Polarized light sensitivity in Pieris rapae is dependent on both color and intensity

Polarized light cues are used by many arthropods but apart from polarized skylight navigation little is known about how these organisms perceive polarized reflections (Heinloth et al., 2018). All organisms with rhabdomeric photoreceptors have the potential to sense polarized light (Horváth and Varjú, 2004). The tubular structure of the microvilli forming the rhabdom results in photopigments aligning more along the long axis of the microvilli. This alignment, in turn, causes these photopigments to be more sensitive to light vibrating in the plane parallel to the long axis of the microvilli (Johnsen, 2011). The plane of polarization with the greatest photoreceptor sensitivity is referred to as φ_max and typically aligns with the microvillar orientation (Horváth and Varjú, 2004). The size of this difference in sensitivity is referred to as polarization sensitivity (PS) and is defined as the the ratio of sensitivity to light vibrating at φ_max, and to light vibrating orthogonal to φ_max. Photoreceptors with a high PS are typically found in a specialized area of the compound eye known as the dorsal rim, allowing for polarized skylight navigation (Labhart and Meyer, 1999). The microvilli of these photoreceptors are aligned, and untwisted, along the length of their relatively short rhabdom, thereby enhancing PS. Additionally, these high PS photoreceptors involved in skylight navigation, which differ in φ_max, all share similar spectral sensitivities. If these photoreceptors differed in both spectral sensitivity and φ_max, the perceived color of an object would depend on both its reflection spectrum and its polarization (Wehner and Bernard, 1993). Many insects avoid polarization-induced false colors by twisting the direction of these microvilli along the length of the rhabdom, because otherwise the perceived color of objects would change as insects navigate through the environment. However, many other insects, especially those in aquatic and semi-aquatic habitats (Horváth and Csabai, 2014), possess photoreceptors with moderate PS throughout their compound eyes, and some of these insects do experience these polarization-induced false colors (Kelber et al., 2001). Histological and electrophysiological work has also revealed evidence for PS in many herbivorous insects (Ilić et al., 2016; Mishra, 2015; Wachmann, 1977).

Recently, small white butterfly (Pieris rapae) females have been shown to discriminate among potential host plants based on the polarization of light reflected from their foliage (Blake et al., 2019a). Like any shiny surface, the leaf surface preferentially reflects light oscillating parallel to that surface (Horváth et al., 2002; Shashar et al., 1998). This axis of polarization (AoP, 0–180 deg), as well as the degree to which the foliar reflection is polarized (degree of linear polarization, DoLP, 0–100%), are both strongly dependent upon the viewing angle and the location of the light source. However, AoP (unlike DoLP) is largely independent of leaf surface characteristics (Blake et al., 2019a). As only the specular component of the reflection is polarized, any leaf characteristics that affect the relative shininess or mattness also affect the DoLP. Decreasing the diffuse reflection through absorbance by pigments, scattering the specular reflection with epicuticular waxes or pigments, or undulations of the plane of the leaf's surface can all affect the DoLP of foliar reflections (Grant et al., 1993). Being dependent on these leaf characteristics, foliar DoLP can convey information about the host plant not conveyed by its color or intensity. Female P. rapae are able to discern cabbage host plants and potato non-host plants based on the lower DoLP of cabbage leaf reflections (Blake et al., 2019a). In choice bioassays, which presented manipulated host plant images, P. rapae females rejected most images with a DoLP dissimilar to that of their cabbage host plant (DoLP of 31%). The informative value of this cue during host plant selection is enhanced by a relative insensitivity of P. rapae females to all but AoP very near to 45 or 135 deg. Both the underlying neurological mechanism and the photoreceptors involved in this discrimination remain unknown.

The visual system of P. rapae resembles that of other butterflies in that each ommatidium contains nine photoreceptors and the three ommatidial types are arranged in a random mosaic throughout the compound eye (Fig. 1A). Similar to the ommatidia of Papilio butterflies (Kelber, 2001), the shortwave-sensitive (UV, violet, blue) R1,2 photoreceptors, with the exception of the polarization-insensitive UV photoreceptor, respond most strongly to vertically polarized light, whereas the longwave-sensitive R3–9 photoreceptors respond most strongly to horizontally polarized light (R3,4) and obliquely polarized light (R5–8) (Blake et al., 2019b; Fig. 1B, Table 1). In the ventral portion of the eye, the sensitivity of the R5–8 photoreceptors, which, like R3,4, express a green-sensitive opsin, are modified by perirhabdomal filtering pigments into three classes of red photoreceptors distinct to the three ommatidial types, with more variation in PS among ommatidial types than reported in Papilio (Table 1). Of the shortwave receptors, only the type I blue photoreceptors show significant PS. There is also a lower PS in type II R3,4 receptors, whose polarization filtering effects on more basal photoreceptors (Snyder, 1973) may explain the difference in the axis of maximal polarization sensitivity (φ_max) of red photoreceptors among ommatidial types (Blake et al., 2019b). The R9 receptor is thought to be red-sensitive (Shimohigashi and Tominaga, 1991), and likely has low PS owing to its bidirectional microvillar arrangement (Qiu et al., 2002).

The compound eye of P. rapae has been extensively characterized, but there is no obvious mechanism that would explain how P. rapae processes the signals from its suite of photoreceptors to discriminate among stimuli with different DoLP. To determine whether P. rapae perceives differential DoLP as differences in stimulus intensity or color, we sought to emulate the work of Kinoshita et al. (2011). In two-choice bioassays, we examined the responses of P. rapae to differences in DoLP or color between stimuli to determine whether intensity differences between the stimuli affected preference in a similar manner. We also determined the photoreceptors involved in DoLP discrimination by minimizing the blue, red, or blue and red light of cabbage images that we presented to P. rapae in bioassays. This type of manipulation is possible through use of our novel monitor bioassay (Blake et al., 2019a). We predicted that if a photoreceptor were involved in DoLP discrimination, then image manipulations of stimuli reducing the photoreceptor's stimulation should alter the behavioral response of P. rapae to DoLP differences. We also modeled the catch of all P. rapae photoreceptors aiming to explain the observed behavioral bioassay responses of P. rapae.

Our laboratory colony of P. rapae rapae (Linnaeus 1758) originated from eggs obtained from the Carolina Biological Supply Company (no. 144100, Burlington, NC, USA) and later from adults collected from cabbage fields near Delta, BC, Canada. Using a well-established protocol (Webb and Shelton, 1988), larvae were maintained either on a wheat-germ diet or on cabbage plants grown in a greenhouse. We housed both male and female adults in indoor cages (60×60×60 cm, BugDorm 2120, MegaView Science Co. Ltd, Taichung, Taiwan) kept at 18–25°C and a photoperiod of 16 h:8 h light:dark. The females we tested in experiments were randomly selected from cohorts of adults 3–14 days post eclosion and were assumed to be gravid. We tested females in multiple bioassays, each bioassay presenting a new pair of experimental plant images. These different bioassays were considered independent.

We used the same experimental arena (31.6×76.5×32.1 cm) and LCD monitor setup as recently described (Blake et al., 2019a; Fig. 2A). The inner surface of the removable arena lid was lined with matt white banner paper (NCR Corp., Duluth, GA, USA). We left the two end sections of the arena facing the monitors (stimulus windows) unobstructed but lined all the other inner surfaces of the arena with a matt brown kraft paper (NCR Corp.). To prevent build-up of any olfactory cues in the arena, we replaced the paper lining the interior surfaces and cleaned exposed glass surfaces with hexane daily.

In all experiments, we displayed cabbage plant images, created through photo polarimetry, as detailed in a recent publication (Blake et al., 2019a). In summary, we photographed cabbage plants, corrected the image color balance using a reflectance standard (SRS-99-010, Labsphere, NH, USA), removed the image background, and then standardized the plant size in each image such that all plant images presented an equal number of pixels. The pixel values of these red/green/blue (RGB) images were then manipulated to create versions that differed in intensity or color (Table S1, Fig. S1D–F). These images were presented on paired liquid crystal display (LCD) monitors (1707FPt, Dell Inc., Round Rock, TX, USA) calibrated to minimize any differences between monitors in the displayed irradiance spectra of pixels with identical RGB values (Fig. S1C). These monitors lack UV irradiance but the absence of UV wavelengths did not affect DoLP-based host plant preferences (Blake et al., 2019a). As LCD monitors produce highly polarized light, we manipulated the AoP by rotating the display and counter-rotating the image. Using a λ/4 retarder film (no. 88-253, Edmund Optics, Barrington, NJ, USA), we were also able to manipulate the plant image DoLP by changing the alignment of the AoP of the display relative to the retarder film (Blake et al., 2019a,c). Using LCD monitors also enabled us to readily manipulate the plant image's color and/or intensity.

The monitors were separated from the stimulus windows of the experimental arena by a stimulus chamber (31×31×47 cm) lined with the same kraft paper as the arena. This separation limited the range of viewing angles of the monitor from within the arena. In order to limit the visible portion of the LCD to that displaying the plant image, we placed a kraft paper plant mask over the display aperture in each stimulus chamber (Fig. 2B). The top of each stimulus chamber had a lighting aperture (27×26 cm) covered with the same white banner paper as the arena lid, thus affording similar illumination of the arena and the stimulus chambers. The arena and the chambers were lit by a fluorescent lamp (Fig. S2B; F32T8/SPX50/ECO GE, Boston, MA, USA) centered 15 cm above the arena.

Using a camera mounted at the top rear of each stimulus chamber, we monitored the response of P. rapae females introduced into the arena. We allowed each female up to 5 min to approach one of the stimulus windows and recorded this approach as a behavioral response to the associated plant image. We considered females making no response as non-responders. Image stimuli were alternated so they appeared equally often on both monitors/sides of the arena. To help minimize any time-of-day (Lazopulo et al., 2019), day/weather (Pellegrino et al., 2013; Roitberg et al., 1993) or cohort-of-butterflies effects, we ran bioassays in blocks that included all stimuli comparisons using butterflies from a single cohort. There were two exceptions to this blocking: (1) the color-removal experiments commenced comparing R+G+B and G bioassay treatments at AoP 90 deg and only later included the remaining AoP 90 deg treatments; and (2) the AoP 0 deg bioassays in the color-removal experiment were a follow-up to the AoP 90 deg experiments and did not proceed concurrently.

To determine whether P. rapae females perceive differential DoLP as differential color or intensity, we performed experiments similar to those of Kinoshita et al. (2011). We presented females with paired stimuli consisting of the same cabbage image but modified to create differences in (A) intensity, (B) color and intensity or (C) DoLP and intensity between the two images (Fig. 3, Table S1). The paired stimuli we presented were (A) two unmodified images both with a DoLP of 31%; (B) one unmodified (treatment) image and one red-shifted (control) image each at a DoLP of 31% (Table S1, Fig. S1D,E); and (C) two unmodified images presented with a DoLP of either 31% (treatment) or 50% (control). The image whose intensity remained constant in each sub-experiment was designated the control, but this control image was not identical in each sub-experiment. In sub-experiments A–C, we presented the treatment image at intensities lower (44%, 87%), equal (100%) and greater (130%) than the original intensity (Table S1, Fig. S1D,E). In sub-experiment A, we did not present a choice between two unmodified images (DoLP 31%, 100% intensity), assuming no preference in response.

To determine the photoreceptors involved in polarized light discrimination, we modified the color of cabbage images and offered P. rapae females a series of choices between these modified images presented at a DoLP of either 31% or 50%, with both images presented at an AoP of both 0 and 90 deg. To minimize the stimulation of the butterflies' red photoreceptors, blue photoreceptors or both simultaneously (within the limits inherent in the RGB color space where each color channel stimulates multiple photoreceptor classes; Fig. S1C), we set the red, blue, or red and blue values, respectively, of all pixels in both stimulus images to 0 (Table S1, Fig. S1F). As a control, we also offered a choice between images with no modification to any pixel values.

We used two-tailed chi-square tests to determine whether the proportion of P. rapae females responding to plant images differed from 0.5, and whether the proportion of females responding differed among the experimental treatments. We excluded non-responding females from statistical analyses.

Unless otherwise noted, all spectra were measured with a calibrated spectrophotometer (HR-4000, Ocean Optics Inc., Dunedin, FL, USA) and were recently reported (Blake et al., 2020). To allow us to calculate the quantum catch of the background, we measured the ambient irradiance of the fluorescent lamps within the arena, the transmission of the arena wall and the λ/4 retarder film, and the reflectance of the brown kraft paper (Fig. S1A,B). We measured reflectance with a JAZ spectrometer (Ocean Optics) calibrated with a 99% Spectralon reflectance standard (SRS-99-010, Labsphere, NH, USA). Using photo polarimetry of the arena's interior (Foster et al., 2018), we approximated the mean DoLP and modal AoP of the background across the human visible spectrum (400–700 nm) to be 10% and 90 deg, respectively.

Using the mean RGB pixel values of the stimulus image, we could then create a mean spectrum for all pixels displayed in the image. The resulting spectrum was corrected for the transmission spectrum of the aquarium wall and the λ/4 retarder film (Fig. S1A).

In general, when treatment and control stimuli differed only in intensity, P. rapae females preferred the more intense stimulus (Fig. 3A). This preference was statistically significant only when the intensity of treatment stimuli was <50% of that of the control stimuli (χ²=8.70, N=1, P=0.0032). When the treatment stimulus had an intensity of 87% relative to the control stimulus, females did not discriminate between these stimuli.

When we presented a choice between a red-shifted cabbage image (control) and an unmodified (treatment) image of varying intensity, females significantly preferred the treatment image only with an intensity of 87% relative to the control image (χ²=12.30, N=1, P=0.0005; Fig. 3B). Treatment images of a higher or a lower intensity were not significantly preferred, although there was a marginal preference for the treatment image when it had an intensity equal to, or greater than, that of the control image.

Similarly, when the treatment and control image differed in DoLP, females significantly preferred the treatment image (DoLP 31%) only when it had an intensity equal to that of the control image (DoLP 50%; χ²=8.32, N=1, P=0.0039; Fig. 3C). Treatment images of a lower intensity were as attractive as the control image while there was a non-significant preference for the control image when it was more intense.

When cabbage images were presented with all color channels intact (R+G+B), P. rapae females preferred the image with the lower DoLP both at an AoP of 0 and 90 deg (AoP of 0 deg: χ²=7.36, N=44, P=0.0067; AoP of 90 deg: χ²=15.25, N=63, P=0.0001; Fig. 4). When the blue color channel was removed (R+G), females shifted their preference towards the image with a higher DoLP, but only at an AoP of 0 deg (χ²=18.75, N=86, P<0.0001). When the red color channel was removed (G+B), females preferred images with the higher DoLP at both AoPs (AoP of 0 deg: χ²=11.72, N=53, P=0.0006; AoP of 90 deg: χ²=7.41, N=39, P=0.0064). When only the green color channel of the image was included, females did not discriminate between images with a high or a low DoLP, when presented at an AoP of 90 deg. However, when these images were presented at an AoP of 0 deg, females chose the lower DoLP images (χ²=9.28, N=57, P=0.0023), similar to their response when all color channels were intact.

Our study refines the possible neurological processing mechanisms for DoLP-based host plant discrimination by female P. rapae. According to our data, P. rapae females are likely not perceiving differences in DoLP as differences purely in intensity or in color. Rather, our data suggest that perception of color, intensity and polarization, at least in the context of host plant discrimination, are all linked and contingent upon one another.

The intensity versus color discrimination experiment revealed that females preferred the plant image with greater intensity when all other factors were equal (Fig. 3). In our study, color preferences shifted in response to intensity changes in one of the two test stimuli, contrasting with results obtained in similar studies with Papilio butterflies (Kelber and Pfaff, 1999; Kinoshita et al., 2011). Although it is possible that P. rapae lacks true color vision (the ability to discriminate between colors independent of intensity), this explanation seems unlikely given the shared evolutionary history of Papilio and Pieris butterflies as members of Papilionoidea (Wahlberg et al., 2005), and the similarities of their respective compound eyes (Kelber et al., 2001). Although our colored stimuli lacked an appreciable UV component (unlike many stimuli tested with Papilio), these stimuli should provide adequate stimulation of the UV photoreceptors to distinguish between stimuli in the color-removal experiment (Fig. S2B). Training of bioassay insects offers a more likely explanation for these contrasting results. Although we tested the innate preferences of P. rapae females, corresponding studies with Papilio involved rewarded training (Kelber and Pfaff, 1999; Kinoshita et al., 2011). The spontaneous color choices of Pieris brassicae also shift in accordance with stimulus intensity (Scherer and Kolb, 1987); however, increases in intensity always have a positive effect on preference, contrasting our color preference data (Fig. 3B). When paired images were similar in color and DoLP, we observed a positive linear relationship between the intensity of the treatment image and the preference of female P. rapae for this image (Fig. 3A). In contrast, when image pairs were dissimilar in color or dissimilar in DoLP, female preference for the treatment image declined when the intensity of the treatment image was greater than that of the control image (Fig. 3B,C). Like in experiments with Papilio, these results suggest that P. rapae butterflies perceive differences in DoLP in a manner similar to their perception of differences in color, albeit not independent of intensity.

The color-removal experiment revealed that blue, green and red photoreceptors are involved in the perception of differential DoLP. This conclusion is based on data showing: (i) preferential responses to images with a lower DoLP (AoP: 0 and 90 deg) when all color channels were present; (ii) a preference shift for images (AoP: 0 or 90 deg) where either the blue or the red channel was removed; and (iii) the reversal of preferences with the green-only channel images (AoP: 90 deg) as compared with R+G or G+B images (AoP: 90 deg).

Contrary to results of the intensity versus color discrimination experiment, modeling of photoreceptor catch does not support the concept that differences in DoLP are perceived as color differences, at least not when modeled as a linear interaction among photoreceptors (Kelber, 1999, 2001). The color triangles represent the modeled P. rapae color space and depict the relative quantum catch of the red, green and shortwave (omitting UV in type I) photoreceptors of the three ommatidial types disregarding intensity (Fig. 5, Fig. S2). In modeling the catch of the red photoreceptors, we assumed the catch of R5–8 are pooled, negating much of the PS of these photoreceptors. If DoLP discrimination could be explained through linear interactions between different photoreceptors, as seen in Papilio and in P. rapae with unpolarized stimuli, we would expect a consistent direction of preference between stimuli. For example, using existing linear color models for Papilio and Pieris, with the catch of green photoreceptors having a positive effect and blue and red receptors having a negative effect, we would expect the stimuli closest to the upper green vertex to be preferred. In our modeling, stimuli differing only in polarization characteristics largely align along the blue to green axis, with the direction of preference among paired stimuli tested converging on no one region of the color space (Fig. 5). This inconsistency applied to all ommatidial types (Fig. S2), albeit with smaller separations among low and high DoLP stimuli owing to lower PS of the photoreceptors. It is unlikely that this inconsistency could be resolved even if downstream opponent processing was considered (Chen et al., 2019) or if photoreceptors were to be compared among different ommatidial types (Takemura and Arikawa, 2005).

Other plausible mechanisms also fail to explain our bioassay results. If polarization discrimination by P. rapae were to be dependent on comparisons between any two polarization-sensitive photoreceptors, or between one polarization-sensitive and one insensitive photoreceptor, we would expect AoP to have a strong effect on preference (Fig. S3A) (How and Marshall, 2014), similar to how Papilio butterflies strongly prefer horizontally over vertically polarized light (Kelber, 2001). We would also expect such comparisons among photoreceptors to result in either a linear increase or decrease in preferential response as DoLP increased (Fig. S3B; How and Marshall, 2014). Yet, we found that the attractiveness of test stimuli was not affected by AoP outside regions near 45 and 135 deg, and that images with a DoLP similar to that of their cabbage host plants (DoLP of 31%) are preferred, with the appeal of stimuli declining both above and below this 31% value (Blake et al., 2019a). Comparisons between two or more pairs of photoreceptors are also unlikely to explain the observed DoLP preferences of P. rapae (Fig. S3E,F). Models that incorporated the absolute value of the differences in responses between photoreceptors (Fig. S3C,D) (Meglič et al., 2019) could explain observed AoP preferences in P. rapae, but again would fail to explain DoLP preferences. The results of the color-removal experiment preclude true polarization vision (the ability to discriminate among stimuli independent of color or intensity), as changes in color prompted large shifts in polarization preference.

Our combined results suggest a new and as of yet undescribed mechanism for the processing of polarized reflections underlying DoLP discrimination in P. rapae. The mechanism likely involves blue, green and red photoreceptor classes, and is affected by intensity, color and polarization. If true, this would be yet another example of unique neural processing of polarization information from object-reflected light. The systems for processing such information differ between all taxa thus far studied, including crabs (Smithers et al., 2019), fruit flies (Wernet et al., 2012), horse flies (Meglič et al., 2019), and backswimmers (Schwind, 1984). There are even as many as three different systems at work in Papilio butterflies depending on the behavioral context (Kelber et al., 2001; Kinoshita et al., 2011; Stewart et al., 2019). This information seems to show that different arthropod taxa have utilized the polarization sensitivity inherent in rhabdomeric photoreceptors to create visual subsystems tuned in accordance to their particular ecology. Further investigations into different arthropod taxa will almost certainly reveal novel combinations and processing of photoreceptor inputs using polarized light for object recognition.

We thank two anonymous reviewers for their helpful and constructive comments on this manuscript.