Hot temperature during development alters iridescence in Morpho butterflies

Abstract

Environmental temperature during critical stages of development can define the phenotype of an individual, especially for ectotherms. Color is a particularly biologically important phenotypically plastic trait but has been primarily studied in the context of pigmentary, rather than structural, coloration. In Morpho butterflies (Nymphalidae), structural coloration created by nanostructures on the wing scales leads to their striking blue iridescent effect. We reared Morpho helenor larvae and exposed the pupae to three temperature regimes during development: control tropics, temperate cool, and hot house, with hotter temperatures simulating those predicted under a future climate change scenario. We hypothesized that the hot house treatment would be the most disruptive to iridescence. In support of this hypothesis, hot conditions reduced the brightness of adult butterfly wings and altered the color profile enough that they would look different from control butterflies to bird predators and conspecifics. For certain bird perceivers, hot conditions would obliterate the iridescent effect entirely. The mechanism underlying these shifts appears to be, at least in part, driven by a decrease in the gap between nanostructures on the wing scale and consequent narrowing of the entire scale. Our study highlights the ways in which structural color is alternately vulnerable or robust to environmental temperature during development. From this work, we can glean insight into the environments that would be most supportive of iridescence in tropical butterflies and make predictions about the effects that changing climate regimes will have on the diversity of some of the earth's most dazzling organisms.

Significance Statement.

During development, organisms are variably susceptible to their environment. Understanding the degree of this vulnerability can provide a window into the organism's evolutionary past and predicted future in a changing world. Here, we test the influence of environmental temperature during pupal development on the iridescence of Morpho butterflies—a tropical, highly iridescent genus. We find that butterflies developing in cool temperatures are visually unchanged from butterflies developing under tropical temperatures, while butterflies developing under hotter conditions exhibit hue- and brightness-shifted iridescence. These differences are driven in part by changes to wing scale nanostructures. Iridescence therefore does not appear to be limited by a temperate environment but is negatively affected by the hotter conditions predicted under climate change.

Introduction

Phenotypic plasticity—the adjustment of an organism's physical, biochemical, or behavioral traits by environmental factors—can be a deciding factor in how the organism fares under variable conditions (1–3). Some of the clearest examples of adaptive phenotypic plasticity are found in temperature-mediated color and pattern traits of ectotherms (4). For instance, colder temperatures during development induce darker morphs of beetles (5, 6) and butterflies (7), with empirical tests documenting higher survival and fitness of darker adults exposed to cold conditions (5, 8). Plasticity engendered by temperature during development is not always a benefit, however, as it can be costly to invest in the genetic machinery to monitor the environment, or lead to a lower-quality phenotype (3, 9). Pierid butterflies, for example, increase melanization when exposed to cold temperatures during pupal development—an adaptation that benefits thermoregulation but has also been linked to lower fecundity in adults (10, 11).

Environmentally mediated color patterns can also have important implications for biotic interactions. Studies with the squinting bush brown butterfly (Nymphalidae: Bicyclus anynana) have described different eyespot patterning among seasonal morphs, driven by the predators that are most active and abundant during these seasonal periods (12). Genetic assays reveal that temperature cues induce the development of these distinct seasonal morphs (13). Thus, environmentally induced plastic coloration can have life-or-death consequences in a dynamic world.

Much of the work on thermally induced developmental color changes in ectotherms has come from butterflies (Lepidoptera: Papilionoidea), and primarily these studies have focused on pigmentary coloration. An alternative coloration mechanism in butterflies, structural coloration, is widely used across the group to produce bright, short-wavelength colors (blues, violets, and ultraviolets [UVs]) or optical effects, including the striking phenomenon of iridescence: a change in hue over viewing angle. The nanostructural mechanisms for creating iridescence are diverse across families (14). Two previous studies analyzed changes in structural coloration as a result of thermal stress during development in butterflies from two different families (Lycaenidae and Pieridae). Both focused on temperate species and found that these structures are susceptible to change. Under experimentally cooled conditions (5 °C), traditionally drab females of the common blue (Lycaenidae: Polyommatus icarus) were induced to produce the saturated blue scales of their male counterparts (15), using similar nanostructure architecture to males (16). In Colias butterflies (Pieridae: Colias eurytheme) both heat and cold shock altered UV-producing nanostructures on the wing scales to a similar degree as larval nutritional stress. Under both forms of stress, peak UV brightness and the angular breadth across which the UV signal was observable declined significantly (17, 18).

Here, we focus on how structural coloration responds to environmental perturbation in the tropical butterfly, the common blue morpho (Nymphalidae: Morpho helenor). This species is a widespread and charismatic member of neotropical forest communities, known for its striking blue iridescent wing patterning and erratic flight. Iridescence is fundamental to the biology of M. helenor and other species in the genus, likely serving as a long-range visual signal to both mates and predators (19). This dynamic color trait appears to be at least a species identification signal, if not a quality signal for female assessment in nymphalids (20). Iridescence in Morpho butterflies also appears to function as an antiaerial predator strategy (21). That is, the bright flashing blue might serve as an evasive aposematic signal, warning predators that the butterfly will be too challenging to catch (or a Batesian mimic of this warning) (19, 22, 23). Alternatively, or in combination with this advertisement, the flashing blue interspersed with the brown ventral wing surface might create a visual illusion, causing the predator to mis-aim its strike (24–26).

In Morpho, similar to the vast number of species in the Nymphalidae, iridescence is produced by a latticework of longitudinal ridges made up of stacks of lamellae that line the upper portion of the wing scales (27). These are interconnected by short crossribs that lie perpendicular to the ridges (28). Together, the multilayer ridges and crossribs produce irregular reflective surfaces, creating light interference effects that reflect predominantly “blue” (∼450–480 nm) light (29, 30). These structures are cemented via trabeculae to the lower layer of the scale (lower lamina), which additionally acts as a thin film reflector—building bright, directional reflectance via constructive interference where reflected wavelengths sum as they travel in phase with each other (14, 30). In addition, many Morpho species have further modifications between two shingled strata of scales that make up their iridescent regions. The lower stratum, the ground scales, tend to have a tighter latticework of ridges and crossribs, with the ridges being more constrained in height (number of lamellae in the stack). They also contain pigments among these nanostructures that serve to absorb extraneous light, further contributing to the blue reflectance (28, 30). Cover scales lay atop these ground scales and are commonly translucent, lacking pigments and containing far fewer crossribs. Instead, their primary structures are wider spaced and heightened ridges containing multiple layers of lamellae (31, 32). The function of the cover scale appears to be to increase the angular spread over which light is reflected, allowing the selectively tuned bright blue coloration to be seen from a wider range of angles (33, 34).

Although the physical mechanisms underlying Nymphalidae butterfly iridescence have been well described, its vulnerability to environmental conditions remains unknown. This is crucial for understanding the kinds of environments in which these elaborate nanostructures can be produced and for predicting how this color trait will fare under changing climate regimes. We specifically focus on the role that the environment plays on the development of iridescent scales in two subspecies of M. helenor: M. h. peleides and M. h. theodorus. Morpho h. peleides (hereafter Panamanian morpho) occurs in lowland Panamá (elevation ∼60 m) and is almost entirely covered in blue iridescence on its dorsal surface. Morpho h. theodorus (hereafter Amazonian morpho) can be found in the Ecuadorian Amazon (elevation ∼450 m) and has thick blue bands of iridescence on its dorsal surface, surrounded by black on either side (Figs. 1–5). These two subspecies are recently diverged (∼2 mya (35)) and live in slightly different environments, with the Amazonian morpho experiencing more variable temperatures across the year (annual mean temperature Panamanian: 26.5 °C, max warmest month: 31 °C, min coldest month: 22 °C; annual mean temperature Amazonian: 24 °C, max warmest month: 30 °C, min coldest month: 19 °C). We therefore elected to capitalize on the differing phenotypes and environmental background of these closely related subspecies to investigate potentially differing degrees of iridescence plasticity. We exposed developing pupae to either a consistent, tropical temperature (control; 28 °C day and night), a temperature swing more common in temperate regions (temperate cool; 28 °C day and 14 °C night), or a temperature swing akin to the warmer extremes the tropics may experience under the IPCC's SSP3-7.0 current-policies model (low mitigation efforts, high greenhouse gas emissions) (36) (hot house; 35 °C day and 28 °C night). This model predicts a possible upper range of ∼4 °C increase in mean global surface temperature by 2100 (36), while simulation studies focused on South America (37) and the Amazon rainforest region (38) predict a possible ∼5°C mean temperature increase in certain localized areas. We hypothesized that the hot house treatment would be the most disruptive to the iridescent structures, as this treatment represents an upper temperature bound of what a nymphalid butterfly might be exposed to during natural development, while the temperate cool treatment would not have a dramatic effect, aside from slowing down development (39). Understanding how environment changes these nanostructures and the resulting iridescence profile will reveal important information about the plasticity embedded in this trait and can help us predict how a meaningful proportion of butterfly diversity will fare under a rapidly changing climate.

Figure 1.

Thermally challenged butterflies maintained their iridescent effect when viewed through the eyes of a congener and a violet-sensing bird predator, but lost their iridescence in the eyes of ultraviolet-sensing birds. Pupae of two different M. helenor subspecies were exposed to three different temperature treatments: Control tropics (simulating the butterflies' native temperature), temperate cool (reflecting common summer temperatures in temperate regions), and hot house (reflecting an IPCC business-as-usual modeled future of ∼3.5 °C increase). Spectral measurements of adult forewings reveal that all treatments remained iridescent to A) an average violet-sensing bird or B) another Morpho, but that iridescence was obliterated in butterflies exposed to abnormally hot temperatures, viewed through the eyes of C) an average UV-sensing bird. The dotted line represents a JND of 1. Coefficient estimates of the bootstrapped color distance between spectra (dot) and 95% CIs (whiskers) above a JND of 1 indicate that the perceiver would detect a visual difference in wing color between two different viewing angles (perpendicular and oblique)—i.e. the butterfly would appear iridescent. n individual M. h. peleides (Panamanian morpho; coded in purple, on left of each set) control = 20, temperate cool = 19, hot house = 16; n individual M. h. theodorus (Amazonian morpho; coded in brown, on right) control = 24, temperate cool = 23, hot house = 28.

Figure 5.

Inter-ridge distance of butterfly scales that developed under hot house conditions was smaller than those that developed under control conditions. While the same pattern was true for the temperate cool treatment, the certainty around this difference is lower. These measurements were taken from A) cover: C, and ground: G scales that we imaged via B) an eSEM at multiple magnifications. A magnification of 500× is shown here to illustrate the whole scale, while measurements were taken from images at 6,000× magnification. C) Our plots illustrate outputs from our Bayesian models for both types of scales from both subspecies' forewings, with Panamanian M. helenor scales on the top row and Amazonian M. helenor scales on the bottom row. Dots are raw data, black circles are central tendencies with thin whiskers showing the 95% credible interval and thicker whiskers showing the 80% credible interval, and mountains are the posterior distributions. n individual Panamanian morpho control = 6, temperate cool = 6, hot house = 5; n individual Amazonian morpho control = 8, temperate cool = 6, hot house = 8.

Results

Pupal development results

Hotter temperatures lead to higher pupal death, while cooler temperatures extend pupal development time

Thermal challenge treatments (temperate cool and hot house) that deviated from typical conditions to which M. helenor pupae are adapted (control treatment) were physiologically stressful. For the Panamanian subspecies, a simple linear regression revealed that both temperate and hot house treatments led to statistically longer pupal development times, with the temperate chamber leading to the longest time in the pupal stage (control: 12.3 days, temperate cool: 18.5 days, hot house: 13.7 days; P < 0.001). There was also a larger discrepancy between the number of pupae that eclosed vs. the number of pupae that died at the pupal stage between the control and hot house chambers and the temperate and hot house chambers (control: pupae that died = 41 [41%], pupae that eclosed = 60 [59%]; temperate: pupae that died = 60 [52%], pupae that eclosed = 56 [48%]; hot house: pupae that died = 130 [69%], pupae that eclosed = 58 [31%]; P < 0.001). For the Amazonian subspecies, only the temperate cool chamber led to longer pupal development (control: 11.8 days, temperate cool: 16.7 days, hot house: 12.3 days; P < 0.001). We also found that the only statistically significant discrepancy in pupae that eclosed vs. pupae that died was between the control and hot house treatments (control: pupae that died = 6 [21%], pupae that eclosed = 23 [79%]; temperate: pupae that died = 12 [38%], pupae that eclosed = 20 [63%]; hot house: pupae that died = 24 [48%], pupae that eclosed = 26 [52%]; P = 0.05). We note that these smaller absolute numbers are due to a change in notation protocol mid-way through the data collection period for the Amazonian subspecies, which led to a lack of data on the number of pupae that died vs. emerged poorly from the first cohort. To account for the higher mortality in the hot house chamber across both subspecies, we increased the in-going number of pupae to this treatment to ensure a roughly equal number of adult butterflies for each treatment.

Spectral measurement results

Hotter temperatures obliterate iridescence in the eyes of a UV-sensitive bird but not a violet-sensing bird or a congener

We first verified that these Morpho subspecies would be perceived as distinct by relevant perceivers and that they met the definition of iridescence (change in hue over viewing angle; see Figs. 1 and S1, Supplementary Results). For all spectral analyses, we used perceiver-based visual models to compare reflectance spectra from butterflies at two selected angles: perpendicular, where the light source and spectrometer sensor were parallel to each other and perpendicular to the butterfly wing, and oblique, where the light source and spectrometer sensor were at a ∼45° angle from each other and ∼68° from the wing. These perceiver-based visual differences are driven by a combination of the hue and saturation of the reflectance from the wing, as well as chromatic receptor sensitivities of the perceiver (40, 41). We found that control butterflies of both subspecies presented visibly different spectra (just noticeable difference (JND) > 1) between perpendicular and oblique angles for three viewers of interest: average violet-sensitive bird (avg-v), average ultraviolet-sensitive bird (avg-UV), and congener (Morpho; Fig. 1). Thus, M. helenor butterflies fit the definition of iridescence. Butterflies from our thermal challenge treatments (both temperate cool [n Panamanian = 19; n Amazonian = 23] and hot house [n Panamanian = 16; n Amazonian = 28]) remained iridescent (JND across angles > 1; Fig. 1). The exception to this was hot house butterflies through the eyes of an avg-UV bird, for whom the spectra did not significantly change across viewing angle (Fig. 1C). Hot temperatures therefore rendered Morpho noniridescent to UV-sensing birds.

Hotter temperatures differentially alter the perceived color of Morpho by subspecies and wing type

While our treatments did not obliterate the iridescence effect for all viewers, we did find that they consistently altered the perceived reflectance of both subspecies of M. helenor (Fig. 2). The most pronounced changes for the Panamanian M. helenor subspecies were apparent when the butterflies were viewed at an oblique angle. In the eyes of all modeled viewers, the forewings of Panamanian morpho individuals that developed under hot house conditions looked significantly different from those of control butterflies (JND ∼ 2–4; Figs. 2 and S2). We found a similar result for hindwings viewed from an oblique angle, although this difference would not be discernible to the avg-UV bird (Fig. S2). Reflectance differences between control and hot house butterflies were minimal or nondetectable at the perpendicular angle, except by a violet-sensing bird predator (Fig. 2). In contrast to this, the Amazonian morpho exhibited the most consistently distinguishable forewing color profile between hot house and control treatments at the perpendicular angle (Fig. 2B and Table S1). Hindwings at this angle could only be discriminated by the violet-sensing bird, and we found limited evidence that perceivers could detect a visual difference in reflectance from the treatments at the oblique angle (Fig. 2 and see Fig. S2 for additional comparison graphs). Across subspecies, wings, and angles, we found that none of the perceivers would be able to distinguish the color of butterflies that developed under temperate conditions from those that developed under control conditions.

Figure 2.

Butterflies exposed to the hot house treatment look significantly different from butterflies that developed under control tropics conditions to both A) a violet-sensing avian predator and B) a potential mate or competitor from at least one angle. The UV-sensing bird model could not distinguish between hot house and control butterflies at almost any angle and therefore is not shown here (but see Fig. S2). The Panamanian subspecies of M. helenor is depicted in purple (left) and the Amazonian subspecies is in brown (right). We denote the measurement angle beneath each coefficient estimate, with two arrows (yellow [left] = light source, blue [right] = sensor) in parallel indicating the perpendicular measurement and the angled arrows indicating the oblique measurement, for both sets of wings.

Abnormal temperature regimes primarily affect absolute brightness, with some subspecies-specific effects on peak hue

To determine how hue and brightness were quantitatively changing among the treatments, we used a Bayesian modeling approach and compared the posterior distributions of our temperate cool and hot house treatments to our control tropics treatment at each angle for both the UV (260–379 nm) and visual (380–700 nm) peaks in the spectrum. We found that for both subspecies at both angles, the greatest difference was a drop in peak brightness for the hot house treatment (probability that hot house butterfly peak brightness < control butterfly peak brightness = ∼1.0; Table S1) and an overall suppression of the reflectance intensity (brightness; Figs. 3 and S3). For the most part, brightness of butterflies emerging from the temperate cool chamber was not affected (probability that temperate cool butterfly brightness < control butterfly brightness = ∼0.5). The one exception was Amazonian morpho forewings at a perpendicular angle (probability that temperate cool butterfly peak brightness < control butterfly peak brightness = 0.94 for the visible portion of the spectrum, 0.96 for the UV portion). Among the Amazonian M. helenor subspecies, we also found that peak hue shifted toward shorter wavelengths under the hot house condition, primarily for the forewings at both angles (probability that hot house butterfly hue < control butterfly hue = ∼0.95 for visible portion of the spectrum, ∼1.0 for the UV portion). Hindwing peak hue change was slightly more variable, as it was only detectable from the perpendicular angle (probability that hot house butterfly hue < control butterfly hue = 0.87 for visible portion of the spectrum, 0.99 for the UV portion; Table S1). The average negative shift in hue for forewings, across both visible and UV portions of the spectrum, was ∼8 nm, which exceeded the natural variation in control butterflies (Cohen's d for visible portion of the spectrum = 0.36; Cohen's d for ultraviolet portion of the spectrum = 0.72). While these changes in peak hue are noteworthy, they do not fully explain perceived differences across treatments, as they cannot capture spectrum shape. We also note that a perceiver-based analysis of brightness comparisons generally did not reveal significant differences between treatments, outside of the oblique angle for the Amazonian morpho (Fig. S4). This is surprising given the strong quantitative decline in brightness, but could indicate a biological reality: birds are known to have relatively poor achromatic sensitivity (42) and to capture this we used literature-informed weber fractions (signal:noise) in our luminance comparison tests rather than the customary defaults in the analysis package (pavo2 (40)), following the methods of Dang et al. (43). We also used relatively insensitive achromatic weber fraction values for our butterfly perceiver model, again following Dang et al. (43). True achromatic weber fraction values have not been determined for nymphalid butterflies, however, and thus this approximated value could lead to a misestimation of luminance contrasts.

Figure 3.

Unusually hot conditions during development lead to a drop in brightness across the visual spectrum. Averaged reflectance measurements from forewings of A) the Panamanian morpho and B) the Amazonian morpho subspecies reveal generally similar reflectance values from control and temperate treatments and depressed reflectance brightness from hot house treatments. Blue lines represent the perpendicular measurement angle, where the light source and spectrometer sensor were parallel to each other and perpendicular to the wing. Red lines represent the oblique measurement angle, where the light source and sensor were ∼45° from each other and ∼68° from the wing. Shaded regions around the lines show the variance as one SD around the mean. The yellow dot on each butterfly wing depicts the spot from which we measured reflectance.

Microscope photograph results

Cover scale width shrinks in hot conditions and expands in temperate conditions

We hypothesized that these changes in reflectance intensity and spectra were induced by modifications to wing scale cell structure and nanostructure. Indeed, subspecies of M. helenor experienced a narrowing of the forewing cover scales under hot house developmental conditions, compared with the control (tropics) conditions (probability of hot house Panamanian morpho scale width < control scale width = 0.93; probability for Amazonian morpho = 0.99). In contrast, both demonstrated a widening of the forewing cover scale width under temperate cool conditions, compared with control conditions (probability of temperate Panamanian morpho scale width > control scale width = 0.94; probability for Amazonian morpho = 0.93; Fig. 4). Hindwing cover scales showed similar trends overall, although the Panamanian subspecies was less responsive to environmental perturbation than Amazonian (probability of hot house Panamanian morpho scale width < control scale width = 0.52, probability for Amazonian morpho = 0.95; probability of temperate cool Panamanian morpho > control scale width = 0.77; probability for Amazonian morpho = 0.94). Ground scales exhibited less consistent responses to environmental conditions, with ground scales of both forewings and hindwings showing functionally no difference from control across both thermal challenge conditions, except for hot house Amazonian morpho (probability of hot house Amazonian forewing ground scale width < control ground scale width = 0.98; probability of hot house Amazonian hindwing ground scale width < control ground scale width = 0.82; Fig. 4). We found no sex-specific effects of our treatments on scale width. Male cover scales were always larger than female cover scales, however, while ground scales showed less consistent sexual dimorphism (see Supplementary material for more details).

Figure 4.

Wing-scale width measurements taken from cover (left, in pink) and ground (right, in green) scales demonstrates a narrowing of the cover scales when M. helenor pupae of both subspecies develop under hot house conditions and a widening when they develop under temperate cool conditions, compared with control scales. Ground scales of the Amazonian M. helenor that developed under hot house conditions were also smaller than control ground scales, but we found no other differences among the ground scales. Data are broken up by sex to show the sexual dimorphism in cover scales, with males consistently having wider cover scales than females. Points are raw data; dots are central tendencies with whiskers representing 95% credible intervals and mountains show the posterior distributions from our Bayesian models. The circle on the butterfly wing shows the wing region from which we imaged the scales, and the microscope picture shows the wing patch with a chunk of cover scales (circled in pink, on left) removed to reveal the ground scales (circled in green, on right) below. n individual Panamanian morpho control = 16, temperate cool = 20, hot house = 11; n individual Amazonian morpho control = 21, temperate cool = 16, hot house = 20.

Scanning electron microscopy results

Hot temperatures narrow distance between ridge nanostructures in both cover and ground scales

To understand the mechanism underlying the reflectance change between our treatments, we analyzed images taken with an environmental scanning electron microscope (eSEM). Due to the resolution of our images, we focused our comparisons on the distance between the lamellar ridges and the crossribs, the widths of ridges and crossribs, and the number of ridges total on a scale. We found that for both subspecies of M. helenor, the distance between the ridges on cover and ground scales was considerably smaller for individuals who had undergone pupal development in hot house conditions vs. those that had developed under control conditions (probability that ridge distance for hot house < control = 0.95 for Panamanian cover scale, 0.99 for Amazonian cover scale; 0.99 for Panamanian ground scale, 0.94 for Amazonian ground scale). Differences between ridge distance in M. helenor that had developed under temperate vs. control conditions were slightly less pronounced (probability that ridge distance of temperate < temperate control = 0.84 for Panamanian cover scale, 0.91 for Amazonian cover scale; 0.96 for Panamanian ground scale, 0.88 for Amazonian cover scale; Fig. 5). However, an analysis of the number of ridges on covers and ground scales revealed no differences. Thus, a shrinking of ridge distance under our experimental temperature conditions cannot be accounted for by a higher number of more tightly packed ridges overall.

Abnormal temperatures increase distance between crossrib nanostructures

Crossrib distance exhibited even smaller differences between the treatments for cover scales (probability that crossrib distance for cover scales < control = 0.88 for hot house Panamanian, 0.64 for hot house Amazonian; 0.67 for temperate cool Panamanian, 0.69 for temperate cool Amazonian). Ground scale crossribs, however, appear to be more susceptible to environment during development, in an opposite direction from the ridges. That is, the distance between ground scale crossribs increased under thermal stress. We found differences between the subspecies, where crossrib distance increased only under the hot house condition for the Panamanian morpho and only under the temperate cool condition for the Amazonian morpho (probability that crossrib distance for treatment ground scales < control = 0.47 for hot house Panamanian, 0.004 for hot house Amazonian; 0.01 for temperate cool Panamanian, 0.56 for temperate cool Amazonian). We note, however, that the temperate cool Panamanian ground scale crossrib distance increase was far smaller than the hot house Amazonian ground scale distance increase (Panamanian control: 0.70 μm, temperate cool: 0.76 μm, hot house: 0.70 μm; Amazonian control: 0.74 μm, temperate cool: 0.73 μm, hot house: 0.94 μm). We predict that the crossrib distance increases we found are related to a reduction in the overall number of crossribs on the ground scale under environmental stress, although we were unable to count crossribs across entire ground scales to verify. In addition to crossrib spacing, previous work has implicated crossrib width in scale color for a noniridescent species (44). We thus measured the width of both crossribs and ridges and found no meaningful differences across the treatments (probability that treatment crossrib or ridge width < control crossrib or ridge width = ∼0.50).

Discussion

The phenotypic plasticity embedded within species traits holds critical information about environmental conditions within which that lineage has evolved and how it may fare under different future conditions. In this study, we aimed to understand how environmental temperature during development influences iridescent scale formation. We exposed two phenotypically distinct subspecies of M. helenor pupae to tropical (control) conditions (28 °C day and night), temperate cool conditions (28 °C day, 14 °C night), or hot house conditions (35 °C day, 28 °C night), simulating the upper range of temperatures predicted by the current-policies/low mitigation effort IPCC3-7.0 model (36). This pupal stage represents a critical window when the iridescent nanostructures are forming (45). We found that although the reflectance curves of butterflies that developed under temperate cool and control tropic conditions were quite similar to each other, developing under hot house conditions significantly changed the perceived color of both forewings and hindwings in the eyes of both violet-sensing avian predators and other Morpho butterflies, from at least one viewing angle (whether the viewer was looking directly down on the butterfly, parallel to the light source, or whether the viewer was observing the butterfly at an angle to the light source; Fig. 2). An average ultraviolet-sensing tetrachromat bird would in general not be able to detect a visual difference between the treatments for a given angle, however (Fig. S2), nor would it identify butterflies from the hot house chamber as iridescent at all (Fig. 1C). The discrepancy between the violet-sensing and UV-sensing birds is likely due to their differing peak sensitivities. In both cases, the visual system is underlain by four single-cone visual pigments. The fourth cone of violet-sensing birds is shifted spectrally upwards, however, creating greater sensitivity in the violet/blue portion of the color spectrum (where M. helenor reflectance tends to peak) and allowing for greater discernability of small hue or intensity changes (46). Importantly, violet-sensing birds are the most likely predators of Morpho butterflies. For instance, jacamars—known Morpho hunters (23)—in addition to many other aerial insectivores, belong to orders that are not sensitive in the ultraviolet (47).

While both subspecies of M. helenor displayed similar color (hue) and brightness shifts for relevant perceivers in response to hot temperatures (Figs. 2 and 3, Table S1), we did find subspecies-specific patterns. Namely, M. h. theodorus (the Amazonian common blue morpho) demonstrated a measurable peak hue shift toward shorter wavelengths in both the visible and ultraviolet portions of the spectrum, while this hue shift was not detectable in M. h. peleides (the Panamanian common blue morpho; Table S1). This difference in effect between the subspecies may be due to the slightly cooler environment from which the Amazonian morpho originates, leading to increased susceptibility to elevated temperatures. We also found that while the overall intensity of the reflectance spectra was suppressed in hot house Morpho of both subspecies (Fig. 3), this reduced brightness would only be observable to bird and Morpho perceivers in the Amazonian subspecies (Fig. S5). Birds are known to be somewhat limited in sensitivity to achromatic differences (differences in intensity, rather than hue) (42) and while specific achromatic sensitivity values are not known for Morpho butterflies, we used a hypothesized value of relatively poor luminance discrimination as well, based on the literature (43) (see Materials and methods and Results). It therefore may be that the general lack of perceived brightness difference among our viewers is due to a ceiling effect, where all Morpho were so bright that even a quantitatively significant reduction in brightness would not be obviously different. Similarly, the peak hue shift analysis, while illustrative of certain temperature-induced changes in our experiment, does not account for holistic shape changes of the reflectance curve and thus cannot fully explain the differences that we detected in our chromatic comparisons (Fig. 2).

Given that M. helenor live in tropical communities containing multiple species of Morpho across their range, often bearing similar iridescence patterns (48), it seems plausible that species are tuned to the specific color profile of their conspecifics to avoid undesirable hybridization. Interestingly, studies investigating wild Morpho interactions with flying dummies of differing phenotypes revealed that males will often indiscriminately approach conspecific and sympatric Morpho butterfly species—apparently showing little regard for species-specific color or intensity differences (49, 50). It may be, however, that a true copulatory attempt by either sex would require higher fidelity color matching to a species norm. Focused mating studies are needed to determine whether a temperature-induced shift away from a species' spectral optimum could therefore have fitness consequences.

We note that while we also found a hue shift in the UV range (Table S1), these peaks were near 300 nm, likely below the UV sensitivity of an avian predator (51) or a conspecific eye (52). Previous behavioral work with a different nymphalid species, Hypolimnas bolina, has demonstrated that UV can be an important ornament in mate selection (18). In this species, the UV peak is ∼400 and therefore well within the expected sensitivity of nymphalid butterfly vision (52). Given how far into the UV we find the M. helenor peak, it may simply be an artifact of the periodic structures on the scales that make up the iridescence. Future behavioral work will reveal whether the ultraviolet component of Morpho iridescence serves a signaling function.

These chromatic and brightness differences appear to be driven, at least in part, by a shortening of the distance between the longitudinal lamellar ridges on the cover and ground scales (Fig. 5). Generally, a reduction in ridge spacing is associated with brighter iridescence reflectance (17, 45) and a shift to shorter wavelengths (53). While the latter was true in our study (Table S1), we found that butterflies emerging from the hot house chamber had the smallest inter-ridge distance and the lowest intensity reflectance (Figs. 3 and 5). A similarly puzzling result was found when Colias butterflies were exposed to heat and cold shock treatments during their pupal phase, resulting in less bright UV iridescence but more densely packed ridges. The authors attribute the reduction in brightness to higher irregularity in the nanostructures (17). We did not find direct evidence of higher inter-ridge distance irregularity among our thermal challenge treatments, nor did we measure all the potential irregularities that may have been incurred by these environmental perturbations (i.e. number of lamellae in the ridge stack, angle of insertion of ridges, etc.). Based on our analyses, however, we posit that the drop in brightness can be at least partially explained by overall scale narrowing due to ridge tightening (Fig. 4). Cover scales are associated with an increased angular spread of reflected light from the butterfly wing, as well as selectively filtering certain wavelengths of the light spectrum, for instance, long wavelengths (e.g. reds) (33, 34). Additionally, the double layering of cover scales stacked atop ground scales has been postulated to contribute to the intensity of the reflectance spectrum (54). The narrowing of cover scales in our study was not extreme enough to lead to exposure of the underlying ground scales; however, it likely did lead to decreased overlap between adjacent cover scales, compared with the overlap that was present among butterflies from the control and temperate cool treatments. It may be that a reduction in overlapping cover scale material would result in a lower intensity signal, as the synergistic benefit of scale stacking would decrease. We also found that the cover scales of butterflies from the temperate cool treatment were wider than those of control butterflies; however, the effect size of this change was smaller than the shrinkage of the hot house butterfly cover scales. Together, the minor decrease in ridge distance and increase in scale width did not meaningfully change temperate cool butterfly reflectance to relevant perceivers (Figs. 1 and 3, Table S1).

Crossribs are implicated in ridge spacing changes, as they bridge the distance between these longitudinal structures. Although less understood, they also play their own role in building iridescence. A study comparing the iridescent blue scales of a Heliconius butterfly to its black scales revealed that a higher crossrib density is associated with the iridescent phenotype (45, 55). This may be due to the greater number of reflective layers and the decreased opportunity for the light to be transmitted through the scale and absorbed by the pigment molecules residing in the interior lumen of the ground scales (28, 45). We did not find consistent outcomes among our crossrib analyses that would lead to strong inference about their role. While Amazonian morpho from the hot house chamber demonstrated increased distance among the crossribs of hot house butterflies—matching expectations that butterflies with smaller inter-ridge distance and larger inter-crossrib distance would create duller iridescence than baseline—Panamanian morpho crossribs from hot house conditions exhibited no difference compared with control crossrib distance, and instead, we recorded a greater crossrib distance among the temperate cool butterflies. As with the Amazonian subspecies, however, Pamanian morpho that developed under the hot house treatment had the dullest iridescence (Fig. 3) and were the most visually different from control butterflies, in the eyes of their predators and other Morpho (Fig. 2). Thus, the mechanistic contribution of crossrib spacing to iridescence effects is not clear from this study. More developmental work is needed on the cellular level to understand how thermal stress would alter nanostructure spacing of the upper scale lamina. Recent research using fluorescence microscopy suggests an intriguing possibility that environmental temperature could perturb the deposition and organization of F-actin during development, potentially disrupting the scaffolding underlying ridge and crossrib formation (45).

We note that there are several elements of scale morphology and nanostructure formation that we were unable to measure in our study. For instance, scale curvature (the degree to which the scale bends downwards from base to apex) has been connected to the brightness of the reflected light (56). While it is possible that our experimental temperatures affected wing scale curvature, we were unable to measure these effects. Ridge and crossrib height also appear to be associated with intensity of the reflected light (55), but we did not obtain enough cross-sectional images to measure ridge height across treatments in this study. Iridescence color and quality can also be influenced by the thickness of the scale lower lamina (57), as well as the spacing of air pockets and chitin that is created by the lamellae stack building the ridges (41). Due to resolution limitations with the eSEM, we were unable to image the lower lamina or the lamellae stacks with enough clarity to reliably measure these elements. However, scatterometry studies on M. helenor scales have revealed that the reflectance spectra from the upper lamina structures (ridges and crossribs) and lower lamina are highly similar and thus contributions from the lower lamina likely augment, rather than alter, the butterfly color (30). We therefore expect that any unmeasured changes in the lower lamina would contribute to the changes we observed, without fully explaining them or changing the directionality of our results. Finally, we note that in the wild, caterpillars may be able to behaviorally buffer themselves from climatic temperature increases by seeking out microclimate refugia in the forest understory in which to pupate (58); an option that the individuals in our study were not given access to.

In sum, we found that environmental temperature during pupal development—the stage at which wing scales are built and ornamented—can have strong effects on the color and brightness of iridescence. This effect was most profound under hotter temperatures, which may occur by 2080 (36). While this is only a two subspecies study, it lays the foundation for a phylogeny-wide analysis of the developmental plasticity of iridescent nanostructures in Nymphalidae. Based on our results, we join the ranks of scientists finding that under a changing climate, we are draining the color from our natural world (59, 60). Many biological interactions rely on tight coordination between an organism's expressed phenotype and the sensory system of a receiver. In the case of conspecific interactions, this could mean that a dulled butterfly cannot be detected as easily amidst a convoluted forest background by a potential mate (61) or that hue shifts would lead to species misrecognition (50). In the case of antagonistic interactions, these reflectance changes may make a butterfly less effective at participating in evasive aposematic mimicry rings or at creating a successful attack-thwarting visual illusion (48, 62). We also note that these differing temperature environments had important implications for the overall survival and phenology of the adult butterfly. Mortality was higher in our hot house treatment than control or temperate cool. We also found that, for the Panamanian subspecies, the hot house treatment delayed eclosion by ∼4 days on average compared with the tropical control. The temperate cool treatment also delayed eclosion for both subspecies (∼6 days), aligning with previous work (18, 63, 64). Even a week-long delay in eclosion can have significant negative impacts on individual fitness in butterflies (65). A delay in eclosion time under the hot house condition is therefore concerning for future generations of M. helenor. This study is an important first step for understanding biotic and abiotic repercussions of global climate change and its effect on some of the natural world's most dazzling diversity.

Materials and methods

Butterfly rearing and experimental design

We initiated two separate M. helenor stocks by (i) capturing M. h. peleides in Gamboa, Panamá and purchasing pupae from The Butterfly Haven in El Valle, Panamá (Panamanian stock; permit#: MiAmbiente ARB-075-2021) and (ii) purchasing M. h. theodorus pupae from Quinta de Goulaine in Pichincha, Ecuador (Amazonian stock; permit#: PA-01-ARB/ARG-188-2024). Our estimates of the natural environmental conditions of each of these subspecies come from BIO1, 5, and 6, respectively from the WorldClim website (https://www.worldclim.org/data/worldclim21.html), using the 2.5 m resolution raster and observations from iNaturalist. We kept these populations separate in large outdoor flight cages at the Smithsonian Tropical Research Institute (STRI) in Gamboa Panamá, in the Butterfly Ecology and Evolution Research Lab's Insectaries. We reared larvae from these parent populations on potted peanut plants (Arachis pintoi) in pop-up butterfly houses in the insectary, protected from the rain by a clear plastic roof but otherwise in ambient conditions (∼28 °C on average) from August 2023 to December 2024. As the larvae neared their fifth instar, we checked host plants every day and collected pupae within 24 h of pupation to be enrolled in our experimental treatments.

To understand the role of environmental temperature in Morpho butterfly development, we created three different temperature treatment groups in Percival I-36LL chambers: control = consistent, average temperature in a lowland tropical region (Gamboa, Panamá; 28 °C day and night), temperate = average lowland tropical temperature during the day, with cooler temperatures at night (28 °C day, 14 °C night), and hot house = average lowland tropical temperature at night, with hotter temperatures during the day (35 °C day, 28 °C night). Day and night periods were 12 h each, with day beginning at 7:30 AM and night beginning at 7:30 PM. All chambers were kept at a consistent humidity range, which was allowed to fluctuate between 60 and 70%. We kept a logger (ExactLog TH: Temperature and Humidity Logger) on the middle shelf of each chamber to ensure no major deviations from our desired temperature and humidity settings. We assigned pupae their treatment group via randomization (see Supplementary Methods for more information). To analyze survival and length of pupal development data, we specified simple linear models in R Studio (66).

Spectral measurements and analyses

We measured the reflectance spectra of butterflies using an Ocean Optics kit (see Supplementary Methods for full description of hardware, software, and settings). To gather iridescence measurements, we used a machined rig (Fig. S5, as in (67)) which allowed us to vary the angle of the light source and the sensor relative to each other and the butterfly wing sample. We placed individual butterfly wings on the stage of the rig and took reflectance recordings from two different angles: both probes parallel to each other and as close to a 90° angle from the butterfly wing as possible (88° and 92°; hereafter “perpendicular”) and both probes ∼45° from each other and ∼68° from the butterfly wing (67.5° and 112.5°; hereafter “oblique”). For each wing, we took measurements with the wing flat and then tilted the stage until we found the maximum reflectance (MR) of the wing. We always measured between the anal and cubitus veins on the right forewing and hindwing (as the left was used for other measurements) and performed three technical replicates of each measurement (see Fig. S5 for the image of the setup). To eliminate any unintended effects by measurement order, we randomized the measurement sequence of the treatments (see Table S2 for n′s per sex per treatment).

We used the pavo2 package (39) in R Studio (68) to analyze our spectral measurements and focused all analyses on our MR measurements, averaged across our three technical replicates for both perpendicular and oblique angles (see Table S2 for n's per sex per treatment). To ready our data for visual modeling, we used the “as.rspec()” function with wavelength limits between 300 and 700 nm and then smoothed the spectra via “procspec()” with the default span of 0.2. We approximated how an ultraviolet-sensing avian predator, a violet-sensing avian predator, and a congener would each perceive the butterfly color from the differing angles using the “vismodel()” function with the visual model set to avg.uv, avg.v, or Nymph, respectively. Nymph was a custom-built perceiver model that defined five sensitivity peaks, according to a neurophysiology study done on Morpho peleides, ranging from the ultraviolet range to the long wavelength range (345, 451, 508, 570, and 602 nm) (52). We then used the output from “vismodel()” in the “bootcoldist()” function (1,000 bootstraps) to estimate the perceived contrasts between butterflies of different treatments, based on our perceivers of interest. For the average UV-sensitive bird, we used a relative receptor density of 1:2:2:4 (UV:short:medium:long wavelengths) and a weber fraction value of 0.10 for chromatic receptors and 0.18 for achromatic receptors. These weber fractions provide an estimate signal:noise ratio of the visual receptors and are used in weighting the photoreceptor quantum catch to generate the color distance results. For the UV-nonsensitive bird, we used a relative receptor density of 1:2:4:4 (short:medium:medium2:long wavelengths), a weber fraction of 0.06 for chromatic receptors and 0.22 for achromatic receptors. We instituted these values for average avian vision following (43) and therein cited papers. For the color contrast analysis using the nymphalid visual model (Nymph), we used a relative receptor density of 1:3:2:1:1 (aligning with the wavelengths described above), as described in Pirih et al., a weber fraction of 0.06 for chromatic receptors and 0.16 for achromatic receptors. These values are informed by work performed on honeybees, as there has not yet been successful documentation of weber fractions for butterflies and thus we follow the lead of Dang et al. in using these estimates (see R scripts on Zenodo (69)).

After acquiring color contrast results, we were interested in understanding what components of the spectral curves were contributing to these differences. We therefore used the function “peakshape” from the pavo2 package (40) to measure the wavelength at peak reflectance intensity and the intensity level of that peak. Given that M. helenor has multiple peaks in their reflectance spectrum, at least one in the UV and one in the blue portion of the spectrum, we broke up the spectral data for each individual into a UV set (260–379 nm) and a visible set (380–700 nm). To do so, we re-imported the spectral data using “as.rspec()” but this time with wavelength limits defining the two spectral sets, UV and visible. Using the R package brms (70), we then ran Bayesian linear models to determine whether these peak hue and brightness measurements differed by environmental temperature treatment and whether there was a difference among the sexes. We used ChatGPT-5 to assist with code optimization.

Microscope slide photography and analyses

We next wanted to get a more mechanistic understanding of these spectral differences. We therefore collected photos of both cover and ground scales on the right forewing and hindwing for ∼8 individuals per sex per treatment for both subspecies using a Leica DM2500 LED light microscope with attached Leica camera and Leica Acquire software (see Table S3 for n's per sex per treatment and data on Zenodo (69)). To keep our measurements consistent, we focused on the region between the anal and cubitus veins. When possible, we photographed ground scales that naturally emerged from small damage spots to the wing, as we found these to be the cleanest removal of cover scales without disturbing the ground scale matrix. In the absence of damage spots, we gently made our own scratch using the tip of a microscope slide to reveal the ground scales. We took two photos of each scale layer, at 5× and 10×. Our analyses focus on the 10× images, which we imported into ImageJ (71) and used the straight-line tool to measure horizontally across the scale just behind the distal apex, at the widest point (see Fig. S6 for example image). We attempted to measure at least six scales per layer per individual; however, in some cases, only two scales were available for accurate measurement due to focus issues (see datasheet on Zenodo for individual measurements). To determine whether temperature treatment had an effect on scale width and to account for any inherent differences between males and females, we built Bayesian linear mixed effects models for each subspecies and scale type, again using the package brms, with treatment and sex as fixed effects. We included butterfly ID as a random effect to account for multiple measurements per individual.

Scanning electron microscopy

To further understand the mechanistic underpinnings of environmentally induced color changes, we gathered images of the nanostructures on the cover and ground scales using an eSEM at El Instituto de Investigaciones Cientificas y Servicios de Alta Tecnología de Panamá (INDICASAT). We again focused on the region between the anal and cubitus veins on the left forewing and imaged at least three individuals per sex per treatment for both subspecies. Due to financial and time restrictions, we focused our imaging on just the forewing, as this is likely to be the most visibly available wing set to a potential viewer. We removed an ∼1 cm × 1 cm square from this interveinal region and directly applied it to a stub, cemented by copper tape, with the dorsal surface of the square pointed up. For the ground scale photos, we disturbed the top cover scale layer using gentle scratching. eSEM's do not require gold sputtering and we therefore proceeded with collecting images at 500×, 1,000×, and 6,000× at 10 kV. To measure the distances between nanostructures, we employed the InterEdgeDistance ImageJ macro on 6,000× magnification images. Using the segmented-line tool (to account for any slight curvature), we traced sequential ridges in each image and had the program measure from 50 locations between sequential ridges with equidistance set to 25, which approximately broke the entire length of visible ridge material into equal parts. We used the straight-line tool to trace sequential crossribs, beginning and ending the line where the crossrib bisected each ridge (see Fig. S7 for example). Here, we measured from five locations with equidistance set to 1. If the macro could not measure the distance between two crossribs (due to slight offsets between the crossribs), we skipped that pair and moved on to the next. We made sure to have at least four distance measurements per structure per scale but usually had quite a few more (see Table S4 for total individual n's per sex per treatment and data sheet on Zenodo (69)).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Competing Interests

The authors declare no competing interests.

Funding

This work was made possible via the Smithsonian Tropical Research Institute Anthony Coates Postdoctoral Fellowship and the National Science Foundation Postdoctoral Research Fellowship in Biology (2305911) to J.J.R. and funds from the Smithsonian Institution and National Science Foundation (IOS 2110532) to W.O.M.

Author Contributions

Juliette J. Rubin (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing—original draft), Leo T. Camino (Formal analysis, Investigation, Methodology, Visualization, Writing—review & editing), Yanileth F. López-Tacoaman (Formal analysis, Investigation, Methodology, Writing—review & editing), Pushkar R. Wagh (Formal analysis, Investigation, Writing—review & editing), Greta Carolina Hernández Campos (Investigation, Writing—review & editing), and W. Owen McMillan (Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing—review & editing)

Data Availability

All study data are included in the article and on the Zenodo repository (DOI: 10.5281/zenodo.21497628).