Ultraviolet exposure has an epigenetic effect on a Batesian mimetic trait in the butterfly Papilio polytes

Wing polymorphism of butterflies provides a good system in which to study adaptation. The Asian Batesian mimic butterfly Papilio polytes has unmelanized, putative mimetic red spots on its black hind wings. The size of those red spots is non-heritable but it is highly polymorphic, the adaptive significance of which is unknown. We hypothesized that under strong ultraviolet (UV) irradiation, butterflies develop a wider melanized black area to protect the wings from UV damage, and as a result express smaller mimetic red spots. Our field survey on Okinawa Island revealed a negative relationship between the sizes of the red spot and the black area in the wings. The size varied seasonally and was negatively correlated with the intensity of solar UV radiation at the time of capture. Laboratory experiments revealed that the size was reduced by strong UV irradiation not only of the eggs and larvae, but also of their mothers through a putative epigenetic mechanism. The flexible phenotypic expression of the red spots in P. polytes suggests a trade-off between protection against UV damage and predation avoidance, and provides a new insight into the evolution of Batesian mimicry.


Results
Phenotypic trade-off between black and red areas. We first tested the hypothesis of a phenotypic trade-off between the red spot size and the background black area on hind wings. We estimated the correlation between those values of f. polytes collected on Okinawa Island in 2014 and 2015. The background black area (%) was defined as 100 × (hind-wing area − [red + white mimetic spot areas])/hind-wing area 4 . There was a significant negative correlation between them (r = −0.82, P < 0.0001, n = 181, Fig. 2).
Seasonal changes in red spot size. The red spot size of females captured in 2014 and 2015 varied seasonally and was significantly negatively correlated with the cumulative intensity of the solar UV-B radiation of the month in which they were captured (2014, P < 0.01, n = 73, Fig. 3a; 2015, P < 0.0001, n = 108, Fig. 3b). The females had smaller red spots during June and July, when the solar UV-B radiation was strongest.
Effect of experimental UV exposure on red spot size. The red spot size of females exposed to UV radiation at the larval stage was significantly smaller than that of unexposed controls (Fig. 4). Remarkably, that of females whose mothers were exposed to UV was also smaller, suggesting an epigenetic effect. Those effects  were additive and independent, as the effect of their interaction was not significant in either the likelihood test or the full-model ANOVA (Fig. 4). The effect was not caused by a change in host plant quality by UV irradiation, because feeding of larvae on UV-exposed leaves did not significantly affect the red spot size (F 1,7 = 0.04, P = 0.84, Fig. 5). We also investigated whether UV exposure in the laboratory caused not only a wider black area but also more-melanized background black areas. UV irradiation slightly reduced the average brightness value (lower brightness = darker wing), but not significantly so (F 1,11 = 1.95, P = 0.19, nested ANOVA; Supplementary Information).

Discussion
Our results clearly show a negative correlation between UV irradiation and red spot size in the field (Fig. 3). In the laboratory, UV irradiation at the egg and larval stages reduced the red spot size at the adult stage (Fig. 4). This was not caused by UV irradiation of the food plants (Fig. 5). Furthermore, UV irradiation of adult females reduced the size of red spots of their daughter butterflies (Fig. 4). These results support the hypothesis that mimetic  Forma polytes has both the white spot and the red spot mimetic traits on the hind wings. Katoh et al. 4 reported that the white spot size had heritable variation, but that the red spot size was not heritable. However, our field data and results of the UV exposure experiment indicate that the red spots can respond flexibly to environmental changes. An important finding is that maternal-experienced environments affected offspring phenotypes through putative epigenetic effects. Papilio polytes (as well as the model, P. aristolochiae) is multivoltine, and on Okinawa the egg-to-adult growth time is about 1 month (M. Katoh, unpublished data). Therefore, it is highly likely that larvae will be exposed to UV of a similar strength to what their mothers experienced. Given this similarity in environmental UV strengths between parent and offspring generations, the observed epigenetic effect could be a result of an adaptive strategy by which offspring can prepare to allocate more melanin to protecting wings from UV when necessary.
This result raises a new question of UV-irradiation experiment. As strong UV radiation reduces the red spot size of f. polytes, the putative aposematic red spot becomes more dissimilar to the model's. Does this response therefore degrade the mimetic effect? One possibility is that the white spot of f. polytes is the most salient mimetic signal to predators, outweighing other imperfect mimetic traits 9 and maintaining the mimetic efficiency. Another possibility is that the increased UV protection function compensates for the fitness cost of the increased predation risk due to imperfect mimicry. To test these ideas, we should empirically examine the effect of wing degradation by UV radiation on fitness of P. polytes in the field. It is important also to note that butterfly wing colour patterns have various functions such as thermoregulation 10,11 and mate choice 12 , besides mimetic signals 13 .
Epigenetic effects in which parent-experienced environments affect phenotype expression of their offspring are known to have some basic mechanisms, such as DNA methylation, histone acetylation, and microRNA expression 14 . The mimetic coloration of P. polytes in the Ryukyu Islands provides a good opportunity to study the relative importance of those developmental and ecological mechanisms in insects.

Methods
Measurement of spot size. The red spot size is defined as the area of red spots relative to the area of the hind wing. We measured this in the following way 4 : The wing of each butterfly (ventral view) was photographed with a digital camera (Nikon D7000, Nikon Corporation, Tokyo, Japan) mounted 50 cm directly above the specimen under invariant light (Fig. S1). The left wing was used unless it was damaged. We analysed the hind-wing area below a horizontal line (Fig. S1) demarcated as follows. We defined landmark 1 as the intersection of the second anterior cubitus vein (the border between space 1b and space 2) and the border of the cell on the hind wing, and landmark 2 as the tip of the vein on the tail of the hind wing. The line connecting landmarks 1 and 2 was rotated in Adobe Photoshop CS5 (Adobe Systems Inc., San Jose, CA, USA) so that it was vertical. Then a horizontal line crossing landmark 1 was drawn perpendicular to the vertical line. The red spot size was defined as the sum of the total area of the red spot on the hind wing divided by the total hind-wing area (the area bounded by a dotted line in Fig. S1) measured in ImageJ software (National Institutes of Health, Bethesda, MD, USA).
We similarly measured the red spot sizes of the model, P. aristolochiae, for comparison. The dorsal side of wings would be more likely to be exposed to UV light than the ventral side. However, we used the ventral side because we did not need to open the wings. There was a strong correlation in the red spot size between the dorsal and ventral sides in mimetic females of P. polytes (r = 0.78, P < 0.0001, n = 66, collected in the field population on Okinawa Island in 2011-2013). Seasonal variation in red spot size. In the daytime (09:00-14:00) at intervals of 4 to 33 days during 2014 to 2015, we captured mimetic P. polytes females in a field of Nakijin Village on Okinawa and measured their red spot sizes to reveal any seasonal pattern.

Comparison of red spot size with the model's.
In 2015 on Okinawa, we measured the red spot sizes of the model, P. aristolochiae, captured in the field (n = 10) from May to October and compared it with that of f. polytes (n = 108) specimens collected in the field from March to October.

Laboratory rearing and UV-irradiation experiment. A laboratory experiment was conducted in 2015
and 2016. We captured mimetic females in Nakijin in June of each year. As those butterflies were collected in the field, their age was unknown. Adult females were kept in paper triangles, brought to the laboratory, and randomly assigned to one of two groups the next morning (between 06:00 and 07:00). The first group was kept in an incubator (25 ± 1 °C, 13-h light/11-h dark, 70 cm × 58 cm × 102 cm high) with a UV lamp (Vivarium Glow Power UVB, 20 W, Pogona-Club, Saitama, Japan). The other (control) was kept in the same condition but with a fluorescent lamp (FL20S.D-EDL-D65, 20 W, Toshiba, Tokyo, Japan). Note that all of these butterflies were always kept individually in paper triangles to avoid damage due to flying in the incubator (except at feeding time). The paper triangles block only about 9.3% of UV radiation (M. Katoh, unpublished data). The air temperature in the incubators was automatically controlled. We confirmed that the difference in lamps did not affect the temperature inside the incubators (M. Katoh unpublished data). Once a day we fed the females with 15% sucrose solution until they stopped feeding. The butterflies in the UV group were exposed to UV from 06:00 to 19:00 for 3 days (except at feeding time). We let the butterflies in which both groups oviposit on fresh leaves of Toddalia asiatica in the laboratory under fluorescent light (25 ± 1 °C, 13-h light/11-h dark) on days 4 and 5. After oviposition, the eggs were immediately transferred to one of the two randomly assigned incubators (UV or control). Hatched larvae were reared in those incubators on fresh T. asiatica leaves. As the final butterfly wing colour patterns are determined during late larval to pupal development [15][16][17] , we timed the UV exposure to fall before this stage. Gut-purged larvae were moved to the laboratory for pupation. The emerged butterflies were killed within 4 h after emergence and stored in a freezer at −30 °C to avoid physical damage to their wings. We assigned 21 larvae from 8 mothers to the unexposed control, 19 larvae from 6 mothers to the mother-only exposure treatment, 33 larvae from 10 mothers to the larva-only exposure treatment, and 32 larvae from 7 mothers to the both-exposure treatment. Note that although we use the term "larval exposure" for simplicity, we exposed both egg and larval stages.
We calculated the standardized red spot size of the females in the two generations as 100× (red spot size of the lab-bred female − red spot size of its field-collected mother)/(red spot size of the field-collected mother) under the four exposure treatments.
As both butterflies and host plant leaves were held under UV, a change in host plant quality caused by UV exposure might have indirectly caused a change in the red spot size. To test this possibility, we exposed leaves of T. asiatica in the incubator to UV as in the main experiment for 1 to 2 days and then provided them to the larvae of mother butterflies collected in Nakijin Village in 2016 and 2017. We assigned 23 larvae from 4 mothers to the host-plant-only exposure treatment and 14 larvae from 5 mothers to the unexposed control.
Statistical analysis. All analyses were performed in R v. 3.2.3 software 18 . We used a generalized additive model (mgcv in R) to test the relationship between the seasonal variation in the red spot size and UV intensity in the field. The red spot size of field-collected mimetic females was the response variable, and the average daily integrated solar UV-B radiation on Okinawa in the month when each female was collected (obtained from the Japan Meteorological Agency) was the explanatory variable. Smoothing splines were fitted to each data set. We tested the influence of UV irradiation on the expression of the red spot size using a generalized linear mixed model (GLMM, lme4 in R) with a Gaussian error structure and an identity link function, using maternal exposure, larval exposure, and their interaction as fixed effects and year as a random effect. We assessed the significance of each effect using the likelihood-ratio test (ANOVA of the model with a particular explanatory variable vs the model without that variable) and ordinal ANOVA of the full model. To test any effect of a change in host plant quality, we used a GLMM with treatment as the fixed factor and mother as the random factor.