Neurons innervating the lamina in the butterfly, Papilio xuthus
- First Online:
- Cite this article as:
- Hamanaka, Y., Shibasaki, H., Kinoshita, M. et al. J Comp Physiol A (2013) 199: 341. doi:10.1007/s00359-013-0798-6
The butterfly Papilio xuthus has compound eyes with three types of ommatidia. Each type houses nine spectrally heterogeneous photoreceptors (R1–R9) that are divided into six spectral classes: ultraviolet, violet, blue, green, red, and broad-band. Analysis of color discrimination has shown that P. xuthus uses the ultraviolet, blue, green, and red receptors for foraging. The ultraviolet and blue receptors are long visual fibers terminating in the medulla, whereas the green and red receptors are short visual fibers terminating in the lamina. This suggests that processing of wavelength information begins in the lamina in P. xuthus, unlike in flies. To establish the anatomical basis of color discrimination mechanisms, we examined neurons innervating the lamina by injecting Neurobiotin into this neuropil. We found that in addition to photoreceptors and lamina monopolar cells, three distinct groups of cells project fibers into the lamina. Their cell bodies are located (1) at the anterior rim of the medulla, (2) between the proximal surface of the medulla and lobula plate, and (3) in the medulla cell body rind. Neurobiotin injection also labeled distinct terminals in medulla layers 1, 2, 3, 4 and 5. Terminals in layer 4 belong to the long visual fibers (R1, 2 and 9), while arbors in layers 1, 2 and 3 probably correspond to terminals of three subtypes of lamina monopolar cells, respectively. Immunocytochemistry coupled with Neurobiotin injection revealed their transmitter candidates; neurons in (1) and a subset of neurons in (2) are immunoreactive to anti-serotonin and anti-γ-aminobutyric acid, respectively.
KeywordsButterfly Color vision GABA Neurobiotin injection Serotonin
Lamina monopolar cell
Long visual fiber
Short visual fiber
Multiple spectral receptors in the retina—or in the case of insects, the compound eye—are a fundamental requirement of color vision. The number of spectral receptor types and their arrangement within the eyes determine the properties of the color vision system. Accumulated data indicate that the spectral organization of insect compound eyes is quite diverse among species, and in some cases even between sexes (Wakakuwa et al. 2007; Friedrich et al. 2011).
Vision requires further processing of light information in the visual ganglia, the optic lobes beneath the compound eyes. The optic lobe is composed of three parts: (from distal to proximal) the lamina, the medulla, and the lobula complex. The first visual neuropil, the lamina, is modular with each module or cartridge retinotopically connected to a particular ommatidium, or in neural superposition eyes, to a fixed number of neighboring ommatidia. To understand how light information is processed, the structure and cellular components of the optic lobe have been studied in several insect species, but most thoroughly in flies (Strausfeld 1970, 1976; Strausfeld and Blest 1970; Trujillo-Cenóz 1985; Fischbach and Dittrich 1989; Nässel 1991; Homberg 1994; Otsuna and Ito 2006; Raghu and Borst 2011).
Each ommatidium in the fly contains eight photoreceptors (R1–R8). Six of these (R1–R6) are spectrally identical and have axon terminals in the lamina, and are therefore called short visual fibers (svfs). The R7 and R8 photoreceptors are spectrally heterogeneous, and have long axons which pass through the lamina without making any synapses with other neurons, before terminating in the second visual neuropil, the medulla. These are therefore termed long visual fibers (lvfs) (Boschek 1971; Strausfeld 1971; Fischbach and Dittrich 1989; Meinertzhagen and O’Neil 1991; Morante and Desplan 2004). Based on their physiological and anatomical characteristics, it is generally thought that the lvfs are responsible for color vision in flies, and thus, that processing of spectral information takes place in the medulla.
However, the situation is different in butterflies, where color vision is probably more important than in flies. The compound eye of the Japanese yellow swallowtail butterfly, Papilio xuthus, is a collection of three types of ommatidia (type I–III), each housing nine spectrally heterogeneous photoreceptors (R1–R9) in one of three fixed combinations (Arikawa 2003). The photoreceptors are divided into six spectral classes: ultraviolet (UV), violet (V), blue (B), green (G), red (R), and broad-band (BB) (Arikawa 2003). Analysis of wavelength discrimination of foraging P. xuthus revealed that their color vision is based on the UV, B, G, and R photoreceptors, and thus tetrachromatic (Koshitaka et al. 2008). Interestingly, the UV and B receptors are lvfs terminating in the medulla while the G and R receptors are svfs terminating in the lamina. We therefore assume that in P. xuthus, unlike in flies, wavelength information processing begins at the level of the lamina.
This hypothesis has motivated us to investigate the anatomy of the lamina in this butterfly. We have already found that photoreceptor terminals are reciprocally connected in the lamina and that the pattern of connections is ommatidium-type specific (Takemura and Arikawa 2006). In the present study, we injected the tracer Neurobiotin into the lamina synaptic layer, and identified centrifugal neurons innervating the lamina as well as terminals of lvfs and LMCs in the medulla. In addition, we identified candidate neurotransmitters of the centrifugal neurons using immunocytochemistry.
Materials and methods
We used adult P. xuthus of both sexes, taken from a laboratory culture, which is derived from eggs laid by females caught around the campus of Sokendai (Hayama, Kanagawa, Japan). We reared the hatched larvae on fresh citrus leaves under either an LD 10:14 h (diapause-inducing) or LD 14:10 h (diapause-averting) photoperiod. The diapausing pupae were kept at 4 °C for at least 3 months before being allowed to emerge at 25 °C under an LD 12:12 photoperiod.
Mass injection of Neurobiotin
Butterflies with their wings removed were mounted in a section of plastic tubing. A small window was made in the cuticle of the posterior side of the head, and the tracheae were removed to expose the lamina. After removing hemolymph with a piece of paper, a sharp glass needle with crystals of Neurobiotin on the tip was manually advanced into the lamina synaptic layer under a dissecting microscope. The exposed lamina was then briefly washed with saline. The window was resealed with the excised cuticle, covered with a piece of paper soaked in saline, and the butterfly was left overnight at 4 °C to allow the injected Neurobiotin to be taken up by neural processes damaged by the needle.
Injection of Neurobiotin into single long visual fibers
A glass microelectrode filled with 4 % Neurobiotin in 1 M KCl (resistance 80–100 MΩ) was inserted into the retina through a hole made in the dorsal cornea. By recording the spectral and polarization sensitivities of the penetrated cells, we could unambiguously identify the position of the photoreceptor within the ommatidium (Arikawa 2003; Kinoshita et al. 2006). Only when we encountered R1 and R2, which are lvfs with terminals in the medulla, we injected Neurobiotin into the cells by applying 1 nA of depolarizing DC current for 3–5 min. The electrophysiological methods have been described in detail elsewhere (Arikawa et al. 2003).
Visualization of Neurobiotin-filled neurons
The Neurobiotin-injected tissues were fixed with 4 % paraformaldehyde (PFA) and 1 % glutaraldehyde (GA) in 0.067 M phosphate buffer (PB, pH 7.4) containing 1 % saturated picric acid for 4 h on ice or overnight at 4 °C. After fixation, the cornea was removed from the compound eye. The tissues were embedded in gelatin/albumin mixture and then post-fixed in 7–8 % PFA in PB overnight at 4° C. The post-fixed tissues were horizontally sectioned at a thickness of 30–50 μm with a vibrating blade microtome, then washed in PB, and incubated with streptavidin-horseradish peroxidase (HRP, 1:200; RPN1231, Amersham Biosciences, Bucks, UK) in 0.01 M phosphate-buffered saline containing 0.5 % Triton X (PBST, pH 7.4) overnight at 4 °C. The sections were developed with a solution of 0.032 % 3,3’-diaminobenzidine tetrahydrochloride (DAB) in 0.1 M Tris-HCl (pH 7.4) containing 0.0145 % H2O2 and 0.3 % nickel ammonium sulfate. After washing in PB, the sections were mounted on a gelatin-coated slide glass, dehydrated in a grading ethanol series, cleared in xylene, and mounted in mount-quick (Daido Sangyo, Tokyo, Japan) beneath a cover slip.
Antibody labeling of Neurobiotin-injected tissues
To identify neurotransmitter candidates of neurons filled with Neurobiotin, immunocytochemistry was performed on the Neurobiotin-injected tissues. For double labeling with rabbit anti-GABA (Product No. A 2052, Sigma, St. Louis, MO, USA), the tissues were fixed for 3–5 h in an ice-cold solution prepared by mixing 1 part 25 % GA and 3 parts saturated picric acid by volume, and adding acetic acid to a final concentration of 1 %. For double labeling with rabbit anti-serotonin (5-HT; Cat. No. 20080, ImmunoStar, Hudson, WI, USA), the tissues were dissected in ice-cold 4 % PFA and 7.5 % saturated picric acid in 0.067 M PB, and fixed for 3 h on ice or overnight at 4° C. The fixed tissues were embedded in gelatin/albumin, post-fixed, and sectioned as described above.
Immunocytochemistry was performed according to the method of Hamanaka et al. (2012). Sections were incubated with 0.25 % sodium borohydride in PB, followed by PB wash, blocked with 5 % normal goat serum (NGS) in PBST for 1 h, and then incubated with primary antibodies (1:1,000) in PBST containing 5 % NGS for 2 days at 4° C. After washing in PBST, sections were incubated with goat anti-rabbit IgG conjugated to Alexa Fluor 488 (1:200; Molecular Probes, Eugene, OR, USA) in PBST containing 5 % NGS overnight at 4° C. After washing in PBST, the sections were incubated with streptavidin conjugated to Texas Red (1:200; SA-5006, Vector, Burlingame, CA, USA) in PBST overnight at 4° C, followed by washing in PBST, then collected on gelatin-coated slide glass, and mounted in Vectashield (H-1000; Vector) beneath a cover slip. The specificity of the antibodies in P. xuthus tissue has been demonstrated (Hamanaka et al. 2012).
To visualize individual neurons innervating the lamina of P. xuthus, a conventional Golgi-impregnation technique was employed. The method has been described in detail elsewhere (Takemura et al. 2005).
Photography and tracing
DAB-labeled sections were digitally imaged with a CCD camera (DP72, Olympus, Tokyo, Japan) mounted on a light microscope (BX60, Olympus). Several images focused on different planes were manually taken from single sections at an interval of 3–5 μm, and then superimposed into a single image with Zerene Stacker software (Zerene Systems LLC, WA, USA). Medulla terminals of lvfs R1 and R2 were manually traced using a light microscope equipped with a camera lucida.
Sections labeled by immunofluorescence were imaged using a laser scanning microscope (Fluoview FV500; Olympus) equipped with UPlanSApo 20 ×/0.85 and UPlanApo 40 ×/1.0 oil immersion objectives (Olympus). Alexa 488 excited with an Ar laser at 488 nm was viewed through a 505–525-nm band-pass filter, and Texas Red excited with a HeNe green laser at 543 nm through a 560-nm long pass filter. Confocal images were acquired at an interval of 1.0–1.5 μm and a size of 2,048 × 2,048 pixels. All confocal images are stacks of 20 to 35 optical sections. The size, contrast, and brightness of the images were adjusted using Photoshop CS4 (Adobe Systems, Tokyo, Japan) and Corel Draw ver 14 (Corel, Ottawa, ON, Canada).
Neurons labeled by Neurobiotin injection into the lamina
Scattered cell bodies in the area between the proximal surface of the medulla and the lobula plate belong to transmedullary lamina centrifugal neurons. Their thin neurites invade the medulla neuropil through the proximal surface and run parallel to medulla column axes (arrowheads in Fig. 2b).
Golgi-impregnated processes in the lamina
Neurobiotin injection into the lamina usually did not permit the identification of processes within the lamina, due to excessive quantities of tracer. Therefore, we carried out Golgi-impregnation to visualize processes innervating the lamina, and identified two types of processes invading the lamina from the proximal surface: (1) fibers branching over a dozen cartridges (Fig. 3a, b) and (2) fibers innervating single cartridges (Fig. 3c). However, their origin remained uncertain.
In addition to these centrifugal fibers, we identified three different types of LMCs according to their branching patterns in the lamina synaptic layer (Fig. 3d–g): (1) LMC bearing many dendritic spines within its own cartridge (Fig. 3d), (2) LMC innervating the neighboring cartridges as well as its own (Fig. 3e, f), and (3) LMC sparsely bearing short branches in its own cartridge (Fig. 3g).
Transmitter phenotypes of neurons projecting to the lamina
Photoreceptors and LMCs
The distal medulla of P. xuthus is densely innervated by lvf and LMC axons. Our previous study shows that the two types of histamine-positive varicose fibers innervate medulla layers 4a and 4b, respectively, in P. xuthus (Hamanaka et al. 2012). Given that histamine is the neurotransmitter of insect photoreceptors (Hardie 1987), both of these fibers are probably the terminals of lvfs (R1, R2, and R9) (Takemura et al. 2005). Neurobiotin injection into the lamina, where the lvfs extend collaterals, labeled these two types of varicose fibers in medulla layer 4 (Fig. 2d). Since R1 and R2 terminate in layer 4b (Fig. 2e), cells innervating layer 4a are probably R9. This is compatible with the situation in D. melanogaster, where lvf R7 terminates proximal to lvf R8, because lepidopteran R1/2 and R9 correspond to dipteran R7 and R8, respectively (Friedrich et al. 2011). For unambiguous demonstration of axon termination layers, single cell dye injection experiments are necessary.
Besides layer 4, layers 1, 2, 3 and 5 of the medulla are innervated by bushy or slender blebby fibers (Figs. 2c, d, 3h, i). Bushy fibers in layers 1–3 are probably LMC terminals. In the Australian Orchard butterfly, Papilio aegeus, four different types of LMCs (L1–L4) have been identified by Golgi-impregnation, and L1, L2, and L3 have single terminals in layers 1, 2, and 3, respectively, while L4 bears small lateral branches in layer 1 and extends a slender fiber into layer 3 (Ribi 1987). The Neurobiotin-filled arbors in medulla layers 1 to 3 of P. xuthus are morphologically similar to L1 to L3 terminals of P. aegeus, and therefore, we assume that these arbors correspond to terminals of LMC subtypes.
We have found that medulla layers 2 and 3 described by Ribi (1987) in P. aegeus in fact correspond to the distal and proximal layer 2 in P. xuthus, respectively (Hamanaka et al. 2012). The proximal medulla layer 2 of P. xuthus is innervated by tyramine-positive LMCs (Hamanaka et al. 2012), which are morphologically similar to P. aegeus L3 but were not filled with Neurobiotin in the present study (Fig. 2c, d). This is probably because their axons are very thin in the lamina (Hamanaka et al. 2012), sparsely bearing dendritic spines. Besides this LMC, P. xuthus seems to have three other LMC subtypes innervating the medulla layer 1, distal layer 2, and layer 3, respectively. These three terminals possibly originate from three types of Golgi-impregnated LMCs (Fig. 3d–g), respectively. The origin of the slender and blebby fibers in layer 5 (arrows in Figs. 2d and 3i) is uncertain, but these could be attributed to another LMC subtype or one of three distinct populations of centrifugal neurons whose cell bodies are located around the medulla.
In several insects so far studied, tangential (including T1 and amacrine) and columnar neurons are known to provide centrifugal inputs to the lamina (Strausfeld 1970, 1971; Fischbach and Dittrich 1989; Nässel 1991; Homberg 1994). The T1 cells of flies have cell bodies in the anterior and posterior medulla cell body rind, and connect medulla layers 1 and 2 to the lamina cartridges, where they exhibit characteristic basket-like terminals (Strausfeld 1970, 1971; Fischbach and Dittrich 1989; Hamanaka and Meinertzhagen 2010). Similar T1 cells were also identified by Golgi-impregnation in the butterfly Pieris brassicae and the moth Sphinx ligustri (Strausfeld and Blest 1970).
We found similar processes in the lamina of P. xuthus (Fig. 3a, b), which probably originate from the counterparts of T1 in this species. Supporting this view, a group of Neurobiotin-filled cell bodies exists near the anterior medulla, where the T1 cells of flies have their cell bodies (Strausfeld 1970, 1971; Fischbach and Dittrich 1989; Hamanaka and Meinertzhagen 2010). Neurons corresponding to fly amacrines have not been identified in any lepidopteran species so far, and we could not find them in P. xuthus either.
The distribution of 5-HT-immunoreactive tangential processes in the lamina is widely identified among insect species. These processes are derived from cell bodies either along the anterior edge of the medulla or in the central brain (Nässel 1988; Homberg and Hildebrand 1989; Leitinger et al. 1999). In P. xuthus, as in the moth Manduca sexta (Homberg and Hildebrand 1989), we found a set of 5-HT-immunoreactive cell bodies near the anterior rim of the medulla providing centrifugal tangential processes into the lamina (Fig. 6; Hamanaka et al. 2012). The 5-HT-positive tangential neurons possibly have a neuromodulatory role, regulating daily activity of cells in the cartridges.
Likewise, GABA-immunoreactive neurons provide centrifugal inputs to the lamina (Homberg 1994; Kolodziejczyk et al. 2008). GABA is a wide-spread inhibitory neurotransmitter (Sattelle 1990). In D. melanogaster, transmedullary C2 and C3 neurons, which have cell bodies proximal to the medulla, collaterals in the medulla, and terminals in the lamina, are immunoreactive to anti-GABA and anti-glutamic acid decarboxylase antibodies (Kolodziejczyk et al. 2008), and also express vesicular GABA transporter (Fei et al. 2010), implying their GABAergic phenotype. Similar GABA-positive neurons were also found in other fly species (Datum et al. 1986; Meyer et al. 1986; Sinakevitch et al. 2003), a moth (Homberg et al. 1987), and a cockroach (Füller et al. 1989).
The Neurobiotin-filled cell bodies between the medulla and lobula plate in P. xuthus are most likely the counterparts of the fly C2/3 neurons. Since the fly C2/3 neurons each innervate a single lamina cartridge (Sinakevitch et al. 2003), the Golgi-impregnated processes innervating single cartridges (Fig. 3c) probably belong to P. xuthus C2/3-like neurons (Fig. 2b). A unique feature of the P. xuthus C2/3-like neurons is that the transmitter phenotype seems to be heterogeneous: a small number of neurons are GABA-positive but most neurons are GABA-negative (Fig. 5g–i). This heterogeneity may reflect their morphological variation. To address this issue, analysis at a single cell level is essential. Transmitter candidates of the GABA-negative C2/3-like cells remain uncertain. An elaborate electron microscopic study of the lamina of D. melanogaster has revealed that C3 neurons are presynaptic to L1 and L2 (Meinertzhagen and O’Neil 1991). Therefore, the GABA-positive C2/3-like cells of P. xuthus may function as inhibitory feedback elements in the lamina.
GABA-positive but Neurobiotin-negative cell bodies are located in both the medulla cell body rind and at the anterior rim of the medulla. The former are probably medulla columnar neurons whereas the latter are medulla tangential neurons. The GABA-positive medulla columnar neurons with cell bodies along the distal surface of the medulla were also found in a cockroach (Füller et al. 1989) and a moth (Homberg et al. 1987; Homberg 1994). Small sets of putatively GABAergic medulla tangential neurons with fibers projecting to the brain were reported in flies and bees (Meyer et al. 1986).
Lamina cartridges of P. xuthus are innervated by at least 16 different cells: nine photoreceptors (R1 to R9), four LMC subtypes and three centrifugal neurons (5-HT-positive lamina tangential, C2/3-like, and T1-like neurons) (Fig. 6).
Spectral inputs from photoreceptors are probably separated into at least two major parallel pathways in the lamina: a color discrimination pathway and a motion detection pathway. Based on behavioral measurements of wavelength discrimination ability, P. xuthus uses UV, blue, green, and red receptors to discriminate colors, but not violet and broad-band receptors (Koshitaka et al. 2008). Green and red receptors are svfs (R3–R8) terminating in the lamina. UV and blue receptors are lvfs (R1 and R2), which extend their axons to the medulla with some processes in the lamina. Therefore, the four spectral receptors essential for color vision are all incorporated in lamina cartridges, where their axons are mutually connected (Takemura and Arikawa 2006). Our present data suggest that none of the lvfs and LMCs of P. xuthus terminates in the same medulla layer, except for R1 and R2 (Fig. 2d, e). Given that svfs synapse to LMCs in the lamina as generally expected, this means that spectral information received by the svfs and lvfs is fed to separate medulla layers. In flies, lvfs R7 and R8, which are thought to be essential for color vision (Morante and Desplan 2004), project to the medulla without any synaptic engagements in the lamina, while svfs provide inputs to the medulla indirectly via LMCs (Boschek 1971; Strausfeld 1971; Meinertzhagen and O’Neil 1991). Chemical synapses between LMCs and lvfs, between LMCs, and between lvfs are formed in the medulla columns (Takemura et al. 2008), suggesting that wavelength information first converges and is processed in the medulla in flies. In P. xuthus, on the other hand, lvfs and LMCs are probably not connected in the medulla (Fig. 2d). Instead, wavelength information first converges in local circuits in the lamina, where the extraction of color information may occur. Color discrimination at the level of LMCs is theoretically possible given that certain LMCs receive spectral inputs from multiple photoreceptors. However, anatomical or physiological evidence concerning spectral inputs to LMCs is not yet available. Detailed studies of synaptic connections within cartridges and spectral sensitivities of single LMCs will be indispensable to address the question of whether and how LMCs contribute to color discrimination.
At present, it is uncertain how and whether C2/3-like and T1-like neurons are involved in color discrimination. Feedback, in general, is one of the most important mechanisms to tune outputs in the nervous system. The subset of putatively GABAergic C2/3-like neurons is an attractive candidate to be involved in some kind of visual opponency phenomena.
We thank Dr. Uwe Homberg and Dr. Finlay Stewart for critically reading the manuscript. This work was supported in part by the JSPS Grants-in-Aid for Scientific Research No. 21247009 to KA, No. 24570084 to MK, the MAFF (Ministry of Agriculture, Forestry and Fisheries of Japan) grant (Elucidation of biological mechanisms of photoresponse and development of advanced technologies utilizing light) No. INSECT-1101 to KA. All experiments were conducted according to the MEXT (Ministry of Education, Culture, Sports, Science and Technology of Japan) guidelines for proper conduct of animal experiments and related activities in academic research institutions.