Feeding Behavior in the Nocturnal Moth Manduca Sexta is Mediated Mainly by Blue Receptors, but where are they Located in the Retina?

The ‘color vision’ of insects has been widely studied in Hymenoptera (Peitsch et al. 1992; Menzel and Shmida, 1993; Chittka et al. 1994) and, to a much lesser extent, in Lepidoptera (Scherer and Kolb, 1987a,b; Bernard and Remington, 1991). Bees and other flower-feeding Hymenoptera exhibit ‘true color vision’; that is, they learn to associate food sources arbitrarily with chromaticity independent of intensity. Butterflies, in contrast, seem to have a more limited capacity for using color signals, which is better characterized as ‘wavelength-specific behavior’ (Menzel, 1979; Goldsmith, 1990). That is, specific behaviors such as feeding, egg laying or mating appear to be evoked by the specific activation of one or more spectral classes of photoreceptor (Scherer and Kolb, 1987a,b; Bernard and Remington, 1991; White et al. 1994). In three species of butterfly, the spectral sensitivity of feeding behavior shows a peak in the blue–violet with a secondary maximum at longer wavelengths (Scherer and Kolb, 1987a,b).

Hawkmoths (Sphingidae) also suck nectar from flowers, but most feed at dusk or at night. We have investigated whether the feeding behavior of these nocturnal Lepidoptera is also wavelength-specific, as it is in diurnal butterflies. Observations of tropical hawkmoths in the field suggested that this might not be the case, since they feed at characteristic ‘hawkmoth flowers’ that are typically white or cream-colored (Baker, 1961; Paige and Whitham, 1985; Haber and Frankie, 1989).

We examined a number of hawkmoth flowers in Costa Rica and found that they lack ultraviolet reflectance. Furthermore, laboratory experiments showed that the hawkmoth Manduca sexta strongly prefers to feed at white artificial flowers, or at back-lit feeding stations, that mimic hawkmoth flowers in reflecting or transmitting only wavelengths longer than 400 nm (White et al. 1992, 1994). The retina of Manduca sexta contains three visual pigments, P520, P450 and P357 (Bennett and Brown, 1985), presumably providing green-, blue-and ultraviolet-sensitive receptors. Thus, our results indicated that feeding behavior does depend on wavelength, and that it is impeded by input from the ultraviolet receptors and activated via the blue and/or green receptors.

We have now measured the action spectrum of the feeding response and find that it is similar to those of butterflies. The behavioral data indicating wavelength-specific feeding behavior in Manduca sexta also lead to the question of how the three spectral classes of photoreceptor are distributed within individual retinulae and across the retina of the compound eye. We present initial results from a strategy for identifying receptor types based on their morphological response to light.

Manduca sexta were reared as previously described (Bennett and White, 1989; White et al. 1994). Adult moths were used for behavioral experiments during the first week after eclosion. A changing population of 25–30 moths was maintained in a flight room (2.8 m×4.4 m×2.8 m) for experimental observation.

The experimental arrangement for spontaneous-choice experiments to determine the spectral sensitivity of feeding behavior was similar to that previously described (White et al. 1994). The flight room was maintained under a 15 h:9 h L:D cycle in which the room lights (5X10¹³ quanta cm^-2 s^-1) were turned on and off, while dim overhead fluorescent lights stayed on continuously to provide ‘moonlight’ for observation (approximately 5X10¹¹ quanta cm^-2 s^-1, measured at the feeding stations) during the ‘dark’ phase. Moths were observed for 2 h after lights-off; they began to fly and seek food within 15 min. Two feeding stations, centered 34 cm apart, simulated ‘flowers’. These were 5 cm×5 cm horizontal windows, illuminated from below, with reservoirs containing 20 % sucrose solution at each corner (Fig. 1).

The method developed for determining spectral sensitivity was based on the following initial observations. Offered a choice between two stimuli of identical spectral quality, i.e. the illuminated feeding stations covered by the same color filters, moths fed more frequently at the brighter stimulus. Intensity–response functions were linear and similar to those plotted in Fig. 2. When offered a choice between stations with differing spectral qualities, a similar linear intensity–response function was generated by adjusting the relative intensity of the two stimuli.

In order to measure the intensity–response functions shown in Fig. 2, the window of one station was covered with a green Wratten 57 filter, maximal transmission at 520 nm, bandwidth 95 nm. This station provided a reference stimulus of constant spectral quality near the center of the spectral range over which measurements were made. The other station was illuminated by a Bausch and Lomb high-intensity grating monochrometer, bandwidth approximately 20 nm. Stimulus intensities were measured with a PIN 10 ultraviolet-calibrated photodiode in quanta cm^-2 s^-1. The monochrometer was changed randomly to a wavelength between 400 and 640 nm for each daily observation. Each separate visit to a window that resulted in probing, whether or not it ended with feeding from an adjacent sucrose reservoir (Fig. 1), was scored as one unit of feeding behavior.

The procedure for measuring intensity–response functions is best described by providing an example. The intensity–response function for 540 nm (Fig. 2) was determined as follows. At this wavelength, approximately equal numbers of moths initially visited each station (540 nm provided by the monochrometer versus 520 nm provided by the broader-band W57 filter) when neither was dimmed by a neutral density (ND) filter. 41 % of the moths fed at the 540 nm stimulus when it was dimmed with a 0.5 ND filter. 55 % fed at that station when the alternative 520 nm stimulus was dimmed with the same filter. The intensity of the unfiltered 540 nm stimulus was measured as logI=13.0; the measured optical density of the 0.5 ND filter over the range 520–540 nm was 0.58. Thus, 41 % of the moths were attracted to the 540 nm stimulus at logI=12.42. A reduction in intensity at one station is equivalent to an increase at the other. Therefore, it was concluded that 55 % of the moths would have visited the 540 nm stimulus at logI=13.58. These results account for the three central data points at 540 nm in Fig. 2. The remaining points were determined in the same manner with higher-density ND filters.

To summarize: neutral density filters were added at one or the other feeding station, depending on the initial distribution of visits, in order to collect several data points on either side of the 50 % value; it was impossible to use a reference stimulus of constant intensity owing to the large differences in relative sensitivity from 400 to 640 nm (Figs 2, 3). Intensity–response functions were plotted by fitting a least-squares curve to the spontaneous choice data for each monochromatic stimulus wavelength. The 50 % values taken from the intensity–response functions were used to plot a spectral sensitivity function of feeding activity (Fig. 3).

Details of the procedure and optical system for measuring the spectral sensitivity of Manduca sexta from the electroretinogram (ERG) are described in White et al. (1983a,b).

The morphological response of Manduca sexta photoreceptor cells to light – associated with shedding of rhabdomeric membrane (White and Bennett, 1992) – was used to identify chromatic classes. To activate shedding, bright heat-filtered light from a high-intensity tungsten–halogen microscope lamp (Labsource QH-150) was directed into eyes of moths for 2 min at the end of the dark phase of the L:D cycle by way of an ultraviolet-transmitting light guide. Moths were then maintained in darkness for 1 h before fixation for electron microscopy (White and Bennett, 1989).

To determine the effects of light at different wavelengths, eyes were irradiated with unfiltered white light (approximately 2X10¹⁶ photons cm^-2 s^-1, measured from 400 to 700 nm with a Licor Li-185A microEinstein meter), with red light provided by a Corning filter 2-58 cutting off at 640 nm (approximately 3X10¹⁵ photons cm^-2 s^-1, measured from 630 to 700 nm with a PIN-10 ultraviolet photodiode), with yellow light provided by a Schott GG475 cutting off at 475 nm (approximately 1.5X10¹⁶ photons cm^-2 s^-1, measured from 450 to 700 nm with the photodiode) or with ultraviolet light provided by a Wratten 18A open from 310 to 400 nm (approximately 6X10¹⁴ photons cm^-2 s^-1, measured from 300 to 400 nm with the photodiode). Absorbance spectra of filters (Fig. 3C) were measured with a Cary 1E recording spectrophotometer. In another experiment, eyes were irradiated with the red stimulus over a range of intensities provided by numbers 1, 2 and 3 Wratten neutral density filters.

Fig. 2 plots intensity–response functions for the feeding response of Manduca sexta at wavelengths from 400 to 640 nm. Data could not be obtained at shorter and longer wavelengths because they were insufficiently attractive to the moths. The spectral sensitivity function derived from the intensity–response curves is dominated by a peak at 440 nm (Fig. 3A; Table 1). A minor peak at 560 nm is set off by a valley at 500 nm. On its short-wavelength side, the action spectrum falls off steeply between 420 and 400 nm. In contrast, the ERG spectral sensitivity function, which measures the mass response of the retina, shows a broad maximum around 520 nm in the green with a flat shoulder extending through the blue into the ultraviolet (Fig. 3B).

Retinulae of Manduca sexta were depicted by White et al. (1983b) as consisting of nine receptor cells; eight long receptors whose cell bodies extend from the distal surface of the retina almost to its proximal margin at the basement membrane, a distance of approximately 190 μm, and a small proximal ninth cell, some 10 μm in diameter, lying just above the basement membrane (Fig. 4). We have designated the cells of the Manduca sexta retinula according to their position and morphology. The eight long cells fall into three groups distinguished by common features of orientation and morphology (Fig. 4). The pair of receptors designated DV face each other along the dorsal–ventral axis of the eye; they have large, distally expanded rhabdomeres that fold into the cell. The AP pair are oriented perpendicular to the DV cells in the anterior–posterior axis: their rhabdomeres are enlarged proximally. The four OB cells are oriented obliquely between the DV and AP cells: their rhabdomeres are more uniformly distributed along the length of the retinula (R. H. White, unpublished observations). The small proximal cell has been designated PR.

We have now found that many retinulae contain only seven long cells (and therefore eight cells altogether): one member of the DV pair is missing (Fig. 5C). The eight-and nine-cell retinulae appear to be clustered separately, not mixed together. The distribution of these clusters across the retina has not been mapped.

Insect photoreceptors exhibit a variety of structural responses to light that have been exploited for identifying the spectral sensitivities of particular cells (e.g. Welsch, 1977; Ribi, 1978; Meinecke and Langer, 1984; Kolb, 1985). In Manduca sexta (White and Bennett, 1992), as in other arthropods (Blest, 1988; Meinecke and Langer, 1984), light induces the ultrastructural features of rhabdomere turnover in dark-adapted photoreceptors: disorganized rhabdomeres with swollen and shortened microvilli, and stacked endomembrane (myeloid bodies) (Fig. 5).

We have used these striking ultrastructural effects of light to develop a systematic approach for distinguishing the three classes of Manduca sexta photoreceptor: green-, blue-and ultraviolet-sensitive. We chose color filters matched to the absorbance spectra of the three visual pigments of Manduca sexta, P520, P450 and P357. The basis for choosing filters (Fig. 3C) was that the absorbances of P520 and P450 differ maximally, by about 2 log units, at wavelengths longer than 570 nm. Thus, the red light provided by a Corning filter 2403 will selectively activate rhabdomere turnover in green-sensitive cells at an appropriate intensity. The yellow light provided by a Schott GG475 filter will activate both green and blue receptors, but leave ultraviolet receptors dark-adapted. The ultraviolet light provided by a Wratten 18A filter will activate all three receptor classes, since the absorbances of the three photopigments are expected to differ by less than 1 log unit in the near ultraviolet (Ichikawa and Tateda, 1982). Eyes were exposed to bright light for 2 min at the end of the dark segment of the light/dark cycle and fixed after an additional hour in darkness. Red and yellow light gave the same result: six of the retinular cells were activated, the AP and OB receptors (Fig. 5A–C). The response was the same when the red stimulus was reduced with neutral density filters by 1 or 2 log units; when dimmed by 3 log units, all cells retained the dark-adapted configuration (data not shown). Thus, the intensity range of the red stimulus was appropriate for distinguishing between green and blue receptors. We conclude that the AP and OB cells are all green receptors.

The DV receptors remained dark-adapted under red and yellow light, (Fig. 5A–C), but were activated by ultraviolet light (Fig. 5D). Thus, the two DV cells must be ultraviolet receptors. In eight-cell retinulae that lack one member of the DV pair, the remaining cell was also an ultraviolet receptor (Fig. 5C).

The appearance of the small basal cell was variable in these experiments, so that dark-and light-adapted states could not be reliably distinguished. Thus, we could not infer its spectral sensitivity.

Bennett and Brown (1985) extracted three visual pigments from Manduca sexta retinas in the ratio of 75 P520: 19 P450: 6 P357. ERG spectral sensitivity measurements (Fig. 3B) confirmed the predominance of green-sensitive photoreceptors containing P520.

The a.-bands of the three Manduca sexta rhodopsins determined from difference spectra (White et al. 1983b; Bennett and Brown, 1985) fit the corresponding portions of intracellular spectral sensitivity curves measured from the larval stemmata of the noctuid moth Mamestra brassicae (Ichikawa and Tateda, 1982). Thus, we have adopted the Mamestra brassicae data, in lieu of nomogram calculations, to model the absorption spectra of the three rhodopsins. Adding the predicted spectra in the ratio of 77 P520: 10 P450: 13 P357 yields a hypothetical sensitivity function that closely matches the ERG spectral sensitivity curve (Fig. 3B). Taken together, the ERG and visual pigment measurements indicate that about 75 % of the retinal cells are green receptors, with the remaining about equally divided between blue and ultraviolet receptors.

The action spectrum of the feeding response indicates that it is predominantly mediated by blue receptors. The peak at 440 nm is narrower on both sides of the maximum than the absorption spectrum of P450. A distinct minimum at 500 nm isolates a low secondary peak around 550 nm that suggests a small contribution from green receptors. Since the feeding response is suppressed by ultraviolet light (White et al. 1994), we were concerned that the low sensitivity at 500 nm might be an artifact resulting from ultraviolet contamination at that setting of the monochrometer. However, the intensity–response function at 500 nm was the same whether or not an ultraviolet blocking filter was put into the monochrometer beam. Furthermore, preliminary experiments with interference filters yielded a qualitatively similar action spectrum.

Unlike ERG spectral sensitivity, the action spectrum of feeding behavior cannot be matched by adding photopigment absorption spectra. Such an addition in the ratio of 9 P450: 1 P520 is plotted for comparison with the action spectrum in Fig. 3A. The narrowed blue peak and the valley that separates it from the green peak suggest a negative interaction between the blue-and green-sensitive mechanisms. This might arise at various levels in the visual system (Menzel et al. 1986). However, measurements from the retina of the butterfly Papilio aegeus aegeus (Horridge et al. 1983; Matic, 1983) indicate that narrowing of receptor spectral sensitivities in this species arises in the retina from mutual electrical inhibition between receptors, as proposed by Shaw (1975). The action spectra of various types of visual behavior are also narrow in butterflies (Scherer and Kolb, 1987a,b).

The spectral sensitivity of feeding falls off steeply below 420 nm, in accordance with previous behavioral experiments showing that the feeding response is hindered by ultraviolet wavelengths (White et al. 1994). The rapidly increasing absorbance of P357 from 440 to 400 nm (Fig. 3A) inversely mirrors the sharp decline in the sensitivity of the feeding response in the far blue. We suggested (White et al. 1994) that the ultraviolet receptors mediate a competing type of behavior, perhaps the ‘open-space’ reaction (Menzel, 1979; Scherer and Kolb, 1987b).

The spectral sensitivity of flower visitation in Manduca sexta agrees with the observations of Knoll (1925), who found that the dusk-feeding hawkmoth Deilephila livornica was preferentially attracted to blue-and violet-colored papers. Retinula organization and visual pigments are very similar in Deilephila livornica (Schwemer and Paulsen, 1973; Welsch, 1977; Schlecht et al. 1978) and Manduca sexta. Although one might expect the feeding response of a nocturnal lepidopteran visiting white flowers to differ from that of diurnal forms attracted to colored flowers, the action spectrum measured in Manduca sexta is similar to those that have been measured in three species of butterfly, Pieris brassicae, Aglais urticae and Pararge aegeria (Scherer and Kolb, 1987a,b). All are bimodal with a peak at 450 nm separated from one at longer wavelengths by a deep minimum. In Pieris brassicae and Aglais urticae, narrow secondary maxima lie at 600 and 582 nm respectively.

The action spectrum of Pararge aegeria is characterized by two broad equivalent maxima at 450 and 650 nm. However, long-wavelength sensitivity is much lower in Manduca sexta, relative to the blue maximum, than it is in the butterfly species.

The number and arrangement of receptor cells in Manduca sexta retinulae (Fig. 4) resemble the patterns found in most of the butterflies and moths whose retinal ultrastructure has been examined (Gordon, 1977; Maida, 1977; Welsch, 1977; Ribi, 1978, 1987; Schlecht et al. 1978; Eguchi, 1982; Kolb, 1977, 1985; Anton-Erxleben and Langer, 1988; Pelzer and Langer, 1990; Shimohigashi and Tominaga, 1991; Bandai et al. 1992). At least in the small samples we have analyzed, taken towards the center of the retina, the morphological response to light indicates that the six AP and OB cells are all green receptors, while the two DV cells are ultraviolet receptors. Light-induced ultrastructural effects and microspectrophotometric measurements from the retina, as well as other considerations, led others (Welsch, 1977; Schlecht et al. 1978; Schlecht, 1979) to make the same sensitivity assignments to the corresponding cells in the hawkmoth Deilephila elpenor. Neither we nor they have found the blue receptors. Welsch (1977) and Schlecht (1979) proposed that the proximal cell of the Deilephila elpenor retinula might be the blue receptor. Unfortunately, we have so far failed to characterize the proximal cell (PR) in Manduca sexta. However, the amount of rhabdom contributed to the retina by the small PR cells, about 5 % (R. H. White, unpublished data), seems too low to account for the measured proportion of P450 in whole eye extracts (Bennett and Brown, 1985).

The retina of Manduca sexta is made up of over 25 000 retinulae (R. H. White, unpublished measurements). When the distribution of receptor classes across the retina has been mapped, we may find regional patterns of the sort that have been seen in other insects including Lepidoptera (Bernard and Remington, 1991; Stavenga, 1992). In the light of the spectral sensitivity of the feeding response, we are looking particularly for a retinal sector in which blue receptors predominate, perhaps the anterior-ventral region, since the moths approach flowers from above.

This work was supported by National Science Foundation grant BNS-9110672 and by an NSF REU grant BIO-9200415, which supported D.C.’s first summer of work. Expert technical assistance was provided by Niel McGaffigan. We thank Thomas Cronin for a helpful critique of initial spectral sensitivity measurements.