How the ladybird got its spots: effects of resource limitation on the honesty of aposematic signals


Correspondence author. E-mail:


1. Prey species often possess defences (e.g. toxins) coupled with warning signals (i.e. aposematism). There is growing evidence that the expression of aposematic signals often varies within species and correlates with the strength of chemical defences. This has led to the speculation that such signals may be ‘honest’, with signal reliability ensured by the costliness of producing or maintaining aposematic traits.

2. We reared larval seven-spot ladybirds (Coccinella septempunctata) on a Low or High aphid diet and measured the effects on warning signal expression (elytral carotenoid pigmentation, conspicuousness, spot size), levels of defensive alkaloids (precoccinelline, coccinelline), and relationships between these traits.

3. High-diet individuals had greater total precoccinelline levels, and elytra carotenoid concentrations at adulthood which was detectable to a typical avian predator. However, larval diet did not significantly affect adult body mass or size, spot size or coccinelline levels.

4. Elytra carotenoid concentrations correlated positively with total precoccinelline levels in both diet groups and sexes. However, the relationship between elytra carotenoid concentrations and total levels of coccinelline depended on sex: in both diet groups, elytra carotenoids and coccinelline levels were positively correlated in females, but negatively correlated in males. Spot size and coccinelline levels correlated positively in Low-diet individuals, but negatively in High-diet individuals.

5. These results point to physiological linkages between components of aposematism, which are modulated by resource (i.e. food) availability and affect the honesty of signals. Developmental diet, but also sex, influenced the relationships between signals and toxin levels. Ladybirds are sexually size dimorphic, and thus in comparison with males, females may be more susceptible to resource limitation and more likely to be honest signallers.


Prey species often possess defences, such as toxins or spines, coupled with warning colours and patterns, which advertise their unprofitability to predators (aposematism; Poulton 1890). Until recently, it has generally been considered that such aposematic signals function only to inform predators that prey are defended and to enhance predator learning; in the interests of efficacy, aposematic signals are expected to be invariable within species (reviewed in Guilford 1990; Ruxton, Sherratt & Speed 2004). However, there is growing evidence that the expression of aposematic signals varies within species (Holloway et al. 1991; de Jong et al. 1991; Grill & Moore 1998; Grill 1999; Summers, Cronin & Kennedy 2003; Bezzerides et al. 2007; Ojala, Lindström & Mappes 2007; Sandre et al. 2007a; Cortesti & Cheney 2010; Lindstedt et al. 2010), and experiments have shown that predators are more averse to aposematic signals that are larger or more saturated in colour (Gamberale & Tullberg 1996a,b; Gamberale-Stille & Tullberg 1999; Lindstedt, Lindström & Mappes 2008). In addition, the strength of chemical defences often varies within species (reviewed in Ruxton, Sherratt & Speed 2004). However, how warning colours and chemical defences correlate remains equivocal. Theoretical models have suggested that the most toxic prey should have the least conspicuous signals, because they are likely to survive an attack unharmed and should therefore reduce the costs incurred by signalling (Leimar, Enquist & Sillén-Tullberg 1986; Speed 2001; Speed & Ruxton 2005). This prediction has been supported by empirical data for Epipedobates species of poison dart frogs (Darst, Cummings & Cannatella 2006). However, other studies have found the opposite pattern, that is, a positive correlation between signal expression and chemical defence, suggesting ‘honest’ signalling: across dendrobatid species of poison dart frogs (Summers & Clough 2001) and marine opisthobranchs (Cortesti & Cheney 2010); across populations of the strawberry poison dart frog Dendrobates pumilio (Maan & Cummings 2012); and within a single population of the Asian ladybird beetle Harmonia axyridis (Bezzerides et al. 2007).

One explanation for why aposematic traits may vary within species is that there are underlying physiological costs associated with aposematism (e.g. Rowell-Rahier & Pasteels 1986; Holloway et al. 1991; Holloway, de Jong & Ottenheim 1993; Grill & Moore 1998; Bezzerides et al. 2007; Ojala, Lindström & Mappes 2007; Sandre et al. 2007a; Lindstedt, Lindström & Mappes 2008, 2009). However, the costs associated with the production of aposematic coloration and chemical defences have largely been considered independently of each other (exceptions: Grill & Moore 1998; Lindstedt et al. 2010). We recently proposed a resource competition model where we hypothesized how physiological linkage between such traits could arise (Blount et al. 2009). We suggested that aposematic coloration and chemical defence may compete for a shared resource within individuals (e.g. energy or a particular nutrient). The model generated the prediction that at relatively low resource states, there should be a positive correlation between aposematic coloration and chemical defence, that is, ‘honest’ signalling; as resource availability increases, prey are predicted to invest increasingly (and equally) in the two components of aposematism because neither signalling nor toxicity on their own can provide sufficient protection from predators. In contrast, when resources are very abundant, individuals are predicted to invest disproportionately highly in toxins and to reduce investment in signalling, giving rise to a negative correlation between aposematic coloration and chemical defence; that is, ‘dishonest’ signalling, consistent with earlier theoretical work (Leimar, Enquist & Sillén-Tullberg 1986; Speed 2001; Speed & Ruxton 2005). This prediction arises because the model includes an implicit assumption that increased investment in signalling incurs additional costs of conspicuousness to predators (Blount et al. 2009).

Here, we present an experimental test of some basic assumptions and predictions of the resource competition model (Blount et al. 2009). We reared larvae of the seven-spot ladybird beetle (Coccinella septempunctata) in standardized conditions that differed only in terms of food supply. Coccinella septempunctata has carotenoid-based elytra coloration (Britton et al. 1977) and black melanic spot patterning. It possesses chemical defence in the form of two endogenously synthesized alkaloids, the N-oxide coccinelline and its free base precoccinelline, which are distributed throughout the body tissues (Pasteels et al. 1973; Holloway et al. 1991; Daloze, Braekman & Pasteels 1995). Coccinella septempunctata is aversive and harmful to avian predators if consumed (Marples, Brakefield & Cowie 1989; Dolenskáet al. 2009), which, coupled with the fact that it has conspicuous coloration and patterning, suggests that this species is aposematic. We measured the effects of a diet manipulation (Low vs. High food supply) on elytra carotenoid pigmentation, coloration and discriminability to a passerine (Hart, Partridge & Cuthill 1998; Endler & Mielke 2005), spot size, and body levels of defensive alkaloids, and relationships between these components of aposematism. For the resource competition model (Blount et al. 2009) to be supported, it would be necessary to demonstrate that individuals with greater resource (i.e. food) supply have (i) increased investment in toxins and (ii) increased investment in signals up to a certain threshold of resource supply, beyond which investment in signals should decrease. In addition, (iii) higher investment in signalling (in terms of carotenoid pigmentation) should result in increased conspicuousness; and (iv) the amount of resources available should determine whether levels of signals and toxins correlate positively (under conditions of resource limitation) or negatively (where resources are abundant and non-limiting).

Materials and methods

Rearing of Ladybird Beetles and Manipulation of Diets

Forty adults were collected in mid-June 2008 on semi-improved grassland at Falmouth, UK [latitude/longitude (WGS84): 50°08′02.29″N/005°05′10·06″W], and then paired at random in individual Petri dishes (9 cm diam.). Black bean aphids (Aphis fabae; see below) were provided ad libitum. Most pairs produced eggs within 48 h; pairs that did not lay within 48 h were separated and paired with new individuals. Petri dishes were checked twice daily; if eggs were present, the adults were removed to prevent filial cannibalism. Petri dishes were then placed in an environmental chamber (MLR-350H; Sanyo, Osaka, Japan) at 23 ± 0·5 °C, 60 ± 5% RH and a 16L:8D light regime. On the day of hatching, larvae were transferred to individual Petri dishes (5·5 cm diam.), maintained in the same environmental conditions as above and allocated randomly to receive either Low or High food (A. fabae) supply. The body mass of newly hatched larvae did not differ significantly between diet groups (GLMM: diet, F1,16·9 = 0·60, = 0·45; sex, F1,27·4 = 0·00, P = 0·99; diet × sex, F1,21·6 = 1·16, = 0·29). Aphids were cultured on broad bean plants (Vicia faba, Sutton dwarf variety) in a climate-controlled glasshouse at 21 ± 0·5 °C and a 16L:8D light regime. There were insufficient cultured aphids to support third- and fourth-instar ladybird larvae, when food intake increases exponentially, and therefore during these instar stages, we provided solely A. fabae which were field collected daily on creeping thistle Cirsium arvense. We reared one full-sib dyad of larvae chosen at random from each of 16 breeding pairs (i.e. families); one individual was allocated to the Low-diet group and the other individual to the High-diet group. One larva died, and therefore the final sample sizes were as follows: Low-diet males, = 10; Low-diet females, = 5; High-diet males, = 8; High-diet females, = 8. Running the experiment was highly labour-intensive because of the voracious appetites of C. septempunctata larvae and the need to weigh aphids daily for the diets. With this design, there was no replication within families for each diet group; the effects of family were statistically controlled for and in the majority of analyses explained little or no variance (see below).

The Low-diet consisted of 1·03 mg ± 10% aphids day−1 (1st instar), 2·62 mg ± 10% aphids day−1 (2nd instar), 5·24 mg ± 10% aphids day−1 (3rd instar) and 33·55 mg ± 10% aphids day−1 (4th instar), respectively, whereas the High diet consisted of 1·71 mg ± 10% aphids day−1 (1st instar), 4·36 mg ± 10% aphids day−1 (2nd instar), 8·73 mg ± 10% aphids day−1 (3rd instar), and 55·92 mg ± 10% aphids day−1 (4th instar). These values are based on previous work, being 25% below (Low diet) and 25% above (High diet), respectively, the mass of A. fabae presented to C. septempunctata larvae in standard culture (Murdoch & Marks 1973). We therefore anticipated that our experimental diets would provide relatively poor (Low diet) and good (High diet) rearing conditions, compared to a regime that is known to allow successful development. Aphids were weighed using an electronic balance (UMX2 ultra-microbalance; Mettler-Toledo Ltd., Leicester, UK), and presented to ladybird larvae once daily in the morning. Aphids were weighed out randomly with respect to their instar stage, except that 1st instar ladybird larvae were not given 4th instar aphids because they are too large to be consumed. All aphids were always consumed within 24 h of presentation.

Measurement of Development and Growth

We measured development time, growth and body mass at eclosion, and body size and sex when individuals were killed 72 h post-eclosion (see below). Development time was calculated as the number of days between hatching and eclosion. On the first day of each larval instar stage, individuals were weighed to the nearest 0·1 μg using an electronic balance (UMX2 ultra-microbalance; Mettler-Toledo Ltd.) and then placed in a clean Petri dish. Individuals were not fed post-eclosion. Newly eclosed adults were weighed (as above), then placed in a clean Petri dish to allow elytra coloration to develop from being pale yellow at eclosion to bright orange within 72 h (Majerus & Kearns 1989). At 72 ± 12 h post-eclosion, individuals were killed by placing them at −20 °C for 30 min, and then carcasses were immediately placed on ice, sexed by examination of the cuticular plates (Majerus & Kearns 1989) and photographed under standardized lighting using a colour mosaic digital camera to record the colour pattern geometry and body size (model 11.2; Diagnostic Instruments Inc., Sterling Heights, Michigan, USA). Elytra were carefully removed using dissecting scissors and precision forceps under a binocular dissecting microscope (Leica MZ12.5; Meyer Instruments, Houston, Texas, USA), and the dorsal surface of individual elytra was photographed (as above) before carcasses and elytra were stored individually in screw-top microtubes under N2 gas and at −80 °C until measurements of colour and biochemical analyses (see below). Digital images were analysed using Scion Image software (version; Scion Corporation, Frederick, Maryland, USA) to obtain measurements of body size (pronotum width) and spot area. The perimeter of each spot was demarcated for area measurements using the ‘threshold’ tool in Scion Image. All measurements were made to the nearest 0·001 mm.

Measurement of Elytra Coloration

Reflectance measurements of elytra were made by HR using an Ocean Optics USB2000 spectrometer (Ocean Optics, Dunedin, Florida, USA), with specimens illuminated at 45° to normal by a DH1000 balanced halogen deuterium light source; five measurements were made for each elytron approximately equidistant between the black spots and mean values for each individual used for analyses. The reflectance data were used to model avian cone photon catches (following Maddocks, Church & Cuthill 2001; Endler & Mielke 2005) using irradiance spectra from clear skies, collected in the field using an Ocean Optics USB2000 spectrometer fitted with a cosine corrector. The starling (Sturnus vulgaris) was chosen as a typical passerine for which there are good data on receptor spectral sensitivity (Hart, Partridge & Cuthill 1998). We calculated avian perceived luminance as a function of the photon catch of a starling’s double cones in the dorsal region of the retina (Hart, Partridge & Cuthill 1998). We modelled the predicted photon catches of the single cone types (longwave or LW, mediumwave or MW, shortwave or SW, and ultraviolet UV) of a starling for each spectrum in tetrahedral colour space. To describe how much the spectra of Low-diet and High-diet elytra were separated in receptor space, chromatic and achromatic contrasts were calculated [see Endler & Mielke (2005) and Vorobyev et al. (1998) for equations] in terms of just noticeable differences (JNDs). A JND value of one is at the threshold of discrimination, and as values increase, objects become easier to discriminate.

Measurement of Elytra Carotenoids

Carotenoids were assayed by JDB. Elytra were weighed to the nearest 0·1 μg using an electronic balance (UMX2 ultra-microbalance; Mettler-Toledo Ltd.) and then added individually to chloroform (1 mL) in a glass vial, capped under N2 gas and left for 24 h to extract carotenoids. Extracts were then dried under N2 gas at 60 °C, before being re-dissolved in 0·5 mL hexane. Absorbance was measured between 350 and 550 nm using a spectrophotometer (Evolution 500; Thermo Electron Corporation, Waltham, Massachusetts, USA), and total carotenoid concentration (μg g−1 tissue) was calculated using the formula: A/E × volume redissolved (mL)/elytra mass (g), where A is the absorbance of the sample (λmax = 445) and E is the average extinction 1% per 1 cm of xanthophylls in hexane (0·2500). Carotenoid concentrations in left and right elytra were positively correlated (Pearson’s = 0·738, = 31, < 0·0001), and therefore mean values are presented.

Measurement of Alkaloids

Alkaloids were measured by FPD as described previously (Sloggett, Obrycki & Haynes 2009) with some modifications. Whole ladybirds (minus elytra) were crushed in chloroform (1 mL), vortexed for 1 min, then centrifuged at room temperature and 12 000 RPM (c. 9600 g) for 5 min. The clear supernatant was separated and analysed with GC-MS for precoccinelline and with LC-MS for coccinelline. In both cases, warfarin was used as an internal standard; however, for the GC-MS analysis, warfarin was analysed as its trimethylsilyl-ether. For GC-MS analysis, samples (2 μL) were injected in the split mode (split ratio 20:1) into a HP 6890 GC (equipped with a HP-5MS column; length: 30 m; ID: 0·25 mm; film thickness: 0·25 μm) connected to a HP5973 quadrupole mass spectrometer with 70-eV electron impact ionization. The oven programme was as follows: 50 °C (1 min hold) to 80 °C at 10 °C min−1, then from 80 to 150 °C at 4 °C min−1 and finally from 150 to 320 °C at 20 °C min−1 with a 2-min hold at 320 °C. Helium was used as carrier gas, at a constant flow rate of 1·0 mL min−1. The injector was set at 250 °C and the transfer line at 280 °C. The MS was used in scan mode from 50 to 650 amu. For LC-MS analysis, samples (20 μL) were injected into an Alliance 2690 HPLC system equipped with a ODS column [ACE C18; 150 × 2·1 mm (i.d.); 3 μm particle size], operated in gradient mode at a flow rate of 0·2 mL min−1. The solvents were water with 0·1% formic acid (A) and acetonitrile (B). A mixture of 95% solvent A was held for 2 min, followed by a gradient over 13 min to 90% solvent B, returning to the original mixture of 95% A over 5 min, which was then held for 13 min before the next injection. Mass spectra were determined using a triple-quadrupole Quattro II system (Micromass UK Ltd, Manchester, UK), with the capillary voltage of the electrospray probe set to +3·5 kV. The nebulizing gas (nitrogen) was set at 20 L h−1, and the drying gas (nitrogen) was set at 350 L h−1. The source temperature was at 120 °C, and the sampling cone voltage was set at 30 V. Results are presented as the total weight of precoccinelline, or coccinelline, in micrograms; this represents the total investment in toxins by prey, and the toxin load that a predator would experience, and as such is pertinent to interpretations of signal honesty.

Data Analyses

Data were examined for normality, homoscedasticity and outliers. Data for development time (days from hatching to eclosion, which ranged from 16 to 19 day) were analysed using ordinal regression with diet group (Low or High) and sex as fixed factors, and a logit link function. Growth rate was calculated as the regression coefficient of an exponential curve fitted to the x(age in days at the start of each instar stage) and y(body mass) for each larva. Data for larval growth rate, body mass and size at eclosion, average spot size, concentrations of carotenoids in elytra, and total body levels of precoccinelline and coccinelline, respectively, were analysed using GLMMs with diet and sex as fixed factors, and dyad as a random factor to control for paired subjects. Relationships between total body levels of alkaloids (either precoccinelline or coccinelline) and elytra carotenoid concentrations were analysed using GLMMs with diet group, sex and elytra carotenoids as fixed factors and dyad as a random factor. Initially, we also included body mass at adulthood as a covariate in all analyses. Typically, body mass was far from significant ( 0·38), whilst it was marginally non-significant (= 0·054) in an analysis together with diet group as a fixed factor, dyad as a random factor and elytra carotenoid concentration as the dependent variable. However, retaining body mass in this model only served to increase the significance (= 0·004) of the diet effect which we report (see Results). Therefore, we did not include body mass as a covariate in the main results presented below. Other statistical tests are introduced in the text of the Results. Model simplification was performed by backward elimination of terms starting with the highest order interaction; all interactions were included in initial models. The significance of terms was calculated using the F distribution and α = 0·05. To assess the importance of the random term, dyad, we calculated model fit using pseudo r2 values (Cox & Snell 1989). We calculated the importance of dyad as the r2 of a model with dyad as the random factor, minus the same model but with an uninformative random variable included instead of dyad, that is, all values = 1. Analyses were carried out using Genstat v. 12, except for pseudo r2 values that were calculated using r (R Development Core Team 2011). Values are presented as mean estimates ±1 SE.


Developmental Rate, Body Mass and Size

High-diet larvae reached eclosion slightly, but significantly earlier than Low-diet larvae (ordinal regression: diet, Wald χ2 = 10·34, d.f. = 1, = 0·001; sex, χ2 = 3·00, d.f. = 1, P = 0·08; diet × sex, χ2 = 3·10, d.f. = 1, = 0·08; Low diet, 17·87 ± 0·17 day (mean ± SE); High diet, 17·00 ± 0·13 day). Similarly, High-diet larvae had a higher rate of growth than Low-diet larvae (Table 1; exponential slope: Low diet, 0·478 ± 0·0013 (mean ± SE); High diet, 0·521 ± 0·0019). At eclosion, females were heavier and larger than males, but body mass and size did not differ significantly between diet groups (Table 1; male body mass, 29·29 ± 0·77 mg (mean ± SE); female body mass, 34·97 ± 0·90 mg; male pronotum width, 3·45 ± 0·03 mm; female pronotum width, 3·60 ± 0·03 mm). Therefore, food supply during development influenced the time taken to reach adulthood and growth rate, but in terms of body mass and size individuals of the two diet groups were statistically indistinguishable as newly emerged adults.

Table 1.   Variation in larval growth rate, and morphology and aposematic traits at adulthood, arising from GLMMs with rearing diet (Low or High) and sex as fixed factors, and controlling for paired subjects by including dyad as a random factor (see Materials and Methods for details). The rearing diet by sex interaction was non-significant in all cases (> 0·05).
 Rearing dietSexDyad (random factor) Cox-Snell r2
Growth rate6·131,14·50·0260·411,25·80·530·00667
Body mass0·641,14·60·6423·991,28·8<0·0010·00227
Pronotum width0·241,14·00·6418·751,26·1<0·0010·08819
Mean spot size1·461,15·00·250·601,25·80·450·00000
Elytra carotenoids5·841,15·00·0291·431,26·70·240·00000
Total precoccinelline4·951,14·00·0431·011,17·70·330·25376
Total coccinelline2·561,14·60·130·951,27·50·340·00110

Signal Expression and Conspicuousness

High-diet individuals had greater concentrations of carotenoids in elytra than Low-diet individuals (Fig. 1a and Table 1). In tetrahedral colour space, High-diet individuals of both sexes were ‘redder’– the average chroma being closer to the LW cone – compared to Low-diet individuals, particularly in comparison with Low-diet males (diet, F1,14·2 = 4·94, = 0·043, sex, F1,19·4 = 1·07, = 0·32; diet × sex, F1,24·5 = 7·02, = 0·014; dyad (random factor), Cox-Snell r2 = 0·00043; Low-diet males, 0·465 ± 0·008 (predicted mean ± SE); Low-diet females, 0·500 ± 0·011; High-diet males, 0·509 ± 0·009; High-diet females, 0·495 ± 0·009). The difference in elytra coloration between Low- and High-diet groups was discriminable to a passerine (starling), both in terms of brightness/achromatic contrast [one-sample t-test, comparing JNDs between paired ladybird larvae (dyads) against a mean of 1: t14 = 4·15, < 0·001; mean JND = 7·54 ± 1·57 SE] and ‘colour’ chromatic contrast (t14 = 3·57, P = 0·003; mean JND = 14·69 ± 3·84 SE). In contrast to the effects of diet on carotenoid coloration, average spot size did not differ significantly between diet groups or sexes (Table 1).

Figure 1.

 Effects of developmental diet and sex on components of aposematism. (a) Elytral carotenoid pigmentation. (b) Total levels of precoccinelline. (c) Total levels of coccinelline. Values are estimates (±SE) from GLMM analyses controlling for dyad (see text of Results). Means denoted by different letters are significantly different to each other (< 0.05).

Body Levels of Alkaloids

In both sexes, total levels of precoccinelline were greater in the High-diet group than in the Low-diet group (Fig. 1b and Table 1), whereas total levels of coccinelline did not differ significantly between diet groups or sexes (Fig. 1c and Table 1). Levels of coccinelline were not significantly correlated with levels of precoccinelline (GLMM controlling for diet and sex; effect of precoccinelline, F1,20·9 = 0·62, = 0·44; all interactions, N.S.; dyad (random factor), Cox-Snell r2 = 0·00204).

Correlations Between Signals and Toxin Levels

Total levels of precoccinelline were positively correlated with elytra carotenoid concentrations in both diet groups and sexes (GLMM: diet, F1,14·0 = 1·58, = 0·23; sex, F1,17·0 = 0·36, = 0·56; elytra carotenoids, F1,17·0 = 5·03, = 0·039; all interactions,  0·50; dyad (random factor), Cox-Snell r2 = 0·22262; Fig. 2a). In contrast, the relationship between elytra carotenoid concentrations and total levels of coccinelline did not differ between diet groups (GLMM: diet, F1,16·3 = 1·00, = 0·33), but depended on sex; females that had greater concentrations of carotenoids in elytra had higher total levels of coccinelline, whereas in males the opposite relationship was found (sex, F1,25·9 = 0·60, = 0·45; elytra carotenoids, F1,26·0 = 0·83, = 0·37; sex × elytra carotenoids, F1,24·4 = 5·76, = 0·024; all other interactions, N.S; dyad (random factor), Cox-Snell r2 = 0·00000; Fig. 2b).

Figure 2.

 Effects of developmental diet (Low diet, circles; High diet, triangles) and sex (males, open symbols; females, filled symbols) on relationships between components of aposematic coloration and pattern, and total levels of alkaloids. (a) Elytra carotenoids and precoccinelline. Single fitted line because of a lack of interaction with sex or diet. (b) Elytra carotenoids and coccinelline. Fitted lines are dashed for males and solid for females. (c) Coccinelline and spot size. Fitted lines are dashed for Low diet and solid for High diet. In each figure, the raw data are scaled up or down the y-axis according to how much each dyad deviated from the ‘average’ dyad (i.e. the random dyad intercept is scaled out of the figures).

Average spot size was not correlated with total levels of precoccinelline in either diet group or sex (F1,15·3 = 0·12, = 0·74; all interactions, N.S.; dyad (random factor), Cox-Snell r2 = 0·26460). However, the relationship between total levels of coccinelline and average spot size differed between diet groups; Low-diet individuals with larger spots had higher total levels of coccinelline, and this was apparent in both sexes, whereas in the High-diet group larger spots were associated with lower coccinelline levels in both sexes (diet, F1,14·0 = 2·73, = 0·12; sex, F1,24·8 = 1·44, = 0·24; spot size, F1,25·5 = 0·02, = 0·89; diet × spot size, F1,26·7 = 4·42, = 0·045; all other interactions, N.S; dyad (random factor), Cox-Snell r2 = 0·00184; Fig. 2c). To explore the potential causality of this relationship further, we tested whether there was a negative relationship, suggestive of a trade-off, between spot size and elytra carotenoid concentrations. Individuals with higher elytra carotenoid concentrations had smaller spots in both diet groups and both sexes (diet, F1,15·2 = 0·45, = 0·51; sex, F1,28·0 = 0·38, = 0·55; elytra carotenoids, F1,25·8 = 4·71, = 0·039; all interactions,  0·28; dyad (random factor), Cox-Snell r2 = 0·003905).


Our findings add to the growing number of studies of phenotypic associations between aposematic coloration and chemical defences (Grill & Moore 1998; Grill 1999; Summers & Clough 2001; Darst, Cummings & Cannatella 2006; Bezzerides et al. 2007; Ojala, Lindström & Mappes 2007; Sandre et al. 2007a; Lindstedt, Lindström & Mappes 2009; Cortesti & Cheney 2010; Lindstedt et al. 2010; Maan & Cummings 2012). We found that diet during larval development influenced the expression of aposematic signals and chemical defences at adulthood in C. septempunctata. Both developmental diet and sex influenced the relationships between aposematic coloration and pattern, and chemical defences. Below we discuss how these results support or contradict the predictions of the resource competition model, which itself attempts to explain how aposematic traits should correlate (Blount et al. 2009).

Do Individuals with More Resources Have Increased Investment in Toxins?

A key prediction of the resource competition model (Blount et al. 2009) is that individuals with more resources (i.e. food) should have increased investment in toxins. Consistent with this, High-diet individuals had elevated body levels of precoccinelline compared to Low-diet individuals. In contrast, levels of coccinelline did not differ significantly between diet groups. A possible explanation for this result is that synthesis of coccinelline has yet to attain full operational capacity in recently eclosed adults. Both these alkaloids are synthesized in the fat body, precoccinelline being the free base of coccinelline (Pasteels et al. 1973). Coccinella septempunctata is aversive and harmful to avian predators if consumed (Marples, Brakefield & Cowie 1989; Dolenskáet al. 2009), although the relative importance of precoccinelline and coccinelline in conferring chemical defence is not known. Both alkaloids could function as chemical deterrents, because the sixteen-spot ladybird (Tytthaspis sedecimpunctata) synthesizes only the free bases hippodamine (the stereoisomer of precoccinelline) and precoccinelline itself, but no N-oxides (Daloze, Braekman & Pasteels 1995). Therefore, correlations between body levels of precoccinelline and signals seem likely to be functionally relevant for aposematism in young adult C. septempunctata.

Does Investment in Aposematic Signalling Vary According to Resource Supply, and Does it Influence Conspicuousness?

The resource competition model (Blount et al. 2009) predicts that investment in signalling will vary according to resource supply; at relatively low resource states, that is, where resources are limiting, we would expect to see a positive correlation between resource supply and resource allocation to signals. In our experiment, this prediction would be supported if High-diet individuals had greater investment in signals than Low-diet individuals. In contrast, where resources are abundant and non-limiting, the theoretical expectation is that individuals should reduce investment in signalling (and thereby avoid detection costs) and instead to invest disproportionately highly in toxins (because very toxic prey can survive attacks unharmed). Under this scenario, signal expression in Low-diet and High-diet individuals may not be discernable. Consistent with the former prediction, that is, resource limitation, we found that individuals that received more aphids during larval development had greater elytral concentrations of carotenoids at adulthood. Indeed, the fact that all individuals always consumed all aphids which were provided to them strongly suggests that both High-diet and Low-diet individuals were resource-limited, albeit to differing extents. Animals must obtain carotenoids through the diet (Goodwin 1984), although aphids are in fact an exception, being capable of carotenoid biosynthesis (Moran & Jarvik 2010). However, there is no reason to expect that aphids fed to Low- and High-diet ladybirds would have differed in their carotenoid concentration. Rather, High-diet individuals must have received more carotenoids by consuming larger numbers of aphids, which they allocated to increase elytra pigmentation and ‘redness’. This finding is in agreement with previous studies of the effects of developmental diet on the size (Ojala, Lindström & Mappes 2007) and colour intensity of warning signals (Grill & Moore 1998; Lindstedt et al. 2010) in other species, confirming that warning signal expression is to some extent plastic and sensitive to developmental conditions. An important assumption of the resource competition model (Blount et al. 2009) is that higher investment in signalling should incur increased costs of conspicuousness. Our results support this assumption, in so far as a typical passerine, the starling, would be able to discriminate between the elytra coloration of Low-diet and High-diet individuals. It therefore seems likely that individuals of the two diet treatments would differ in conspicuousness when presented against natural backgrounds. However, this requires empirical verification in the field, together with measurement of whether starlings (or other potential predators) show a greater propensity to attack less conspicuous prey once they have been detected.

Does Resource Supply Determine Whether Signals are ‘Honest’ or ‘Dishonest’?

Body levels of precoccinelline were positively correlated with elytra carotenoid concentrations in males and females of both diet groups. As elytral carotenoids influenced detectability to predators (see above), this suggests that carotenoid pigmentation was an ‘honest’ signal of toxin levels in both diet groups, as would be predicted if resources were limiting (Blount et al. 2009).

The relationship between body levels of coccinelline and elytra carotenoid concentrations showed a different pattern, being influenced by sex. In both diet groups, females that had greater carotenoid pigmentation had more coccinelline, whereas in males greater carotenoid pigmentation was associated with lower levels of coccinelline. As discussed above, we found no evidence that experimental diet influenced the production of coccinelline, which could simply be because we studied recently eclosed adults. However, it also seems possible that the effect of sexual size dimorphism on resource states masked any influence of experimental diets. Females were relatively large and weighed more than males in both diet groups, indicating that sexual size dimorphism was largely determined by genetic and/or pre-experimental maternal effects. Relatively large body size in female C. septempunctata and other ladybird beetles is an adaptation for egg production (Pasteels et al. 1973). The resource competition model (Blount et al. 2009) did not address potential sex differences in resource limitation, but given that ladybirds are sexually size dimorphic, it is plausible that females should be more susceptible to resource limitation, and hence more likely to signal honestly, than males.

An alternative possibility is that the optimal strategy for females is nearly always to signal honestly (at least in respect of aposematic coloration), because their relatively large body size means they offer a greater nutritional reward to predators and, potentially, makes them more conspicuous than males (Mänd, Tammaru & Mappes 2007; Sandre, Tammaru & Mänd 2007b). In fact, whether females are more conspicuous than males is equivocal, because males tend to be the more active sex which may increase their probability of detection. Nevertheless, predators may be more likely to encounter females because ladybird populations often exhibit female-biased sex ratios, as a consequence of male-biased mortality during over-wintering and because of male-killing bacteria (Ottenheim, Holloway & de Jong 1992; Hurst & Jiggins 2000). It would therefore be interesting to investigate whether aposematic signals of females are more ‘honest’ in populations that have female-biased sex ratios.

There were no significant relationships between body levels of precoccinelline and spot size in either diet group or sex. However, coccinelline levels were positively correlated with spot size in Low-diet individuals, whereas these traits were negatively correlated in High-diet individuals. The importance of spot size in ladybirds as a determinant of detectability and deterrence to predators is not known, although work using artificial stimuli has shown that large pattern elements confer an enhanced probability of surviving encounters with predators (Forsman & Merilaita 1999). Therefore, one possible interpretation of our results is that, as predicted by theory (Blount et al. 2009), Low-diet individuals signalled ‘honestly’ whereas High-diet individuals signalled ‘dishonestly’ with respect to body coccinelline levels through the expression of spot size. However, this seems unlikely because the experimental diets did not affect coccinelline levels or spot size per se. An alternative possibility is that spot size was influenced by the level of carotenoid deposition in elytra; we found that spot size and elytra carotenoid concentrations were negatively correlated, suggestive of a trade-off. Intriguingly, in their correlational study of H. axyridis, Bezzerides et al. (2007) found that fainter spots (i.e. lower melanin content) were associated with higher total levels of the defensive alkaloid harmonine in females, but not in males, as may be expected if females are more resource constrained than males. However, they did not measure spot size, whereas we did not measure spot brightness, and therefore whether spot size and spot coloration are themselves correlated is not known. Ultimately, the importance of variation in spot size and colour for aposematic signalling in ladybirds need to be experimentally tested, to inform our understanding of signal honesty.

Physiological Linkage Between Aposematic Traits

One explanation for why aposematic traits may vary within species is that there are underlying physiological costs associated with the acquisition or biosynthesis of resources required to produce coloration, or that investment in coloration trades against other functions (e.g. Grill & Moore 1998; Ojala, Lindström & Mappes 2007; Lindstedt, Lindström & Mappes 2008, 2009; Lindstedt et al. 2010). Similarly, variation in body levels of toxins may indicate that the sequestration or production of toxins is costly (e.g. Rowell-Rahier & Pasteels 1986; Holloway, de Jong & Ottenheim 1993; reviewed in Ruxton, Sherratt & Speed 2004). Some empirical studies have revealed phenotypic linkage between aposematic signals and chemical defences (Summers & Clough 2001; Darst, Cummings & Cannatella 2006; Bezzerides et al. 2007; Cortesti & Cheney 2010; Maan & Cummings 2012), but whether investment in such traits is physiologically linked remains equivocal (Grill & Moore 1998; Grill 1999; Lindstedt et al. 2010). For example, in H. axyridis, investment in chemical defence (reflex bleeding) during the larval stage of development has been shown to have strong deleterious effects on life-history traits such as growth rate, but no (Grill 1999) or relatively subtle effects on elytra coloration at adulthood (Grill & Moore 1998). Similarly, in wood tiger moths, the detoxification of iridoid glycosides (IGs) sequestered in the diet was clearly costly; larvae reared on a high IG diet produced fewer offspring and had reduced signal expression (Lindstedt et al. 2010). However, body levels of IGs and the size or coloration of signals were not correlated, leading to the conclusion that costs of warning colour production and toxin sequestration were not physiologically linked (Lindstedt et al. 2010).

Our results support the suggestion that production of toxins and signals is susceptible to resource limitation; total levels of precoccinelline and elytra carotenoid concentrations were constrained on the Low diet. Furthermore, our observation that certain relationships between signals and toxin levels differed according to diet or sex points to physiological linkages between components of aposematism which are modulated by food availability. We do not know whether these effects were attributable to variation in the supply of energy, or alternatively, a particular nutrient(s); this would be interesting to study in the future. Compounds that have dual potential roles as pigments and as antioxidants (e.g. carotenoids) are candidate resources, because the sequestration of toxins (Ahmad & Pardini 1992) or their biosynthesis may impose an oxidative challenge that depletes antioxidant reserves and therefore trades against the capacity to produce aposematic coloration (Blount et al. 2009). Antioxidants have multiple biological functions, and it is conceivable that individuals of aposematic species must trade antioxidants amongst the production of aposematic traits, as well as growth, immune defence, future reproductive capacity and survival. Such relationships have begun to be explored at the phenotypic level and using manipulations of whole foods to alter physiological state (Grill & Moore 1998; Ojala, Lindström & Mappes 2007; Sandre et al. 2007a; Lindstedt et al. 2010; Lindstedt, Lindström & Mappes 2008, 2009; this study). A valuable direction for further study will be to combine such measures of phenotypic correlations at the organismal level and detailed investigations of the underlying physiological mechanisms, for example through manipulation of specific dietary nutrients or oxidative balance.


We thank D. Newcombe and J. Phillips for field assistance, and the editors and anonymous reviewers for helpful comments on an earlier draft. This work was funded by a Royal Society Research Fellowship to JDB and a Natural Environment Research Council grant NE/D010667/1 to MPS and HMR.