britton and Davidowitz, 2023), genetically based (Ellers and Boggs, 2002), or related to dietary resources (Lee et al., 2008), and whether it is plastic or condition dependent (Britton and Davidowitz, 2023) remains poorly understood. To date, we still do not really understand how diet influences melanin synthesis (Britton and Davidowitz, 2023); thus, it is essential to investigate which diet components can influence melanin.

Fatty acids such as PUFAs (polyunsaturated fatty acids) influence the eumelanin pigmentation in fish; for example, in the Japanese flounder, Paralichthys olivaceus, a diet low in PUFAs reduced pigmentation and eye abnormalities (Estévez et al., 1997). Micronutrients, vitamins and metals can also interfere with and influence the melanogenesis process (Britton and Davidowitz, 2023). Proteins are known for their effect on melanin, and melanogenesis relies on amino acids (Britton and Davidowitz, 2023) and the proteins necessary for the synthesis of melanin pigments can constrain the expression of life-history traits and chemical defences (Lindstedt et al., 2020; Galarza, 2021).

Aposematic signals that rely on melanin are strong candidates for testing the resource-allocation hypothesis. Experimental studies testing both the melanin component of warning signals and the chemical defences in the same system are lacking. Only a few studies have focused on testing how the production of secondary defences may imply trade-offs in the allocation of resources needed for the primary defences. For example, previous work by Lindstedt et al. (2010) showed that the aposematic, generalist herbivore Arctia plantaginis (hereafter referred to as the wood tiger moth) develops a paler, orange warning signal when reared on Plantago lanceolata with high concentrations of iridoid glycosides (IG) in comparison to the more conspicuous, dark red warning signal of individuals fed on a strain of the same plant with low IG concentrations. This suggests that the cost of conspicuousness arose via higher excretion costs rather than via resource-allocation costs (Lindstedt et al., 2010). Recent work on monarch butterflies, Danaus plexippus, showed a link between toxin sequestration and warning signals, where male conspicuousness was inversely correlated with oxidative damage (due to an increase in concentrations of sequestered cardenolides) (Blount et al., 2023). Similar to the red and yellow pigments often seen in warning signals, the production of melanin may incur trade-offs with other traits, such as developmental time (e.g. melanised juvenile individuals take longer to develop during the early stages; Windig, 1999) or mate choice (for example, the less melanised females of Colias philodice butterflies attract a higher number of males than the most melanised females; Ellers and Boggs, 2004). The maintenance of costly traits, such as melanin pigments and chemical defences, can be affected by the quality of the diet. Thus, it is crucial to understand the costs associated with these traits and whether they are linked during development.

Here, we used the aposematic wood tiger moth, which has a black melanin pattern that covers approximately 20–70% of its hindwings and 50–80% of its forewings (Hegna et al., 2013), to investigate how early-life resource availability influences melanin expression in the wings, the efficacy of chemical defences, and life-history traits. In this species, melanin synthesis competes against the production of chemical defences for energy and resources from diet precursors (nitrogen from proteins). While the metabolic pathway in the wood tiger moth chemical defence has not been identified yet, the thoracic defence fluid presents two methoxypyrazines which, similar to the eumelanin components, are heterocyclic compounds based on nitrogen (Higasio and Shoji, 2001) and produced from the same amino acids: phenylalanine and tyrosine (Hearing, 1993; Wang et al., 2021). We manipulated the protein content of the diet of male wood tiger moths (high, low), to investigate how wild-caught, natural predators (blue tits, Cyanistes caeruleus) respond to their chemical defences, and whether the production of such defences leads to trade-offs with the production of other traits, such as wing melanisation.

Based on previous studies (Burdfield-Steel et al., 2019; Lindstedt et al., 2020), we can predict that in an environment with limited resources (i.e. low-protein diet), traits that are costly to produce, such as chemical defences and melanin, would be highly affected. Thus, we assumed that compared with males raised on a low-protein diet, males raised on a high-protein diet would have (1) higher melanisation; (2) more effective chemical defences, rendering them less palatable to bird predators; and (3) lower life-history costs in terms of size or developmental time. Alternatively, if melanin, which plays an important role in several traits relevant to individual fitness, is under strict genetic control and prioritised in investment, we would expect that in resource-limited conditions (low-protein diet), the amount of melanin in the hindwings would not be affected. Likewise, we would expect that those individuals with high amounts of melanin in the hindwings would have less deterrent chemical defences, meaning that the investment of resources in chemical defences, but not in melanin synthesis, would be reduced.

The aposematic wood tiger moth Arctia plantaginis (Linnaeus 1758), formerly Parasemia plantaginis (Rönkä et al., 2016), displays a conspicuous colour polymorphism (Watson and Goodger, 1986; Chinery, 1993; Nokelainen, 2013) throughout the Holarctic region (Hegna et al., 2015). Male hindwings across the species distribution range can have a red, yellow or white background with black patterning that covers variable proportions (ca. 20–70%) of the wing. White and yellow colourations are partly produced by pheomelanin, whereas black is a dopamine-derived eumelanin (Brien et al., 2022). Previous research has scored melanisation (the amount of eumelanin) of the hindwings as ‘plus’ (with stripes, high melanin) or ‘minus’ (without stripes, low melanin) (see Fig. 1), but this arbitrary categorisation does not necessarily represent the true variation, in which subtle differences are not detectable by the human eye. Long-term natural frequencies of male wood tiger moth melanin morphs in Estonia (classified by the human eye) are about 47% high-melanin and 53% low-melanin (O.N., unpublished data; see Fig. 1).

As capital breeders, wood tiger moth adults do not feed; thus, resource acquisition only occurs during the larval stage. To avoid predation by avian and terrestrial predators, these moths produce two different types of defensive fluids: abdominal secretions deter predators such as ants, and thorax secretions deter birds with no adverse effects on ants (Rojas et al., 2017). The thoracic fluid contains two methoxypyrazines: 2-sec-butyl-3-methoxypyrazine (SBMP) and 2-isobutyl-3-methoxypyrazine (IBMP), which are produced de novo (Burdfield-Steel et al., 2018).

Male wood tiger moths were obtained from a laboratory stock founded in 2013 with wild-caught individuals from Estonia and kept at the greenhouse of the University of Jyväskylä. The stock is supplemented annually with Estonian wild-caught individuals to preserve genetic diversity. The greenhouse conditions approximately followed the outdoor temperatures and natural daylight from May to August in Central Finland: between 20°C and 30°C during the day, decreasing to 15–20°C during the night. We picked 10 families that contained both low- and high-melanin male morphs in the parental and grandparental generations, as classified by eye (see Fig. 1). Laboratory crosses show that wing melanisation (both pattern and amount) is strongly heritable (C.O., personal observation). After hatching, larvae of each of the 10 families were kept together for 14 days and fed with lettuce and dandelion (Taraxacum spp.). Then, using a split family design, they were divided into two artificial diet treatments that differed in protein content. In the high-protein diet, Teklad Vitamin-Free casein was used as the predominant source of protein (32.4 g in high-protein diet), while in the low-protein diet, a lower amount was used (10.8 g) and non-nutritive cellulose was added to replace the casein (see Supplementary Materials and Methods, ‘High- and low-protein diet’, for recipe). For each diet treatment, we used 240 individuals which were kept in boxes of 10 individuals until reaching the pupal stage. Each box was checked and watered daily, and cleaned when needed. Larvae were fed daily, ad libitum, with their corresponding diet treatment. Life-history traits (time to pupation, pupal mass) and the degree of melanin (high +, low −) were recorded for each individual. In total, 480 larvae were followed from eggs to adulthood or death, 177 of which emerged as females and 184 as males. All larvae were grown by Furlanetto (2017) from May to August 2017. When the male adults emerged from pupation, they were given water and stored at 4°C to slow their metabolic rate and to maintain their condition; thoracic chemical fluid was then collected in June and July 2017. Prior to fluid collection, moths were kept at 20–25°C for 30 min. Then, fluid was extracted by squeezing just below the prothoracic section with tweezers and collecting expelled fluid in a 10 μl glass capillary. Fluid samples were stored in glass vials at −18°C (Furlanetto, 2017).

To measure variation in wing melanisation, we photographed 53 male moths from 10 different families (23 raised on low-protein diet and 30 raised on high-protein diet) whose defensive fluids were used in the predator-response assay. Because the degree of melanisation varies continuously, we measured the proportion of melanin on the wings on a continuous scale using image analysis.

Spread male moths were photographed using an established protocol (Nokelainen et al., 2017) and scaled to the same resolution (pixels mm⁻¹). We used the MICA toolbox (Troscianko and Stevens 2015) in ImageJ (v.1.50f) for image analysis. From every image, a set of regions of interest (ROI) were hand-selected from the dorsal side of the wings: (a) whole forewing area, (b) forewing melanised region, (c) whole hindwing area, and (d) hindwing melanised region. The relative proportion of the wings that was melanised was calculated by dividing the area of the melanised ROI by the whole wing ROI, for the forewings (forewing melanised region area/whole forewing area) and for the hindwings: (hindwing melanised region area/whole hindwings area).

Blue tits (Cyanistes caeruleus) are generalist feeders that have a similar distribution to that of A.plantaginis in Europe and are known to attack this species. Blue tits are common in Finland and can be kept briefly in captivity for experiments (e.g. Rojas et al., 2017). For the predator-response assay, we used fluids from 53 male moths fed on high-protein (n=30) and low-protein (n=23) diets; bait soaked with water was offered to 9 birds as a positive control (we use in total 62 birds). The volume of wood tiger moth thoracic fluid has no effect on blue tit responses (Burdfield-Steel et al., 2019), so we diluted each fluid sample with water to reach a total volume of 15 μl. The 15 µl was then divided into two samples of 7 μl each, which were used as bait for the same bird. The same amount of water was offered to control birds. The birds used for the experiment were caught at Konnevesi Research Station (Central Finland) from February to April 2018 using feeders with peanuts as bait (Ham et al., 2006), and then housed individually in plywood cages for the duration of the assays (see Ottocento et al., 2022).

Birds were first familiarised with the experimental boxes and trained to eat a bait (oat flakes) (see Ottocento et al., 2022). Each bird then experienced four sequential trials, at 5 min intervals, in which they were presented with a plate containing one bait (see Fig. 2A). The first and last trials were done with bait soaked in tap water to ensure that the bird was motivated to eat (first) and that the bird was still hungry (last). In the second and third trials, the bird was presented with a bait soaked in 7 μl of the defensive fluid of the same moth (30 raised under a high-protein diet, 23 raised under a low-protein diet). The trial ended two min after the bird had eaten the whole bait, or after a maximum duration of 5 min if the bird did not eat the whole bait. In each trial, we recorded the latency to approach (how long it took for the bird to get close to the plate on which the bait was offered), the latency to eat after approaching (how long it took for the bird to ‘attack’ after approaching the bait) and the latency to eat (the total time from when the bird saw the bait on the plate until it started eating it). These measurements were taken as proxies of how repellent the odour of the moth's fluids was. Additionally, the beak cleaning frequency (the number of times the bird wiped its beak against a surface, e.g. the perch), and how much of the bait was eaten to the nearest 0%, 50% and 100% were recorded. Cases where the bait was tasted but no significant portion of it was eaten were classed as 10%, to differentiate them from instances in which the bait was not even tasted. Likewise, cases where almost all the bait was eaten, but some crumbs were left over the cage floor were designated as 90% (see Fig. 2B).

Wild birds were used with permission from the Central Finland Centre for Economic Development, Transport and Environment and licence from the National Animal Experiment Board (ESAVI/9114/ 04.10.07/2014) and the Central Finland Regional Environment Centre (VARELY/294/2015). All experimental birds were used according to the ASAB/ABS Guidelines for the treatment of animals in behavioural research and teaching (Association for the Study of Animal Behaviour 2020).

All statistical analyses were done with the software R v.4.1.2 (http://www.R-project.org/) using the RStudio v.1.2.1335 interface (http://www.rstudio.com/). The level of significance in all analyses was set at P<0.05.

Our results show that the amount of melanin in the hindwings of wood tiger moths is best measured as a continuous variable (see Supplementary Materials and Methods, ‘Is hindwing melanisation continuous or discrete?’ and Fig. S1), so we analysed how protein content in the diet influences the proportion of the wing that is melanised and the size of the wings using linear mixed models with treatment as a fixed factor and family as a random factor, using the lme4 package (Bates et al., 2015). The total area of the forewings, the total area of the hindwings, the amount of melanin in the forewings (melanised area of forewing/total area of forewing), and the amount of melanin in the hindwings (melanised area of hindwing/total area of hindwing) were set as response variables (in separate models); the different types of diet (high/low-protein content) were set as predictors.

To test the proportion of bait eaten, we used beta regression models using the package glmmTMB (Brooks et al., 2017) with family=beta_family(link= “logit); diet (low protein, high protein), the amount of melanin and trial set as fixed effects; bird ID as a random effect; and an offset included to account for differences in observation time. To test for differences in the number of beak wiping events per minute, we used generalised linear mixed-effects models (GLMM; https://CRAN.R-project.org/package=glmm) with a log link and Poisson distribution, fitted by maximum likelihood (Laplace approximation). The diet treatments (low protein, high protein), the amount of melanin, and trial were set as fixed effects; beak wiping frequency was set as the response variable; and bird ID was set as a random effect. To test the latency to approach, the latency to eat after approaching and the latency to eat, we used a Cox mixed-effects model using the package coxme (https://CRAN.R-project.org/package=coxme), fitted by maximum likelihood. These three behaviours were set as response variables; the interaction between diet (low protein, high protein), the amount of melanin and trial were set as fixed effects; and bird ID was set as a random effect. The behaviours of the predators were first compared with a water-only control to determine whether moth chemical defences elicited adverse predator reactions. We used Spearman's rank correlations (Ojala et al., 2005) to test the correlation between the amount of melanin and predator responses to chemical defences (proportion of bait eaten, beak wiping, latency to approach, latency to eat after approaching, latency to eat). For all the models, we selected the best model fit by comparing the Akaike information criterion (AIC) values.

To test whether male moths’ development time is affected by dietary protein and the amount of melanin, we used a mixed-effects Cox model (package coxme, https://CRAN.R-project.org/package=coxme), with development time included as the response variable, diet treatment and melanisation included as the predictor variables, and family as a random factor. We assessed the effect of diet treatment on pupae mass with a generalised mixed-effects model (package lme4, Bates et al., 2015) with a Gamma (link="log") distribution, with diet treatment and melanisation as the predictor variables and family as a random factor. We used a linear mixed-effects model (package lme4, Bates et al., 2015) with diet and melanisation as predictor variables and family as a random factor to test the effect of diet on the volume of the thoracic fluid. We tested the correlation between pupal mass and the volume of the thoracic fluid using a Pearson correlation. Spearman's rank correlations (Ojala et al., 2005) were used to test the relationships between the amount of melanin and the developmental time, pupal mass, and volume of thoracic fluid.

There was no significant effect of early-life protein availability, either on the content of melanin in the wings (hindwings: coef±s.e.=0.001±0.32, t=0.265, P=0.96; forewings: coef±s.e.=−0.02±0.03, t=−0.85, P=0.40) or on wing size (hindwings: coef±s.e.=−95,528±81,245, t=−1.18, P=0.246; forewings: coef±s.e.=−104,081±89,682, z=−1.16, P=0.25).

To test whether the thoracic fluid of the moth evoked an adverse reaction in the birds, we first examined predator behaviour in response to bait soaked in water (control). Bait soaked in the chemical defences of moths raised on a high-protein diet were eaten by the predators in lower proportions than was bait soaked in water (control) (coef±s.e.=−2.54±0.87, z=−2.91, P=0.004), while the predators’ response to bait soaked in fluid from moths fed with a low-protein diet did not differ from that to bait soaked in water (coef±s.e.=−1.56±0.87, z=−1.79, P=0.074). We found no significant differences in the other predator behaviours recorded (i.e. latency to approach; latency to eat after approach; latency to eat; frequency of beak wiping) between birds exposed to bait soaked in fluid from moths raised on either high-protein or low-protein diet, and those exposed to water-soaked bait (P>0.05 for all comparisons, see Supplementary Materials and Methods, ‘Effect of diet manipulation and hindwing melanin on the predator response to male's defensive fluid’).

The chemical defences of the most melanised individuals raised on the low-protein diet were also more deterrent against predators (Spearman's rank correlation, r_s=−0.50; P=0.00045; Fig. 3A). Although not significant, the chemical defences of the most melanised individuals raised on the high-protein diet followed a similar trend (Spearman's rank correlation, r_s=−0.13; P=0.31; Fig. 3A). When the two diets were pooled together, we observed a clear trade-off between melanin and chemical defence: bait soaked in the defensive fluid of highly melanised moths was consumed in lower proportions than that soaked in fluid from less melanised males (coef±s.e.=−5.20±1.96, z=−2.66, P=0.007; Fig. 3A). Similarly, the predators' latency to eat the bait increased with higher amounts of melanin, but only for the high-protein diet (Spearman's rank correlation, r_s=0.28; P=0.04; Fig. 3B), not the low-protein diet (Spearman's rank correlation, r_s=0.235; P=0.14; Fig. 3B). There was no correlation between the amount of melanin and other behavioural variables (latency to eat after approaching; frequency of beak wiping; P>0.05 for all comparisons; see Supplementary Materials and Methods, ‘Effect of diet manipulation and hindwing melanin on the predator response to male's defensive fluid’).

The developmental time from egg to pupa varied among diet treatments: individuals raised on the low-protein diet took significantly longer to pupate than those raised on the high-protein diet (coef±s.e.=−0.92±0.32, z=−2.87, P=0.004). There were no differences in body mass (coef±s.e.=0.03±0.04, t=0.75, P=0.47) or in the volume of the thoracic defensive fluid (coef±s.e.=1.05±0.87, t=1.2, P=0.23) between individuals raised on the high- and low-protein diets. The volume of defensive fluid did not correlate with pupae mass (t=0.96, d.f.=188, P=0.34). There was no correlation between the amount of melanin and the larval developmental time (r_s=0.11; P=0.435), body mass (r_s=0.003; P=0.98), or the volume of defensive fluid (r_s=−0.03; P=0.79).

Investigating how resource allocation affects life-history traits and defences against predators is vital for understanding which factors drive phenotypic variation in animals. In aposematic organisms, the phenotypic correlations among traits such as melanin and chemical defences do not necessarily prove they are linked, as interactions between such complex traits can be shaped by genetic correlations, pleiotropic effects or simply variable environmental conditions (e.g. food availability, environmental stochasticity, presence of predators). Therefore, to test the resource allocation hypothesis, we need to first test whether the production of defences is costly and then whether their development is linked to other traits. To investigate whether those traits are costly, and whether a specific trait (melanin or chemical defence) is prioritised over others or robust, we manipulated specific constituents of the early-life resources of male wood tiger moths. The diet used in the study varied in the amount of protein, which is an essential compound in different aspects of the larval life history of this species, and which has been shown to influence the efficacy of the species' warning signal, immunity and life-history traits (Lindstedt et al., 2020).

Surprisingly, dietary protein levels had no effect on the amount of melanin in the forewings or hindwings, suggesting that this trait may not be directly influenced by the early-life environment. Across different organisms, the effect of diet on melanin is varied and complex, and specific components, such as proteins, fatty acids, metals and micronutrients, can play a crucial part in melanin pigmentation (Britton and Davidowitz, 2023). Studies in invertebrates have shown that melanisation is influenced by the quality of proteins in the diet (Lee et al., 2008; Ethier et al., 2015). For example, in the African cotton leafworm, Spodoptera littoralis, the high nutritional quality of the protein diet led to more melanised cuticles, faster growth and better antibacterial activity and survival (Lee et al., 2008). In the forest moth Malacosoma disstria, individuals raised under low-nitrogen availability had decreased melanic pigmentation and smaller size, highlighting the high costs of melanisation (Ethier et al., 2015). In vertebrates such as the tawny owl, Strix aluco, the genetic control of melanin deposition appears to be strictly regulated (Roulin and Dijkstra, 2003; Mundy et al., 2003; Bize et al., 2006; Hoekstra, 2006; Emaresi et al., 2011). However, this is not always the case, and environmental conditions, such as stress, may influence melanin pigmentation in vertebrates (Britton and Davidowitz, 2023). For example, the abundance of food influences the production of pheomelanin in juvenile goshawks, Accipiter gentilis, but not in adults, which present different levels of oxidative stress (Galván et al., 2019).

In a recent study, Lindstedt et al. (2020) found that the protein content of the diet fed to larval wood tiger moths directly influenced the amount of forewing melanin of adult males in the closely related Finnish population. The apparent mismatch between Lindstedt et al.’s (2020) and our results could be due to the study design: Lindstedt et al. (2020) used larvae from selection lines for low and high larval melanisation, such that the phenotypic variation was larger than the natural variation (and, consequently, the average phenotypes were either rare or missing completely), probably making the possible costs of melanin easier to detect. In this study, by contrast, we used the natural variation of phenotypes. Interestingly, Lindstedt et al. (2020) did not find differences in the amount of hindwing melanin between individuals fed with high- and low-protein diets. Thus, hindwing melanisation and patterning seem to be under strong genetic control, which is unsurprising because moth hindwings commonly have an important signalling function (Sargent, 1978; Kang et al., 2017; Rönkä et al., 2018). It is also possible that the Finnish and Estonian populations differ in their degree of plasticity of melanin production, but this hypothesis requires further investigation.

We hypothesised that the thoracic fluid of males raised on a high-protein diet would elicit a stronger predator response than the fluids from moths raised on a low-protein diet, as shown in a previous study (Furlanetto, 2017), where the abundance of pyrazines was higher in fluids from moths raised on the high-protein artificial diet. An increase in protein content means an increase in the concentration of nitrogen, an essential element in the synthesis of pyrazines (Hodge et al., 1972; Wong and Bernhard, 1988), a family of compounds with a characteristic repulsive odour (Rothschild et al., 1984; Guilford et al., 1987; Kaye et al., 1989; Moore et al., 1990) and a deterrent effect on birds (Marples and Roper, 1996; Lindström et al., 2001; Siddall and Marples, 2011). In this study, predators found the chemical secretions from male moths raised on the high-protein diet more unpalatable (i.e. they tasted worse) than those from males raised on the low-protein diet, which confirms that the chemical defence is costly. However, we did not find any differences between the predators' hesitation time (i.e. latency) to ‘attack’ the bait soaked with fluid from male moths raised on the high-protein diet and the latency to attack bait soaked in water (control), which suggests that the predators may not perceive a difference in odour (volatile compounds) between the two. These results agree with previous studies suggesting that the defensive fluids of wood tiger moths may contain several repulsive compounds, not only pyrazines, that influence defence efficacy in concert (see also Winters et al., 2021; Ottocento et al., 2022) and affect predator response in different ways (Rojas et al., 2019; Winters et al., 2021). Moreover, as Ottocento et al. (2022) recently showed, the relative amount of the two methoxypyrazines (SBMP, IBMP) present in the wood tiger moth defensive fluids is more relevant to predator deterrence than the total amount of pyrazines. Unfortunately, however, it is not possible to quantify the amount of pyrazines and conduct bird assays with the defensive fluids of the same individual. Interestingly, the same study (Ottocento et al., 2022) showed that predator responses are stronger towards the chemical defences of individuals originating from populations with high predation pressure (e.g. Scotland; Rönkä et al., 2020) than towards individuals from populations with lower predation pressure (e.g. Estonia; Rönkä et al., 2020), even if the total amount of pyrazines in the defensive fluids does not differ among populations.

Our results show that highly melanised male moths have more effective chemical defences than those with less melanin, hinting at a positive correlation between costly melanin pigments and chemical defences. While this is not what we predicted, particularly in a scenario of low resource availability in early life, environmental conditions may have a greater influence on chemical defences than the pigmentary composition of the melanin hindwings, which is less sensitive to the variability of resources than other fitness-related traits. Although in this species the chemical defence compounds are not sequestered directly from plants but produced de novo (Burdfield-Steel et al., 2018), the main elements that the moths require to build their defences are still collected from their food intake. As melanin is a potent antioxidant pigment (Sarna and Swartz, 2006), another interesting possibility is that in A. plantaginis, high levels of melanin are necessary to withstand the oxidative stress resulting from the high amounts of chemical defence compounds produced in their bodies (McGraw, 2005; Blount et al., 2009). Therefore, this could also explain the positive correlation between melanin and chemical defences. It is also possible that, independently of the amount of protein in their diet, some individuals, are more efficient at converting and allocating the protein intake for building melanin pigment and pyrazine compounds, while others may experience greater challenges as a consequence of competition and nutritional imbalances, resulting in reduced digestion and absorption of amino acids (Lee et al., 2008).

When looking at how early-life resource availability influences wood tiger moth size, we hypothesised that males raised on a low-protein diet would have both smaller pupae and smaller hindwings and forewings than those raised on a high-protein diet, as environmental fluctuations and environmental stress (in this case due to the low amount of protein in the diet) may strongly affect both insect wing (Bitner-Mathé and Klaczko, 1999) and pupal size (Nguyen et al., 2019). Our findings, however, reveal no differences in size between males raised on high- and low-protein diets. This could be because individuals raised on a low-protein diet compensate by increasing their food intake (Simpson and Simpson, 2017), or had a longer development time than those raised on a high-protein diet, which might facilitate the acquisition of more resources at the larval stage, even if they are of poorer quality, allowing these individuals to ultimately reach the same size as those on a richer diet, as reported by Lindstedt et al. (2020). A longer developmental time, however, may also lead to increased predation risk (Häggström, and Larsson, 1995), higher vulnerability to environmental perturbations (Tammaru et al., 2001), reduced fecundity (Saastamoinen et al., 2013) and difficulties in escaping from a risky environment (Cowan et al., 1996).

In sum, experimental evidence suggests that the production and maintenance of chemical defences are affected and limited by the resources available in early life, but melanin synthesis is not. This indicates that, at least in wood tiger moths, melanin synthesis in adult wings might be less environmentally regulated than pyrazine production, implying that the expression of either of these traits is not constrained by the same resource pool. Another possible explanation is that, if they are constrained by the same resource pool, melanin synthesis is prioritised (canalised) by selection, and thus any remaining resources contribute to the production of chemical defences, which depends on the total amount of resources in the diet. Our findings also confirm that resource-dependent variation in chemical defences is perceived by natural predators, which has seldom been shown when variation in chemical defences is investigated (White and Umbers, 2021). Studying the response of relevant predators, which are the selective agents for defended prey, is key when aiming to understand how variation in defences is maintained in natural populations.

We are grateful to Helinä Nisu for assisting with the catching and training of wild birds. We thank Konnevesi Research Station for use of the facilities. Kaisa Suisto and other greenhouse workers were invaluable in the rearing of the moths. We thank the members of the ‘Ecology and Evolution of Interactions’ group for the useful discussion on an early version of the manuscript, and two anonymous reviewers for their thorough and constructive feedback. The Materials and Methods, Results, and Discussion sections in this paper have been reproduced from the PhD thesis of O.C. (Ottocento, 2023).

This article has an associated ECR Spotlight interview with Cristina Ottocento.