Predation and the relative importance of larval colour polymorphisms and colour polyphenism in a damselfly
- First Online:
- Cite this article as:
- Johansson, F. & Nilsson-Örtman, V. Evol Ecol (2013) 27: 579. doi:10.1007/s10682-012-9617-8
- 311 Views
Intraspecific body colour variation is common in many animal species. Predation could be a key selective agent giving rise to variation in body colour, and such variation could be due to genetics (polymorphisms) or phenotypic plasticity (polyphenisms). In this study we examined the degree of colour polymorphism and polyphenism in background colour matching in larvae of the damselfly Coenagrionarmatum. In addition, we tested if predation risk is reduced when larvae are exposed to a matching compared to a non-matching background. By raising families of larvae at three different background colours we showed that polymorphism explained about 20 % of the total variation and polyphenism about 35 %. In a predation experiment with fish, we showed that larvae with a body colour matching the background had a higher survival success compared to larvae with a non-matching background colour. We suggest that the background matching is adaptive in terms of survival from predation and that colour diversity is maintained because of spatial and temporal variation in the background experienced by damselfly larvae under field conditions.
KeywordsColour polymorphismPredationPhenotypic plasticityPolyphenismCoenagrion armatum
Many organisms display striking intraspecific variation in body colour. Such variation can reflect either genetic polymorphisms or colour polyphenisms (Ramachandran et al. 1996; Hargeby et al. 2004; Ruxton et al. 2004; Gray and McKinnon 2006; Simpson et al. 2011). A colour polymorphism is usually defined as the occurrence of two or more distinct, genetically determined colour morphs in a population (Huxley 1955). A colour polyphenism, in contrast, is environmentally determined and the term is used when two or more distinct phenotypes are produced by the same genotype in different environments (Simpson et al. 2011). Hence, colour polymorphisms arise through changes in allele frequencies while colour polyphenisms are induced by the environment (phenotypic plasticity). Many classical examples of colour polymorphisms and polyphenisms refer to categorical phenotypes. Here we will use a somewhat broader definition where the colour morph variation among individuals in a population shows a more continuous variation. We note that for many study systems it has been pointed out that it is somewhat arbitrary to define if colour variation is discrete or continuous (Karlsson et al. 2008; Mullen and Hoekstra 2008; Canfield and Greene 2009).
Predation is a key selective agent giving rise to variation in body colour. By matching their body colour to the colour and/or pattern of their background, prey organisms can appear more cryptic to predators, thereby reducing the risks of predation (Endler 1978; Storfer et al. 1999; Ruxton et al. 2004; Svanbäck and Eklöv 2011). Whether such adaptive variation in body colour is the result of fixed, genetic factors (polymorphisms) or phenotypic plasticity (polyphenisms) is thought to depend on several physiological and environmental factors, such as the cost of maintaining phenotypic plasticity, the distribution and frequency of suitable habitats, the selective advantage of colour-matching phenotypes and migration rates between environments (Moran 1992; Sultan and Spencer 2002). For example, in habitats where the spatial distribution of background colours is homogeneous and/or costs of plasticity are high, we should expect polymorphism to evolve. On the other hand, if the frequency of background colours is heterogeneous and/or the costs of plasticity are low, we should expect body colour polyphenism to evolve (Moran 1992; Ruxton et al. 2004). Therefore, in order to better understand how cryptic colouration evolves, we need to determine (1) the frequency distribution of background colours across both temporal and spatial scales; (2) the selective advantage of background colour-matching in different habitats; and (3) how much of the observed colour diversity is due to polymorphisms and polyphenism. In this article we examine the two latter aspects and discuss these in relation to the former using a damselfly species that shows striking variation in larval body colour as a model organism.
In this study we raised full-sib families of the damselfly Coenagrion armatum (Charpentier 1840) against three differently coloured backgrounds, which allowed us to determine how much of the colour diversity observed among larvae is due to colour polymorphism and polyphenism, respectively. In addition, we tested the crypsis hypothesis by assessing whether plastically induced colour change lead to reduced predation risks when larvae were exposed to a matching compared to non-matching background.
Materials and methods
Genetic architecture of larval colour variation
The damselfly C. armatum is a boreal species restricted to the north-east of Europe and it occurs mainly in standing mesotrophic waters (Askew 2004). Of the 5–6 species of coenagrionid damselflies occurring in north-east Europe, the larvae of this species shows the highest diversity in body colour (F. J. personal observation and Fig. 1).
To examine the relative contribution of polymorphism (full-sib family) and polyphenism (bottom substrate colour) effects on the body colour of C. armatum larvae, we collected eggs from 7 mated females in June 2010 from lake Rovsundet (63°45′N, 20°32′E), in the vicinity of Umeå, northern Sweden. The study area is situated close to the Baltic sea and displays low elevation differences (0–200 m a.s.l. within a 50 km radius) and is mostly covered by nutrient-poor moraine soils, but more nutrient-rich fluvial deposits are found locally along rivers and streams. Eggs were acquired from the females by placing each female in a glass jar with moist filter paper attached to the side of the jar. The females oviposited in the filter paper within 48 h. Thereafter the filter papers with eggs were submerged in glass jars with non-chlorinated tap water and kept at a water temperature of 22 °C until hatching occured. The eggs hatched after 2–3 weeks. Immediately follow hatching, the larvae were transferred to individual rearing cups. The small plastic cups (diameter 7 cm, height 4 cm) were filled with 40 ml non-chlorinated tap water. To simulate different bottom substrate colours, the bottom and sides of the cups (which were transparent before painting) were painted on the outside using three colours: white, green and brown. The offspring from each female were then randomly assigned one of the three treatment groups (background colours). We used four replicate full-sib offspring from each female per treatment group, producing a total of 84 individuals: 7 females × 3 colours × 4 replicates.
The larvae were reared at a constant water temperature of 26 °C in a walk-in climate chamber at Umeå University, and the cups were randomly distributed within the climate chamber. A 4 cm piece of dry grass was added as a perch in each cup. Larvae were fed 6 days a week on laboratory-reared brine shrimp, Artemia sp., nauplii ad libitum (mean ± 1 SD = 282 ± 62 individual nauplii; n = 25 food doses). Artemia nauplii not consumed within 2–3 h from the feeding event inevitably died in the fresh water of the rearing cups. This minimized the variation in food availability among rearing cups. The photoperiod was maintained at 14:10 L:D, representative of typical late summer photoperiod conditions in north Sweden.
The experiment was terminated after 58 days and larvae were not fed the day prior to the termination. At day 58 all larvae had reached the 4th or 5th instar. C. armatum has about 10 instars when reared in the laboratory at a temperature of 26 °C (unpublished data). On day 58, all larvae were put individually into a petri dish and photographed with a Canon EOS 350D digital camera and a Tamron 90 mm macro lens. A millimeter grid paper was placed under the petri dish for size calibration. The photographing took place in a closed room with constant illumination consisting of one wall-mounted and four roof-mounted fluorescent tubes. The camera position was fixed and the same shutter speed and aperture was used throughout. The resulting photographs were used to perform the colour analysis and to measure larval size. We measured size as head width, the maximum distance between the distal parts of the eyes using ImageJ 1.45 s (National Institutes of Health, Bethesda, Maryland, USA). This size estimate has been shown to be the most relevant for size estimates in damselflies (Benke 1970).
The colour analysis was performed as follows. First, we colour-corrected all photographs based on reference photographs of a QPcard 201 colour chart (QPcard AB, Gothenburg, Sweden) taken under identical conditions as those described above. A colour correction profile was established from the photographs of the colour card and applied to all other photographs using the batch mode in the QPcoloursoft 501 software (QPcard AB, Gothenburg, Sweden). Using the colour-corrected photographs, the colour analysis was then performed in ImageJ 1.45 s. As a first step, each RGB photograph was converted to a HSB stack, where each pixel is represented by three values: hue, saturation and brightness (HSB, explained below). The HSB colour space is an approximation of how humans perceive colour. This is potentially problematic in studies of animal ecology, as animals may differ from humans in their perception of colour (Bennett et al. 1994; Stevens et al. 2007). However, here we study the effect of a fish predator and fish have a tetrachromatic colour vision similar to humans (Kelber and Osorio 2010). In addition, the HSB colour space has proved highly useful for understanding evolutionary changes involving visual perception in fish, as demonstrated by the numerous studies on sexual selection of body colour in several fishes (Deutsch 1997) including sticklebacks, the predator genera we investigate here (Candolin et al. 2007). Thus, we feel that the benefits of using the HSB colour space (ecological relevance and straightforward interpretability) can sufficiently justify its use in this case. We will bring up this issue in the light of our results in the discussion.
On each individual, we outlined three regions of interest (ROIs), defined as the maximally sized circles centred at the centre of the head, the median abdominal region and the mediodistal abdominal region, respectively (Fig. 1a; yellow circles). If undigested food items were visible through an individual’s integument (one food item is visible in Fig. 1a, just behind the rear ROI) at any of these positions, the nearest food-free position along the median axis was used instead for the ROI. Using the Measure stack tool in ImageJ 1.45 s, we then quantified the average hue (colour tone), saturation (colour intensity) and brightness (overall lightness) in the area covered by the three ROIs. We used the average value from these three regions in the analysis below. Note that in ImageJ, this method produces measurements at a common scale (0–255) for hue, saturation and brightness. These values are easily converted to the more commonly seen degrees and percent, but we favoured retaining them at a common scale for the statistical analysis.
In this experiment we examined whether environmentally induced plasticity in body colour (polyphenism) affected the susceptibility of larvae to predation from a fish predator. Eggs from five mated females were collected as in the colour diversity experiment. After hatching, larvae from all five females were mixed and larvae were divided into 6 equally sized groups. To induce phenotypic colour change in these larvae, three of these groups were reared in dark blue plastic tubs and the other three in white tubs (inducing dark and light larval body colour, respectively). The plastic tubs (height 13 cm, diameter 33 cm) were filled with non-chlorinated tap water to a level of 11 cm, and had a water temperature of 22 °C. Larvae were fed brine shrimp in excess for 6 days per week. When larvae had reached a size of about 6–8 mm the predation experiment started. The predation experiment was performed in five plastic aquaria (size: 30 × 18 × height 20 cm) filled with 8 l of non-chlorinated tap water. The bottoms of the aquaria were covered with a 1 cm deep sand layer simulating two different bottom substrates: dark and light. In the dark substrate aquaria we used black sand (gravel size 2 mm) and in the light substrate we used whitish sand (gravel size 4 mm). The damselfly larvae from the white tubs represented the light phenotypic group and the damselfly larvae from the dark blue tubs represented the dark phenotypic group. To the human eye, the two groups clearly differed in pigmentation, being notably dark and pale, respectively. In order to establish that the two groups did differ significantly in this respect we photographed them and performed a colour analysis as described above in order to statistically compare the colour of the two groups of larvae used in the predation trials.
At the start of the experiment, we introduced either 10 larvae from the Light phenotypic group or 10 larvae from the dark phenotypic group into each aquarium with dark and light bottom respectively. From the dark phenotypic group we discharged a handful of larvae that appeared dark greenish and used only larvae that appeared dark brownish to a human eye. The larvae were allowed to acclimatize to the novel environment during 10 min before a single fish predator was introduced into each aquarium. Nine-spined sticklebacks (Pungitius pungitius: size 3–4 cm) were used as predators and they were allowed to feed on the larvae for 10 min, after which the fish were removed from the aquaria. We then counted the number of surviving larvae in each aquarium. No dead larvae were found after the predation trial. Three replicate trials with the light bottom substrate and two with the dark bottom substrate were performed for each larval phenotype. An alternative approach to estimate predation would have been to add 10 light and 10 dark phenotype larvae into the same aquaria and then estimate mortality from predation. However, such an approach requires that each individual can be assigned without error to the correct treatment group after the trial, which we judged to be difficult in the present case.
To analyse the outcome of the colour variation experiment we first assessed whether larval body colour differed between full-sib families (polymorphism) and between larvae reared under different background colours (polyphenism). To do this, we performed three separate two-way ANOVAs using the colour measurements (hue, saturation or brightness) as the response variable and including bottom substrate, family and the interaction between substrate x family as predictor variables. The final models were reached by removing non-significant, higher-order terms. Colour was used as a Fixed factor and family as a Random factor in the analyses which were run in Statistica (StatSoft, Inc. 2011) using the Variance Estimation and Precision module. Secondly, we were interested in assessing the relative contribution of polymorphism and polyphenism effects on total phenotypic variation in body colour. To do this, we assumed that the total phenotypic variation, VTOTAL, can be partitioned into plastic, genetic and residual variance components, e.g. VTOTAL = VPLASTIC + VGENETIC + VRESIDUAL (Scheiner 1993). To do this, we used linear mixed-effects models implemented in the function lmer in the R package lme4 (Pinheiro and Bates 2000), considering both bottom substrate colour and full-sib family as random factors. All colour variables were first standardized by their means and variances, following Schielzeth (2010). It should be noted here that we thereby assume that the bottom substrate colours (brown, green, white) represents a random sample of the full range of natural background colours and the full-sib families represent a random sample of the genotypes present in the population. Although family groups were sampled at random, this is not true for background colours. However, the three background colours represent both extreme and typical values along a dark-to-light background gradient and are thus at least broadly representative of the natural range of environmental conditions. Finally, although we cannot exclude the possibility that some of the offspring of the same female represent paternal half-sibs, we will refer to these as full-sib families as damselfly males typically remove all of the sperm from previous matings (Miller and Miller 1981; Waage 1986).
To analyze the outcome of the predation experiment we used a generalized linear mixed effect model (GLMM) with a binary response variable in order to assess differences in survival between Light and Dark phenotypes when exposed to different backgrounds. In the full GLMM model, each individual’s survival was used as the binary response variable and background colour and colour phenotype was included, together with their interaction, as fixed factors and individual aquaria was included as a random factor. Analyses were also run using the function lmer. Significance levels were calculated using likelihood-ratio tests (LRT), providing the χ2 and P values presented in the text. It should be noted that LRTs are non-conservative for fixed effects (Baayen et al. 2008). For ease of interpretation, we also report results from Wald Z-tests in table form and the analysis were run in R using the function lmer from the lme4 package (Bates et al. 2008).
Results from separate ANOVAs analyzing the effect of bottom colour and family on the hue, saturation and brightness of C. armatum larvae
Results from a binomial generalized linear mixed model (GLMM) with binomial errors and a logit link function
Dark morph: dark background
The larvae in the dark and light phenotypic groups used for the predation experiment did not differ in size (F-test: F1,40 = 0.092, P = 0.764). However they did differ significantly in hue, saturation and brightness (F-test: F1,91 = 38.5, P < 0.001, F1,91 = 60.0, P < 0.001, F1,91 = 121.5, P < 0.001 respectively).
We have showed that both polymorphism and polyphenism contribute to body colour variation in C. armatum larvae and have quantified the relative contribution of each. We note that the observed colour morph variation represent a continuous variation rather than discrete morphs. When reared against different background colours, C. armatum larvae can induce a change in body colour to better match the background: a dark brown background induced darker larval body colour and the lighter backgrounds induced lighter body colour. The observed changes represents a shift from dark brown to lighter brown with a tint of green (Fig. 2), and the observed changes thus matches the variation in body colour that we have previously observed in field-collected larvae (Johansson, F. unpublished). Hence, we have demonstrated the presence of polyphenism (phenotypic plasticity) in body colour in C. armatum. Such a plastic background matching in body colour has been shown previously in dragonflies as well as other organisms (Moum and Baker 1990; Henriksson 1993; Hanlon et al. 1999; Wente and Phillips 2003; Clarke and Schluter 2011).
However, we also found support for polymorphisms of larval colour variation as we found significant differences between families in hue, saturation and brightness. These differences were independent of plasticity (no significant interaction effects; Table 1), but judging by the shape of the family reaction norms in Fig. 2b and c, this may reflects low statistical power and deserves to be investigated further. Similar genetic colour polymorphisms have previously been found in other organism (Majerus 1998; Hoffmann and Blouin 2000; Nachman et al. 2003; Wente and Phillips 2003; Hargeby et al. 2004), but this is the first time that the presence of genetic colour polymorphisms has been demonstrated in dragonfly larvae. Furthermore, very few earlier studies have simultaneously examined and found evidence for both polyphenism and polymorphism in body colour acting in the same species (but see Wente and Phillips 2003).
We found that polymorphism explained about 20 % of the total variation (across the three colour traits) and polyphenism explained somewhat more, about 35 %. Based on these results, we suggest that both genetic and plastic effects should be looked for—and quantified—simultaneously in many more systems. In fact, we expect both mechanisms to be present in many more organisms, as most organisms’ habitats can be described in terms of a matrix consisting of homogeneous and heterogeneous elements (Manríquez et al. 2009). The most rewarding challenge for researchers will thus not be to demonstrate if either of these mechanisms is operating, but to understand factors influencing the relative contributions of genetic versus plastic effects.
We also tested the hypothesis that cryptic colouration is adaptive in terms of escape from predation. The results showed that larvae matching their background colour experience drastically reduced mortality rates (Fig. 3), supporting the hypothesis that predation may be a dominant force behind the observed colour-matching mechanism. A similar reduction in predation with increased background matching has been shown for many species previously, including damselflies (Kettlewell 1955; Moum and Baker 1990; Manríquez et al. 2009; Svanbäck and Eklöv 2011). It is worth noting, however, that other selective forces may also be responsible for, or contribute to, the background matching. For example, damselflies are themselves predators of other invertebrates and a better match between background and body colour could lead to higher predation success. In support of this hypothesis, Moum and Baker (1990) found that background colour had an influence on the predation success of the damselfly larva Ischnura verticalis. However, the direction of this effect was not the expected one, as brown larvae had higher success rates against a green compared to a brown background. However, in black bears a background matching coat colour has been shown to result in a higher capture success on salmon prey (Klinka and Reimchen 2009).
Coenagrionid larvae appear to be able to continuously adjust their body colour throughout development. We demonstrated that plastic changes occur from hatching up to instar 4–5. Moum and Baker (1990) on the other hand used field-collected larvae of the damselfly I. verticalis that had already reached instar 4–6 and observed background-matching changes in body colour after two moults in the lab, i.e. at instar 6–8. As we did not observe any changes in body colour during the minutes up to half an hour of handling time (when photographing larvae or preparing the predation experiment), we suggest that the induced colour changes takes days rather than hours to be induced. In support of this time frame, Henriksson (1993) found that larvae of the dragonfly Leucorrhinia dubia could change their body colour to match the background in eight days. It would have been interesting to see if larvae had induced even larger differences in body colour among treatment groups if we had continued our experiment beyond instar 4–5. Interestingly, the study by Henriksson (1993) also demonstrated that body colour plasticity in odonates can be reversible, as the larvae in that study changed from a green to a brown body colour and vice versa. Further colour change beyond instar 4–5 could either enforce differences between genotypic groups or increase the magnitude of phenotypic plasticity, so it is not clear how prolonging the experiment would affect our estimates of the relative contribution of these forces.
Our method of estimating colour was based on techniques that are based on human vision and fish including nine-spined sticklebacks differs somewhat from humans in their colour vision (Smith et al. 2004; Rick et al. 2012). However, since the sticklebacks in our experiment selected prey according to our prediction, we argue that our results are relevant in a quantitative manner. The purpose of the predation experiment was not to understand the exact colour vision mechanism. It was simply to show that background matching seem to be important. In addition we note that: (1) Studies on the closely related three-spined stickleback (Gasterosteus aculeatus) suggest that stickleback can distinguish prey easily under four different spectral conditions: UV, short-wave, mid-wave and long-wave (Rick et al. 2012). (2) In a comparative study Smith et al. (2004) found that three-spined and nine-spined sticklebacks responded similar with respect to the red colour light spectra. (3) Bakker et al. (1997) found that three-spined sticklebacks selected orange coloured prey as expected from a human eyes perspective. Hence although humans and sticklebacks do not have the same colour visions, experimental studies suggest that they corroborate in a qualitative manner.
The fact that we revealed that both genetic polymorphisms and polyphenism act on larval body colour and have about equal effects, suggest that the selective pressure for cryptic colouration varies spatially and temporally in this system (Moran 1992; Sultan and Spencer 2002). Based on the ecology of the studied species, we hypothesize that spatial heterogeneity is likely to be the more important force in this case and suggest the following scenario. Soil chemistry and topology will cause certain areas to contain nearby lakes with similar background colour characteristics, whereby a polymorphic strategy will have a selective advantage locally. Here, the colour morphs that best match the average background colour of nearby lakes have a selective advantage and the frequency of the favoured colour morph will be dependent on how uniform the characteristics are of nearby lakes and how much migration there is among lakes. In such a scenario nearby lakes in one area have the same characteristics and nearby lakes in another area have other characteristics. Without too much migration between lakes among generations, a polymorphic strategy should be at a selective advantage, over a polyphenic strategy in such an area (Sultan and Spencer 2002). In contrast, in areas where adjacent lakes display drastically different background colour characteristics, selection for colour polyphenism should instead be strong, as the environment that larvae experience will become increasingly heterogeneous as the among-lake heterogeneity increase in an area (provided that adults do not discriminate between habitat types). We consider it very unlikely that damselflies actively select where to lay eggs based on the habitat’s colour characteristics. The reason why we believe this is that even dragonfly species that are extremely sensitive to fish predation are apparently not able to distinguish between fish and fishless waters (McGuffin et al. 2006), possibly because the costs of moving between sites are high. Presumably, the selective benefits of choosing breeding site based on habitat colour will be even weaker. Finally, there is also variation in background colour within a lake (e.g. Hargeby et al. 2004), and since eggs from ovipositing female might end up in a dark or light environments, such environmental variation within a lake could be viewed upon as heterogeneous and will hence select for colour polymorphism in body colour (Moran 1992; Sultan and Spencer 2002).
Our results point to two promising ways forward: firstly, more research should be directed towards quantifying the relative contribution of colour polymorphism and polyphenism in many more species, especially in those that differ in dispersal. Secondly, the genetic architecture of crypsis should be investigated at the landscape scale. By contrasting topographically homogenous areas with areas with a more complex landscape structure, our data suggest that we can gain a deeper understanding of the spatial and temporal scale at which crypsis evolves.
We thank Richard Svanbäck and Eleanor Jones for comments on earlier draft of this article. The research was funded by the Swedish Research Council to F.J.