Plastic background colour matching in the springbok mantis
Abstract
- Within-species variation in colour phenotypes is widespread in animals. One mechanism by which such variation can be maintained is plastic background matching, where individuals plastically develop a similar colour to that of their surroundings. A few examples are known from insects that exhibit green–brown colour polyphenisms. But the extent to which plastic colour responses are shaped by other factors, such as genetic variation in plasticity or the interaction of other environmental cues, is poorly understood.
- Here, we investigate the plasticity of body coloration in the springbok mantis, Miomantis caffra—a species where hatchlings emerge brown in colour and typically change to green but sometimes remain entirely or partly brown through successive moults. We reared 350 mantises from 10 full-sib families on a green or brown background under a high or low temperature and a high or low humidity using a fully factorial, split-brood design, and recorded colour phenotypes (all green, all brown or mixed coloration) after 14 weeks of development.
- We found very strong evidence of developmental plasticity for background matching: The green background induced a higher incidence of the all-green phenotype, whereas the brown background produced more of the all-brown and mixed phenotypes. The all-green phenotype was also universally more common under higher humidity, and under higher temperature when the background was green. However, not all body parts showed the same level of environmental sensitivity: The steepest reaction norms were observed in the mid-legs and hindlegs, potentially reflecting selection for disruptive coloration of the body outline in browner environments. Using model comparison techniques, we found little evidence of genotype-level variation in colour plasticity—a pattern likely the result of strong viability selection for camouflage.
- Our study shows how developmental plasticity in coloration can be triggered directly by the colour of the environment and indirectly by climatic cues associated with habitat coloration. We argue that this high level of developmental plasticity has likely evolved due to the diversity of habitats but sedentary lifestyle of this sit-and-wait predator.
Read the free Plain Language Summary for this article on the Journal blog.
1 INTRODUCTION
Individuals of the same species often express strikingly different colour phenotypes (Mclean & Stuart-Fox, 2014; Sapir et al., 2021). In animals, such phenotypes can manifest as discrete colour morphs or continuous differences in colour (Nijhout, 2003). Both forms of variation can provide camouflage benefits by expanding the breadth of the habitat niche in which a species is cryptic (Cuthill et al., 2005; Forsman & Åberg, 2008; Troscianko et al., 2018; Wennersten & Forsman, 2012). Because distinct colour phenotypes can be genetically programmed polymorphisms (Roulin et al., 2004; Wellenreuther et al., 2014) or environmentally induced polyphenisms (Booth, 1990; Nijhout, 1999), the evolutionary maintenance of such colour differences can be achieved either through disruptive selection on alleles for coloration (Allen & Greenwood, 1988; Ford, 1945) or alleles for habitat matching through plastic colour change (Duarte et al., 2017; Pfennig, 2021). Knowing the genetic versus environmental basis of colour differences is therefore important for understanding evolutionary dynamics, since the response of populations to selection will depend on the heritability of coloration if it is genetically determined (Falconer & Mackay, 1996) and on the genetic variation underpinning colour plasticity if it is environmentally induced (Goldstein & Ehrenreich, 2021).
A notable example of intra-specific colour variation is the widespread green–brown colour morphology of many insects (Bedford, 1978; Castner & Nickle, 1995; Eacock et al., 2017; Edmunds, 1972; Goodman, 2021; Köhler et al., 2017; Musolin & Numata, 2003; Winter et al., 2021; Yumnam et al., 2021), which likely corresponds to the dominance of various shades of green and brown in terrestrial habitats. Green–brown colour differences are typically characterised by the co-occurrence of green morphs and brown morphs, and can be phenotypically plastic when colour differences are induced by environmental cues (Booth, 1990; Nijhout, 1999). Individuals can either respond directly to the colour cue by perceiving their surroundings and expressing a similar colour (Boyle & Start, 2020; Eacock et al., 2017), or by perceiving other environmental factors that predict habitat colour, such as temperature (Valverde & Schielzeth, 2015; van Bergen & Beldade, 2019), humidity (Mayekar & Kodandaramaiah, 2017), day length (Musolin & Numata, 2003) or even substrate texture (Hiraga, 2006). Indirect colour matching through plastic responses to correlated environmental variables often occurs in habitats that vary seasonally in coloration, such as in temperate zones (see Mayekar & Kodandaramaiah, 2017). But the extent to which green–brown colour morphs are induced by direct versus indirect cues acting independently or in concert is poorly known. Understanding which combination of cues induces colour change is important for determining not only how camouflage through background matching is mechanistically achieved but also how plastic responses are likely to play out in the complexity of real-world habitats.
Colour morphs do not always consist of a single colour, however. In intermediate morphs, coloration can vary between body regions, such that some sections of the body are more or less likely to be one colour or another (e.g. Winter et al., 2021). Such intra-individual heterogeneity in body coloration could be favoured if it enhances crypsis, such as by disguising the body outline through disruptive coloration (Merilaita & Lind, 2005; Price et al., 2019; Stevens & Merilaita, 2009). While much research has focused on the cryptic function of contrasting coloration, the mechanisms that produce heterogeneous colour arrangements within individuals are less well understood. A simple genetic architecture underlying the inheritance of green versus brown body segments has been inferred for the steppe grasshopper (Winter et al., 2021), suggesting that intra-individual colour variation can be genetically determined. Plasticity could also play an important role in producing the mixed colour patterns characteristic of intermediate colour morphs if certain body regions are more sensitive to environmental cues than others. Comparing levels of plasticity among different body regions would be one way to test this idea, but such tests are scarce.
Even when colour traits are environmentally induced, plasticity itself can show genetic variability, with different genotypes responding to the same environmental cues in different ways (Sultan, 2021; Touchon & Robertson, 2018). Such variation is often characterised as intersecting norms of reaction, or ‘reaction norms’ (Woltereck, 1909; see Figure 1d). Genotype-by-environment interactions are expected to be widespread in nature (Stearns, 1989) and may play an important role in facilitating rapid evolution of plasticity in altered environments (Corl et al., 2018; Ghalambor et al., 2007). However, genetic variation in plasticity can potentially be depleted if strong selection drives alleles for plasticity to fixation (Saltz et al., 2018), which can result in genotypes showing almost identical overlap in reaction norms (see Figure 1b). This may be especially true for alleles underpinning plasticity in viability traits (Saltz et al., 2018), such as those involved in camouflage for predator avoidance. Although accumulating evidence indicates that plasticity in coloration can vary by genotype (Bergstrom et al., 2012; Edelaar et al., 2017), the extent to which genetic variation underpins colour change in response to multiple interacting cues is poorly understood.
Here, we investigate the interactive effects of different environmental cues on body coloration in the springbok mantis, Miomantis caffra, and the genetic variation that underpins plastic colour change. This praying mantis is native to South Africa but was introduced to New Zealand 40 years ago and has become naturalised there (Ramsay, 1984). Both juveniles and adults vary in coloration, ranging from all green to all brown, including mixed phenotypes that consist of various combinations of green and brown body parts (see Figure 2). The lifecycle is annual: Hatchlings emerge from overwintered oothecae in the New Zealand spring and summer and go through six to eight instars before reaching adulthood in summer and autumn (pers. obs.). Very few, if any, mantises survive the winter months. While this species mostly occurs in suburban parks and gardens in its introduced habitat, it can be found on a wide variety of plants and foliage types, and on the ground (pers. obs.). Females are flightless their whole lives, whereas males can fly as adults. As sit-and-wait predators, these mantises are typically inactive, moving only when searching for a mate, struggling to copulate, capturing prey or when disturbed (pers. obs.).
To investigate colour plasticity, we reared mantis offspring from full-sib families on a green or brown background at a high or low temperature and a high or low humidity and recorded changes in body colour. If M. caffra changes colour to directly match the background, we predicted that background colour would induce a phenotype of similar colour. However, because this species occupies temperate latitudes in its home and invasive ranges, we anticipated that climatic variables that predict foliage coloration might also affect body colour. Based on the typical patterns of colour plasticity observed in other temperate zone insects (see Mayekar & Kodandaramaiah, 2017), we anticipated that high temperature and humidity associated with actively growing foliage might promote green matching, and low temperature and humidity associated with senescent foliage might promote brown matching. We also predicted that variable plasticity would be apparent when seasonal variables mismatched the background, since mantises would receive contradictory information about how to develop. If M. caffra shows sensitivity to background coloration, we anticipated that variation in the mixed phenotype might be explained by different levels of plasticity in the different body regions, and that such differences would be evident in the steepness of the reaction norms of these regions. Finally, if colour plasticity in M. caffra is genotype dependent, we expected to see strong evidence of genotype-by-environment interactions.
2 MATERIALS AND METHODS
2.1 Experimental methodology
A total of 350 hatchlings from 10 full-sib families derived from mothers and fathers of various colour morphs were used in the experiment. We initially started with 428 hatchlings, but 78 died during the experiment and so were excluded from subsequent analyses. Parents were obtained from 10 different populations in Auckland, New Zealand. Each mother was crossed with a father from a different population to avoid potential inbreeding. Immediately following hatching, each mantis offspring was placed inside a separate clear plastic enclosure (50 mm diameter, 95 mm high) that had a mesh ceiling for airflow and a pipe cleaner glued to the inside for perching. Thus, all offspring were reared one individual per enclosure. Mantises were fed several fruit flies three times per week during early instars and one to two houseflies three times per week during later instars, with occasional supplementary feeding of size-appropriate mealworms. Enclosures were sprayed with water every other day for mantises to drink. Ethical approvals were not required for research on these insects.
Mantises were reared in climate-controlled chambers under a 12:12 light:dark cycle and different interacting conditions of temperature (low/18°C vs. high/28°C), humidity (low/50% vs. high/100%) and background coloration (green vs. brown). The colour backgrounds were achieved by attaching either brown or green card around the outside half of each enclosure, with pipe cleaner perches colour-matched to the cardboard. To ensure mantises were never exposed to the opposite colour of their treatment, we grouped enclosures by background colour and separated them from opposite-coloured enclosures with white partitions. The location of each grouping within the chambers was randomised at each feeding event. Equivalent numbers of hatchlings from each full-sib family were allocated to each treatment combination in a split-brood design which allowed us to assess the contribution of genotype to the development of the mantises' colour phenotype. Realised sample sizes for each treatment combination are provided in Table S1.
Springbok mantis hatchlings emerge completely brown in colour (pers. obs.). But as they moult, their different body parts either change to green or remain brown. To assess changes in coloration, we recorded the colour (green vs. brown; scored as 1 vs. 0) of the dorsal integument of the head, thorax, abdomen, foreleg femur, mid-leg femur and hindleg femur for each mantis. Dorsal and ventral coloration did not differ, so we recorded the colour based on the dorsal appearance. For simplicity, the very few body parts that appeared dark yellow, tan or pink were classified as ‘brown’, since all such colours represent alternatives to green in green–brown polymorphisms. When both green and brown were present on a body part, the majority colour was recorded. Because coloration was symmetrical (i.e. for each leg position, the left leg of an individual was always the same colour as the right leg), each leg type was recorded as a single score. These data were collected at 14 weeks when mantises were in the fifth or sixth instar. The overall colour phenotype of each mantis was then recorded as ‘all green’ (i.e. all body parts majority green), ‘all brown’ (i.e. all body parts majority brown) or ‘mixed coloration’ (i.e. at least one body part majority brown and the others majority green). We chose to quantify colour as discrete phenotypes rather than as a continuous response because we suspected colour in this species to be a polyphenism (Nijhout, 2003). This was based on the observation that mantises with mixed body coloration in wild populations typically show stereotyped colour arrangement in the body segments with very little variability (pers. obs.), suggesting thresholds of colour difference rather than scales. We did not formally quantify the coloration of the mantises or their background using spectrometry because we were not interested in assessing the precise degree of match between the colour of the mantis and the colour of the background. Rather, we aimed to determine whether or not mantises alter their coloration in response to interacting environmental cues, whether body segments differ in their level of plasticity, and whether genotype affects plastic responses.
2.2 Population-level effects of environment on colour phenotype
To assess the influence of the environment on body coloration, we analysed colour phenotype as an ordinal response in a cumulative ordinal hierarchical Bayesian model using Hamiltonian Monte Carlo (HMC) no U-turn sampling (NUTS). The model was fitted with a flexible threshold and a probit link. Cumulative ordinal models are useful for analysing categorical data binned across a latent continuous response (Bürkner & Vuorre, 2019). Our colour phenotype response fits such a definition since its three levels (all brown, mixed coloration, all green) are categorisations of the proportion of the body that is green versus brown. The interacting predictors in our model were temperature, humidity and background colour, with family modelled hierarchically (i.e. as a ‘random effect’) to account for the split-brood design. We used weakly informative priors with distributions of N(0, 1.5) for intercepts, beta coefficients and standard errors. This distribution was chosen because it provides a relatively flat prior on the probability scale for logistic and ordinal regressions (McElreath, 2018) and therefore can account for our lack of prior knowledge about how mantis coloration responds to different environmental variables. We ran four chains for 5000 iterations with a warmup period of 2500 and no thinning, resulting in an iteration sample size of 10,000 and effective sample sizes >1000 for all model parameters.
2.3 Population-level effects of environment on the coloration of individual body parts
To assess levels of developmental plasticity in different body parts, we analysed the colour (green vs. brown; scored as 1 vs. 0) of the head, thorax, abdomen, forelegs, mid-legs and hindlegs in separate Bayesian hierarchical logistic regression models using NUTS. The family was Bernoulli, with a logit link. While it is possible that colour development covaries between body parts, for modelling simplicity, we have assumed independent responses for each body part. Temperature, humidity and background colour were the interacting predictors, with intercepts allowed to vary by genotype to account for the split-brood design. For the reasons described above, we used weakly informative priors with distributions of N(0, 1.5) for all intercepts, beta coefficients and standard errors. All other details of the models (iterations, warmup period, thinning) were the same as described above for the model of colour phenotype.
2.4 Genotype-level effects
To investigate the influence of genotype on colour plasticity, we built versions of the above-mentioned Bayesian models with and without genotype-varying intercepts and/or slopes and used model comparison techniques to identify the model with the best fit to the data (see Statistical methods for details). To model the null expectation (i.e. the absence of any environment or genotype effects; Figure 1a), we fitted an intercept-only model that possessed no predictors or grouping term (model 0). To model the effect of environment in the absence of genotype effects (Figure 1b), we fitted a model that possessed all environmental predictors but no hierarchical grouping term (model 1). To model the additive effect of environment and genotype (Figure 1c), we created a model that possessed all predictors and a hierarchical term that allowed intercepts to vary by genotype (model 2). And to model the multiplicative effect of environment and genotype (Figure 1d), we fitted a model that included all predictors and a hierarchical term that allowed intercepts to vary by genotype but also allowed slopes to vary by genotype within each combination of interacting environments (model 3). Note that the correlation between slope and intercept was not parameterised for model 3. We expected model 0 to have the best fit if colour development was unaffected by either environment or genotype, model 1 to have the best fit if colour was mostly determined by environment, model 2 to have the best fit if colour was characterised by additive effects of environment and genotype and model 3 to have the best fit if colour responses were the outcome of genotype-by-environment interactions.
2.5 Replication statement
| Scale of inference | Scale at which the factor of interest is applied | Number of replicates at the appropriate scale |
|---|---|---|
| Individuals | Individuals | 38, 40, 45, 49, 38, 46, 44, 50 (for each treatment combination of background, temperature and humidity) |
2.6 Statistical methods
All Bayesian models were fitted in Stan (Carpenter et al., 2017) via the statistical computing platform R, version 4.3.1 (R Core Team, 2023) using the brm function from the brms R package (Bürkner & Vuorre, 2019). We inspected diagnostic trace plots, effective sample sizes and R-hat statistics to confirm that chains converged and were not autocorrelated. All models were run with adapt_delta set to 0.9999 and a maximum tree depth of 20 to avoid false-positive divergences. To assess effects of environment versus genotype on colour plasticity, we compared models 0 through 3 using leave-one-out information criteria (LOOIC) which identifies the best fit model in a similar way to Akaike information criteria (AIC; Akaike, 1998). LOOIC cross-validation is a recommended technique for assessing statistical support for estimates of variance components in hierarchical HMC models (Gelman et al., 2021; Vehtari et al., 2017). We ran model comparisons using the kfold and loo_compare functions from the loo package (Vehtari et al., 2017). Best fit models were considered substantially different to other models if the 95% confidence interval (95% CI) of ΔLOOIC (calculated as ΔLOOIC ± 1.96 × SE) did not overlap zero (Bürkner & Vuorre, 2019). Regardless of the outcome of these model comparisons, all reported results are from our base split-brood models that appropriately account for the experimental design. For our analysis of environmental effects on phenotype and body part coloration, we report posterior mean estimates, estimate errors and 95% credible intervals (95% CrI) (Table 1). For genotype-specific effects, we report posterior mean deviations of genotype intercepts from the population intercept, along with estimate errors and 95% CrIs (Table 2). For pairwise comparisons of significant interaction effects, we report median point estimates and highest posterior density intervals (HPDI) (Table S3). We also report phenotype and colour counts and proportions for all treatment combinations (Tables S1, S2 and S4). Data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.nzs7h44xk (Burke & Holwell, 2023).
| Model parameter | Colour phenotype | Head colour | Thorax colour | Abdomen colour | Forelegs colour | Mid-legs colour | Hindlegs colour |
|---|---|---|---|---|---|---|---|
| Intercept 1 |
−0.48 (0.22) [−0.92, −0.04] |
0.49 (0.24) [0.01, 0.97] |
0.30 (0.26) [−0.23, 0.81] |
0.30 (0.22) [−0.13, 0.73] |
0.50 (0.25) [0.001, 0.99] |
−1.23 (0.26) [−1.77, −0.73] |
−1.43 (0.30) [−2.06, −0.87] |
| Intercept 2 | 1.47 (0.24) [1.01, 1.95] | NA | NA | NA | NA | NA | NA |
| Background |
2.02 (0.26) [1.51, 2.53] |
2.49 (0.69) [1.34, 4.04] |
1.81 (0.41) [1.05, 2.66] |
1.98 (0.47) [1.16, 2.99] |
2.54 (0.69) [1.37, 4.06] |
1.85 (0.29) [1.27, 2.42] |
1.99 (0.31) [1.41, 2.60] |
| Temperature |
0.03 (0.23) [−0.41, 0.47] |
0.21 (0.27) [−0.33, 0.75] |
0.40 (0.27) [−0.12, 0.94] |
0.23 (0.26) [−0.28, 0.74] |
0.13 (0.27) [−0.40, 0.67] |
0.15 (0.31) [−0.45, 0.75] |
−0.38 (0.37) [−1.13, 0.34] |
| Humidity | 0.69 (0.23) [0.24, 1.15] | 1.01 (0.33) [0.38, 1.66] |
1.00 (0.31) [0.41, 1.62] |
1.05 (0.31) [0.46, 1.67] |
1.03 (0.33) [0.40, 1.69] |
0.28 (0.30) [−0.30, 0.88] |
0.45 (0.32) [−0.16, 1.07] |
| Background × Temperature | 0.98 (0.40) [0.21, 1.77] |
0.64 (1.06) [−1.30, 2.93] |
1.21 (0.96) [−0.46, 3.26] |
−0.18 (0.63) [−1.40, 1.06] |
0.69 (1.08) [−1.28, 2.97] |
1.04 (0.48) [0.11, 1.99] |
1.60 (0.51) [0.62, 2.63] |
| Background × Humidity |
−0.04 (0.36) [−0.75, 0.67] |
0.32 (1.15) [−1.71, 2.77] |
−0.64 (0.63) [−1.83, 0.64] |
−1.29 (0.60) [−2.48, −0.14] |
0.29 (1.16) [−1.77, 2.74] |
0.29 (0.41) [−0.53, 1.10] |
0.16 (0.43) [−0.67, 1.00] |
| Temperature × Humidity |
0.25 (0.33) [−0.40, 0.91] |
−0.78 (0.44) [−1.64, 0.09] |
−0.73 (0.43) [−1.57, 0.10] |
−0.74 (0.42) [−1.54, 0.07] |
−0.72 (0.45) [−1.61, 0.13] |
0.92 (0.41) [0.12, 1.74] |
1.27 (0.47) [0.38, 2.22] |
| Background × Temperature × Humidity |
0.94 (1.01) [−0.80, 3.13] |
0.20 (1.30) [−2.29, 2.79] |
0.57 (1.21) [−1.70, 3.01] |
1.66 (0.98) [−0.17, 3.71] |
0.19 (1.31) [−2.31, 2.82] |
0.65 (1.15) [−1.37, 3.10] |
0.50 (1.20) [−1.61, 3.05] |
- Note: The same data are visualised in Figure 5. Reference levels for comparisons are brown background, low temperature and low humidity. Parameters whose 95% credible intervals do not straddle 0 are in bold. Note that models of ordered responses consisting of n levels are characterised by n − 1 thresholds (Bürkner & Vuorre, 2019), hence the two intercepts for colour phenotype.
| Model parameter | Colour phenotype | Head colour | Thorax colour | Abdomen colour | Forelegs colour | Mid-legs colour | Hindlegs colour |
|---|---|---|---|---|---|---|---|
| Family 1 (intercept) | 0.68 (0.29) [0.16, 1.27] |
0.42 (0.37) [−0.17, 1.27] |
0.57 (0.40) [−0.08, 1.45] |
0.31 (0.31) [−0.15, 1.01] |
0.49 (0.38) [−0.13, 1.35] |
0.56 (0.30) [0.02, 1.19] |
0.74 (0.34) [0.13, 1.46] |
| Family 2 (intercept) |
0.20 (0.24) [−0.24, 0.69] |
0.28 (0.32) [−0.26, 1.01] |
0.41 (0.36) [−0.21, 1.20] |
0.21 (0.27) [−0.24, 0.84] |
0.32 (0.34) [−0.29, 1.05] |
0.14 (0.24) [−0.32, 0.63] |
0.10 (0.28) [−0.44, 0.68] |
| Family 3 (intercept) |
−0.27 (0.23) [−0.74, 0.16] |
−0.46 (0.32) [−1.14, 0.07] |
−0.49 (0.31) [−1.14, 0.07] |
−0.23 (0.26) [−0.81, 0.19] |
−0.60 (0.33) [−1.30, −0.01] |
−0.15 (0.25) [−0.66, 0,32] |
−0.09 (0.28) [−0.64, 0.47] |
| Family 4 (intercept) |
−0.20 (0.24) [−0.69, 0.27] |
−0.02 (0.29) [−0.61, 0.55] |
−0.13 (0.31) [−0.76, 0.46] |
−0.02 (0.24) [−0.51, 0.47] |
−0.01 (0.31) [−0.62, 0.61] |
−0.19 (0.25) [−0.74, 0.28] |
−0.21 (0.29) [−0.81, 0.34] |
| Family 5 (intercept) |
−0.27 (0.22) [−0.72, 0.14] |
−0.35 (0.29) [−0.95, 0.15] |
−0.27 (0.29) [−0.86, 0.26] |
−0.16 (0.23) [−0.66, 0.26] |
−0.38 (0.29) [−0.99, 0.15] |
−0.24 (0.23) [−0.74, 0.19] |
−0.18 (0.26) [−0.72, 0.32] |
| Family 6 (intercept) |
0.16 (0.27) [−0.36, 0.71] |
0.21 (0.36) [−0.42, 1.04] |
0.12 (0.37) [−0.59, 0.92] |
0.10 (0.28) [−0.42, 0.74] |
0.24 (0.38) [−0.43, 1.08] |
−0.01 (0.28) [−0.57, 0.55] |
0.07 (0.33) [−0.58, 0.74] |
| Family 7 (intercept) |
0.16 (0.24) [−0.29, 0.65] |
0.07 (0.28) [−0.47, 0.66] |
0.05 (0.31) [−0.55, 0.69] |
−0.10 (0.24) [−0.62, 0.34] |
0.10 (0.30) [−0.49, 0.72] |
0.07 (0.25) [−0.41, 0.60] |
0.26 (0.31) [−0.30, 0.91] |
| Family 8 (intercept) |
−0.26 (0.27) [−0.80, 0.25] |
−0.22 (0.32) [−0.93, 0.36] |
−0.56 (0.37) [−1.33, 0.09] |
−0.34 (0.31) [−1.02, 0.12] |
−0.24 (0.35) [−0.98, 0.40] |
−0.16 (0.27) [−0.74, 0.35] |
−0.24 (0.32) [−0.91, 0.36] |
| Family 9 (intercept) |
−0.22 (0.24) [−0.72, 0.23] |
0.19 (0.31) [−0.38, 0.87] |
0.30 (0.35) [−0.33, 1.05] |
0.23 (0.28) [−0.23, 0.88] |
0.22 (0.33) [−0.38, 0.92] |
−0.30 (0.27) [−0.88, 0.15] |
−0.61 (0.34) [−1.33, −0.004] |
| Family 10 (intercept) |
0.28 (0.27) [−0.20, 0.84] |
0.15 (0.31) [−0.43, 0.83] |
0.26 (0.35) [−0.38, 1.01] |
0.10 (0.26) [−0.39, 0.69] |
0.18 (0.33) [−0.43, 0.90] |
0.29 (0.29) [−0.23, 0.90] |
0.25 (0.32) [−0.34, 0.93] |
- Note: Estimates are mean differences between genotypes and the population mean. The same data are visualised in Figure 9. Parameters whose 95% credible intervals do not straddle 0 are in bold.
3 RESULTS
3.1 Population-level effects of environment on colour phenotype
Mantises developmentally altered their colour phenotype to match the colour of their background: More mantises developed an all-green phenotype when raised in a green environment, and more mantises developed an all-brown or mixed phenotype when raised in a brown environment (Table 1; Table S1; Figures 3-5). Interestingly, the all-brown phenotype was never produced on the green background (Figure 3). Humidity also affected colour phenotype: More mantises had an all-green phenotype and fewer had an all-brown phenotype when raised under high humidity compared to low humidity (Table 1; Table S1; Figures 3-5). Surprisingly, temperature alone had little effect on colour phenotype (Table 1; Table S1; Figures 3-5). But our analysis revealed an interaction effect between temperature and background (Table 1; Figure 5), which was driven by high temperature increasing the incidence of the all-green phenotype when the background was green, but not when it was brown (Table S3).
3.2 Population-level effects of environment on the coloration of individual body parts
All body parts were plastic in response to background colour: The head, thorax, abdomen, forelegs, mid-legs and hindlegs were all more likely to develop a matching green colour when mantises were reared on a green background (Table 1; Table S2; Figure 5). Higher humidity also triggered a higher incidence of green in all of the body parts except the mid-legs and hindlegs (Table 1; Figure 5). More complex patterns of plasticity were observed in these legs, as well as in the abdomen. The mid-legs and hindlegs were both affected by an interaction between background and temperature (Table 1; Figure 5). For the mid-legs, this effect was due to differences between all combinations of background and temperature, whereas for the hindlegs, the effect was driven by increased temperature elevating green body coloration on the green background but not the brown background (Table S3). Mid-leg colour and hindleg colour were additionally affected by an interaction between temperature and humidity (Table 1; Figure 5). This occurred because the incidence of green increased in the mid-legs as temperature increased and as humidity increased but showed no difference when temperature and humidity were in opposing high and low combinations (i.e. high temperature and low humidity vs. low temperature and high humidity); whereas increased temperature produced greener hindlegs at high humidity but not at low humidity (Table S3). Abdomen coloration was affected by an interaction between background colour and humidity (Table 1; Figure 5). Pairwise comparisons indicated that this effect was due to higher humidity increasing the incidence of green in the abdomen when the environment was brown but not when it was green (Table S3).
Background coloration induced the greatest effects on all body parts (Table 1; Table S2; Figure 5). However, the mid-legs and hindlegs exhibited the steepest reaction norms of all the body parts and had very similar slopes (Figure 6), indicating strong correlated sensitivity to background colour. This greater plasticity in the second and third pairs of legs was also reflected in the combination of body part colours that made up the mixed phenotype. Although there were, in principle, 62 possible combinations of green and brown body parts that could have made up the mixed phenotype, only seven combinations were observed (Table S4). By far the most common of these phenotypes was the one in which mantises' mid-legs and hindlegs were brown and the rest of the body was green (see Figure 2b), which occurred in 82% of individuals that exhibited the mixed phenotype (Table S4).
3.3 Genotype-level effects
All families produced offspring of each of the three colour phenotypes (Figure 7). Some environment combinations induced more varied phenotypic responses than others, but there was broad consistency. All offspring that experienced a combination of green background, high temperature and high humidity developed an all-green phenotype regardless of their family of origin (Figure 8). Most families continued producing exclusively all-green offspring on the green background even when the temperature and humidity were lower, although some families produced some offspring of the mixed phenotype, but never the all-brown phenotype (Figure 8). When the background was brown, most families produced a higher proportion of the mixed phenotype than either the all-brown or all-green phenotypes, but there was also a greater diversity of responses on this background (Figure 8). The greatest inter-family variation in colour phenotype was observed when both the temperature and humidity were high, but the background colour was brown (Figure 8).
If colour plasticity is shaped by genotype, we expected models that allowed intercepts and/or slopes to vary by genotype to fit the data substantially better than models that did not. While model comparisons suggested that models with varying intercepts and slopes (model 3) provided a better fit for mid-leg and hindleg coloration, and models with varying intercepts (model 2) provided a better fit for all other colour responses, these best fit models did not differ substantially from simpler models that incorporated only environmental effects and no genotype effects (model 1) (see Table 3). This suggests that genotype had little influence on plastic colour responses (as per Figure 1b). Indeed, families responded remarkably similarly to the environmental cues we manipulated. Posterior estimates of genotype-varying intercepts from our base split-brood models showed little deviation from the population average (Table 2; Figure 9). Only one genotype (family 1) showed a discernible difference in colour phenotype (Table 2), producing slightly more of the all-green phenotype than the average (see Figures 7 and 9). This difference was due to a greater incidence in members of this family of green in the mid-legs and hindlegs (Table 2; Figure 9). Families 3 and 9 also showed browner forelegs and hindlegs, respectively, than average (Table 2; Figure 9), but this did not affect the level of plasticity in the colour phenotype of these families (Table 2). Taken together, our results are indicative of environmentally induced plasticity in body coloration mediated only weakly by genotype.
| Model | Effect captured by model | Response variable | ||||||
|---|---|---|---|---|---|---|---|---|
| Colour phenotype | Head | Thorax | Abdomen | Forelegs | Mid-legs | Hindlegs | ||
| Model 0 | Null |
−114.83 [−137.66, −91.99] |
−33.32 [−46.22, −20.41] |
−28.68 [−45.34, −12.02] |
−34.30 [−49.74, −18.85] |
−31.88 [−45.62, −18.14] |
−31.88 [−45.62, −18.14] |
−111.74 [−134.80, −88.68] |
| Model 1 | Environment |
−6.25 [−14.62, 2.11] |
−3.33 [−9.24, 2.57] |
−1.65 [−11.47, 8.17] |
−6.32 [−16.46, 3.81] |
−0.58 [−7.05, 5.89] |
−1.57 [−7.41, 4.27] |
−2.94 [−10.55, 4.68] |
| Model 2 | Environment + Genotype | 0 | 0 |
−0.25 [−7.67, 7.17] |
−5.22 [−14.15, 3.71] |
0 | 0 | 0 |
| Model 3 | Environment × Genotype |
−0.81 [−14.88, 13.25] |
−6.43 [−13.24, 0.38] |
0 | 0 |
−1.27 [−8.41, 5.88] |
−5.71 [−11.88, 0.46] |
−2.49 [−11.19, 6.21] |
- Note: Reported are the pairwise differences (ΔLOOIC) and 95% confidence intervals (in brackets) between the best fit model (ΔLOOIC = 0) and each of the other models. Models that are substantially different from best fit models (i.e. 95% CIs do not overlap zero) are in bold. The ‘environment + genotype’ model (model 2) was the best-fitting model for all colour responses except the thorax and abdomen which were best explained by the ‘environment × genotype’ model (model 3). However, none of the best fit models differed substantially from simpler ‘environment’ models that allowed no genotype-level variation (model 1), suggesting that genotype has rather weak effects on colour plasticity in Miomantis caffra.
4 DISCUSSION
We found strong effects of environment on colour development in M. caffra. While humidity and to a lesser extent temperature influenced colour development to some degree, background colour showed the largest effects, with green background inducing more of the all-green phenotype, and brown background inducing more of the all-brown and mixed phenotypes. Mixed phenotype outcomes were largely mediated by the very high sensitivity of the mid-legs and hindlegs to the brown background. All phenotypic responses were, for the most part, consistent across families, suggesting little genetic variation for colour plasticity. The strong and stable effect of background suggests that M. caffra develops a colour to match its surroundings in at least some of its body regions. While green–brown morphotypes have been documented in mantises previously (Battiston & Fontana, 2010; Edmunds, 1972), our results represent, to the best of our knowledge, the first experimental evidence of plastic background matching in a mantis.
Background matching is a predicted response to predation (Duarte et al., 2017) and occurs at varying response speeds depending on the mechanism of colour change (Umbers et al., 2014). The evolution of developmentally plastic responses that take place over extended ontogenetic periods requires that environments be heterogeneous in quality but predictable through time (Taborsky, 2017). Environmental heterogeneity is expected for habitat generalists that occupy a range of habitat types, and environmental predictability may arise if such generalists go through a more-or-less sessile stage of their lifecycle. Some classic examples of developmental colour matching come from lepidopteran caterpillars and pupae (Eacock et al., 2017; Yumnam et al., 2021), and this may be because such species typically feed on a variety of different plant taxa but individuals remain on the same specific plant throughout development (Braby, 1995). A similar lifestyle may apply to mantises. As generalist hunters, many mantis species occupy a range of different habitat types (Cassar, 2020; Edmunds, 1972), and their sit-and-wait approach to hunting likely limits their motility (see Bartley, 1982; Christensen & Brown, 2018), especially during flightless stages of development. Limited movement may restrict home ranges and generate strong fidelity to natal habitats, generating the necessary predictability for developmental plasticity to evolve. Whether developmental plasticity underpins green–brown colour variation in other mantis species (e.g. Battiston & Fontana, 2010; Edmunds, 1972) is yet to be determined. But there is great potential for such effects.
Our results show that M. caffra pays attention to more than just colour cues in the development of its colour phenotype. High humidity induced more of the all-green phenotype, and temperature interacted with background coloration in complex ways, with high temperature and green background inducing a greater amount of green in the phenotype than any other combination of temperature and background. Many other examples of green–brown colour morphology in insects are induced by climatic variables. For example, low temperature induces melanised coloration in the alpine grasshopper (Valverde & Schielzeth, 2015), and high humidity triggers green phenotypes in the pupae of the tropical satyrine butterfly (Mayekar & Kodandaramaiah, 2017). Climatic variables are probably not selective agents per se, but token stimuli that predict the colour of the background (Nijhout, 1999) and as such, developing a green body is unlikely to be an adaptation to high temperature or humidity. That background colour functions as the primary cue for colour development in M. caffra, with temperature and humidity acting as secondary cues, is supported by two lines of evidence. First, background colour had a universally strong effect on the coloration of all body parts, whereas humidity and temperature had variable effects. Second, all families produced 100% all-green offspring when high temperature and humidity were coupled with a green background, and only a few families produced low numbers of the mixed phenotype when temperature and humidity were decreased. However, the greatest variation in family reaction norms occurred when high temperature and humidity were coupled with a brown background. This suggests that high temperature and humidity predict a green environment but result in conflicting signals when coupled with a background that is not green. This may be because higher temperatures and humidities lead to a greater proportion of green foliage in M. caffra's natural range. Sensitivity to climatic variables means that body coloration in this species might also shift predictably as the seasons change, a phenomenon known as seasonal colour plasticity (van Bergen & Beldade, 2019). Overall, our results suggest that temperature interacts multiplicatively with background to induce changes to the colour phenotype, perhaps to improve matching, and that humidity induces colour change independently of background. This pattern is consistent with other examples of colour matching from temperate zones (Mayekar & Kodandaramaiah, 2017), and highlights how multiple environmental cues can operate in tandem to shape colour responses.
One of the interesting findings from our study was that colour matching occurred more in certain body parts than in others, especially in response to the brown background which triggered colour matching mostly in the mid-legs and hindlegs rather than in the entire body. This could have occurred if producing and/or maintaining all-brown body parts is costly or synthesising or breaking down pigments is constrained by life-history trade-offs (Van Der Veen, 2005), chemical availability (Olson & Owens, 1998) or physiological processes (Nappi & Vass, 1993). In this case, partial colour matching may be the best compromise. Alternatively, the mixed phenotype may be just as cryptic against predation in brown environments as the all-brown phenotype, or even more so. Such an outcome could be facilitated by disruptive coloration, where coloration or patterning functions to disguise the body outline rather than to camouflage the whole body (Cuthill et al., 2005). For M. caffra, full colour matching may be superfluous if matching only part of the body is sufficient to maintain crypsis. The second and third pairs of legs may be most plastic in their colour because they best disguise the outline of the mantis body plan when viewed from above by visually guided predators such as birds. This could explain why the colour responses of these legs were so strongly correlated. It is also interesting to note that mixed phenotypes regularly exhibited a large brown spot at the junction of the abdomen and thorax (see Figure 2b), and this feature could serve the same disruptive function by making the abdomen appear as a leaf shape distinct from the rest of the body. The largely repeatable makeup of the mixed phenotype (brown mid-legs and hindlegs, green elsewhere) suggests that the three colour phenotypes we document here (all-brown, all-green and mixed coloration) are discrete alternative phenotypes (i.e. a polyphenism; sensu Nijhout, 2003). Whether the brown legs of the mixed phenotype provide adaptive benefits by disrupting the search image of visually guided predators in browner environments is currently unknown. It is also unknown how camouflaged each colour phenotype is in natural habitats. But such effects could be assessed with predator vision models (e.g. Kelber et al., 2003) or analyses that compare the survival of phenotypes in matched versus mismatched environments (e.g. Troscianko et al., 2016).
Our model comparisons suggested that the most parsimonious explanation for the colour responses we observed was environment-induced background matching with little to no influence of genotype. Genotype-specific reaction norms are expected to be pervasive in nature (Stearns, 1989), yet their absence is not uncommon (e.g. Gupta & Lewontin, 1982; Husby et al., 2010; Sirovy et al., 2021). Failure to detect a genotypic signature in plastic responses can occur if strong stabilising or directional selection depletes genetic variation in reaction norms (Saltz et al., 2018). This might be especially true for viability-related traits like colour matching that facilitate predator avoidance. Drift could also deplete standing genetic variation for plasticity through bottlenecks or founding effects (Saltz et al., 2018). Interestingly, the mantises we used in our study come from a recently established invasive population in New Zealand (Ramsay, 1984). Whether genetic variation, including variation for plasticity, is lower due to drift in this invasive population than in the species' native range of South Africa is unknown. If genetic variation in colour reaction norms is naturally low due to selection or drift (or both), then our sampling effort could have additionally contributed to the weak genotype effects we observed. Although our sampling effort was, in principle, sufficient to detect genotype-level effects on reaction norms (Martin et al., 2011), it is possible that the range of genotypic variation captured by our 10 families was too narrow. Future work could usefully investigate how colour responses vary over a broader range and number of genotypes and populations.
Mantises possess only one photoreceptor in their eyes which means they are essentially colour-blind and see the world only in shades of green (Van Der Kooi et al., 2021). It is therefore unlikely that M. caffra achieves colour matching by visually perceiving the colour of its surroundings. An alternative explanation is that the level of brightness or luminance of the colour environment, rather than the colour per se, is what triggers colour development in this species. Similar non colour-based mechanisms facilitate camouflage in other colour-blind taxa, most notably in cephalopods (Chiao et al., 2005). If duller (less bright) habitats tend to be browner, then cryptic colour matching could evolve in M. caffra through visual perception of brightness alone. Experiments that manipulate both colour and brightness (e.g. Eacock et al., 2017) could help to provide a clearer picture. Another possibility is that colour change is not mediated by signals received through the eyes, but through some other mechanism, such as photoreceptors in non-ocular cells. Such a phenomenon occurs in the peppered moth caterpillar, which possesses cells in the integument that detect substrate colours and trigger the development of a matching colour without the need for vision (Eacock et al., 2019). Whether similar photoreceptive capabilities exist in the non-ocular cells of mantises is unknown. Regardless, it is clear from our results that M. caffra uses more than just vision to perceive environmental cues. The induction of different phenotypes by non-visual cues (temperature and humidity) suggests that integument coloration in this species is also controlled, at least partially, by proximate mechanisms unrelated to visual processing. Hormones are the most likely candidates. A hormone called bursicon mediates integument tanning following moulting in many insects (Fraenkel & Hsiao, 1965), and juvenile hormone induces green coloration in locusts (Tanaka, 2000). Swallowtail butterfly larvae change colour in response to numerous environmental factors via the mediating effects of an endocrine factor called pupal cuticle-melanising hormone (Yamanaka et al., 1999). Similar hormones have been shown to induce pupal colour change in a number of other butterflies (Jones et al., 2007; Yamanaka et al., 2009). The molecular mechanisms that mediate colour change in mantises, however, remain unexplored.
Our results suggest an important role for the environment in determining colour variation at the population level in M. caffra, with little evidence of genotype-dependent plasticity. We found that mantis coloration was sensitive to temperature and humidity to some degree, but the largest and most consistent effects were induced by background colour, suggesting that the function of colour plasticity is habitat matching for crypsis, with climatic variables functioning as secondary cues. Our work here represents the first experimental assessment of background matching in a mantis, and the first evidence of such plasticity in this group. We suggest that many other mantises may show similar degrees of plasticity due to their unique life history as generalist sit-and-wait predators.
AUTHOR CONTRIBUTIONS
Nathan W. Burke and Gregory Holwell conceived the ideas and designed methodology; Nathan W. Burke collected the data; Nathan W. Burke analysed the data; Nathan W. Burke led the writing of the manuscript. Both authors contributed critically to the drafts and gave final approval for publication.
ACKNOWLEDGEMENTS
Open access publishing facilitated by The University of Auckland, as part of the Wiley - The University of Auckland agreement via the Council of Australian University Librarians.
CONFLICT OF INTEREST STATEMENT
The authors have no conflicts of interest in the production of this work.
Open Research
DATA AVAILABILITY STATEMENT
Data are available at Dryad Digital Repository: https://doi.org/10.5061/dryad.nzs7h44xk.
